Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-04-24 23:54:16 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-16 22:58:11 +01:00
|
|
|
|
2019-04-24 23:54:16 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-10-03 07:22:02 -06:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-01-11 09:43:02 -07:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-01-18 20:22:41 +03:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add per-task callback handler
For poll requests, it's not uncommon to link a read (or write) after
the poll to execute immediately after the file is marked as ready.
Since the poll completion is called inside the waitqueue wake up handler,
we have to punt that linked request to async context. This slows down
the processing, and actually means it's faster to not use a link for this
use case.
We also run into problems if the completion_lock is contended, as we're
doing a different lock ordering than the issue side is. Hence we have
to do trylock for completion, and if that fails, go async. Poll removal
needs to go async as well, for the same reason.
eventfd notification needs special case as well, to avoid stack blowing
recursion or deadlocks.
These are all deficiencies that were inherited from the aio poll
implementation, but I think we can do better. When a poll completes,
simply queue it up in the task poll list. When the task completes the
list, we can run dependent links inline as well. This means we never
have to go async, and we can remove a bunch of code associated with
that, and optimizations to try and make that run faster. The diffstat
speaks for itself.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-02-17 09:52:41 -07:00
|
|
|
|
2020-09-13 13:09:39 -06:00
|
|
|
|
2023-12-01 00:57:35 +00:00
|
|
|
|
2021-02-16 19:46:48 -05:00
|
|
|
|
lsm,io_uring: add LSM hooks to io_uring
A full expalantion of io_uring is beyond the scope of this commit
description, but in summary it is an asynchronous I/O mechanism
which allows for I/O requests and the resulting data to be queued
in memory mapped "rings" which are shared between the kernel and
userspace. Optionally, io_uring offers the ability for applications
to spawn kernel threads to dequeue I/O requests from the ring and
submit the requests in the kernel, helping to minimize the syscall
overhead. Rings are accessed in userspace by memory mapping a file
descriptor provided by the io_uring_setup(2), and can be shared
between applications as one might do with any open file descriptor.
Finally, process credentials can be registered with a given ring
and any process with access to that ring can submit I/O requests
using any of the registered credentials.
While the io_uring functionality is widely recognized as offering a
vastly improved, and high performing asynchronous I/O mechanism, its
ability to allow processes to submit I/O requests with credentials
other than its own presents a challenge to LSMs. When a process
creates a new io_uring ring the ring's credentials are inhertied
from the calling process; if this ring is shared with another
process operating with different credentials there is the potential
to bypass the LSMs security policy. Similarly, registering
credentials with a given ring allows any process with access to that
ring to submit I/O requests with those credentials.
In an effort to allow LSMs to apply security policy to io_uring I/O
operations, this patch adds two new LSM hooks. These hooks, in
conjunction with the LSM anonymous inode support previously
submitted, allow an LSM to apply access control policy to the
sharing of io_uring rings as well as any io_uring credential changes
requested by a process.
The new LSM hooks are described below:
* int security_uring_override_creds(cred)
Controls if the current task, executing an io_uring operation,
is allowed to override it's credentials with @cred. In cases
where the current task is a user application, the current
credentials will be those of the user application. In cases
where the current task is a kernel thread servicing io_uring
requests the current credentials will be those of the io_uring
ring (inherited from the process that created the ring).
* int security_uring_sqpoll(void)
Controls if the current task is allowed to create an io_uring
polling thread (IORING_SETUP_SQPOLL). Without a SQPOLL thread
in the kernel processes must submit I/O requests via
io_uring_enter(2) which allows us to compare any requested
credential changes against the application making the request.
With a SQPOLL thread, we can no longer compare requested
credential changes against the application making the request,
the comparison is made against the ring's credentials.
Signed-off-by: Paul Moore <paul@paul-moore.com>
2021-02-01 19:56:49 -05:00
|
|
|
|
2024-10-18 17:07:59 +01:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
io_uring: add set of tracing events
To trace io_uring activity one can get an information from workqueue and
io trace events, but looks like some parts could be hard to identify via
this approach. Making what happens inside io_uring more transparent is
important to be able to reason about many aspects of it, hence introduce
the set of tracing events.
All such events could be roughly divided into two categories:
* those, that are helping to understand correctness (from both kernel
and an application point of view). E.g. a ring creation, file
registration, or waiting for available CQE. Proposed approach is to
get a pointer to an original structure of interest (ring context, or
request), and then find relevant events. io_uring_queue_async_work
also exposes a pointer to work_struct, to be able to track down
corresponding workqueue events.
* those, that provide performance related information. Mostly it's about
events that change the flow of requests, e.g. whether an async work
was queued, or delayed due to some dependencies. Another important
case is how io_uring optimizations (e.g. registered files) are
utilized.
Signed-off-by: Dmitrii Dolgov <9erthalion6@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-15 19:02:01 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2019-10-24 07:25:42 -06:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2025-09-04 11:08:58 -06:00
|
|
|
|
2022-05-24 12:45:38 -06:00
|
|
|
|
2022-05-25 20:31:09 -06:00
|
|
|
|
2022-05-25 08:56:52 -06:00
|
|
|
|
2022-05-25 11:01:04 -06:00
|
|
|
|
2023-12-19 08:54:20 -07:00
|
|
|
|
2022-05-25 09:13:39 -06:00
|
|
|
|
2022-05-25 10:40:19 -06:00
|
|
|
|
2022-06-13 07:07:23 -06:00
|
|
|
|
2022-06-13 07:12:45 -06:00
|
|
|
|
2022-06-16 10:22:02 +01:00
|
|
|
|
2022-07-07 14:30:09 -06:00
|
|
|
|
2022-07-12 21:52:38 +01:00
|
|
|
|
2023-07-10 16:14:37 -06:00
|
|
|
|
io_uring: add support for futex wake and wait
Add support for FUTEX_WAKE/WAIT primitives.
IORING_OP_FUTEX_WAKE is mix of FUTEX_WAKE and FUTEX_WAKE_BITSET, as
it does support passing in a bitset.
Similary, IORING_OP_FUTEX_WAIT is a mix of FUTEX_WAIT and
FUTEX_WAIT_BITSET.
For both of them, they are using the futex2 interface.
FUTEX_WAKE is straight forward, as those can always be done directly from
the io_uring submission without needing async handling. For FUTEX_WAIT,
things are a bit more complicated. If the futex isn't ready, then we
rely on a callback via futex_queue->wake() when someone wakes up the
futex. From that calback, we queue up task_work with the original task,
which will post a CQE and wake it, if necessary.
Cancelations are supported, both from the application point-of-view,
but also to be able to cancel pending waits if the ring exits before
all events have occurred. The return value of futex_unqueue() is used
to gate who wins the potential race between cancelation and futex
wakeups. Whomever gets a 'ret == 1' return from that claims ownership
of the io_uring futex request.
This is just the barebones wait/wake support. PI or REQUEUE support is
not added at this point, unclear if we might look into that later.
Likewise, explicit timeouts are not supported either. It is expected
that users that need timeouts would do so via the usual io_uring
mechanism to do that using linked timeouts.
The SQE format is as follows:
`addr` Address of futex
`fd` futex2(2) FUTEX2_* flags
`futex_flags` io_uring specific command flags. None valid now.
`addr2` Value of futex
`addr3` Mask to wake/wait
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2023-06-08 11:57:40 -06:00
|
|
|
|
2023-06-08 09:38:36 -07:00
|
|
|
|
2024-03-18 22:00:23 +00:00
|
|
|
|
2024-06-06 12:25:01 -06:00
|
|
|
|
2024-03-27 14:59:09 -06:00
|
|
|
|
2025-02-14 16:09:36 -08:00
|
|
|
|
2026-02-26 12:48:39 +00:00
|
|
|
|
2022-05-24 10:56:14 -06:00
|
|
|
|
2022-05-25 08:57:27 -06:00
|
|
|
|
2022-05-25 20:31:09 -06:00
|
|
|
|
2023-06-02 08:41:46 -06:00
|
|
|
|
2022-07-07 14:16:20 -06:00
|
|
|
|
2024-06-03 11:51:19 -06:00
|
|
|
|
2026-01-22 08:40:19 -07:00
|
|
|
|
2026-01-15 08:24:02 -07:00
|
|
|
|
io_uring: introduce callback driven main loop
The io_uring_enter() has a fixed order of execution: it submits
requests, waits for completions, and returns to the user. Allow to
optionally replace it with a custom loop driven by a callback called
loop_step. The basic requirements to the callback is that it should be
able to submit requests, wait for completions, parse them and repeat.
Most of the communication including parameter passing can be implemented
via shared memory.
The callback should return IOU_LOOP_CONTINUE to continue execution or
IOU_LOOP_STOP to return to the user space. Note that the kernel may
decide to prematurely terminate it as well, e.g. in case the process was
signalled or killed.
The hook takes a structure with parameters. It can be used to ask the
kernel to wait for CQEs by setting cq_wait_idx to the CQE index it wants
to wait for. Spurious wake ups are possible and even likely, the callback
is expected to handle it. There will be more parameters in the future
like timeout.
It can be used with kernel callbacks, for example, as a slow path
deprecation mechanism overwiting SQEs and emulating the wanted
behaviour, however it's more useful together with BPF programs
implemented in following patches.
Note that keeping it separately from the normal io_uring wait loop
makes things much simpler and cleaner. It keeps it in one place instead
of spreading a bunch of checks in different places including disabling
the submission path. It holds the lock by default, which is a better fit
for BPF synchronisation and the loop execution model. It nicely avoids
existing quirks like forced wake ups on timeout request completion. And
it should be easier to implement new features.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://patch.msgid.link/a2d369aa1c9dd23ad7edac9220cffc563abcaed6.1772109579.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2026-02-26 12:48:38 +00:00
|
|
|
|
2022-05-24 11:46:43 -06:00
|
|
|
|
2021-09-15 12:03:38 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-02-11 13:19:56 -07:00
|
|
|
|
|
|
|
|
|
2021-06-17 18:14:04 +01:00
|
|
|
|
2025-07-11 15:41:02 -06:00
|
|
|
|
2021-02-18 18:29:40 +00:00
|
|
|
|
2025-02-11 13:19:56 -07:00
|
|
|
|
2025-07-11 15:41:02 -06:00
|
|
|
|
|
|
|
|
|
2022-03-21 22:02:22 +00:00
|
|
|
|
2021-06-14 02:36:22 +01:00
|
|
|
|
|
|
|
|
|
2021-02-10 00:03:13 +00:00
|
|
|
|
2021-02-10 00:03:17 +00:00
|
|
|
|
2021-02-10 00:03:10 +00:00
|
|
|
|
2021-08-15 10:40:25 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-06-05 11:48:33 -06:00
|
|
|
|
2025-05-09 12:12:54 +01:00
|
|
|
|
2019-04-06 21:51:27 -06:00
|
|
|
|
2026-02-15 22:06:52 +00:00
|
|
|
|
2024-10-18 17:07:59 +01:00
|
|
|
|
2023-01-18 07:56:30 -08:00
|
|
|
|
2024-04-01 15:16:19 -06:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2023-08-21 17:15:52 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-01-28 13:48:37 +01:00
|
|
|
|
2023-08-21 17:15:52 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-08-17 23:09:18 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-05-19 12:44:32 +01:00
|
|
|
|
2025-08-17 23:09:18 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-08-27 17:46:09 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-05-24 15:21:00 -06:00
|
|
|
|
2021-08-27 17:46:09 +08:00
|
|
|
|
|
|
|
|
|
2022-04-12 15:09:48 +01:00
|
|
|
|
|
|
|
|
|
2025-08-17 23:09:18 +01:00
|
|
|
|
|
|
|
|
|
2022-04-12 15:09:48 +01:00
|
|
|
|
2021-08-27 17:46:09 +08:00
|
|
|
|
|
|
|
|
|
2021-10-04 20:02:54 +01:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-05-14 17:18:39 -06:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2022-06-16 10:22:10 +01:00
|
|
|
|
|
|
|
|
|
2024-09-30 17:11:32 -06:00
|
|
|
|
2024-09-30 14:30:39 -06:00
|
|
|
|
2022-06-16 10:22:10 +01:00
|
|
|
|
2024-09-30 17:11:32 -06:00
|
|
|
|
|
|
|
|
|
treewide: Replace kmalloc with kmalloc_obj for non-scalar types
This is the result of running the Coccinelle script from
scripts/coccinelle/api/kmalloc_objs.cocci. The script is designed to
avoid scalar types (which need careful case-by-case checking), and
instead replace kmalloc-family calls that allocate struct or union
object instances:
Single allocations: kmalloc(sizeof(TYPE), ...)
are replaced with: kmalloc_obj(TYPE, ...)
Array allocations: kmalloc_array(COUNT, sizeof(TYPE), ...)
are replaced with: kmalloc_objs(TYPE, COUNT, ...)
Flex array allocations: kmalloc(struct_size(PTR, FAM, COUNT), ...)
are replaced with: kmalloc_flex(*PTR, FAM, COUNT, ...)
(where TYPE may also be *VAR)
The resulting allocations no longer return "void *", instead returning
"TYPE *".
Signed-off-by: Kees Cook <kees@kernel.org>
2026-02-20 23:49:23 -08:00
|
|
|
|
|
|
|
|
|
2024-09-30 17:11:32 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 10:22:10 +01:00
|
|
|
|
|
|
|
|
|
2024-09-30 14:30:39 -06:00
|
|
|
|
|
|
|
|
|
2022-06-16 10:22:10 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-01-31 17:27:02 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-03-21 18:04:33 +00:00
|
|
|
|
2025-01-31 17:27:02 +00:00
|
|
|
|
2025-02-27 14:39:16 -08:00
|
|
|
|
2025-01-31 17:27:02 +00:00
|
|
|
|
|
|
|
|
|
2021-10-04 20:02:54 +01:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2022-05-01 10:52:44 -06:00
|
|
|
|
2024-03-20 15:19:44 -06:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2026-02-21 16:37:42 -08:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-05-01 10:52:44 -06:00
|
|
|
|
io_uring/rsrc: add huge page accounting for registered buffers
Track huge page references in a per-ring xarray to prevent double
accounting when the same huge page is used by multiple registered
buffers, either within the same ring or across cloned rings.
When registering buffers backed by huge pages, we need to account for
RLIMIT_MEMLOCK. But if multiple buffers share the same huge page (common
with cloned buffers), we must not account for the same page multiple
times. Similarly, we must only unaccount when the last reference to a
huge page is released.
Maintain a per-ring xarray (hpage_acct) that tracks reference counts for
each huge page. When registering a buffer, for each unique huge page,
increment its accounting reference count, and only account pages that
are newly added.
When unregistering a buffer, for each unique huge page, decrement its
refcount. Once the refcount hits zero, the page is unaccounted.
Note: any account is done against the ctx->user that was assigned when
the ring was setup. As before, if root is running the operation, no
accounting is done.
With these changes, any use of imu->acct_pages is also dead, hence kill
it from struct io_mapped_ubuf. This shrinks it from 56b to 48b on a
64-bit arch. Additionally, hpage_already_acct() is gone, which was an
O(M*M) scan over current + previous registrations.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2026-01-24 10:02:41 -07:00
|
|
|
|
2022-05-01 10:52:44 -06:00
|
|
|
|
2019-12-04 19:56:40 -07:00
|
|
|
|
|
|
|
|
|
2022-06-16 10:22:05 +01:00
|
|
|
|
|
|
|
|
|
2019-12-04 19:56:40 -07:00
|
|
|
|
2022-06-16 10:22:05 +01:00
|
|
|
|
|
|
|
|
|
2022-06-16 10:22:10 +01:00
|
|
|
|
2019-12-04 19:56:40 -07:00
|
|
|
|
2019-05-07 10:01:48 -07:00
|
|
|
|
2022-07-15 19:45:01 +02:00
|
|
|
|
2019-11-07 18:27:42 -07:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2024-11-01 17:19:57 +08:00
|
|
|
|
2024-01-17 00:57:29 +00:00
|
|
|
|
2020-09-03 12:12:41 -06:00
|
|
|
|
2020-09-14 11:16:23 -06:00
|
|
|
|
io_uring: add support for backlogged CQ ring
Currently we drop completion events, if the CQ ring is full. That's fine
for requests with bounded completion times, but it may make it harder or
impossible to use io_uring with networked IO where request completion
times are generally unbounded. Or with POLL, for example, which is also
unbounded.
After this patch, we never overflow the ring, we simply store requests
in a backlog for later flushing. This flushing is done automatically by
the kernel. To prevent the backlog from growing indefinitely, if the
backlog is non-empty, we apply back pressure on IO submissions. Any
attempt to submit new IO with a non-empty backlog will get an -EBUSY
return from the kernel. This is a signal to the application that it has
backlogged CQ events, and that it must reap those before being allowed
to submit more IO.
Note that if we do return -EBUSY, we will have filled whatever
backlogged events into the CQ ring first, if there's room. This means
the application can safely reap events WITHOUT entering the kernel and
waiting for them, they are already available in the CQ ring.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-06 11:31:17 -07:00
|
|
|
|
2024-10-26 06:43:44 -06:00
|
|
|
|
2025-01-22 20:00:57 -07:00
|
|
|
|
2024-03-20 15:19:44 -06:00
|
|
|
|
2025-01-22 20:00:57 -07:00
|
|
|
|
|
|
|
|
|
2024-03-20 15:19:44 -06:00
|
|
|
|
2025-01-22 20:00:57 -07:00
|
|
|
|
|
|
|
|
|
2025-03-19 06:12:47 +00:00
|
|
|
|
2025-03-21 18:04:33 +00:00
|
|
|
|
|
|
|
|
|
2024-03-20 15:19:44 -06:00
|
|
|
|
2025-02-27 14:39:16 -08:00
|
|
|
|
2024-03-20 15:19:44 -06:00
|
|
|
|
2024-09-23 18:05:12 +08:00
|
|
|
|
2020-05-14 17:18:39 -06:00
|
|
|
|
2021-03-08 14:16:16 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2021-06-14 23:37:28 +01:00
|
|
|
|
2023-01-09 14:46:08 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2024-12-17 08:21:46 -07:00
|
|
|
|
2025-12-11 03:25:41 -07:00
|
|
|
|
2019-04-06 21:51:27 -06:00
|
|
|
|
2019-09-17 12:26:57 -06:00
|
|
|
|
2021-08-28 19:54:38 -06:00
|
|
|
|
io_uring: switch local task_work to a mpscq
The local (DEFER_TASKRUN) task_work list is an llist, which is LIFO
ordered, and hence __io_run_local_work() has to restore the right
running order with an O(n) llist_reverse_order() pass first. On top of
that, a batch that gets capped by max_events needs the leftover entries
parked on a separate ->retry_llist, as they can't be pushed back to the
shared list.
Switch it to the FIFO mpscq. Adds are wait-free instead of a cmpxchg
retry loop, entries are popped in queue order with no reversal pass,
capping a run simply leaves the remainder on the queue, and
->retry_llist goes away entirely. The consumer cursor, ->work_head,
lives with the rest of the ->uring_lock protected state rather than
next to the queue, so that popping entries doesn't dirty the producer
side cacheline.
For low amounts of task_work, this ends up being a bit more efficient
than the existing scheme. As an example of that, doing multishot
receives for 8 clients has the following task_work overhead:
1.02% sock-test [kernel.kallsyms] [k] io_req_local_work_add
0.88% sock-test [kernel.kallsyms] [k] __io_run_local_work_loop
0.60% sock-test [kernel.kallsyms] [k] llist_reverse_order
0.14% sock-test [kernel.kallsyms] [k] __io_run_local_work
2.64% at ~46Gb/sec
and after this change:
1.08% sock-test [kernel.kallsyms] [k] io_req_local_work_add
1.03% sock-test [kernel.kallsyms] [k] __io_run_local_work
2.11% at ~53Gb/sec
which has less overhead even though that test run was faster. For a case
of having 1024 clients on a single ring:
2.22% sock-test [kernel.kallsyms] [k] llist_reverse_order
0.84% sock-test [kernel.kallsyms] [k] __io_run_local_work_loop
0.42% sock-test [kernel.kallsyms] [k] io_req_local_work_add
0.02% sock-test [kernel.kallsyms] [k] __io_run_local_work
3.50% at ~24Gb/sec
we start to see the llist reversing taking a considerable amount of
time, and the total add+run task_work overhead is around 3.5%. After
the change:
0.90% sock-test [kernel.kallsyms] [k] __io_run_local_work
0.42% sock-test [kernel.kallsyms] [k] io_req_local_work_add
1.32% at ~26Gb/sec
most of that overhead is gone, and performance is better as well.
Caleb Sander Mateos <csander@purestorage.com> reports that it improves
the performance of a ublk 4kb workload by 4% [1], while testing v1 of
this patchset.
[1] https://lore.kernel.org/io-uring/CADUfDZr-MMYBaP-e+y9+xuRhuiunO2sBTUCmwZyd7AgT8sVtiQ@mail.gmail.com/
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2026-06-10 15:19:35 -06:00
|
|
|
|
2021-03-06 11:02:12 +00:00
|
|
|
|
io_uring/tctx: add separate lock for list of tctx's in ctx
ctx->tcxt_list holds the tasks using this ring, and it's currently
protected by the normal ctx->uring_lock. However, this can cause a
circular locking issue, as reported by syzbot, where cancelations off
exec end up needing to remove an entry from this list:
======================================================
WARNING: possible circular locking dependency detected
syzkaller #0 Tainted: G L
------------------------------------------------------
syz.0.9999/12287 is trying to acquire lock:
ffff88805851c0a8 (&ctx->uring_lock){+.+.}-{4:4}, at: io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
but task is already holding lock:
ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: prepare_bprm_creds fs/exec.c:1360 [inline]
ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: bprm_execve+0xb9/0x1400 fs/exec.c:1733
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #2 (&sig->cred_guard_mutex){+.+.}-{4:4}:
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
proc_pid_attr_write+0x547/0x630 fs/proc/base.c:2837
vfs_write+0x27e/0xb30 fs/read_write.c:684
ksys_write+0x145/0x250 fs/read_write.c:738
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
-> #1 (sb_writers#3){.+.+}-{0:0}:
percpu_down_read_internal include/linux/percpu-rwsem.h:53 [inline]
percpu_down_read_freezable include/linux/percpu-rwsem.h:83 [inline]
__sb_start_write include/linux/fs/super.h:19 [inline]
sb_start_write+0x4d/0x1c0 include/linux/fs/super.h:125
mnt_want_write+0x41/0x90 fs/namespace.c:499
open_last_lookups fs/namei.c:4529 [inline]
path_openat+0xadd/0x3dd0 fs/namei.c:4784
do_filp_open+0x1fa/0x410 fs/namei.c:4814
io_openat2+0x3e0/0x5c0 io_uring/openclose.c:143
__io_issue_sqe+0x181/0x4b0 io_uring/io_uring.c:1792
io_issue_sqe+0x165/0x1060 io_uring/io_uring.c:1815
io_queue_sqe io_uring/io_uring.c:2042 [inline]
io_submit_sqe io_uring/io_uring.c:2320 [inline]
io_submit_sqes+0xbf4/0x2140 io_uring/io_uring.c:2434
__do_sys_io_uring_enter io_uring/io_uring.c:3280 [inline]
__se_sys_io_uring_enter+0x2e0/0x2b60 io_uring/io_uring.c:3219
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
-> #0 (&ctx->uring_lock){+.+.}-{4:4}:
check_prev_add kernel/locking/lockdep.c:3165 [inline]
check_prevs_add kernel/locking/lockdep.c:3284 [inline]
validate_chain kernel/locking/lockdep.c:3908 [inline]
__lock_acquire+0x15a6/0x2cf0 kernel/locking/lockdep.c:5237
lock_acquire+0x107/0x340 kernel/locking/lockdep.c:5868
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
io_uring_clean_tctx+0xd4/0x1a0 io_uring/tctx.c:195
io_uring_cancel_generic+0x6ca/0x7d0 io_uring/cancel.c:646
io_uring_task_cancel include/linux/io_uring.h:24 [inline]
begin_new_exec+0x10ed/0x2440 fs/exec.c:1131
load_elf_binary+0x9f8/0x2d70 fs/binfmt_elf.c:1010
search_binary_handler fs/exec.c:1669 [inline]
exec_binprm fs/exec.c:1701 [inline]
bprm_execve+0x92e/0x1400 fs/exec.c:1753
do_execveat_common+0x510/0x6a0 fs/exec.c:1859
do_execve fs/exec.c:1933 [inline]
__do_sys_execve fs/exec.c:2009 [inline]
__se_sys_execve fs/exec.c:2004 [inline]
__x64_sys_execve+0x94/0xb0 fs/exec.c:2004
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
other info that might help us debug this:
Chain exists of:
&ctx->uring_lock --> sb_writers#3 --> &sig->cred_guard_mutex
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&sig->cred_guard_mutex);
lock(sb_writers#3);
lock(&sig->cred_guard_mutex);
lock(&ctx->uring_lock);
*** DEADLOCK ***
1 lock held by syz.0.9999/12287:
#0: ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: prepare_bprm_creds fs/exec.c:1360 [inline]
#0: ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: bprm_execve+0xb9/0x1400 fs/exec.c:1733
stack backtrace:
CPU: 0 UID: 0 PID: 12287 Comm: syz.0.9999 Tainted: G L syzkaller #0 PREEMPT(full)
Tainted: [L]=SOFTLOCKUP
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/25/2025
Call Trace:
<TASK>
dump_stack_lvl+0xe8/0x150 lib/dump_stack.c:120
print_circular_bug+0x2e2/0x300 kernel/locking/lockdep.c:2043
check_noncircular+0x12e/0x150 kernel/locking/lockdep.c:2175
check_prev_add kernel/locking/lockdep.c:3165 [inline]
check_prevs_add kernel/locking/lockdep.c:3284 [inline]
validate_chain kernel/locking/lockdep.c:3908 [inline]
__lock_acquire+0x15a6/0x2cf0 kernel/locking/lockdep.c:5237
lock_acquire+0x107/0x340 kernel/locking/lockdep.c:5868
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
io_uring_clean_tctx+0xd4/0x1a0 io_uring/tctx.c:195
io_uring_cancel_generic+0x6ca/0x7d0 io_uring/cancel.c:646
io_uring_task_cancel include/linux/io_uring.h:24 [inline]
begin_new_exec+0x10ed/0x2440 fs/exec.c:1131
load_elf_binary+0x9f8/0x2d70 fs/binfmt_elf.c:1010
search_binary_handler fs/exec.c:1669 [inline]
exec_binprm fs/exec.c:1701 [inline]
bprm_execve+0x92e/0x1400 fs/exec.c:1753
do_execveat_common+0x510/0x6a0 fs/exec.c:1859
do_execve fs/exec.c:1933 [inline]
__do_sys_execve fs/exec.c:2009 [inline]
__se_sys_execve fs/exec.c:2004 [inline]
__x64_sys_execve+0x94/0xb0 fs/exec.c:2004
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7ff3a8b8f749
Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 a8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007ff3a9a97038 EFLAGS: 00000246 ORIG_RAX: 000000000000003b
RAX: ffffffffffffffda RBX: 00007ff3a8de5fa0 RCX: 00007ff3a8b8f749
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000200000000400
RBP: 00007ff3a8c13f91 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 00007ff3a8de6038 R14: 00007ff3a8de5fa0 R15: 00007ff3a8f0fa28
</TASK>
Add a separate lock just for the tctx_list, tctx_lock. This can nest
under ->uring_lock, where necessary, and be used separately for list
manipulation. For the cancelation off exec side, this removes the
need to grab ->uring_lock, hence fixing the circular locking
dependency.
Reported-by: syzbot+b0e3b77ffaa8a4067ce5@syzkaller.appspotmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-12-31 08:12:46 -07:00
|
|
|
|
2021-09-24 21:59:47 +01:00
|
|
|
|
2023-07-10 16:14:37 -06:00
|
|
|
|
2025-04-20 10:31:20 +01:00
|
|
|
|
io_uring: add support for futex wake and wait
Add support for FUTEX_WAKE/WAIT primitives.
IORING_OP_FUTEX_WAKE is mix of FUTEX_WAKE and FUTEX_WAKE_BITSET, as
it does support passing in a bitset.
Similary, IORING_OP_FUTEX_WAIT is a mix of FUTEX_WAIT and
FUTEX_WAIT_BITSET.
For both of them, they are using the futex2 interface.
FUTEX_WAKE is straight forward, as those can always be done directly from
the io_uring submission without needing async handling. For FUTEX_WAIT,
things are a bit more complicated. If the futex isn't ready, then we
rely on a callback via futex_queue->wake() when someone wakes up the
futex. From that calback, we queue up task_work with the original task,
which will post a CQE and wake it, if necessary.
Cancelations are supported, both from the application point-of-view,
but also to be able to cancel pending waits if the ring exits before
all events have occurred. The return value of futex_unqueue() is used
to gate who wins the potential race between cancelation and futex
wakeups. Whomever gets a 'ret == 1' return from that claims ownership
of the io_uring futex request.
This is just the barebones wait/wake support. PI or REQUEUE support is
not added at this point, unclear if we might look into that later.
Likewise, explicit timeouts are not supported either. It is expected
that users that need timeouts would do so via the usual io_uring
mechanism to do that using linked timeouts.
The SQE format is as follows:
`addr` Address of futex
`fd` futex2(2) FUTEX2_* flags
`futex_flags` io_uring specific command flags. None valid now.
`addr2` Value of futex
`addr3` Mask to wake/wait
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2023-06-08 11:57:40 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-09-24 21:59:44 +01:00
|
|
|
|
2023-09-28 20:43:25 +08:00
|
|
|
|
2023-06-08 09:38:36 -07:00
|
|
|
|
2024-11-29 13:34:22 +00:00
|
|
|
|
2023-06-08 09:38:36 -07:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2024-09-23 18:05:12 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-11-07 18:27:42 -07:00
|
|
|
|
2025-01-31 17:27:02 +00:00
|
|
|
|
2024-09-30 17:11:32 -06:00
|
|
|
|
2022-05-01 10:52:44 -06:00
|
|
|
|
io_uring/rsrc: add huge page accounting for registered buffers
Track huge page references in a per-ring xarray to prevent double
accounting when the same huge page is used by multiple registered
buffers, either within the same ring or across cloned rings.
When registering buffers backed by huge pages, we need to account for
RLIMIT_MEMLOCK. But if multiple buffers share the same huge page (common
with cloned buffers), we must not account for the same page multiple
times. Similarly, we must only unaccount when the last reference to a
huge page is released.
Maintain a per-ring xarray (hpage_acct) that tracks reference counts for
each huge page. When registering a buffer, for each unique huge page,
increment its accounting reference count, and only account pages that
are newly added.
When unregistering a buffer, for each unique huge page, decrement its
refcount. Once the refcount hits zero, the page is unaccounted.
Note: any account is done against the ctx->user that was assigned when
the ring was setup. As before, if root is running the operation, no
accounting is done.
With these changes, any use of imu->acct_pages is also dead, hence kill
it from struct io_mapped_ubuf. This shrinks it from 56b to 48b on a
64-bit arch. Additionally, hpage_already_acct() is gone, which was an
O(M*M) scan over current + previous registrations.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2026-01-24 10:02:41 -07:00
|
|
|
|
2019-11-07 18:27:42 -07:00
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2023-06-23 12:23:24 +01:00
|
|
|
|
|
|
|
|
|
2025-02-05 11:36:44 +00:00
|
|
|
|
2025-02-05 11:36:48 +00:00
|
|
|
|
2023-06-23 12:23:24 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: move struct io_kiocb from task_struct to io_uring_task
Rather than store the task_struct itself in struct io_kiocb, store
the io_uring specific task_struct. The life times are the same in terms
of io_uring, and this avoids doing some dereferences through the
task_struct. For the hot path of putting local task references, we can
deref req->tctx instead, which we'll need anyway in that function
regardless of whether it's local or remote references.
This is mostly straight forward, except the original task PF_EXITING
check needs a bit of tweaking. task_work is _always_ run from the
originating task, except in the fallback case, where it's run from a
kernel thread. Replace the potentially racy (in case of fallback work)
checks for req->task->flags with current->flags. It's either the still
the original task, in which case PF_EXITING will be sane, or it has
PF_KTHREAD set, in which case it's fallback work. Both cases should
prevent moving forward with the given request.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-11-03 10:23:38 -07:00
|
|
|
|
|
|
|
|
|
2023-06-23 12:23:24 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-06-04 10:25:42 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-01 23:57:02 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: move struct io_kiocb from task_struct to io_uring_task
Rather than store the task_struct itself in struct io_kiocb, store
the io_uring specific task_struct. The life times are the same in terms
of io_uring, and this avoids doing some dereferences through the
task_struct. For the hot path of putting local task references, we can
deref req->tctx instead, which we'll need anyway in that function
regardless of whether it's local or remote references.
This is mostly straight forward, except the original task PF_EXITING
check needs a bit of tweaking. task_work is _always_ run from the
originating task, except in the fallback case, where it's run from a
kernel thread. Replace the potentially racy (in case of fallback work)
checks for req->task->flags with current->flags. It's either the still
the original task, in which case PF_EXITING will be sane, or it has
PF_KTHREAD set, in which case it's fallback work. Both cases should
prevent moving forward with the given request.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-11-03 10:23:38 -07:00
|
|
|
|
2022-06-01 23:57:02 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-08-11 19:28:31 +01:00
|
|
|
|
|
|
|
|
|
2021-08-15 10:40:26 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-08-15 10:40:24 +01:00
|
|
|
|
|
|
|
|
|
2021-08-11 19:28:31 +01:00
|
|
|
|
|
|
|
|
|
2021-08-15 10:40:18 +01:00
|
|
|
|
2021-08-15 10:40:24 +01:00
|
|
|
|
|
|
|
|
|
2021-08-11 19:28:31 +01:00
|
|
|
|
|
|
|
|
|
2020-06-29 19:18:43 +03:00
|
|
|
|
2019-10-24 07:25:42 -06:00
|
|
|
|
2023-01-12 06:44:10 -08:00
|
|
|
|
2019-09-10 09:15:04 -06:00
|
|
|
|
2021-06-17 18:14:02 +01:00
|
|
|
|
|
|
|
|
|
2021-06-17 18:14:01 +01:00
|
|
|
|
2021-06-17 18:14:02 +01:00
|
|
|
|
2021-03-06 09:22:27 -07:00
|
|
|
|
2021-03-22 01:58:29 +00:00
|
|
|
|
2024-06-13 19:28:27 +00:00
|
|
|
|
2020-10-22 16:47:16 +01:00
|
|
|
|
2024-06-13 19:28:27 +00:00
|
|
|
|
2020-10-22 16:47:16 +01:00
|
|
|
|
2023-06-20 13:32:31 +02:00
|
|
|
|
2023-06-20 13:32:32 +02:00
|
|
|
|
2022-07-21 09:06:47 -06:00
|
|
|
|
2023-04-11 12:06:01 +01:00
|
|
|
|
2023-03-07 09:47:20 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-03-28 13:27:24 +01:00
|
|
|
|
2023-03-07 09:47:20 -07:00
|
|
|
|
2026-03-02 10:29:10 -07:00
|
|
|
|
2020-03-14 00:31:04 +03:00
|
|
|
|
2021-04-01 08:38:34 -06:00
|
|
|
|
2019-12-18 09:50:26 -07:00
|
|
|
|
2024-06-13 19:28:27 +00:00
|
|
|
|
2019-09-10 09:15:04 -06:00
|
|
|
|
2019-10-24 07:25:42 -06:00
|
|
|
|
|
|
|
|
|
2020-06-29 19:18:43 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-07-26 14:14:31 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-12-17 08:21:46 -07:00
|
|
|
|
2021-07-26 14:14:31 +01:00
|
|
|
|
|
|
|
|
|
2024-12-17 08:21:46 -07:00
|
|
|
|
2021-07-26 14:14:31 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-06-29 19:18:43 +03:00
|
|
|
|
|
|
|
|
|
2024-03-18 22:00:28 +00:00
|
|
|
|
2019-10-24 07:25:42 -06:00
|
|
|
|
io_uring: move struct io_kiocb from task_struct to io_uring_task
Rather than store the task_struct itself in struct io_kiocb, store
the io_uring specific task_struct. The life times are the same in terms
of io_uring, and this avoids doing some dereferences through the
task_struct. For the hot path of putting local task references, we can
deref req->tctx instead, which we'll need anyway in that function
regardless of whether it's local or remote references.
This is mostly straight forward, except the original task PF_EXITING
check needs a bit of tweaking. task_work is _always_ run from the
originating task, except in the fallback case, where it's run from a
kernel thread. Replace the potentially racy (in case of fallback work)
checks for req->task->flags with current->flags. It's either the still
the original task, in which case PF_EXITING will be sane, or it has
PF_KTHREAD set, in which case it's fallback work. Both cases should
prevent moving forward with the given request.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-11-03 10:23:38 -07:00
|
|
|
|
2019-10-24 07:25:42 -06:00
|
|
|
|
2021-02-16 14:15:30 -07:00
|
|
|
|
2024-12-19 19:52:58 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-10-24 07:25:42 -06:00
|
|
|
|
2020-06-29 19:18:43 +03:00
|
|
|
|
|
|
|
|
|
2021-07-23 11:53:54 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: move struct io_kiocb from task_struct to io_uring_task
Rather than store the task_struct itself in struct io_kiocb, store
the io_uring specific task_struct. The life times are the same in terms
of io_uring, and this avoids doing some dereferences through the
task_struct. For the hot path of putting local task references, we can
deref req->tctx instead, which we'll need anyway in that function
regardless of whether it's local or remote references.
This is mostly straight forward, except the original task PF_EXITING
check needs a bit of tweaking. task_work is _always_ run from the
originating task, except in the fallback case, where it's run from a
kernel thread. Replace the potentially racy (in case of fallback work)
checks for req->task->flags with current->flags. It's either the still
the original task, in which case PF_EXITING will be sane, or it has
PF_KTHREAD set, in which case it's fallback work. Both cases should
prevent moving forward with the given request.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-11-03 10:23:38 -07:00
|
|
|
|
2024-06-13 19:28:27 +00:00
|
|
|
|
2021-07-23 11:53:54 -06:00
|
|
|
|
2022-06-16 13:57:20 +01:00
|
|
|
|
2021-03-01 18:20:47 +00:00
|
|
|
|
2020-06-29 19:18:43 +03:00
|
|
|
|
|
|
|
|
|
2025-10-31 14:34:29 -06:00
|
|
|
|
2024-09-11 17:34:37 +01:00
|
|
|
|
2025-10-31 14:34:29 -06:00
|
|
|
|
2024-09-11 17:34:37 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-03 12:39:05 -07:00
|
|
|
|
2025-05-09 12:12:48 +01:00
|
|
|
|
2025-05-09 12:12:54 +01:00
|
|
|
|
|
|
|
|
|
2025-05-09 12:12:48 +01:00
|
|
|
|
2025-05-09 12:12:54 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-05-09 12:12:48 +01:00
|
|
|
|
|
|
|
|
|
2024-12-30 17:34:41 -07:00
|
|
|
|
2020-05-26 20:34:05 +03:00
|
|
|
|
2025-05-09 12:12:48 +01:00
|
|
|
|
|
|
|
|
|
2025-05-09 12:12:54 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-06-14 23:37:31 +01:00
|
|
|
|
2020-07-13 23:37:14 +03:00
|
|
|
|
|
|
|
|
|
2020-05-26 20:34:05 +03:00
|
|
|
|
2025-05-09 12:12:48 +01:00
|
|
|
|
2025-05-09 12:12:54 +01:00
|
|
|
|
|
|
|
|
|
2025-05-09 12:12:48 +01:00
|
|
|
|
2020-07-13 23:37:14 +03:00
|
|
|
|
2025-05-09 12:12:54 +01:00
|
|
|
|
2021-01-26 23:35:10 +00:00
|
|
|
|
2020-07-13 23:37:14 +03:00
|
|
|
|
2025-05-09 12:12:48 +01:00
|
|
|
|
2021-06-14 23:37:31 +01:00
|
|
|
|
2020-05-26 20:34:05 +03:00
|
|
|
|
|
|
|
|
|
2022-06-19 12:26:06 +01:00
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2023-01-09 14:46:09 +00:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2022-12-02 17:47:24 +00:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2025-04-24 12:31:16 +01:00
|
|
|
|
2022-06-19 12:26:06 +01:00
|
|
|
|
|
|
|
|
|
2022-12-07 08:50:01 -07:00
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2022-12-07 08:50:01 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-12-02 17:47:23 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-12-07 08:50:01 -07:00
|
|
|
|
2023-01-09 14:46:10 +00:00
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
|
|
|
|
|
2023-08-24 23:53:29 +01:00
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2023-08-24 23:53:29 +01:00
|
|
|
|
2023-01-09 14:46:10 +00:00
|
|
|
|
2023-08-24 23:53:28 +01:00
|
|
|
|
2023-01-09 14:46:10 +00:00
|
|
|
|
|
|
|
|
|
2023-06-23 12:23:30 +01:00
|
|
|
|
2022-11-24 19:46:41 +00:00
|
|
|
|
2022-06-20 01:25:56 +01:00
|
|
|
|
2022-12-07 08:50:01 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-24 23:53:28 +01:00
|
|
|
|
2022-06-20 01:25:56 +01:00
|
|
|
|
|
|
|
|
|
2024-04-10 02:26:55 +01:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2024-04-10 02:26:54 +01:00
|
|
|
|
|
|
|
|
|
2024-04-12 13:16:20 -06:00
|
|
|
|
|
|
|
|
|
2022-12-07 03:53:29 +00:00
|
|
|
|
io_uring: add support for backlogged CQ ring
Currently we drop completion events, if the CQ ring is full. That's fine
for requests with bounded completion times, but it may make it harder or
impossible to use io_uring with networked IO where request completion
times are generally unbounded. Or with POLL, for example, which is also
unbounded.
After this patch, we never overflow the ring, we simply store requests
in a backlog for later flushing. This flushing is done automatically by
the kernel. To prevent the backlog from growing indefinitely, if the
backlog is non-empty, we apply back pressure on IO submissions. Any
attempt to submit new IO with a non-empty backlog will get an -EBUSY
return from the kernel. This is a signal to the application that it has
backlogged CQ events, and that it must reap those before being allowed
to submit more IO.
Note that if we do return -EBUSY, we will have filled whatever
backlogged events into the CQ ring first, if there's room. This means
the application can safely reap events WITHOUT entering the kernel and
waiting for them, they are already available in the CQ ring.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-06 11:31:17 -07:00
|
|
|
|
2022-06-20 01:25:56 +01:00
|
|
|
|
2021-02-23 12:40:22 +00:00
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
2023-08-24 23:53:27 +01:00
|
|
|
|
2021-02-23 12:40:22 +00:00
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
2020-09-28 13:10:13 -06:00
|
|
|
|
2021-02-23 12:40:22 +00:00
|
|
|
|
|
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-25 12:33:06 +00:00
|
|
|
|
|
|
|
|
|
2024-04-10 02:26:55 +01:00
|
|
|
|
|
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
2024-04-10 02:26:55 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-23 12:40:22 +00:00
|
|
|
|
|
|
|
|
|
2024-09-20 02:51:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-05-17 13:27:37 +01:00
|
|
|
|
2024-09-20 02:51:20 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add support for backlogged CQ ring
Currently we drop completion events, if the CQ ring is full. That's fine
for requests with bounded completion times, but it may make it harder or
impossible to use io_uring with networked IO where request completion
times are generally unbounded. Or with POLL, for example, which is also
unbounded.
After this patch, we never overflow the ring, we simply store requests
in a backlog for later flushing. This flushing is done automatically by
the kernel. To prevent the backlog from growing indefinitely, if the
backlog is non-empty, we apply back pressure on IO submissions. Any
attempt to submit new IO with a non-empty backlog will get an -EBUSY
return from the kernel. This is a signal to the application that it has
backlogged CQ events, and that it must reap those before being allowed
to submit more IO.
Note that if we do return -EBUSY, we will have filled whatever
backlogged events into the CQ ring first, if there's room. This means
the application can safely reap events WITHOUT entering the kernel and
waiting for them, they are already available in the CQ ring.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-06 11:31:17 -07:00
|
|
|
|
|
|
|
|
|
2022-12-07 03:53:29 +00:00
|
|
|
|
2022-04-21 02:13:43 -07:00
|
|
|
|
2022-04-25 19:49:00 -06:00
|
|
|
|
2020-12-17 00:24:38 +00:00
|
|
|
|
2022-06-20 01:25:56 +01:00
|
|
|
|
io_uring: add support for backlogged CQ ring
Currently we drop completion events, if the CQ ring is full. That's fine
for requests with bounded completion times, but it may make it harder or
impossible to use io_uring with networked IO where request completion
times are generally unbounded. Or with POLL, for example, which is also
unbounded.
After this patch, we never overflow the ring, we simply store requests
in a backlog for later flushing. This flushing is done automatically by
the kernel. To prevent the backlog from growing indefinitely, if the
backlog is non-empty, we apply back pressure on IO submissions. Any
attempt to submit new IO with a non-empty backlog will get an -EBUSY
return from the kernel. This is a signal to the application that it has
backlogged CQ events, and that it must reap those before being allowed
to submit more IO.
Note that if we do return -EBUSY, we will have filled whatever
backlogged events into the CQ ring first, if there's room. This means
the application can safely reap events WITHOUT entering the kernel and
waiting for them, they are already available in the CQ ring.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-06 11:31:17 -07:00
|
|
|
|
|
|
|
|
|
2024-04-10 02:26:55 +01:00
|
|
|
|
2022-12-21 07:05:09 -07:00
|
|
|
|
2024-04-10 02:26:55 +01:00
|
|
|
|
|
|
|
|
|
2022-12-21 07:05:09 -07:00
|
|
|
|
|
|
|
|
|
2026-01-22 08:40:19 -07:00
|
|
|
|
2021-01-04 20:36:36 +00:00
|
|
|
|
2024-04-10 02:26:54 +01:00
|
|
|
|
2024-04-10 02:26:55 +01:00
|
|
|
|
2024-04-10 02:26:54 +01:00
|
|
|
|
2021-01-04 20:36:36 +00:00
|
|
|
|
|
|
|
|
|
io_uring: introduce callback driven main loop
The io_uring_enter() has a fixed order of execution: it submits
requests, waits for completions, and returns to the user. Allow to
optionally replace it with a custom loop driven by a callback called
loop_step. The basic requirements to the callback is that it should be
able to submit requests, wait for completions, parse them and repeat.
Most of the communication including parameter passing can be implemented
via shared memory.
The callback should return IOU_LOOP_CONTINUE to continue execution or
IOU_LOOP_STOP to return to the user space. Note that the kernel may
decide to prematurely terminate it as well, e.g. in case the process was
signalled or killed.
The hook takes a structure with parameters. It can be used to ask the
kernel to wait for CQEs by setting cq_wait_idx to the CQE index it wants
to wait for. Spurious wake ups are possible and even likely, the callback
is expected to handle it. There will be more parameters in the future
like timeout.
It can be used with kernel callbacks, for example, as a slow path
deprecation mechanism overwiting SQEs and emulating the wanted
behaviour, however it's more useful together with BPF programs
implemented in following patches.
Note that keeping it separately from the normal io_uring wait loop
makes things much simpler and cleaner. It keeps it in one place instead
of spreading a bunch of checks in different places including disabling
the submission path. It holds the lock by default, which is a better fit
for BPF synchronisation and the loop execution model. It nicely avoids
existing quirks like forced wake ups on timeout request completion. And
it should be easier to implement new features.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://patch.msgid.link/a2d369aa1c9dd23ad7edac9220cffc563abcaed6.1772109579.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2026-02-26 12:48:38 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-01-16 16:48:58 +00:00
|
|
|
|
io_uring: move struct io_kiocb from task_struct to io_uring_task
Rather than store the task_struct itself in struct io_kiocb, store
the io_uring specific task_struct. The life times are the same in terms
of io_uring, and this avoids doing some dereferences through the
task_struct. For the hot path of putting local task references, we can
deref req->tctx instead, which we'll need anyway in that function
regardless of whether it's local or remote references.
This is mostly straight forward, except the original task PF_EXITING
check needs a bit of tweaking. task_work is _always_ run from the
originating task, except in the fallback case, where it's run from a
kernel thread. Replace the potentially racy (in case of fallback work)
checks for req->task->flags with current->flags. It's either the still
the original task, in which case PF_EXITING will be sane, or it has
PF_KTHREAD set, in which case it's fallback work. Both cases should
prevent moving forward with the given request.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-11-03 10:23:38 -07:00
|
|
|
|
2023-01-16 16:48:58 +00:00
|
|
|
|
io_uring: move struct io_kiocb from task_struct to io_uring_task
Rather than store the task_struct itself in struct io_kiocb, store
the io_uring specific task_struct. The life times are the same in terms
of io_uring, and this avoids doing some dereferences through the
task_struct. For the hot path of putting local task references, we can
deref req->tctx instead, which we'll need anyway in that function
regardless of whether it's local or remote references.
This is mostly straight forward, except the original task PF_EXITING
check needs a bit of tweaking. task_work is _always_ run from the
originating task, except in the fallback case, where it's run from a
kernel thread. Replace the potentially racy (in case of fallback work)
checks for req->task->flags with current->flags. It's either the still
the original task, in which case PF_EXITING will be sane, or it has
PF_KTHREAD set, in which case it's fallback work. Both cases should
prevent moving forward with the given request.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-11-03 10:23:38 -07:00
|
|
|
|
2024-11-02 21:26:16 -06:00
|
|
|
|
io_uring: move struct io_kiocb from task_struct to io_uring_task
Rather than store the task_struct itself in struct io_kiocb, store
the io_uring specific task_struct. The life times are the same in terms
of io_uring, and this avoids doing some dereferences through the
task_struct. For the hot path of putting local task references, we can
deref req->tctx instead, which we'll need anyway in that function
regardless of whether it's local or remote references.
This is mostly straight forward, except the original task PF_EXITING
check needs a bit of tweaking. task_work is _always_ run from the
originating task, except in the fallback case, where it's run from a
kernel thread. Replace the potentially racy (in case of fallback work)
checks for req->task->flags with current->flags. It's either the still
the original task, in which case PF_EXITING will be sane, or it has
PF_KTHREAD set, in which case it's fallback work. Both cases should
prevent moving forward with the given request.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-11-03 10:23:38 -07:00
|
|
|
|
2024-11-02 21:26:16 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: move struct io_kiocb from task_struct to io_uring_task
Rather than store the task_struct itself in struct io_kiocb, store
the io_uring specific task_struct. The life times are the same in terms
of io_uring, and this avoids doing some dereferences through the
task_struct. For the hot path of putting local task references, we can
deref req->tctx instead, which we'll need anyway in that function
regardless of whether it's local or remote references.
This is mostly straight forward, except the original task PF_EXITING
check needs a bit of tweaking. task_work is _always_ run from the
originating task, except in the fallback case, where it's run from a
kernel thread. Replace the potentially racy (in case of fallback work)
checks for req->task->flags with current->flags. It's either the still
the original task, in which case PF_EXITING will be sane, or it has
PF_KTHREAD set, in which case it's fallback work. Both cases should
prevent moving forward with the given request.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-11-03 10:23:38 -07:00
|
|
|
|
2024-11-02 21:26:16 -06:00
|
|
|
|
2023-01-16 16:48:58 +00:00
|
|
|
|
|
|
|
|
|
2022-07-12 21:52:47 +01:00
|
|
|
|
2021-08-27 11:55:01 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-03 12:39:05 -07:00
|
|
|
|
2022-01-09 00:53:22 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-05-16 13:39:32 -06:00
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2023-01-04 01:34:57 +00:00
|
|
|
|
|
|
|
|
|
2021-04-13 02:58:44 +01:00
|
|
|
|
2025-05-09 12:12:52 +01:00
|
|
|
|
|
|
|
|
|
2021-04-13 02:58:44 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-05-09 12:12:52 +01:00
|
|
|
|
2022-04-21 02:13:44 -07:00
|
|
|
|
2021-04-13 02:58:44 +01:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2021-04-13 02:58:44 +01:00
|
|
|
|
2022-04-21 02:13:43 -07:00
|
|
|
|
2022-04-25 19:49:00 -06:00
|
|
|
|
2021-08-07 17:13:42 -07:00
|
|
|
|
2021-04-13 02:58:44 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2025-05-16 09:39:14 -06:00
|
|
|
|
2025-05-16 13:16:44 -06:00
|
|
|
|
|
|
|
|
|
2022-06-17 09:48:02 +01:00
|
|
|
|
2025-05-16 09:39:14 -06:00
|
|
|
|
|
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
2025-05-16 09:39:14 -06:00
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
|
|
|
|
|
2025-05-16 09:39:14 -06:00
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
2025-05-16 09:39:14 -06:00
|
|
|
|
2025-05-16 13:16:44 -06:00
|
|
|
|
2025-05-16 09:45:24 -06:00
|
|
|
|
2025-05-16 09:39:14 -06:00
|
|
|
|
2025-05-16 09:45:24 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-05-16 13:16:44 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-05-16 09:39:14 -06:00
|
|
|
|
|
|
|
|
|
2025-05-16 13:16:44 -06:00
|
|
|
|
|
|
|
|
|
2025-05-16 09:39:14 -06:00
|
|
|
|
2022-06-17 09:48:02 +01:00
|
|
|
|
|
|
|
|
|
io_uring: validate user-controlled cq.head in io_cqe_cache_refill()
A fuzzing run reproduced an unkillable io_uring task stuck at ~100% CPU:
[root@fedora io_uring_stress]# ps -ef | grep io_uring
root 1240 1 99 13:36 ? 00:01:35 [io_uring_stress] <defunct>
The task loops inside io_cqring_wait() and never returns to userspace,
and SIGKILL has no effect.
This is caused by the CQ ring exposing rings->cq.head to userspace as
writable, while the authoritative tail lives in kernel-private
ctx->cached_cq_tail. io_cqe_cache_refill() computes free space as an
unsigned subtraction:
free = ctx->cq_entries - min(tail - head, ctx->cq_entries);
If userspace keeps head within [0, tail], the subtraction is well
defined and min() just acts as a defensive clamp. But if userspace
advances head past tail, (tail - head) wraps to a huge value, free
becomes 0, and io_cqe_cache_refill() fails. The CQE is pushed onto the
overflow list and IO_CHECK_CQ_OVERFLOW_BIT is set.
The wait loop in io_cqring_wait() relies on an invariant: refill() only
fails when the CQ is *physically* full, in which case rings->cq.tail has
been advanced to iowq->cq_tail and io_should_wake() returns true. The
tampered head breaks this: refill() fails while the ring is not full, no
OCQE is copied in, rings->cq.tail never catches up, io_should_wake()
stays false, and io_cqring_wait_schedule() keeps returning early because
IO_CHECK_CQ_OVERFLOW_BIT is still set. The result is a tight retry loop
that never returns to userspace.
Introduce io_cqring_queued() as the single point that converts the
(tail, head) pair into a trustworthy queued count. Since the real
head/tail distance is bounded by cq_entries (far below 2^31), a signed
comparison reliably detects userspace moving head past tail; in that
case treat the queue as empty so callers see the full cache as free and
forward progress is preserved.
Suggested-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Zizhi Wo <wozizhi@huawei.com>
Link: https://patch.msgid.link/20260514021847.4062782-1-wozizhi@huaweicloud.com
[axboe: fixup commit message, kill 'queued' var, and keep it all in
io_uring.c]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2026-05-14 10:18:47 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: validate user-controlled cq.head in io_cqe_cache_refill()
A fuzzing run reproduced an unkillable io_uring task stuck at ~100% CPU:
[root@fedora io_uring_stress]# ps -ef | grep io_uring
root 1240 1 99 13:36 ? 00:01:35 [io_uring_stress] <defunct>
The task loops inside io_cqring_wait() and never returns to userspace,
and SIGKILL has no effect.
This is caused by the CQ ring exposing rings->cq.head to userspace as
writable, while the authoritative tail lives in kernel-private
ctx->cached_cq_tail. io_cqe_cache_refill() computes free space as an
unsigned subtraction:
free = ctx->cq_entries - min(tail - head, ctx->cq_entries);
If userspace keeps head within [0, tail], the subtraction is well
defined and min() just acts as a defensive clamp. But if userspace
advances head past tail, (tail - head) wraps to a huge value, free
becomes 0, and io_cqe_cache_refill() fails. The CQE is pushed onto the
overflow list and IO_CHECK_CQ_OVERFLOW_BIT is set.
The wait loop in io_cqring_wait() relies on an invariant: refill() only
fails when the CQ is *physically* full, in which case rings->cq.tail has
been advanced to iowq->cq_tail and io_should_wake() returns true. The
tampered head breaks this: refill() fails while the ring is not full, no
OCQE is copied in, rings->cq.tail never catches up, io_should_wake()
stays false, and io_cqring_wait_schedule() keeps returning early because
IO_CHECK_CQ_OVERFLOW_BIT is still set. The result is a tight retry loop
that never returns to userspace.
Introduce io_cqring_queued() as the single point that converts the
(tail, head) pair into a trustworthy queued count. Since the real
head/tail distance is bounded by cq_entries (far below 2^31), a signed
comparison reliably detects userspace moving head past tail; in that
case treat the queue as empty so callers see the full cache as free and
forward progress is preserved.
Suggested-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Zizhi Wo <wozizhi@huawei.com>
Link: https://patch.msgid.link/20260514021847.4062782-1-wozizhi@huaweicloud.com
[axboe: fixup commit message, kill 'queued' var, and keep it all in
io_uring.c]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2026-05-14 10:18:47 +08:00
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-17 09:48:01 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
2022-06-17 09:48:01 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: validate user-controlled cq.head in io_cqe_cache_refill()
A fuzzing run reproduced an unkillable io_uring task stuck at ~100% CPU:
[root@fedora io_uring_stress]# ps -ef | grep io_uring
root 1240 1 99 13:36 ? 00:01:35 [io_uring_stress] <defunct>
The task loops inside io_cqring_wait() and never returns to userspace,
and SIGKILL has no effect.
This is caused by the CQ ring exposing rings->cq.head to userspace as
writable, while the authoritative tail lives in kernel-private
ctx->cached_cq_tail. io_cqe_cache_refill() computes free space as an
unsigned subtraction:
free = ctx->cq_entries - min(tail - head, ctx->cq_entries);
If userspace keeps head within [0, tail], the subtraction is well
defined and min() just acts as a defensive clamp. But if userspace
advances head past tail, (tail - head) wraps to a huge value, free
becomes 0, and io_cqe_cache_refill() fails. The CQE is pushed onto the
overflow list and IO_CHECK_CQ_OVERFLOW_BIT is set.
The wait loop in io_cqring_wait() relies on an invariant: refill() only
fails when the CQ is *physically* full, in which case rings->cq.tail has
been advanced to iowq->cq_tail and io_should_wake() returns true. The
tampered head breaks this: refill() fails while the ring is not full, no
OCQE is copied in, rings->cq.tail never catches up, io_should_wake()
stays false, and io_cqring_wait_schedule() keeps returning early because
IO_CHECK_CQ_OVERFLOW_BIT is still set. The result is a tight retry loop
that never returns to userspace.
Introduce io_cqring_queued() as the single point that converts the
(tail, head) pair into a trustworthy queued count. Since the real
head/tail distance is bounded by cq_entries (far below 2^31), a signed
comparison reliably detects userspace moving head past tail; in that
case treat the queue as empty so callers see the full cache as free and
forward progress is preserved.
Suggested-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Zizhi Wo <wozizhi@huawei.com>
Link: https://patch.msgid.link/20260514021847.4062782-1-wozizhi@huaweicloud.com
[axboe: fixup commit message, kill 'queued' var, and keep it all in
io_uring.c]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2026-05-14 10:18:47 +08:00
|
|
|
|
2022-06-17 09:48:01 +01:00
|
|
|
|
2022-09-23 14:53:25 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-24 23:53:26 +01:00
|
|
|
|
2022-06-17 09:48:01 +01:00
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: validate user-controlled cq.head in io_cqe_cache_refill()
A fuzzing run reproduced an unkillable io_uring task stuck at ~100% CPU:
[root@fedora io_uring_stress]# ps -ef | grep io_uring
root 1240 1 99 13:36 ? 00:01:35 [io_uring_stress] <defunct>
The task loops inside io_cqring_wait() and never returns to userspace,
and SIGKILL has no effect.
This is caused by the CQ ring exposing rings->cq.head to userspace as
writable, while the authoritative tail lives in kernel-private
ctx->cached_cq_tail. io_cqe_cache_refill() computes free space as an
unsigned subtraction:
free = ctx->cq_entries - min(tail - head, ctx->cq_entries);
If userspace keeps head within [0, tail], the subtraction is well
defined and min() just acts as a defensive clamp. But if userspace
advances head past tail, (tail - head) wraps to a huge value, free
becomes 0, and io_cqe_cache_refill() fails. The CQE is pushed onto the
overflow list and IO_CHECK_CQ_OVERFLOW_BIT is set.
The wait loop in io_cqring_wait() relies on an invariant: refill() only
fails when the CQ is *physically* full, in which case rings->cq.tail has
been advanced to iowq->cq_tail and io_should_wake() returns true. The
tampered head breaks this: refill() fails while the ring is not full, no
OCQE is copied in, rings->cq.tail never catches up, io_should_wake()
stays false, and io_cqring_wait_schedule() keeps returning early because
IO_CHECK_CQ_OVERFLOW_BIT is still set. The result is a tight retry loop
that never returns to userspace.
Introduce io_cqring_queued() as the single point that converts the
(tail, head) pair into a trustworthy queued count. Since the real
head/tail distance is bounded by cq_entries (far below 2^31), a signed
comparison reliably detects userspace moving head past tail; in that
case treat the queue as empty so callers see the full cache as free and
forward progress is preserved.
Suggested-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Zizhi Wo <wozizhi@huawei.com>
Link: https://patch.msgid.link/20260514021847.4062782-1-wozizhi@huaweicloud.com
[axboe: fixup commit message, kill 'queued' var, and keep it all in
io_uring.c]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2026-05-14 10:18:47 +08:00
|
|
|
|
2022-06-17 09:48:01 +01:00
|
|
|
|
|
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
2023-08-24 23:53:26 +01:00
|
|
|
|
2022-06-17 09:48:01 +01:00
|
|
|
|
2022-06-17 09:48:05 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-17 09:48:01 +01:00
|
|
|
|
|
|
|
|
|
2023-08-24 23:53:26 +01:00
|
|
|
|
2022-06-17 09:48:01 +01:00
|
|
|
|
|
|
|
|
|
2025-06-16 10:46:28 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
2025-06-16 10:46:28 +01:00
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
2025-06-16 10:46:28 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-12-07 08:50:01 -07:00
|
|
|
|
|
|
|
|
|
2021-11-10 15:49:31 +00:00
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
2022-06-15 11:23:06 +01:00
|
|
|
|
|
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
2022-06-15 11:23:06 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-15 11:23:07 +01:00
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
2022-06-15 11:23:07 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-10-18 17:14:00 +01:00
|
|
|
|
|
|
|
|
|
2022-06-15 11:23:06 +01:00
|
|
|
|
|
|
|
|
|
2022-06-30 02:12:26 -07:00
|
|
|
|
io_uring: support buffer selection for OP_READ and OP_RECV
If a server process has tons of pending socket connections, generally
it uses epoll to wait for activity. When the socket is ready for reading
(or writing), the task can select a buffer and issue a recv/send on the
given fd.
Now that we have fast (non-async thread) support, a task can have tons
of pending reads or writes pending. But that means they need buffers to
back that data, and if the number of connections is high enough, having
them preallocated for all possible connections is unfeasible.
With IORING_OP_PROVIDE_BUFFERS, an application can register buffers to
use for any request. The request then sets IOSQE_BUFFER_SELECT in the
sqe, and a given group ID in sqe->buf_group. When the fd becomes ready,
a free buffer from the specified group is selected. If none are
available, the request is terminated with -ENOBUFS. If successful, the
CQE on completion will contain the buffer ID chosen in the cqe->flags
member, encoded as:
(buffer_id << IORING_CQE_BUFFER_SHIFT) | IORING_CQE_F_BUFFER;
Once a buffer has been consumed by a request, it is no longer available
and must be registered again with IORING_OP_PROVIDE_BUFFERS.
Requests need to support this feature. For now, IORING_OP_READ and
IORING_OP_RECV support it. This is checked on SQE submission, a CQE with
res == -EOPNOTSUPP will be posted if attempted on unsupported requests.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-02-23 16:42:51 -07:00
|
|
|
|
|
|
|
|
|
2025-05-16 09:45:24 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-05-16 13:39:32 -06:00
|
|
|
|
2025-10-21 22:59:30 +05:30
|
|
|
|
2025-05-16 13:39:32 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-12-30 19:57:28 +01:00
|
|
|
|
2025-05-16 13:39:32 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-03-18 22:00:31 +00:00
|
|
|
|
2022-06-17 09:48:00 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-20 01:25:56 +01:00
|
|
|
|
2025-03-24 15:32:34 +00:00
|
|
|
|
2025-05-16 09:39:14 -06:00
|
|
|
|
2025-05-16 09:45:24 -06:00
|
|
|
|
2025-05-16 09:39:14 -06:00
|
|
|
|
2025-05-16 13:39:32 -06:00
|
|
|
|
2025-05-16 09:39:14 -06:00
|
|
|
|
2022-06-20 01:25:56 +01:00
|
|
|
|
2022-06-17 09:48:00 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-06-06 10:28:26 -06:00
|
|
|
|
2025-11-04 21:01:09 -08:00
|
|
|
|
2024-06-06 10:28:26 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-05-09 12:03:43 +01:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2025-05-09 12:03:43 +01:00
|
|
|
|
2024-07-02 14:38:11 +01:00
|
|
|
|
2025-05-16 09:45:24 -06:00
|
|
|
|
2025-05-16 09:39:14 -06:00
|
|
|
|
2025-05-16 13:39:32 -06:00
|
|
|
|
2024-07-02 14:38:11 +01:00
|
|
|
|
2024-06-06 10:28:26 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-11 13:53:45 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-03-18 22:00:31 +00:00
|
|
|
|
2022-11-24 01:35:55 -08:00
|
|
|
|
2023-06-07 14:41:20 -06:00
|
|
|
|
2024-03-18 22:00:32 +00:00
|
|
|
|
2022-11-24 01:35:55 -08:00
|
|
|
|
2025-05-07 07:34:24 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-03-08 13:55:57 +00:00
|
|
|
|
2022-11-24 01:35:55 -08:00
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2025-04-24 12:28:39 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-03-18 22:00:32 +00:00
|
|
|
|
|
|
|
|
|
2022-11-24 01:35:55 -08:00
|
|
|
|
|
|
|
|
|
2025-06-16 10:46:28 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2025-06-16 10:46:28 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-03-18 22:00:34 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2022-11-23 11:33:40 +00:00
|
|
|
|
2025-03-24 15:32:33 +00:00
|
|
|
|
2022-11-23 11:33:40 +00:00
|
|
|
|
2024-04-05 16:50:03 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-03-18 22:00:34 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2025-03-24 15:32:33 +00:00
|
|
|
|
2024-03-18 22:00:34 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-11-23 11:33:40 +00:00
|
|
|
|
|
|
|
|
|
2025-03-24 15:32:33 +00:00
|
|
|
|
|
|
|
|
|
2022-06-20 01:25:56 +01:00
|
|
|
|
2023-04-04 13:39:47 +01:00
|
|
|
|
2025-03-24 15:32:33 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-04-05 16:50:03 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-04-05 16:50:02 +01:00
|
|
|
|
io_uring: support buffer selection for OP_READ and OP_RECV
If a server process has tons of pending socket connections, generally
it uses epoll to wait for activity. When the socket is ready for reading
(or writing), the task can select a buffer and issue a recv/send on the
given fd.
Now that we have fast (non-async thread) support, a task can have tons
of pending reads or writes pending. But that means they need buffers to
back that data, and if the number of connections is high enough, having
them preallocated for all possible connections is unfeasible.
With IORING_OP_PROVIDE_BUFFERS, an application can register buffers to
use for any request. The request then sets IOSQE_BUFFER_SELECT in the
sqe, and a given group ID in sqe->buf_group. When the fd becomes ready,
a free buffer from the specified group is selected. If none are
available, the request is terminated with -ENOBUFS. If successful, the
CQE on completion will contain the buffer ID chosen in the cqe->flags
member, encoded as:
(buffer_id << IORING_CQE_BUFFER_SHIFT) | IORING_CQE_F_BUFFER;
Once a buffer has been consumed by a request, it is no longer available
and must be registered again with IORING_OP_PROVIDE_BUFFERS.
Requests need to support this feature. For now, IORING_OP_READ and
IORING_OP_RECV support it. This is checked on SQE submission, a CQE with
res == -EOPNOTSUPP will be posted if attempted on unsupported requests.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-02-23 16:42:51 -07:00
|
|
|
|
|
|
|
|
|
2022-11-24 01:35:53 -08:00
|
|
|
|
2022-11-23 11:33:37 +00:00
|
|
|
|
2021-02-28 22:35:12 +00:00
|
|
|
|
2023-01-12 06:44:11 -08:00
|
|
|
|
2022-09-21 12:17:46 +01:00
|
|
|
|
2022-11-23 11:33:37 +00:00
|
|
|
|
|
|
|
|
|
2021-05-16 22:58:05 +01:00
|
|
|
|
io_uring/kbuf: switch to storing struct io_buffer_list locally
Currently the buffer list is stored in struct io_kiocb. The buffer list
can be of two types:
1) Classic/legacy buffer list. These don't need to get referenced after
a buffer pick, and hence storing them in struct io_kiocb is perfectly
fine.
2) Ring provided buffer lists. These DO need to be referenced after the
initial buffer pick, as they need to get consumed later on. This can
be either just incrementing the head of the ring, or it can be
consuming parts of a buffer if incremental buffer consumptions has
been configured.
For case 2, io_uring needs to be careful not to access the buffer list
after the initial pick-and-execute context. The core does recycling of
these, but it's easy to make a mistake, because it's stored in the
io_kiocb which does persist across multiple execution contexts. Either
because it's a multishot request, or simply because it needed some kind
of async trigger (eg poll) for retry purposes.
Add a struct io_buffer_list to struct io_br_sel, which is always on
stack for the various users of it. This prevents the buffer list from
leaking outside of that execution context, and additionally it enables
kbuf to not even pass back the struct io_buffer_list if the given
context isn't appropriately locked already.
This doesn't fix any bugs, it's simply a defensive measure to prevent
any issues with reuse of a buffer list.
Link: https://lore.kernel.org/r/20250821020750.598432-12-axboe@kernel.dk
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-20 20:03:39 -06:00
|
|
|
|
2022-09-21 12:17:46 +01:00
|
|
|
|
|
|
|
|
|
2022-11-24 01:35:53 -08:00
|
|
|
|
2021-02-28 22:35:12 +00:00
|
|
|
|
|
|
|
|
|
io_uring: remove submission references
Requests are by default given with two references, submission and
completion. Completion references are straightforward, they represent
request ownership and are put when a request is completed or so.
Submission references are a bit more trickier. They're needed when
io_issue_sqe() followed deep into the submission stack (e.g. in fs,
block, drivers, etc.), request may have given away for concurrent
execution or already completed, and the code unwinding back to
io_issue_sqe() may be accessing some pieces of our requests, e.g.
file or iov.
Now, we prevent such async/in-depth completions by pushing requests
through task_work. Punting to io-wq is also done through task_works,
apart from a couple of cases with a pretty well known context. So,
there're two cases:
1) io_issue_sqe() from the task context and protected by ->uring_lock.
Either requests return back to io_uring or handed to task_work, which
won't be executed because we're currently controlling that task. So,
we can be sure that requests are staying alive all the time and we don't
need submission references to pin them.
2) io_issue_sqe() from io-wq, which doesn't hold the mutex. The role of
submission reference is played by io-wq reference, which is put by
io_wq_submit_work(). Hence, it should be fine.
Considering that, we can carefully kill the submission reference.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/6b68f1c763229a590f2a27148aee77767a8d7750.1628705069.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2021-08-11 19:28:29 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-07-27 10:30:40 +01:00
|
|
|
|
io_uring: remove submission references
Requests are by default given with two references, submission and
completion. Completion references are straightforward, they represent
request ownership and are put when a request is completed or so.
Submission references are a bit more trickier. They're needed when
io_issue_sqe() followed deep into the submission stack (e.g. in fs,
block, drivers, etc.), request may have given away for concurrent
execution or already completed, and the code unwinding back to
io_issue_sqe() may be accessing some pieces of our requests, e.g.
file or iov.
Now, we prevent such async/in-depth completions by pushing requests
through task_work. Punting to io-wq is also done through task_works,
apart from a couple of cases with a pretty well known context. So,
there're two cases:
1) io_issue_sqe() from the task context and protected by ->uring_lock.
Either requests return back to io_uring or handed to task_work, which
won't be executed because we're currently controlling that task. So,
we can be sure that requests are staying alive all the time and we don't
need submission references to pin them.
2) io_issue_sqe() from io-wq, which doesn't hold the mutex. The role of
submission reference is played by io-wq reference, which is put by
io_wq_submit_work(). Hence, it should be fine.
Considering that, we can carefully kill the submission reference.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/6b68f1c763229a590f2a27148aee77767a8d7750.1628705069.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2021-08-11 19:28:29 +01:00
|
|
|
|
2020-06-22 10:13:11 -06:00
|
|
|
|
2025-05-06 13:31:07 +01:00
|
|
|
|
2021-09-24 21:59:45 +01:00
|
|
|
|
2026-05-28 11:34:32 +02:00
|
|
|
|
2021-02-10 00:03:23 +00:00
|
|
|
|
2021-08-09 13:04:08 +01:00
|
|
|
|
2026-05-28 11:34:32 +02:00
|
|
|
|
|
|
|
|
|
2021-08-09 13:04:08 +01:00
|
|
|
|
2026-05-28 11:34:32 +02:00
|
|
|
|
2021-09-24 21:59:45 +01:00
|
|
|
|
|
|
|
|
|
2021-10-04 20:02:49 +01:00
|
|
|
|
2026-05-28 11:34:32 +02:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2021-08-09 13:04:08 +01:00
|
|
|
|
2026-05-28 11:34:32 +02:00
|
|
|
|
|
|
|
|
|
2025-05-09 12:12:53 +01:00
|
|
|
|
2026-05-28 11:34:32 +02:00
|
|
|
|
|
|
|
|
|
2021-10-04 20:02:49 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-04-04 13:39:48 +01:00
|
|
|
|
|
|
|
|
|
2023-06-23 12:23:22 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-04-04 13:39:48 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-09-08 16:40:51 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-12-02 17:47:23 +00:00
|
|
|
|
2022-06-20 01:25:55 +01:00
|
|
|
|
2022-12-02 17:47:23 +00:00
|
|
|
|
2021-09-08 16:40:51 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-05-10 16:07:28 -06:00
|
|
|
|
2021-03-09 00:37:58 +00:00
|
|
|
|
2019-11-05 12:40:47 -07:00
|
|
|
|
2019-05-10 16:07:28 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-09-08 16:40:51 +01:00
|
|
|
|
|
|
|
|
|
2021-03-09 00:37:58 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-11-20 13:03:52 -07:00
|
|
|
|
2019-11-05 12:40:47 -07:00
|
|
|
|
2025-10-31 14:34:29 -06:00
|
|
|
|
io_uring: use task_work for links if possible
Currently links are always done in an async fashion, unless we catch them
inline after we successfully complete a request without having to resort
to blocking. This isn't necessarily the most efficient approach, it'd be
more ideal if we could just use the task_work handling for this.
Outside of saving an async jump, we can also do less prep work for these
kinds of requests.
Running dependent links from the task_work handler yields some nice
performance benefits. As an example, examples/link-cp from the liburing
repository uses read+write links to implement a copy operation. Without
this patch, the a cache fold 4G file read from a VM runs in about 3
seconds:
$ time examples/link-cp /data/file /dev/null
real 0m2.986s
user 0m0.051s
sys 0m2.843s
and a subsequent cache hot run looks like this:
$ time examples/link-cp /data/file /dev/null
real 0m0.898s
user 0m0.069s
sys 0m0.797s
With this patch in place, the cold case takes about 2.4 seconds:
$ time examples/link-cp /data/file /dev/null
real 0m2.400s
user 0m0.020s
sys 0m2.366s
and the cache hot case looks like this:
$ time examples/link-cp /data/file /dev/null
real 0m0.676s
user 0m0.010s
sys 0m0.665s
As expected, the (mostly) cache hot case yields the biggest improvement,
running about 25% faster with this change, while the cache cold case
yields about a 20% increase in performance. Outside of the performance
increase, we're using less CPU as well, as we're not using the async
offload threads at all for this anymore.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-25 15:39:59 -06:00
|
|
|
|
2025-10-31 14:34:29 -06:00
|
|
|
|
|
|
|
|
|
2025-02-16 19:25:04 -07:00
|
|
|
|
2022-11-24 01:35:53 -08:00
|
|
|
|
io_uring: use task_work for links if possible
Currently links are always done in an async fashion, unless we catch them
inline after we successfully complete a request without having to resort
to blocking. This isn't necessarily the most efficient approach, it'd be
more ideal if we could just use the task_work handling for this.
Outside of saving an async jump, we can also do less prep work for these
kinds of requests.
Running dependent links from the task_work handler yields some nice
performance benefits. As an example, examples/link-cp from the liburing
repository uses read+write links to implement a copy operation. Without
this patch, the a cache fold 4G file read from a VM runs in about 3
seconds:
$ time examples/link-cp /data/file /dev/null
real 0m2.986s
user 0m0.051s
sys 0m2.843s
and a subsequent cache hot run looks like this:
$ time examples/link-cp /data/file /dev/null
real 0m0.898s
user 0m0.069s
sys 0m0.797s
With this patch in place, the cold case takes about 2.4 seconds:
$ time examples/link-cp /data/file /dev/null
real 0m2.400s
user 0m0.020s
sys 0m2.366s
and the cache hot case looks like this:
$ time examples/link-cp /data/file /dev/null
real 0m0.676s
user 0m0.010s
sys 0m0.665s
As expected, the (mostly) cache hot case yields the biggest improvement,
running about 25% faster with this change, while the cache cold case
yields about a 20% increase in performance. Outside of the performance
increase, we're using less CPU as well, as we're not using the async
offload threads at all for this anymore.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-25 15:39:59 -06:00
|
|
|
|
|
|
|
|
|
2025-10-31 14:34:29 -06:00
|
|
|
|
io_uring: use task_work for links if possible
Currently links are always done in an async fashion, unless we catch them
inline after we successfully complete a request without having to resort
to blocking. This isn't necessarily the most efficient approach, it'd be
more ideal if we could just use the task_work handling for this.
Outside of saving an async jump, we can also do less prep work for these
kinds of requests.
Running dependent links from the task_work handler yields some nice
performance benefits. As an example, examples/link-cp from the liburing
repository uses read+write links to implement a copy operation. Without
this patch, the a cache fold 4G file read from a VM runs in about 3
seconds:
$ time examples/link-cp /data/file /dev/null
real 0m2.986s
user 0m0.051s
sys 0m2.843s
and a subsequent cache hot run looks like this:
$ time examples/link-cp /data/file /dev/null
real 0m0.898s
user 0m0.069s
sys 0m0.797s
With this patch in place, the cold case takes about 2.4 seconds:
$ time examples/link-cp /data/file /dev/null
real 0m2.400s
user 0m0.020s
sys 0m2.366s
and the cache hot case looks like this:
$ time examples/link-cp /data/file /dev/null
real 0m0.676s
user 0m0.010s
sys 0m0.665s
As expected, the (mostly) cache hot case yields the biggest improvement,
running about 25% faster with this change, while the cache cold case
yields about a 20% increase in performance. Outside of the performance
increase, we're using less CPU as well, as we're not using the async
offload threads at all for this anymore.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-25 15:39:59 -06:00
|
|
|
|
2025-10-31 14:34:29 -06:00
|
|
|
|
2025-09-18 10:21:14 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-09-23 04:25:22 -06:00
|
|
|
|
2022-11-24 01:35:53 -08:00
|
|
|
|
2023-01-27 05:52:24 -08:00
|
|
|
|
2024-03-18 22:00:28 +00:00
|
|
|
|
2023-01-27 05:52:24 -08:00
|
|
|
|
2025-06-05 11:48:33 -06:00
|
|
|
|
2019-11-20 13:03:52 -07:00
|
|
|
|
2019-05-10 16:07:28 -06:00
|
|
|
|
2022-05-25 08:57:27 -06:00
|
|
|
|
2021-02-18 22:32:52 +00:00
|
|
|
|
2022-05-24 15:21:00 -06:00
|
|
|
|
2021-06-30 21:54:04 +01:00
|
|
|
|
2022-05-21 09:17:05 -06:00
|
|
|
|
2021-02-18 22:32:52 +00:00
|
|
|
|
|
|
|
|
|
2022-06-13 07:27:03 -06:00
|
|
|
|
2021-02-28 22:35:10 +00:00
|
|
|
|
2021-06-30 21:54:04 +01:00
|
|
|
|
2022-05-21 09:17:05 -06:00
|
|
|
|
2021-02-28 22:35:10 +00:00
|
|
|
|
|
|
|
|
|
2022-05-25 08:57:27 -06:00
|
|
|
|
2019-11-09 11:00:08 +08:00
|
|
|
|
2020-06-29 13:13:00 +03:00
|
|
|
|
2019-11-21 23:21:01 +03:00
|
|
|
|
2020-06-27 14:04:55 +03:00
|
|
|
|
|
|
|
|
|
2019-11-09 11:00:08 +08:00
|
|
|
|
|
|
|
|
|
2025-05-06 13:31:16 +01:00
|
|
|
|
|
|
|
|
|
2026-03-17 14:28:44 -06:00
|
|
|
|
|
|
|
|
|
2025-05-06 13:31:16 +01:00
|
|
|
|
2026-03-17 14:28:44 -06:00
|
|
|
|
2025-05-06 13:31:16 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-17 14:28:44 -06:00
|
|
|
|
2025-05-06 13:31:16 +01:00
|
|
|
|
|
|
|
|
|
2023-08-24 23:53:29 +01:00
|
|
|
|
|
|
|
|
|
2021-08-12 12:48:34 -06:00
|
|
|
|
2021-02-10 00:03:14 +00:00
|
|
|
|
2021-09-24 21:59:50 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-03-21 22:02:22 +00:00
|
|
|
|
2025-01-07 10:59:35 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-03-21 22:02:22 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-03-21 22:02:23 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-03-04 12:48:12 -07:00
|
|
|
|
2022-03-21 22:02:23 +00:00
|
|
|
|
|
|
|
|
|
2022-04-15 22:08:29 +01:00
|
|
|
|
2022-03-21 22:02:24 +00:00
|
|
|
|
2022-03-21 22:02:22 +00:00
|
|
|
|
|
|
|
|
|
2021-10-04 20:02:55 +01:00
|
|
|
|
2023-07-07 11:14:40 -06:00
|
|
|
|
io_uring/rsrc: get rid of per-ring io_rsrc_node list
Work in progress, but get rid of the per-ring serialization of resource
nodes, like registered buffers and files. Main issue here is that one
node can otherwise hold up a bunch of other nodes from getting freed,
which is especially a problem for file resource nodes and networked
workloads where some descriptors may not see activity in a long time.
As an example, instantiate an io_uring ring fd and create a sparse
registered file table. Even 2 will do. Then create a socket and register
it as fixed file 0, F0. The number of open files in the app is now 5,
with 0/1/2 being the usual stdin/out/err, 3 being the ring fd, and 4
being the socket. Register this socket (eg "the listener") in slot 0 of
the registered file table. Now add an operation on the socket that uses
slot 0. Finally, loop N times, where each loop creates a new socket,
registers said socket as a file, then unregisters the socket, and
finally closes the socket. This is roughly similar to what a basic
accept loop would look like.
At the end of this loop, it's not unreasonable to expect that there
would still be 5 open files. Each socket created and registered in the
loop is also unregistered and closed. But since the listener socket
registered first still has references to its resource node due to still
being active, each subsequent socket unregistration is stuck behind it
for reclaim. Hence 5 + N files are still open at that point, where N is
awaiting the final put held up by the listener socket.
Rewrite the io_rsrc_node handling to NOT rely on serialization. Struct
io_kiocb now gets explicit resource nodes assigned, with each holding a
reference to the parent node. A parent node is either of type FILE or
BUFFER, which are the two types of nodes that exist. A request can have
two nodes assigned, if it's using both registered files and buffers.
Since request issue and task_work completion is both under the ring
private lock, no atomics are needed to handle these references. It's a
simple unlocked inc/dec. As before, the registered buffer or file table
each hold a reference as well to the registered nodes. Final put of the
node will remove the node and free the underlying resource, eg unmap the
buffer or put the file.
Outside of removing the stall in resource reclaim described above, it
has the following advantages:
1) It's a lot simpler than the previous scheme, and easier to follow.
No need to specific quiesce handling anymore.
2) There are no resource node allocations in the fast path, all of that
happens at resource registration time.
3) The structs related to resource handling can all get simplified
quite a bit, like io_rsrc_node and io_rsrc_data. io_rsrc_put can
go away completely.
4) Handling of resource tags is much simpler, and doesn't require
persistent storage as it can simply get assigned up front at
registration time. Just copy them in one-by-one at registration time
and assign to the resource node.
The only real downside is that a request is now explicitly limited to
pinning 2 resources, one file and one buffer, where before just
assigning a resource node to a request would pin all of them. The upside
is that it's easier to follow now, as an individual resource is
explicitly referenced and assigned to the request.
With this in place, the above mentioned example will be using exactly 5
files at the end of the loop, not N.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-10-25 19:27:39 -06:00
|
|
|
|
io_uring: move struct io_kiocb from task_struct to io_uring_task
Rather than store the task_struct itself in struct io_kiocb, store
the io_uring specific task_struct. The life times are the same in terms
of io_uring, and this avoids doing some dereferences through the
task_struct. For the hot path of putting local task references, we can
deref req->tctx instead, which we'll need anyway in that function
regardless of whether it's local or remote references.
This is mostly straight forward, except the original task PF_EXITING
check needs a bit of tweaking. task_work is _always_ run from the
originating task, except in the fallback case, where it's run from a
kernel thread. Replace the potentially racy (in case of fallback work)
checks for req->task->flags with current->flags. It's either the still
the original task, in which case PF_EXITING will be sane, or it has
PF_KTHREAD set, in which case it's fallback work. Both cases should
prevent moving forward with the given request.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-11-03 10:23:38 -07:00
|
|
|
|
2024-04-05 16:50:05 +01:00
|
|
|
|
2021-10-04 20:02:55 +01:00
|
|
|
|
2022-04-12 15:09:48 +01:00
|
|
|
|
2021-09-24 21:59:50 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-24 23:53:29 +01:00
|
|
|
|
2021-08-12 12:48:34 -06:00
|
|
|
|
2021-02-10 00:03:14 +00:00
|
|
|
|
2021-08-09 20:18:11 +01:00
|
|
|
|
2023-03-09 09:51:13 -07:00
|
|
|
|
2021-02-10 00:03:14 +00:00
|
|
|
|
2022-12-07 08:50:01 -07:00
|
|
|
|
2023-03-09 09:51:13 -07:00
|
|
|
|
2022-06-19 12:26:08 +01:00
|
|
|
|
|
|
|
|
|
2021-06-26 21:40:48 +01:00
|
|
|
|
2025-01-07 10:59:35 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-11 13:53:43 +01:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2025-05-16 13:39:32 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-12-07 08:50:01 -07:00
|
|
|
|
2021-02-10 00:03:14 +00:00
|
|
|
|
2023-06-23 12:23:31 +01:00
|
|
|
|
2022-06-19 12:26:08 +01:00
|
|
|
|
2024-06-14 10:57:03 -06:00
|
|
|
|
2022-11-24 01:35:54 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-05-09 12:12:54 +01:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2025-05-09 12:12:54 +01:00
|
|
|
|
|
|
|
|
|
2024-03-18 22:00:32 +00:00
|
|
|
|
2020-03-03 21:33:13 +03:00
|
|
|
|
|
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-03 12:39:05 -07:00
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-12-11 03:25:41 -07:00
|
|
|
|
2020-07-07 16:36:22 +03:00
|
|
|
|
2021-09-24 21:59:43 +01:00
|
|
|
|
2020-07-07 16:36:22 +03:00
|
|
|
|
2019-08-21 22:19:11 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-07-06 17:59:31 +03:00
|
|
|
|
2019-08-21 22:19:11 -06:00
|
|
|
|
2020-07-06 17:59:31 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
|
|
|
|
|
2025-06-13 15:24:53 -06:00
|
|
|
|
|
|
|
|
|
2026-06-11 11:44:47 -06:00
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
|
|
|
|
|
2025-03-24 15:32:36 +00:00
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
2022-04-21 02:13:44 -07:00
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
2025-03-24 15:32:36 +00:00
|
|
|
|
|
|
|
|
|
2024-04-10 02:26:54 +01:00
|
|
|
|
|
|
|
|
|
2022-09-08 16:56:52 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-15 17:33:55 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-04-10 02:26:55 +01:00
|
|
|
|
2022-06-15 17:33:55 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-04-13 02:58:46 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-03-22 14:07:58 +00:00
|
|
|
|
2022-04-21 02:13:44 -07:00
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
2023-08-09 10:03:00 -06:00
|
|
|
|
|
|
|
|
|
2019-08-19 12:15:59 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-12-11 03:25:41 -07:00
|
|
|
|
2025-03-24 15:32:35 +00:00
|
|
|
|
2021-04-13 02:58:45 +01:00
|
|
|
|
2025-12-11 03:25:41 -07:00
|
|
|
|
2022-09-03 09:52:01 -06:00
|
|
|
|
2022-09-08 16:56:54 +01:00
|
|
|
|
2022-09-03 09:52:01 -06:00
|
|
|
|
|
|
|
|
|
2021-07-08 13:37:06 +01:00
|
|
|
|
2026-03-02 10:29:12 -07:00
|
|
|
|
2021-04-13 02:58:45 +01:00
|
|
|
|
2019-08-19 12:15:59 -06:00
|
|
|
|
2025-03-24 15:32:35 +00:00
|
|
|
|
2023-08-09 10:03:00 -06:00
|
|
|
|
|
|
|
|
|
2022-03-22 14:07:58 +00:00
|
|
|
|
2023-08-09 16:20:21 +01:00
|
|
|
|
|
|
|
|
|
2023-08-09 10:03:00 -06:00
|
|
|
|
|
|
|
|
|
2026-03-02 10:29:12 -07:00
|
|
|
|
2023-08-09 10:03:00 -06:00
|
|
|
|
|
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
2022-06-16 10:21:59 +01:00
|
|
|
|
2025-10-31 14:34:29 -06:00
|
|
|
|
2021-08-10 15:15:25 -06:00
|
|
|
|
2025-10-31 14:34:29 -06:00
|
|
|
|
2021-08-10 15:15:25 -06:00
|
|
|
|
|
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-04-13 02:58:46 +01:00
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
|
|
|
|
|
2021-10-15 17:09:12 +01:00
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
|
|
|
|
|
2021-10-18 21:34:31 +08:00
|
|
|
|
2021-06-14 02:36:14 +01:00
|
|
|
|
|
|
|
|
|
2021-10-18 21:34:31 +08:00
|
|
|
|
2021-06-14 02:36:14 +01:00
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-12-11 03:25:41 -07:00
|
|
|
|
2021-06-28 05:37:30 +08:00
|
|
|
|
|
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
|
|
|
|
|
2025-12-11 03:25:41 -07:00
|
|
|
|
2021-10-12 13:12:14 +02:00
|
|
|
|
2021-06-28 05:37:30 +08:00
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
|
|
|
|
|
2025-12-11 03:25:41 -07:00
|
|
|
|
io_uring: fix poll_list race for SETUP_IOPOLL|SETUP_SQPOLL
After making ext4 support iopoll method:
let ext4_file_operations's iopoll method be iomap_dio_iopoll(),
we found fio can easily hang in fio_ioring_getevents() with below fio
job:
rm -f testfile; sync;
sudo fio -name=fiotest -filename=testfile -iodepth=128 -thread
-rw=write -ioengine=io_uring -hipri=1 -sqthread_poll=1 -direct=1
-bs=4k -size=10G -numjobs=8 -runtime=2000 -group_reporting
with IORING_SETUP_SQPOLL and IORING_SETUP_IOPOLL enabled.
There are two issues that results in this hang, one reason is that
when IORING_SETUP_SQPOLL and IORING_SETUP_IOPOLL are enabled, fio
does not use io_uring_enter to get completed events, it relies on
kernel io_sq_thread to poll for completed events.
Another reason is that there is a race: when io_submit_sqes() in
io_sq_thread() submits a batch of sqes, variable 'inflight' will
record the number of submitted reqs, then io_sq_thread will poll for
reqs which have been added to poll_list. But note, if some previous
reqs have been punted to io worker, these reqs will won't be in
poll_list timely. io_sq_thread() will only poll for a part of previous
submitted reqs, and then find poll_list is empty, reset variable
'inflight' to be zero. If app just waits these deferred reqs and does
not wake up io_sq_thread again, then hang happens.
For app that entirely relies on io_sq_thread to poll completed requests,
let io_iopoll_req_issued() wake up io_sq_thread properly when adding new
element to poll_list, and when io_sq_thread prepares to sleep, check
whether poll_list is empty again, if not empty, continue to poll.
Signed-off-by: Xiaoguang Wang <xiaoguang.wang@linux.alibaba.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-02-25 22:12:08 +08:00
|
|
|
|
2021-10-18 21:34:31 +08:00
|
|
|
|
2021-06-14 02:36:14 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
|
|
|
|
|
2024-01-28 20:05:47 -07:00
|
|
|
|
2021-03-12 08:30:14 -07:00
|
|
|
|
2024-01-28 20:05:47 -07:00
|
|
|
|
2021-03-12 08:30:14 -07:00
|
|
|
|
2025-03-31 08:54:00 +01:00
|
|
|
|
|
|
|
|
|
2025-07-08 11:04:45 -06:00
|
|
|
|
2023-06-20 13:32:32 +02:00
|
|
|
|
2023-06-20 13:32:28 +02:00
|
|
|
|
2023-06-20 13:32:32 +02:00
|
|
|
|
2021-10-17 00:07:10 +01:00
|
|
|
|
2021-03-12 08:30:14 -07:00
|
|
|
|
|
|
|
|
|
2021-10-04 20:02:54 +01:00
|
|
|
|
2022-11-23 11:33:37 +00:00
|
|
|
|
2019-04-06 21:51:27 -06:00
|
|
|
|
2019-11-08 08:09:12 -07:00
|
|
|
|
2025-05-09 12:12:51 +01:00
|
|
|
|
2020-07-13 23:37:14 +03:00
|
|
|
|
2021-06-15 16:47:57 +01:00
|
|
|
|
treewide: Replace kmalloc with kmalloc_obj for non-scalar types
This is the result of running the Coccinelle script from
scripts/coccinelle/api/kmalloc_objs.cocci. The script is designed to
avoid scalar types (which need careful case-by-case checking), and
instead replace kmalloc-family calls that allocate struct or union
object instances:
Single allocations: kmalloc(sizeof(TYPE), ...)
are replaced with: kmalloc_obj(TYPE, ...)
Array allocations: kmalloc_array(COUNT, sizeof(TYPE), ...)
are replaced with: kmalloc_objs(TYPE, COUNT, ...)
Flex array allocations: kmalloc(struct_size(PTR, FAM, COUNT), ...)
are replaced with: kmalloc_flex(*PTR, FAM, COUNT, ...)
(where TYPE may also be *VAR)
The resulting allocations no longer return "void *", instead returning
"TYPE *".
Signed-off-by: Kees Cook <kees@kernel.org>
2026-02-20 23:49:23 -08:00
|
|
|
|
2021-06-14 23:37:30 +01:00
|
|
|
|
2025-05-09 12:12:49 +01:00
|
|
|
|
2021-10-01 18:07:01 +01:00
|
|
|
|
2021-06-15 16:47:56 +01:00
|
|
|
|
2020-07-13 23:37:15 +03:00
|
|
|
|
2020-06-29 19:18:43 +03:00
|
|
|
|
2022-06-16 13:57:20 +01:00
|
|
|
|
2020-07-13 23:37:14 +03:00
|
|
|
|
2025-05-09 12:12:51 +01:00
|
|
|
|
2025-05-09 12:12:54 +01:00
|
|
|
|
2020-07-13 23:37:14 +03:00
|
|
|
|
2025-05-09 12:12:54 +01:00
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2019-04-06 21:51:27 -06:00
|
|
|
|
|
|
|
|
|
2023-01-20 09:10:30 -07:00
|
|
|
|
|
|
|
|
|
io_uring: defer file assignment
If an application uses direct open or accept, it knows in advance what
direct descriptor value it will get as it picks it itself. This allows
combined requests such as:
sqe = io_uring_get_sqe(ring);
io_uring_prep_openat_direct(sqe, ..., file_slot);
sqe->flags |= IOSQE_IO_LINK | IOSQE_CQE_SKIP_SUCCESS;
sqe = io_uring_get_sqe(ring);
io_uring_prep_read(sqe,file_slot, buf, buf_size, 0);
sqe->flags |= IOSQE_FIXED_FILE;
io_uring_submit(ring);
where we prepare both a file open and read, and only get a completion
event for the read when both have completed successfully.
Currently links are fully prepared before the head is issued, but that
fails if the dependent link needs a file assigned that isn't valid until
the head has completed.
Conversely, if the same chain is performed but the fixed file slot is
already valid, then we would be unexpectedly returning data from the
old file slot rather than the newly opened one. Make sure we're
consistent here.
Allow deferral of file setup, which makes this documented case work.
Cc: stable@vger.kernel.org # v5.15+
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2022-03-29 10:10:08 -06:00
|
|
|
|
2023-01-20 09:10:30 -07:00
|
|
|
|
io_uring: defer file assignment
If an application uses direct open or accept, it knows in advance what
direct descriptor value it will get as it picks it itself. This allows
combined requests such as:
sqe = io_uring_get_sqe(ring);
io_uring_prep_openat_direct(sqe, ..., file_slot);
sqe->flags |= IOSQE_IO_LINK | IOSQE_CQE_SKIP_SUCCESS;
sqe = io_uring_get_sqe(ring);
io_uring_prep_read(sqe,file_slot, buf, buf_size, 0);
sqe->flags |= IOSQE_FIXED_FILE;
io_uring_submit(ring);
where we prepare both a file open and read, and only get a completion
event for the read when both have completed successfully.
Currently links are fully prepared before the head is issued, but that
fails if the dependent link needs a file assigned that isn't valid until
the head has completed.
Conversely, if the same chain is performed but the fixed file slot is
already valid, then we would be unexpectedly returning data from the
old file slot rather than the newly opened one. Make sure we're
consistent here.
Allow deferral of file setup, which makes this documented case work.
Cc: stable@vger.kernel.org # v5.15+
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2022-03-29 10:10:08 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-04-12 15:09:43 +01:00
|
|
|
|
io_uring: defer file assignment
If an application uses direct open or accept, it knows in advance what
direct descriptor value it will get as it picks it itself. This allows
combined requests such as:
sqe = io_uring_get_sqe(ring);
io_uring_prep_openat_direct(sqe, ..., file_slot);
sqe->flags |= IOSQE_IO_LINK | IOSQE_CQE_SKIP_SUCCESS;
sqe = io_uring_get_sqe(ring);
io_uring_prep_read(sqe,file_slot, buf, buf_size, 0);
sqe->flags |= IOSQE_FIXED_FILE;
io_uring_submit(ring);
where we prepare both a file open and read, and only get a completion
event for the read when both have completed successfully.
Currently links are fully prepared before the head is issued, but that
fails if the dependent link needs a file assigned that isn't valid until
the head has completed.
Conversely, if the same chain is performed but the fixed file slot is
already valid, then we would be unexpectedly returning data from the
old file slot rather than the newly opened one. Make sure we're
consistent here.
Allow deferral of file setup, which makes this documented case work.
Cc: stable@vger.kernel.org # v5.15+
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2022-03-29 10:10:08 -06:00
|
|
|
|
2022-04-12 15:09:43 +01:00
|
|
|
|
io_uring: defer file assignment
If an application uses direct open or accept, it knows in advance what
direct descriptor value it will get as it picks it itself. This allows
combined requests such as:
sqe = io_uring_get_sqe(ring);
io_uring_prep_openat_direct(sqe, ..., file_slot);
sqe->flags |= IOSQE_IO_LINK | IOSQE_CQE_SKIP_SUCCESS;
sqe = io_uring_get_sqe(ring);
io_uring_prep_read(sqe,file_slot, buf, buf_size, 0);
sqe->flags |= IOSQE_FIXED_FILE;
io_uring_submit(ring);
where we prepare both a file open and read, and only get a completion
event for the read when both have completed successfully.
Currently links are fully prepared before the head is issued, but that
fails if the dependent link needs a file assigned that isn't valid until
the head has completed.
Conversely, if the same chain is performed but the fixed file slot is
already valid, then we would be unexpectedly returning data from the
old file slot rather than the newly opened one. Make sure we're
consistent here.
Allow deferral of file setup, which makes this documented case work.
Cc: stable@vger.kernel.org # v5.15+
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2022-03-29 10:10:08 -06:00
|
|
|
|
2022-04-18 20:51:12 +01:00
|
|
|
|
io_uring: defer file assignment
If an application uses direct open or accept, it knows in advance what
direct descriptor value it will get as it picks it itself. This allows
combined requests such as:
sqe = io_uring_get_sqe(ring);
io_uring_prep_openat_direct(sqe, ..., file_slot);
sqe->flags |= IOSQE_IO_LINK | IOSQE_CQE_SKIP_SUCCESS;
sqe = io_uring_get_sqe(ring);
io_uring_prep_read(sqe,file_slot, buf, buf_size, 0);
sqe->flags |= IOSQE_FIXED_FILE;
io_uring_submit(ring);
where we prepare both a file open and read, and only get a completion
event for the read when both have completed successfully.
Currently links are fully prepared before the head is issued, but that
fails if the dependent link needs a file assigned that isn't valid until
the head has completed.
Conversely, if the same chain is performed but the fixed file slot is
already valid, then we would be unexpectedly returning data from the
old file slot rather than the newly opened one. Make sure we're
consistent here.
Allow deferral of file setup, which makes this documented case work.
Cc: stable@vger.kernel.org # v5.15+
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2022-03-29 10:10:08 -06:00
|
|
|
|
|
|
|
|
|
2025-05-04 08:06:28 -06:00
|
|
|
|
|
|
|
|
|
2025-02-23 17:22:31 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2021-02-27 15:57:30 -07:00
|
|
|
|
2025-05-04 08:06:28 -06:00
|
|
|
|
2019-12-17 19:53:05 -07:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2025-05-04 08:06:28 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-27 15:57:30 -07:00
|
|
|
|
2022-05-23 16:53:15 -06:00
|
|
|
|
2021-02-16 19:46:48 -05:00
|
|
|
|
|
|
|
|
|
2022-05-23 16:56:21 -06:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2022-05-23 16:53:15 -06:00
|
|
|
|
2021-02-16 19:46:48 -05:00
|
|
|
|
|
|
|
|
|
2025-05-04 08:06:28 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-05-24 15:21:00 -06:00
|
|
|
|
2025-02-23 17:22:31 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-05-08 14:48:33 -06:00
|
|
|
|
2022-06-16 10:21:58 +01:00
|
|
|
|
2022-06-20 01:26:00 +01:00
|
|
|
|
2022-06-16 10:21:58 +01:00
|
|
|
|
2022-11-23 11:33:41 +00:00
|
|
|
|
2022-05-24 15:21:00 -06:00
|
|
|
|
2023-12-01 00:38:52 +00:00
|
|
|
|
|
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
2023-12-01 00:38:53 +00:00
|
|
|
|
|
|
|
|
|
2022-05-24 15:21:00 -06:00
|
|
|
|
2026-03-02 10:29:11 -07:00
|
|
|
|
2023-12-01 00:38:53 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2025-02-16 19:25:04 -07:00
|
|
|
|
2022-05-25 20:31:09 -06:00
|
|
|
|
2025-02-23 17:22:31 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-02-16 19:25:04 -07:00
|
|
|
|
2025-02-23 17:22:31 +00:00
|
|
|
|
|
|
|
|
|
2026-03-02 10:29:10 -07:00
|
|
|
|
2025-02-23 17:22:31 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-03-08 17:19:32 +00:00
|
|
|
|
2025-02-23 17:22:31 +00:00
|
|
|
|
2022-05-25 20:31:09 -06:00
|
|
|
|
|
|
|
|
|
2022-05-25 11:01:04 -06:00
|
|
|
|
2021-08-09 13:04:05 +01:00
|
|
|
|
|
|
|
|
|
2023-06-23 12:23:21 +01:00
|
|
|
|
2021-08-09 13:04:05 +01:00
|
|
|
|
2025-04-03 12:29:30 +01:00
|
|
|
|
2026-05-11 10:58:38 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-06-23 12:23:21 +01:00
|
|
|
|
2026-05-11 10:58:38 -06:00
|
|
|
|
|
|
|
|
|
2023-06-23 12:23:21 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-08-09 13:04:05 +01:00
|
|
|
|
|
|
|
|
|
2022-05-25 11:01:04 -06:00
|
|
|
|
2020-06-08 21:08:19 +03:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2023-01-12 06:44:10 -08:00
|
|
|
|
2022-12-07 03:53:30 +00:00
|
|
|
|
2021-10-23 12:13:57 +01:00
|
|
|
|
io_uring: defer file assignment
If an application uses direct open or accept, it knows in advance what
direct descriptor value it will get as it picks it itself. This allows
combined requests such as:
sqe = io_uring_get_sqe(ring);
io_uring_prep_openat_direct(sqe, ..., file_slot);
sqe->flags |= IOSQE_IO_LINK | IOSQE_CQE_SKIP_SUCCESS;
sqe = io_uring_get_sqe(ring);
io_uring_prep_read(sqe,file_slot, buf, buf_size, 0);
sqe->flags |= IOSQE_FIXED_FILE;
io_uring_submit(ring);
where we prepare both a file open and read, and only get a completion
event for the read when both have completed successfully.
Currently links are fully prepared before the head is issued, but that
fails if the dependent link needs a file assigned that isn't valid until
the head has completed.
Conversely, if the same chain is performed but the fixed file slot is
already valid, then we would be unexpectedly returning data from the
old file slot rather than the newly opened one. Make sure we're
consistent here.
Allow deferral of file setup, which makes this documented case work.
Cc: stable@vger.kernel.org # v5.15+
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2022-03-29 10:10:08 -06:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2025-03-29 10:15:24 -06:00
|
|
|
|
2021-08-15 10:40:18 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: remove submission references
Requests are by default given with two references, submission and
completion. Completion references are straightforward, they represent
request ownership and are put when a request is completed or so.
Submission references are a bit more trickier. They're needed when
io_issue_sqe() followed deep into the submission stack (e.g. in fs,
block, drivers, etc.), request may have given away for concurrent
execution or already completed, and the code unwinding back to
io_issue_sqe() may be accessing some pieces of our requests, e.g.
file or iov.
Now, we prevent such async/in-depth completions by pushing requests
through task_work. Punting to io-wq is also done through task_works,
apart from a couple of cases with a pretty well known context. So,
there're two cases:
1) io_issue_sqe() from the task context and protected by ->uring_lock.
Either requests return back to io_uring or handed to task_work, which
won't be executed because we're currently controlling that task. So,
we can be sure that requests are staying alive all the time and we don't
need submission references to pin them.
2) io_issue_sqe() from io-wq, which doesn't hold the mutex. The role of
submission reference is played by io-wq reference, which is put by
io_wq_submit_work(). Hence, it should be fine.
Considering that, we can carefully kill the submission reference.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/6b68f1c763229a590f2a27148aee77767a8d7750.1628705069.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2021-08-11 19:28:29 +01:00
|
|
|
|
2021-08-23 13:30:44 +01:00
|
|
|
|
2024-06-13 19:28:27 +00:00
|
|
|
|
2022-04-12 15:24:43 +01:00
|
|
|
|
io_uring: defer file assignment
If an application uses direct open or accept, it knows in advance what
direct descriptor value it will get as it picks it itself. This allows
combined requests such as:
sqe = io_uring_get_sqe(ring);
io_uring_prep_openat_direct(sqe, ..., file_slot);
sqe->flags |= IOSQE_IO_LINK | IOSQE_CQE_SKIP_SUCCESS;
sqe = io_uring_get_sqe(ring);
io_uring_prep_read(sqe,file_slot, buf, buf_size, 0);
sqe->flags |= IOSQE_FIXED_FILE;
io_uring_submit(ring);
where we prepare both a file open and read, and only get a completion
event for the read when both have completed successfully.
Currently links are fully prepared before the head is issued, but that
fails if the dependent link needs a file assigned that isn't valid until
the head has completed.
Conversely, if the same chain is performed but the fixed file slot is
already valid, then we would be unexpectedly returning data from the
old file slot rather than the newly opened one. Make sure we're
consistent here.
Allow deferral of file setup, which makes this documented case work.
Cc: stable@vger.kernel.org # v5.15+
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2022-03-29 10:10:08 -06:00
|
|
|
|
2021-10-23 12:13:57 +01:00
|
|
|
|
2019-10-24 07:25:42 -06:00
|
|
|
|
2023-01-20 09:10:30 -07:00
|
|
|
|
2022-04-12 15:24:43 +01:00
|
|
|
|
2024-06-13 19:28:27 +00:00
|
|
|
|
2022-04-12 15:24:43 +01:00
|
|
|
|
|
|
|
|
|
2019-01-18 22:56:34 -07:00
|
|
|
|
2024-03-08 13:55:57 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-03-27 09:57:27 +00:00
|
|
|
|
2024-03-08 13:55:57 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-04-01 11:30:06 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-03-27 09:57:27 +00:00
|
|
|
|
2024-04-01 11:30:06 -06:00
|
|
|
|
2024-03-08 13:55:57 +00:00
|
|
|
|
|
|
|
|
|
2021-10-23 12:13:57 +01:00
|
|
|
|
2021-10-23 12:13:59 +01:00
|
|
|
|
|
|
|
|
|
2024-01-28 20:08:24 -07:00
|
|
|
|
2021-10-23 12:13:59 +01:00
|
|
|
|
2021-10-23 12:13:57 +01:00
|
|
|
|
2021-10-23 12:13:59 +01:00
|
|
|
|
2021-10-23 12:13:57 +01:00
|
|
|
|
io_uring: implement async hybrid mode for pollable requests
The current logic of requests with IOSQE_ASYNC is first queueing it to
io-worker, then execute it in a synchronous way. For unbound works like
pollable requests(e.g. read/write a socketfd), the io-worker may stuck
there waiting for events for a long time. And thus other works wait in
the list for a long time too.
Let's introduce a new way for unbound works (currently pollable
requests), with this a request will first be queued to io-worker, then
executed in a nonblock try rather than a synchronous way. Failure of
that leads it to arm poll stuff and then the worker can begin to handle
other works.
The detail process of this kind of requests is:
step1: original context:
queue it to io-worker
step2: io-worker context:
nonblock try(the old logic is a synchronous try here)
|
|--fail--> arm poll
|
|--(fail/ready)-->synchronous issue
|
|--(succeed)-->worker finish it's job, tw
take over the req
This works much better than the old IOSQE_ASYNC logic in cases where
unbound max_worker is relatively small. In this case, number of
io-worker eazily increments to max_worker, new worker cannot be created
and running workers stuck there handling old works in IOSQE_ASYNC mode.
In my 64-core machine, set unbound max_worker to 20, run echo-server,
turns out:
(arguments: register_file, connetion number is 1000, message size is 12
Byte)
original IOSQE_ASYNC: 76664.151 tps
after this patch: 166934.985 tps
Suggested-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Hao Xu <haoxu@linux.alibaba.com>
Link: https://lore.kernel.org/r/20211018133445.103438-1-haoxu@linux.alibaba.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2021-10-18 21:34:45 +08:00
|
|
|
|
2021-10-23 12:13:57 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-07-20 13:16:53 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-10-23 12:13:57 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-02 10:29:10 -07:00
|
|
|
|
2022-05-13 11:24:56 +01:00
|
|
|
|
2023-09-07 13:50:07 +01:00
|
|
|
|
|
|
|
|
|
2021-10-23 12:13:57 +01:00
|
|
|
|
|
|
|
|
|
io_uring: implement async hybrid mode for pollable requests
The current logic of requests with IOSQE_ASYNC is first queueing it to
io-worker, then execute it in a synchronous way. For unbound works like
pollable requests(e.g. read/write a socketfd), the io-worker may stuck
there waiting for events for a long time. And thus other works wait in
the list for a long time too.
Let's introduce a new way for unbound works (currently pollable
requests), with this a request will first be queued to io-worker, then
executed in a nonblock try rather than a synchronous way. Failure of
that leads it to arm poll stuff and then the worker can begin to handle
other works.
The detail process of this kind of requests is:
step1: original context:
queue it to io-worker
step2: io-worker context:
nonblock try(the old logic is a synchronous try here)
|
|--fail--> arm poll
|
|--(fail/ready)-->synchronous issue
|
|--(succeed)-->worker finish it's job, tw
take over the req
This works much better than the old IOSQE_ASYNC logic in cases where
unbound max_worker is relatively small. In this case, number of
io-worker eazily increments to max_worker, new worker cannot be created
and running workers stuck there handling old works in IOSQE_ASYNC mode.
In my 64-core machine, set unbound max_worker to 20, run echo-server,
turns out:
(arguments: register_file, connetion number is 1000, message size is 12
Byte)
original IOSQE_ASYNC: 76664.151 tps
after this patch: 166934.985 tps
Suggested-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Hao Xu <haoxu@linux.alibaba.com>
Link: https://lore.kernel.org/r/20211018133445.103438-1-haoxu@linux.alibaba.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2021-10-18 21:34:45 +08:00
|
|
|
|
|
|
|
|
|
2022-03-15 10:54:08 -06:00
|
|
|
|
2021-10-23 12:13:57 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-01-18 22:56:34 -07:00
|
|
|
|
2021-02-18 22:32:52 +00:00
|
|
|
|
2024-07-24 12:16:21 +01:00
|
|
|
|
2021-02-18 22:32:52 +00:00
|
|
|
|
2019-01-18 22:56:34 -07:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2022-05-24 21:19:47 -06:00
|
|
|
|
|
|
|
|
|
2020-02-24 11:32:44 +03:00
|
|
|
|
2022-04-04 17:18:43 -06:00
|
|
|
|
io_uring/rsrc: get rid of per-ring io_rsrc_node list
Work in progress, but get rid of the per-ring serialization of resource
nodes, like registered buffers and files. Main issue here is that one
node can otherwise hold up a bunch of other nodes from getting freed,
which is especially a problem for file resource nodes and networked
workloads where some descriptors may not see activity in a long time.
As an example, instantiate an io_uring ring fd and create a sparse
registered file table. Even 2 will do. Then create a socket and register
it as fixed file 0, F0. The number of open files in the app is now 5,
with 0/1/2 being the usual stdin/out/err, 3 being the ring fd, and 4
being the socket. Register this socket (eg "the listener") in slot 0 of
the registered file table. Now add an operation on the socket that uses
slot 0. Finally, loop N times, where each loop creates a new socket,
registers said socket as a file, then unregisters the socket, and
finally closes the socket. This is roughly similar to what a basic
accept loop would look like.
At the end of this loop, it's not unreasonable to expect that there
would still be 5 open files. Each socket created and registered in the
loop is also unregistered and closed. But since the listener socket
registered first still has references to its resource node due to still
being active, each subsequent socket unregistration is stuck behind it
for reclaim. Hence 5 + N files are still open at that point, where N is
awaiting the final put held up by the listener socket.
Rewrite the io_rsrc_node handling to NOT rely on serialization. Struct
io_kiocb now gets explicit resource nodes assigned, with each holding a
reference to the parent node. A parent node is either of type FILE or
BUFFER, which are the two types of nodes that exist. A request can have
two nodes assigned, if it's using both registered files and buffers.
Since request issue and task_work completion is both under the ring
private lock, no atomics are needed to handle these references. It's a
simple unlocked inc/dec. As before, the registered buffer or file table
each hold a reference as well to the registered nodes. Final put of the
node will remove the node and free the underlying resource, eg unmap the
buffer or put the file.
Outside of removing the stall in resource reclaim described above, it
has the following advantages:
1) It's a lot simpler than the previous scheme, and easier to follow.
No need to specific quiesce handling anymore.
2) There are no resource node allocations in the fast path, all of that
happens at resource registration time.
3) The structs related to resource handling can all get simplified
quite a bit, like io_rsrc_node and io_rsrc_data. io_rsrc_put can
go away completely.
4) Handling of resource tags is much simpler, and doesn't require
persistent storage as it can simply get assigned up front at
registration time. Just copy them in one-by-one at registration time
and assign to the resource node.
The only real downside is that a request is now explicitly limited to
pinning 2 resources, one file and one buffer, where before just
assigning a resource node to a request would pin all of them. The upside
is that it's easier to follow now, as an individual resource is
explicitly referenced and assigned to the request.
With this in place, the above mentioned example will be using exactly 5
files at the end of the loop, not N.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-10-25 19:27:39 -06:00
|
|
|
|
2022-04-04 17:18:43 -06:00
|
|
|
|
2020-02-24 11:32:44 +03:00
|
|
|
|
2022-04-18 20:51:11 +01:00
|
|
|
|
2024-10-27 09:08:31 -06:00
|
|
|
|
io_uring/rsrc: get rid of per-ring io_rsrc_node list
Work in progress, but get rid of the per-ring serialization of resource
nodes, like registered buffers and files. Main issue here is that one
node can otherwise hold up a bunch of other nodes from getting freed,
which is especially a problem for file resource nodes and networked
workloads where some descriptors may not see activity in a long time.
As an example, instantiate an io_uring ring fd and create a sparse
registered file table. Even 2 will do. Then create a socket and register
it as fixed file 0, F0. The number of open files in the app is now 5,
with 0/1/2 being the usual stdin/out/err, 3 being the ring fd, and 4
being the socket. Register this socket (eg "the listener") in slot 0 of
the registered file table. Now add an operation on the socket that uses
slot 0. Finally, loop N times, where each loop creates a new socket,
registers said socket as a file, then unregisters the socket, and
finally closes the socket. This is roughly similar to what a basic
accept loop would look like.
At the end of this loop, it's not unreasonable to expect that there
would still be 5 open files. Each socket created and registered in the
loop is also unregistered and closed. But since the listener socket
registered first still has references to its resource node due to still
being active, each subsequent socket unregistration is stuck behind it
for reclaim. Hence 5 + N files are still open at that point, where N is
awaiting the final put held up by the listener socket.
Rewrite the io_rsrc_node handling to NOT rely on serialization. Struct
io_kiocb now gets explicit resource nodes assigned, with each holding a
reference to the parent node. A parent node is either of type FILE or
BUFFER, which are the two types of nodes that exist. A request can have
two nodes assigned, if it's using both registered files and buffers.
Since request issue and task_work completion is both under the ring
private lock, no atomics are needed to handle these references. It's a
simple unlocked inc/dec. As before, the registered buffer or file table
each hold a reference as well to the registered nodes. Final put of the
node will remove the node and free the underlying resource, eg unmap the
buffer or put the file.
Outside of removing the stall in resource reclaim described above, it
has the following advantages:
1) It's a lot simpler than the previous scheme, and easier to follow.
No need to specific quiesce handling anymore.
2) There are no resource node allocations in the fast path, all of that
happens at resource registration time.
3) The structs related to resource handling can all get simplified
quite a bit, like io_rsrc_node and io_rsrc_data. io_rsrc_put can
go away completely.
4) Handling of resource tags is much simpler, and doesn't require
persistent storage as it can simply get assigned up front at
registration time. Just copy them in one-by-one at registration time
and assign to the resource node.
The only real downside is that a request is now explicitly limited to
pinning 2 resources, one file and one buffer, where before just
assigning a resource node to a request would pin all of them. The upside
is that it's easier to follow now, as an individual resource is
explicitly referenced and assigned to the request.
With this in place, the above mentioned example will be using exactly 5
files at the end of the loop, not N.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-10-25 19:27:39 -06:00
|
|
|
|
2025-04-16 13:25:03 -06:00
|
|
|
|
|
|
|
|
|
io_uring/rsrc: get rid of per-ring io_rsrc_node list
Work in progress, but get rid of the per-ring serialization of resource
nodes, like registered buffers and files. Main issue here is that one
node can otherwise hold up a bunch of other nodes from getting freed,
which is especially a problem for file resource nodes and networked
workloads where some descriptors may not see activity in a long time.
As an example, instantiate an io_uring ring fd and create a sparse
registered file table. Even 2 will do. Then create a socket and register
it as fixed file 0, F0. The number of open files in the app is now 5,
with 0/1/2 being the usual stdin/out/err, 3 being the ring fd, and 4
being the socket. Register this socket (eg "the listener") in slot 0 of
the registered file table. Now add an operation on the socket that uses
slot 0. Finally, loop N times, where each loop creates a new socket,
registers said socket as a file, then unregisters the socket, and
finally closes the socket. This is roughly similar to what a basic
accept loop would look like.
At the end of this loop, it's not unreasonable to expect that there
would still be 5 open files. Each socket created and registered in the
loop is also unregistered and closed. But since the listener socket
registered first still has references to its resource node due to still
being active, each subsequent socket unregistration is stuck behind it
for reclaim. Hence 5 + N files are still open at that point, where N is
awaiting the final put held up by the listener socket.
Rewrite the io_rsrc_node handling to NOT rely on serialization. Struct
io_kiocb now gets explicit resource nodes assigned, with each holding a
reference to the parent node. A parent node is either of type FILE or
BUFFER, which are the two types of nodes that exist. A request can have
two nodes assigned, if it's using both registered files and buffers.
Since request issue and task_work completion is both under the ring
private lock, no atomics are needed to handle these references. It's a
simple unlocked inc/dec. As before, the registered buffer or file table
each hold a reference as well to the registered nodes. Final put of the
node will remove the node and free the underlying resource, eg unmap the
buffer or put the file.
Outside of removing the stall in resource reclaim described above, it
has the following advantages:
1) It's a lot simpler than the previous scheme, and easier to follow.
No need to specific quiesce handling anymore.
2) There are no resource node allocations in the fast path, all of that
happens at resource registration time.
3) The structs related to resource handling can all get simplified
quite a bit, like io_rsrc_node and io_rsrc_data. io_rsrc_put can
go away completely.
4) Handling of resource tags is much simpler, and doesn't require
persistent storage as it can simply get assigned up front at
registration time. Just copy them in one-by-one at registration time
and assign to the resource node.
The only real downside is that a request is now explicitly limited to
pinning 2 resources, one file and one buffer, where before just
assigning a resource node to a request would pin all of them. The upside
is that it's easier to follow now, as an individual resource is
explicitly referenced and assigned to the request.
With this in place, the above mentioned example will be using exactly 5
files at the end of the loop, not N.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-10-25 19:27:39 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-04-18 20:51:11 +01:00
|
|
|
|
2020-10-10 18:34:08 +01:00
|
|
|
|
2020-02-24 11:32:44 +03:00
|
|
|
|
|
|
|
|
|
2022-05-24 21:19:47 -06:00
|
|
|
|
2021-08-09 13:04:02 +01:00
|
|
|
|
io_uring: remove file batch-get optimisation
For requests with non-fixed files, instead of grabbing just one
reference, we get by the number of left requests, so the following
requests using the same file can take it without atomics.
However, it's not all win. If there is one request in the middle
not using files or having a fixed file, we'll need to put back the left
references. Even worse if an application submits requests dealing with
different files, it will do a put for each new request, so doubling the
number of atomics needed. Also, even if not used, it's still takes some
cycles in the submission path.
If a file used many times, it rather makes sense to pre-register it, if
not, we may fall in the described pitfall. So, this optimisation is a
matter of use case. Go with the simpliest code-wise way, remove it.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2021-08-10 14:52:47 +01:00
|
|
|
|
2021-08-09 13:04:02 +01:00
|
|
|
|
2022-06-16 13:57:20 +01:00
|
|
|
|
2021-08-09 13:04:02 +01:00
|
|
|
|
|
|
|
|
|
2022-05-25 10:28:04 -06:00
|
|
|
|
2022-06-01 23:57:02 -06:00
|
|
|
|
2021-08-09 13:04:02 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-06-05 11:33:52 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-09-24 21:59:59 +01:00
|
|
|
|
|
|
|
|
|
2022-04-15 22:08:28 +01:00
|
|
|
|
2025-06-05 11:33:52 -06:00
|
|
|
|
2022-11-24 01:35:53 -08:00
|
|
|
|
2022-04-15 22:08:28 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-06-05 11:33:52 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-03-15 10:54:08 -06:00
|
|
|
|
2021-09-24 21:59:59 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-03-18 22:00:28 +00:00
|
|
|
|
2021-09-24 21:59:59 +01:00
|
|
|
|
2022-03-09 11:27:52 -07:00
|
|
|
|
|
|
|
|
|
2021-09-24 21:59:59 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-06-05 11:48:33 -06:00
|
|
|
|
2021-08-09 13:04:10 +01:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2025-06-05 11:48:33 -06:00
|
|
|
|
|
|
|
|
|
2019-03-12 10:18:47 -06:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2025-06-05 11:48:33 -06:00
|
|
|
|
2019-10-17 09:20:46 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-12-01 00:38:53 +00:00
|
|
|
|
2025-06-05 11:33:52 -06:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2021-09-24 21:59:58 +01:00
|
|
|
|
2021-08-09 13:04:10 +01:00
|
|
|
|
2019-09-09 20:50:40 +08:00
|
|
|
|
2022-04-15 22:08:32 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-11-24 01:35:53 -08:00
|
|
|
|
2021-06-14 23:37:30 +01:00
|
|
|
|
2025-06-05 11:33:52 -06:00
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2023-01-27 02:59:11 -08:00
|
|
|
|
2021-06-14 23:37:30 +01:00
|
|
|
|
2024-03-18 22:00:28 +00:00
|
|
|
|
2019-12-17 08:04:44 -07:00
|
|
|
|
2019-09-09 20:50:40 +08:00
|
|
|
|
|
|
|
|
|
2020-08-27 16:58:30 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2026-01-12 08:14:44 -07:00
|
|
|
|
2020-08-27 16:58:30 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-02-12 09:48:05 -07:00
|
|
|
|
2021-10-01 18:07:00 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2021-10-01 18:07:00 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-11-25 17:21:03 +08:00
|
|
|
|
2021-10-01 18:07:00 +01:00
|
|
|
|
|
|
|
|
|
2021-11-25 17:21:03 +08:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2021-10-01 18:07:00 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-03-16 09:51:40 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-04-12 02:05:05 +03:00
|
|
|
|
io_uring: add support for IORING_SETUP_SQE_MIXED
Normal rings support 64b SQEs for posting submissions, while certain
features require the ring to be configured with IORING_SETUP_SQE128, as
they need to convey more information per submission. This, in turn,
makes ALL the SQEs be 128b in size. This is somewhat wasteful and
inefficient, particularly when only certain SQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed SQE sizes, using
IORING_SETUP_SQE_MIXED. When setup in this mode, SQEs posted to the ring
may be either 64b or 128b in size. If a SQE is 128b in size, then opcode
will be set to a variante to indicate that this is the case. Any other
non-128b opcode will assume the SQ's default size.
SQEs on these types of mixed rings may also utilize NOP with skip
success set. This can happen if the ring is one (small) SQE entry away
from wrapping, and an attempt is made to get a 128b SQE. As SQEs must be
contiguous in the SQ ring, a 128b SQE cannot wrap the ring. For this
case, a single NOP SQE should be inserted with the SKIP_SUCCESS flag
set. The kernel will process this as a normal NOP and without posting a
CQE.
Signed-off-by: Keith Busch <kbusch@kernel.org>
[axboe: {} style fix and assign sqe before opcode read]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-10-16 11:09:38 -07:00
|
|
|
|
2021-08-09 13:04:10 +01:00
|
|
|
|
2020-04-08 08:58:45 +03:00
|
|
|
|
2023-01-12 06:44:10 -08:00
|
|
|
|
2020-04-12 02:05:05 +03:00
|
|
|
|
2021-10-01 18:07:02 +01:00
|
|
|
|
2020-04-12 02:05:05 +03:00
|
|
|
|
2025-05-06 13:31:07 +01:00
|
|
|
|
2026-05-17 17:30:10 -04:00
|
|
|
|
2021-02-12 18:41:17 +00:00
|
|
|
|
2024-01-28 20:05:47 -07:00
|
|
|
|
2024-09-22 18:41:29 +08:00
|
|
|
|
2022-04-12 15:09:43 +01:00
|
|
|
|
2020-04-08 08:58:45 +03:00
|
|
|
|
io_uring: move struct io_kiocb from task_struct to io_uring_task
Rather than store the task_struct itself in struct io_kiocb, store
the io_uring specific task_struct. The life times are the same in terms
of io_uring, and this avoids doing some dereferences through the
task_struct. For the hot path of putting local task references, we can
deref req->tctx instead, which we'll need anyway in that function
regardless of whether it's local or remote references.
This is mostly straight forward, except the original task PF_EXITING
check needs a bit of tweaking. task_work is _always_ run from the
originating task, except in the fallback case, where it's run from a
kernel thread. Replace the potentially racy (in case of fallback work)
checks for req->task->flags with current->flags. It's either the still
the original task, in which case PF_EXITING will be sane, or it has
PF_KTHREAD set, in which case it's fallback work. Both cases should
prevent moving forward with the given request.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-11-03 10:23:38 -07:00
|
|
|
|
2024-06-14 01:04:29 +01:00
|
|
|
|
2025-08-21 13:24:57 -06:00
|
|
|
|
2020-04-12 02:05:05 +03:00
|
|
|
|
2026-05-17 17:30:10 -04:00
|
|
|
|
2021-10-06 16:06:49 +01:00
|
|
|
|
2024-03-16 09:51:40 -06:00
|
|
|
|
2021-10-06 16:06:49 +01:00
|
|
|
|
2026-05-17 17:30:10 -04:00
|
|
|
|
2025-02-14 22:48:15 +00:00
|
|
|
|
2026-05-17 17:30:10 -04:00
|
|
|
|
io_uring: add support for IORING_SETUP_SQE_MIXED
Normal rings support 64b SQEs for posting submissions, while certain
features require the ring to be configured with IORING_SETUP_SQE128, as
they need to convey more information per submission. This, in turn,
makes ALL the SQEs be 128b in size. This is somewhat wasteful and
inefficient, particularly when only certain SQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed SQE sizes, using
IORING_SETUP_SQE_MIXED. When setup in this mode, SQEs posted to the ring
may be either 64b or 128b in size. If a SQE is 128b in size, then opcode
will be set to a variante to indicate that this is the case. Any other
non-128b opcode will assume the SQ's default size.
SQEs on these types of mixed rings may also utilize NOP with skip
success set. This can happen if the ring is one (small) SQE entry away
from wrapping, and an attempt is made to get a 128b SQE. As SQEs must be
contiguous in the SQ ring, a 128b SQE cannot wrap the ring. For this
case, a single NOP SQE should be inserted with the SKIP_SUCCESS flag
set. The kernel will process this as a normal NOP and without posting a
CQE.
Signed-off-by: Keith Busch <kbusch@kernel.org>
[axboe: {} style fix and assign sqe before opcode read]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-10-16 11:09:38 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-09 22:20:02 -07:00
|
|
|
|
io_uring: add support for IORING_SETUP_SQE_MIXED
Normal rings support 64b SQEs for posting submissions, while certain
features require the ring to be configured with IORING_SETUP_SQE128, as
they need to convey more information per submission. This, in turn,
makes ALL the SQEs be 128b in size. This is somewhat wasteful and
inefficient, particularly when only certain SQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed SQE sizes, using
IORING_SETUP_SQE_MIXED. When setup in this mode, SQEs posted to the ring
may be either 64b or 128b in size. If a SQE is 128b in size, then opcode
will be set to a variante to indicate that this is the case. Any other
non-128b opcode will assume the SQ's default size.
SQEs on these types of mixed rings may also utilize NOP with skip
success set. This can happen if the ring is one (small) SQE entry away
from wrapping, and an attempt is made to get a 128b SQE. As SQEs must be
contiguous in the SQ ring, a 128b SQE cannot wrap the ring. For this
case, a single NOP SQE should be inserted with the SKIP_SUCCESS flag
set. The kernel will process this as a normal NOP and without posting a
CQE.
Signed-off-by: Keith Busch <kbusch@kernel.org>
[axboe: {} style fix and assign sqe before opcode read]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-10-16 11:09:38 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-09-15 12:03:38 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-03-16 09:51:40 -06:00
|
|
|
|
2022-04-28 19:09:43 -06:00
|
|
|
|
2022-05-23 16:53:15 -06:00
|
|
|
|
2024-03-16 09:51:40 -06:00
|
|
|
|
2022-04-28 19:09:43 -06:00
|
|
|
|
|
|
|
|
|
2021-11-10 15:49:34 +00:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2021-11-10 15:49:34 +00:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2024-03-16 09:51:40 -06:00
|
|
|
|
2025-02-12 09:48:05 -07:00
|
|
|
|
2021-11-10 15:49:34 +00:00
|
|
|
|
2021-09-24 21:59:57 +01:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2026-01-12 08:14:44 -07:00
|
|
|
|
2024-03-16 09:51:40 -06:00
|
|
|
|
2021-09-24 21:59:57 +01:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2021-09-24 21:59:57 +01:00
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-11-25 17:21:03 +08:00
|
|
|
|
2021-09-24 21:59:57 +01:00
|
|
|
|
2021-09-15 12:03:38 +01:00
|
|
|
|
2020-08-27 16:58:30 +02:00
|
|
|
|
2022-05-23 16:53:15 -06:00
|
|
|
|
2024-03-16 09:51:40 -06:00
|
|
|
|
2022-05-23 16:53:15 -06:00
|
|
|
|
2024-03-16 09:51:40 -06:00
|
|
|
|
2022-04-26 11:34:56 -06:00
|
|
|
|
2022-05-23 16:53:15 -06:00
|
|
|
|
2021-10-06 16:06:46 +01:00
|
|
|
|
|
|
|
|
|
2022-04-12 15:09:43 +01:00
|
|
|
|
io_uring: defer file assignment
If an application uses direct open or accept, it knows in advance what
direct descriptor value it will get as it picks it itself. This allows
combined requests such as:
sqe = io_uring_get_sqe(ring);
io_uring_prep_openat_direct(sqe, ..., file_slot);
sqe->flags |= IOSQE_IO_LINK | IOSQE_CQE_SKIP_SUCCESS;
sqe = io_uring_get_sqe(ring);
io_uring_prep_read(sqe,file_slot, buf, buf_size, 0);
sqe->flags |= IOSQE_FIXED_FILE;
io_uring_submit(ring);
where we prepare both a file open and read, and only get a completion
event for the read when both have completed successfully.
Currently links are fully prepared before the head is issued, but that
fails if the dependent link needs a file assigned that isn't valid until
the head has completed.
Conversely, if the same chain is performed but the fixed file slot is
already valid, then we would be unexpectedly returning data from the
old file slot rather than the newly opened one. Make sure we're
consistent here.
Allow deferral of file setup, which makes this documented case work.
Cc: stable@vger.kernel.org # v5.15+
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2022-03-29 10:10:08 -06:00
|
|
|
|
2021-10-06 16:06:46 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-05-23 16:53:15 -06:00
|
|
|
|
2021-10-06 16:06:46 +01:00
|
|
|
|
|
|
|
|
|
2021-10-06 11:01:42 -06:00
|
|
|
|
2021-10-06 16:06:46 +01:00
|
|
|
|
2020-11-20 15:50:51 +00:00
|
|
|
|
2020-04-12 02:05:05 +03:00
|
|
|
|
2021-03-06 09:22:27 -07:00
|
|
|
|
|
|
|
|
|
2021-11-01 21:06:18 -07:00
|
|
|
|
|
|
|
|
|
2021-06-17 18:14:01 +01:00
|
|
|
|
|
|
|
|
|
2024-03-16 09:51:40 -06:00
|
|
|
|
2021-06-17 18:14:01 +01:00
|
|
|
|
lsm,io_uring: add LSM hooks to io_uring
A full expalantion of io_uring is beyond the scope of this commit
description, but in summary it is an asynchronous I/O mechanism
which allows for I/O requests and the resulting data to be queued
in memory mapped "rings" which are shared between the kernel and
userspace. Optionally, io_uring offers the ability for applications
to spawn kernel threads to dequeue I/O requests from the ring and
submit the requests in the kernel, helping to minimize the syscall
overhead. Rings are accessed in userspace by memory mapping a file
descriptor provided by the io_uring_setup(2), and can be shared
between applications as one might do with any open file descriptor.
Finally, process credentials can be registered with a given ring
and any process with access to that ring can submit I/O requests
using any of the registered credentials.
While the io_uring functionality is widely recognized as offering a
vastly improved, and high performing asynchronous I/O mechanism, its
ability to allow processes to submit I/O requests with credentials
other than its own presents a challenge to LSMs. When a process
creates a new io_uring ring the ring's credentials are inhertied
from the calling process; if this ring is shared with another
process operating with different credentials there is the potential
to bypass the LSMs security policy. Similarly, registering
credentials with a given ring allows any process with access to that
ring to submit I/O requests with those credentials.
In an effort to allow LSMs to apply security policy to io_uring I/O
operations, this patch adds two new LSM hooks. These hooks, in
conjunction with the LSM anonymous inode support previously
submitted, allow an LSM to apply access control policy to the
sharing of io_uring rings as well as any io_uring credential changes
requested by a process.
The new LSM hooks are described below:
* int security_uring_override_creds(cred)
Controls if the current task, executing an io_uring operation,
is allowed to override it's credentials with @cred. In cases
where the current task is a user application, the current
credentials will be those of the user application. In cases
where the current task is a kernel thread servicing io_uring
requests the current credentials will be those of the io_uring
ring (inherited from the process that created the ring).
* int security_uring_sqpoll(void)
Controls if the current task is allowed to create an io_uring
polling thread (IORING_SETUP_SQPOLL). Without a SQPOLL thread
in the kernel processes must submit I/O requests via
io_uring_enter(2) which allows us to compare any requested
credential changes against the application making the request.
With a SQPOLL thread, we can no longer compare requested
credential changes against the application making the request,
the comparison is made against the ring's credentials.
Signed-off-by: Paul Moore <paul@paul-moore.com>
2021-02-01 19:56:49 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-03-16 09:51:40 -06:00
|
|
|
|
lsm,io_uring: add LSM hooks to io_uring
A full expalantion of io_uring is beyond the scope of this commit
description, but in summary it is an asynchronous I/O mechanism
which allows for I/O requests and the resulting data to be queued
in memory mapped "rings" which are shared between the kernel and
userspace. Optionally, io_uring offers the ability for applications
to spawn kernel threads to dequeue I/O requests from the ring and
submit the requests in the kernel, helping to minimize the syscall
overhead. Rings are accessed in userspace by memory mapping a file
descriptor provided by the io_uring_setup(2), and can be shared
between applications as one might do with any open file descriptor.
Finally, process credentials can be registered with a given ring
and any process with access to that ring can submit I/O requests
using any of the registered credentials.
While the io_uring functionality is widely recognized as offering a
vastly improved, and high performing asynchronous I/O mechanism, its
ability to allow processes to submit I/O requests with credentials
other than its own presents a challenge to LSMs. When a process
creates a new io_uring ring the ring's credentials are inhertied
from the calling process; if this ring is shared with another
process operating with different credentials there is the potential
to bypass the LSMs security policy. Similarly, registering
credentials with a given ring allows any process with access to that
ring to submit I/O requests with those credentials.
In an effort to allow LSMs to apply security policy to io_uring I/O
operations, this patch adds two new LSM hooks. These hooks, in
conjunction with the LSM anonymous inode support previously
submitted, allow an LSM to apply access control policy to the
sharing of io_uring rings as well as any io_uring credential changes
requested by a process.
The new LSM hooks are described below:
* int security_uring_override_creds(cred)
Controls if the current task, executing an io_uring operation,
is allowed to override it's credentials with @cred. In cases
where the current task is a user application, the current
credentials will be those of the user application. In cases
where the current task is a kernel thread servicing io_uring
requests the current credentials will be those of the io_uring
ring (inherited from the process that created the ring).
* int security_uring_sqpoll(void)
Controls if the current task is allowed to create an io_uring
polling thread (IORING_SETUP_SQPOLL). Without a SQPOLL thread
in the kernel processes must submit I/O requests via
io_uring_enter(2) which allows us to compare any requested
credential changes against the application making the request.
With a SQPOLL thread, we can no longer compare requested
credential changes against the application making the request,
the comparison is made against the ring's credentials.
Signed-off-by: Paul Moore <paul@paul-moore.com>
2021-02-01 19:56:49 -05:00
|
|
|
|
2021-06-17 18:14:02 +01:00
|
|
|
|
2021-03-06 09:22:27 -07:00
|
|
|
|
2020-04-12 02:05:05 +03:00
|
|
|
|
2022-05-23 16:56:21 -06:00
|
|
|
|
2020-04-08 08:58:45 +03:00
|
|
|
|
|
|
|
|
|
2022-04-15 22:08:30 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-16 13:57:20 +01:00
|
|
|
|
2022-04-15 22:08:30 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add support for IORING_SETUP_SQE_MIXED
Normal rings support 64b SQEs for posting submissions, while certain
features require the ring to be configured with IORING_SETUP_SQE128, as
they need to convey more information per submission. This, in turn,
makes ALL the SQEs be 128b in size. This is somewhat wasteful and
inefficient, particularly when only certain SQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed SQE sizes, using
IORING_SETUP_SQE_MIXED. When setup in this mode, SQEs posted to the ring
may be either 64b or 128b in size. If a SQE is 128b in size, then opcode
will be set to a variante to indicate that this is the case. Any other
non-128b opcode will assume the SQ's default size.
SQEs on these types of mixed rings may also utilize NOP with skip
success set. This can happen if the ring is one (small) SQE entry away
from wrapping, and an attempt is made to get a 128b SQE. As SQEs must be
contiguous in the SQ ring, a 128b SQE cannot wrap the ring. For this
case, a single NOP SQE should be inserted with the SKIP_SUCCESS flag
set. The kernel will process this as a normal NOP and without posting a
CQE.
Signed-off-by: Keith Busch <kbusch@kernel.org>
[axboe: {} style fix and assign sqe before opcode read]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-10-16 11:09:38 -07:00
|
|
|
|
2021-08-09 13:04:10 +01:00
|
|
|
|
2019-05-10 16:07:28 -06:00
|
|
|
|
2021-02-18 18:29:42 +00:00
|
|
|
|
2020-04-12 02:05:05 +03:00
|
|
|
|
2019-05-10 16:07:28 -06:00
|
|
|
|
io_uring: add support for IORING_SETUP_SQE_MIXED
Normal rings support 64b SQEs for posting submissions, while certain
features require the ring to be configured with IORING_SETUP_SQE128, as
they need to convey more information per submission. This, in turn,
makes ALL the SQEs be 128b in size. This is somewhat wasteful and
inefficient, particularly when only certain SQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed SQE sizes, using
IORING_SETUP_SQE_MIXED. When setup in this mode, SQEs posted to the ring
may be either 64b or 128b in size. If a SQE is 128b in size, then opcode
will be set to a variante to indicate that this is the case. Any other
non-128b opcode will assume the SQ's default size.
SQEs on these types of mixed rings may also utilize NOP with skip
success set. This can happen if the ring is one (small) SQE entry away
from wrapping, and an attempt is made to get a 128b SQE. As SQEs must be
contiguous in the SQ ring, a 128b SQE cannot wrap the ring. For this
case, a single NOP SQE should be inserted with the SKIP_SUCCESS flag
set. The kernel will process this as a normal NOP and without posting a
CQE.
Signed-off-by: Keith Busch <kbusch@kernel.org>
[axboe: {} style fix and assign sqe before opcode read]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-10-16 11:09:38 -07:00
|
|
|
|
2022-04-15 22:08:30 +01:00
|
|
|
|
|
|
|
|
|
2021-06-14 23:37:31 +01:00
|
|
|
|
2026-01-17 08:27:23 -07:00
|
|
|
|
|
|
|
|
|
2026-01-15 08:24:02 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-03-30 10:03:41 -06:00
|
|
|
|
2021-02-18 18:29:41 +00:00
|
|
|
|
2019-05-10 16:07:28 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-04-15 22:08:31 +01:00
|
|
|
|
2024-09-19 23:38:01 -06:00
|
|
|
|
2025-06-05 11:33:52 -06:00
|
|
|
|
2020-10-27 23:25:37 +00:00
|
|
|
|
2020-10-27 23:25:35 +00:00
|
|
|
|
2019-05-10 16:07:28 -06:00
|
|
|
|
2022-04-15 22:08:29 +01:00
|
|
|
|
2021-09-24 21:59:56 +01:00
|
|
|
|
2022-04-15 22:08:30 +01:00
|
|
|
|
|
|
|
|
|
2021-09-24 21:59:56 +01:00
|
|
|
|
2022-04-15 22:08:31 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-09-24 21:59:56 +01:00
|
|
|
|
2019-05-10 16:07:28 -06:00
|
|
|
|
2019-12-05 16:15:45 +03:00
|
|
|
|
2025-06-05 11:48:33 -06:00
|
|
|
|
2020-04-12 02:05:03 +03:00
|
|
|
|
2019-05-10 16:07:28 -06:00
|
|
|
|
|
|
|
|
|
2019-01-09 09:06:50 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-09-24 21:59:55 +01:00
|
|
|
|
2019-01-09 09:06:50 -07:00
|
|
|
|
2021-09-24 21:59:55 +01:00
|
|
|
|
|
|
|
|
|
2022-04-12 15:09:45 +01:00
|
|
|
|
|
|
|
|
|
2021-09-24 21:59:55 +01:00
|
|
|
|
2021-09-08 16:40:52 +01:00
|
|
|
|
2020-10-28 09:33:23 -06:00
|
|
|
|
|
|
|
|
|
2019-01-09 09:06:50 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-10 00:03:11 +00:00
|
|
|
|
2019-01-09 09:06:50 -07:00
|
|
|
|
2020-10-28 09:33:23 -06:00
|
|
|
|
2021-09-08 16:40:49 +01:00
|
|
|
|
2021-10-06 11:01:42 -06:00
|
|
|
|
2021-02-18 18:29:42 +00:00
|
|
|
|
|
|
|
|
|
2019-01-09 09:06:50 -07:00
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2019-08-26 17:23:46 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2026-01-20 20:47:40 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-06-04 17:42:56 +01:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-01-23 14:37:15 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2023-08-24 23:53:32 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-02-15 22:06:52 +00:00
|
|
|
|
2024-10-18 17:07:59 +01:00
|
|
|
|
2023-08-24 23:53:32 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-11-04 12:02:47 +00:00
|
|
|
|
2023-08-24 23:53:32 +01:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-24 23:53:32 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-04-08 08:58:43 +03:00
|
|
|
|
|
|
|
|
|
2022-05-25 09:13:39 -06:00
|
|
|
|
2021-08-09 13:04:10 +01:00
|
|
|
|
io_uring: add submission polling
This enables an application to do IO, without ever entering the kernel.
By using the SQ ring to fill in new sqes and watching for completions
on the CQ ring, we can submit and reap IOs without doing a single system
call. The kernel side thread will poll for new submissions, and in case
of HIPRI/polled IO, it'll also poll for completions.
By default, we allow 1 second of active spinning. This can by changed
by passing in a different grace period at io_uring_register(2) time.
If the thread exceeds this idle time without having any work to do, it
will set:
sq_ring->flags |= IORING_SQ_NEED_WAKEUP.
The application will have to call io_uring_enter() to start things back
up again. If IO is kept busy, that will never be needed. Basically an
application that has this feature enabled will guard it's
io_uring_enter(2) call with:
read_barrier();
if (*sq_ring->flags & IORING_SQ_NEED_WAKEUP)
io_uring_enter(fd, 0, 0, IORING_ENTER_SQ_WAKEUP);
instead of calling it unconditionally.
It's mandatory to use fixed files with this feature. Failure to do so
will result in the application getting an -EBADF CQ entry when
submitting IO.
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-10 11:22:30 -07:00
|
|
|
|
2026-01-20 20:47:40 +00:00
|
|
|
|
2022-04-12 15:09:49 +01:00
|
|
|
|
|
|
|
|
|
io_uring: add submission polling
This enables an application to do IO, without ever entering the kernel.
By using the SQ ring to fill in new sqes and watching for completions
on the CQ ring, we can submit and reap IOs without doing a single system
call. The kernel side thread will poll for new submissions, and in case
of HIPRI/polled IO, it'll also poll for completions.
By default, we allow 1 second of active spinning. This can by changed
by passing in a different grace period at io_uring_register(2) time.
If the thread exceeds this idle time without having any work to do, it
will set:
sq_ring->flags |= IORING_SQ_NEED_WAKEUP.
The application will have to call io_uring_enter() to start things back
up again. If IO is kept busy, that will never be needed. Basically an
application that has this feature enabled will guard it's
io_uring_enter(2) call with:
read_barrier();
if (*sq_ring->flags & IORING_SQ_NEED_WAKEUP)
io_uring_enter(fd, 0, 0, IORING_ENTER_SQ_WAKEUP);
instead of calling it unconditionally.
It's mandatory to use fixed files with this feature. Failure to do so
will result in the application getting an -EBADF CQ entry when
submitting IO.
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-10 11:22:30 -07:00
|
|
|
|
2026-01-20 20:47:40 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-31 07:07:47 -06:00
|
|
|
|
2026-01-20 20:47:40 +00:00
|
|
|
|
2025-10-16 12:20:31 +01:00
|
|
|
|
2021-10-04 20:02:47 +01:00
|
|
|
|
2021-09-24 22:00:01 +01:00
|
|
|
|
2025-10-16 12:20:31 +01:00
|
|
|
|
|
|
|
|
|
2022-04-12 15:09:49 +01:00
|
|
|
|
|
|
|
|
|
io_uring: add submission polling
This enables an application to do IO, without ever entering the kernel.
By using the SQ ring to fill in new sqes and watching for completions
on the CQ ring, we can submit and reap IOs without doing a single system
call. The kernel side thread will poll for new submissions, and in case
of HIPRI/polled IO, it'll also poll for completions.
By default, we allow 1 second of active spinning. This can by changed
by passing in a different grace period at io_uring_register(2) time.
If the thread exceeds this idle time without having any work to do, it
will set:
sq_ring->flags |= IORING_SQ_NEED_WAKEUP.
The application will have to call io_uring_enter() to start things back
up again. If IO is kept busy, that will never be needed. Basically an
application that has this feature enabled will guard it's
io_uring_enter(2) call with:
read_barrier();
if (*sq_ring->flags & IORING_SQ_NEED_WAKEUP)
io_uring_enter(fd, 0, 0, IORING_ENTER_SQ_WAKEUP);
instead of calling it unconditionally.
It's mandatory to use fixed files with this feature. Failure to do so
will result in the application getting an -EBADF CQ entry when
submitting IO.
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-10 11:22:30 -07:00
|
|
|
|
2021-09-24 22:00:01 +01:00
|
|
|
|
2019-12-19 18:24:38 -07:00
|
|
|
|
2019-11-07 01:41:06 +03:00
|
|
|
|
2019-10-25 12:31:30 +03:00
|
|
|
|
2023-01-23 14:37:16 +00:00
|
|
|
|
2019-10-25 12:31:30 +03:00
|
|
|
|
2023-01-23 14:37:15 +00:00
|
|
|
|
2022-04-12 15:09:48 +01:00
|
|
|
|
2021-02-12 11:55:17 +00:00
|
|
|
|
|
|
|
|
|
2022-04-12 15:09:50 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add support for IORING_SETUP_SQE_MIXED
Normal rings support 64b SQEs for posting submissions, while certain
features require the ring to be configured with IORING_SETUP_SQE128, as
they need to convey more information per submission. This, in turn,
makes ALL the SQEs be 128b in size. This is somewhat wasteful and
inefficient, particularly when only certain SQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed SQE sizes, using
IORING_SETUP_SQE_MIXED. When setup in this mode, SQEs posted to the ring
may be either 64b or 128b in size. If a SQE is 128b in size, then opcode
will be set to a variante to indicate that this is the case. Any other
non-128b opcode will assume the SQ's default size.
SQEs on these types of mixed rings may also utilize NOP with skip
success set. This can happen if the ring is one (small) SQE entry away
from wrapping, and an attempt is made to get a 128b SQE. As SQEs must be
contiguous in the SQ ring, a 128b SQE cannot wrap the ring. For this
case, a single NOP SQE should be inserted with the SKIP_SUCCESS flag
set. The kernel will process this as a normal NOP and without posting a
CQE.
Signed-off-by: Keith Busch <kbusch@kernel.org>
[axboe: {} style fix and assign sqe before opcode read]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-10-16 11:09:38 -07:00
|
|
|
|
2022-04-12 15:09:50 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-03-10 12:59:35 -07:00
|
|
|
|
2022-04-12 15:09:50 +01:00
|
|
|
|
io_uring: add submission polling
This enables an application to do IO, without ever entering the kernel.
By using the SQ ring to fill in new sqes and watching for completions
on the CQ ring, we can submit and reap IOs without doing a single system
call. The kernel side thread will poll for new submissions, and in case
of HIPRI/polled IO, it'll also poll for completions.
By default, we allow 1 second of active spinning. This can by changed
by passing in a different grace period at io_uring_register(2) time.
If the thread exceeds this idle time without having any work to do, it
will set:
sq_ring->flags |= IORING_SQ_NEED_WAKEUP.
The application will have to call io_uring_enter() to start things back
up again. If IO is kept busy, that will never be needed. Basically an
application that has this feature enabled will guard it's
io_uring_enter(2) call with:
read_barrier();
if (*sq_ring->flags & IORING_SQ_NEED_WAKEUP)
io_uring_enter(fd, 0, 0, IORING_ENTER_SQ_WAKEUP);
instead of calling it unconditionally.
It's mandatory to use fixed files with this feature. Failure to do so
will result in the application getting an -EBADF CQ entry when
submitting IO.
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-10 11:22:30 -07:00
|
|
|
|
2022-04-12 15:09:49 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-01-25 22:34:01 +03:00
|
|
|
|
io_uring: add submission polling
This enables an application to do IO, without ever entering the kernel.
By using the SQ ring to fill in new sqes and watching for completions
on the CQ ring, we can submit and reap IOs without doing a single system
call. The kernel side thread will poll for new submissions, and in case
of HIPRI/polled IO, it'll also poll for completions.
By default, we allow 1 second of active spinning. This can by changed
by passing in a different grace period at io_uring_register(2) time.
If the thread exceeds this idle time without having any work to do, it
will set:
sq_ring->flags |= IORING_SQ_NEED_WAKEUP.
The application will have to call io_uring_enter() to start things back
up again. If IO is kept busy, that will never be needed. Basically an
application that has this feature enabled will guard it's
io_uring_enter(2) call with:
read_barrier();
if (*sq_ring->flags & IORING_SQ_NEED_WAKEUP)
io_uring_enter(fd, 0, 0, IORING_ENTER_SQ_WAKEUP);
instead of calling it unconditionally.
It's mandatory to use fixed files with this feature. Failure to do so
will result in the application getting an -EBADF CQ entry when
submitting IO.
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-10 11:22:30 -07:00
|
|
|
|
2021-09-24 21:59:55 +01:00
|
|
|
|
2019-11-06 00:22:14 +03:00
|
|
|
|
|
|
|
|
|
2022-04-12 15:09:49 +01:00
|
|
|
|
io_uring: add submission polling
This enables an application to do IO, without ever entering the kernel.
By using the SQ ring to fill in new sqes and watching for completions
on the CQ ring, we can submit and reap IOs without doing a single system
call. The kernel side thread will poll for new submissions, and in case
of HIPRI/polled IO, it'll also poll for completions.
By default, we allow 1 second of active spinning. This can by changed
by passing in a different grace period at io_uring_register(2) time.
If the thread exceeds this idle time without having any work to do, it
will set:
sq_ring->flags |= IORING_SQ_NEED_WAKEUP.
The application will have to call io_uring_enter() to start things back
up again. If IO is kept busy, that will never be needed. Basically an
application that has this feature enabled will guard it's
io_uring_enter(2) call with:
read_barrier();
if (*sq_ring->flags & IORING_SQ_NEED_WAKEUP)
io_uring_enter(fd, 0, 0, IORING_ENTER_SQ_WAKEUP);
instead of calling it unconditionally.
It's mandatory to use fixed files with this feature. Failure to do so
will result in the application getting an -EBADF CQ entry when
submitting IO.
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-10 11:22:30 -07:00
|
|
|
|
|
|
|
|
|
2021-11-05 17:15:46 -06:00
|
|
|
|
|
|
|
|
|
2025-11-04 14:44:53 -08:00
|
|
|
|
|
|
|
|
|
2024-03-12 14:56:27 +00:00
|
|
|
|
2026-03-09 14:21:37 -06:00
|
|
|
|
2024-03-12 14:56:27 +00:00
|
|
|
|
2021-11-05 17:15:46 -06:00
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:58 +00:00
|
|
|
|
|
|
|
|
|
2019-08-26 17:23:46 +00:00
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:57 +00:00
|
|
|
|
2025-11-12 12:45:53 +00:00
|
|
|
|
2019-08-26 17:23:46 +00:00
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:57 +00:00
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
io_uring: add support for IORING_SETUP_SQE_MIXED
Normal rings support 64b SQEs for posting submissions, while certain
features require the ring to be configured with IORING_SETUP_SQE128, as
they need to convey more information per submission. This, in turn,
makes ALL the SQEs be 128b in size. This is somewhat wasteful and
inefficient, particularly when only certain SQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed SQE sizes, using
IORING_SETUP_SQE_MIXED. When setup in this mode, SQEs posted to the ring
may be either 64b or 128b in size. If a SQE is 128b in size, then opcode
will be set to a variante to indicate that this is the case. Any other
non-128b opcode will assume the SQ's default size.
SQEs on these types of mixed rings may also utilize NOP with skip
success set. This can happen if the ring is one (small) SQE entry away
from wrapping, and an attempt is made to get a 128b SQE. As SQEs must be
contiguous in the SQ ring, a 128b SQE cannot wrap the ring. For this
case, a single NOP SQE should be inserted with the SKIP_SUCCESS flag
set. The kernel will process this as a normal NOP and without posting a
CQE.
Signed-off-by: Keith Busch <kbusch@kernel.org>
[axboe: {} style fix and assign sqe before opcode read]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-10-16 11:09:38 -07:00
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:57 +00:00
|
|
|
|
io_uring: add support for IORING_SETUP_SQE_MIXED
Normal rings support 64b SQEs for posting submissions, while certain
features require the ring to be configured with IORING_SETUP_SQE128, as
they need to convey more information per submission. This, in turn,
makes ALL the SQEs be 128b in size. This is somewhat wasteful and
inefficient, particularly when only certain SQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed SQE sizes, using
IORING_SETUP_SQE_MIXED. When setup in this mode, SQEs posted to the ring
may be either 64b or 128b in size. If a SQE is 128b in size, then opcode
will be set to a variante to indicate that this is the case. Any other
non-128b opcode will assume the SQ's default size.
SQEs on these types of mixed rings may also utilize NOP with skip
success set. This can happen if the ring is one (small) SQE entry away
from wrapping, and an attempt is made to get a 128b SQE. As SQEs must be
contiguous in the SQ ring, a 128b SQE cannot wrap the ring. For this
case, a single NOP SQE should be inserted with the SKIP_SUCCESS flag
set. The kernel will process this as a normal NOP and without posting a
CQE.
Signed-off-by: Keith Busch <kbusch@kernel.org>
[axboe: {} style fix and assign sqe before opcode read]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-10-16 11:09:38 -07:00
|
|
|
|
2019-08-26 17:23:46 +00:00
|
|
|
|
2025-11-12 12:45:57 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:54 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:57 +00:00
|
|
|
|
2019-08-26 17:23:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:57 +00:00
|
|
|
|
2019-08-26 17:23:46 +00:00
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:53 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:57 +00:00
|
|
|
|
2025-11-12 12:45:53 +00:00
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:54 +00:00
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:57 +00:00
|
|
|
|
2023-08-24 23:53:32 +01:00
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:57 +00:00
|
|
|
|
|
|
|
|
|
2019-08-26 17:23:46 +00:00
|
|
|
|
|
|
|
|
|
2025-05-09 12:12:54 +01:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2023-01-23 14:37:16 +00:00
|
|
|
|
2021-10-04 20:02:53 +01:00
|
|
|
|
2021-02-10 00:03:17 +00:00
|
|
|
|
2022-04-12 15:09:47 +01:00
|
|
|
|
2023-01-23 14:37:16 +00:00
|
|
|
|
2025-08-17 23:09:18 +01:00
|
|
|
|
2021-09-24 21:59:47 +01:00
|
|
|
|
2021-10-04 20:02:53 +01:00
|
|
|
|
2021-09-24 21:59:47 +01:00
|
|
|
|
2025-05-09 12:12:53 +01:00
|
|
|
|
|
|
|
|
|
2021-10-04 20:02:53 +01:00
|
|
|
|
2025-05-09 12:12:53 +01:00
|
|
|
|
2025-05-09 12:12:54 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-13 09:09:44 -07:00
|
|
|
|
|
|
|
|
|
2021-10-04 20:02:54 +01:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2026-02-26 12:48:41 +00:00
|
|
|
|
2021-02-17 21:03:43 -07:00
|
|
|
|
2021-08-10 02:44:23 +01:00
|
|
|
|
2021-02-19 17:19:36 +08:00
|
|
|
|
io_uring/rsrc: get rid of per-ring io_rsrc_node list
Work in progress, but get rid of the per-ring serialization of resource
nodes, like registered buffers and files. Main issue here is that one
node can otherwise hold up a bunch of other nodes from getting freed,
which is especially a problem for file resource nodes and networked
workloads where some descriptors may not see activity in a long time.
As an example, instantiate an io_uring ring fd and create a sparse
registered file table. Even 2 will do. Then create a socket and register
it as fixed file 0, F0. The number of open files in the app is now 5,
with 0/1/2 being the usual stdin/out/err, 3 being the ring fd, and 4
being the socket. Register this socket (eg "the listener") in slot 0 of
the registered file table. Now add an operation on the socket that uses
slot 0. Finally, loop N times, where each loop creates a new socket,
registers said socket as a file, then unregisters the socket, and
finally closes the socket. This is roughly similar to what a basic
accept loop would look like.
At the end of this loop, it's not unreasonable to expect that there
would still be 5 open files. Each socket created and registered in the
loop is also unregistered and closed. But since the listener socket
registered first still has references to its resource node due to still
being active, each subsequent socket unregistration is stuck behind it
for reclaim. Hence 5 + N files are still open at that point, where N is
awaiting the final put held up by the listener socket.
Rewrite the io_rsrc_node handling to NOT rely on serialization. Struct
io_kiocb now gets explicit resource nodes assigned, with each holding a
reference to the parent node. A parent node is either of type FILE or
BUFFER, which are the two types of nodes that exist. A request can have
two nodes assigned, if it's using both registered files and buffers.
Since request issue and task_work completion is both under the ring
private lock, no atomics are needed to handle these references. It's a
simple unlocked inc/dec. As before, the registered buffer or file table
each hold a reference as well to the registered nodes. Final put of the
node will remove the node and free the underlying resource, eg unmap the
buffer or put the file.
Outside of removing the stall in resource reclaim described above, it
has the following advantages:
1) It's a lot simpler than the previous scheme, and easier to follow.
No need to specific quiesce handling anymore.
2) There are no resource node allocations in the fast path, all of that
happens at resource registration time.
3) The structs related to resource handling can all get simplified
quite a bit, like io_rsrc_node and io_rsrc_data. io_rsrc_put can
go away completely.
4) Handling of resource tags is much simpler, and doesn't require
persistent storage as it can simply get assigned up front at
registration time. Just copy them in one-by-one at registration time
and assign to the resource node.
The only real downside is that a request is now explicitly limited to
pinning 2 resources, one file and one buffer, where before just
assigning a resource node to a request would pin all of them. The upside
is that it's easier to follow now, as an individual resource is
explicitly referenced and assigned to the request.
With this in place, the above mentioned example will be using exactly 5
files at the end of the loop, not N.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-10-25 19:27:39 -06:00
|
|
|
|
|
|
|
|
|
2026-03-23 12:44:05 +00:00
|
|
|
|
2022-12-07 03:53:28 +00:00
|
|
|
|
2019-04-11 11:45:41 -06:00
|
|
|
|
2025-01-31 17:27:02 +00:00
|
|
|
|
2020-02-23 16:23:11 -07:00
|
|
|
|
2025-11-04 14:44:53 -08:00
|
|
|
|
2023-04-01 20:50:39 +01:00
|
|
|
|
2021-04-20 12:03:32 +01:00
|
|
|
|
|
|
|
|
|
2022-06-16 10:22:08 +01:00
|
|
|
|
|
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
2021-08-28 19:54:38 -06:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2022-10-04 03:19:08 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-11-05 17:15:46 -06:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2024-10-18 17:07:59 +01:00
|
|
|
|
2026-02-15 22:06:52 +00:00
|
|
|
|
2024-10-18 17:07:59 +01:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2021-02-27 15:04:18 -07:00
|
|
|
|
2025-05-09 12:12:53 +01:00
|
|
|
|
2026-01-17 08:27:23 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-01-15 08:24:02 -07:00
|
|
|
|
2025-05-09 12:12:53 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-02-19 12:33:30 -07:00
|
|
|
|
|
|
|
|
|
2023-06-08 09:38:36 -07:00
|
|
|
|
2024-09-30 17:11:32 -06:00
|
|
|
|
2022-05-01 10:52:44 -06:00
|
|
|
|
io_uring/rsrc: add huge page accounting for registered buffers
Track huge page references in a per-ring xarray to prevent double
accounting when the same huge page is used by multiple registered
buffers, either within the same ring or across cloned rings.
When registering buffers backed by huge pages, we need to account for
RLIMIT_MEMLOCK. But if multiple buffers share the same huge page (common
with cloned buffers), we must not account for the same page multiple
times. Similarly, we must only unaccount when the last reference to a
huge page is released.
Maintain a per-ring xarray (hpage_acct) that tracks reference counts for
each huge page. When registering a buffer, for each unique huge page,
increment its accounting reference count, and only account pages that
are newly added.
When unregistering a buffer, for each unique huge page, decrement its
refcount. Once the refcount hits zero, the page is unaccounted.
Note: any account is done against the ctx->user that was assigned when
the ring was setup. As before, if root is running the operation, no
accounting is done.
With these changes, any use of imu->acct_pages is also dead, hence kill
it from struct io_mapped_ubuf. This shrinks it from 56b to 48b on a
64-bit arch. Additionally, hpage_already_acct() is gone, which was an
O(M*M) scan over current + previous registrations.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2026-01-24 10:02:41 -07:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-01-09 14:46:09 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2023-01-09 14:46:09 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-12-19 08:54:20 -07:00
|
|
|
|
2023-01-09 14:46:09 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2023-01-09 14:46:09 +00:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2023-01-09 14:46:09 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-14 08:46:17 -06:00
|
|
|
|
2023-01-09 14:46:09 +00:00
|
|
|
|
2019-04-24 23:54:17 +02:00
|
|
|
|
2025-01-07 17:27:30 +01:00
|
|
|
|
|
|
|
|
|
2019-04-24 23:54:17 +02:00
|
|
|
|
2025-01-07 17:27:30 +01:00
|
|
|
|
|
|
|
|
|
2026-03-31 07:07:47 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2021-02-05 16:34:21 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-11-25 02:34:11 -08:00
|
|
|
|
2021-02-05 16:34:21 +08:00
|
|
|
|
2022-08-30 05:50:08 -07:00
|
|
|
|
2023-01-23 14:37:13 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2026-03-31 07:07:47 -06:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-03-06 11:02:15 +00:00
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
|
|
|
|
|
2021-10-04 20:02:54 +01:00
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-02-17 08:27:23 -07:00
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
2022-12-06 01:38:32 -08:00
|
|
|
|
|
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
2023-02-17 08:27:23 -07:00
|
|
|
|
2021-06-14 02:36:15 +01:00
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-10-04 20:02:54 +01:00
|
|
|
|
2020-04-09 18:14:00 -06:00
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
2026-01-21 16:06:41 -07:00
|
|
|
|
2021-08-09 13:04:17 +01:00
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-04-09 18:14:00 -06:00
|
|
|
|
2026-03-23 12:43:50 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-06-17 15:00:04 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-07-07 16:36:22 +03:00
|
|
|
|
2022-12-07 03:53:28 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-01-13 23:03:31 +07:00
|
|
|
|
io_uring: move local task_work in exit cancel loop
With IORING_SETUP_DEFER_TASKRUN, task work is queued to ctx->work_llist
(local work) rather than the fallback list. During io_ring_exit_work(),
io_move_task_work_from_local() was called once before the cancel loop,
moving work from work_llist to fallback_llist.
However, task work can be added to work_llist during the cancel loop
itself. There are two cases:
1) io_kill_timeouts() is called from io_uring_try_cancel_requests() to
cancel pending timeouts, and it adds task work via io_req_queue_tw_complete()
for each cancelled timeout:
2) URING_CMD requests like ublk can be completed via
io_uring_cmd_complete_in_task() from ublk_queue_rq() during canceling,
given ublk request queue is only quiesced when canceling the 1st uring_cmd.
Since io_allowed_defer_tw_run() returns false in io_ring_exit_work()
(kworker != submitter_task), io_run_local_work() is never invoked,
and the work_llist entries are never processed. This causes
io_uring_try_cancel_requests() to loop indefinitely, resulting in
100% CPU usage in kworker threads.
Fix this by moving io_move_task_work_from_local() inside the cancel
loop, ensuring any work on work_llist is moved to fallback before
each cancel attempt.
Cc: stable@vger.kernel.org
Fixes: c0e0d6ba25f1 ("io_uring: add IORING_SETUP_DEFER_TASKRUN")
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2026-01-14 16:54:05 +08:00
|
|
|
|
|
|
|
|
|
2026-06-11 11:44:47 -06:00
|
|
|
|
2022-06-20 01:25:52 +01:00
|
|
|
|
io_uring: move local task_work in exit cancel loop
With IORING_SETUP_DEFER_TASKRUN, task work is queued to ctx->work_llist
(local work) rather than the fallback list. During io_ring_exit_work(),
io_move_task_work_from_local() was called once before the cancel loop,
moving work from work_llist to fallback_llist.
However, task work can be added to work_llist during the cancel loop
itself. There are two cases:
1) io_kill_timeouts() is called from io_uring_try_cancel_requests() to
cancel pending timeouts, and it adds task work via io_req_queue_tw_complete()
for each cancelled timeout:
2) URING_CMD requests like ublk can be completed via
io_uring_cmd_complete_in_task() from ublk_queue_rq() during canceling,
given ublk request queue is only quiesced when canceling the 1st uring_cmd.
Since io_allowed_defer_tw_run() returns false in io_ring_exit_work()
(kworker != submitter_task), io_run_local_work() is never invoked,
and the work_llist entries are never processed. This causes
io_uring_try_cancel_requests() to loop indefinitely, resulting in
100% CPU usage in kworker threads.
Fix this by moving io_move_task_work_from_local() inside the cancel
loop, ensuring any work on work_llist is moved to fallback before
each cancel attempt.
Cc: stable@vger.kernel.org
Fixes: c0e0d6ba25f1 ("io_uring: add IORING_SETUP_DEFER_TASKRUN")
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2026-01-14 16:54:05 +08:00
|
|
|
|
2022-06-20 01:25:52 +01:00
|
|
|
|
2021-04-25 23:34:45 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-06-11 13:53:43 -07:00
|
|
|
|
2021-04-25 23:34:45 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-03-06 11:02:16 +00:00
|
|
|
|
2021-10-04 20:02:53 +01:00
|
|
|
|
|
|
|
|
|
2021-08-09 13:04:17 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: wait interruptibly for request completions on exit
WHen the ring exits, cleanup is done and the final cancelation and
waiting on completions is done by io_ring_exit_work. That function is
invoked by kworker, which doesn't take any signals. Because of that, it
doesn't really matter if we wait for completions in TASK_INTERRUPTIBLE
or TASK_UNINTERRUPTIBLE state. However, it does matter to the hung task
detection checker!
Normally we expect cancelations and completions to happen rather
quickly. Some test cases, however, will exit the ring and park the
owning task stopped (eg via SIGSTOP). If the owning task needs to run
task_work to complete requests, then io_ring_exit_work won't make any
progress until the task is runnable again. Hence io_ring_exit_work can
trigger the hung task detection, which is particularly problematic if
panic-on-hung-task is enabled.
As the ring exit doesn't take signals to begin with, have it wait
interruptibly rather than uninterruptibly. io_uring has a separate
stuck-exit warning that triggers independently anyway, so we're not
really missing anything by making this switch.
Cc: stable@vger.kernel.org # 5.10+
Link: https://lore.kernel.org/r/b0e4aaef-7088-56ce-244c-976edeac0e66@kernel.dk
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2023-06-11 21:14:09 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
2021-04-14 13:38:34 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-12-03 15:37:53 +00:00
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
io_uring/tctx: add separate lock for list of tctx's in ctx
ctx->tcxt_list holds the tasks using this ring, and it's currently
protected by the normal ctx->uring_lock. However, this can cause a
circular locking issue, as reported by syzbot, where cancelations off
exec end up needing to remove an entry from this list:
======================================================
WARNING: possible circular locking dependency detected
syzkaller #0 Tainted: G L
------------------------------------------------------
syz.0.9999/12287 is trying to acquire lock:
ffff88805851c0a8 (&ctx->uring_lock){+.+.}-{4:4}, at: io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
but task is already holding lock:
ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: prepare_bprm_creds fs/exec.c:1360 [inline]
ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: bprm_execve+0xb9/0x1400 fs/exec.c:1733
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #2 (&sig->cred_guard_mutex){+.+.}-{4:4}:
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
proc_pid_attr_write+0x547/0x630 fs/proc/base.c:2837
vfs_write+0x27e/0xb30 fs/read_write.c:684
ksys_write+0x145/0x250 fs/read_write.c:738
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
-> #1 (sb_writers#3){.+.+}-{0:0}:
percpu_down_read_internal include/linux/percpu-rwsem.h:53 [inline]
percpu_down_read_freezable include/linux/percpu-rwsem.h:83 [inline]
__sb_start_write include/linux/fs/super.h:19 [inline]
sb_start_write+0x4d/0x1c0 include/linux/fs/super.h:125
mnt_want_write+0x41/0x90 fs/namespace.c:499
open_last_lookups fs/namei.c:4529 [inline]
path_openat+0xadd/0x3dd0 fs/namei.c:4784
do_filp_open+0x1fa/0x410 fs/namei.c:4814
io_openat2+0x3e0/0x5c0 io_uring/openclose.c:143
__io_issue_sqe+0x181/0x4b0 io_uring/io_uring.c:1792
io_issue_sqe+0x165/0x1060 io_uring/io_uring.c:1815
io_queue_sqe io_uring/io_uring.c:2042 [inline]
io_submit_sqe io_uring/io_uring.c:2320 [inline]
io_submit_sqes+0xbf4/0x2140 io_uring/io_uring.c:2434
__do_sys_io_uring_enter io_uring/io_uring.c:3280 [inline]
__se_sys_io_uring_enter+0x2e0/0x2b60 io_uring/io_uring.c:3219
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
-> #0 (&ctx->uring_lock){+.+.}-{4:4}:
check_prev_add kernel/locking/lockdep.c:3165 [inline]
check_prevs_add kernel/locking/lockdep.c:3284 [inline]
validate_chain kernel/locking/lockdep.c:3908 [inline]
__lock_acquire+0x15a6/0x2cf0 kernel/locking/lockdep.c:5237
lock_acquire+0x107/0x340 kernel/locking/lockdep.c:5868
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
io_uring_clean_tctx+0xd4/0x1a0 io_uring/tctx.c:195
io_uring_cancel_generic+0x6ca/0x7d0 io_uring/cancel.c:646
io_uring_task_cancel include/linux/io_uring.h:24 [inline]
begin_new_exec+0x10ed/0x2440 fs/exec.c:1131
load_elf_binary+0x9f8/0x2d70 fs/binfmt_elf.c:1010
search_binary_handler fs/exec.c:1669 [inline]
exec_binprm fs/exec.c:1701 [inline]
bprm_execve+0x92e/0x1400 fs/exec.c:1753
do_execveat_common+0x510/0x6a0 fs/exec.c:1859
do_execve fs/exec.c:1933 [inline]
__do_sys_execve fs/exec.c:2009 [inline]
__se_sys_execve fs/exec.c:2004 [inline]
__x64_sys_execve+0x94/0xb0 fs/exec.c:2004
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
other info that might help us debug this:
Chain exists of:
&ctx->uring_lock --> sb_writers#3 --> &sig->cred_guard_mutex
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&sig->cred_guard_mutex);
lock(sb_writers#3);
lock(&sig->cred_guard_mutex);
lock(&ctx->uring_lock);
*** DEADLOCK ***
1 lock held by syz.0.9999/12287:
#0: ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: prepare_bprm_creds fs/exec.c:1360 [inline]
#0: ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: bprm_execve+0xb9/0x1400 fs/exec.c:1733
stack backtrace:
CPU: 0 UID: 0 PID: 12287 Comm: syz.0.9999 Tainted: G L syzkaller #0 PREEMPT(full)
Tainted: [L]=SOFTLOCKUP
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/25/2025
Call Trace:
<TASK>
dump_stack_lvl+0xe8/0x150 lib/dump_stack.c:120
print_circular_bug+0x2e2/0x300 kernel/locking/lockdep.c:2043
check_noncircular+0x12e/0x150 kernel/locking/lockdep.c:2175
check_prev_add kernel/locking/lockdep.c:3165 [inline]
check_prevs_add kernel/locking/lockdep.c:3284 [inline]
validate_chain kernel/locking/lockdep.c:3908 [inline]
__lock_acquire+0x15a6/0x2cf0 kernel/locking/lockdep.c:5237
lock_acquire+0x107/0x340 kernel/locking/lockdep.c:5868
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
io_uring_clean_tctx+0xd4/0x1a0 io_uring/tctx.c:195
io_uring_cancel_generic+0x6ca/0x7d0 io_uring/cancel.c:646
io_uring_task_cancel include/linux/io_uring.h:24 [inline]
begin_new_exec+0x10ed/0x2440 fs/exec.c:1131
load_elf_binary+0x9f8/0x2d70 fs/binfmt_elf.c:1010
search_binary_handler fs/exec.c:1669 [inline]
exec_binprm fs/exec.c:1701 [inline]
bprm_execve+0x92e/0x1400 fs/exec.c:1753
do_execveat_common+0x510/0x6a0 fs/exec.c:1859
do_execve fs/exec.c:1933 [inline]
__do_sys_execve fs/exec.c:2009 [inline]
__se_sys_execve fs/exec.c:2004 [inline]
__x64_sys_execve+0x94/0xb0 fs/exec.c:2004
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7ff3a8b8f749
Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 a8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007ff3a9a97038 EFLAGS: 00000246 ORIG_RAX: 000000000000003b
RAX: ffffffffffffffda RBX: 00007ff3a8de5fa0 RCX: 00007ff3a8b8f749
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000200000000400
RBP: 00007ff3a8c13f91 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 00007ff3a8de6038 R14: 00007ff3a8de5fa0 R15: 00007ff3a8f0fa28
</TASK>
Add a separate lock just for the tctx_list, tctx_lock. This can nest
under ->uring_lock, where necessary, and be used separately for list
manipulation. For the cancelation off exec side, this removes the
need to grab ->uring_lock, hence fixing the circular locking
dependency.
Reported-by: syzbot+b0e3b77ffaa8a4067ce5@syzkaller.appspotmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-12-31 08:12:46 -07:00
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
2021-03-06 11:02:16 +00:00
|
|
|
|
|
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
|
|
|
|
|
2021-04-14 13:38:34 +01:00
|
|
|
|
|
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring/tctx: add separate lock for list of tctx's in ctx
ctx->tcxt_list holds the tasks using this ring, and it's currently
protected by the normal ctx->uring_lock. However, this can cause a
circular locking issue, as reported by syzbot, where cancelations off
exec end up needing to remove an entry from this list:
======================================================
WARNING: possible circular locking dependency detected
syzkaller #0 Tainted: G L
------------------------------------------------------
syz.0.9999/12287 is trying to acquire lock:
ffff88805851c0a8 (&ctx->uring_lock){+.+.}-{4:4}, at: io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
but task is already holding lock:
ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: prepare_bprm_creds fs/exec.c:1360 [inline]
ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: bprm_execve+0xb9/0x1400 fs/exec.c:1733
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #2 (&sig->cred_guard_mutex){+.+.}-{4:4}:
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
proc_pid_attr_write+0x547/0x630 fs/proc/base.c:2837
vfs_write+0x27e/0xb30 fs/read_write.c:684
ksys_write+0x145/0x250 fs/read_write.c:738
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
-> #1 (sb_writers#3){.+.+}-{0:0}:
percpu_down_read_internal include/linux/percpu-rwsem.h:53 [inline]
percpu_down_read_freezable include/linux/percpu-rwsem.h:83 [inline]
__sb_start_write include/linux/fs/super.h:19 [inline]
sb_start_write+0x4d/0x1c0 include/linux/fs/super.h:125
mnt_want_write+0x41/0x90 fs/namespace.c:499
open_last_lookups fs/namei.c:4529 [inline]
path_openat+0xadd/0x3dd0 fs/namei.c:4784
do_filp_open+0x1fa/0x410 fs/namei.c:4814
io_openat2+0x3e0/0x5c0 io_uring/openclose.c:143
__io_issue_sqe+0x181/0x4b0 io_uring/io_uring.c:1792
io_issue_sqe+0x165/0x1060 io_uring/io_uring.c:1815
io_queue_sqe io_uring/io_uring.c:2042 [inline]
io_submit_sqe io_uring/io_uring.c:2320 [inline]
io_submit_sqes+0xbf4/0x2140 io_uring/io_uring.c:2434
__do_sys_io_uring_enter io_uring/io_uring.c:3280 [inline]
__se_sys_io_uring_enter+0x2e0/0x2b60 io_uring/io_uring.c:3219
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
-> #0 (&ctx->uring_lock){+.+.}-{4:4}:
check_prev_add kernel/locking/lockdep.c:3165 [inline]
check_prevs_add kernel/locking/lockdep.c:3284 [inline]
validate_chain kernel/locking/lockdep.c:3908 [inline]
__lock_acquire+0x15a6/0x2cf0 kernel/locking/lockdep.c:5237
lock_acquire+0x107/0x340 kernel/locking/lockdep.c:5868
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
io_uring_clean_tctx+0xd4/0x1a0 io_uring/tctx.c:195
io_uring_cancel_generic+0x6ca/0x7d0 io_uring/cancel.c:646
io_uring_task_cancel include/linux/io_uring.h:24 [inline]
begin_new_exec+0x10ed/0x2440 fs/exec.c:1131
load_elf_binary+0x9f8/0x2d70 fs/binfmt_elf.c:1010
search_binary_handler fs/exec.c:1669 [inline]
exec_binprm fs/exec.c:1701 [inline]
bprm_execve+0x92e/0x1400 fs/exec.c:1753
do_execveat_common+0x510/0x6a0 fs/exec.c:1859
do_execve fs/exec.c:1933 [inline]
__do_sys_execve fs/exec.c:2009 [inline]
__se_sys_execve fs/exec.c:2004 [inline]
__x64_sys_execve+0x94/0xb0 fs/exec.c:2004
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
other info that might help us debug this:
Chain exists of:
&ctx->uring_lock --> sb_writers#3 --> &sig->cred_guard_mutex
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&sig->cred_guard_mutex);
lock(sb_writers#3);
lock(&sig->cred_guard_mutex);
lock(&ctx->uring_lock);
*** DEADLOCK ***
1 lock held by syz.0.9999/12287:
#0: ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: prepare_bprm_creds fs/exec.c:1360 [inline]
#0: ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: bprm_execve+0xb9/0x1400 fs/exec.c:1733
stack backtrace:
CPU: 0 UID: 0 PID: 12287 Comm: syz.0.9999 Tainted: G L syzkaller #0 PREEMPT(full)
Tainted: [L]=SOFTLOCKUP
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/25/2025
Call Trace:
<TASK>
dump_stack_lvl+0xe8/0x150 lib/dump_stack.c:120
print_circular_bug+0x2e2/0x300 kernel/locking/lockdep.c:2043
check_noncircular+0x12e/0x150 kernel/locking/lockdep.c:2175
check_prev_add kernel/locking/lockdep.c:3165 [inline]
check_prevs_add kernel/locking/lockdep.c:3284 [inline]
validate_chain kernel/locking/lockdep.c:3908 [inline]
__lock_acquire+0x15a6/0x2cf0 kernel/locking/lockdep.c:5237
lock_acquire+0x107/0x340 kernel/locking/lockdep.c:5868
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
io_uring_clean_tctx+0xd4/0x1a0 io_uring/tctx.c:195
io_uring_cancel_generic+0x6ca/0x7d0 io_uring/cancel.c:646
io_uring_task_cancel include/linux/io_uring.h:24 [inline]
begin_new_exec+0x10ed/0x2440 fs/exec.c:1131
load_elf_binary+0x9f8/0x2d70 fs/binfmt_elf.c:1010
search_binary_handler fs/exec.c:1669 [inline]
exec_binprm fs/exec.c:1701 [inline]
bprm_execve+0x92e/0x1400 fs/exec.c:1753
do_execveat_common+0x510/0x6a0 fs/exec.c:1859
do_execve fs/exec.c:1933 [inline]
__do_sys_execve fs/exec.c:2009 [inline]
__se_sys_execve fs/exec.c:2004 [inline]
__x64_sys_execve+0x94/0xb0 fs/exec.c:2004
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7ff3a8b8f749
Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 a8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007ff3a9a97038 EFLAGS: 00000246 ORIG_RAX: 000000000000003b
RAX: ffffffffffffffda RBX: 00007ff3a8de5fa0 RCX: 00007ff3a8b8f749
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000200000000400
RBP: 00007ff3a8c13f91 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 00007ff3a8de6038 R14: 00007ff3a8de5fa0 R15: 00007ff3a8f0fa28
</TASK>
Add a separate lock just for the tctx_list, tctx_lock. This can nest
under ->uring_lock, where necessary, and be used separately for list
manipulation. For the cancelation off exec side, this removes the
need to grab ->uring_lock, hence fixing the circular locking
dependency.
Reported-by: syzbot+b0e3b77ffaa8a4067ce5@syzkaller.appspotmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-12-31 08:12:46 -07:00
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
io_uring: wait interruptibly for request completions on exit
WHen the ring exits, cleanup is done and the final cancelation and
waiting on completions is done by io_ring_exit_work. That function is
invoked by kworker, which doesn't take any signals. Because of that, it
doesn't really matter if we wait for completions in TASK_INTERRUPTIBLE
or TASK_UNINTERRUPTIBLE state. However, it does matter to the hung task
detection checker!
Normally we expect cancelations and completions to happen rather
quickly. Some test cases, however, will exit the ring and park the
owning task stopped (eg via SIGSTOP). If the owning task needs to run
task_work to complete requests, then io_ring_exit_work won't make any
progress until the task is runnable again. Hence io_ring_exit_work can
trigger the hung task detection, which is particularly problematic if
panic-on-hung-task is enabled.
As the ring exit doesn't take signals to begin with, have it wait
interruptibly rather than uninterruptibly. io_uring has a separate
stuck-exit warning that triggers independently anyway, so we're not
really missing anything by making this switch.
Cc: stable@vger.kernel.org # 5.10+
Link: https://lore.kernel.org/r/b0e4aaef-7088-56ce-244c-976edeac0e66@kernel.dk
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2023-06-11 21:14:09 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
io_uring/tctx: add separate lock for list of tctx's in ctx
ctx->tcxt_list holds the tasks using this ring, and it's currently
protected by the normal ctx->uring_lock. However, this can cause a
circular locking issue, as reported by syzbot, where cancelations off
exec end up needing to remove an entry from this list:
======================================================
WARNING: possible circular locking dependency detected
syzkaller #0 Tainted: G L
------------------------------------------------------
syz.0.9999/12287 is trying to acquire lock:
ffff88805851c0a8 (&ctx->uring_lock){+.+.}-{4:4}, at: io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
but task is already holding lock:
ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: prepare_bprm_creds fs/exec.c:1360 [inline]
ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: bprm_execve+0xb9/0x1400 fs/exec.c:1733
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #2 (&sig->cred_guard_mutex){+.+.}-{4:4}:
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
proc_pid_attr_write+0x547/0x630 fs/proc/base.c:2837
vfs_write+0x27e/0xb30 fs/read_write.c:684
ksys_write+0x145/0x250 fs/read_write.c:738
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
-> #1 (sb_writers#3){.+.+}-{0:0}:
percpu_down_read_internal include/linux/percpu-rwsem.h:53 [inline]
percpu_down_read_freezable include/linux/percpu-rwsem.h:83 [inline]
__sb_start_write include/linux/fs/super.h:19 [inline]
sb_start_write+0x4d/0x1c0 include/linux/fs/super.h:125
mnt_want_write+0x41/0x90 fs/namespace.c:499
open_last_lookups fs/namei.c:4529 [inline]
path_openat+0xadd/0x3dd0 fs/namei.c:4784
do_filp_open+0x1fa/0x410 fs/namei.c:4814
io_openat2+0x3e0/0x5c0 io_uring/openclose.c:143
__io_issue_sqe+0x181/0x4b0 io_uring/io_uring.c:1792
io_issue_sqe+0x165/0x1060 io_uring/io_uring.c:1815
io_queue_sqe io_uring/io_uring.c:2042 [inline]
io_submit_sqe io_uring/io_uring.c:2320 [inline]
io_submit_sqes+0xbf4/0x2140 io_uring/io_uring.c:2434
__do_sys_io_uring_enter io_uring/io_uring.c:3280 [inline]
__se_sys_io_uring_enter+0x2e0/0x2b60 io_uring/io_uring.c:3219
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
-> #0 (&ctx->uring_lock){+.+.}-{4:4}:
check_prev_add kernel/locking/lockdep.c:3165 [inline]
check_prevs_add kernel/locking/lockdep.c:3284 [inline]
validate_chain kernel/locking/lockdep.c:3908 [inline]
__lock_acquire+0x15a6/0x2cf0 kernel/locking/lockdep.c:5237
lock_acquire+0x107/0x340 kernel/locking/lockdep.c:5868
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
io_uring_clean_tctx+0xd4/0x1a0 io_uring/tctx.c:195
io_uring_cancel_generic+0x6ca/0x7d0 io_uring/cancel.c:646
io_uring_task_cancel include/linux/io_uring.h:24 [inline]
begin_new_exec+0x10ed/0x2440 fs/exec.c:1131
load_elf_binary+0x9f8/0x2d70 fs/binfmt_elf.c:1010
search_binary_handler fs/exec.c:1669 [inline]
exec_binprm fs/exec.c:1701 [inline]
bprm_execve+0x92e/0x1400 fs/exec.c:1753
do_execveat_common+0x510/0x6a0 fs/exec.c:1859
do_execve fs/exec.c:1933 [inline]
__do_sys_execve fs/exec.c:2009 [inline]
__se_sys_execve fs/exec.c:2004 [inline]
__x64_sys_execve+0x94/0xb0 fs/exec.c:2004
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
other info that might help us debug this:
Chain exists of:
&ctx->uring_lock --> sb_writers#3 --> &sig->cred_guard_mutex
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&sig->cred_guard_mutex);
lock(sb_writers#3);
lock(&sig->cred_guard_mutex);
lock(&ctx->uring_lock);
*** DEADLOCK ***
1 lock held by syz.0.9999/12287:
#0: ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: prepare_bprm_creds fs/exec.c:1360 [inline]
#0: ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: bprm_execve+0xb9/0x1400 fs/exec.c:1733
stack backtrace:
CPU: 0 UID: 0 PID: 12287 Comm: syz.0.9999 Tainted: G L syzkaller #0 PREEMPT(full)
Tainted: [L]=SOFTLOCKUP
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/25/2025
Call Trace:
<TASK>
dump_stack_lvl+0xe8/0x150 lib/dump_stack.c:120
print_circular_bug+0x2e2/0x300 kernel/locking/lockdep.c:2043
check_noncircular+0x12e/0x150 kernel/locking/lockdep.c:2175
check_prev_add kernel/locking/lockdep.c:3165 [inline]
check_prevs_add kernel/locking/lockdep.c:3284 [inline]
validate_chain kernel/locking/lockdep.c:3908 [inline]
__lock_acquire+0x15a6/0x2cf0 kernel/locking/lockdep.c:5237
lock_acquire+0x107/0x340 kernel/locking/lockdep.c:5868
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
io_uring_clean_tctx+0xd4/0x1a0 io_uring/tctx.c:195
io_uring_cancel_generic+0x6ca/0x7d0 io_uring/cancel.c:646
io_uring_task_cancel include/linux/io_uring.h:24 [inline]
begin_new_exec+0x10ed/0x2440 fs/exec.c:1131
load_elf_binary+0x9f8/0x2d70 fs/binfmt_elf.c:1010
search_binary_handler fs/exec.c:1669 [inline]
exec_binprm fs/exec.c:1701 [inline]
bprm_execve+0x92e/0x1400 fs/exec.c:1753
do_execveat_common+0x510/0x6a0 fs/exec.c:1859
do_execve fs/exec.c:1933 [inline]
__do_sys_execve fs/exec.c:2009 [inline]
__se_sys_execve fs/exec.c:2004 [inline]
__x64_sys_execve+0x94/0xb0 fs/exec.c:2004
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7ff3a8b8f749
Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 a8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007ff3a9a97038 EFLAGS: 00000246 ORIG_RAX: 000000000000003b
RAX: ffffffffffffffda RBX: 00007ff3a8de5fa0 RCX: 00007ff3a8b8f749
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000200000000400
RBP: 00007ff3a8c13f91 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 00007ff3a8de6038 R14: 00007ff3a8de5fa0 R15: 00007ff3a8f0fa28
</TASK>
Add a separate lock just for the tctx_list, tctx_lock. This can nest
under ->uring_lock, where necessary, and be used separately for list
manipulation. For the cancelation off exec side, this removes the
need to grab ->uring_lock, hence fixing the circular locking
dependency.
Reported-by: syzbot+b0e3b77ffaa8a4067ce5@syzkaller.appspotmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-12-31 08:12:46 -07:00
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
io_uring/tctx: add separate lock for list of tctx's in ctx
ctx->tcxt_list holds the tasks using this ring, and it's currently
protected by the normal ctx->uring_lock. However, this can cause a
circular locking issue, as reported by syzbot, where cancelations off
exec end up needing to remove an entry from this list:
======================================================
WARNING: possible circular locking dependency detected
syzkaller #0 Tainted: G L
------------------------------------------------------
syz.0.9999/12287 is trying to acquire lock:
ffff88805851c0a8 (&ctx->uring_lock){+.+.}-{4:4}, at: io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
but task is already holding lock:
ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: prepare_bprm_creds fs/exec.c:1360 [inline]
ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: bprm_execve+0xb9/0x1400 fs/exec.c:1733
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #2 (&sig->cred_guard_mutex){+.+.}-{4:4}:
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
proc_pid_attr_write+0x547/0x630 fs/proc/base.c:2837
vfs_write+0x27e/0xb30 fs/read_write.c:684
ksys_write+0x145/0x250 fs/read_write.c:738
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
-> #1 (sb_writers#3){.+.+}-{0:0}:
percpu_down_read_internal include/linux/percpu-rwsem.h:53 [inline]
percpu_down_read_freezable include/linux/percpu-rwsem.h:83 [inline]
__sb_start_write include/linux/fs/super.h:19 [inline]
sb_start_write+0x4d/0x1c0 include/linux/fs/super.h:125
mnt_want_write+0x41/0x90 fs/namespace.c:499
open_last_lookups fs/namei.c:4529 [inline]
path_openat+0xadd/0x3dd0 fs/namei.c:4784
do_filp_open+0x1fa/0x410 fs/namei.c:4814
io_openat2+0x3e0/0x5c0 io_uring/openclose.c:143
__io_issue_sqe+0x181/0x4b0 io_uring/io_uring.c:1792
io_issue_sqe+0x165/0x1060 io_uring/io_uring.c:1815
io_queue_sqe io_uring/io_uring.c:2042 [inline]
io_submit_sqe io_uring/io_uring.c:2320 [inline]
io_submit_sqes+0xbf4/0x2140 io_uring/io_uring.c:2434
__do_sys_io_uring_enter io_uring/io_uring.c:3280 [inline]
__se_sys_io_uring_enter+0x2e0/0x2b60 io_uring/io_uring.c:3219
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
-> #0 (&ctx->uring_lock){+.+.}-{4:4}:
check_prev_add kernel/locking/lockdep.c:3165 [inline]
check_prevs_add kernel/locking/lockdep.c:3284 [inline]
validate_chain kernel/locking/lockdep.c:3908 [inline]
__lock_acquire+0x15a6/0x2cf0 kernel/locking/lockdep.c:5237
lock_acquire+0x107/0x340 kernel/locking/lockdep.c:5868
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
io_uring_clean_tctx+0xd4/0x1a0 io_uring/tctx.c:195
io_uring_cancel_generic+0x6ca/0x7d0 io_uring/cancel.c:646
io_uring_task_cancel include/linux/io_uring.h:24 [inline]
begin_new_exec+0x10ed/0x2440 fs/exec.c:1131
load_elf_binary+0x9f8/0x2d70 fs/binfmt_elf.c:1010
search_binary_handler fs/exec.c:1669 [inline]
exec_binprm fs/exec.c:1701 [inline]
bprm_execve+0x92e/0x1400 fs/exec.c:1753
do_execveat_common+0x510/0x6a0 fs/exec.c:1859
do_execve fs/exec.c:1933 [inline]
__do_sys_execve fs/exec.c:2009 [inline]
__se_sys_execve fs/exec.c:2004 [inline]
__x64_sys_execve+0x94/0xb0 fs/exec.c:2004
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
other info that might help us debug this:
Chain exists of:
&ctx->uring_lock --> sb_writers#3 --> &sig->cred_guard_mutex
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&sig->cred_guard_mutex);
lock(sb_writers#3);
lock(&sig->cred_guard_mutex);
lock(&ctx->uring_lock);
*** DEADLOCK ***
1 lock held by syz.0.9999/12287:
#0: ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: prepare_bprm_creds fs/exec.c:1360 [inline]
#0: ffff88802db5a2e0 (&sig->cred_guard_mutex){+.+.}-{4:4}, at: bprm_execve+0xb9/0x1400 fs/exec.c:1733
stack backtrace:
CPU: 0 UID: 0 PID: 12287 Comm: syz.0.9999 Tainted: G L syzkaller #0 PREEMPT(full)
Tainted: [L]=SOFTLOCKUP
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 10/25/2025
Call Trace:
<TASK>
dump_stack_lvl+0xe8/0x150 lib/dump_stack.c:120
print_circular_bug+0x2e2/0x300 kernel/locking/lockdep.c:2043
check_noncircular+0x12e/0x150 kernel/locking/lockdep.c:2175
check_prev_add kernel/locking/lockdep.c:3165 [inline]
check_prevs_add kernel/locking/lockdep.c:3284 [inline]
validate_chain kernel/locking/lockdep.c:3908 [inline]
__lock_acquire+0x15a6/0x2cf0 kernel/locking/lockdep.c:5237
lock_acquire+0x107/0x340 kernel/locking/lockdep.c:5868
__mutex_lock_common kernel/locking/mutex.c:614 [inline]
__mutex_lock+0x187/0x1350 kernel/locking/mutex.c:776
io_uring_del_tctx_node+0xf0/0x2c0 io_uring/tctx.c:179
io_uring_clean_tctx+0xd4/0x1a0 io_uring/tctx.c:195
io_uring_cancel_generic+0x6ca/0x7d0 io_uring/cancel.c:646
io_uring_task_cancel include/linux/io_uring.h:24 [inline]
begin_new_exec+0x10ed/0x2440 fs/exec.c:1131
load_elf_binary+0x9f8/0x2d70 fs/binfmt_elf.c:1010
search_binary_handler fs/exec.c:1669 [inline]
exec_binprm fs/exec.c:1701 [inline]
bprm_execve+0x92e/0x1400 fs/exec.c:1753
do_execveat_common+0x510/0x6a0 fs/exec.c:1859
do_execve fs/exec.c:1933 [inline]
__do_sys_execve fs/exec.c:2009 [inline]
__se_sys_execve fs/exec.c:2004 [inline]
__x64_sys_execve+0x94/0xb0 fs/exec.c:2004
do_syscall_x64 arch/x86/entry/syscall_64.c:63 [inline]
do_syscall_64+0xec/0xf80 arch/x86/entry/syscall_64.c:94
entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7ff3a8b8f749
Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 a8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007ff3a9a97038 EFLAGS: 00000246 ORIG_RAX: 000000000000003b
RAX: ffffffffffffffda RBX: 00007ff3a8de5fa0 RCX: 00007ff3a8b8f749
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000200000000400
RBP: 00007ff3a8c13f91 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 00007ff3a8de6038 R14: 00007ff3a8de5fa0 R15: 00007ff3a8f0fa28
</TASK>
Add a separate lock just for the tctx_list, tctx_lock. This can nest
under ->uring_lock, where necessary, and be used separately for list
manipulation. For the cancelation off exec side, this removes the
need to grab ->uring_lock, hence fixing the circular locking
dependency.
Reported-by: syzbot+b0e3b77ffaa8a4067ce5@syzkaller.appspotmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-12-31 08:12:46 -07:00
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
2021-08-10 15:18:27 -06:00
|
|
|
|
|
|
|
|
|
2021-03-06 11:02:13 +00:00
|
|
|
|
2023-04-06 14:20:08 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-04-09 18:14:00 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-10-04 20:02:54 +01:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2021-03-08 14:16:16 +00:00
|
|
|
|
2026-02-11 19:47:38 -07:00
|
|
|
|
2021-03-08 14:16:16 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2021-03-08 14:16:16 +00:00
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2020-04-09 18:14:00 -06:00
|
|
|
|
2020-08-19 11:10:51 -06:00
|
|
|
|
2025-09-05 11:02:40 +02:00
|
|
|
|
2020-08-19 11:10:51 -06:00
|
|
|
|
|
|
|
|
|
2025-09-05 11:02:39 +02:00
|
|
|
|
2020-08-19 11:10:51 -06:00
|
|
|
|
2024-04-01 15:16:19 -06:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add support for fixed wait regions
Generally applications have 1 or a few waits of waiting, yet they pass
in a struct io_uring_getevents_arg every time. This needs to get copied
and, in turn, the timeout value needs to get copied.
Rather than do this for every invocation, allow the application to
register a fixed set of wait regions that can simply be indexed when
asking the kernel to wait on events.
At ring setup time, the application can register a number of these wait
regions and initialize region/index 0 upfront:
struct io_uring_reg_wait *reg;
reg = io_uring_setup_reg_wait(ring, nr_regions, &ret);
/* set timeout and mark as set, sigmask/sigmask_sz as needed */
reg->ts.tv_sec = 0;
reg->ts.tv_nsec = 100000;
reg->flags = IORING_REG_WAIT_TS;
where nr_regions >= 1 && nr_regions <= PAGE_SIZE / sizeof(*reg). The
above initializes index 0, but 63 other regions can be initialized,
if needed. Now, instead of doing:
struct __kernel_timespec timeout = { .tv_nsec = 100000, };
io_uring_submit_and_wait_timeout(ring, &cqe, nr, &t, NULL);
to wait for events for each submit_and_wait, or just wait, operation, it
can just reference the above region at offset 0 and do:
io_uring_submit_and_wait_reg(ring, &cqe, nr, 0);
to achieve the same goal of waiting 100usec without needing to copy
both struct io_uring_getevents_arg (24b) and struct __kernel_timeout
(16b) for each invocation. Struct io_uring_reg_wait looks as follows:
struct io_uring_reg_wait {
struct __kernel_timespec ts;
__u32 min_wait_usec;
__u32 flags;
__u64 sigmask;
__u32 sigmask_sz;
__u32 pad[3];
__u64 pad2[2];
};
embedding the timeout itself in the region, rather than passing it as
a pointer as well. Note that the signal mask is still passed as a
pointer, both for compatability reasons, but also because there doesn't
seem to be a lot of high frequency waits scenarios that involve setting
and resetting the signal mask for each wait.
The application is free to modify any region before a wait call, or it
can use keep multiple regions with different settings to avoid needing to
modify the same one for wait calls. Up to a page size of regions is mapped
by default, allowing PAGE_SIZE / 64 available regions for use.
The registered region must fit within a page. On a 4kb page size system,
that allows for 64 wait regions if a full page is used, as the size of
struct io_uring_reg_wait is 64b. The region registered must be aligned
to io_uring_reg_wait in size. It's valid to register less than 64
entries.
In network performance testing with zero-copy, this reduced the time
spent waiting on the TX side from 3.12% to 0.3% and the RX side from 4.4%
to 0.3%.
Wait regions are fixed for the lifetime of the ring - once registered,
they are persistent until the ring is torn down. The regions support
minimum wait timeout as well as the regular waits.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-10-22 13:47:00 -06:00
|
|
|
|
|
|
|
|
|
2022-03-22 14:07:56 +00:00
|
|
|
|
2024-11-15 16:54:43 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-03-22 14:07:56 +00:00
|
|
|
|
2024-11-15 16:54:43 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-12-08 21:43:22 +00:00
|
|
|
|
2024-11-15 16:54:43 +00:00
|
|
|
|
io_uring: add support for fixed wait regions
Generally applications have 1 or a few waits of waiting, yet they pass
in a struct io_uring_getevents_arg every time. This needs to get copied
and, in turn, the timeout value needs to get copied.
Rather than do this for every invocation, allow the application to
register a fixed set of wait regions that can simply be indexed when
asking the kernel to wait on events.
At ring setup time, the application can register a number of these wait
regions and initialize region/index 0 upfront:
struct io_uring_reg_wait *reg;
reg = io_uring_setup_reg_wait(ring, nr_regions, &ret);
/* set timeout and mark as set, sigmask/sigmask_sz as needed */
reg->ts.tv_sec = 0;
reg->ts.tv_nsec = 100000;
reg->flags = IORING_REG_WAIT_TS;
where nr_regions >= 1 && nr_regions <= PAGE_SIZE / sizeof(*reg). The
above initializes index 0, but 63 other regions can be initialized,
if needed. Now, instead of doing:
struct __kernel_timespec timeout = { .tv_nsec = 100000, };
io_uring_submit_and_wait_timeout(ring, &cqe, nr, &t, NULL);
to wait for events for each submit_and_wait, or just wait, operation, it
can just reference the above region at offset 0 and do:
io_uring_submit_and_wait_reg(ring, &cqe, nr, 0);
to achieve the same goal of waiting 100usec without needing to copy
both struct io_uring_getevents_arg (24b) and struct __kernel_timeout
(16b) for each invocation. Struct io_uring_reg_wait looks as follows:
struct io_uring_reg_wait {
struct __kernel_timespec ts;
__u32 min_wait_usec;
__u32 flags;
__u64 sigmask;
__u32 sigmask_sz;
__u32 pad[3];
__u64 pad2[2];
};
embedding the timeout itself in the region, rather than passing it as
a pointer as well. Note that the signal mask is still passed as a
pointer, both for compatability reasons, but also because there doesn't
seem to be a lot of high frequency waits scenarios that involve setting
and resetting the signal mask for each wait.
The application is free to modify any region before a wait call, or it
can use keep multiple regions with different settings to avoid needing to
modify the same one for wait calls. Up to a page size of regions is mapped
by default, allowing PAGE_SIZE / 64 available regions for use.
The registered region must fit within a page. On a 4kb page size system,
that allows for 64 wait regions if a full page is used, as the size of
struct io_uring_reg_wait is 64b. The region registered must be aligned
to io_uring_reg_wait in size. It's valid to register less than 64
entries.
In network performance testing with zero-copy, this reduced the time
spent waiting on the TX side from 3.12% to 0.3% and the RX side from 4.4%
to 0.3%.
Wait regions are fixed for the lifetime of the ring - once registered,
they are persistent until the ring is torn down. The regions support
minimum wait timeout as well as the regular waits.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-10-22 13:47:00 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-11-15 16:54:39 +00:00
|
|
|
|
|
|
|
|
|
io_uring: add support for fixed wait regions
Generally applications have 1 or a few waits of waiting, yet they pass
in a struct io_uring_getevents_arg every time. This needs to get copied
and, in turn, the timeout value needs to get copied.
Rather than do this for every invocation, allow the application to
register a fixed set of wait regions that can simply be indexed when
asking the kernel to wait on events.
At ring setup time, the application can register a number of these wait
regions and initialize region/index 0 upfront:
struct io_uring_reg_wait *reg;
reg = io_uring_setup_reg_wait(ring, nr_regions, &ret);
/* set timeout and mark as set, sigmask/sigmask_sz as needed */
reg->ts.tv_sec = 0;
reg->ts.tv_nsec = 100000;
reg->flags = IORING_REG_WAIT_TS;
where nr_regions >= 1 && nr_regions <= PAGE_SIZE / sizeof(*reg). The
above initializes index 0, but 63 other regions can be initialized,
if needed. Now, instead of doing:
struct __kernel_timespec timeout = { .tv_nsec = 100000, };
io_uring_submit_and_wait_timeout(ring, &cqe, nr, &t, NULL);
to wait for events for each submit_and_wait, or just wait, operation, it
can just reference the above region at offset 0 and do:
io_uring_submit_and_wait_reg(ring, &cqe, nr, 0);
to achieve the same goal of waiting 100usec without needing to copy
both struct io_uring_getevents_arg (24b) and struct __kernel_timeout
(16b) for each invocation. Struct io_uring_reg_wait looks as follows:
struct io_uring_reg_wait {
struct __kernel_timespec ts;
__u32 min_wait_usec;
__u32 flags;
__u64 sigmask;
__u32 sigmask_sz;
__u32 pad[3];
__u64 pad2[2];
};
embedding the timeout itself in the region, rather than passing it as
a pointer as well. Note that the signal mask is still passed as a
pointer, both for compatability reasons, but also because there doesn't
seem to be a lot of high frequency waits scenarios that involve setting
and resetting the signal mask for each wait.
The application is free to modify any region before a wait call, or it
can use keep multiple regions with different settings to avoid needing to
modify the same one for wait calls. Up to a page size of regions is mapped
by default, allowing PAGE_SIZE / 64 available regions for use.
The registered region must fit within a page. On a 4kb page size system,
that allows for 64 wait regions if a full page is used, as the size of
struct io_uring_reg_wait is 64b. The region registered must be aligned
to io_uring_reg_wait in size. It's valid to register less than 64
entries.
In network performance testing with zero-copy, this reduced the time
spent waiting on the TX side from 3.12% to 0.3% and the RX side from 4.4%
to 0.3%.
Wait regions are fixed for the lifetime of the ring - once registered,
they are persistent until the ring is torn down. The regions support
minimum wait timeout as well as the regular waits.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-10-22 13:47:00 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-03-22 14:07:56 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add support for fixed wait regions
Generally applications have 1 or a few waits of waiting, yet they pass
in a struct io_uring_getevents_arg every time. This needs to get copied
and, in turn, the timeout value needs to get copied.
Rather than do this for every invocation, allow the application to
register a fixed set of wait regions that can simply be indexed when
asking the kernel to wait on events.
At ring setup time, the application can register a number of these wait
regions and initialize region/index 0 upfront:
struct io_uring_reg_wait *reg;
reg = io_uring_setup_reg_wait(ring, nr_regions, &ret);
/* set timeout and mark as set, sigmask/sigmask_sz as needed */
reg->ts.tv_sec = 0;
reg->ts.tv_nsec = 100000;
reg->flags = IORING_REG_WAIT_TS;
where nr_regions >= 1 && nr_regions <= PAGE_SIZE / sizeof(*reg). The
above initializes index 0, but 63 other regions can be initialized,
if needed. Now, instead of doing:
struct __kernel_timespec timeout = { .tv_nsec = 100000, };
io_uring_submit_and_wait_timeout(ring, &cqe, nr, &t, NULL);
to wait for events for each submit_and_wait, or just wait, operation, it
can just reference the above region at offset 0 and do:
io_uring_submit_and_wait_reg(ring, &cqe, nr, 0);
to achieve the same goal of waiting 100usec without needing to copy
both struct io_uring_getevents_arg (24b) and struct __kernel_timeout
(16b) for each invocation. Struct io_uring_reg_wait looks as follows:
struct io_uring_reg_wait {
struct __kernel_timespec ts;
__u32 min_wait_usec;
__u32 flags;
__u64 sigmask;
__u32 sigmask_sz;
__u32 pad[3];
__u64 pad2[2];
};
embedding the timeout itself in the region, rather than passing it as
a pointer as well. Note that the signal mask is still passed as a
pointer, both for compatability reasons, but also because there doesn't
seem to be a lot of high frequency waits scenarios that involve setting
and resetting the signal mask for each wait.
The application is free to modify any region before a wait call, or it
can use keep multiple regions with different settings to avoid needing to
modify the same one for wait calls. Up to a page size of regions is mapped
by default, allowing PAGE_SIZE / 64 available regions for use.
The registered region must fit within a page. On a 4kb page size system,
that allows for 64 wait regions if a full page is used, as the size of
struct io_uring_reg_wait is 64b. The region registered must be aligned
to io_uring_reg_wait in size. It's valid to register less than 64
entries.
In network performance testing with zero-copy, this reduced the time
spent waiting on the TX side from 3.12% to 0.3% and the RX side from 4.4%
to 0.3%.
Wait regions are fixed for the lifetime of the ring - once registered,
they are persistent until the ring is torn down. The regions support
minimum wait timeout as well as the regular waits.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-10-22 13:47:00 -06:00
|
|
|
|
|
|
|
|
|
2020-11-03 10:54:37 +08:00
|
|
|
|
2024-10-22 13:41:42 -06:00
|
|
|
|
2020-11-03 10:54:37 +08:00
|
|
|
|
|
|
|
|
|
io_uring: enable toggle of iowait usage when waiting on CQEs
By default, io_uring marks a waiting task as being in iowait, if it's
sleeping waiting on events and there are pending requests. This isn't
necessarily always useful, and may be confusing on non-storage setups
where iowait isn't expected. It can also cause extra power usage, by
preventing the CPU from entering lower sleep states.
This adds a new enter flag, IORING_ENTER_NO_IOWAIT. If set, then
io_uring will not account the sleeping task as being in iowait. If the
kernel supports this feature, then it will be marked by having the
IORING_FEAT_NO_IOWAIT feature flag set.
As the kernel currently does not support separating the iowait
accounting and CPU frequency boosting, the IORING_ENTER_NO_IOWAIT
controls both of these at the same time. In the future, if those do end
up being split, then it'd be possible to control them separately.
However, it seems more likely that the kernel will decouple iowait and
CPU frequency boosting anyway.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-03-14 10:19:01 -06:00
|
|
|
|
|
|
|
|
|
2020-11-03 10:54:37 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-08-15 16:13:32 -06:00
|
|
|
|
2020-11-03 10:54:37 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add support for fixed wait regions
Generally applications have 1 or a few waits of waiting, yet they pass
in a struct io_uring_getevents_arg every time. This needs to get copied
and, in turn, the timeout value needs to get copied.
Rather than do this for every invocation, allow the application to
register a fixed set of wait regions that can simply be indexed when
asking the kernel to wait on events.
At ring setup time, the application can register a number of these wait
regions and initialize region/index 0 upfront:
struct io_uring_reg_wait *reg;
reg = io_uring_setup_reg_wait(ring, nr_regions, &ret);
/* set timeout and mark as set, sigmask/sigmask_sz as needed */
reg->ts.tv_sec = 0;
reg->ts.tv_nsec = 100000;
reg->flags = IORING_REG_WAIT_TS;
where nr_regions >= 1 && nr_regions <= PAGE_SIZE / sizeof(*reg). The
above initializes index 0, but 63 other regions can be initialized,
if needed. Now, instead of doing:
struct __kernel_timespec timeout = { .tv_nsec = 100000, };
io_uring_submit_and_wait_timeout(ring, &cqe, nr, &t, NULL);
to wait for events for each submit_and_wait, or just wait, operation, it
can just reference the above region at offset 0 and do:
io_uring_submit_and_wait_reg(ring, &cqe, nr, 0);
to achieve the same goal of waiting 100usec without needing to copy
both struct io_uring_getevents_arg (24b) and struct __kernel_timeout
(16b) for each invocation. Struct io_uring_reg_wait looks as follows:
struct io_uring_reg_wait {
struct __kernel_timespec ts;
__u32 min_wait_usec;
__u32 flags;
__u64 sigmask;
__u32 sigmask_sz;
__u32 pad[3];
__u64 pad2[2];
};
embedding the timeout itself in the region, rather than passing it as
a pointer as well. Note that the signal mask is still passed as a
pointer, both for compatability reasons, but also because there doesn't
seem to be a lot of high frequency waits scenarios that involve setting
and resetting the signal mask for each wait.
The application is free to modify any region before a wait call, or it
can use keep multiple regions with different settings to avoid needing to
modify the same one for wait calls. Up to a page size of regions is mapped
by default, allowing PAGE_SIZE / 64 available regions for use.
The registered region must fit within a page. On a 4kb page size system,
that allows for 64 wait regions if a full page is used, as the size of
struct io_uring_reg_wait is 64b. The region registered must be aligned
to io_uring_reg_wait in size. It's valid to register less than 64
entries.
In network performance testing with zero-copy, this reduced the time
spent waiting on the TX side from 3.12% to 0.3% and the RX side from 4.4%
to 0.3%.
Wait regions are fixed for the lifetime of the ring - once registered,
they are persistent until the ring is torn down. The regions support
minimum wait timeout as well as the regular waits.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-10-22 13:47:00 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-11-03 10:54:37 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-08-15 16:13:32 -06:00
|
|
|
|
2020-11-03 10:54:37 +08:00
|
|
|
|
2024-10-22 13:41:42 -06:00
|
|
|
|
|
|
|
|
|
2020-11-03 10:54:37 +08:00
|
|
|
|
2024-10-22 13:41:42 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-01-04 10:46:30 -07:00
|
|
|
|
2024-08-15 16:13:32 -06:00
|
|
|
|
|
|
|
|
|
2024-10-22 13:37:00 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-11-03 10:54:37 +08:00
|
|
|
|
2024-10-22 13:41:42 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-11-03 10:54:37 +08:00
|
|
|
|
|
|
|
|
|
2026-04-08 11:56:02 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-04-20 14:06:00 +00:00
|
|
|
|
|
|
|
|
|
2026-04-08 11:56:02 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2020-11-03 10:54:37 +08:00
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2023-11-28 10:29:58 -07:00
|
|
|
|
2021-03-19 17:22:30 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2025-09-08 00:02:59 +01:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2026-04-08 11:56:02 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-11-28 10:29:58 -07:00
|
|
|
|
2020-08-27 16:58:31 +02:00
|
|
|
|
2026-01-05 14:05:40 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-08-27 16:58:31 +02:00
|
|
|
|
|
|
|
|
|
io_uring: introduce callback driven main loop
The io_uring_enter() has a fixed order of execution: it submits
requests, waits for completions, and returns to the user. Allow to
optionally replace it with a custom loop driven by a callback called
loop_step. The basic requirements to the callback is that it should be
able to submit requests, wait for completions, parse them and repeat.
Most of the communication including parameter passing can be implemented
via shared memory.
The callback should return IOU_LOOP_CONTINUE to continue execution or
IOU_LOOP_STOP to return to the user space. Note that the kernel may
decide to prematurely terminate it as well, e.g. in case the process was
signalled or killed.
The hook takes a structure with parameters. It can be used to ask the
kernel to wait for CQEs by setting cq_wait_idx to the CQE index it wants
to wait for. Spurious wake ups are possible and even likely, the callback
is expected to handle it. There will be more parameters in the future
like timeout.
It can be used with kernel callbacks, for example, as a slow path
deprecation mechanism overwiting SQEs and emulating the wanted
behaviour, however it's more useful together with BPF programs
implemented in following patches.
Note that keeping it separately from the normal io_uring wait loop
makes things much simpler and cleaner. It keeps it in one place instead
of spreading a bunch of checks in different places including disabling
the submission path. It holds the lock by default, which is a better fit
for BPF synchronisation and the loop execution model. It nicely avoids
existing quirks like forced wake ups on timeout request completion. And
it should be easier to implement new features.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://patch.msgid.link/a2d369aa1c9dd23ad7edac9220cffc563abcaed6.1772109579.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2026-02-26 12:48:38 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add submission polling
This enables an application to do IO, without ever entering the kernel.
By using the SQ ring to fill in new sqes and watching for completions
on the CQ ring, we can submit and reap IOs without doing a single system
call. The kernel side thread will poll for new submissions, and in case
of HIPRI/polled IO, it'll also poll for completions.
By default, we allow 1 second of active spinning. This can by changed
by passing in a different grace period at io_uring_register(2) time.
If the thread exceeds this idle time without having any work to do, it
will set:
sq_ring->flags |= IORING_SQ_NEED_WAKEUP.
The application will have to call io_uring_enter() to start things back
up again. If IO is kept busy, that will never be needed. Basically an
application that has this feature enabled will guard it's
io_uring_enter(2) call with:
read_barrier();
if (*sq_ring->flags & IORING_SQ_NEED_WAKEUP)
io_uring_enter(fd, 0, 0, IORING_ENTER_SQ_WAKEUP);
instead of calling it unconditionally.
It's mandatory to use fixed files with this feature. Failure to do so
will result in the application getting an -EBADF CQ entry when
submitting IO.
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-10 11:22:30 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-09-12 14:19:16 -06:00
|
|
|
|
io_uring: add submission polling
This enables an application to do IO, without ever entering the kernel.
By using the SQ ring to fill in new sqes and watching for completions
on the CQ ring, we can submit and reap IOs without doing a single system
call. The kernel side thread will poll for new submissions, and in case
of HIPRI/polled IO, it'll also poll for completions.
By default, we allow 1 second of active spinning. This can by changed
by passing in a different grace period at io_uring_register(2) time.
If the thread exceeds this idle time without having any work to do, it
will set:
sq_ring->flags |= IORING_SQ_NEED_WAKEUP.
The application will have to call io_uring_enter() to start things back
up again. If IO is kept busy, that will never be needed. Basically an
application that has this feature enabled will guard it's
io_uring_enter(2) call with:
read_barrier();
if (*sq_ring->flags & IORING_SQ_NEED_WAKEUP)
io_uring_enter(fd, 0, 0, IORING_ENTER_SQ_WAKEUP);
instead of calling it unconditionally.
It's mandatory to use fixed files with this feature. Failure to do so
will result in the application getting an -EBADF CQ entry when
submitting IO.
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-10 11:22:30 -07:00
|
|
|
|
2021-08-14 09:04:40 -06:00
|
|
|
|
|
|
|
|
|
2021-03-07 11:54:29 +01:00
|
|
|
|
2021-08-14 09:04:40 -06:00
|
|
|
|
io_uring: add submission polling
This enables an application to do IO, without ever entering the kernel.
By using the SQ ring to fill in new sqes and watching for completions
on the CQ ring, we can submit and reap IOs without doing a single system
call. The kernel side thread will poll for new submissions, and in case
of HIPRI/polled IO, it'll also poll for completions.
By default, we allow 1 second of active spinning. This can by changed
by passing in a different grace period at io_uring_register(2) time.
If the thread exceeds this idle time without having any work to do, it
will set:
sq_ring->flags |= IORING_SQ_NEED_WAKEUP.
The application will have to call io_uring_enter() to start things back
up again. If IO is kept busy, that will never be needed. Basically an
application that has this feature enabled will guard it's
io_uring_enter(2) call with:
read_barrier();
if (*sq_ring->flags & IORING_SQ_NEED_WAKEUP)
io_uring_enter(fd, 0, 0, IORING_ENTER_SQ_WAKEUP);
instead of calling it unconditionally.
It's mandatory to use fixed files with this feature. Failure to do so
will result in the application getting an -EBADF CQ entry when
submitting IO.
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-10 11:22:30 -07:00
|
|
|
|
2020-09-02 13:52:19 -06:00
|
|
|
|
2023-01-15 15:15:19 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-04-21 02:13:42 -07:00
|
|
|
|
2019-09-12 14:19:16 -06:00
|
|
|
|
2021-06-14 02:36:15 +01:00
|
|
|
|
2020-09-13 13:09:39 -06:00
|
|
|
|
|
|
|
|
|
2019-12-18 19:53:45 +03:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2022-04-21 02:13:42 -07:00
|
|
|
|
|
|
|
|
|
2022-03-22 14:07:58 +00:00
|
|
|
|
2019-12-18 19:53:45 +03:00
|
|
|
|
2022-03-22 14:07:58 +00:00
|
|
|
|
2022-10-06 21:42:33 +01:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2022-10-06 21:42:33 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-01-31 10:50:08 -07:00
|
|
|
|
2022-10-06 21:42:33 +01:00
|
|
|
|
2022-03-22 14:07:58 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2022-08-30 05:50:10 -07:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2022-04-21 02:13:42 -07:00
|
|
|
|
2022-08-30 05:50:10 -07:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2022-03-22 14:07:58 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add support for fixed wait regions
Generally applications have 1 or a few waits of waiting, yet they pass
in a struct io_uring_getevents_arg every time. This needs to get copied
and, in turn, the timeout value needs to get copied.
Rather than do this for every invocation, allow the application to
register a fixed set of wait regions that can simply be indexed when
asking the kernel to wait on events.
At ring setup time, the application can register a number of these wait
regions and initialize region/index 0 upfront:
struct io_uring_reg_wait *reg;
reg = io_uring_setup_reg_wait(ring, nr_regions, &ret);
/* set timeout and mark as set, sigmask/sigmask_sz as needed */
reg->ts.tv_sec = 0;
reg->ts.tv_nsec = 100000;
reg->flags = IORING_REG_WAIT_TS;
where nr_regions >= 1 && nr_regions <= PAGE_SIZE / sizeof(*reg). The
above initializes index 0, but 63 other regions can be initialized,
if needed. Now, instead of doing:
struct __kernel_timespec timeout = { .tv_nsec = 100000, };
io_uring_submit_and_wait_timeout(ring, &cqe, nr, &t, NULL);
to wait for events for each submit_and_wait, or just wait, operation, it
can just reference the above region at offset 0 and do:
io_uring_submit_and_wait_reg(ring, &cqe, nr, 0);
to achieve the same goal of waiting 100usec without needing to copy
both struct io_uring_getevents_arg (24b) and struct __kernel_timeout
(16b) for each invocation. Struct io_uring_reg_wait looks as follows:
struct io_uring_reg_wait {
struct __kernel_timespec ts;
__u32 min_wait_usec;
__u32 flags;
__u64 sigmask;
__u32 sigmask_sz;
__u32 pad[3];
__u64 pad2[2];
};
embedding the timeout itself in the region, rather than passing it as
a pointer as well. Note that the signal mask is still passed as a
pointer, both for compatability reasons, but also because there doesn't
seem to be a lot of high frequency waits scenarios that involve setting
and resetting the signal mask for each wait.
The application is free to modify any region before a wait call, or it
can use keep multiple regions with different settings to avoid needing to
modify the same one for wait calls. Up to a page size of regions is mapped
by default, allowing PAGE_SIZE / 64 available regions for use.
The registered region must fit within a page. On a 4kb page size system,
that allows for 64 wait regions if a full page is used, as the size of
struct io_uring_reg_wait is 64b. The region registered must be aligned
to io_uring_reg_wait in size. It's valid to register less than 64
entries.
In network performance testing with zero-copy, this reduced the time
spent waiting on the TX side from 3.12% to 0.3% and the RX side from 4.4%
to 0.3%.
Wait regions are fixed for the lifetime of the ring - once registered,
they are persistent until the ring is torn down. The regions support
minimum wait timeout as well as the regular waits.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-10-22 13:47:00 -06:00
|
|
|
|
2025-03-24 15:32:36 +00:00
|
|
|
|
2022-04-21 02:13:42 -07:00
|
|
|
|
2022-03-22 14:07:58 +00:00
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
2024-08-15 16:13:32 -06:00
|
|
|
|
2022-03-22 14:07:56 +00:00
|
|
|
|
io_uring: add support for fixed wait regions
Generally applications have 1 or a few waits of waiting, yet they pass
in a struct io_uring_getevents_arg every time. This needs to get copied
and, in turn, the timeout value needs to get copied.
Rather than do this for every invocation, allow the application to
register a fixed set of wait regions that can simply be indexed when
asking the kernel to wait on events.
At ring setup time, the application can register a number of these wait
regions and initialize region/index 0 upfront:
struct io_uring_reg_wait *reg;
reg = io_uring_setup_reg_wait(ring, nr_regions, &ret);
/* set timeout and mark as set, sigmask/sigmask_sz as needed */
reg->ts.tv_sec = 0;
reg->ts.tv_nsec = 100000;
reg->flags = IORING_REG_WAIT_TS;
where nr_regions >= 1 && nr_regions <= PAGE_SIZE / sizeof(*reg). The
above initializes index 0, but 63 other regions can be initialized,
if needed. Now, instead of doing:
struct __kernel_timespec timeout = { .tv_nsec = 100000, };
io_uring_submit_and_wait_timeout(ring, &cqe, nr, &t, NULL);
to wait for events for each submit_and_wait, or just wait, operation, it
can just reference the above region at offset 0 and do:
io_uring_submit_and_wait_reg(ring, &cqe, nr, 0);
to achieve the same goal of waiting 100usec without needing to copy
both struct io_uring_getevents_arg (24b) and struct __kernel_timeout
(16b) for each invocation. Struct io_uring_reg_wait looks as follows:
struct io_uring_reg_wait {
struct __kernel_timespec ts;
__u32 min_wait_usec;
__u32 flags;
__u64 sigmask;
__u32 sigmask_sz;
__u32 pad[3];
__u64 pad2[2];
};
embedding the timeout itself in the region, rather than passing it as
a pointer as well. Note that the signal mask is still passed as a
pointer, both for compatability reasons, but also because there doesn't
seem to be a lot of high frequency waits scenarios that involve setting
and resetting the signal mask for each wait.
The application is free to modify any region before a wait call, or it
can use keep multiple regions with different settings to avoid needing to
modify the same one for wait calls. Up to a page size of regions is mapped
by default, allowing PAGE_SIZE / 64 available regions for use.
The registered region must fit within a page. On a 4kb page size system,
that allows for 64 wait regions if a full page is used, as the size of
struct io_uring_reg_wait is 64b. The region registered must be aligned
to io_uring_reg_wait in size. It's valid to register less than 64
entries.
In network performance testing with zero-copy, this reduced the time
spent waiting on the TX side from 3.12% to 0.3% and the RX side from 4.4%
to 0.3%.
Wait regions are fixed for the lifetime of the ring - once registered,
they are persistent until the ring is torn down. The regions support
minimum wait timeout as well as the regular waits.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2024-10-22 13:47:00 -06:00
|
|
|
|
2025-03-24 15:32:36 +00:00
|
|
|
|
2024-08-07 15:18:13 +01:00
|
|
|
|
2024-08-15 16:13:32 -06:00
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
2022-04-21 02:13:42 -07:00
|
|
|
|
2022-04-21 02:13:44 -07:00
|
|
|
|
2022-04-21 02:13:42 -07:00
|
|
|
|
|
|
|
|
|
2022-04-21 02:13:44 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-01-09 08:59:42 -07:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2019-12-18 19:53:45 +03:00
|
|
|
|
2023-11-28 10:29:58 -07:00
|
|
|
|
|
|
|
|
|
2022-04-21 02:13:42 -07:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-03-27 14:59:09 -06:00
|
|
|
|
2019-11-28 12:53:22 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2020-02-26 18:38:32 +01:00
|
|
|
|
2020-01-30 08:25:34 -07:00
|
|
|
|
2020-02-26 18:38:32 +01:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2022-05-25 11:48:35 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-10-04 20:02:54 +01:00
|
|
|
|
2025-11-12 12:45:58 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2025-11-12 12:45:58 +00:00
|
|
|
|
|
|
|
|
|
2024-11-29 13:34:34 +00:00
|
|
|
|
2019-08-26 17:23:46 +00:00
|
|
|
|
2024-11-29 13:34:34 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2020-08-05 12:58:23 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-11-29 13:34:35 +00:00
|
|
|
|
2025-11-12 12:45:57 +00:00
|
|
|
|
2024-11-29 13:34:35 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-09 14:21:37 -06:00
|
|
|
|
2023-08-24 23:53:32 +01:00
|
|
|
|
2025-11-12 12:45:57 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2024-11-29 13:34:34 +00:00
|
|
|
|
2025-11-12 12:45:57 +00:00
|
|
|
|
2024-11-29 13:34:34 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-11-20 09:26:29 -07:00
|
|
|
|
2024-11-29 13:34:34 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-06 12:58:19 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-04-29 01:40:30 +09:00
|
|
|
|
2020-12-21 18:34:05 +00:00
|
|
|
|
2023-04-29 01:40:30 +09:00
|
|
|
|
2020-12-21 18:34:05 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-12-19 12:36:34 -07:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2020-12-21 18:34:05 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
fs: Rename anon_inode_getfile_secure() and anon_inode_getfd_secure()
The call to the inode_init_security_anon() LSM hook is not the sole
reason to use anon_inode_getfile_secure() or anon_inode_getfd_secure().
For example, the functions also allow one to create a file with non-zero
size, without needing a full-blown filesystem. In this case, you don't
need a "secure" version, just unique inodes; the current name of the
functions is confusing and does not explain well the difference with
the more "standard" anon_inode_getfile() and anon_inode_getfd().
Of course, there is another side of the coin; neither io_uring nor
userfaultfd strictly speaking need distinct inodes, and it is not
that clear anymore that anon_inode_create_get{file,fd}() allow the LSM
to intercept and block the inode's creation. If one was so inclined,
anon_inode_getfile_secure() and anon_inode_getfd_secure() could be kept,
using the shared inode or a new one depending on CONFIG_SECURITY.
However, this is probably overkill, and potentially a cause of bugs in
different configurations. Therefore, just add a comment to io_uring
and userfaultfd explaining the choice of the function.
While at it, remove the export for what is now anon_inode_create_getfd().
There is no in-tree module that uses it, and the old name is gone anyway.
If anybody actually needs the symbol, they can ask or they can just use
anon_inode_create_getfile(), which will be exported very soon for use
in KVM.
Suggested-by: Christian Brauner <brauner@kernel.org>
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2023-11-03 06:47:51 -04:00
|
|
|
|
Generic:
- Use memdup_array_user() to harden against overflow.
- Unconditionally advertise KVM_CAP_DEVICE_CTRL for all architectures.
- Clean up Kconfigs that all KVM architectures were selecting
- New functionality around "guest_memfd", a new userspace API that
creates an anonymous file and returns a file descriptor that refers
to it. guest_memfd files are bound to their owning virtual machine,
cannot be mapped, read, or written by userspace, and cannot be resized.
guest_memfd files do however support PUNCH_HOLE, which can be used to
switch a memory area between guest_memfd and regular anonymous memory.
- New ioctl KVM_SET_MEMORY_ATTRIBUTES allowing userspace to specify
per-page attributes for a given page of guest memory; right now the
only attribute is whether the guest expects to access memory via
guest_memfd or not, which in Confidential SVMs backed by SEV-SNP,
TDX or ARM64 pKVM is checked by firmware or hypervisor that guarantees
confidentiality (AMD PSP, Intel TDX module, or EL2 in the case of pKVM).
x86:
- Support for "software-protected VMs" that can use the new guest_memfd
and page attributes infrastructure. This is mostly useful for testing,
since there is no pKVM-like infrastructure to provide a meaningfully
reduced TCB.
- Fix a relatively benign off-by-one error when splitting huge pages during
CLEAR_DIRTY_LOG.
- Fix a bug where KVM could incorrectly test-and-clear dirty bits in non-leaf
TDP MMU SPTEs if a racing thread replaces a huge SPTE with a non-huge SPTE.
- Use more generic lockdep assertions in paths that don't actually care
about whether the caller is a reader or a writer.
- let Xen guests opt out of having PV clock reported as "based on a stable TSC",
because some of them don't expect the "TSC stable" bit (added to the pvclock
ABI by KVM, but never set by Xen) to be set.
- Revert a bogus, made-up nested SVM consistency check for TLB_CONTROL.
- Advertise flush-by-ASID support for nSVM unconditionally, as KVM always
flushes on nested transitions, i.e. always satisfies flush requests. This
allows running bleeding edge versions of VMware Workstation on top of KVM.
- Sanity check that the CPU supports flush-by-ASID when enabling SEV support.
- On AMD machines with vNMI, always rely on hardware instead of intercepting
IRET in some cases to detect unmasking of NMIs
- Support for virtualizing Linear Address Masking (LAM)
- Fix a variety of vPMU bugs where KVM fail to stop/reset counters and other state
prior to refreshing the vPMU model.
- Fix a double-overflow PMU bug by tracking emulated counter events using a
dedicated field instead of snapshotting the "previous" counter. If the
hardware PMC count triggers overflow that is recognized in the same VM-Exit
that KVM manually bumps an event count, KVM would pend PMIs for both the
hardware-triggered overflow and for KVM-triggered overflow.
- Turn off KVM_WERROR by default for all configs so that it's not
inadvertantly enabled by non-KVM developers, which can be problematic for
subsystems that require no regressions for W=1 builds.
- Advertise all of the host-supported CPUID bits that enumerate IA32_SPEC_CTRL
"features".
- Don't force a masterclock update when a vCPU synchronizes to the current TSC
generation, as updating the masterclock can cause kvmclock's time to "jump"
unexpectedly, e.g. when userspace hotplugs a pre-created vCPU.
- Use RIP-relative address to read kvm_rebooting in the VM-Enter fault paths,
partly as a super minor optimization, but mostly to make KVM play nice with
position independent executable builds.
- Guard KVM-on-HyperV's range-based TLB flush hooks with an #ifdef on
CONFIG_HYPERV as a minor optimization, and to self-document the code.
- Add CONFIG_KVM_HYPERV to allow disabling KVM support for HyperV "emulation"
at build time.
ARM64:
- LPA2 support, adding 52bit IPA/PA capability for 4kB and 16kB
base granule sizes. Branch shared with the arm64 tree.
- Large Fine-Grained Trap rework, bringing some sanity to the
feature, although there is more to come. This comes with
a prefix branch shared with the arm64 tree.
- Some additional Nested Virtualization groundwork, mostly
introducing the NV2 VNCR support and retargetting the NV
support to that version of the architecture.
- A small set of vgic fixes and associated cleanups.
Loongarch:
- Optimization for memslot hugepage checking
- Cleanup and fix some HW/SW timer issues
- Add LSX/LASX (128bit/256bit SIMD) support
RISC-V:
- KVM_GET_REG_LIST improvement for vector registers
- Generate ISA extension reg_list using macros in get-reg-list selftest
- Support for reporting steal time along with selftest
s390:
- Bugfixes
Selftests:
- Fix an annoying goof where the NX hugepage test prints out garbage
instead of the magic token needed to run the test.
- Fix build errors when a header is delete/moved due to a missing flag
in the Makefile.
- Detect if KVM bugged/killed a selftest's VM and print out a helpful
message instead of complaining that a random ioctl() failed.
- Annotate the guest printf/assert helpers with __printf(), and fix the
various bugs that were lurking due to lack of said annotation.
There are two non-KVM patches buried in the middle of guest_memfd support:
fs: Rename anon_inode_getfile_secure() and anon_inode_getfd_secure()
mm: Add AS_UNMOVABLE to mark mapping as completely unmovable
The first is small and mostly suggested-by Christian Brauner; the second
a bit less so but it was written by an mm person (Vlastimil Babka).
-----BEGIN PGP SIGNATURE-----
iQFIBAABCAAyFiEE8TM4V0tmI4mGbHaCv/vSX3jHroMFAmWcMWkUHHBib256aW5p
QHJlZGhhdC5jb20ACgkQv/vSX3jHroO15gf/WLmmg3SET6Uzw9iEq2xo28831ZA+
6kpILfIDGKozV5safDmMvcInlc/PTnqOFrsKyyN4kDZ+rIJiafJdg/loE0kPXBML
wdR+2ix5kYI1FucCDaGTahskBDz8Lb/xTpwGg9BFLYFNmuUeHc74o6GoNvr1uliE
4kLZL2K6w0cSMPybUD+HqGaET80ZqPwecv+s1JL+Ia0kYZJONJifoHnvOUJ7DpEi
rgudVdgzt3EPjG0y1z6MjvDBXTCOLDjXajErlYuZD3Ej8N8s59Dh2TxOiDNTLdP4
a4zjRvDmgyr6H6sz+upvwc7f4M4p+DBvf+TkWF54mbeObHUYliStqURIoA==
=66Ws
-----END PGP SIGNATURE-----
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Pull kvm updates from Paolo Bonzini:
"Generic:
- Use memdup_array_user() to harden against overflow.
- Unconditionally advertise KVM_CAP_DEVICE_CTRL for all
architectures.
- Clean up Kconfigs that all KVM architectures were selecting
- New functionality around "guest_memfd", a new userspace API that
creates an anonymous file and returns a file descriptor that refers
to it. guest_memfd files are bound to their owning virtual machine,
cannot be mapped, read, or written by userspace, and cannot be
resized. guest_memfd files do however support PUNCH_HOLE, which can
be used to switch a memory area between guest_memfd and regular
anonymous memory.
- New ioctl KVM_SET_MEMORY_ATTRIBUTES allowing userspace to specify
per-page attributes for a given page of guest memory; right now the
only attribute is whether the guest expects to access memory via
guest_memfd or not, which in Confidential SVMs backed by SEV-SNP,
TDX or ARM64 pKVM is checked by firmware or hypervisor that
guarantees confidentiality (AMD PSP, Intel TDX module, or EL2 in
the case of pKVM).
x86:
- Support for "software-protected VMs" that can use the new
guest_memfd and page attributes infrastructure. This is mostly
useful for testing, since there is no pKVM-like infrastructure to
provide a meaningfully reduced TCB.
- Fix a relatively benign off-by-one error when splitting huge pages
during CLEAR_DIRTY_LOG.
- Fix a bug where KVM could incorrectly test-and-clear dirty bits in
non-leaf TDP MMU SPTEs if a racing thread replaces a huge SPTE with
a non-huge SPTE.
- Use more generic lockdep assertions in paths that don't actually
care about whether the caller is a reader or a writer.
- let Xen guests opt out of having PV clock reported as "based on a
stable TSC", because some of them don't expect the "TSC stable" bit
(added to the pvclock ABI by KVM, but never set by Xen) to be set.
- Revert a bogus, made-up nested SVM consistency check for
TLB_CONTROL.
- Advertise flush-by-ASID support for nSVM unconditionally, as KVM
always flushes on nested transitions, i.e. always satisfies flush
requests. This allows running bleeding edge versions of VMware
Workstation on top of KVM.
- Sanity check that the CPU supports flush-by-ASID when enabling SEV
support.
- On AMD machines with vNMI, always rely on hardware instead of
intercepting IRET in some cases to detect unmasking of NMIs
- Support for virtualizing Linear Address Masking (LAM)
- Fix a variety of vPMU bugs where KVM fail to stop/reset counters
and other state prior to refreshing the vPMU model.
- Fix a double-overflow PMU bug by tracking emulated counter events
using a dedicated field instead of snapshotting the "previous"
counter. If the hardware PMC count triggers overflow that is
recognized in the same VM-Exit that KVM manually bumps an event
count, KVM would pend PMIs for both the hardware-triggered overflow
and for KVM-triggered overflow.
- Turn off KVM_WERROR by default for all configs so that it's not
inadvertantly enabled by non-KVM developers, which can be
problematic for subsystems that require no regressions for W=1
builds.
- Advertise all of the host-supported CPUID bits that enumerate
IA32_SPEC_CTRL "features".
- Don't force a masterclock update when a vCPU synchronizes to the
current TSC generation, as updating the masterclock can cause
kvmclock's time to "jump" unexpectedly, e.g. when userspace
hotplugs a pre-created vCPU.
- Use RIP-relative address to read kvm_rebooting in the VM-Enter
fault paths, partly as a super minor optimization, but mostly to
make KVM play nice with position independent executable builds.
- Guard KVM-on-HyperV's range-based TLB flush hooks with an #ifdef on
CONFIG_HYPERV as a minor optimization, and to self-document the
code.
- Add CONFIG_KVM_HYPERV to allow disabling KVM support for HyperV
"emulation" at build time.
ARM64:
- LPA2 support, adding 52bit IPA/PA capability for 4kB and 16kB base
granule sizes. Branch shared with the arm64 tree.
- Large Fine-Grained Trap rework, bringing some sanity to the
feature, although there is more to come. This comes with a prefix
branch shared with the arm64 tree.
- Some additional Nested Virtualization groundwork, mostly
introducing the NV2 VNCR support and retargetting the NV support to
that version of the architecture.
- A small set of vgic fixes and associated cleanups.
Loongarch:
- Optimization for memslot hugepage checking
- Cleanup and fix some HW/SW timer issues
- Add LSX/LASX (128bit/256bit SIMD) support
RISC-V:
- KVM_GET_REG_LIST improvement for vector registers
- Generate ISA extension reg_list using macros in get-reg-list
selftest
- Support for reporting steal time along with selftest
s390:
- Bugfixes
Selftests:
- Fix an annoying goof where the NX hugepage test prints out garbage
instead of the magic token needed to run the test.
- Fix build errors when a header is delete/moved due to a missing
flag in the Makefile.
- Detect if KVM bugged/killed a selftest's VM and print out a helpful
message instead of complaining that a random ioctl() failed.
- Annotate the guest printf/assert helpers with __printf(), and fix
the various bugs that were lurking due to lack of said annotation"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (185 commits)
x86/kvm: Do not try to disable kvmclock if it was not enabled
KVM: x86: add missing "depends on KVM"
KVM: fix direction of dependency on MMU notifiers
KVM: introduce CONFIG_KVM_COMMON
KVM: arm64: Add missing memory barriers when switching to pKVM's hyp pgd
KVM: arm64: vgic-its: Avoid potential UAF in LPI translation cache
RISC-V: KVM: selftests: Add get-reg-list test for STA registers
RISC-V: KVM: selftests: Add steal_time test support
RISC-V: KVM: selftests: Add guest_sbi_probe_extension
RISC-V: KVM: selftests: Move sbi_ecall to processor.c
RISC-V: KVM: Implement SBI STA extension
RISC-V: KVM: Add support for SBI STA registers
RISC-V: KVM: Add support for SBI extension registers
RISC-V: KVM: Add SBI STA info to vcpu_arch
RISC-V: KVM: Add steal-update vcpu request
RISC-V: KVM: Add SBI STA extension skeleton
RISC-V: paravirt: Implement steal-time support
RISC-V: Add SBI STA extension definitions
RISC-V: paravirt: Add skeleton for pv-time support
RISC-V: KVM: Fix indentation in kvm_riscv_vcpu_set_reg_csr()
...
2024-01-17 13:03:37 -08:00
|
|
|
|
2021-02-01 19:33:52 -05:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2025-01-31 17:31:03 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-10 13:04:52 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-01-20 20:47:40 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-01-23 05:09:08 -07:00
|
|
|
|
2026-01-20 20:47:40 +00:00
|
|
|
|
|
|
|
|
|
2025-01-31 17:31:03 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add support for IORING_SETUP_SQE_MIXED
Normal rings support 64b SQEs for posting submissions, while certain
features require the ring to be configured with IORING_SETUP_SQE128, as
they need to convey more information per submission. This, in turn,
makes ALL the SQEs be 128b in size. This is somewhat wasteful and
inefficient, particularly when only certain SQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed SQE sizes, using
IORING_SETUP_SQE_MIXED. When setup in this mode, SQEs posted to the ring
may be either 64b or 128b in size. If a SQE is 128b in size, then opcode
will be set to a variante to indicate that this is the case. Any other
non-128b opcode will assume the SQ's default size.
SQEs on these types of mixed rings may also utilize NOP with skip
success set. This can happen if the ring is one (small) SQE entry away
from wrapping, and an attempt is made to get a 128b SQE. As SQEs must be
contiguous in the SQ ring, a 128b SQE cannot wrap the ring. For this
case, a single NOP SQE should be inserted with the SKIP_SUCCESS flag
set. The kernel will process this as a normal NOP and without posting a
CQE.
Signed-off-by: Keith Busch <kbusch@kernel.org>
[axboe: {} style fix and assign sqe before opcode read]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-10-16 11:09:38 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
io_uring: add support for IORING_SETUP_CQE_MIXED
Normal rings support 16b CQEs for posting completions, while certain
features require the ring to be configured with IORING_SETUP_CQE32, as
they need to convey more information per completion. This, in turn,
makes ALL the CQEs be 32b in size. This is somewhat wasteful and
inefficient, particularly when only certain CQEs need to be of the
bigger variant.
This adds support for setting up a ring with mixed CQE sizes, using
IORING_SETUP_CQE_MIXED. When setup in this mode, CQEs posted to the ring
may be either 16b or 32b in size. If a CQE is 32b in size, then
IORING_CQE_F_32 is set in the CQE flags to indicate that this is the
case. If this flag isn't set, the CQE is the normal 16b variant.
CQEs on these types of mixed rings may also have IORING_CQE_F_SKIP set.
This can happen if the ring is one (small) CQE entry away from wrapping,
and an attempt is made to post a 32b CQE. As CQEs must be contigious in
the CQ ring, a 32b CQE cannot wrap the ring. For this case, a single
dummy CQE is posted with the SKIP flag set. The application should
simply ignore those.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2025-08-07 14:14:41 -06:00
|
|
|
|
2025-01-31 17:31:03 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:56 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2025-11-10 13:04:50 +00:00
|
|
|
|
|
|
|
|
|
2019-12-28 15:39:54 -07:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2019-12-28 15:39:54 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-10-04 12:10:03 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2019-10-04 12:10:03 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-11-24 15:03:03 +08:00
|
|
|
|
2019-10-04 12:10:03 -06:00
|
|
|
|
2019-12-28 15:39:54 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-11-24 15:03:03 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-10-04 12:10:03 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2024-10-21 13:32:19 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:56 +00:00
|
|
|
|
2024-10-21 13:32:19 -06:00
|
|
|
|
2025-11-12 12:45:56 +00:00
|
|
|
|
2024-10-21 13:32:19 -06:00
|
|
|
|
|
|
|
|
|
2025-01-31 17:31:03 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-10 13:04:50 +00:00
|
|
|
|
2025-11-12 12:45:56 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:58 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-10-21 13:32:19 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:58 +00:00
|
|
|
|
|
|
|
|
|
2024-10-21 13:32:19 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-01-08 10:18:31 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2026-01-08 10:18:31 -07:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2026-01-08 10:18:31 -07:00
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:56 +00:00
|
|
|
|
2024-10-21 13:32:19 -06:00
|
|
|
|
2025-11-12 12:45:56 +00:00
|
|
|
|
2024-10-21 13:32:19 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:56 +00:00
|
|
|
|
2025-01-31 17:31:03 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2021-02-21 16:19:37 -07:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2022-03-22 14:07:57 +00:00
|
|
|
|
2024-08-07 15:18:14 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-10-18 17:07:59 +01:00
|
|
|
|
2026-02-15 22:06:52 +00:00
|
|
|
|
2024-10-18 17:07:59 +01:00
|
|
|
|
2022-12-07 03:53:30 +00:00
|
|
|
|
2026-02-06 14:58:04 -07:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2022-12-07 03:53:30 +00:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-24 23:53:29 +01:00
|
|
|
|
2023-01-09 14:46:09 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
|
|
|
|
|
2023-01-09 14:46:09 +00:00
|
|
|
|
2022-03-22 14:07:57 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
2022-03-22 14:07:57 +00:00
|
|
|
|
2026-03-14 08:41:04 -06:00
|
|
|
|
|
|
|
|
|
2023-07-18 13:56:07 +02:00
|
|
|
|
2021-02-21 16:19:37 -07:00
|
|
|
|
2020-09-14 10:45:53 -06:00
|
|
|
|
2022-04-25 19:49:02 -06:00
|
|
|
|
2022-04-25 19:49:03 -06:00
|
|
|
|
|
|
|
|
|
2022-04-25 19:49:02 -06:00
|
|
|
|
2025-01-31 17:31:03 +00:00
|
|
|
|
2022-04-25 19:49:02 -06:00
|
|
|
|
2025-01-31 17:31:03 +00:00
|
|
|
|
2022-04-25 19:49:02 -06:00
|
|
|
|
2022-08-30 05:50:10 -07:00
|
|
|
|
2026-01-08 10:18:31 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-09-14 10:45:53 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-08-25 07:58:00 -06:00
|
|
|
|
2020-09-14 10:45:53 -06:00
|
|
|
|
2020-08-25 07:58:00 -06:00
|
|
|
|
2025-11-12 12:45:58 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-08-27 16:58:31 +02:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2023-04-11 12:06:07 +01:00
|
|
|
|
2025-09-08 00:02:59 +01:00
|
|
|
|
2020-05-05 16:28:53 +08:00
|
|
|
|
2025-11-12 12:45:56 +00:00
|
|
|
|
2020-05-05 16:28:53 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-07-30 13:43:53 -06:00
|
|
|
|
2022-09-26 10:09:25 -07:00
|
|
|
|
2026-01-05 14:05:41 -07:00
|
|
|
|
2025-09-04 10:12:22 -06:00
|
|
|
|
2022-09-26 10:09:25 -07:00
|
|
|
|
2020-12-21 18:34:05 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-04-29 01:40:30 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-10-28 09:15:33 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-04-29 01:40:30 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-10-28 09:15:33 -06:00
|
|
|
|
io_uring: add set of tracing events
To trace io_uring activity one can get an information from workqueue and
io trace events, but looks like some parts could be hard to identify via
this approach. Making what happens inside io_uring more transparent is
important to be able to reason about many aspects of it, hence introduce
the set of tracing events.
All such events could be roughly divided into two categories:
* those, that are helping to understand correctness (from both kernel
and an application point of view). E.g. a ring creation, file
registration, or waiting for available CQE. Proposed approach is to
get a pointer to an original structure of interest (ring context, or
request), and then find relevant events. io_uring_queue_async_work
also exposes a pointer to work_struct, to be able to track down
corresponding workqueue events.
* those, that provide performance related information. Mostly it's about
events that change the flow of requests, e.g. whether an async work
was queued, or delayed due to some dependencies. Another important
case is how io_uring optimizations (e.g. registered files) are
utilized.
Signed-off-by: Dmitrii Dolgov <9erthalion6@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-15 19:02:01 +02:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-04-29 01:40:30 +09:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:56 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
2025-11-12 12:45:56 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:56 +00:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2025-11-12 12:45:56 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
2025-01-27 10:57:17 -05:00
|
|
|
|
2023-08-21 17:15:52 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-01-27 10:57:17 -05:00
|
|
|
|
2023-08-21 17:15:52 -04:00
|
|
|
|
|
|
|
|
|
2025-01-27 10:57:18 -05:00
|
|
|
|
2023-08-21 17:15:52 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-01-27 10:57:17 -05:00
|
|
|
|
2023-08-21 17:15:52 -04:00
|
|
|
|
2025-01-27 10:57:17 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-01-27 10:57:18 -05:00
|
|
|
|
|
|
|
|
|
2023-08-21 17:15:52 -04:00
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2025-01-27 10:57:17 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-21 17:15:52 -04:00
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-09-05 09:57:00 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-10-15 12:19:33 -06:00
|
|
|
|
|
|
|
|
|
2024-09-05 09:57:00 +02:00
|
|
|
|
|
|
|
|
|
2022-08-11 09:11:16 +02:00
|
|
|
|
2020-01-29 14:39:41 +01:00
|
|
|
|
2022-08-11 09:11:16 +02:00
|
|
|
|
2020-01-29 14:39:41 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-08-11 09:11:16 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-01-29 14:39:41 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-08-11 09:11:16 +02:00
|
|
|
|
|
|
|
|
|
2020-01-29 14:39:41 +01:00
|
|
|
|
2020-02-24 11:32:45 +03:00
|
|
|
|
2020-01-29 14:39:41 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-06-17 17:53:55 +08:00
|
|
|
|
|
|
|
|
|
2020-01-29 14:39:41 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-02-24 11:32:45 +03:00
|
|
|
|
2022-08-11 09:11:16 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-01-29 14:39:41 +01:00
|
|
|
|
|
|
|
|
|
2021-06-24 15:09:58 +01:00
|
|
|
|
2020-01-29 14:39:41 +01:00
|
|
|
|
2020-02-24 11:32:45 +03:00
|
|
|
|
2021-08-25 12:25:45 +01:00
|
|
|
|
2022-09-01 11:54:04 +01:00
|
|
|
|
2025-05-06 17:47:27 +05:30
|
|
|
|
|
|
|
|
|
2022-09-01 11:54:04 +01:00
|
|
|
|
2022-03-23 08:44:19 -07:00
|
|
|
|
2022-08-11 09:11:16 +02:00
|
|
|
|
2024-11-28 16:52:36 +05:30
|
|
|
|
|
|
|
|
|
2022-08-11 09:11:16 +02:00
|
|
|
|
2020-01-29 14:39:41 +01:00
|
|
|
|
2021-04-27 16:13:53 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-08-25 20:51:40 +01:00
|
|
|
|
|
|
|
|
|
io_uring: add support for ring mapped supplied buffers
Provided buffers allow an application to supply io_uring with buffers
that can then be grabbed for a read/receive request, when the data
source is ready to deliver data. The existing scheme relies on using
IORING_OP_PROVIDE_BUFFERS to do that, but it can be difficult to use
in real world applications. It's pretty efficient if the application
is able to supply back batches of provided buffers when they have been
consumed and the application is ready to recycle them, but if
fragmentation occurs in the buffer space, it can become difficult to
supply enough buffers at the time. This hurts efficiency.
Add a register op, IORING_REGISTER_PBUF_RING, which allows an application
to setup a shared queue for each buffer group of provided buffers. The
application can then supply buffers simply by adding them to this ring,
and the kernel can consume then just as easily. The ring shares the head
with the application, the tail remains private in the kernel.
Provided buffers setup with IORING_REGISTER_PBUF_RING cannot use
IORING_OP_{PROVIDE,REMOVE}_BUFFERS for adding or removing entries to the
ring, they must use the mapped ring. Mapped provided buffer rings can
co-exist with normal provided buffers, just not within the same group ID.
To gauge overhead of the existing scheme and evaluate the mapped ring
approach, a simple NOP benchmark was written. It uses a ring of 128
entries, and submits/completes 32 at the time. 'Replenish' is how
many buffers are provided back at the time after they have been
consumed:
Test Replenish NOPs/sec
================================================================
No provided buffers NA ~30M
Provided buffers 32 ~16M
Provided buffers 1 ~10M
Ring buffers 32 ~27M
Ring buffers 1 ~27M
The ring mapped buffers perform almost as well as not using provided
buffers at all, and they don't care if you provided 1 or more back at
the same time. This means application can just replenish as they go,
rather than need to batch and compact, further reducing overhead in the
application. The NOP benchmark above doesn't need to do any compaction,
so that overhead isn't even reflected in the above test.
Co-developed-by: Dylan Yudaken <dylany@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2022-04-30 14:38:53 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-08-25 20:51:40 +01:00
|
|
|
|
2021-04-27 16:13:53 +01:00
|
|
|
|
|
|
|
|
|
2021-09-15 12:03:38 +01:00
|
|
|
|
|
|
|
|
|
2021-04-27 16:13:53 +01:00
|
|
|
|
2024-01-28 20:05:47 -07:00
|
|
|
|
2021-06-24 15:09:58 +01:00
|
|
|
|
2022-04-25 19:49:00 -06:00
|
|
|
|
|
|
|
|
|
2023-09-28 20:43:24 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-06-15 16:27:42 -06:00
|
|
|
|
2022-05-23 16:56:21 -06:00
|
|
|
|
2025-02-27 14:39:14 -08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-02 14:38:01 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2024-09-05 09:57:00 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-02 14:38:01 -06:00
|
|
|
|
2024-04-01 15:16:19 -06:00
|
|
|
|
2025-01-31 17:28:21 +00:00
|
|
|
|
2024-04-01 15:16:19 -06:00
|
|
|
|
2023-08-21 17:15:52 -04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Add io_uring IO interface
The submission queue (SQ) and completion queue (CQ) rings are shared
between the application and the kernel. This eliminates the need to
copy data back and forth to submit and complete IO.
IO submissions use the io_uring_sqe data structure, and completions
are generated in the form of io_uring_cqe data structures. The SQ
ring is an index into the io_uring_sqe array, which makes it possible
to submit a batch of IOs without them being contiguous in the ring.
The CQ ring is always contiguous, as completion events are inherently
unordered, and hence any io_uring_cqe entry can point back to an
arbitrary submission.
Two new system calls are added for this:
io_uring_setup(entries, params)
Sets up an io_uring instance for doing async IO. On success,
returns a file descriptor that the application can mmap to
gain access to the SQ ring, CQ ring, and io_uring_sqes.
io_uring_enter(fd, to_submit, min_complete, flags, sigset, sigsetsize)
Initiates IO against the rings mapped to this fd, or waits for
them to complete, or both. The behavior is controlled by the
parameters passed in. If 'to_submit' is non-zero, then we'll
try and submit new IO. If IORING_ENTER_GETEVENTS is set, the
kernel will wait for 'min_complete' events, if they aren't
already available. It's valid to set IORING_ENTER_GETEVENTS
and 'min_complete' == 0 at the same time, this allows the
kernel to return already completed events without waiting
for them. This is useful only for polling, as for IRQ
driven IO, the application can just check the CQ ring
without entering the kernel.
With this setup, it's possible to do async IO with a single system
call. Future developments will enable polled IO with this interface,
and polled submission as well. The latter will enable an application
to do IO without doing ANY system calls at all.
For IRQ driven IO, an application only needs to enter the kernel for
completions if it wants to wait for them to occur.
Each io_uring is backed by a workqueue, to support buffered async IO
as well. We will only punt to an async context if the command would
need to wait for IO on the device side. Any data that can be accessed
directly in the page cache is done inline. This avoids the slowness
issue of usual threadpools, since cached data is accessed as quickly
as a sync interface.
Sample application: http://git.kernel.dk/cgit/fio/plain/t/io_uring.c
Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-01-07 10:46:33 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
