[ Upstream commit 6622b76fe9 ]
Mixing AER Event Type and Event Info has masking clashes. Just print the
event type, but also include the event info of the AER result in the
trace.
Fixes: 09bd1ff4b1 ("nvme-core: add async event trace helper")
Reported-by: Nate Thornton <nate.thornton@samsung.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Minwoo Im <minwoo.im@samsung.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Keith Busch <kbusch@kernel.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 88eaba8032 ]
When we allocate a nvme-tcp queue, we set the data_ready callback before
we actually need to use it. This creates the potential that if a stray
controller sends us data on the socket before we connect, we can trigger
the io_work and start consuming the socket.
In this case reported: we failed to allocate one of the io queues, and
as we start releasing the queues that we already allocated, we get
a UAF [1] from the io_work which is running before it should really.
Fix this by setting the socket ops callbacks only before we start the
queue, so that we can't accidentally schedule the io_work in the
initialization phase before the queue started. While we are at it,
rename nvme_tcp_restore_sock_calls to pair with nvme_tcp_setup_sock_ops.
[1]:
[16802.107284] nvme nvme4: starting error recovery
[16802.109166] nvme nvme4: Reconnecting in 10 seconds...
[16812.173535] nvme nvme4: failed to connect socket: -111
[16812.173745] nvme nvme4: Failed reconnect attempt 1
[16812.173747] nvme nvme4: Reconnecting in 10 seconds...
[16822.413555] nvme nvme4: failed to connect socket: -111
[16822.413762] nvme nvme4: Failed reconnect attempt 2
[16822.413765] nvme nvme4: Reconnecting in 10 seconds...
[16832.661274] nvme nvme4: creating 32 I/O queues.
[16833.919887] BUG: kernel NULL pointer dereference, address: 0000000000000088
[16833.920068] nvme nvme4: Failed reconnect attempt 3
[16833.920094] #PF: supervisor write access in kernel mode
[16833.920261] nvme nvme4: Reconnecting in 10 seconds...
[16833.920368] #PF: error_code(0x0002) - not-present page
[16833.921086] Workqueue: nvme_tcp_wq nvme_tcp_io_work [nvme_tcp]
[16833.921191] RIP: 0010:_raw_spin_lock_bh+0x17/0x30
...
[16833.923138] Call Trace:
[16833.923271] <TASK>
[16833.923402] lock_sock_nested+0x1e/0x50
[16833.923545] nvme_tcp_try_recv+0x40/0xa0 [nvme_tcp]
[16833.923685] nvme_tcp_io_work+0x68/0xa0 [nvme_tcp]
[16833.923824] process_one_work+0x1e8/0x390
[16833.923969] worker_thread+0x53/0x3d0
[16833.924104] ? process_one_work+0x390/0x390
[16833.924240] kthread+0x124/0x150
[16833.924376] ? set_kthread_struct+0x50/0x50
[16833.924518] ret_from_fork+0x1f/0x30
[16833.924655] </TASK>
Reported-by: Yanjun Zhang <zhangyanjun@cestc.cn>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Tested-by: Yanjun Zhang <zhangyanjun@cestc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit def84ab600 ]
Identify CNS 06h (I/O Command Set Specific Identify Controller data
structure) is supported only on i/o controllers.
But nvme_init_non_mdts_limits() currently invokes this on all
controllers. Correct this by ensuring this is sent to I/O
controllers only.
Signed-off-by: Martin George <marting@netapp.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit d3205ab75e ]
The device can report discard support without setting the ONCS DSM bit.
When not set, the driver clears max_discard_size expecting it to be set
later. We don't know the size until we have the namespace format,
though, so setting it is deferred until configuring one, but the driver
was abandoning the discard settings due to that initial clearing.
Move the max_discard_size calculation above the check for a '0' discard
size.
Fixes: 1a86924e4f ("nvme: fix interpretation of DMRSL")
Reported-by: Laurence Oberman <loberman@redhat.com>
Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Niklas Cassel <niklas.cassel@wdc.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Tested-by: Laurence Oberman <loberman@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 9d2789ac9d upstream.
io_uring_cmd_done() currently assumes that the uring_lock is held
when invoked, and while it generally is, this is not guaranteed.
Pass in the issue_flags associated with it, so that we have
IO_URING_F_UNLOCKED available to be able to lock the CQ ring
appropriately when completing events.
Cc: stable@vger.kernel.org
Fixes: ee692a21e9 ("fs,io_uring: add infrastructure for uring-cmd")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 37f0dc2ec7 ]
When investigating one customer report on warning in nvme_setup_discard,
we observed the controller(nvme/tcp) actually exposes
queue_max_discard_segments(req->q) == 1.
Obviously the current code can't handle this situation, since contiguity
merge like normal RW request is taken.
Fix the issue by building range from request sector/nr_sectors directly.
Fixes: b35ba01ea6 ("nvme: support ranged discard requests")
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 26a57cb355 ]
The kernel always logs the unique subsystem name for a discovery
controller, even in the case user space asked for the well known.
This has lead to confusion as the logs of nvme-cli and the kernel
logs didn't match.
First, nvme-cli connects to the well known discovery controller to
figure out if it supports TP8013. If so then nvme-cli disconnects and
connects to the unique discovery controller. Currently, the kernel show
that user space connected twice to the unique one.
To avoid further confusion, show the well known discovery controller if
user space asked for it:
$ nvme connect-all -v -t tcp -a 192.168.0.1
nvme0: nqn.2014-08.org.nvmexpress.discovery connected
nvme0: nqn.2014-08.org.nvmexpress.discovery disconnected
nvme0: nqn.discovery connected
kernel log:
nvme nvme0: new ctrl: NQN "nqn.2014-08.org.nvmexpress.discovery", addr 192.168.0.1:8009
nvme nvme0: Removing ctrl: NQN "nqn.2014-08.org.nvmexpress.discovery"
nvme nvme0: new ctrl: NQN "nqn.discovery", addr 192.168.0.1:8009
Fixes: e5ea42faa7 ("nvme: display correct subsystem NQN")
Signed-off-by: Daniel Wagner <dwagner@suse.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 76d54bf20c ]
While the error recovery work is temporarily failing reconnect attempts,
running the 'nvme list' command causes a kernel NULL pointer dereference
by calling getsockname() with a released socket.
During error recovery work, the nvme tcp socket is released and a new one
created, so it is not safe to access the socket without proper check.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Fixes: 02c57a82c0 ("nvme-tcp: print actual source IP address through sysfs "address" attr")
Reviewed-by: Martin Belanger <martin.belanger@dell.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 0dd6fff2aa ]
Bring back the check of the Identify Namespace return value for the
legacy NVMe 1.0-style sequential scanning. While NVMe 1.0 does not
support namespace management, there are "modern" cloud solutions like
Google Cloud Platform that claim the obsolete 1.0 compliance for no
good reason while supporting proprietary sideband namespace management.
Fixes: 1a893c2bfe ("nvme: refactor namespace probing")
Reported-by: Nils Hanke <nh@edgeless.systems>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Tested-by: Nils Hanke <nh@edgeless.systems>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit e917a849c3 upstream.
The sysfs group containing the cmb attributes is registered before the
driver knows if they need to be visible or not. Update the group when
cmb attributes are known to exist so the visibility setting is correct.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=217037
Fixes: 86adbf0cdb ("nvme: simplify transport specific device attribute handling")
Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 91c11d5f32 ]
when starting error recovery there might be a authentication work
running, and it involves I/O commands. Given the controller is tearing
down there is no chance for the I/O to complete other than timing out
which may unnecessarily take a full io timeout.
So first tear down the queues, fail/cancel all inflight I/O (including
potentially authentication) and only then stop authentication. This
ensures that failover is not stalled due to blocked authentication I/O.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 1f1a4f8956 ]
when starting error recovery there might be a authentication work
running, and it involves I/O commands. Given the controller is tearing
down there is no chance for the I/O to complete other than timing out
which may unnecessarily take a full io timeout.
So first tear down the queues, fail/cancel all inflight I/O (including
potentially authentication) and only then stop authentication. This
ensures that failover is not stalled due to blocked authentication I/O.
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 6fbf13c0e2 ]
In nvme_alloc_io_tag_set(), the connect_q pointer should be set to NULL
in case of error to avoid potential invalid pointer dereferences.
Signed-off-by: Maurizio Lombardi <mlombard@redhat.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit fd62678ab5 ]
If nvme_alloc_admin_tag_set() fails, the admin_q and fabrics_q pointers
are left with an invalid, non-NULL value. Other functions may then check
the pointers and dereference them, e.g. in
nvme_probe() -> out_disable: -> nvme_dev_remove_admin().
Fix the bug by setting admin_q and fabrics_q to NULL in case of error.
Also use the set variable to free the tag_set as ctrl->admin_tagset isn't
initialized yet.
Signed-off-by: Maurizio Lombardi <mlombard@redhat.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit de4eda9de2 ]
READ/WRITE proved to be actively confusing - the meanings are
"data destination, as used with read(2)" and "data source, as
used with write(2)", but people keep interpreting those as
"we read data from it" and "we write data to it", i.e. exactly
the wrong way.
Call them ITER_DEST and ITER_SOURCE - at least that is harder
to misinterpret...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Stable-dep-of: 6dd88fd59d ("vhost-scsi: unbreak any layout for response")
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit c0a4a1eafb ]
NVMe controller register access hangs indefinitely when the co-processor
is not running. A missed reset is preferable over a hanging thread since
it could be recoverable.
Signed-off-by: Janne Grunau <j@jannau.net>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 85eee6341a ]
The namespace head saves the Command Set Indicator enum, so use that
instead of the Command Set Selected. The two values are not the same.
Fixes: 831ed60c2a ("nvme: also return I/O command effects from nvme_command_effects")
Signed-off-by: Keith Busch <kbusch@kernel.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 98e3528012 ]
ctrl->ops is used by nvme_alloc_admin_tag_set() but set by
nvme_init_ctrl() so reorder the calls to avoid a NULL pointer
dereference.
Fixes: 6dfba1c09c ("nvme-fc: use the tagset alloc/free helpers")
Signed-off-by: Ross Lagerwall <ross.lagerwall@citrix.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit db45e1a5dd ]
All nvme transports should be using the same flags for their tagsets,
with the exception for the blocking flag that should only be set for
transports that can block in ->queue_rq.
Add a NVME_F_BLOCKING flag to nvme_ctrl_ops to control the blocking
behavior and lift setting the flags into nvme_alloc_{admin,io}_tag_set.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Stable-dep-of: 98e3528012 ("nvme-fc: fix initialization order")
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 86adbf0cdb ]
Allow the transport driver to override the attribute groups for the
control device, so that the PCIe driver doesn't manually have to add a
group after device creation and keep track of it.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Tested-by Gerd Bayer <gbayer@linxu.ibm.com>
Stable-dep-of: 98e3528012 ("nvme-fc: fix initialization order")
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 1c58420858 ]
Polling the completion can progress the request state to IDLE, either
inline with the completion, or through softirq. Either way, the state
may not be COMPLETED, so don't check for that. We only care if the state
isn't IN_FLIGHT.
This is fixing an issue where the driver aborts an IO that we just
completed. Seeing the "aborting" message instead of "polled" is very
misleading as to where the timeout problem resides.
Fixes: bf392a5dc0 ("nvme-pci: Remove tag from process cq")
Signed-off-by: Keith Busch <kbusch@kernel.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 613b14884b upstream.
This can't happen right now, but in preparation for allowing
bio_split_to_limits() returning NULL if it ended the bio, check for it
in all the callers.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 831ed60c2a ]
To be able to use the Commands Supported and Effects Log for allowing
unprivileged passtrough, it needs to be corretly reported for I/O
commands as well. Return the I/O command effects from
nvme_command_effects, and also add a default list of effects for the
NVM command set. For other command sets, the Commands Supported and
Effects log is required to be present already.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Kanchan Joshi <joshi.k@samsung.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 841734234a ]
The size allocated out of the dma pool is at most NVME_CTRL_PAGE_SIZE,
which may be smaller than the PAGE_SIZE.
Fixes: c61b82c7b7 ("nvme-pci: fix PRP pool size")
Signed-off-by: Keith Busch <kbusch@kernel.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit c89a529e82 ]
Convert the max size to bytes to match the units of the divisor that
calculates the worst-case number of PRP entries.
The result is used to determine how many PRP Lists are required. The
code was previously rounding this to 1 list, but we can require 2 in the
worst case. In that scenario, the driver would corrupt memory beyond the
size provided by the mempool.
While unlikely to occur (you'd need a 4MB in exactly 127 phys segments
on a queue that doesn't support SGLs), this memory corruption has been
observed by kfence.
Cc: Jens Axboe <axboe@kernel.dk>
Fixes: 943e942e62 ("nvme-pci: limit max IO size and segments to avoid high order allocations")
Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Kanchan Joshi <joshi.k@samsung.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit b5f96cb719 ]
When using shadow doorbells, the event index and the doorbell values are
written to host memory. Prior to this patch, the values written would
erroneously be written in host endianness. This causes trouble on
big-endian platforms. Fix this by adding missing endian conversions.
This issue was noticed by Guenter while testing various big-endian
platforms under QEMU[1]. A similar fix required for hw/nvme in QEMU is
up for review as well[2].
[1]: https://lore.kernel.org/qemu-devel/20221209110022.GA3396194@roeck-us.net/
[2]: https://lore.kernel.org/qemu-devel/20221212114409.34972-4-its@irrelevant.dk/
Fixes: f9f38e3338 ("nvme: improve performance for virtual NVMe devices")
Reported-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 01604350e1 ]
Replace ctrl ctrl_key/host_key only after nvme_auth_generate_key is successful.
Also, this fixes a bug where the keys are leaked.
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit dcef77274a ]
Don't look at ctrl->ops as only RDMA and TCP actually support multiple
maps.
Fixes: 6dfba1c09c ("nvme-fc: use the tagset alloc/free helpers")
Fixes: ceee1953f9 ("nvme-loop: use the tagset alloc/free helpers")
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit bcaf434b8f ]
In nvme_init_non_mdts_limits function we were returning 0 when kzalloc
failed; it now returns -ENOMEM.
Fixes: 5befc7c26e ("nvme: implement non-mdts command limits")
Signed-off-by: Joel Granados <j.granados@samsung.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
A device might have a core quirk for NVME_QUIRK_IGNORE_DEV_SUBNQN
(such as Samsung X5) but it would still give a:
"missing or invalid SUBNQN field"
warning as core quirks are filled after calling nvme_init_subnqn. Fill
ctrl->quirks from struct core_quirks before calling nvme_init_subsystem
to fix this.
Tested on a Samsung X5.
Fixes: ab9e00cc72 ("nvme: track subsystems")
Signed-off-by: Pankaj Raghav <p.raghav@samsung.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
If the prp2 field is not filled in nvme_setup_prp_simple(), the prp2
field is garbage data. According to nvme spec, the prp2 is reserved if
the data transfer does not cross a memory page boundary, so clear it to
zero if it is not used.
Signed-off-by: Lei Rao <lei.rao@intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
The driver is spamming the kernel logs for entirely harmless errors from
user space submitting unsupported commands. Just silence the errors.
The application has direct access to command status, so there's no need
to log these.
And since every passthrough command now uses the quiet flag, move the
setting to the common initializer.
Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Alan Adamson <alan.adamson@oracle.com>
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Kanchan Joshi <joshi.k@samsung.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Daniel Wagner <dwagner@suse.de>
Tested-by: Alan Adamson <alan.adamson@oracle.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
NVMe spec requires all transports support dword aligned addresses, which
is already set in the namespace request_queue. Set the same limit in the
multipath device's request_queue as well.
Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
When destroying a queue, when calling sock_release, the network stack
might need to allocate an skb to send a FIN/RST. When that happens
during memory pressure, there is a need to reclaim memory, which
in turn may ask the nvme-tcp device to write out dirty pages, however
this is not possible due to a ctrl teardown that is going on.
Set PF_MEMALLOC to the task that releases the socket to grant access
to PF_MEMALLOC reserves. In addition, do the same for the nvme-tcp
thread as this may also originate from the swap itself and should
be more resilient to memory pressure situations.
This fixes the following lockdep complaint:
--
======================================================
WARNING: possible circular locking dependency detected
6.0.0-rc2+ #25 Tainted: G W
------------------------------------------------------
kswapd0/92 is trying to acquire lock:
ffff888114003240 (sk_lock-AF_INET-NVME){+.+.}-{0:0}, at: tcp_sendpage+0x23/0xa0
but task is already holding lock:
ffffffff97e95ca0 (fs_reclaim){+.+.}-{0:0}, at: balance_pgdat+0x987/0x10d0
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #1 (fs_reclaim){+.+.}-{0:0}:
fs_reclaim_acquire+0x11e/0x160
kmem_cache_alloc_node+0x44/0x530
__alloc_skb+0x158/0x230
tcp_send_active_reset+0x7e/0x730
tcp_disconnect+0x1272/0x1ae0
__tcp_close+0x707/0xd90
tcp_close+0x26/0x80
inet_release+0xfa/0x220
sock_release+0x85/0x1a0
nvme_tcp_free_queue+0x1fd/0x470 [nvme_tcp]
nvme_do_delete_ctrl+0x130/0x13d [nvme_core]
nvme_sysfs_delete.cold+0x8/0xd [nvme_core]
kernfs_fop_write_iter+0x356/0x530
vfs_write+0x4e8/0xce0
ksys_write+0xfd/0x1d0
do_syscall_64+0x58/0x80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
-> #0 (sk_lock-AF_INET-NVME){+.+.}-{0:0}:
__lock_acquire+0x2a0c/0x5690
lock_acquire+0x18e/0x4f0
lock_sock_nested+0x37/0xc0
tcp_sendpage+0x23/0xa0
inet_sendpage+0xad/0x120
kernel_sendpage+0x156/0x440
nvme_tcp_try_send+0x48a/0x2630 [nvme_tcp]
nvme_tcp_queue_rq+0xefb/0x17e0 [nvme_tcp]
__blk_mq_try_issue_directly+0x452/0x660
blk_mq_plug_issue_direct.constprop.0+0x207/0x700
blk_mq_flush_plug_list+0x6f5/0xc70
__blk_flush_plug+0x264/0x410
blk_finish_plug+0x4b/0xa0
shrink_lruvec+0x1263/0x1ea0
shrink_node+0x736/0x1a80
balance_pgdat+0x740/0x10d0
kswapd+0x5f2/0xaf0
kthread+0x256/0x2f0
ret_from_fork+0x1f/0x30
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(fs_reclaim);
lock(sk_lock-AF_INET-NVME);
lock(fs_reclaim);
lock(sk_lock-AF_INET-NVME);
*** DEADLOCK ***
3 locks held by kswapd0/92:
#0: ffffffff97e95ca0 (fs_reclaim){+.+.}-{0:0}, at: balance_pgdat+0x987/0x10d0
#1: ffff88811f21b0b0 (q->srcu){....}-{0:0}, at: blk_mq_flush_plug_list+0x6b3/0xc70
#2: ffff888170b11470 (&queue->send_mutex){+.+.}-{3:3}, at: nvme_tcp_queue_rq+0xeb9/0x17e0 [nvme_tcp]
Fixes: 3f2304f8c6 ("nvme-tcp: add NVMe over TCP host driver")
Reported-by: Daniel Wagner <dwagner@suse.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Tested-by: Daniel Wagner <dwagner@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
In nvme_tcp_ddgst_update(), sg_init_marker() is called with an
uninitialized scatterlist. This is probably fine, but gcc complains:
CC [M] drivers/nvme/host/tcp.o
In file included from ./include/linux/dma-mapping.h:10,
from ./include/linux/skbuff.h:31,
from ./include/net/net_namespace.h:43,
from ./include/linux/netdevice.h:38,
from ./include/net/sock.h:46,
from drivers/nvme/host/tcp.c:12:
In function ‘sg_mark_end’,
inlined from ‘sg_init_marker’ at ./include/linux/scatterlist.h:356:2,
inlined from ‘nvme_tcp_ddgst_update’ at drivers/nvme/host/tcp.c:390:2:
./include/linux/scatterlist.h:234:11: error: ‘sg.page_link’ is used uninitialized [-Werror=uninitialized]
234 | sg->page_link |= SG_END;
| ~~^~~~~~~~~~~
drivers/nvme/host/tcp.c: In function ‘nvme_tcp_ddgst_update’:
drivers/nvme/host/tcp.c:388:28: note: ‘sg’ declared here
388 | struct scatterlist sg;
| ^~
cc1: all warnings being treated as errors
Use sg_init_table() instead, which basically memset the scatterlist to
zero first before calling sg_init_marker().
Signed-off-by: Nam Cao <namcaov@gmail.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Recent commit 52fde2c07d ("nvme: set dma alignment to dword") has
caused a regression on our platform.
It turned out that the nvme_get_log() method invocation caused the
nvme_hwmon_data structure instance corruption. In particular the
nvme_hwmon_data.ctrl pointer was overwritten either with zeros or with
garbage. After some research we discovered that the problem happened
even before the actual NVME DMA execution, but during the buffer mapping.
Since our platform is DMA-noncoherent, the mapping implied the cache-line
invalidations or write-backs depending on the DMA-direction parameter.
In case of the NVME SMART log getting the DMA was performed
from-device-to-memory, thus the cache-invalidation was activated during
the buffer mapping. Since the log-buffer isn't cache-line aligned, the
cache-invalidation caused the neighbour data to be discarded. The
neighbouring data turned to be the data surrounding the buffer in the
framework of the nvme_hwmon_data structure.
In order to fix that we need to make sure that the whole log-buffer is
defined within the cache-line-aligned memory region so the
cache-invalidation procedure wouldn't involve the adjacent data. One of
the option to guarantee that is to kmalloc the DMA-buffer [1]. Seeing the
rest of the NVME core driver prefer that method it has been chosen to fix
this problem too.
Note after a deeper researches we found out that the denoted commit wasn't
a root cause of the problem. It just revealed the invalidity by activating
the DMA-based NVME SMART log getting performed in the framework of the
NVME hwmon driver. The problem was here since the initial commit of the
driver.
[1] Documentation/core-api/dma-api-howto.rst
Fixes: 400b6a7b13 ("nvme: Add hardware monitoring support")
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Christoph Hellwig <hch@lst.de>
An NVMe controller works perfectly fine even when the hwmon
initialization fails. Stop returning errors that do not come from a
controller reset from nvme_hwmon_init to handle this case consistently.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Serge Semin <fancer.lancer@gmail.com>
NVMe uses PRPs for data transfers and has no specific limit for a single
DMA segement. Limiting the size will cause problems because the block
layer assumes PRP-ish devices using a virt boundary mask don't have a
segment limit. And while this is true, we also really need to tell the
DMA mapping layer about it, otherwise dma-debug will trip over it.
Fixes: 5bd2927ace ("nvme-apple: Add initial Apple SoC NVMe driver")
Suggested-by: Sven Peter <sven@svenpeter.dev>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
[hch: rewrote the commit message based on the PCIe commit]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Curtin <ecurtin@redhat.com>
Reviewed-by: Sven Peter <sven@svenpeter.dev>
Kingston SSDs do support NVMe Write_Zeroes cmd but take long time to
process. The firmware version is locked by these SSDs, we can not expect
firmware improvement, so disable Write_Zeroes cmd.
Signed-off-by: Xander Li <xander_li@kingston.com.tw>
Signed-off-by: Christoph Hellwig <hch@lst.de>
There is typo here so it releases the wrong variable. "ctrl->admin_q"
was intended instead of "ctrl->fabrics_q".
Fixes: fe60e8c534 ("nvme: add common helpers to allocate and free tagsets")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
When we revalidate paths as part of ns size change (as of commit
e7d65803e2), it is possible that during the path revalidation, the
only paths that is IO capable (i.e. optimized/non-optimized) are the
ones that ns resize was not yet informed to the host, which will cause
inflight requests to be requeued (as we have available paths but none
are IO capable). These requests on the requeue list are waiting for
someone to resubmit them at some point.
The IO capable paths will eventually notify the ns resize change to the
host, but there is nothing that will kick the requeue list to resubmit
the queued requests.
Fix this by always kicking the requeue list, and if no IO capable path
exists, these requests will be queued again.
A typical log that indicates that IOs are requeued:
--
nvme nvme1: creating 4 I/O queues.
nvme nvme1: new ctrl: "testnqn1"
nvme nvme2: creating 4 I/O queues.
nvme nvme2: mapped 4/0/0 default/read/poll queues.
nvme nvme2: new ctrl: NQN "testnqn1", addr 127.0.0.1:8009
nvme nvme1: rescanning namespaces.
nvme1n1: detected capacity change from 2097152 to 4194304
block nvme1n1: no usable path - requeuing I/O
block nvme1n1: no usable path - requeuing I/O
block nvme1n1: no usable path - requeuing I/O
block nvme1n1: no usable path - requeuing I/O
block nvme1n1: no usable path - requeuing I/O
block nvme1n1: no usable path - requeuing I/O
block nvme1n1: no usable path - requeuing I/O
block nvme1n1: no usable path - requeuing I/O
block nvme1n1: no usable path - requeuing I/O
block nvme1n1: no usable path - requeuing I/O
nvme nvme2: rescanning namespaces.
--
Reported-by: Yogev Cohen <yogev@lightbitslabs.com>
Fixes: e7d65803e2 ("nvme-multipath: revalidate paths during rescan")
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Cc: <stable@vger.kernel.org> # v5.15+
Signed-off-by: Christoph Hellwig <hch@lst.de>
