If an IIO driver uses callbacks from another IIO driver and calls
iio_channel_start_all_cb() from one of its buffer setup ops, then
lockdep complains due to the lock nesting, as in the below example with
lmp91000.
Since the locks are being taken on different IIO devices, there is no
actual deadlock. Fix the warning by telling lockdep to use a different
class for each iio_device.
============================================
WARNING: possible recursive locking detected
--------------------------------------------
python3/23 is trying to acquire lock:
(&indio_dev->mlock){+.+.}-{3:3}, at: iio_update_buffers
but task is already holding lock:
(&indio_dev->mlock){+.+.}-{3:3}, at: enable_store
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0
----
lock(&indio_dev->mlock);
lock(&indio_dev->mlock);
*** DEADLOCK ***
May be due to missing lock nesting notation
5 locks held by python3/23:
#0: (sb_writers#5){.+.+}-{0:0}, at: ksys_write
#1: (&of->mutex){+.+.}-{3:3}, at: kernfs_fop_write_iter
#2: (kn->active#14){.+.+}-{0:0}, at: kernfs_fop_write_iter
#3: (&indio_dev->mlock){+.+.}-{3:3}, at: enable_store
#4: (&iio_dev_opaque->info_exist_lock){+.+.}-{3:3}, at: iio_update_buffers
Call Trace:
__mutex_lock
iio_update_buffers
iio_channel_start_all_cb
lmp91000_buffer_postenable
__iio_update_buffers
enable_store
Fixes: 67e17300dc ("iio: potentiostat: add LMP91000 support")
Signed-off-by: Vincent Whitchurch <vincent.whitchurch@axis.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Link: https://lore.kernel.org/r/20220829091840.2791846-1-vincent.whitchurch@axis.com
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Adding support for max11205 16-bit single-channel ultra-low power
delta-sigma adc.
The MAX11205 is compatible with the 2-wire interface and uses
SCLK and RDY/DOUT for serial communications. In this mode, all
controls are implemented by timing the high or low phase of the SCLK.
The 2-wire serial interface only allows for data to be read out through
the RDY/DOUT output.
Datasheet: https://datasheets.maximintegrated.com/en/ds/MAX11205.pdf
Signed-off-by: Ramona Bolboaca <ramona.bolboaca@analog.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Link: https://lore.kernel.org/r/20220831133021.215625-2-ramona.bolboaca@analog.com
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Passed through PCI device sometimes misbehave on Gen1 VMs when Hyper-V
DRM driver is also loaded. Looking at IOMEM assignment, we can see e.g.
$ cat /proc/iomem
...
f8000000-fffbffff : PCI Bus 0000:00
f8000000-fbffffff : 0000:00:08.0
f8000000-f8001fff : bb8c4f33-2ba2-4808-9f7f-02f3b4da22fe
...
fe0000000-fffffffff : PCI Bus 0000:00
fe0000000-fe07fffff : bb8c4f33-2ba2-4808-9f7f-02f3b4da22fe
fe0000000-fe07fffff : 2ba2:00:02.0
fe0000000-fe07fffff : mlx4_core
the interesting part is the 'f8000000' region as it is actually the
VM's framebuffer:
$ lspci -v
...
0000:00:08.0 VGA compatible controller: Microsoft Corporation Hyper-V virtual VGA (prog-if 00 [VGA controller])
Flags: bus master, fast devsel, latency 0, IRQ 11
Memory at f8000000 (32-bit, non-prefetchable) [size=64M]
...
hv_vmbus: registering driver hyperv_drm
hyperv_drm 5620e0c7-8062-4dce-aeb7-520c7ef76171: [drm] Synthvid Version major 3, minor 5
hyperv_drm 0000:00:08.0: vgaarb: deactivate vga console
hyperv_drm 0000:00:08.0: BAR 0: can't reserve [mem 0xf8000000-0xfbffffff]
hyperv_drm 5620e0c7-8062-4dce-aeb7-520c7ef76171: [drm] Cannot request framebuffer, boot fb still active?
Note: "Cannot request framebuffer" is not a fatal error in
hyperv_setup_gen1() as the code assumes there's some other framebuffer
device there but we actually have some other PCI device (mlx4 in this
case) config space there!
The problem appears to be that vmbus_allocate_mmio() can use dedicated
framebuffer region to serve any MMIO request from any device. The
semantics one might assume of a parameter named "fb_overlap_ok"
aren't implemented because !fb_overlap_ok essentially has no effect.
The existing semantics are really "prefer_fb_overlap". This patch
implements the expected and needed semantics, which is to not allocate
from the frame buffer space when !fb_overlap_ok.
Note, Gen2 VMs are usually unaffected by the issue because
framebuffer region is already taken by EFI fb (in case kernel supports
it) but Gen1 VMs may have this region unclaimed by the time Hyper-V PCI
pass-through driver tries allocating MMIO space if Hyper-V DRM/FB drivers
load after it. Devices can be brought up in any sequence so let's
resolve the issue by always ignoring 'fb_mmio' region for non-FB
requests, even if the region is unclaimed.
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Link: https://lore.kernel.org/r/20220827130345.1320254-4-vkuznets@redhat.com
Signed-off-by: Wei Liu <wei.liu@kernel.org>
vmbus_reserve_fb() tries reserving framebuffer region iff
'screen_info.lfb_base' is set. Gen2 VMs seem to have it set by EFI
and/or by the kernel EFI FB driver (or, in some edge cases like kexec,
the address where the buffer was moved, see
https://lore.kernel.org/all/20201014092429.1415040-1-kasong@redhat.com/)
but on Gen1 VM it depends on bootloader behavior. With grub, it depends
on 'gfxpayload=' setting but in some cases it is observed to be zero.
That being said, relying on 'screen_info.lfb_base' to reserve
framebuffer region is risky. For Gen1 VMs, it should always be
possible to get the address from the dedicated PCI device instead.
Check for legacy PCI video device presence and reserve the whole
region for framebuffer on Gen1 VMs.
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Link: https://lore.kernel.org/r/20220827130345.1320254-3-vkuznets@redhat.com
Signed-off-by: Wei Liu <wei.liu@kernel.org>
There are already three places in kernel which define
PCI_VENDOR_ID_MICROSOFT and two for PCI_DEVICE_ID_HYPERV_VIDEO and
there's a need to use these from core VMBus code. Move the defines where
they belong.
No functional change.
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com> # pci_ids.h
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Link: https://lore.kernel.org/r/20220827130345.1320254-2-vkuznets@redhat.com
Signed-off-by: Wei Liu <wei.liu@kernel.org>
Merge series from Cezary Rojewski <cezary.rojewski@intel.com>:
Continuation of recent upstream discussion [1] regarding user string
tokenization.
First, parse_int_array_user() is introduced to allow for splitting
specified user string into a sequence of integers. Makes use of
get_options() internally so the parsing logic is not duplicated.
With that done, redundant parts of the sound driver are removed.
Originally similar functionality was added for the SOF sound driver. As
more users are on the horizon, it is desirable to update existing
string_helpers code and provide a unified solution.
During stress testing with CONFIG_SMP disabled, KASAN reports as below:
==================================================================
BUG: KASAN: use-after-free in __mutex_lock+0xe5/0xc30
Read of size 8 at addr ffff8881094223f8 by task stress/7789
CPU: 0 PID: 7789 Comm: stress Not tainted 6.0.0-rc1-00002-g0d53d2e882f9 #3
Hardware name: Red Hat KVM, BIOS 0.5.1 01/01/2011
Call Trace:
<TASK>
..
__mutex_lock+0xe5/0xc30
..
z_erofs_do_read_page+0x8ce/0x1560
..
z_erofs_readahead+0x31c/0x580
..
Freed by task 7787
kasan_save_stack+0x1e/0x40
kasan_set_track+0x20/0x30
kasan_set_free_info+0x20/0x40
__kasan_slab_free+0x10c/0x190
kmem_cache_free+0xed/0x380
rcu_core+0x3d5/0xc90
__do_softirq+0x12d/0x389
Last potentially related work creation:
kasan_save_stack+0x1e/0x40
__kasan_record_aux_stack+0x97/0xb0
call_rcu+0x3d/0x3f0
erofs_shrink_workstation+0x11f/0x210
erofs_shrink_scan+0xdc/0x170
shrink_slab.constprop.0+0x296/0x530
drop_slab+0x1c/0x70
drop_caches_sysctl_handler+0x70/0x80
proc_sys_call_handler+0x20a/0x2f0
vfs_write+0x555/0x6c0
ksys_write+0xbe/0x160
do_syscall_64+0x3b/0x90
The root cause is that erofs_workgroup_unfreeze() doesn't reset to
orig_val thus it causes a race that the pcluster reuses unexpectedly
before freeing.
Since UP platforms are quite rare now, such path becomes unnecessary.
Let's drop such specific-designed path directly instead.
Fixes: 73f5c66df3 ("staging: erofs: fix `erofs_workgroup_{try_to_freeze, unfreeze}'")
Reviewed-by: Yue Hu <huyue2@coolpad.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Link: https://lore.kernel.org/r/20220902045710.109530-1-hsiangkao@linux.alibaba.com
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Actually, 'compressedlcs' stores compressed block count rather than
lcluster count. Therefore, the number of bits for shifting the count
should be 'LOG_BLOCK_SIZE' rather than 'lclusterbits' although current
lcluster size is 4K.
The value of 'm_plen' will be wrong once we enable the non 4K-sized
lcluster.
Signed-off-by: Yue Hu <huyue2@coolpad.com>
Reviewed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Link: https://lore.kernel.org/r/20220812060150.8510-1-huyue2@coolpad.com
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
If erofs_fscache_alloc_request fail and then goto out, it will return 0.
it should return a negative error code instead of 0.
Fixes: d435d53228 ("erofs: change to use asynchronous io for fscache readpage/readahead")
Signed-off-by: Sun Ke <sunke32@huawei.com>
Reviewed-by: Jingbo Xu <jefflexu@linux.alibaba.com>
Reviewed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Link: https://lore.kernel.org/r/20220815034829.3940803-1-sunke32@huawei.com
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Remove the CONFIG_PREEMPT_RT symbol from the ifdef around
do_softirq_own_stack() and move it to Kconfig instead.
Enable softirq stacks based on SOFTIRQ_ON_OWN_STACK which depends on
HAVE_SOFTIRQ_ON_OWN_STACK and its default value is set to !PREEMPT_RT.
This ensures that softirq stacks are not used on PREEMPT_RT and avoids
a 'select' statement on an option which has a 'depends' statement.
Link: https://lore.kernel.org/YvN5E%2FPrHfUhggr7@linutronix.de
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
In the commit 617d32938d ("rpmsg: Move the rpmsg control device
from rpmsg_char to rpmsg_ctrl"), we split the rpmsg_char driver in two.
By default give everyone who had the old driver enabled the rpmsg_ctrl
driver too.
Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@foss.st.com>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
In the commit 617d32938d ("rpmsg: Move the rpmsg control device
from rpmsg_char to rpmsg_ctrl"), we split the rpmsg_char driver in two.
By default give everyone who had the old driver enabled the rpmsg_ctrl
driver too.
Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@foss.st.com>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Reviewed-by: Mukesh Ojha <quic_mojha@quicinc.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Make use of global integer-array parsing helper instead of the internal
one as both serve same purpose. With that, both strsplit_u32() and
tokenize_input() become unused so remove them.
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Cezary Rojewski <cezary.rojewski@intel.com>
Link: https://lore.kernel.org/r/20220904102840.862395-3-cezary.rojewski@intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Add new helper function to allow for splitting specified user string
into a sequence of integers. Internally it makes use of get_options() so
the returned sequence contains the integers extracted plus an additional
element that begins the sequence and specifies the integers count.
Suggested-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Cezary Rojewski <cezary.rojewski@intel.com>
Link: https://lore.kernel.org/r/20220904102840.862395-2-cezary.rojewski@intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
We're not in a hot path and don't want to miss this message,
therefore remove the net_ratelimit() check.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In preparation for FORTIFY_SOURCE doing bounds-check on memcpy(),
switch from __nlmsg_put to nlmsg_put(), and explain the bounds check
for dealing with the memcpy() across a composite flexible array struct.
Avoids this future run-time warning:
memcpy: detected field-spanning write (size 32) of single field "&errmsg->msg" at net/netlink/af_netlink.c:2447 (size 16)
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: Jozsef Kadlecsik <kadlec@netfilter.org>
Cc: Florian Westphal <fw@strlen.de>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: syzbot <syzkaller@googlegroups.com>
Cc: netfilter-devel@vger.kernel.org
Cc: coreteam@netfilter.org
Cc: netdev@vger.kernel.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20220901071336.1418572-1-keescook@chromium.org
Signed-off-by: David S. Miller <davem@davemloft.net>
Since commit 6a921de589 ("btrfs: zoned: introduce
space_info->active_total_bytes"), we're only counting the bytes of a
block group on an active zone as usable for metadata writes. But on a
SMR drive, we don't have active zones and short circuit some of the
logic.
This leads to an error on mount, because we cannot reserve space for
metadata writes.
Fix this by also setting the BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE bit in the
block-group's runtime flag if the zone is a conventional zone.
Fixes: 6a921de589 ("btrfs: zoned: introduce space_info->active_total_bytes")
Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Alexei Starovoitov says:
====================
Introduce any context BPF specific memory allocator.
Tracing BPF programs can attach to kprobe and fentry. Hence they run in
unknown context where calling plain kmalloc() might not be safe. Front-end
kmalloc() with per-cpu cache of free elements. Refill this cache asynchronously
from irq_work.
Major achievements enabled by bpf_mem_alloc:
- Dynamically allocated hash maps used to be 10 times slower than fully
preallocated. With bpf_mem_alloc and subsequent optimizations the speed
of dynamic maps is equal to full prealloc.
- Tracing bpf programs can use dynamically allocated hash maps. Potentially
saving lots of memory. Typical hash map is sparsely populated.
- Sleepable bpf programs can used dynamically allocated hash maps.
Future work:
- Expose bpf_mem_alloc as uapi FD to be used in dynptr_alloc, kptr_alloc
- Convert lru map to bpf_mem_alloc
- Further cleanup htab code. Example: htab_use_raw_lock can be removed.
Changelog:
v5->v6:
- Debugged the reason for selftests/bpf/test_maps ooming in a small VM that BPF CI is using.
Added patch 16 that optimizes the usage of rcu_barrier-s between bpf_mem_alloc and
hash map. It drastically improved the speed of htab destruction.
v4->v5:
- Fixed missing migrate_disable in hash tab free path (Daniel)
- Replaced impossible "memory leak" with WARN_ON_ONCE (Martin)
- Dropped sysctl kernel.bpf_force_dyn_alloc patch (Daniel)
- Added Andrii's ack
- Added new patch 15 that removes kmem_cache usage from bpf_mem_alloc.
It saves memory, speeds up map create/destroy operations
while maintains hash map update/delete performance.
v3->v4:
- fix build issue due to missing local.h on 32-bit arch
- add Kumar's ack
- proposal for next steps from Delyan:
https://lore.kernel.org/bpf/d3f76b27f4e55ec9e400ae8dcaecbb702a4932e8.camel@fb.com/
v2->v3:
- Rewrote the free_list algorithm based on discussions with Kumar. Patch 1.
- Allowed sleepable bpf progs use dynamically allocated maps. Patches 13 and 14.
- Added sysctl to force bpf_mem_alloc in hash map even if pre-alloc is
requested to reduce memory consumption. Patch 15.
- Fix: zero-fill percpu allocation
- Single rcu_barrier at the end instead of each cpu during bpf_mem_alloc destruction
v2 thread:
https://lore.kernel.org/bpf/20220817210419.95560-1-alexei.starovoitov@gmail.com/
v1->v2:
- Moved unsafe direct call_rcu() from hash map into safe place inside bpf_mem_alloc. Patches 7 and 9.
- Optimized atomic_inc/dec in hash map with percpu_counter. Patch 6.
- Tuned watermarks per allocation size. Patch 8
- Adopted this approach to per-cpu allocation. Patch 10.
- Fully converted hash map to bpf_mem_alloc. Patch 11.
- Removed tracing prog restriction on map types. Combination of all patches and final patch 12.
v1 thread:
https://lore.kernel.org/bpf/20220623003230.37497-1-alexei.starovoitov@gmail.com/
LWN article:
https://lwn.net/Articles/899274/
====================
Link: https://lore.kernel.org/r/
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
User space might be creating and destroying a lot of hash maps. Synchronous
rcu_barrier-s in a destruction path of hash map delay freeing of hash buckets
and other map memory and may cause artificial OOM situation under stress.
Optimize rcu_barrier usage between bpf hash map and bpf_mem_alloc:
- remove rcu_barrier from hash map, since htab doesn't use call_rcu
directly and there are no callback to wait for.
- bpf_mem_alloc has call_rcu_in_progress flag that indicates pending callbacks.
Use it to avoid barriers in fast path.
- When barriers are needed copy bpf_mem_alloc into temp structure
and wait for rcu barrier-s in the worker to let the rest of
hash map freeing to proceed.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20220902211058.60789-17-alexei.starovoitov@gmail.com
For bpf_mem_cache based hash maps the following stress test:
for (i = 1; i <= 512; i <<= 1)
for (j = 1; j <= 1 << 18; j <<= 1)
fd = bpf_map_create(BPF_MAP_TYPE_HASH, NULL, i, j, 2, 0);
creates many kmem_cache-s that are not mergeable in debug kernels
and consume unnecessary amount of memory.
Turned out bpf_mem_cache's free_list logic does batching well,
so usage of kmem_cache for fixes size allocations doesn't bring
any performance benefits vs normal kmalloc.
Hence get rid of kmem_cache in bpf_mem_cache.
That saves memory, speeds up map create/destroy operations,
while maintains hash map update/delete performance.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20220902211058.60789-16-alexei.starovoitov@gmail.com
Since hash map is now converted to bpf_mem_alloc and it's waiting for rcu and
rcu_tasks_trace GPs before freeing elements into global memory slabs it's safe
to use dynamically allocated hash maps in sleepable bpf programs.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220902211058.60789-15-alexei.starovoitov@gmail.com
Use call_rcu_tasks_trace() to wait for sleepable progs to finish.
Then use call_rcu() to wait for normal progs to finish
and finally do free_one() on each element when freeing objects
into global memory pool.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220902211058.60789-14-alexei.starovoitov@gmail.com
The hash map is now fully converted to bpf_mem_alloc. Its implementation is not
allocating synchronously and not calling call_rcu() directly. It's now safe to
use non-preallocated hash maps in all types of tracing programs including
BPF_PROG_TYPE_PERF_EVENT that runs out of NMI context.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220902211058.60789-13-alexei.starovoitov@gmail.com
Convert dynamic allocations in percpu hash map from alloc_percpu() to
bpf_mem_cache_alloc() from per-cpu bpf_mem_alloc. Since bpf_mem_alloc frees
objects after RCU gp the call_rcu() is removed. pcpu_init_value() now needs to
zero-fill per-cpu allocations, since dynamically allocated map elements are now
similar to full prealloc, since alloc_percpu() is not called inline and the
elements are reused in the freelist.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220902211058.60789-12-alexei.starovoitov@gmail.com
Extend bpf_mem_alloc to cache free list of fixed size per-cpu allocations.
Once such cache is created bpf_mem_cache_alloc() will return per-cpu objects.
bpf_mem_cache_free() will free them back into global per-cpu pool after
observing RCU grace period.
per-cpu flavor of bpf_mem_alloc is going to be used by per-cpu hash maps.
The free list cache consists of tuples { llist_node, per-cpu pointer }
Unlike alloc_percpu() that returns per-cpu pointer
the bpf_mem_cache_alloc() returns a pointer to per-cpu pointer and
bpf_mem_cache_free() expects to receive it back.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220902211058.60789-11-alexei.starovoitov@gmail.com
SLAB_TYPESAFE_BY_RCU makes kmem_caches non mergeable and slows down
kmem_cache_destroy. All bpf_mem_cache are safe to share across different maps
and programs. Convert SLAB_TYPESAFE_BY_RCU to batched call_rcu. This change
solves the memory consumption issue, avoids kmem_cache_destroy latency and
keeps bpf hash map performance the same.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220902211058.60789-10-alexei.starovoitov@gmail.com
The same low/high watermarks for every bucket in bpf_mem_cache consume
significant amount of memory. Preallocating 64 elements of 4096 bytes each in
the free list is not efficient. Make low/high watermarks and batching value
dependent on element size. This change brings significant memory savings.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220902211058.60789-9-alexei.starovoitov@gmail.com
Doing call_rcu() million times a second becomes a bottle neck.
Convert non-preallocated hash map from call_rcu to SLAB_TYPESAFE_BY_RCU.
The rcu critical section is no longer observed for one htab element
which makes non-preallocated hash map behave just like preallocated hash map.
The map elements are released back to kernel memory after observing
rcu critical section.
This improves 'map_perf_test 4' performance from 100k events per second
to 250k events per second.
bpf_mem_alloc + percpu_counter + typesafe_by_rcu provide 10x performance
boost to non-preallocated hash map and make it within few % of preallocated map
while consuming fraction of memory.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220902211058.60789-8-alexei.starovoitov@gmail.com
The atomic_inc/dec might cause extreme cache line bouncing when multiple cpus
access the same bpf map. Based on specified max_entries for the hash map
calculate when percpu_counter becomes faster than atomic_t and use it for such
maps. For example samples/bpf/map_perf_test is using hash map with max_entries
1000. On a system with 16 cpus the 'map_perf_test 4' shows 14k events per
second using atomic_t. On a system with 15 cpus it shows 100k events per second
using percpu. map_perf_test is an extreme case where all cpus colliding on
atomic_t which causes extreme cache bouncing. Note that the slow path of
percpu_counter is 5k events per secound vs 14k for atomic, so the heuristic is
necessary. See comment in the code why the heuristic is based on
num_online_cpus().
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220902211058.60789-7-alexei.starovoitov@gmail.com
Since bpf hash map was converted to use bpf_mem_alloc it is safe to use
from tracing programs and in RT kernels.
But per-cpu hash map is still using dynamic allocation for per-cpu map
values, hence keep the warning for this map type.
In the future alloc_percpu_gfp can be front-end-ed with bpf_mem_cache
and this restriction will be completely lifted.
perf_event (NMI) bpf programs have to use preallocated hash maps,
because free_htab_elem() is using call_rcu which might crash if re-entered.
Sleepable bpf programs have to use preallocated hash maps, because
life time of the map elements is not protected by rcu_read_lock/unlock.
This restriction can be lifted in the future as well.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220902211058.60789-6-alexei.starovoitov@gmail.com
Make test_maps more stressful with more parallelism in
update/delete/lookup/walk including different value sizes.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220902211058.60789-4-alexei.starovoitov@gmail.com
Tracing BPF programs can attach to kprobe and fentry. Hence they
run in unknown context where calling plain kmalloc() might not be safe.
Front-end kmalloc() with minimal per-cpu cache of free elements.
Refill this cache asynchronously from irq_work.
BPF programs always run with migration disabled.
It's safe to allocate from cache of the current cpu with irqs disabled.
Free-ing is always done into bucket of the current cpu as well.
irq_work trims extra free elements from buckets with kfree
and refills them with kmalloc, so global kmalloc logic takes care
of freeing objects allocated by one cpu and freed on another.
struct bpf_mem_alloc supports two modes:
- When size != 0 create kmem_cache and bpf_mem_cache for each cpu.
This is typical bpf hash map use case when all elements have equal size.
- When size == 0 allocate 11 bpf_mem_cache-s for each cpu, then rely on
kmalloc/kfree. Max allocation size is 4096 in this case.
This is bpf_dynptr and bpf_kptr use case.
bpf_mem_alloc/bpf_mem_free are bpf specific 'wrappers' of kmalloc/kfree.
bpf_mem_cache_alloc/bpf_mem_cache_free are 'wrappers' of kmem_cache_alloc/kmem_cache_free.
The allocators are NMI-safe from bpf programs only. They are not NMI-safe in general.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20220902211058.60789-2-alexei.starovoitov@gmail.com
The commit 7d7672bc5d ("btrfs: convert count_max_extents() to use
fs_info->max_extent_size") introduced a division by
fs_info->max_extent_size. This max_extent_size is initialized with max
zone append limit size of the device btrfs runs on. However, in zone
emulation mode, the device is not zoned then its zone append limit is
zero. This resulted in zero value of fs_info->max_extent_size and caused
zero division error.
Fix the error by setting non-zero pseudo value to max append zone limit
in zone emulation mode. Set the pseudo value based on max_segments as
suggested in the commit c2ae7b772e ("btrfs: zoned: revive
max_zone_append_bytes").
Fixes: 7d7672bc5d ("btrfs: convert count_max_extents() to use fs_info->max_extent_size")
CC: stable@vger.kernel.org # 5.12+
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The commit 2ce543f478 ("btrfs: zoned: wait until zone is finished when
allocation didn't progress") implemented a zone finish waiting mechanism
to the write path of zoned mode. However, using
wait_var_event()/wake_up_all() on fs_info->zone_finish_wait is wrong and
wait_var_event() just hangs because no one ever wakes it up once it goes
into sleep.
Instead, we can simply use wait_on_bit_io() and clear_and_wake_up_bit()
on fs_info->flags with a proper barrier installed.
Fixes: 2ce543f478 ("btrfs: zoned: wait until zone is finished when allocation didn't progress")
CC: stable@vger.kernel.org # 5.16+
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Add 1000BASE-KX interface mode. This 1G backplane ethernet as described in
clause 70. Clause 73 autonegotiation is mandatory, and only full duplex
operation is supported.
Although at the PMA level this interface mode is identical to
1000BASE-X, it uses a different form of in-band autonegation. This
justifies a separate interface mode, since the interface mode (along
with the MLO_AN_* autonegotiation mode) sets the type of autonegotiation
which will be used on a link. This results in more than just electrical
differences between the link modes.
With regard to 1000BASE-X, 1000BASE-KX holds a similar position to
SGMII: same signaling, but different autonegotiation. PCS drivers
(which typically handle in-band autonegotiation) may only support
1000BASE-X, and not 1000BASE-KX. Similarly, the phy mode is used to
configure serdes phys with phy_set_mode_ext. Due to the different
electrical standards (SFI or XFI vs Clause 70), they will likely want to
use different configuration. Adding a phy interface mode for
1000BASE-KX helps simplify configuration in these areas.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Sean Anderson says:
====================
net: dpaa: Cleanups in preparation for phylink conversion (part 2)
This series contains several cleanup patches for dpaa/fman. While they
are intended to prepare for a phylink conversion, they stand on their
own. This series was originally submitted as part of [1].
[1] https://lore.kernel.org/netdev/20220715215954.1449214-1-sean.anderson@seco.com
Changes in v5:
- Reduce line length of tgec_config
- Reduce line length of qman_update_cgr_safe
- Rebase onto net-next/master
Changes in v4:
- weer -> were
- tricy -> tricky
- Use mac_dev for calling change_addr
- qman_cgr_create -> qman_create_cgr
Changes in v2:
- Fix prototype for dtsec_initialization
- Fix warning if sizeof(void *) != sizeof(resource_size_t)
- Specify type of mac_dev for exception_cb
- Add helper for sanity checking cgr ops
- Add CGR update function
- Adjust queue depth on rate change
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of setting the queue depth once during probe, adjust it on the
fly whenever we configure the link. This is a bit unusal, since usually
the DPAA driver calls into the FMAN driver, but here we do the opposite.
We need to add a netdev to struct mac_device for this, but it will soon
live in the phylink config.
I haven't tested this extensively, but it doesn't seem to break
anything. We could possibly optimize this a bit by keeping track of the
last rate, but for now we just update every time. 10GEC probably doesn't
need to call into this at all, but I've added it for consistency.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Acked-by: Camelia Groza <camelia.groza@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This adds a function to update a CGR with new parameters. qman_create_cgr
can almost be used for this (with flags=0), but it's not suitable because
it also registers the callback function. The _safe variant was modeled off
of qman_cgr_delete_safe. However, we handle multiple arguments and a return
value.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Acked-by: Camelia Groza <camelia.groza@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This breaks out/combines get_affine_portal and the cgr sanity check in
preparation for the next commit. No functional change intended.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Acked-by: Camelia Groza <camelia.groza@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There are several references to mac_dev in dpaa_netdev_init. Make things a
bit more concise by adding a local variable for it.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Acked-by: Camelia Groza <camelia.groza@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When disabling, there is nothing we can do about errors. In fact, the
only error which can occur is misuse of the API. Just warn in the mac
driver instead.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Acked-by: Camelia Groza <camelia.groza@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This removes the _return label, since something like
err = -EFOO;
goto _return;
can be replaced by the briefer
return -EFOO;
Additionally, this skips going to _return_of_node_put when dev_node has
already been put (preventing a double put).
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Acked-by: Camelia Groza <camelia.groza@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Instead of using a void pointer for mac_dev, specify its type
explicitly.
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Acked-by: Camelia Groza <camelia.groza@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>