There is a possibility of having just one DMA window available with
a limited capacity which the existing code does not handle that well.
If the window is big enough for the system RAM but less than
MAX_PHYSMEM_BITS (which we want when persistent memory is present),
we create 1:1 window and leave persistent memory without DMA.
This disables 1:1 mapping entirely if there is persistent memory and
either:
- the huge DMA window does not cover the entire address space;
- the default DMA window is removed.
This relies on reverted 54fc3c681ded
("powerpc/pseries/ddw: Extend upper limit for huge DMA window for persistent memory")
to return the actual amount RAM in ddw_memory_hotplug_max() (posted
separately).
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://lore.kernel.org/r/20211108040320.3857636-4-aik@ozlabs.ru Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Paolo Pisati <paolo.pisati@canonical.com>
Fix the length of holes reported at the end of a file: the length is
relative to the beginning of the extent, not the seek position which is
rounded down to the filesystem block size.
This bug went unnoticed for some time, but is now caught by the
following assertion in iomap_iter_done():
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Paolo Pisati <paolo.pisati@canonical.com>
Before this patch, evict would clear the iopen glock's gl_object after
releasing the inode glock. In the meantime, another process could reuse
the same block and thus glocks for a new inode. It would lock the inode
glock (exclusively), and then the iopen glock (shared). The shared
locking mode doesn't provide any ordering against the evict, so by the
time the iopen glock is reused, evict may not have gotten to setting
gl_object to NULL.
Fix that by releasing the iopen glock before the inode glock in
gfs2_evict_inode.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>gl_object Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Paolo Pisati <paolo.pisati@canonical.com>
The recent regression report revealed that the judgment of the
low-latency playback mode based on the runtime->stop_threshold cannot
work reliably at the prepare stage, as sw_params call may happen at
any time, and PCM dmix actually sets it up after the prepare call.
This ended up with the stall of the stream as PCM ack won't be issued
at all.
For addressing this, check the free-wheeling mode again at the PCM
trigger right before starting the stream again, and allow switching to
the non-LL mode at a late stage.
The recent support for the improved low-latency playback mode applied
the SNDRV_PCM_INFO_EXPLICIT_SYNC flag for the target streams, but this
was a slight overkill. The use of the flag above disables effectively
both PCM status and control mmaps, while basically what we want to
track is only about the appl_ptr update.
For less restriction, use a more proper flag,
SNDRV_PCM_INFO_SYNC_APPLPTR instead, which disables only the control
mmap.
The commit d215f63d49da ("ALSA: usb-audio: Check available frames for
the next packet size") introduced the available frame size check, but
the conversion forgot to initialize the temporary variable properly,
and it resulted in a bogus calculation. This patch fixes it.
While draining a stream, ALSA PCM core stops the stream by issuing
snd_pcm_stop() after all data has been sent out. And, at PCM trigger
stop, currently USB-audio driver kills the in-flight URBs explicitly,
then at sync-stop ops, sync with the finish of all remaining URBs.
This might result in a drop of the drained samples as most of
USB-audio devices / hosts allow relatively long in-flight samples (as
a sort of FIFO).
For avoiding the trimming, this patch changes the stream-stop behavior
during PCM draining state. Under that condition, the pending URBs
won't be killed. The leftover in-flight URBs are caught by the
sync-stop operation that shall be performed after the trigger-stop
operation.
This is another attempt to improve further the handling of playback
stream in the low latency mode. The latest workaround in commit 4267c5a8f313 ("ALSA: usb-audio: Work around for XRUN with low latency
playback") revealed that submitting URBs forcibly in advance may
trigger XRUN easily. In the classical mode, this problem was avoided
by practically delaying the submission of the actual data with the
pre-submissions of silent data before triggering the stream start.
But that is exactly what we want to avoid.
Now, in this patch, instead of the previous workaround, we take a
similar approach as used in the implicit feedback mode. The URBs are
queued at the PCM trigger start like before, but we check whether the
buffer has been already filled enough before each submission, and
stop queuing if the data overcomes the threshold. The remaining URBs
are kept in the ready list, and they will be retrieved in the URB
complete callback of other (already queued) URBs. In the complete
callback, we try to fill the data and submit as much as possible
again. When there is no more available in-flight URBs that may handle
the pending data, we'll check in PCM ack callback and submit and
process URBs there in addition. In this way, the amount of in-flight
URBs may vary dynamically and flexibly depending on the available data
without hitting XRUN.
The following things are changed to achieve the behavior above:
* The endpoint prepare callback is changed to return an error code;
when there is no enough data available, it may return -EAGAIN.
Currently only prepare_playback_urb() returns the error.
The evaluation of the available data is a bit messy here; we can't
check with snd_pcm_avail() at the point of prepare callback (as
runtime->status->hwptr hasn't been updated yet), hence we manually
estimate the appl_ptr and compare with the internal hwptr_done to
calculate the available frames.
* snd_usb_endpoint_start() doesn't submit full URBs if the prepare
callback returns -EAGAIN, and puts the remaining URBs to the ready
list for the later submission.
* snd_complete_urb() treats the URBs in the low-latency mode similarly
like the implicit feedback mode, and submissions are done in
(now exported) snd_usb_queue_pending_output_urbs().
* snd_usb_queue_pending_output_urbs() again checks the error value
from the prepare callback. If it's -EAGAIN for the normal stream
(i.e. not implicit feedback mode), we push it back to the ready list
again.
* PCM ack callback is introduced for the playback stream, and it calls
snd_usb_queue_pending_output_urbs() if there is no in-flight URB
while the stream is running. This corresponds to the case where the
system needs the appl_ptr update for re-submitting a new URB.
* snd_usb_queue_pending_output_urbs() and the prepare EP callback
receive in_stream_lock argument, which is a bool flag indicating the
call path from PCM ack. It's needed for avoiding the deadlock of
snd_pcm_period_elapsed() calls.
* Set the new SNDRV_PCM_INFO_EXPLICIT_SYNC flag when the new
low-latency mode is deployed. This assures catching each applptr
update even in the mmap mode.
Fixes: 4267c5a8f313 ("ALSA: usb-audio: Work around for XRUN with low latency playback") Link: https://lore.kernel.org/r/20210929080844.11583-9-tiwai@suse.de Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Paolo Pisati <paolo.pisati@canonical.com>
In theory, stop_urbs() may be called concurrently.
Although we have the state check beforehand, it's safer to apply
ep->lock during the critical list head manipulations.
This is yet more preparation for the upcoming changes.
Extend snd_usb_endpoint_next_packet_size() to check the available
frames and return -EAGAIN if the next packet size is equal or exceeds
the given size. This will be needed for avoiding XRUN during the low
latency operation.
As of this patch, avail=0 is passed, i.e. the check is skipped and no
behavior change.
When a playback stream runs in the implicit feedback mode, its
operation is passive and won't start unless the capture packet is
received. This behavior contradicts with the low-latency playback
mode, and we should turn off lowlatency_playback flag accordingly.
In theory, we may take the low-latency mode when the playback-first
quirk is set, but it still conflicts with the later operation with the
fixed packet numbers, so it's disabled all together for now.
The free-wheel stream operation like dmix may not update the appl_ptr
appropriately, and it doesn't fit with the low-latency playback mode.
Disable the low-latency playback operation when the stream is set up
in such a mode.
This is a preparation patch for the upcoming low-latency improvement
changes.
Rename early_playback_start flag with lowlatency_playback as it's more
intuitive. The new flag is basically a reverse meaning.
Along with the rename, factor out the code to set the flag to a
function. This makes the complex condition checks simpler.
Also, the same flag is introduced to snd_usb_endpoint, too, that is
carried from the snd_usb_substream flag. Currently the endpoint flag
isn't still referred, but will be used in later patches.
When a single clock source is shared among several endpoints, we have
to keep the same rate on all active endpoints as long as the clock is
being used. For dealing with such a case, this patch adds one more
check in the hw params constraint for the rate to take the shared
clocks into account. The current rate is evaluated from the endpoint
list that applies the same clock source.
BugLink: https://bugs.launchpad.net/bugs/1947206
Additional patch to fix the build errors after applying the
ib_peer_memory patch set forward-ported from impish.
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Jason Gunthorpe [Wed, 17 Nov 2021 15:49:00 +0000 (16:49 +0100)]
UBUNTU: SAUCE: RDMA/core: Updated ib_peer_memory
BugLink: https://launchpad.net/bugs/1947206
- Allow clients to opt out of unmap during invalidation
- Fix some bugs in the sequencing of mlx5 MRs
- Enable ATS for peer memory
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
(provided by Nvidia via private email) Signed-off-by: dann frazier <dann.frazier@canonical.com> Acked-by: Stefan Bader <stefan.bader@canonical.com> Acked-by: Tim Gardner <tim.gardner@canonical.com> Signed-off-by: Stefan Bader <stefan.bader@canonical.com> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The peer_memory_client scheme allows a driver to register with the ib_umem
system that it has the ability to understand user virtual address ranges
that are not compatible with get_user_pages(). For instance VMAs created
with io_remap_pfn_range(), or other driver special VMA.
For ranges the interface understands it can provide a DMA mapped sg_table
for use by the ib_umem, allowing user virtual ranges that cannot be
supported by get_user_pages() to be used as umems for RDMA.
This is designed to preserve the kABI, no functions or structures are
changed, only new symbols are added:
This interface is compatible with the two out of tree GPU drivers:
https://github.com/RadeonOpenCompute/ROCK-Kernel-Driver/blob/master/drivers/gpu/drm/amd/amdkfd/kfd_peerdirect.c
https://github.com/Mellanox/nv_peer_memory/blob/master/nv_peer_mem.c
Signed-off-by: Yishai Hadas <yishaih@mellanox.com> Signed-off-by: Feras Daoud <ferasda@mellanox.com> Signed-off-by: Jason Gunthorpe <jgg@mellanox.com> Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
(cherry picked from commit a42989294cf39d6e829424734ab0e7ec48bebcef
git://git.kernel.org/pub/scm/linux/kernel/git/leon/linux-rdma.git) Signed-off-by: dann frazier <dann.frazier@canonical.com>
[ saf: resolve conflicts with updates in 5.12 ] Signed-off-by: Seth Forshee <seth.forshee@canonical.com> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
elevator_init_mq() is only called before adding disk, when there isn't
any FS I/O, only passthrough requests can be queued, so freezing queue
plus canceling dispatch work is enough to drain any dispatch activities,
then we can avoid synchronize_srcu() in blk_mq_quiesce_queue().
Long boot latency issue can be fixed in case of lots of disks added
during booting.
Fixes: 737eb78e82d5 ("block: Delay default elevator initialization") Reported-by: yangerkun <yangerkun@huawei.com> Cc: Damien Le Moal <damien.lemoal@wdc.com> Signed-off-by: Ming Lei <ming.lei@redhat.com> Link: https://lore.kernel.org/r/20211117115502.1600950-1-ming.lei@redhat.com Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
For avoiding to slow down queue destroy, we don't call
blk_mq_quiesce_queue() in blk_cleanup_queue(), instead of delaying to
cancel dispatch work in blk_release_queue().
However, this way has caused kernel oops[1], reported by Changhui. The log
shows that scsi_device can be freed before running blk_release_queue(),
which is expected too since scsi_device is released after the scsi disk
is closed and the scsi_device is removed.
Fixes the issue by canceling blk-mq dispatch work in both blk_cleanup_queue()
and disk_release():
1) when disk_release() is run, the disk has been closed, and any sync
dispatch activities have been done, so canceling dispatch work is enough to
quiesce filesystem I/O dispatch activity.
2) in blk_cleanup_queue(), we only focus on passthrough request, and
passthrough request is always explicitly allocated & freed by
its caller, so once queue is frozen, all sync dispatch activity
for passthrough request has been done, then it is enough to just cancel
dispatch work for avoiding any dispatch activity.
drivers/firmware/arm_scmi/sensors.c:640:28: warning: incorrect type in argument 1 (different base types)
drivers/firmware/arm_scmi/sensors.c:640:28: expected unsigned int [usertype] val
drivers/firmware/arm_scmi/sensors.c:640:28: got restricted __le32 [usertype]
The reason above issue is 'buf->list' has 2,100,000 nodes, occupied cpu lead
to soft lockup.
To solve this issue, we need add schedule point when do while loop in
'__io_remove_buffers'.
After add schedule point we do regression, get follow data.
[ 240.141864] __io_remove_buffers: [1]start ctx=0xffff888170603000 bgid=65533 buf=0xffff8881116fcb00
[ 268.408260] __io_remove_buffers: [1]start ctx=0xffff8881b92d2000 bgid=65533 buf=0xffff888130c83180
[ 275.899234] __io_remove_buffers: [2099199]start ctx=0xffff888170603000 bgid=65533 buf=0xffff8881116fcb00
[ 296.741404] __io_remove_buffers: [1]start ctx=0xffff8881b659c000 bgid=65533 buf=0xffff8881010fe380
[ 305.090059] __io_remove_buffers: [2099199]start ctx=0xffff8881b92d2000 bgid=65533 buf=0xffff888130c83180
[ 325.415746] __io_remove_buffers: [1]start ctx=0xffff8881b92d1000 bgid=65533 buf=0xffff8881a17d8f00
[ 333.160318] __io_remove_buffers: [2099199]start ctx=0xffff8881b659c000 bgid=65533 buf=0xffff8881010fe380
...
Fixes:8bab4c09f24e("io_uring: allow conditional reschedule for intensive iterators") Signed-off-by: Ye Bin <yebin10@huawei.com> Link: https://lore.kernel.org/r/20211122024737.2198530-1-yebin10@huawei.com Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Recent fix to maintain a nosharesock state on the
server struct caused a regression. It updated this
field in the old tcp session, and not the new one.
This caused the multichannel scenario to misbehave.
Fixes: c9f1c19cf7c5 (cifs: nosharesock should not share socket with future sessions) Signed-off-by: Shyam Prasad N <sprasad@microsoft.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
When pid filtering is activated in an instance, all of the events trace
files for that instance has the PID_FILTER flag set. This determines
whether or not pid filtering needs to be done on the event, otherwise the
event is executed as normal.
If pid filtering is enabled when an event is created (via a dynamic event
or modules), its flag is not updated to reflect the current state, and the
events are not filtered properly.
The "used length" reported by calling vhost_add_used() must be the
number of bytes written by the device (using "in" buffers).
In vhost_vsock_handle_tx_kick() the device only reads the guest
buffers (they are all "out" buffers), without writing anything,
so we must pass 0 as "used length" to comply virtio spec.
Fixes: 433fc58e6bf2 ("VSOCK: Introduce vhost_vsock.ko") Cc: stable@vger.kernel.org Reported-by: Halil Pasic <pasic@linux.ibm.com> Suggested-by: Jason Wang <jasowang@redhat.com> Signed-off-by: Stefano Garzarella <sgarzare@redhat.com> Link: https://lore.kernel.org/r/20211122163525.294024-2-sgarzare@redhat.com Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Reviewed-by: Halil Pasic <pasic@linux.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The system will crash if we put an uninitialized iova_domain, this
could happen when an error occurs before initializing the iova_domain
in vdpasim_create().
The messages printed on the initialization of the AMD IOMMUv2 driver
have caused some confusion in the past. Clarify the messages to lower
the confusion in the future.
ceph_statfs currently stuffs the cluster fsid into the f_fsid field.
This was fine when we only had a single filesystem per cluster, but now
that we have multiples we need to use something that will vary between
them.
Change ceph_statfs to xor each 32-bit chunk of the fsid (aka cluster id)
into the lower bits of the statfs->f_fsid. Change the lower bits to hold
the fscid (filesystem ID within the cluster).
That should give us a value that is guaranteed to be unique between
filesystems within a cluster, and should minimize the chance of
collisions between mounts of different clusters.
URL: https://tracker.ceph.com/issues/52812 Reported-by: Sachin Prabhu <sprabhu@redhat.com> Signed-off-by: Jeff Layton <jlayton@kernel.org> Reviewed-by: Xiubo Li <xiubli@redhat.com> Signed-off-by: Ilya Dryomov <idryomov@gmail.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Today, when a new mount is done with nosharesock, we ensure
that we don't select an existing matching session. However,
we don't mark the connection as nosharesock, which means that
those could be shared with future sessions.
Fixed it with this commit. Also printing this info in DebugData.
Signed-off-by: Shyam Prasad N <sprasad@microsoft.com> Reviewed-by: Paulo Alcantara (SUSE) <pc@cjr.nz> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Devicetree source is a description of hardware and hardware has only one
block @20008000 which can be configured either as eMMC or SDHC. Having
two node for different modes is an obscure, unusual and confusing way to
configure it. Instead the board file is supposed to customize the block
to its needs, e.g. to SDHC mode.
This fixes dtbs_check warning:
arch/riscv/boot/dts/microchip/microchip-mpfs-icicle-kit.dt.yaml: sdhc@20008000: $nodename:0: 'sdhc@20008000' does not match '^mmc(@.*)?$'
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com> Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Inconsistent node block will cause a file fail to open or read,
which could make the user process crashes or stucks. Let's mark
SBI_NEED_FSCK flag to trigger a fix at next fsck time. After
unlinking the corrupted file, the user process could regenerate
a new one and work correctly.
Signed-off-by: Weichao Guo <guoweichao@oppo.com> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
This patch changes block_operations() to use trylock, if it fails,
it means there is potential quota data updater, in this condition,
let's flush quota data first and then trylock again to check dirty
status of quota data.
The side effect is: in heavy race condition (e.g. multi quota data
upaters vs quota data flusher), it may decrease the probability of
synchronizing quota data successfully in checkpoint() due to limited
retry time of quota flush.
Reported-by: Yi Zhuang <zhuangyi1@huawei.com> Signed-off-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
When supporting only the .map and .unmap callbacks of iommu_ops,
the IOMMU driver can make assumptions about the size and alignment
used for mappings based on the driver provided pgsize_bitmap. VT-d
previously used essentially PAGE_MASK for this bitmap as any power
of two mapping was acceptably filled by native page sizes.
However, with the .map_pages and .unmap_pages interface we're now
getting page-size and count arguments. If we simply combine these
as (page-size * count) and make use of the previous map/unmap
functions internally, any size and alignment assumptions are very
different.
As an example, a given vfio device assignment VM will often create
a 4MB mapping at IOVA pfn [0x3fe00 - 0x401ff]. On a system that
does not support IOMMU super pages, the unmap_pages interface will
ask to unmap 1024 4KB pages at the base IOVA. dma_pte_clear_level()
will recurse down to level 2 of the page table where the first half
of the pfn range exactly matches the entire pte level. We clear the
pte, increment the pfn by the level size, but (oops) the next pte is
on a new page, so we exit the loop an pop back up a level. When we
then update the pfn based on that higher level, we seem to assume
that the previous pfn value was at the start of the level. In this
case the level size is 256K pfns, which we add to the base pfn and
get a results of 0x7fe00, which is clearly greater than 0x401ff,
so we're done. Meanwhile we never cleared the ptes for the remainder
of the range. When the VM remaps this range, we're overwriting valid
ptes and the VT-d driver complains loudly, as reported by the user
report linked below.
The fix for this seems relatively simple, if each iteration of the
loop in dma_pte_clear_level() is assumed to clear to the end of the
level pte page, then our next pfn should be calculated from level_pfn
rather than our working pfn.
With the submission of iommu driver for RK3568 a subtle bug was
introduced: PAGE_DESC_HI_MASK1 and PAGE_DESC_HI_MASK2 have to be
the other way arround - that leads to random errors, especially when
addresses beyond 32 bit are used.
Fix it.
Fixes: c55356c534aa ("iommu: rockchip: Add support for iommu v2") Signed-off-by: Alex Bee <knaerzche@gmail.com> Tested-by: Peter Geis <pgwipeout@gmail.com> Reviewed-by: Heiko Stuebner <heiko@sntech.de> Tested-by: Dan Johansen <strit@manjaro.org> Reviewed-by: Benjamin Gaignard <benjamin.gaignard@collabora.com> Link: https://lore.kernel.org/r/20211124021325.858139-1-knaerzche@gmail.com Signed-off-by: Joerg Roedel <jroedel@suse.de> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
To hot unplug a CPU, the idle task on that CPU calls a few layers of C
code before finally leaving the kernel. When KASAN is in use, poisoned
shadow is left around for each of the active stack frames, and when
shadow call stacks are in use. When shadow call stacks (SCS) are in use
the task's saved SCS SP is left pointing at an arbitrary point within
the task's shadow call stack.
When a CPU is offlined than onlined back into the kernel, this stale
state can adversely affect execution. Stale KASAN shadow can alias new
stackframes and result in bogus KASAN warnings. A stale SCS SP is
effectively a memory leak, and prevents a portion of the shadow call
stack being used. Across a number of hotplug cycles the idle task's
entire shadow call stack can become unusable.
We previously fixed the KASAN issue in commit:
e1b77c92981a5222 ("sched/kasan: remove stale KASAN poison after hotplug")
... by removing any stale KASAN stack poison immediately prior to
onlining a CPU.
Subsequently in commit:
f1a0a376ca0c4ef1 ("sched/core: Initialize the idle task with preemption disabled")
... the refactoring left the KASAN and SCS cleanup in one-time idle
thread initialization code rather than something invoked prior to each
CPU being onlined, breaking both as above.
We fixed SCS (but not KASAN) in commit:
63acd42c0d4942f7 ("sched/scs: Reset the shadow stack when idle_task_exit")
... but as this runs in the context of the idle task being offlined it's
potentially fragile.
To fix these consistently and more robustly, reset the SCS SP and KASAN
shadow of a CPU's idle task immediately before we online that CPU in
bringup_cpu(). This ensures the idle task always has a consistent state
when it is running, and removes the need to so so when exiting an idle
task.
Whenever any thread is created, dup_task_struct() will give the task a
stack which is free of KASAN shadow, and initialize the task's SCS SP,
so there's no need to specially initialize either for idle thread within
init_idle(), as this was only necessary to handle hotplug cycles.
On x86 this shouldn't happen, which has arch_irq_work_raise().
The test program sets up a perf event with sigtrap set to fire on the
'sched_wakeup' tracepoint, which fired in ttwu_do_wakeup().
This happened because the 'sched_wakeup' tracepoint also takes a task
argument passed on to perf_tp_event(), which is used to deliver the
event to that other task.
Since we cannot deliver synchronous signals to other tasks, skip an event if
perf_tp_event() is targeted at another task and perf_event_attr::sigtrap is
set, which will avoid ever entering perf_sigtrap() for such events.
Fixes: 97ba62b27867 ("perf: Add support for SIGTRAP on perf events") Reported-by: syzbot+663359e32ce6f1a305ad@syzkaller.appspotmail.com Signed-off-by: Marco Elver <elver@google.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/YYpoCOBmC/kJWfmI@elver.google.com Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
There are some inconsistency in the way that the handoff bit is being
handled in readers and writers that lead to a race condition.
Firstly, when a queue head writer set the handoff bit, it will clear
it when the writer is being killed or interrupted on its way out
without acquiring the lock. That is not the case for a queue head
reader. The handoff bit will simply be inherited by the next waiter.
Secondly, in the out_nolock path of rwsem_down_read_slowpath(), both
the waiter and handoff bits are cleared if the wait queue becomes
empty. For rwsem_down_write_slowpath(), however, the handoff bit is
not checked and cleared if the wait queue is empty. This can
potentially make the handoff bit set with empty wait queue.
Worse, the situation in rwsem_down_write_slowpath() relies on wstate,
a variable set outside of the critical section containing the ->count
manipulation, this leads to race condition where RWSEM_FLAG_HANDOFF
can be double subtracted, corrupting ->count.
To make the handoff bit handling more consistent and robust, extract
out handoff bit clearing code into the new rwsem_del_waiter() helper
function. Also, completely eradicate wstate; always evaluate
everything inside the same critical section.
The common function will only use atomic_long_andnot() to clear bits
when the wait queue is empty to avoid possible race condition. If the
first waiter with handoff bit set is killed or interrupted to exit the
slowpath without acquiring the lock, the next waiter will inherit the
handoff bit.
While at it, simplify the trylock for loop in
rwsem_down_write_slowpath() to make it easier to read.
Fixes: 4f23dbc1e657 ("locking/rwsem: Implement lock handoff to prevent lock starvation") Reported-by: Zhenhua Ma <mazhenhua@xiaomi.com> Suggested-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Waiman Long <longman@redhat.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Link: https://lkml.kernel.org/r/20211116012912.723980-1-longman@redhat.com Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The driver doesn't support RX timestamping for non-PTP packets, but it
declares that it does. Restrict the reported RX filters to PTP v2 over
L2 and over L4.
Fixes: 4e3b0468e6d7 ("net: mscc: PTP Hardware Clock (PHC) support") Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The ocelot driver, when asked to timestamp all receiving packets, 1588
v1 or NTP, says "nah, here's 1588 v2 for you".
According to this discussion:
https://patchwork.kernel.org/project/netdevbpf/patch/20211104133204.19757-8-martin.kaistra@linutronix.de/#24577647
drivers that downgrade from a wider request to a narrower response (or
even a response where the intersection with the request is empty) are
buggy, and should return -ERANGE instead. This patch fixes that.
Fixes: 4e3b0468e6d7 ("net: mscc: PTP Hardware Clock (PHC) support") Suggested-by: Richard Cochran <richardcochran@gmail.com> Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Acked-by: Richard Cochran <richardcochran@gmail.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Currently, HNS3 driver doesn't clear the reset flags of components after
successfully executing reset, it causes userspace info of
"Components reset" and "Components not reset" is incorrect.
So fix this problem by clear corresponding reset flag after reset process.
Fixes: ddccc5e368a3 ("net: hns3: add support for triggering reset by ethtool") Signed-off-by: Jie Wang <wangjie125@huawei.com> Signed-off-by: Guangbin Huang <huangguangbin2@huawei.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
When PF is set to multi-TCs and configured mapping relationship between
priorities and TCs, the hardware will active these settings for this PF
and its VFs.
In this case when VF just uses one TC and its rx packets contain priority,
and if the priority is not mapped to TC0, as other TCs of VF is not valid,
hardware always put this kind of packets to the queue 0. It cause this kind
of packets of VF can not be used RSS function.
To fix this problem, set tc mode of all unused TCs of VF to the setting of
TC0, then rx packet with priority which map to unused TC will be direct to
TC0.
When applications call shutdown() with SHUT_RDWR in userspace,
smc_close_active() calls kernel_sock_shutdown(), and it is called
twice in smc_shutdown().
This fixes this by checking sk_state before do clcsock shutdown, and
avoids missing the application's call of smc_shutdown().
Inject error before dev_hold(real_dev) in register_vlan_dev(),
and execute the following testcase:
ip link add dev dummy1 type dummy
ip link add name dummy1.100 link dummy1 type vlan id 100
ip link del dev dummy1
When the dummy netdevice is removed, we will get a WARNING as following:
=======================================================================
refcount_t: decrement hit 0; leaking memory.
WARNING: CPU: 2 PID: 0 at lib/refcount.c:31 refcount_warn_saturate+0xbf/0x1e0
and an endless loop of:
=======================================================================
unregister_netdevice: waiting for dummy1 to become free. Usage count = -1073741824
That is because dev_put(real_dev) in vlan_dev_free() be called without
dev_hold(real_dev) in register_vlan_dev(). It makes the refcnt of real_dev
underflow.
Move the dev_hold(real_dev) to vlan_dev_init() which is the call-back of
ndo_init(). That makes dev_hold() and dev_put() for vlan's real_dev
symmetrical.
Fixes: 563bcbae3ba2 ("net: vlan: fix a UAF in vlan_dev_real_dev()") Reported-by: Petr Machata <petrm@nvidia.com> Suggested-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Ziyang Xuan <william.xuanziyang@huawei.com> Link: https://lore.kernel.org/r/20211126015942.2918542-1-william.xuanziyang@huawei.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
ethtool_set_coalesce() now uses both the .get_coalesce() and
.set_coalesce() callbacks. But the check for their availability is
buggy, so changing the coalesce settings on a device where the driver
provides only _one_ of the callbacks results in a NULL pointer
dereference instead of an -EOPNOTSUPP.
Fix the condition so that the availability of both callbacks is
ensured. This also matches the netlink code.
Note that reproducing this requires some effort - it only affects the
legacy ioctl path, and needs a specific combination of driver options:
- have .get_coalesce() and .coalesce_supported but no
.set_coalesce(), or
- have .set_coalesce() but no .get_coalesce(). Here eg. ethtool doesn't
cause the crash as it first attempts to call ethtool_get_coalesce()
and bails out on error.
Fixes: f3ccfda19319 ("ethtool: extend coalesce setting uAPI with CQE mode") Cc: Yufeng Mo <moyufeng@huawei.com> Cc: Huazhong Tan <tanhuazhong@huawei.com> Cc: Andrew Lunn <andrew@lunn.ch> Cc: Heiner Kallweit <hkallweit1@gmail.com> Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com> Link: https://lore.kernel.org/r/20211126175543.28000-1-jwi@linux.ibm.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
when the number of DRR classes decreases, the round-robin active list can
contain elements that have already been freed in ets_qdisc_change(). As a
consequence, it's possible to see a NULL dereference crash, caused by the
attempt to call cl->qdisc->ops->peek(cl->qdisc) when cl->qdisc is NULL:
v3: fix race between ets_qdisc_change() and ets_qdisc_dequeue() delisting
DRR classes beyond 'nbands' in ets_qdisc_change() with the qdisc lock
acquired, thanks to Cong Wang.
v2: when a NULL qdisc is found in the DRR active list, try to dequeue skb
from the next list item.
Reported-by: Hangbin Liu <liuhangbin@gmail.com> Fixes: dcc68b4d8084 ("net: sch_ets: Add a new Qdisc") Signed-off-by: Davide Caratti <dcaratti@redhat.com> Link: https://lore.kernel.org/r/7a5c496eed2d62241620bdbb83eb03fb9d571c99.1637762721.git.dcaratti@redhat.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The Tx queues were not disabled in situations where the driver needed to
stop the interface to apply a new configuration. This could result in a
kernel panic when doing any of the 3 following actions:
* reconfiguring the number of queues (ethtool -L)
* reconfiguring the size of the ring buffers (ethtool -G)
* installing/removing an XDP program (ip l set dev ethX xdp)
Prevent the panic by making sure netif_tx_disable is called when stopping
an interface.
Without this patch, the following kernel panic can be observed when doing
any of the actions above:
Unable to handle kernel paging request at virtual address ffff80001238d040
[....]
Call trace:
dwmac4_set_addr+0x8/0x10
dev_hard_start_xmit+0xe4/0x1ac
sch_direct_xmit+0xe8/0x39c
__dev_queue_xmit+0x3ec/0xaf0
dev_queue_xmit+0x14/0x20
[...]
[ end trace 0000000000000002 ]---
Fixes: 5fabb01207a2d ("net: stmmac: Add initial XDP support") Fixes: aa042f60e4961 ("net: stmmac: Add support to Ethtool get/set ring parameters") Fixes: 0366f7e06a6be ("net: stmmac: add ethtool support for get/set channels") Signed-off-by: Yannick Vignon <yannick.vignon@nxp.com> Link: https://lore.kernel.org/r/20211124154731.1676949-1-yannick.vignon@oss.nxp.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
We replace proto_ops whenever TLS is configured for RX. But our
replacement also overrides sendpage_locked, which will crash
unless TX is also configured. Similarly we plug both of those
in for TLS_HW (NIC crypto offload) even tho TLS_HW has a completely
different implementation for TX.
Last but not least we always plug in something based on inet_stream_ops
even though a few of the callbacks differ for IPv6 (getname, release,
bind).
Use a callback building method similar to what we do for struct proto.
Fixes: c46234ebb4d1 ("tls: RX path for ktls") Fixes: d4ffb02dee2f ("net/tls: enable sk_msg redirect to tls socket egress") Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
recvmsg() will put peek()ed and partially read records onto the rx_list.
splice_read() needs to consult that list otherwise it may miss data.
Align with recvmsg() and also put partially-read records onto rx_list.
tls_sw_advance_skb() is pretty pointless now and will be removed in
net-next.
Fixes: 692d7b5d1f91 ("tls: Fix recvmsg() to be able to peek across multiple records") Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
We don't support splicing control records. TLS 1.3 changes moved
the record type check into the decrypt if(). The skb may already
be decrypted and still be an alert.
Note that decrypt_skb_update() is idempotent and updates ctx->decrypted
so the if() is pointless.
Reorder the check for decryption errors with the content type check
while touching them. This part is not really a bug, because if
decryption failed in TLS 1.3 content type will be DATA, and for
TLS 1.2 it will be correct. Nevertheless its strange to touch output
before checking if the function has failed.
Fixes: fedf201e1296 ("net: tls: Refactor control message handling on recv") Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
It hangup when booting Loongson 3A1000 with BOTH
CONFIG_PAGE_SIZE_64KB and CONFIG_MIPS_VA_BITS_48, that it turn
out to use 2-level pgtable instead of 3-level. 64KB page size
with 2-level pgtable only cover 42 bits VA, use 3-level pgtable
to cover all 48 bits VA(55 bits)
Fixes: 1e321fa917fb ("MIPS64: Support of at least 48 bits of SEGBITS) Signed-off-by: Huang Pei <huangpei@loongson.cn> Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
It turns out that 'decode_configs' -> 'set_ftlb_enable' is called under
c->cputype unset, which leaves FTLB disabled on BOTH 3A2000 and 3A3000
Fix it by calling "decode_configs" after c->cputype is initialized
Fixes: da1bd29742b1 ("MIPS: Loongson64: Probe CPU features via CPUCFG") Signed-off-by: Huang Pei <huangpei@loongson.cn> Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Oleksandr brought a bug report where netpoll causes trace
messages in the log on igb.
Danielle brought this back up as still occurring, so we'll try
again.
[22038.710800] ------------[ cut here ]------------
[22038.710801] igb_poll+0x0/0x1440 [igb] exceeded budget in poll
[22038.710802] WARNING: CPU: 12 PID: 40362 at net/core/netpoll.c:155 netpoll_poll_dev+0x18a/0x1a0
As Alex suggested, change the driver to return work_done at the
exit of napi_poll, which should be safe to do in this driver
because it is not polling multiple queues in this single napi
context (multiple queues attached to one MSI-X vector). Several
other drivers contain the same simple sequence, so I hope
this will not create new problems.
Fixes: 16eb8815c235 ("igb: Refactor clean_rx_irq to reduce overhead and improve performance") Reported-by: Oleksandr Natalenko <oleksandr@natalenko.name> Reported-by: Danielle Ratson <danieller@nvidia.com> Suggested-by: Alexander Duyck <alexander.duyck@gmail.com> Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com> Tested-by: Oleksandr Natalenko <oleksandr@natalenko.name> Tested-by: Danielle Ratson <danieller@nvidia.com> Link: https://lore.kernel.org/r/20211123204000.1597971-1-jesse.brandeburg@intel.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The kernel_listen function in smc_listen will fail when all the available
ports are occupied. At this point smc->clcsock->sk->sk_data_ready has
been changed to smc_clcsock_data_ready. When we call smc_listen again,
now both smc->clcsock->sk->sk_data_ready and smc->clcsk_data_ready point
to the smc_clcsock_data_ready function.
The smc_clcsock_data_ready() function calls lsmc->clcsk_data_ready which
now points to itself resulting in an infinite loop.
This patch restores smc->clcsock->sk->sk_data_ready with the old value.
Fixes: a60a2b1e0af1 ("net/smc: reduce active tcp_listen workers") Signed-off-by: Guo DaXing <guodaxing@huawei.com> Acked-by: Tony Lu <tonylu@linux.alibaba.com> Signed-off-by: Karsten Graul <kgraul@linux.ibm.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Coverity reports a possible NULL dereferencing problem:
in smc_vlan_by_tcpsk():
6. returned_null: netdev_lower_get_next returns NULL (checked 29 out of 30 times).
7. var_assigned: Assigning: ndev = NULL return value from netdev_lower_get_next.
1623 ndev = (struct net_device *)netdev_lower_get_next(ndev, &lower);
CID 1468509 (#1 of 1): Dereference null return value (NULL_RETURNS)
8. dereference: Dereferencing a pointer that might be NULL ndev when calling is_vlan_dev.
1624 if (is_vlan_dev(ndev)) {
Remove the manual implementation and use netdev_walk_all_lower_dev() to
iterate over the lower devices. While on it remove an obsolete function
parameter comment.
Fixes: cb9d43f67754 ("net/smc: determine vlan_id of stacked net_device") Suggested-by: Julian Wiedmann <jwi@linux.ibm.com> Signed-off-by: Karsten Graul <kgraul@linux.ibm.com> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
On mv88e6xxx 1G/2.5G PCS, the SerDes register 4.2001.2 has the following
description:
This register bit indicates when link was lost since the last
read. For the current link status, read this register
back-to-back.
Thus to get current link state, we need to read the register twice.
But doing that in the link change interrupt handler would lead to
potentially ignoring link down events, which we really want to avoid.
Thus this needs to be solved in phylink's resolve, by retriggering
another resolve in the event when PCS reports link down and previous
link was up, and by re-reading PCS state if the previous link was down.
The wrong value is read when phylink requests change from sgmii to
2500base-x mode, and link won't come up. This fixes the bug.
Fixes: 9525ae83959b ("phylink: add phylink infrastructure") Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Marek Behún <kabel@kernel.org> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
On PHY state change the phylink_resolve() function can read stale
information from the MAC and report incorrect link speed and duplex to
the kernel message log.
Example with a Marvell 88X3310 PHY connected to a SerDes port on Marvell
88E6393X switch:
- PHY driver triggers state change due to PHY interface mode being
changed from 10gbase-r to 2500base-x due to copper change in speed
from 10Gbps to 2.5Gbps, but the PHY itself either hasn't yet changed
its interface to the host, or the interrupt about loss of SerDes link
hadn't arrived yet (there can be a delay of several milliseconds for
this), so we still think that the 10gbase-r mode is up
- phylink_resolve()
- phylink_mac_pcs_get_state()
- this fills in speed=10g link=up
- interface mode is updated to 2500base-x but speed is left at 10Gbps
- phylink_major_config()
- interface is changed to 2500base-x
- phylink_link_up()
- mv88e6xxx_mac_link_up()
- .port_set_speed_duplex()
- speed is set to 10Gbps
- reports "Link is Up - 10Gbps/Full" to dmesg
Afterwards when the interrupt finally arrives for mv88e6xxx, another
resolve is forced in which we get the correct speed from
phylink_mac_pcs_get_state(), but since the interface is not being
changed anymore, we don't call phylink_major_config() but only
phylink_mac_config(), which does not set speed/duplex anymore.
To fix this, we need to force the link down and trigger another resolve
on PHY interface change event.
Fixes: 9525ae83959b ("phylink: add phylink infrastructure") Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Marek Behún <kabel@kernel.org> Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Usage of phy_ethtool_get_link_ksettings() in the link status change
handler isn't needed, and in combination with the referenced change
it results in a deadlock. Simply remove the call and replace it with
direct access to phydev->speed. The duplex argument of
lan743x_phy_update_flowcontrol() isn't used and can be removed.
Fixes: c10a485c3de5 ("phy: phy_ethtool_ksettings_get: Lock the phy for consistency") Reported-by: Alessandro B Maurici <abmaurici@gmail.com> Tested-by: Alessandro B Maurici <abmaurici@gmail.com> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Link: https://lore.kernel.org/r/40e27f76-0ba3-dcef-ee32-a78b9df38b0f@gmail.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
While testing BIG TCP patch series, I was expecting that TCP_RR workloads
with 80KB requests/answers would send one 80KB TSO packet,
then being received as a single GRO packet.
It turns out this was not happening, and the root cause was that
cubic Hystart ACK train was triggering after a few (2 or 3) rounds of RPC.
Hystart was wrongly setting CWND/SSTHRESH to 30, while my RPC
needed a budget of ~20 segments.
Ideally these TCP_RR flows should not exit slow start.
Cubic Hystart should reset itself at each round, instead of assuming
every TCP flow is a bulk one.
Note that even after this patch, Hystart can still trigger, depending
on scheduling artifacts, but at a higher CWND/SSTHRESH threshold,
keeping optimal TSO packet sizes.
Tested:
ip link set dev eth0 gro_ipv6_max_size 131072 gso_ipv6_max_size 131072
nstat -n; netperf -H ... -t TCP_RR -l 5 -- -r 80000,80000 -K cubic; nstat|egrep "Ip6InReceives|Hystart|Ip6OutRequests"
[Why]
The HW interrupt gets disabled after GPU reset so we don't receive
notifications for HPD or AUX from DMUB - leading to timeout and
black screen with (or without) DPIA links connected.
[How]
Re-enable the interrupt after GPU reset like we do for the other
DC interrupts.
Fixes: 81927e2808be ("drm/amd/display: Support for DMUB AUX") Reviewed-by: Jude Shih <Jude.Shih@amd.com> Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com> Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Commit 39fbef4b0f77 ("PM: hibernate: Get block device exclusively in
swsusp_check()") changed the opening mode of the block device to
(FMODE_READ | FMODE_EXCL).
In the corresponding calls to swsusp_close(), the mode is still just
FMODE_READ which triggers the warning in blkdev_flush_mapping() on
resume from hibernate.
So, use the mode (FMODE_READ | FMODE_EXCL) also when closing the
device.
Fixes: 39fbef4b0f77 ("PM: hibernate: Get block device exclusively in swsusp_check()") Signed-off-by: Thomas Zeitlhofer <thomas.zeitlhofer+lkml@ze-it.at> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Update NC-SI command handler (both standard and OEM) to take into
account of payload paddings in allocating skb (in case of payload
size is not 32-bit aligned).
The checksum field follows payload field, without taking payload
padding into account can cause checksum being truncated, leading to
dropped packets.
Fixes: fb4ee67529ff ("net/ncsi: Add NCSI OEM command support") Signed-off-by: Kumar Thangavel <thangavel.k@hcl.com> Acked-by: Samuel Mendoza-Jonas <sam@mendozajonas.com> Reviewed-by: Paul Menzel <pmenzel@molgen.mpg.de> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The put_user() in schedule_tail() can get stuck in a livelock, similar
to a problem recently fixed on riscv in commit:
285a76bb2cf51b0c ("riscv: evaluate put_user() arg before enabling user access")
In __raw_put_user() we have a critical section between
uaccess_ttbr0_enable() and uaccess_ttbr0_disable() where we cannot
safely call into the scheduler without having taken an exception, as
schedule() and other scheduling functions will not save/restore the
TTBR0 state. If either of the `x` or `ptr` arguments to __raw_put_user()
contain a blocking call, we may call into the scheduler within the
critical section. This can result in two problems:
1) The access within the critical section will occur without the
required TTBR0 tables installed. This will fault, and where the
required tables permit access, the access will be retried without the
required tables, resulting in a livelock.
2) When TTBR0 SW PAN is in use, check_and_switch_context() does not
modify TTBR0, leaving a stale value installed. The mappings of the
blocked task will erroneously be accessible to regular accesses in
the context of the new task. Additionally, if the tables are
subsequently freed, local TLB maintenance required to reuse the ASID
may be lost, potentially resulting in TLB corruption (e.g. in the
presence of CnP).
The same issue exists for __raw_get_user() in the critical section
between uaccess_ttbr0_enable() and uaccess_ttbr0_disable().
A similar issue exists for __get_kernel_nofault() and
__put_kernel_nofault() for the critical section between
__uaccess_enable_tco_async() and __uaccess_disable_tco_async(), as the
TCO state is not context-switched by direct calls into the scheduler.
Here the TCO state may be lost from the context of the current task,
resulting in unexpected asynchronous tag check faults. It may also be
leaked to another task, suppressing expected tag check faults.
To fix all of these cases, we must ensure that we do not directly call
into the scheduler in their respective critical sections. This patch
reworks __raw_put_user(), __raw_get_user(), __get_kernel_nofault(), and
__put_kernel_nofault(), ensuring that parameters are evaluated outside
of the critical sections. To make this requirement clear, comments are
added describing the problem, and line spaces added to separate the
critical sections from other portions of the macros.
For __raw_get_user() and __raw_put_user() the `err` parameter is
conditionally assigned to, and we must currently evaluate this in the
critical section. This behaviour is relied upon by the signal code,
which uses chains of put_user_error() and get_user_error(), checking the
return value at the end. In all cases, the `err` parameter is a plain
int rather than a more complex expression with a blocking call, so this
is safe.
In future we should try to clean up the `err` usage to remove the
potential for this to be a problem.
Aside from the changes to time of evaluation, there should be no
functional change as a result of this patch.
The Hyper-V DRM driver tries to free MMIO region on removing
the device regardless of VM type, while Gen1 VMs don't use MMIO
and hence causing the kernel to crash on a NULL pointer dereference.
Fix this by making deallocating MMIO only on Gen2 machines and implement
removal for Gen1
Fixes: 76c56a5affeb ("drm/hyperv: Add DRM driver for hyperv synthetic video device") Signed-off-by: Mohammed Gamal <mgamal@redhat.com> Reviewed-by: Deepak Rawat <drawat.floss@gmail.com> Signed-off-by: Deepak Rawat <drawat.floss@gmail.com> Link: https://patchwork.freedesktop.org/patch/msgid/20211119112900.300537-1-mgamal@redhat.com Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Current nvmet_try_send_ddgst() code does not check whether
all data digest bytes are transmitted, fix this by returning
-EAGAIN if all data digest bytes are not transmitted.
Fixes: 872d26a391da ("nvmet-tcp: add NVMe over TCP target driver") Signed-off-by: Varun Prakash <varun@chelsio.com> Reviewed-by: Sagi Grimberg <sagi@grimberg.me> Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Commit fbdc21e9b038 ("cpufreq: intel_pstate: Add Icelake servers
support in no-HWP mode") enabled the use of Intel P-State driver
for Ice Lake servers.
But it doesn't cover the case when OS can't control P-States.
Therefore, for Ice Lake server, if MSR_MISC_PWR_MGMT bits 8 or 18
are enabled, then the Intel P-State driver should exit as OS can't
control P-States.
Fixes: fbdc21e9b038 ("cpufreq: intel_pstate: Add Icelake servers support in no-HWP mode") Signed-off-by: Adamos Ttofari <attofari@amazon.de> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Currently mvpp2_xdp_setup won't allow attaching XDP program if
mtu > ETH_DATA_LEN (1500).
The mvpp2_change_mtu on the other hand checks whether
MVPP2_RX_PKT_SIZE(mtu) > MVPP2_BM_LONG_PKT_SIZE.
These two checks are semantically different.
Moreover this limit can be increased to MVPP2_MAX_RX_BUF_SIZE, since in
mvpp2_rx we have
xdp.data = data + MVPP2_MH_SIZE + MVPP2_SKB_HEADROOM;
xdp.frame_sz = PAGE_SIZE;
Change the checks to check whether
mtu > MVPP2_MAX_RX_BUF_SIZE
Fixes: 07dd0a7aae7f ("mvpp2: add basic XDP support") Signed-off-by: Marek Behún <kabel@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Calling ipa_cmd_pipeline_clear() after stopping the channel
underlying the AP<-modem RX endpoint can lead to a deadlock.
This occurs in the ->runtime_suspend device power operation for the
IPA driver. While this callback is in progress, any other requests
for power will block until the callback returns.
Stopping the AP<-modem RX channel does not prevent the modem from
sending another packet to this endpoint. If a packet arrives for an
RX channel when the channel is stopped, an SUSPEND IPA interrupt
condition will be pending. Handling an IPA interrupt requires
power, so ipa_isr_thread() calls pm_runtime_get_sync() first thing.
The problem occurs because a "pipeline clear" command will not
complete while such a SUSPEND interrupt condition exists. So the
SUSPEND IPA interrupt handler won't proceed until it gets power;
that won't happen until the ->runtime_suspend callback (and its
"pipeline clear" command) completes; and that can't happen while
the SUSPEND interrupt condition exists.
It turns out that in this case there is no need to use the "pipeline
clear" command. There are scenarios in which clearing the pipeline
is required while suspending, but those are not (yet) supported
upstream. So a simple fix, avoiding the potential deadlock, is to
stop calling ipa_cmd_pipeline_clear() in ipa_endpoint_suspend().
This removes the only user of ipa_cmd_pipeline_clear(), so get rid
of that function. It can be restored again whenever it's needed.
This is basically a manual revert along with an explanation for
commit 6cb63ea6a39ea ("net: ipa: introduce ipa_cmd_tag_process()").
Fixes: 6cb63ea6a39ea ("net: ipa: introduce ipa_cmd_tag_process()") Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The IPA setup_complete flag is set at the end of ipa_setup(), when
the setup phase of initialization has completed successfully. This
occurs as part of driver probe processing, or (if "modem-init" is
specified in the DTS file) it is triggered by the "ipa-setup-ready"
SMP2P interrupt generated by the modem.
In the latter case, it's possible for driver shutdown (or remove) to
begin while setup processing is underway, and this can't be allowed.
The problem is that the setup_complete flag is not adequate to signal
that setup is underway.
If setup_complete is set, it will never be un-set, so that case is
not a problem. But if setup_complete is false, there's a chance
setup is underway.
Because setup is triggered by an interrupt on a "modem-init" system,
there is a simple way to ensure the value of setup_complete is safe
to read. The threaded handler--if it is executing--will complete as
part of a request to disable the "ipa-modem-ready" interrupt. This
means that ipa_setup() (which is called from the handler) will run
to completion if it was underway, or will never be called otherwise.
The request to disable the "ipa-setup-ready" interrupt is currently
made within ipa_modem_stop(). Instead, disable the interrupt
outside that function in the two places it's called. In the case of
ipa_remove(), this ensures the setup_complete flag is safe to read
before we read it.
Rename ipa_smp2p_disable() to be ipa_smp2p_irq_disable_setup(), to be
more specific about its effect.
Fixes: 530f9216a953 ("soc: qcom: ipa: AP/modem communications") Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
We currently maintain a "disabled" Boolean flag to determine whether
the "ipa-setup-ready" SMP2P IRQ handler does anything. That flag
must be accessed under protection of a mutex.
Instead, disable the SMP2P interrupt when requested, which prevents
the interrupt handler from ever being called. More importantly, it
synchronizes a thread disabling the interrupt with the completion of
the interrupt handler in case they run concurrently.
Use the IPA setup_complete flag rather than the disabled flag in the
handler to determine whether to ignore any interrupts arriving after
the first.
Rename the "disabled" flag to be "setup_disabled", to be specific
about its purpose.
Fixes: 530f9216a953 ("soc: qcom: ipa: AP/modem communications") Signed-off-by: Alex Elder <elder@linaro.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
When processing port up/down events generated by the device's firmware,
the driver protects itself from events reported for non-existent local
ports, but not the CPU port (local port 0), which exists, but lacks a
netdev.
This can result in a NULL pointer dereference when calling
netif_carrier_{on,off}().
Fix this by bailing early when processing an event reported for the CPU
port. Problem was only observed when running on top of a buggy emulator.
Fixes: 28b1987ef506 ("mlxsw: spectrum: Register CPU port with devlink") Signed-off-by: Amit Cohen <amcohen@nvidia.com> Signed-off-by: Ido Schimmel <idosch@nvidia.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The side that actively closed socket, it's clcsock doesn't enter
TIME_WAIT state, but the passive side does it. It should show the same
behavior as TCP sockets.
Consider this, when client actively closes the socket, the clcsock in
server enters TIME_WAIT state, which means the address is occupied and
won't be reused before TIME_WAIT dismissing. If we restarted server, the
service would be unavailable for a long time.
To solve this issue, shutdown the clcsock in [A], perform the TCP active
close progress first, before the passive closed side closing it. So that
the actively closed side enters TIME_WAIT, not the passive one.
Client | Server
close() // client actively close |
smc_release() |
smc_close_active() // PEERCLOSEWAIT1 |
smc_close_final() // abort or closed = 1|
smc_cdc_get_slot_and_msg_send() |
[A] |
|smc_cdc_msg_recv_action() // ACTIVE
| queue_work(smc_close_wq, &conn->close_work)
| smc_close_passive_work() // PROCESSABORT or APPCLOSEWAIT1
| smc_close_passive_abort_received() // only in abort
|
|close() // server recv zero, close
| smc_release() // PROCESSABORT or APPCLOSEWAIT1
| smc_close_active()
| smc_close_abort() or smc_close_final() // CLOSED
| smc_cdc_get_slot_and_msg_send() // abort or closed = 1
smc_cdc_msg_recv_action() | smc_clcsock_release()
queue_work(smc_close_wq, &conn->close_work) | sock_release(tcp) // actively close clc, enter TIME_WAIT
smc_close_passive_work() // PEERCLOSEWAIT1 | smc_conn_free()
smc_close_passive_abort_received() // CLOSED|
smc_conn_free() |
smc_clcsock_release() |
sock_release(tcp) // passive close clc |
Link: https://www.spinics.net/lists/netdev/msg780407.html Fixes: b38d732477e4 ("smc: socket closing and linkgroup cleanup") Signed-off-by: Tony Lu <tonylu@linux.alibaba.com> Reviewed-by: Wen Gu <guwen@linux.alibaba.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
If a timeout is hit, it can result is incorrect data on the I2C bus
and/or memory corruptions in the guest since the device can still be
operating on the buffers it was given while the guest has freed them.
Here is, for example, the start of a slub_debug splat which was
triggered on the next transfer after one transfer was forced to timeout
by setting a breakpoint in the backend (rust-vmm/vhost-device):
BUG kmalloc-1k (Not tainted): Poison overwritten
First byte 0x1 instead of 0x6b
Allocated in virtio_i2c_xfer+0x65/0x35c age=350 cpu=0 pid=29
__kmalloc+0xc2/0x1c9
virtio_i2c_xfer+0x65/0x35c
__i2c_transfer+0x429/0x57d
i2c_transfer+0x115/0x134
i2cdev_ioctl_rdwr+0x16a/0x1de
i2cdev_ioctl+0x247/0x2ed
vfs_ioctl+0x21/0x30
sys_ioctl+0xb18/0xb41
Freed in virtio_i2c_xfer+0x32e/0x35c age=244 cpu=0 pid=29
kfree+0x1bd/0x1cc
virtio_i2c_xfer+0x32e/0x35c
__i2c_transfer+0x429/0x57d
i2c_transfer+0x115/0x134
i2cdev_ioctl_rdwr+0x16a/0x1de
i2cdev_ioctl+0x247/0x2ed
vfs_ioctl+0x21/0x30
sys_ioctl+0xb18/0xb41
There is no simple fix for this (the driver would have to always create
bounce buffers and hold on to them until the device eventually returns
the buffers), so just disable the timeout support for now.
Fixes: 3cfc88380413d20f ("i2c: virtio: add a virtio i2c frontend driver") Acked-by: Jie Deng <jie.deng@intel.com> Signed-off-by: Vincent Whitchurch <vincent.whitchurch@axis.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Wolfram Sang <wsa@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
We observed the following deadlock in the stress test under low
memory scenario:
Thread A Thread B
- erofs_shrink_scan
- erofs_try_to_release_workgroup
- erofs_workgroup_try_to_freeze -- A
- z_erofs_do_read_page
- z_erofs_collection_begin
- z_erofs_register_collection
- erofs_insert_workgroup
- xa_lock(&sbi->managed_pslots) -- B
- erofs_workgroup_get
- erofs_wait_on_workgroup_freezed -- A
- xa_erase
- xa_lock(&sbi->managed_pslots) -- B
To fix this, it needs to hold xa_lock before freezing the workgroup
since xarray will be touched then. So let's hold the lock before
accessing each workgroup, just like what we did with the radix tree
before.
When a reset is requested the position of the write pointer is updated but
the data in the corresponding zone is not cleared. Instead scsi_debug
returns any data written before the write pointer was reset. This is an
error and prevents using scsi_debug for stale page cache testing of the
BLKRESETZONE ioctl.
Zero written data in the zone when resetting the write pointer.
Link: https://lore.kernel.org/r/20211122061223.298890-1-shinichiro.kawasaki@wdc.com Fixes: f0d1cf9378bd ("scsi: scsi_debug: Add ZBC zone commands") Reviewed-by: Damien Le Moal <damien.lemoal@opensource.wdc.com> Acked-by: Douglas Gilbert <dgilbert@interlog.com> Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
This fixes an issue added in commit 4edd8cd4e86d ("scsi: core: sysfs: Fix
hang when device state is set via sysfs") where if userspace is requesting
to set the device state to SDEV_RUNNING when the state is already
SDEV_RUNNING, we return -EINVAL instead of count. The commmit above set ret
to count for this case, when it should have set it to 0.
Link: https://lore.kernel.org/r/20211120164917.4924-1-michael.christie@oracle.com Fixes: 4edd8cd4e86d ("scsi: core: sysfs: Fix hang when device state is set via sysfs") Reviewed-by: Lee Duncan <lduncan@suse.com> Signed-off-by: Mike Christie <michael.christie@oracle.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Ice driver has the routines for managing XDP resources that are shared
between ndo_bpf op and VSI rebuild flow. The latter takes place for
example when user changes queue count on an interface via ethtool's
set_channels().
There is an issue around the bpf_prog refcounting when VSI is being
rebuilt - since ice_prepare_xdp_rings() is called with vsi->xdp_prog as
an argument that is used later on by ice_vsi_assign_bpf_prog(), same
bpf_prog pointers are swapped with each other. Then it is also
interpreted as an 'old_prog' which in turn causes us to call
bpf_prog_put on it that will decrement its refcount.
Below splat can be interpreted in a way that due to zero refcount of a
bpf_prog it is wiped out from the system while kernel still tries to
refer to it:
Fix this by only calling ice_vsi_assign_bpf_prog() inside
ice_prepare_xdp_rings() when current vsi->xdp_prog pointer is NULL.
This way set_channels() flow will not attempt to swap the vsi->xdp_prog
pointers with itself.
Also, sprinkle around some comments that provide a reasoning about
correlation between driver and kernel in terms of bpf_prog refcount.
Fixes: efc2214b6047 ("ice: Add support for XDP") Reviewed-by: Alexander Lobakin <alexandr.lobakin@intel.com> Signed-off-by: Marta Plantykow <marta.a.plantykow@intel.com> Co-developed-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com> Signed-off-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com> Tested-by: Kiran Bhandare <kiranx.bhandare@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The approach of having XDP queue per CPU regardless of user's setting
exposed a hidden bug that could occur in case when Rx queue count differ
from Tx queue count. Currently vsi->txq_map's size is equal to the
doubled vsi->alloc_txq, which is not correct due to the fact that XDP
rings were previously based on the Rx queue count. Below splat can be
seen when ethtool -L is used and XDP rings are configured:
When replacing a nexthop group, we must release the IPv6 per-cpu dsts of
the removed nexthop entries after an RCU grace period because they
contain references to the nexthop's net device and to the fib6 info.
With specific series of events[1] we can reach net device refcount
imbalance which is unrecoverable. IPv4 is not affected because dsts
don't take a refcount on the route.
[1]
$ ip nexthop list
id 200 via 2002:db8::2 dev bridge.10 scope link onlink
id 201 via 2002:db8::3 dev bridge scope link onlink
id 203 group 201/200
$ ip -6 route
2001:db8::10 nhid 203 metric 1024 pref medium
nexthop via 2002:db8::3 dev bridge weight 1 onlink
nexthop via 2002:db8::2 dev bridge.10 weight 1 onlink
Create rt6_info through one of the multipath legs, e.g.:
$ taskset -a -c 1 ./pkt_inj 24 bridge.10 2001:db8::10
(pkt_inj is just a custom packet generator, nothing special)
Then remove that leg from the group by replace (let's assume it is id
200 in this case):
$ ip nexthop replace id 203 group 201
Now remove the IPv6 route:
$ ip -6 route del 2001:db8::10/128
The route won't be really deleted due to the stale rt6_info holding 1
refcnt in nexthop id 200.
At this point we have the following reference count dependency:
(deleted) IPv6 route holds 1 reference over nhid 203
nh 203 holds 1 ref over id 201
nh 200 holds 1 ref over the net device and the route due to the stale
rt6_info
Now to create circular dependency between nh 200 and the IPv6 route, and
also to get a reference over nh 200, restore nhid 200 in the group:
$ ip nexthop replace id 203 group 201/200
And now we have a permanent circular dependncy because nhid 203 holds a
reference over nh 200 and 201, but the route holds a ref over nh 203 and
is deleted.
To trigger the bug just delete the group (nhid 203):
$ ip nexthop del id 203
It won't really be deleted due to the IPv6 route dependency, and now we
have 2 unlinked and deleted objects that reference each other: the group
and the IPv6 route. Since the group drops the reference it holds over its
entries at free time (i.e. its own refcount needs to drop to 0) that will
never happen and we get a permanent ref on them, since one of the entries
holds a reference over the IPv6 route it will also never be released.
At this point the dependencies are:
(deleted, only unlinked) IPv6 route holds reference over group nh 203
(deleted, only unlinked) group nh 203 holds reference over nh 201 and 200
nh 200 holds 1 ref over the net device and the route due to the stale
rt6_info
This is the last point where it can be fixed by running traffic through
nh 200, and specifically through the same CPU so the rt6_info (dst) will
get released due to the IPv6 genid, that in turn will free the IPv6
route, which in turn will free the ref count over the group nh 203.
If nh 200 is deleted at this point, it will never be released due to the
ref from the unlinked group 203, it will only be unlinked:
$ ip nexthop del id 200
$ ip nexthop
$
Now we can never release that stale rt6_info, we have IPv6 route with ref
over group nh 203, group nh 203 with ref over nh 200 and 201, nh 200 with
rt6_info (dst) with ref over the net device and the IPv6 route. All of
these objects are only unlinked, and cannot be released, thus they can't
release their ref counts.
Message from syslogd@dev at Nov 19 14:04:10 ...
kernel:[73501.828730] unregister_netdevice: waiting for bridge.10 to become free. Usage count = 3
Message from syslogd@dev at Nov 19 14:04:20 ...
kernel:[73512.068811] unregister_netdevice: waiting for bridge.10 to become free. Usage count = 3
Fixes: 7bf4796dd099 ("nexthops: add support for replace") Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
We need a way to release a fib6_nh's per-cpu dsts when replacing
nexthops otherwise we can end up with stale per-cpu dsts which hold net
device references, so add a new IPv6 stub called fib6_nh_release_dsts.
It must be used after an RCU grace period, so no new dsts can be created
through a group's nexthop entry.
Similar to fib6_nh_release it shouldn't be used if fib6_nh_init has failed
so it doesn't need a dummy stub when IPv6 is not enabled.
Fixes: 7bf4796dd099 ("nexthops: add support for replace") Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Currently, when user space emits SIOCSHWTSTAMP ioctl calls such as
enabling/disabling timestamping or changing filter settings, the driver
reads the current CLOCK_REALTIME value and programming this into the
NIC's hardware clock. This might be necessary during system
initialization, but at runtime, when the PTP clock has already been
synchronized to a grandmaster, a reset of the timestamp settings might
result in a clock jump. Furthermore, if the clock is also controlled by
phc2sys in automatic mode (where the UTC offset is queried from ptp4l),
that UTC-to-TAI offset (currently 37 seconds in 2021) would be
temporarily reset to 0, and it would take a long time for phc2sys to
readjust so that CLOCK_REALTIME and the PHC are apart by 37 seconds
again.
To address the issue, we introduce a new function called
stmmac_init_tstamp_counter(), which gets called during ndo_open().
It contains the code snippet moved from stmmac_hwtstamp_set() that
manages the time synchronization. Besides, the sub second increment
configuration is also moved here since the related values are hardware
dependent and runtime invariant.
Furthermore, the hardware clock must be kept running even when no time
stamping mode is selected in order to retain the synchronized time base.
That way, timestamping can be enabled again at any time only with the
need to compensate the clock's natural drifting.
As a side effect, this patch fixes the issue that ptp_clock_info::enable
can be called before SIOCSHWTSTAMP and the driver (which looks at
priv->systime_flags) was not prepared to handle that ordering.
Fixes: 92ba6888510c ("stmmac: add the support for PTP hw clock driver") Reported-by: Michael Olbrich <m.olbrich@pengutronix.de> Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de> Signed-off-by: Holger Assmann <h.assmann@pengutronix.de> Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Use nn->tlv_caps.me_freq_mhz instead of nn->me_freq_mhz to check whether
rx-usecs/tx-usecs is valid.
This is because nn->tlv_caps.me_freq_mhz represents the clock_freq (MHz) of
the flow processing cores (FPC) on the NIC. While nn->me_freq_mhz is not
be set.
Fixes: ce991ab6662a ("nfp: read ME frequency from vNIC ctrl memory") Signed-off-by: Diana Wang <na.wang@corigine.com> Signed-off-by: Simon Horman <simon.horman@corigine.com> Reviewed-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The ARCH_FEATURES function ID is a 32-bit SMC call, which returns
a 32-bit result per the SMCCC spec. Current code is doing a 64-bit
comparison against -1 (SMCCC_RET_NOT_SUPPORTED) to detect that the
feature is unimplemented. That check doesn't work in a Hyper-V VM,
where the upper 32-bits are zero as allowed by the spec.
Cast the result as an 'int' so the comparison works. The change also
makes the code consistent with other similar checks in this file.
On kernels before v5.15, calling read() on a unix socket after
shutdown(SHUT_RD) or shutdown(SHUT_RDWR) would return the data
previously written or EOF. But now, while read() after
shutdown(SHUT_RD) still behaves the same way, read() after
shutdown(SHUT_RDWR) always fails with -EINVAL.
This behaviour change was apparently inadvertently introduced as part of
a bug fix for a different regression caused by the commit adding sockmap
support to af_unix, commit 94531cfcbe79c359 ("af_unix: Add
unix_stream_proto for sockmap"). Those commits, for unclear reasons,
started setting the socket state to TCP_CLOSE on shutdown(SHUT_RDWR),
while this state change had previously only been done in
unix_release_sock().
Restore the original behaviour. The sockmap tests in
tests/selftests/bpf continue to pass after this patch.
Fixes: d0c6416bd7091647f60 ("unix: Fix an issue in unix_shutdown causing the other end read/write failures") Link: https://lore.kernel.org/lkml/20211111140000.GA10779@axis.com/ Signed-off-by: Vincent Whitchurch <vincent.whitchurch@axis.com> Tested-by: Casey Schaufler <casey@schaufler-ca.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Scheduling a delack in mptcp_established_options_mp() is
not a good idea: such function is called by tcp_send_ack() and
the pending delayed ack will be cleared shortly after by the
tcp_event_ack_sent() call in __tcp_transmit_skb().
Instead use the mptcp delegated action infrastructure to
schedule the delayed ack after the current bh processing completes.
Additionally moves the schedule_3rdack_retransmission() helper
into protocol.c to avoid making it visible in a different compilation
unit.
Fixes: ec3edaa7ca6ce02f ("mptcp: Add handling of outgoing MP_JOIN requests") Reviewed-by: Mat Martineau <mathew.j.martineau>@linux.intel.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
To compute the rtx timeout schedule_3rdack_retransmission() does multiple
things in the wrong way: srtt_us is measured in usec/8 and the timeout
itself is an absolute value.
Fixes: ec3edaa7ca6ce02f ("mptcp: Add handling of outgoing MP_JOIN requests") Acked-by: Paolo Abeni <pabeni@redhat.com> Reviewed-by: Mat Martineau <mathew.j.martineau>@linux.intel.com> Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
These devices are based on an I2C/I2S device, we need to force the use
of the SOF driver otherwise the legacy HDaudio driver will be loaded -
only HDMI will be supported.
We previously added support for other Intel platforms but missed
JasperLake.
BugLink: https://github.com/thesofproject/linux/issues/3210 Fixes: 9d36ceab9415 ('ALSA: intel-dsp-config: add quirk for APL/GLK/TGL devices based on ES8336 codec') Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com> Reviewed-by: Kai Vehmanen <kai.vehmanen@intel.com> Signed-off-by: Bard Liao <yung-chuan.liao@linux.intel.com> Link: https://lore.kernel.org/r/20211027023254.24955-1-yung-chuan.liao@linux.intel.com Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
The prototype of mem_map_via_hcall() is missing in its header, so add
it.
Reported-by: kernel test robot <lkp@intel.com> Fixes: a43fb7da53007e67ad ("xen/pvh: Move Xen code for getting mem map via hcall out of common file") Signed-off-by: Juergen Gross <jgross@suse.com> Link: https://lore.kernel.org/r/20211119153913.21678-1-jgross@suse.com Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
When a VF goes through a reset, it's possible for the VF's feature set
to change. For example it may lose the VIRTCHNL_VF_OFFLOAD_VLAN
capability after VF reset. Unfortunately, the driver doesn't correctly
deal with this situation and errors are seen from downing/upping the
interface and/or moving the interface in/out of a network namespace.
When setting the interface down/up we see the following errors after the
VIRTCHNL_VF_OFFLOAD_VLAN capability was taken away from the VF:
These add/delete errors are happening because the VLAN filters are
tracked internally to the driver and regardless of the VLAN_ALLOWED()
setting the driver tries to delete/re-add them over virtchnl.
Fix the delete failure by making sure to delete any VLAN filter tracking
in the driver when a removal request is made, while preventing the
virtchnl request. This makes it so the driver's VLAN list is up to date
and the errors are
Fix the add failure by making sure the check for VLAN_ALLOWED() during
reset is done after the VF receives its capability list from the PF via
VIRTCHNL_OP_GET_VF_RESOURCES. If VLAN functionality is not allowed, then
prevent requesting re-adding the filters over virtchnl.
When moving the interface into a network namespace we see the following
errors after the VIRTCHNL_VF_OFFLOAD_VLAN capability was taken away from
the VF:
iavf 0000:51:09.1 enp81s0f1v1: NIC Link is Up Speed is 25 Gbps Full Duplex
iavf 0000:51:09.1 temp_27: renamed from enp81s0f1v1
iavf 0000:51:09.1 mgmt: renamed from temp_27
iavf 0000:51:09.1 dev27: set_features() failed (-22); wanted 0x020190001fd54833, left 0x020190001fd54bb3
These errors are happening because we aren't correctly updating the
netdev capabilities and dealing with ndo_fix_features() and
ndo_set_features() correctly.
Fix this by only reporting errors in the driver's ndo_set_features()
callback when VIRTCHNL_VF_OFFLOAD_VLAN is not allowed and any attempt to
enable the VLAN features is made. Also, make sure to disable VLAN
insertion, filtering, and stripping since the VIRTCHNL_VF_OFFLOAD_VLAN
flag applies to all of them and not just VLAN stripping.
Also, after we process the capabilities in the VF reset path, make sure
to call netdev_update_features() in case the capabilities have changed
in order to update the netdev's feature set to match the VF's actual
capabilities.
Lastly, make sure to always report success on VLAN filter delete when
VIRTCHNL_VF_OFFLOAD_VLAN is not supported. The changed flow in
iavf_del_vlans() allows the stack to delete previosly existing VLAN
filters even if VLAN filtering is not allowed. This makes it so the VLAN
filter list is up to date.
Fixes: 8774370d268f ("i40e/i40evf: support for VF VLAN tag stripping control") Signed-off-by: Brett Creeley <brett.creeley@intel.com> Tested-by: Konrad Jankowski <konrad0.jankowski@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Andrea Righi <andrea.righi@canonical.com>