ARM exclusively uses GENERIC_IRQ_MULTI_HANDLER, so at some point
set_handle_irq() needs to be called to handle system-wide
interrupts.
For all DT-enabled boards, this call happens down in the
drivers/irqchip subsystem, after locating the target irqchip
driver from the device tree.
We still have a few instances of the boardfiles with machine
descriptors passing a machine-specific .handle_irq() to the
ARM kernel core.
Get rid of this by letting the few remaining machines consistently
call set_handle_irq() from the end of the .init_irq() callback
instead and diet down one member from the machine descriptor.
Cc: Marc Zyngier <maz@kernel.org>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
The HASHKEYR register contains a secret per-process key to enable unique
hashes per process. In general it should not be exposed to userspace
at all and a regular process has no need to know its key.
However, checkpoint restore in userspace (CRIU) functionality requires
that a process be able to set the HASHKEYR of another process, otherwise
existing hashes on the stack would be invalidated by a new random key.
Exposing HASHKEYR in this way also makes it appear in core dumps, which
is a security concern. Multiple threads may share a key, for example
just after a fork() call, where the kernel cannot know if the child is
going to return back along the parent's stack. If such a thread is
coerced into making a core dump, then the HASHKEYR value will be
readable and able to be used against all other threads sharing that key,
effectively undoing any protection offered by hashst/hashchk.
Therefore we expose HASHKEYR to ptrace when CONFIG_CHECKPOINT_RESTORE is
enabled, providing a choice of increased security or migratable ROP
protected processes. This is similar to how ARM exposes its PAC keys.
Signed-off-by: Benjamin Gray <bgray@linux.ibm.com>
Reviewed-by: Russell Currey <ruscur@russell.cc>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://msgid.link/20230616034846.311705-8-bgray@linux.ibm.com
The DEXCR register is of interest when ptracing processes. Currently it
is static, but eventually will be dynamically controllable by a process.
If a process can control its own, then it is useful for it to be
ptrace-able to (e.g., for checkpoint-restore functionality).
It is also relevant to core dumps (the NPHIE aspect in particular),
which use the ptrace mechanism (or is it the other way around?) to
decide what to dump. The HDEXCR is useful here too, as the NPHIE aspect
may be set in the HDEXCR without being set in the DEXCR. Although the
HDEXCR is per-cpu and we don't track it in the task struct (it's useless
in normal operation), it would be difficult to imagine why a hypervisor
would set it to different values within a guest. A hypervisor cannot
safely set NPHIE differently at least, as that would break programs.
Expose a read-only view of the userspace DEXCR and HDEXCR to ptrace.
The HDEXCR is always readonly, and is useful for diagnosing the core
dumps (as the HDEXCR may set NPHIE without the DEXCR setting it).
Signed-off-by: Benjamin Gray <bgray@linux.ibm.com>
Reviewed-by: Russell Currey <ruscur@russell.cc>
[mpe: Use lower_32_bits() rather than open coding]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://msgid.link/20230616034846.311705-7-bgray@linux.ibm.com
The trace output for the HRTIMER_MODE_.*_HARD modes is seen as a number
since these modes are not decoded. The author was not aware of the fancy
decoding function which makes the life easier.
Extend decode_hrtimer_mode() with the additional HRTIMER_MODE_.*_HARD
modes.
Fixes: ae6683d815 ("hrtimer: Introduce HARD expiry mode")
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Mukesh Ojha <quic_mojha@quicinc.com>
Acked-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Link: https://lore.kernel.org/r/20230418143854.8vHWQKLM@linutronix.de
The VIA fbdev exposes a custom GPIO chip for its GPIOs, these
are in turn looked up the camera driver using a custom API.
Drop the custom API, provide a look-up table and convert to
GPIO descriptors. Note proper polarity on the RESET line.
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Helge Deller <deller@gmx.de>
Group some variables based on their sizes to reduce hole and avoid padding.
On x86_64, this shrinks the size of 'struct hdmi_avi_infoframe'
from 68 to 60 bytes.
It saves a few bytes of memory and is more cache-line friendly.
This also reduces the union hdmi_infoframe the same way.
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Helge Deller <deller@gmx.de>
Updates for v6.5.. this includes a backmerg of drm-next tree to be able
to use new DRM DSC helpers.
Core:
+ Add Marijn Suijten as drm/msm reviewer
+ Adreno A660 bindings
+ SM8350 MDSS bindings fix
+ Fix adreno_is_a690() warnings
+ More generic (DRM) and MSM-specific DSC helpers
DP:
+ Removed obsolete USB-PD remains
+ Documented DP compatible string for sm8550 platform
DPU:
+ Enable missing features (DSPP, DSC, split display) on sc8180x,
sc8280xp, sm8450
+ Enabled writeback on sc7280
+ Implemented tearcheck support to support vsync on SM150 and
newer platforms
+ Native HDMI output support
+ Dropped unused features: regdma, GC, IGC
+ Fixed the DSC flush operations
+ Simplified QoS handling, removing obsolete and unused features
and merging SSPP and WB code paths
+ Reworked dpu_encoder initialisation path
+ Enabled DSPP support on sdm845
+ Disabled color-management if DSPP blocks are not available
+ Added support for DSC 1.2 blocks found on sm8350 and later
+ Added .fb_dirty to fix CMD panels
DSI:
+ Drop powerup quirks in favour of using pre_enable_prev_first for
downstream bridges
+ Fixed 14nm DSI PHY programming
+ Added support for DSI and 28nm DSI PHY on MSM8226 platform
+ Make use of DRM and MSM DSC helpers
MDP5:
+ Added support for display controller on MSM8226 platform
GPU:
+ A690 support
+ Don't set IO_PGTABLE_QUIRK_ARM_OUTER_WBWA on devices with coherent SMMU
(like A690)
+ Move cmdstream dumping out of fence signaling path
+ Cleanups
+ Support for a6xx devices without GMU (aka "GMU wrapper"
+ a610 support
+ a619_holi support (a619 variant without GMU)
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rob Clark <robdclark@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/CAF6AEGsUB=tRB4nR6ZCJMuLhro5zN3BQWUSywVYbaipqqDZ_cQ@mail.gmail.com
The sys_ni_posix_timers() definition causes a warning when the declaration
is missing, so this needs to be added along with the normal syscalls,
outside of the #ifdef.
kernel/time/posix-stubs.c:26:17: error: no previous prototype for 'sys_ni_posix_timers' [-Werror=missing-prototypes]
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20230607142925.3126422-1-arnd@kernel.org
posix_timer_add() tries to allocate a posix timer ID by starting from the
cached ID which was stored by the last successful allocation.
This is done in a loop searching the ID space for a free slot one by
one. The loop has to terminate when the search wrapped around to the
starting point.
But that's racy vs. establishing the starting point. That is read out
lockless, which leads to the following problem:
CPU0 CPU1
posix_timer_add()
start = sig->posix_timer_id;
lock(hash_lock);
... posix_timer_add()
if (++sig->posix_timer_id < 0)
start = sig->posix_timer_id;
sig->posix_timer_id = 0;
So CPU1 can observe a negative start value, i.e. -1, and the loop break
never happens because the condition can never be true:
if (sig->posix_timer_id == start)
break;
While this is unlikely to ever turn into an endless loop as the ID space is
huge (INT_MAX), the racy read of the start value caught the attention of
KCSAN and Dmitry unearthed that incorrectness.
Rewrite it so that all id operations are under the hash lock.
Reported-by: syzbot+5c54bd3eb218bb595aa9@syzkaller.appspotmail.com
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Link: https://lore.kernel.org/r/87bkhzdn6g.ffs@tglx
mlx5-updates-2023-06-16
1) Added a new event handler to firmware sync reset, which is used to
support firmware sync reset flow on smart NIC. Adding this new stage to
the flow enables the firmware to ensure host PFs unload before ECPFs
unload, to avoid race of PFs recovery.
2) Debugfs for mlx5 eswitch bridge offloads
3) Added two new counters for vport stats
4) Minor Fixups and cleanups for net-next branch
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull ata fix from Damien Le Moal:
- Avoid deadlocks on resume from sleep by delaying scsi rescan until
the scsi device is also fully resumed.
* tag 'ata-6.4-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/dlemoal/libata:
ata: libata-scsi: Avoid deadlock on rescan after device resume
We implement KVM device interface for in-kernel AIA irqchip so that
user-space can use KVM device ioctls to create, configure, and destroy
in-kernel AIA irqchip.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Atish Patra <atishp@rivosinc.com>
Signed-off-by: Anup Patel <anup@brainfault.org>
Per-VMA locking allows us to lock a struct vm_area_struct without
taking the process-wide mmap lock in read mode.
Consider a process workload where the mmap lock is taken constantly in
write mode. In this scenario, all zerocopy receives are periodically
blocked during that period of time - though in principle, the memory
ranges being used by TCP are not touched by the operations that need
the mmap write lock. This results in performance degradation.
Now consider another workload where the mmap lock is never taken in
write mode, but there are many TCP connections using receive zerocopy
that are concurrently receiving. These connections all take the mmap
lock in read mode, but this does induce a lot of contention and atomic
ops for this process-wide lock. This results in additional CPU
overhead caused by contending on the cache line for this lock.
However, with per-vma locking, both of these problems can be avoided.
As a test, I ran an RPC-style request/response workload with 4KB
payloads and receive zerocopy enabled, with 100 simultaneous TCP
connections. I measured perf cycles within the
find_tcp_vma/mmap_read_lock/mmap_read_unlock codepath, with and
without per-vma locking enabled.
When using process-wide mmap semaphore read locking, about 1% of
measured perf cycles were within this path. With per-VMA locking, this
value dropped to about 0.45%.
Signed-off-by: Arjun Roy <arjunroy@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is part of the effort to remove the empty element at the end of
ctl_table structs. "child" was a deprecated elem in this struct and was
being used to differentiate between two types of ctl_tables: "normal"
and "permanently emtpy".
What changed?:
* Replace "child" with an enumeration that will have two values: the
default (0) and the permanently empty (1). The latter is left at zero
so when struct ctl_table is created with kzalloc or in a local
context, it will have the zero value by default. We document the
new enum with kdoc.
* Remove the "empty child" check from sysctl_check_table
* Remove count_subheaders function as there is no longer a need to
calculate how many headers there are for every child
* Remove the recursive call to unregister_sysctl_table as there is no
need to traverse down the child tree any longer
* Add a new SYSCTL_PERM_EMPTY_DIR binary flag
* Remove the last remanence of child from partport/procfs.c
Signed-off-by: Joel Granados <j.granados@samsung.com>
Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
When an ATA port is resumed from sleep, the port is reset and a power
management request issued to libata EH to reset the port and rescanning
the device(s) attached to the port. Device rescanning is done by
scheduling an ata_scsi_dev_rescan() work, which will execute
scsi_rescan_device().
However, scsi_rescan_device() takes the generic device lock, which is
also taken by dpm_resume() when the SCSI device is resumed as well. If
a device rescan execution starts before the completion of the SCSI
device resume, the rcu locking used to refresh the cached VPD pages of
the device, combined with the generic device locking from
scsi_rescan_device() and from dpm_resume() can cause a deadlock.
Avoid this situation by changing struct ata_port scsi_rescan_task to be
a delayed work instead of a simple work_struct. ata_scsi_dev_rescan() is
modified to check if the SCSI device associated with the ATA device that
must be rescanned is not suspended. If the SCSI device is still
suspended, ata_scsi_dev_rescan() returns early and reschedule itself for
execution after an arbitrary delay of 5ms.
Reported-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Reported-by: Joe Breuer <linux-kernel@jmbreuer.net>
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217530
Fixes: a19a93e4c6 ("scsi: core: pm: Rely on the device driver core for async power management")
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Tested-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Tested-by: Joe Breuer <linux-kernel@jmbreuer.net>
These commits
a494aef23d ("PCI: hv: Replace retarget_msi_interrupt_params with hyperv_pcpu_input_arg")
2c6ba42168 ("PCI: hv: Enable PCI pass-thru devices in Confidential VMs")
update the Hyper-V virtual PCI driver to use the hyperv_pcpu_input_arg
because that memory will be correctly marked as decrypted or encrypted
for all VM types (CoCo or normal). But problems ensue when CPUs in the
VM go online or offline after virtual PCI devices have been configured.
When a CPU is brought online, the hyperv_pcpu_input_arg for that CPU is
initialized by hv_cpu_init() running under state CPUHP_AP_ONLINE_DYN.
But this state occurs after state CPUHP_AP_IRQ_AFFINITY_ONLINE, which
may call the virtual PCI driver and fault trying to use the as yet
uninitialized hyperv_pcpu_input_arg. A similar problem occurs in a CoCo
VM if the MMIO read and write hypercalls are used from state
CPUHP_AP_IRQ_AFFINITY_ONLINE.
When a CPU is taken offline, IRQs may be reassigned in state
CPUHP_TEARDOWN_CPU. Again, the virtual PCI driver may fault trying to
use the hyperv_pcpu_input_arg that has already been freed by a
higher state.
Fix the onlining problem by adding state CPUHP_AP_HYPERV_ONLINE
immediately after CPUHP_AP_ONLINE_IDLE (similar to CPUHP_AP_KVM_ONLINE)
and before CPUHP_AP_IRQ_AFFINITY_ONLINE. Use this new state for
Hyper-V initialization so that hyperv_pcpu_input_arg is allocated
early enough.
Fix the offlining problem by not freeing hyperv_pcpu_input_arg when
a CPU goes offline. Retain the allocated memory, and reuse it if
the CPU comes back online later.
Signed-off-by: Michael Kelley <mikelley@microsoft.com>
Reviewed-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Acked-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Dexuan Cui <decui@microsoft.com>
Link: https://lore.kernel.org/r/1684862062-51576-1-git-send-email-mikelley@microsoft.com
Signed-off-by: Wei Liu <wei.liu@kernel.org>
This event brackets the svcrdma_post_* trace points. If this trace
event is enabled but does not appear as expected, that indicates a
chunk_ctxt leak.
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Tom Talpey <tom@talpey.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Under certain circumstances, the tcp receive buffer memory limit
set by autotuning (sk_rcvbuf) is increased due to incoming data
packets as a result of the window not closing when it should be.
This can result in the receive buffer growing all the way up to
tcp_rmem[2], even for tcp sessions with a low BDP.
To reproduce: Connect a TCP session with the receiver doing
nothing and the sender sending small packets (an infinite loop
of socket send() with 4 bytes of payload with a sleep of 1 ms
in between each send()). This will cause the tcp receive buffer
to grow all the way up to tcp_rmem[2].
As a result, a host can have individual tcp sessions with receive
buffers of size tcp_rmem[2], and the host itself can reach tcp_mem
limits, causing the host to go into tcp memory pressure mode.
The fundamental issue is the relationship between the granularity
of the window scaling factor and the number of byte ACKed back
to the sender. This problem has previously been identified in
RFC 7323, appendix F [1].
The Linux kernel currently adheres to never shrinking the window.
In addition to the overallocation of memory mentioned above, the
current behavior is functionally incorrect, because once tcp_rmem[2]
is reached when no remediations remain (i.e. tcp collapse fails to
free up any more memory and there are no packets to prune from the
out-of-order queue), the receiver will drop in-window packets
resulting in retransmissions and an eventual timeout of the tcp
session. A receive buffer full condition should instead result
in a zero window and an indefinite wait.
In practice, this problem is largely hidden for most flows. It
is not applicable to mice flows. Elephant flows can send data
fast enough to "overrun" the sk_rcvbuf limit (in a single ACK),
triggering a zero window.
But this problem does show up for other types of flows. Examples
are websockets and other type of flows that send small amounts of
data spaced apart slightly in time. In these cases, we directly
encounter the problem described in [1].
RFC 7323, section 2.4 [2], says there are instances when a retracted
window can be offered, and that TCP implementations MUST ensure
that they handle a shrinking window, as specified in RFC 1122,
section 4.2.2.16 [3]. All prior RFCs on the topic of tcp window
management have made clear that sender must accept a shrunk window
from the receiver, including RFC 793 [4] and RFC 1323 [5].
This patch implements the functionality to shrink the tcp window
when necessary to keep the right edge within the memory limit by
autotuning (sk_rcvbuf). This new functionality is enabled with
the new sysctl: net.ipv4.tcp_shrink_window
Additional information can be found at:
https://blog.cloudflare.com/unbounded-memory-usage-by-tcp-for-receive-buffers-and-how-we-fixed-it/
[1] https://www.rfc-editor.org/rfc/rfc7323#appendix-F
[2] https://www.rfc-editor.org/rfc/rfc7323#section-2.4
[3] https://www.rfc-editor.org/rfc/rfc1122#page-91
[4] https://www.rfc-editor.org/rfc/rfc793
[5] https://www.rfc-editor.org/rfc/rfc1323
Signed-off-by: Mike Freemon <mfreemon@cloudflare.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
iort_pmsi_get_dev_id() has a __weak definition in the driver, and
an override in arm64 specific code, but the declaration is conditional
and not always seen when the copy in the driver gets built:
drivers/irqchip/irq-gic-v3-its-platform-msi.c:41:12: error: no previous prototype for 'iort_pmsi_get_dev_id' [-Werror=missing-prototypes]
Move the existing declaration out of the #ifdef block to ensure
it can be seen in all configurations.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Hanjun Guo <guohanjun@huawei.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20230516200516.554663-5-arnd@kernel.org
Building with "W=1" warns about missing declarations for
two functions in the mmp irqchip driver:
drivers/irqchip/irq-mmp.c:248:13: error: no previous prototype for 'icu_init_irq'
drivers/irqchip/irq-mmp.c:271:13: error: no previous prototype for 'mmp2_init_icu'
The declarations are present in an unused header, but since there is no
caller, it's best to just remove the functions and the header completely,
making the driver DT-only to match the state of the platform.
Fixes: 77acc85ce7 ("ARM: mmp: remove device definitions")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20230516200516.554663-2-arnd@kernel.org
As described in commit 8111964f1b ("dm thin: Fix ABBA deadlock between
shrink_slab and dm_pool_abort_metadata"), ABBA deadlocks will be
triggered because shrinker_rwsem currently needs to held by
dm_pool_abort_metadata() as a side-effect of thin-pool metadata
operation failure.
The following three problem scenarios have been noticed:
1) Described by commit 8111964f1b ("dm thin: Fix ABBA deadlock between
shrink_slab and dm_pool_abort_metadata")
2) shrinker_rwsem and throttle->lock
P1(drop cache) P2(kworker)
drop_caches_sysctl_handler
drop_slab
shrink_slab
down_read(&shrinker_rwsem) - LOCK A
do_shrink_slab
super_cache_scan
prune_icache_sb
dispose_list
evict
ext4_evict_inode
ext4_clear_inode
ext4_discard_preallocations
ext4_mb_load_buddy_gfp
ext4_mb_init_cache
ext4_wait_block_bitmap
__ext4_error
ext4_handle_error
ext4_commit_super
...
dm_submit_bio
do_worker
throttle_work_update
down_write(&t->lock) -- LOCK B
process_deferred_bios
commit
metadata_operation_failed
dm_pool_abort_metadata
dm_block_manager_create
dm_bufio_client_create
register_shrinker
down_write(&shrinker_rwsem)
-- LOCK A
thin_map
thin_bio_map
thin_defer_bio_with_throttle
throttle_lock
down_read(&t->lock) - LOCK B
3) shrinker_rwsem and wait_on_buffer
P1(drop cache) P2(kworker)
drop_caches_sysctl_handler
drop_slab
shrink_slab
down_read(&shrinker_rwsem) - LOCK A
do_shrink_slab
...
ext4_wait_block_bitmap
__ext4_error
ext4_handle_error
jbd2_journal_abort
jbd2_journal_update_sb_errno
jbd2_write_superblock
submit_bh
// LOCK B
// RELEASE B
do_worker
throttle_work_update
down_write(&t->lock) - LOCK B
process_deferred_bios
process_bio
commit
metadata_operation_failed
dm_pool_abort_metadata
dm_block_manager_create
dm_bufio_client_create
register_shrinker
register_shrinker_prepared
down_write(&shrinker_rwsem) - LOCK A
bio_endio
wait_on_buffer
__wait_on_buffer
Fix these by resetting dm_bufio_client without holding shrinker_rwsem.
Fixes: 8111964f1b ("dm thin: Fix ABBA deadlock between shrink_slab and dm_pool_abort_metadata")
Cc: stable@vger.kernel.org
Signed-off-by: Li Lingfeng <lilingfeng3@huawei.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Add a tracepoint for when a CSD is queued to a remote CPU's
call_single_queue. This allows finding exactly which CPU queued a given CSD
when looking at a csd_function_{entry,exit} event, and also enables us to
accurately measure IPI delivery time with e.g. a synthetic event:
$ echo 'hist:keys=cpu,csd.hex:ts=common_timestamp.usecs' >\
/sys/kernel/tracing/events/smp/csd_queue_cpu/trigger
$ echo 'csd_latency unsigned int dst_cpu; unsigned long csd; u64 time' >\
/sys/kernel/tracing/synthetic_events
$ echo \
'hist:keys=common_cpu,csd.hex:'\
'time=common_timestamp.usecs-$ts:'\
'onmatch(smp.csd_queue_cpu).trace(csd_latency,common_cpu,csd,$time)' >\
/sys/kernel/tracing/events/smp/csd_function_entry/trigger
$ trace-cmd record -e 'synthetic:csd_latency' hackbench
$ trace-cmd report
<...>-467 [001] 21.824263: csd_queue_cpu: cpu=0 callsite=try_to_wake_up+0x2ea func=sched_ttwu_pending csd=0xffff8880076148b8
<...>-467 [001] 21.824280: ipi_send_cpu: cpu=0 callsite=try_to_wake_up+0x2ea callback=generic_smp_call_function_single_interrupt+0x0
<...>-489 [000] 21.824299: csd_function_entry: func=sched_ttwu_pending csd=0xffff8880076148b8
<...>-489 [000] 21.824320: csd_latency: dst_cpu=0, csd=18446612682193848504, time=36
Suggested-by: Valentin Schneider <vschneid@redhat.com>
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Tested-and-reviewed-by: Valentin Schneider <vschneid@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20230615065944.188876-7-leobras@redhat.com
The recently added ipi_send_{cpu,cpumask} tracepoints allow finding sources
of IPIs targeting CPUs running latency-sensitive applications.
For NOHZ_FULL CPUs, all IPIs are interference, and those tracepoints are
sufficient to find them and work on getting rid of them. In some setups
however, not *all* IPIs are to be suppressed, but long-running IPI
callbacks can still be problematic.
Add a pair of tracepoints to mark the start and end of processing a CSD IPI
callback, similar to what exists for softirq, workqueue or timer callbacks.
Signed-off-by: Leonardo Bras <leobras@redhat.com>
Tested-and-reviewed-by: Valentin Schneider <vschneid@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20230615065944.188876-5-leobras@redhat.com
Added a new event handler to firmware sync reset, which is used to
support firmware sync reset flow on smart NIC. Adding this new stage to
the flow enables the firmware to ensure host PFs unload before ECPFs
unload, to avoid race of PFs recovery.
If firmware sends sync_reset_unload event to driver the driver should
unload and close all HW resources of the function. Once the driver
finishes unloading part, it can't get any more events from firmware as
event queues are closed, so it polls the reset state field to know when
to continue to next stage of the sync reset flow.
Added capability bit for supporting sync_reset_unload event.
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Expose new timoueout in Default Timeouts Register to be used on sync
reset flow running on smart NIC. In this flow the driver should know how
much time to wait from getting unload request till firmware will ask the
PF to continue to next stage of the flow.
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Pull RCU fix from Paul McKenney:
"This fixes a spinlock-initialization regression in SRCU that causes
the SRCU notifier to fail.
The fix simply adds the initialization, but introduces a #ifdef
because there is no spinlock to initialize for the Tiny SRCU used in
!SMP builds.
Yes, it would be nice to abstract this somehow in order to hide it in
SRCU, but I still don't see a good way of doing this"
* tag 'urgent-rcu.2023.06.11a' of git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu:
notifier: Initialize new struct srcu_usage field
vfio-cdx driver enables IOCTLs for user space to query
MMIO regions for CDX devices and mmap them. This change
also adds support for reset of CDX devices. With VFIO
enabled on CDX devices, user-space applications can also
exercise DMA securely via IOMMU on these devices.
This change adds the VFIO CDX driver and enables the following
ioctls for CDX devices:
- VFIO_DEVICE_GET_INFO:
- VFIO_DEVICE_GET_REGION_INFO
- VFIO_DEVICE_RESET
Signed-off-by: Nipun Gupta <nipun.gupta@amd.com>
Reviewed-by: Pieter Jansen van Vuuren <pieter.jansen-van-vuuren@amd.com>
Tested-by: Nikhil Agarwal <nikhil.agarwal@amd.com>
Link: https://lore.kernel.org/r/20230531124557.11009-1-nipun.gupta@amd.com
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Martin Wilck <mwilck@suse.com> says:
This patch series addresses some issues we saw in a test setup with a
large number of SCSI LUNs. The first two patches simply increase the
number of available sg and bsg devices. 3-5 fix a large delay we
encountered between blocking a Fibre Channel remote port and the
dev_loss_tmo. 6 renames scsi_target_block() to scsi_block_targets(),
and makes additional changes to this API, as suggested in the review
of the v2 series. 7 improves a warning message.
Link: https://lore.kernel.org/r/20230614103616.31857-1-mwilck@suse.com
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
All callers (fc_remote_port_delete(), __iscsi_block_session(),
__srp_start_tl_fail_timers(), srp_reconnect_rport(), snic_tgt_del()) pass
parent devices of scsi_target devices to scsi_target_block().
Rename the function to scsi_block_targets(), and simplify it by assuming
that it is always passed a parent device. Also, have callers pass the
Scsi_Host pointer to scsi_block_targets(), as every caller has this pointer
readily available.
Suggested-by: Christoph Hellwig <hch@lst.de>
Suggested-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin Wilck <mwilck@suse.com>
Link: https://lore.kernel.org/r/20230614103616.31857-7-mwilck@suse.com
Cc: Karan Tilak Kumar <kartilak@cisco.com>
Cc: Sesidhar Baddela <sebaddel@cisco.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
In MCQ mode, when a device command uses a hardware queue shared with other
commands, a race condition may occur in the following scenario:
1. A device command is completed in CQx with CQE entry "e".
2. The interrupt handler copies the "cqe" pointer to "hba->dev_cmd.cqe"
and completes "hba->dev_cmd.complete".
3. The "ufshcd_wait_for_dev_cmd()" function is awakened and retrieves the
OCS value from "hba->dev_cmd.cqe".
However, there is a possibility that the CQE entry "e" will be overwritten
by newly completed commands in CQx, resulting in an incorrect OCS value
being received by "ufshcd_wait_for_dev_cmd()".
To avoid this race condition, the OCS value should be immediately copied to
the struct "lrb" of the device command. Then "ufshcd_wait_for_dev_cmd()"
can retrieve the OCS value from the struct "lrb".
Fixes: 57b1c0ef89 ("scsi: ufs: core: mcq: Add support to allocate multiple queues")
Suggested-by: Can Guo <quic_cang@quicinc.com>
Signed-off-by: Stanley Chu <stanley.chu@mediatek.com>
Link: https://lore.kernel.org/r/20230610021553.1213-2-powen.kao@mediatek.com
Tested-by: Po-Wen Kao <powen.kao@mediatek.com>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
The introduction of the macro IOPRIO_PRIO_LEVEL() in commit eca2040972
("scsi: block: ioprio: Clean up interface definition") results in an
iopriority level to always be masked using the macro IOPRIO_LEVEL_MASK, and
thus to the kernel always seeing an acceptable value for an I/O priority
level when checked in ioprio_check_cap(). Before this patch, this function
would return an error for some (but not all) invalid values for a level
valid range of [0..7].
Restore and improve the detection of invalid priority levels by introducing
the inline function ioprio_value() to check an ioprio class, level and hint
value before combining these fields into a single value to be used with
ioprio_set() or AIOs. If an invalid value for the class, level or hint of
an ioprio is detected, ioprio_value() returns an ioprio using the class
IOPRIO_CLASS_INVALID, indicating an invalid value and causing
ioprio_check_cap() to return -EINVAL.
Fixes: 6c91325722 ("scsi: block: Introduce ioprio hints")
Fixes: eca2040972 ("scsi: block: ioprio: Clean up interface definition")
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Link: https://lore.kernel.org/r/20230608095556.124001-1-dlemoal@kernel.org
Reviewed-by: Niklas Cassel <niklas.cassel@wdc.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
