linux-stable.git/kernel, branch v5.4.261

genirq/matrix: Exclude managed interrupts in irq_matrix_allocated()

2023-11-20T09:30:08+00:00

[ Upstream commit a0b0bad10587ae2948a7c36ca4ffc206007fbcf3 ]

When a CPU is about to be offlined, x86 validates that all active
interrupts which are targeted to this CPU can be migrated to the remaining
online CPUs. If not, the offline operation is aborted.

The validation uses irq_matrix_allocated() to retrieve the number of
vectors which are allocated on the outgoing CPU. The returned number of
allocated vectors includes also vectors which are associated to managed
interrupts.

That's overaccounting because managed interrupts are:

  - not migrated when the affinity mask of the interrupt targets only
    the outgoing CPU

  - migrated to another CPU, but in that case the vector is already
    pre-allocated on the potential target CPUs and must not be taken into
    account.

As a consequence the check whether the remaining online CPUs have enough
capacity for migrating the allocated vectors from the outgoing CPU might
fail incorrectly.

Let irq_matrix_allocated() return only the number of allocated non-managed
interrupts to make this validation check correct.

[ tglx: Amend changelog and fixup kernel-doc comment ]

Fixes: 2f75d9e1c905 ("genirq: Implement bitmap matrix allocator")
Reported-by: Wendy Wang 
Signed-off-by: Chen Yu 
Signed-off-by: Thomas Gleixner 
Link: https://lore.kernel.org/r/20231020072522.557846-1-yu.c.chen@intel.com
Signed-off-by: Sasha Levin

perf/core: Fix potential NULL deref

2023-11-08T10:23:37+00:00

commit a71ef31485bb51b846e8db8b3a35e432cc15afb5 upstream.

Smatch is awesome.

Fixes: 32671e3799ca ("perf: Disallow mis-matched inherited group reads")
Reported-by: Dan Carpenter 
Signed-off-by: Peter Zijlstra (Intel) 
Signed-off-by: Ingo Molnar 
Signed-off-by: Greg Kroah-Hartman

perf: Disallow mis-matched inherited group reads

2023-10-25T09:53:24+00:00

commit 32671e3799ca2e4590773fd0e63aaa4229e50c06 upstream.

Because group consistency is non-atomic between parent (filedesc) and children
(inherited) events, it is possible for PERF_FORMAT_GROUP read() to try and sum
non-matching counter groups -- with non-sensical results.

Add group_generation to distinguish the case where a parent group removes and
adds an event and thus has the same number, but a different configuration of
events as inherited groups.

This became a problem when commit fa8c269353d5 ("perf/core: Invert
perf_read_group() loops") flipped the order of child_list and sibling_list.
Previously it would iterate the group (sibling_list) first, and for each
sibling traverse the child_list. In this order, only the group composition of
the parent is relevant. By flipping the order the group composition of the
child (inherited) events becomes an issue and the mis-match in group
composition becomes evident.

That said; even prior to this commit, while reading of a group that is not
equally inherited was not broken, it still made no sense.

(Ab)use ECHILD as error return to indicate issues with child process group
composition.

Fixes: fa8c269353d5 ("perf/core: Invert perf_read_group() loops")
Reported-by: Budimir Markovic 
Signed-off-by: Peter Zijlstra (Intel) 
Link: https://lkml.kernel.org/r/20231018115654.GK33217@noisy.programming.kicks-ass.net
Signed-off-by: Greg Kroah-Hartman

tracing: relax trace_event_eval_update() execution with cond_resched()

2023-10-25T09:53:23+00:00

[ Upstream commit 23cce5f25491968b23fb9c399bbfb25f13870cd9 ]

When kernel is compiled without preemption, the eval_map_work_func()
(which calls trace_event_eval_update()) will not be preempted up to its
complete execution. This can actually cause a problem since if another
CPU call stop_machine(), the call will have to wait for the
eval_map_work_func() function to finish executing in the workqueue
before being able to be scheduled. This problem was observe on a SMP
system at boot time, when the CPU calling the initcalls executed
clocksource_done_booting() which in the end calls stop_machine(). We
observed a 1 second delay because one CPU was executing
eval_map_work_func() and was not preempted by the stop_machine() task.

Adding a call to cond_resched() in trace_event_eval_update() allows
other tasks to be executed and thus continue working asynchronously
like before without blocking any pending task at boot time.

Link: https://lore.kernel.org/linux-trace-kernel/20230929191637.416931-1-cleger@rivosinc.com

Cc: Masami Hiramatsu 
Signed-off-by: Clément Léger 
Tested-by: Atish Patra 
Reviewed-by: Atish Patra 
Signed-off-by: Steven Rostedt (Google) 
Signed-off-by: Sasha Levin

cgroup: Remove duplicates in cgroup v1 tasks file

2023-10-25T09:53:19+00:00

commit 1ca0b605150501b7dc59f3016271da4eb3e96fce upstream.

One PID may appear multiple times in a preloaded pidlist.
(Possibly due to PID recycling but we have reports of the same
task_struct appearing with different PIDs, thus possibly involving
transfer of PID via de_thread().)

Because v1 seq_file iterator uses PIDs as position, it leads to
a message:
> seq_file: buggy .next function kernfs_seq_next did not update position index

Conservative and quick fix consists of removing duplicates from `tasks`
file (as opposed to removing pidlists altogether). It doesn't affect
correctness (it's sufficient to show a PID once), performance impact
would be hidden by unconditional sorting of the pidlist already in place
(asymptotically).

Link: https://lore.kernel.org/r/20230823174804.23632-1-mkoutny@suse.com/
Suggested-by: Firo Yang 
Signed-off-by: Michal Koutný 
Signed-off-by: Tejun Heo 
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman

workqueue: Override implicit ordered attribute in workqueue_apply_unbound_cpumask()

2023-10-25T09:53:18+00:00

[ Upstream commit ca10d851b9ad0338c19e8e3089e24d565ebfffd7 ]

Commit 5c0338c68706 ("workqueue: restore WQ_UNBOUND/max_active==1
to be ordered") enabled implicit ordered attribute to be added to
WQ_UNBOUND workqueues with max_active of 1. This prevented the changing
of attributes to these workqueues leading to fix commit 0a94efb5acbb
("workqueue: implicit ordered attribute should be overridable").

However, workqueue_apply_unbound_cpumask() was not updated at that time.
So sysfs changes to wq_unbound_cpumask has no effect on WQ_UNBOUND
workqueues with implicit ordered attribute. Since not all WQ_UNBOUND
workqueues are visible on sysfs, we are not able to make all the
necessary cpumask changes even if we iterates all the workqueue cpumasks
in sysfs and changing them one by one.

Fix this problem by applying the corresponding change made
to apply_workqueue_attrs_locked() in the fix commit to
workqueue_apply_unbound_cpumask().

Fixes: 5c0338c68706 ("workqueue: restore WQ_UNBOUND/max_active==1 to be ordered")
Signed-off-by: Waiman Long 
Signed-off-by: Tejun Heo 
Signed-off-by: Sasha Levin

ring-buffer: Update "shortest_full" in polling

2023-10-10T19:46:41+00:00

commit 1e0cb399c7653462d9dadf8ab9425337c355d358 upstream.

It was discovered that the ring buffer polling was incorrectly stating
that read would not block, but that's because polling did not take into
account that reads will block if the "buffer-percent" was set. Instead,
the ring buffer polling would say reads would not block if there was any
data in the ring buffer. This was incorrect behavior from a user space
point of view. This was fixed by commit 42fb0a1e84ff by having the polling
code check if the ring buffer had more data than what the user specified
"buffer percent" had.

The problem now is that the polling code did not register itself to the
writer that it wanted to wait for a specific "full" value of the ring
buffer. The result was that the writer would wake the polling waiter
whenever there was a new event. The polling waiter would then wake up, see
that there's not enough data in the ring buffer to notify user space and
then go back to sleep. The next event would wake it up again.

Before the polling fix was added, the code would wake up around 100 times
for a hackbench 30 benchmark. After the "fix", due to the constant waking
of the writer, it would wake up over 11,0000 times! It would never leave
the kernel, so the user space behavior was still "correct", but this
definitely is not the desired effect.

To fix this, have the polling code add what it's waiting for to the
"shortest_full" variable, to tell the writer not to wake it up if the
buffer is not as full as it expects to be.

Note, after this fix, it appears that the waiter is now woken up around 2x
the times it was before (~200). This is a tremendous improvement from the
11,000 times, but I will need to spend some time to see why polling is
more aggressive in its wakeups than the read blocking code.

Link: https://lore.kernel.org/linux-trace-kernel/20230929180113.01c2cae3@rorschach.local.home

Cc: stable@vger.kernel.org
Cc: Masami Hiramatsu 
Cc: Mark Rutland 
Fixes: 42fb0a1e84ff ("tracing/ring-buffer: Have polling block on watermark")
Reported-by: Julia Lawall 
Tested-by: Julia Lawall 
Signed-off-by: Steven Rostedt (Google) 
Signed-off-by: Greg Kroah-Hartman

ring-buffer: Avoid softlockup in ring_buffer_resize()

2023-10-10T19:46:40+00:00

[ Upstream commit f6bd2c92488c30ef53b5bd80c52f0a7eee9d545a ]

When user resize all trace ring buffer through file 'buffer_size_kb',
then in ring_buffer_resize(), kernel allocates buffer pages for each
cpu in a loop.

If the kernel preemption model is PREEMPT_NONE and there are many cpus
and there are many buffer pages to be allocated, it may not give up cpu
for a long time and finally cause a softlockup.

To avoid it, call cond_resched() after each cpu buffer allocation.

Link: https://lore.kernel.org/linux-trace-kernel/20230906081930.3939106-1-zhengyejian1@huawei.com

Cc: 
Signed-off-by: Zheng Yejian 
Signed-off-by: Steven Rostedt (Google) 
Signed-off-by: Sasha Levin

bpf: Avoid deadlock when using queue and stack maps from NMI

2023-10-10T19:46:36+00:00

[ Upstream commit a34a9f1a19afe9c60ca0ea61dfeee63a1c2baac8 ]

Sysbot discovered that the queue and stack maps can deadlock if they are
being used from a BPF program that can be called from NMI context (such as
one that is attached to a perf HW counter event). To fix this, add an
in_nmi() check and use raw_spin_trylock() in NMI context, erroring out if
grabbing the lock fails.

Fixes: f1a2e44a3aec ("bpf: add queue and stack maps")
Reported-by: Hsin-Wei Hung 
Tested-by: Hsin-Wei Hung 
Co-developed-by: Hsin-Wei Hung 
Signed-off-by: Toke Høiland-Jørgensen 
Link: https://lore.kernel.org/r/20230911132815.717240-1-toke@redhat.com
Signed-off-by: Alexei Starovoitov 
Signed-off-by: Sasha Levin

tracing: Increase trace array ref count on enable and filter files

2023-10-10T19:46:35+00:00

[ Upstream commit f5ca233e2e66dc1c249bf07eefa37e34a6c9346a ]

When the trace event enable and filter files are opened, increment the
trace array ref counter, otherwise they can be accessed when the trace
array is being deleted. The ref counter keeps the trace array from being
deleted while those files are opened.

Link: https://lkml.kernel.org/r/20230907024803.456187066@goodmis.org
Link: https://lore.kernel.org/all/1cb3aee2-19af-c472-e265-05176fe9bd84@huawei.com/

Cc: stable@vger.kernel.org
Cc: Masami Hiramatsu 
Cc: Mark Rutland 
Cc: Andrew Morton 
Fixes: 8530dec63e7b4 ("tracing: Add tracing_check_open_get_tr()")
Tested-by: Linux Kernel Functional Testing 
Tested-by: Naresh Kamboju 
Reported-by: Zheng Yejian 
Signed-off-by: Steven Rostedt (Google) 
Signed-off-by: Sasha Levin