linux.git/kernel/trace/trace_osnoise.c, branch v5.15.71

tracing/osnoise: Force quiescent states while tracing

2022-03-16T13:23:43+00:00

commit caf4c86bf136845982c5103b2661751b40c474c0 upstream.

At the moment running osnoise on a nohz_full CPU or uncontested FIFO
priority and a PREEMPT_RCU kernel might have the side effect of
extending grace periods too much. This will entice RCU to force a
context switch on the wayward CPU to end the grace period, all while
introducing unwarranted noise into the tracer. This behaviour is
unavoidable as overly extending grace periods might exhaust the system's
memory.

This same exact problem is what extended quiescent states (EQS) were
created for, conversely, rcu_momentary_dyntick_idle() emulates them by
performing a zero duration EQS. So let's make use of it.

In the common case rcu_momentary_dyntick_idle() is fairly inexpensive:
atomically incrementing a local per-CPU counter and doing a store. So it
shouldn't affect osnoise's measurements (which has a 1us granularity),
so we'll call it unanimously.

The uncommon case involve calling rcu_momentary_dyntick_idle() after
having the osnoise process:

 - Receive an expedited quiescent state IPI with preemption disabled or
   during an RCU critical section. (activates rdp->cpu_no_qs.b.exp
   code-path).

 - Being preempted within in an RCU critical section and having the
   subsequent outermost rcu_read_unlock() called with interrupts
   disabled. (t->rcu_read_unlock_special.b.blocked code-path).

Neither of those are possible at the moment, and are unlikely to be in
the future given the osnoise's loop design. On top of this, the noise
generated by the situations described above is unavoidable, and if not
exposed by rcu_momentary_dyntick_idle() will be eventually seen in
subsequent rcu_read_unlock() calls or schedule operations.

Link: https://lkml.kernel.org/r/20220307180740.577607-1-nsaenzju@redhat.com

Cc: stable@vger.kernel.org
Fixes: bce29ac9ce0b ("trace: Add osnoise tracer")
Signed-off-by: Nicolas Saenz Julienne 
Acked-by: Paul E. McKenney 
Acked-by: Daniel Bristot de Oliveira 
Signed-off-by: Steven Rostedt (Google) 
Signed-off-by: Greg Kroah-Hartman

tracing/osnoise: Make osnoise_main to sleep for microseconds

2022-03-16T13:23:42+00:00

[ Upstream commit dd990352f01ee9a6c6eee152e5d11c021caccfe4 ]

osnoise's runtime and period are in the microseconds scale, but it is
currently sleeping in the millisecond's scale. This behavior roots in the
usage of hwlat as the skeleton for osnoise.

Make osnoise to sleep in the microseconds scale. Also, move the sleep to
a specialized function.

Link: https://lkml.kernel.org/r/302aa6c7bdf2d131719b22901905e9da122a11b2.1645197336.git.bristot@kernel.org

Cc: Ingo Molnar 
Signed-off-by: Daniel Bristot de Oliveira 
Signed-off-by: Steven Rostedt (Google) 
Signed-off-by: Sasha Levin

tracing/osnoise: Properly unhook events if start_per_cpu_kthreads() fails

2022-01-27T10:05:18+00:00

commit 0878355b51f5f26632e652c848a8e174bb02d22d upstream.

If start_per_cpu_kthreads() called from osnoise_workload_start() returns
error, event hooks are left in broken state: unhook_irq_events() called
but unhook_thread_events() and unhook_softirq_events() not called, and
trace_osnoise_callback_enabled flag not cleared.

On the next tracer enable, hooks get not installed due to
trace_osnoise_callback_enabled flag.

And on the further tracer disable an attempt to remove non-installed
hooks happened, hitting a WARN_ON_ONCE() in tracepoint_remove_func().

Fix the error path by adding the missing part of cleanup.
While at this, introduce osnoise_unhook_events() to avoid code
duplication between this error path and normal tracer disable.

Link: https://lkml.kernel.org/r/20220109153459.3701773-1-nikita.yushchenko@virtuozzo.com

Cc: stable@vger.kernel.org
Fixes: bce29ac9ce0b ("trace: Add osnoise tracer")
Acked-by: Daniel Bristot de Oliveira 
Signed-off-by: Nikita Yushchenko 
Signed-off-by: Steven Rostedt 
Signed-off-by: Greg Kroah-Hartman

tracing: Disable "other" permission bits in the tracefs files

2021-11-18T18:16:15+00:00

[ Upstream commit 21ccc9cd72116289469e5519b6159c675a2fa58f ]

When building the files in the tracefs file system, do not by default set
any permissions for OTH (other). This will make it easier for admins who
want to define a group for accessing tracefs and not having to first
disable all the permission bits for "other" in the file system.

As tracing can leak sensitive information, it should never by default
allowing all users access. An admin can still set the permission bits for
others to have access, which may be useful for creating a honeypot and
seeing who takes advantage of it and roots the machine.

Link: https://lkml.kernel.org/r/20210818153038.864149276@goodmis.org

Signed-off-by: Steven Rostedt (VMware) 
Signed-off-by: Sasha Levin

tracing/osnoise: Fix missed cpus_read_unlock() in start_per_cpu_kthreads()

2021-09-08T19:10:24+00:00

When start_kthread() return error, the cpus_read_unlock() need
to be called.

Link: https://lkml.kernel.org/r/20210831022919.27630-1-qiang.zhang@windriver.com

Cc: 
Fixes: c8895e271f79 ("trace/osnoise: Support hotplug operations")
Acked-by: Daniel Bristot de Oliveira 
Signed-off-by: Qiang.Zhang 
Signed-off-by: Steven Rostedt (VMware)

tracing: Replace deprecated CPU-hotplug functions.

2021-08-17T19:47:14+00:00

The functions get_online_cpus() and put_online_cpus() have been
deprecated during the CPU hotplug rework. They map directly to
cpus_read_lock() and cpus_read_unlock().

Replace deprecated CPU-hotplug functions with the official version.
The behavior remains unchanged.

Link: https://lkml.kernel.org/r/20210803141621.780504-37-bigeasy@linutronix.de

Cc: Peter Zijlstra 
Cc: Ingo Molnar 
Acked-by: Daniel Bristot de Oliveira 
Signed-off-by: Sebastian Andrzej Siewior 
Signed-off-by: Steven Rostedt (VMware)

trace/osnoise: Print a stop tracing message

2021-08-12T17:35:56+00:00

When using osnoise/timerlat with stop tracing, sometimes it is
not clear in which CPU the stop condition was hit, mainly
when using some extra events.

Print a message informing in which CPU the trace stopped, like
in the example below:

          -0       [006] d.h.  2932.676616: #1672599 context    irq timer_latency     34689 ns
          -0       [006] dNh.  2932.676618: irq_noise: local_timer:236 start 2932.676615639 duration 2391 ns
          -0       [006] dNh.  2932.676620: irq_noise: virtio0-output.0:47 start 2932.676620180 duration 86 ns
          -0       [003] d.h.  2932.676621: #1673374 context    irq timer_latency      1200 ns
          -0       [006] d...  2932.676623: thread_noise: swapper/6:0 start 2932.676615964 duration 4339 ns
          -0       [003] dNh.  2932.676623: irq_noise: local_timer:236 start 2932.676620597 duration 1881 ns
          -0       [006] d...  2932.676623: sched_switch: prev_comm=swapper/6 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=timerlat/6 next_pid=852 next_prio=4
      timerlat/6-852     [006] ....  2932.676623: #1672599 context thread timer_latency     41931 ns
          -0       [003] d...  2932.676623: thread_noise: swapper/3:0 start 2932.676620854 duration 880 ns
          -0       [003] d...  2932.676624: sched_switch: prev_comm=swapper/3 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=timerlat/3 next_pid=849 next_prio=4
      timerlat/6-852     [006] ....  2932.676624: timerlat_main: stop tracing hit on cpu 6
      timerlat/3-849     [003] ....  2932.676624: #1673374 context thread timer_latency      4310 ns

Link: https://lkml.kernel.org/r/b30a0d7542adba019185f44ee648e60e14923b11.1626598844.git.bristot@kernel.org

Cc: Tom Zanussi 
Cc: Namhyung Kim 
Cc: Masami Hiramatsu 
Signed-off-by: Daniel Bristot de Oliveira 
Signed-off-by: Steven Rostedt (VMware)

trace/timerlat: Add a header with PREEMPT_RT additional fields

2021-08-12T17:35:56+00:00

Some extra flags are printed to the trace header when using the
PREEMPT_RT config. The extra flags are: need-resched-lazy,
preempt-lazy-depth, and migrate-disable.

Without printing these fields, the timerlat specific fields are
shifted by three positions, for example:

 # tracer: timerlat
 #
 #                                _-----=> irqs-off
 #                               / _----=> need-resched
 #                              | / _---=> hardirq/softirq
 #                              || / _--=> preempt-depth
 #                              || /
 #                              ||||             ACTIVATION
 #           TASK-PID      CPU# ||||   TIMESTAMP    ID            CONTEXT                LATENCY
 #              | |         |   ||||      |         |                  |                       |
           -0       [000] d..h...  3279.798871: #1     context    irq timer_latency       830 ns
            <...>-807     [000] .......  3279.798881: #1     context thread timer_latency     11301 ns

Add a new header for timerlat with the missing fields, to be used
when the PREEMPT_RT is enabled.

Link: https://lkml.kernel.org/r/babb83529a3211bd0805be0b8c21608230202c55.1626598844.git.bristot@kernel.org

Cc: Tom Zanussi 
Cc: Namhyung Kim 
Cc: Masami Hiramatsu 
Signed-off-by: Daniel Bristot de Oliveira 
Signed-off-by: Steven Rostedt (VMware)

trace/osnoise: Add a header with PREEMPT_RT additional fields

2021-08-12T17:35:56+00:00

Some extra flags are printed to the trace header when using the
PREEMPT_RT config. The extra flags are: need-resched-lazy,
preempt-lazy-depth, and migrate-disable.

Without printing these fields, the osnoise specific fields are
shifted by three positions, for example:

 # tracer: osnoise
 #
 #                                _-----=> irqs-off
 #                               / _----=> need-resched
 #                              | / _---=> hardirq/softirq
 #                              || / _--=> preempt-depth                            MAX
 #                              || /                                             SINGLE      Interference counters:
 #                              ||||               RUNTIME      NOISE  %% OF CPU  NOISE    +-----------------------------+
 #           TASK-PID      CPU# ||||   TIMESTAMP    IN US       IN US  AVAILABLE  IN US     HW    NMI    IRQ   SIRQ THREAD
 #              | |         |   ||||      |           |             |    |            |      |      |      |      |      |
            <...>-741     [000] .......  1105.690909: 1000000        234  99.97660      36     21      0   1001     22      3
            <...>-742     [001] .......  1105.691923: 1000000        281  99.97190     197      7      0   1012     35     14
            <...>-743     [002] .......  1105.691958: 1000000       1324  99.86760     118     11      0   1016    155    143
            <...>-744     [003] .......  1105.691998: 1000000        109  99.98910      21      4      0   1004     33      7
            <...>-745     [004] .......  1105.692015: 1000000       2023  99.79770      97     37      0   1023     52     18

Add a new header for osnoise with the missing fields, to be used
when the PREEMPT_RT is enabled.

Link: https://lkml.kernel.org/r/1f03289d2a51fde5a58c2e7def063dc630820ad1.1626598844.git.bristot@kernel.org

Cc: Tom Zanussi 
Cc: Namhyung Kim 
Cc: Masami Hiramatsu 
Signed-off-by: Daniel Bristot de Oliveira 
Signed-off-by: Steven Rostedt (VMware)

trace/timerlat: Fix indentation on timerlat_main()

2021-06-29T20:37:50+00:00

Dan Carpenter reported that:

 The patch a955d7eac177: "trace: Add timerlat tracer" from Jun 22,
 2021, leads to the following static checker warning:

	kernel/trace/trace_osnoise.c:1400 timerlat_main()
	warn: inconsistent indenting

here:
  1389          while (!kthread_should_stop()) {
  1390                  now = ktime_to_ns(hrtimer_cb_get_time(&tlat->timer));
  1391                  diff = now - tlat->abs_period;
  1392
  1393                  s.seqnum = tlat->count;
  1394                  s.timer_latency = diff;
  1395                  s.context = THREAD_CONTEXT;
  1396
  1397                  trace_timerlat_sample(&s);
  1398
  1399  #ifdef CONFIG_STACKTRACE
  1400          if (osnoise_data.print_stack)
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
	    This should be indented another tab?

  1401                  if (osnoise_data.print_stack <= time_to_us(diff))
  1402                          timerlat_dump_stack();
  1403  #endif /* CONFIG_STACKTRACE */
  1404
  1405                  tlat->tracing_thread = false;
  1406                  if (osnoise_data.stop_tracing_total)
  1407                          if (time_to_us(diff) >= osnoise_data.stop_tracing_total)
  1408                                  osnoise_stop_tracing();
  1409
  1410                  wait_next_period(tlat);
  1411          }

And the static checker is right. Fix the indentation.

Link: https://lkml.kernel.org/r/3d5d8c9258fbdcfa9d3c7362941b3d13a2a28d9d.1624986368.git.bristot@redhat.com

Cc: Ingo Molnar 
Cc: Daniel Bristot de Oliveira 
Cc: Dan Carpenter 
Cc: linux-kernel@vger.kernel.org
Fixes: a955d7eac177 ("trace: Add timerlat tracer")
Reported-by: Dan Carpenter 
Signed-off-by: Daniel Bristot de Oliveira 
Signed-off-by: Steven Rostedt (VMware)