[PATCH v1 3/4] perf record: Skip don't fail for events that don't open

[Ian Rogers]

On Tue, Nov 12, 2024 at 11:53 AM Leo Yan <leo.yan at arm.com> wrote:
>
> On Sat, Oct 26, 2024 at 05:17:57AM -0700, Ian Rogers wrote:
> >
> > Whilst for many tools it is an expected behavior that failure to open
> > a perf event is a failure, ARM decided to name PMU events the same as
> > legacy events and then failed to rename such events on a server uncore
> > SLC PMU. As perf's default behavior when no PMU is specified is to
> > open the event on all PMUs that advertise/"have" the event, this
> > yielded failures when trying to make the priority of legacy and
> > sysfs/json events uniform - something requested by RISC-V and ARM. A
> > legacy event user on ARM hardware may find their event opened on an
> > uncore PMU which for perf record will fail. Arnaldo suggested skipping
> > such events which this patch implements. Rather than have the skipping
> > conditional on running on ARM, the skipping is done on all
> > architectures as such a fundamental behavioral difference could lead
> > to problems with tools built/depending on perf.
> >
> > An example of perf record failing to open events on x86 is:
> > ```
> > $ perf record -e data_read,cycles,LLC-prefetch-read -a sleep 0.1
> > Error:
> > Failure to open event 'data_read' on PMU 'uncore_imc_free_running_0' which will be removed.
> > The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (data_read).
> > "dmesg | grep -i perf" may provide additional information.
> >
> > Error:
> > Failure to open event 'data_read' on PMU 'uncore_imc_free_running_1' which will be removed.
> > The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (data_read).
> > "dmesg | grep -i perf" may provide additional information.
> >
> > Error:
> > Failure to open event 'LLC-prefetch-read' on PMU 'cpu' which will be removed.
> > The LLC-prefetch-read event is not supported.
> > [ perf record: Woken up 1 times to write data ]
> > [ perf record: Captured and wrote 2.188 MB perf.data (87 samples) ]
> >
> > $ perf report --stats
> > Aggregated stats:
> >                TOTAL events:      17255
> >                 MMAP events:        284  ( 1.6%)
> >                 COMM events:       1961  (11.4%)
> >                 EXIT events:          1  ( 0.0%)
> >                 FORK events:       1960  (11.4%)
> >               SAMPLE events:         87  ( 0.5%)
> >                MMAP2 events:      12836  (74.4%)
> >              KSYMBOL events:         83  ( 0.5%)
> >            BPF_EVENT events:         36  ( 0.2%)
> >       FINISHED_ROUND events:          2  ( 0.0%)
> >             ID_INDEX events:          1  ( 0.0%)
> >           THREAD_MAP events:          1  ( 0.0%)
> >              CPU_MAP events:          1  ( 0.0%)
> >            TIME_CONV events:          1  ( 0.0%)
> >        FINISHED_INIT events:          1  ( 0.0%)
> > cycles stats:
> >               SAMPLE events:         87
> > ```
>
> Thanks for James reminding me.  Tested on AVA platform:
>
> # tree /sys/bus/event_source/devices/arm_dsu_*/events
>   ...
>   /sys/bus/event_source/devices/arm_dsu_9/events
>   ├── bus_access
>   ├── bus_cycles
>   ├── cycles
>   ├── l3d_cache
>   ├── l3d_cache_allocate
>   ├── l3d_cache_refill
>   ├── l3d_cache_wb
>   └── memory_error
>
> # ./perf record -- sleep 0.1
>  Error:
>  Failure to open event 'cycles:PH' on PMU 'arm_dsu_0' which will be
>  removed.
>  cycles:PH: PMU Hardware doesn't support sampling/overflow-interrupts.
>  Try 'perf stat'
>  Error:
>  Failure to open event 'cycles:PH' on PMU 'arm_dsu_1' which will be
>  removed.
>  cycles:PH: PMU Hardware doesn't support sampling/overflow-interrupts.
>  Try 'perf stat'
>  ...
>  Error:
>  Failure to open event 'cycles:PH' on PMU 'arm_dsu_15' which will be
>  removed.
>  cycles:PH: PMU Hardware doesn't support sampling/overflow-interrupts.
>  Try 'perf stat'
>  [ perf record: Woken up 1 times to write data ]
>  [ perf record: Captured and wrote 0.008 MB perf.data (8 samples) ]
>
> # ./perf report --stats
>
>  Aggregated stats:
>                 TOTAL events:         67
>                  MMAP events:         40  (59.7%)
>                  COMM events:          1  ( 1.5%)
>                SAMPLE events:          8  (11.9%)
>               KSYMBOL events:          6  ( 9.0%)
>             BPF_EVENT events:          6  ( 9.0%)
>        FINISHED_ROUND events:          1  ( 1.5%)
>              ID_INDEX events:          1  ( 1.5%)
>            THREAD_MAP events:          1  ( 1.5%)
>               CPU_MAP events:          1  ( 1.5%)
>             TIME_CONV events:          1  ( 1.5%)
>         FINISHED_INIT events:          1  ( 1.5%)
>  cycles:P stats:
>                SAMPLE events:          8
>
> # ./perf stat -- sleep 0.1
>
>  Performance counter stats for 'sleep 0.1':
>
>               0.87 msec task-clock                       #    0.009 CPUs utilized
>                  1      context-switches                 #    1.148 K/sec
>                  0      cpu-migrations                   #    0.000 /sec
>                 52      page-faults                      #   59.685 K/sec
>            877,835      instructions                     #    1.14  insn per cycle
>                                                   #    0.25  stalled cycles per insn
>            772,102      cycles                           #  886.210 M/sec
>            191,914      stalled-cycles-frontend          #   24.86% frontend cycles idle
>            219,183      stalled-cycles-backend           #   28.39% backend cycles idle
>            184,099      branches                         #  211.307 M/sec
>              8,548      branch-misses                    #    4.64% of all branches
>
>        0.101623529 seconds time elapsed
>
>        0.001645000 seconds user
>        0.000000000 seconds sys
>
> Tested-by: Leo Yan <leo.yan at arm.com>

Thanks Leo! As the Tested-by makes sense only if you've applied all 4
patches, which your testing and James' testing shows you've both done,
I'll add the tags to all 4 patches. I'll do likewise with Atish,
rebase and resend the patches.

Thanks,
Ian