pá 17. 1. 2025 v 2:18 odesílatel Steven Rostedt [off-list ref] napsal:

I tested this out, and it failed the last test:

Thank you for testing this.

make[1]: Leaving directory '/work/git/linux.git/tools/tracing/rtla'
RTLA=/work/git/linux.git/tools/tracing/rtla/rtla prove -o -f tests/
tests/hwnoise.t ... ok
tests/osnoise.t ... ok
tests/timerlat.t .. 9/9
not ok 9 - hist test in nanoseconds
#
tests/timerlat.t .. Failed 1/9 subtests

Test Summary Report
-------------------
tests/timerlat.t (Wstat: 0 Tests: 9 Failed: 1)
  Failed test:  9
Files=3, Tests=20, 451 wallclock secs ( 0.07 usr  0.11 sys + 10.57 cusr 74.09 csys = 84.84 CPU)
Result: FAIL
make: *** [Makefile:89: check] Error 1

I have not seen this when testing the patch. It looks like rtla exits
without any output, I will send a v2 that makes the test engine print
the error code in that case, it will be useful not only for this.

That being said, it's not like nanosecond resolution rtla-timerlat is
without bugs:

# RTLA timerlat histogram
# Time unit is nanoseconds (ns)
# Duration:   0 00:00:02
Index   IRQ-000   Thr-000   Usr-000
over:         1         0         0
count:        1         0         0
min:       6102         -         -
avg:       6102         -         -
max:       6102         -         -
ALL:        IRQ       Thr       Usr
count:        1         0         0
min:       6102         -         -
avg:       6102         -         -
max:       6102         -         -
rtla timerlat hit stop tracing
## CPU 0 hit stop tracing, analyzing it ##
 IRQ handler delay:
1442069690324.42 us (23632738287847.00 %)
 IRQ latency:                                               6.10 us
 Blocking thread:
                               :0
------------------------------------------------------------------------
    IRQ latency:                                            6.10 us (100%)
Max timerlat IRQ latency from idle: 6.10 us in cpu 0

Maybe after enough time repeating the command I'm also going to hit
the bug you saw.

Tomas