[bolt][tests] Skip tests that use perf when perf counters are unavailable

[tstellar]

On the GitHub Action runners, perf always fails with the error below , so we need to skip the perf tests on platforms like this that have limited access to the perf counters.

Access to performance monitoring and observability operations is limited.
Consider adjusting /proc/sys/kernel/perf_event_paranoid setting to open
access to performance monitoring and observability operations for processes
without CAP_PERFMON, CAP_SYS_PTRACE or CAP_SYS_ADMIN Linux capability.
More information can be found at 'Perf events and tool security' document:
https://www.kernel.org/doc/html/latest/admin-guide/perf-security.html
perf_event_paranoid setting is 4:
  -1: Allow use of (almost) all events by all users
      Ignore mlock limit after perf_event_mlock_kb without CAP_IPC_LOCK
>= 0: Disallow raw and ftrace function tracepoint access
>= 1: Disallow CPU event access
>= 2: Disallow kernel profiling
To make the adjusted perf_event_paranoid setting permanent preserve it
in /etc/sysctl.conf (e.g. kernel.perf_event_paranoid = <setting>)

[llvmbot]

@llvm/pr-subscribers-bolt

Author: Tom Stellard (tstellar)

Changes

On the GitHub Action runners, perf always fails with the error below , so we need to skip the perf tests on platforms like this that have limited access to the perf counters.

Access to performance monitoring and observability operations is limited.
Consider adjusting /proc/sys/kernel/perf_event_paranoid setting to open
access to performance monitoring and observability operations for processes
without CAP_PERFMON, CAP_SYS_PTRACE or CAP_SYS_ADMIN Linux capability.
More information can be found at 'Perf events and tool security' document:
https://www.kernel.org/doc/html/latest/admin-guide/perf-security.html
perf_event_paranoid setting is 4:
  -1: Allow use of (almost) all events by all users
      Ignore mlock limit after perf_event_mlock_kb without CAP_IPC_LOCK
>= 0: Disallow raw and ftrace function tracepoint access
>= 1: Disallow CPU event access
>= 2: Disallow kernel profiling
To make the adjusted perf_event_paranoid setting permanent preserve it
in /etc/sysctl.conf (e.g. kernel.perf_event_paranoid = <setting>)

---

Full diff: https://github.com/llvm/llvm-project/pull/107892.diff

1 Files Affected:

• (modified) bolt/test/perf2bolt/lit.local.cfg (+3-2)

diff --git a/bolt/test/perf2bolt/lit.local.cfg b/bolt/test/perf2bolt/lit.local.cfg
index 4ee9ad08cc78a0..9f20f37c6a1d7d 100644
--- a/bolt/test/perf2bolt/lit.local.cfg
+++ b/bolt/test/perf2bolt/lit.local.cfg
@@ -1,4 +1,5 @@
 import shutil
+import subprocess
 
-if shutil.which("perf") is not None:
-    config.available_features.add("perf")
\ No newline at end of file
+if shutil.which("perf") is not None and subprocess.run(["perf", "record", "-e", "cycles:u", "-o", "/dev/null", "--", "perf"], capture_output=True).returncode == 0:
+    config.available_features.add("perf")

[rafaelauler]

Thanks Tom

[tstellar]

/cherry-pick 773353b

[llvmbot]

/pull-request #109072