| Summary: | Warnings when running DTrace testsuite under Valgrind (WARNING: unhandled eBPF command 23) | ||
|---|---|---|---|
| Product: | [Developer tools] valgrind | Reporter: | Sam James <sam> |
| Component: | general | Assignee: | Julian Seward <jseward> |
| Status: | REPORTED --- | ||
| Severity: | normal | CC: | mark |
| Priority: | NOR | ||
| Version First Reported In: | 3.24 GIT | ||
| Target Milestone: | --- | ||
| Platform: | Other | ||
| OS: | Linux | ||
| See Also: |
https://bugs.kde.org/show_bug.cgi?id=492050 https://bugs.kde.org/show_bug.cgi?id=478774 |
||
| Latest Commit: | Version Fixed/Implemented In: | ||
| Sentry Crash Report: | |||
|
Description
Sam James
2024-08-24 10:10:03 UTC
From looking at /usr/include/linux/bpf.h's bpf_cmd in linux-headers-6.10, it might be BPF_MAP_FREEZE? I'm not yet sure how to produce a standalone testcase as I'm not familiar enough with BPF yet. Mark, I think you're right and it's actually https://bugs.kde.org/show_bug.cgi?id=478774, because I investigated the uninit var warning and it turns out to be the same thing (Valgrind didn't know it was init'd). Let me try the patch from https://bugs.kde.org/show_bug.cgi?id=478774#c4... I'm surprised but it *didn't* help: ``` $ sudo valgrind --error-exitcode=1 --exit-on-first-error=yes -q --track-origins=yes dtrace -DARCH_x86_64 -I/usr/lib64/dtrace/include -xerrtags -s test/unittest/multiaggs/err.D_PRINTA_AGGKEY.d --2155-- WARNING: unhandled eBPF command 23 --2155-- WARNING: unhandled eBPF command 23 ==2155== Conditional jump or move depends on uninitialised value(s) ==2155== at 0x4852415: strlen (vg_replace_strmem.c:505) ==2155== by 0x48E5164: dt_module_lookup_by_name (dt_module.c:173) ==2155== by 0x48AADDC: dt_btf_get_module_ids (dt_btf.c:944) ==2155== by 0x48E95E6: dt_vopen (dt_open.c:1160) ==2155== by 0x1098F7: main (dtrace.c:1102) ==2155== Uninitialised value was created by a stack allocation ==2155== at 0x48AACEA: dt_btf_get_module_ids (dt_btf.c:908) ==2155== ==2155== ==2155== Exit program on first error (--exit-on-first-error=yes) ``` (From my primitive knowledge of BPF, I think we do need that patch though.) |