Re: aarch64 missing from Revert "perf/x86: Always store regs->ip in perf_callchain_kernel()"

[Jiri Olsa]

On Mon, Jan 12, 2026 at 01:17:53AM -0500, Eugene Loh wrote:
> Thanks for reverting that patch:
> https://lore.kernel.org/r/20251104215405.168643-2-jolsa@xxxxxxxxxx
> https://github.com/torvalds/linux/commit/6d08340d1e354787d6c65a8c3cdd4d41ffb8a5ed
>     Revert "perf/x86: Always store regs->ip in perf_callchain_kernel()"
>     This reverts commit 83f44ae.

hi,
thanks for the report, I did not check aarch64, just x86_64
and there's even more fun with it:
  https://lore.kernel.org/bpf/20260112214940.1222115-1-jolsa@xxxxxxxxxx/T/#mac207867799058c9b0cb273b09e9a2aaf552eebf

> 
> But it seems to me that this issue is not specific to x86.  For example,
> here is aarch64, using a 6.18 kernel:
> 
>     # bpftrace -e 'tracepoint:sched:sched_process_exec { print(kstack()); }'
> 
>         perf_trace_sched_process_exec+16
>         perf_trace_sched_process_exec+16
>         __traceiter_sched_process_exec+80
>         exec_binprm+908
>         bprm_execve.part.0+464
>         bprm_execve+68
>         do_execveat_common.isra.0+456
>         __arm64_sys_execve+76
>         invoke_syscall+112
>         el0_svc_common.constprop.0+264
>         do_el0_svc+36
>         el0_svc+56
>         el0t_64_sync_handler+184
>         el0t_64_sync+412
> 
> The leaf IP is repeated.  Are corresponding changes needed for processors
> other than just x86?
> 

probably.. I think it's very arch code that will need investigation
from somebody fluent in aarch64, adding Will ;-)

thanks,
jirka