Re: [PATCH] tracing/filters: Check perf permissions before resolving .function

[Masami Hiramatsu (Google) <mhiramat@kernel.org>]

On Tue, 06 Oct 2026 15:51:12 -0700, Kyle Zeng <kylebot@openai.com> wrote:
> perf_trace_event_perm() allows tracepoint counters that do not request
> PERF_SAMPLE_RAW without raw tracepoint permissions. A self-targeted,
> disabled event with exclude_kernel=1 can therefore reach the filter
> compiler even at perf_event_paranoid=2.
>
> The .function suffix accepts any field of sizeof(long) and resolves its
> operand through kallsyms_lookup_name() and kallsyms_lookup_size_offset().
> The success or failure of a numeric filter discloses whether an address
> belongs to a known kernel symbol range. On x86-64 this can be used to
> recover the randomized kernel image base. A named filter also exposes
> the resolved symbol range through the counter when the tracepoint field
> is controlled by the caller, as with a syscall argument.
>
> Pass the filter's perf origin to the predicate parser and require
> perf_allow_tracepoint() before resolving a .function operand. This uses
> the same sysctl, initial-namespace capability and LSM policy as raw
> tracepoint access, and closes both the numeric and named-symbol oracles.
> Do not change ordinary perf counting filters or filters created through
> the separately controlled tracefs interfaces.

Good catch!

>
> Fixes: e6745a4da964 ("tracing: Add a way to filter function addresses to function names")
> Cc: stable@vger.kernel.org
> Assisted-by: Codex:gpt-6-astra

nit: This should be 

Assisted-by: LLM

> Signed-off-by: Kyle Zeng <kylebot@openai.com>

Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>

Thanks!


-- 
Masami Hiramatsu (Google) <mhiramat@kernel.org>