]> git.ipfire.org Git - thirdparty/linux.git/commit
perf bpf: Reject oversized BPF metadata events that truncate header.size
authorArnaldo Carvalho de Melo <acme@redhat.com>
Thu, 11 Jun 2026 00:02:25 +0000 (21:02 -0300)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Wed, 17 Jun 2026 11:28:57 +0000 (08:28 -0300)
commit2d6ea0875093da9033fcb62c09a9e2f1de49fe91
tree5e3c7a47a0b9592c11793600ce9dcd7d95a0a6f0
parent10b3c3d63ecc17c6acb855bac5f40367f1115765
perf bpf: Reject oversized BPF metadata events that truncate header.size

bpf_metadata_alloc() computes event_size from the number of BPF metadata
variables and stores it in header.size, which is __u16.  With 204 or
more .rodata variables prefixed "bpf_metadata_", event_size exceeds
65535 and silently truncates.

The truncated header.size causes synthesize_perf_record_bpf_metadata()
to allocate a buffer sized by the truncated value, then memcpy the full
event data into it — a heap buffer overflow.

Add a check that event_size fits in __u16 before proceeding.  BPF
programs with that many metadata variables are exotic enough that
silently dropping the metadata is acceptable.

Reported-by: sashiko-bot <sashiko-bot@kernel.org>
Fixes: ab38e84ba9a80581 ("perf record: collect BPF metadata from existing BPF programs")
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Blake Jones <blakejones@google.com>
Assisted-by: Claude:claude-opus-4.6
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
tools/perf/util/bpf-event.c