On 4/7/22 9:38 AM, Yuntao Wang wrote:

The bpf_get_stackid() function may also return 0 on success.

Can you add couple of sentences to describe what this patch does? bpf_get_stackid() may also return doesn't really say anything about why this patch is needed.

Signed-off-by: Yuntao Wang ytcoode@gmail.com

tools/testing/selftests/bpf/progs/perf_event_stackmap.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/bpf/progs/perf_event_stackmap.c b/tools/testing/selftests/bpf/progs/perf_event_stackmap.c index b3fcb5274ee0..f793280a3238 100644 --- a/tools/testing/selftests/bpf/progs/perf_event_stackmap.c +++ b/tools/testing/selftests/bpf/progs/perf_event_stackmap.c @@ -35,10 +35,10 @@ int oncpu(void *ctx) long val; val = bpf_get_stackid(ctx, &stackmap, 0);

• if (val > 0)

• if (val >= 0) stackid_kernel = 2; val = bpf_get_stackid(ctx, &stackmap, BPF_F_USER_STACK);

• if (val > 0)

• if (val >= 0) stackid_user = 2;

trace = bpf_map_lookup_elem(&stackdata_map, &key);

Linux 5.18-rc1 shows a couple of more bpf_get_stackid() in this function. Removed in bpf-next - I assume.

The change is good. I would like to see it explained better in the commit log.

With the commit log fixed to explain why this change is needed and what happens if val equals to 0 condition isn't checked:

Reviewed-by: Shuah Khan skhan@linuxfoundation.org

thanks, -- Shuah