On 8/31/2023 11:03 PM, Eduard Zingerman wrote:

On Thu, 2023-08-31 at 13:52 +0300, Eduard Zingerman wrote:

quoted

On Wed, 2023-08-30 at 23:07 +0200, Jiri Olsa wrote:

quoted

On Wed, Aug 30, 2023 at 08:58:11PM +0200, Jiri Olsa wrote:

quoted

hi,
I'm hitting crash below on bpf-next/master when running selftests,
full log and config attached

it seems to be 'test_progs -t sockmap_listen' triggering that

Hi,

I hit it as well, use the following command to reproduce:

   for i in $(seq 1 100); do \
     ./test_progs -a 'sockmap_listen/sockmap VSOCK test_vsock_redir' \
     | grep Summary; \
   done

For what its worth, bisect points to the following commit:
147f3efaa241 ("sched/fair: Implement an EEVDF-like scheduling policy")

Which was merged into bpf-next 3 days ago as a part of:
3ca9a836ff53 ("Merge tag 'sched-core-2023-08-28' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip")

Scheduling changes uncovered some old race condition?

As replied in another mail, I think the issue is introduced by this commit:

405df89dd52c ("bpf, sockmap: Improved check for empty queue")

[...]


.