On Wed, Jul 08, 2026 at 01:31:45AM +0200, Thomas Gleixner wrote: > On Tue, Jul 07 2026 at 23:28, Mark Brown wrote:
> > I'm seeing failures on arm64 (running a system with 4xA53) in -next in > > this test which bisect to this commit, the commit seems obviously > > plausible but I didn't actually investigate: > > # Starting 2 tests from 1 test cases. > > # RUN global.requeue_single ... > > # futex_requeue.c:28:requeue_single:Expected res (-1) == 0 (0) > > # futex_requeue.c:29:requeue_single:waiter failed errno 110: Connection > > timed out > > # futex_requeue.c:49:requeue_single:Expected > > futex_wait_for_thread(&waiter, _metadata) (110) == 0 (0) > > # futex_requeue.c:50:requeue_single:Wait for thread failed > Do you have /proc/ disabled by chance or is it not accessible for the > test case? I've had another poke at this, I'm still seeing it with today's -next: https://lava.sirena.org.uk/scheduler/job/2973947#L5558 It looks like we start seeing connection timeout reports a bit earlier on: https://lava.sirena.org.uk/scheduler/job/2973947#L4231 though the test succeeds. On another arm64 system I can reproduce with a current kernel in a kvmtool VM (so a rather distinct userspace, though also Debian based), but not directly in the host where the kernel is Debian's 6.12.95+deb13-cloud-arm64. The kvmtool VM does mount /proc read only, remounting it read/write doesn't seem to help. At the very least it feels like we need better diagnostics here.
signature.asc
Description: PGP signature
