On Tue, Mar 3, 2026 at 8:30 PM Samuel Thibault <[email protected]> wrote:
> Re, > > [email protected], le mar. 03 mars 2026 04:08:15 +0000, a ecrit: > > 3. rwlock timeout/wakeup races (H-3, H-4): When a timed rwlock > > operation times out simultaneously with being woken by an unlock, > > reader counts or writer ownership leak, permanently locking the > > rwlock. > > Indeed! > > > This could explain rumpnet hangs. > > Errr... rumpnet doesn't use pthread_rwlock_* It was just guessing, as it often does. I hadn't given it access to the hurd source tree, but it could see the mailing list archives and IRC logs. Also - I forgot to mention this - I haven't reviewed any of the issues it flagged as "LOW" severity. It was almost all memory load/store ordering stuff that doesn't matter on x86, but of course, we have an ARM port. But like I said, I didn't review it, so some of that might be bogus. agape brent
