Thank you for your contribution! Sashiko AI review found 1 potential issue to consider: - [Low] selftests/ve: CPU count taken from sysconf() not the possible mask --
commit e3f3b30a760161024a96299f38de8b9682310200 Author: Pavel Tikhomirov <[email protected]> fs/aio: restore the AIO ring of the dumped size in a pseudosuper VE In a pseudosuper VE, ioctx_alloc() skips the num_possible_cpus() * 4 minimum ring size. This lets CRIU restore a ring of the size it had when dumped. A ring below the node minimum does not use the per-CPU caches of free slots. > diff --git a/fs/aio.c b/fs/aio.c > index 901cb622fe2c..2ff175850761 100644 > --- a/fs/aio.c > +++ b/fs/aio.c [ ... ] > - nr_events = max(nr_events, num_possible_cpus() * 4); > + if (ve_is_super(ve) || !ve->is_pseudosuper) > + nr_events = max(nr_events, num_possible_cpus() * 4); > + else if (nr_events < num_possible_cpus() * 4) > + no_percpu_reqs = true; [Severity: Low] This isn't a bug in this hunk, but the selftest added later in the series (tools/testing/selftests/ve/ve_aio_ring_test.c) takes its CPU count from sysconf(_SC_NPROCESSORS_CONF) in FIXTURE_SETUP(ve_aio_ring). The kernel sizes the ring with num_possible_cpus() here. The sysconf() value comes from the cpuN entries in sysfs, which need not cover the possible mask. Can the test compute wrong expected ring sizes, and so make a wrong SKIP decision, on a host where the possible CPU mask is larger than the set of present CPUs? As reported by Konstantin Khorenko on v1 [1], reading /sys/devices/system/cpu/possible would match what the kernel uses. The v2 code does not do this. [1] https://lore.kernel.org/r/179459890134.57567.3765669749365305224.b4-review@b4 -- Sashiko AI review ยท https://sashiko.vzint.dev/#/patchset/20261116-pr-ptikhomirov-vstor-154129-aio-pseudosuper-ring-v2-0-532d149b4...@virtuozzo.com?part=1 _______________________________________________ Devel mailing list [email protected] https://lists.openvz.org/mailman/listinfo/devel
