On Tue, 2025-07-22 at 16:19 +0400, Alex Markuze wrote: > Hi Slava, > > Thanks for the patch. > > The fix for the race condition in ceph_check_delayed_caps() is > correct > and necessary. The systematic change to use atomic bit operations > like > set_bit() and clear_bit() with the proper memory barriers is a > significant improvement for safety and readability. > > One minor critique for a follow-up patch: > > The refactoring in fs/ceph/super.h to use named _BIT definitions is a > great idea, but the cleanup is incomplete. Several definitions were > not converted and still use hardcoded bit-shift numbers . For > example, > > CEPH_I_POOL_RD, CEPH_I_POOL_WR, and CEPH_I_ODIRECT still use (1 << > 4), > (1 << 5), and (1 << 11) respectively. It would be good to finish this > refactoring for consistency. > > Makes sense. Let me rework the patch. Thanks, Slava.
