Hi all,
This series fixes some assorted bugs in the inodegc code that have been
lurking for a while.
The first is that inodes that are ready to be inactivated are put on
per-cpu lockless lists and a per-cpu worker is scheduled to clear that
list eventually. Unfortunately, WORK_CPU_UNBOUND doesn't guarantee that
a worker will actually be scheduled on that CPU, so we need to force
this by specifying the CPU explicitly.
The second problem is that that xfs_inodegc_stop races with other
threads that are trying to add inodes to the percpu list and schedule
the inodegc workers. The solution here is to drain the inodegc lists
by scheduling workers immediately, flushing the workqueue, and
scheduling if any new inodes have appeared.
We also disable the broken fscounters usage of inodegc_stop by neutering
the whole scrubber for now because the proper fixes for it are in the
next batch of online repair patches for 6.5, and the code is still
marked experimental.
v2: document the new locking requirements of xfs_inodegc_{start,stop}
and why we cannot use drain_workqueue
If you're going to start using this mess, you probably ought to just
pull from my git trees, which are linked below.
This is an extraordinary way to destroy everything. Enjoy!
Comments and questions are, as always, welcome.
--D
kernel git tree:
https://git.kernel.org/cgit/linux/kernel/git/djwong/xfs-linux.git/log/?h=inodegc-fixes-6.4
---
fs/xfs/scrub/common.c | 26 --------------------------
fs/xfs/scrub/common.h | 2 --
fs/xfs/scrub/fscounters.c | 13 ++++++-------
fs/xfs/scrub/scrub.c | 2 --
fs/xfs/scrub/scrub.h | 1 -
fs/xfs/scrub/trace.h | 1 -
fs/xfs/xfs_icache.c | 40 +++++++++++++++++++++++++++++++++-------
fs/xfs/xfs_mount.h | 3 +++
fs/xfs/xfs_super.c | 3 +++
9 files changed, 45 insertions(+), 46 deletions(-)
