Hi all,
Ritesh and Eric separately reported crashes in XFS's hook function for
CPU hot remove if the remove event races with a filesystem being
mounted. I also noticed via generic/650 that once in a while the log
will shut down over an apparent overrun of a transaction reservation;
this turned out to be due to CIL percpu list aggregation failing to pick
up the percpu list items from a dying CPU.
Either way, the solution here is to eliminate the need for a CPU dying
hook by using a private cpumask to track which CPUs have added to their
percpu lists directly, and iterating with that mask. This fixes the log
problems and (I think) solves a theoretical UAF bug in the inodegc code
too.
v2: fix a few put_cpu uses, add necessary memory barriers, and use
atomic cpumask operations
If you're going to start using this code, I strongly recommend pulling
from my git trees, which are linked below.
This has been lightly tested with fstests. 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=fix-percpu-lists-6.6
---
fs/xfs/xfs_icache.c | 78 ++++++++++++++--------------------------
fs/xfs/xfs_icache.h | 1 -
fs/xfs/xfs_log_cil.c | 52 ++++++++-------------------
fs/xfs/xfs_log_priv.h | 14 +++----
fs/xfs/xfs_mount.h | 7 ++--
fs/xfs/xfs_super.c | 86 +-------------------------------------------
include/linux/cpuhotplug.h | 1 -
7 files changed, 56 insertions(+), 183 deletions(-)
