On 9/17/21 6:30 PM, Darrick J. Wong wrote:
Hi all,
During review of Allison's logged xattrs patchset last cycle, I noticed
that there was an opportunity to clean up some code structure
differences between how regular runtime deferred attributes hold on to
resources across a transaction roll, and how it's done during log
recovery. This series, in cleaning that up, should shorten her
patchset and simplify it a bit.
During regular operation, transactions are allowed to hold up to two
inodes and two buffers across a transaction roll to finish deferred log
items. This implies that log recovery of a log intent item ought to be
able to do the same. However, current log recovery code open-codes
saving only a single inode, because that was all that was required.
With atomic extent swapping and logged extended attributes upon us, it
has become evident that we need to use the same runtime mechanisms
during recovery. Refactor the deferred ops code to use the same
resource capture mechanisms for both.
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
Hi Darrick,
Sorry to for the delay, thanks for putting this together. These
improvements look good to me, I will see if I can get this worked in
underneath delayed attrs and give it a test run.
Allison
kernel git tree:
https://urldefense.com/v3/__https://git.kernel.org/cgit/linux/kernel/git/djwong/xfs-linux.git/log/?h=log-recovery-defer-capture-5.16__;!!ACWV5N9M2RV99hQ!ZyRkandC0OnbO6-EW6AVFwC62a54u8Ki6doYqvl2vJDfgBmkzB4GFZe5s25s8_0XfN1E$
---
fs/xfs/libxfs/xfs_defer.c | 171 +++++++++++++++++++++++++++++++-------------
fs/xfs/libxfs/xfs_defer.h | 38 ++++++++--
fs/xfs/xfs_bmap_item.c | 2 -
fs/xfs/xfs_extfree_item.c | 2 -
fs/xfs/xfs_log_recover.c | 12 +--
fs/xfs/xfs_refcount_item.c | 2 -
fs/xfs/xfs_rmap_item.c | 2 -
fs/xfs/xfs_trans.h | 6 --
8 files changed, 157 insertions(+), 78 deletions(-)
