diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2018-08-14 18:56:02 +0300 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2018-08-14 18:56:02 +0300 |
commit | 781fca5b104693bc9242199cc47c690dcaf6a4cb (patch) | |
tree | d216d4299ae5715331a535c84bab390a907bebd6 /fs/xfs/xfs_trans.c | |
parent | 10f3e23f07cb0c20f9bcb77a5b5a7eb2a1b2a2fe (diff) | |
parent | 01239d77b9dd978863d1a75f0d095ab942a1fe66 (diff) | |
download | linux-781fca5b104693bc9242199cc47c690dcaf6a4cb.tar.xz |
Merge tag 'xfs-4.19-merge-6' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Pull xfs updates from Darrick Wong:
"This is the second part of the XFS changes for 4.19.
The biggest changes are the removal of buffer heads frm XFS, a massive
reworking of the deferred transaction operations handling code, the
removal of the long defunct barrier/nobarrier mount options, and the
addition of a few more online repair functions.
Summary:
- Use extent maps to track pagecache page status instead of
bufferhead state.
- Refactor pagecache read and write paths to use the new iomap
library functions, which enable us to drop the old bufferhead code
for pagesize == blocksize filesystems.
- Set up parallel per-block-per-page metadata to track subpage
information that was tracked by buffer heads, which enables us to
drop the old bufferhead code for pagesize > blocksize filesystems.
- Tie a deferred ops control structure to a transaction so that we
can take advantage of an upper-level dfops without having to plumb
pointer passing through the code.
- Refactor the deferred ops code to track deferred ops as part of the
transaction structure (instead of as a separate data structure) so
that we can simplify the scoping rules around defer_ops.
- Refactor twisty delwri buffer submission code to avoid deadlocks.
- Shorten and fix indenting problems in the scrub code.
- Detect obviously bad summary counts at mount and fix them.
- Directly associate deferred ops control structure with a
transaction so that callers no longer have to manage it themselves.
- Remove a couple of IRIX-era inode macros.
- Remove the long-deprecated 'barrier' and 'nobarrier' mount options.
- Clean up the inode fork structure a bit.
- Check for bad fs summary counter values in the superblock.
- Reduce COW fork lookups during writeback.
- Refactor the deferred ops control structures into the transaction
structure, thereby eliminating the need for transaction users to
handle the deferred ops as a separate data structure.
- Add the ability to repair AG headers online.
- Fix a crash due to insufficient return value checking.
- Various fixes and cleanups"
* tag 'xfs-4.19-merge-6' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux: (155 commits)
xfs: fix a null pointer dereference in xfs_bmap_extents_to_btree
xfs: remove b_last_holder & associated macros
iomap: Switch to offset_in_page for clarity
xfs: Close race between direct IO and xfs_break_layouts()
xfs: repair the AGI
xfs: repair the AGFL
xfs: repair the AGF
xfs: remove dead error handling code in xfs_dquot_disk_alloc()
xfs: use WRITE_ONCE to update if_seq
xfs: fix a comment in xfs_log_reserve
xfs: only validate summary counts on primary superblock
xfs: substitute spaces with tabs
xfs: fold dfops into the transaction
xfs: always defer agfl block frees
xfs: pass transaction to xfs_defer_add()
xfs: replace xfs_defer_ops ->dop_pending with on-stack list
xfs: cancel dfops on xfs_defer_finish() error
xfs: clean out superfluous dfops dop params/vars
xfs: drop dop param from xfs_defer_op_type ->finish_item() callback
xfs: automatic dfops inode relogging
...
Diffstat (limited to 'fs/xfs/xfs_trans.c')
-rw-r--r-- | fs/xfs/xfs_trans.c | 26 |
1 files changed, 22 insertions, 4 deletions
diff --git a/fs/xfs/xfs_trans.c b/fs/xfs/xfs_trans.c index 524f543c5b82..bedc5a5133a5 100644 --- a/fs/xfs/xfs_trans.c +++ b/fs/xfs/xfs_trans.c @@ -100,6 +100,8 @@ xfs_trans_dup( ntp->t_mountp = tp->t_mountp; INIT_LIST_HEAD(&ntp->t_items); INIT_LIST_HEAD(&ntp->t_busy); + INIT_LIST_HEAD(&ntp->t_dfops); + ntp->t_firstblock = NULLFSBLOCK; ASSERT(tp->t_flags & XFS_TRANS_PERM_LOG_RES); ASSERT(tp->t_ticket != NULL); @@ -118,7 +120,9 @@ xfs_trans_dup( ntp->t_rtx_res = tp->t_rtx_res - tp->t_rtx_res_used; tp->t_rtx_res = tp->t_rtx_res_used; ntp->t_pflags = tp->t_pflags; - ntp->t_agfl_dfops = tp->t_agfl_dfops; + + /* move deferred ops over to the new tp */ + xfs_defer_move(ntp, tp); xfs_trans_dup_dqinfo(tp, ntp); @@ -273,6 +277,8 @@ xfs_trans_alloc( tp->t_mountp = mp; INIT_LIST_HEAD(&tp->t_items); INIT_LIST_HEAD(&tp->t_busy); + INIT_LIST_HEAD(&tp->t_dfops); + tp->t_firstblock = NULLFSBLOCK; error = xfs_trans_reserve(tp, resp, blocks, rtextents); if (error) { @@ -914,12 +920,21 @@ __xfs_trans_commit( int error = 0; int sync = tp->t_flags & XFS_TRANS_SYNC; - ASSERT(!tp->t_agfl_dfops || - !xfs_defer_has_unfinished_work(tp->t_agfl_dfops) || regrant); - trace_xfs_trans_commit(tp, _RET_IP_); /* + * Finish deferred items on final commit. Only permanent transactions + * should ever have deferred ops. + */ + WARN_ON_ONCE(!list_empty(&tp->t_dfops) && + !(tp->t_flags & XFS_TRANS_PERM_LOG_RES)); + if (!regrant && (tp->t_flags & XFS_TRANS_PERM_LOG_RES)) { + error = xfs_defer_finish_noroll(&tp); + if (error) + goto out_unreserve; + } + + /* * If there is nothing to be logged by the transaction, * then unlock all of the items associated with the * transaction and free the transaction structure. @@ -1008,6 +1023,9 @@ xfs_trans_cancel( trace_xfs_trans_cancel(tp, _RET_IP_); + if (tp->t_flags & XFS_TRANS_PERM_LOG_RES) + xfs_defer_cancel(tp); + /* * See if the caller is relying on us to shut down the * filesystem. This happens in paths where we detect |