The patch titled
Subject: fsdax: force clear dirty mark if CoW
has been added to the -mm mm-hotfixes-unstable branch. Its filename is
fsdax-force-clear-dirty-mark-if-cow.patch
This patch will shortly appear at
https://git.kernel.org/pub/scm/linux/kernel/git/akpm/25-new.git/tree/patches/fsdax-force-clear-dirty-mark-if-cow.patch
This patch will later appear in the mm-hotfixes-unstable branch at
git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***
The -mm tree is included into linux-next via the mm-everything
branch at git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
and is updated there every 2-3 working days
------------------------------------------------------
From: Shiyang Ruan <ruansy.fnst@xxxxxxxxxxx>
Subject: fsdax: force clear dirty mark if CoW
Date: Fri, 24 Mar 2023 10:28:00 +0000
XFS allows CoW on non-shared extents to combat fragmentation[1]. The old
non-shared extent could be mwrited before, its dax entry is marked dirty.
To be able to delete this entry, clear its dirty mark before
invalidate_inode_pages2_range().
[1] https://lore.kernel.org/linux-xfs/20230321151339.GA11376@frogsfrogsfrogs/
Link: https://lkml.kernel.org/r/1679653680-2-1-git-send-email-ruansy.fnst@xxxxxxxxxxx
Signed-off-by: Shiyang Ruan <ruansy.fnst@xxxxxxxxxxx>
Cc: Dan Williams <dan.j.williams@xxxxxxxxx>
Cc: Darrick J. Wong <djwong@xxxxxxxxxx>
Cc: Jan Kara <jack@xxxxxxx>
Cc: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
fs/dax.c | 37 +++++++++++++++++++++++++++++++++++++
1 file changed, 37 insertions(+)
--- a/fs/dax.c~fsdax-force-clear-dirty-mark-if-cow
+++ a/fs/dax.c
@@ -781,6 +781,33 @@ out:
return ret;
}
+static int __dax_clear_dirty_range(struct address_space *mapping,
+ pgoff_t start, pgoff_t end)
+{
+ XA_STATE(xas, &mapping->i_pages, start);
+ unsigned int scanned = 0;
+ void *entry;
+
+ xas_lock_irq(&xas);
+ xas_for_each(&xas, entry, end) {
+ entry = get_unlocked_entry(&xas, 0);
+ xas_clear_mark(&xas, PAGECACHE_TAG_DIRTY);
+ xas_clear_mark(&xas, PAGECACHE_TAG_TOWRITE);
+ put_unlocked_entry(&xas, entry, WAKE_NEXT);
+
+ if (++scanned % XA_CHECK_SCHED)
+ continue;
+
+ xas_pause(&xas);
+ xas_unlock_irq(&xas);
+ cond_resched();
+ xas_lock_irq(&xas);
+ }
+ xas_unlock_irq(&xas);
+
+ return 0;
+}
+
/*
* Delete DAX entry at @index from @mapping. Wait for it
* to be unlocked before deleting it.
@@ -1440,6 +1467,16 @@ static loff_t dax_iomap_iter(const struc
* written by write(2) is visible in mmap.
*/
if (iomap->flags & IOMAP_F_NEW || cow) {
+ /*
+ * Filesystem allows CoW on non-shared extents. The src extents
+ * may have been mmapped with dirty mark before. To be able to
+ * invalidate its dax entries, we need to clear the dirty mark
+ * in advance.
+ */
+ if (cow)
+ __dax_clear_dirty_range(iomi->inode->i_mapping,
+ pos >> PAGE_SHIFT,
+ (end - 1) >> PAGE_SHIFT);
invalidate_inode_pages2_range(iomi->inode->i_mapping,
pos >> PAGE_SHIFT,
(end - 1) >> PAGE_SHIFT);
_
Patches currently in -mm which might be from ruansy.fnst@xxxxxxxxxxx are
fsdax-unshare-zero-destination-if-srcmap-is-hole-or-unwritten.patch
fsdax-dedupe-should-compare-the-min-of-two-iters-length.patch
fsdax-force-clear-dirty-mark-if-cow.patch
