Skip to contentSkip to conversation

fixes for uncached IO

PATCH Applied Jingbo Xu 5 participants 11 messages Feb 18, 2025 – Feb 21, 2025

Subsystems (3)
lore
Share & tools
lore

Jingbo Xu

author
#0 lore raw

cover fixes for uncached IO

Jingbo Xu (2):
mm/filemap: fix miscalculated file range for
filemap_fdatawrite_range_kick()
mm/truncate: don't skip dirty page in folio_unmap_invalidate()

 include/linux/fs.h | 4 ++--
 mm/filemap.c       | 2 +-
 mm/truncate.c      | 2 --
 3 files changed, 3 insertions(+), 5 deletions(-)
Signature · 1 line
2.19.1.6.gb485710b
author
#1 lore raw

PATCH 1/2 mm/filemap: fix miscalculated file range for filemap_fdatawrite_range_kick()

iocb->ki_pos has been updated with the number of written bytes since generic_perform_write().

Besides __filemap_fdatawrite_range() accepts the inclusive end of the data range.

Fixes 1d4457576570 ("mm: call filemap_fdatawrite_range_kick() after IOCB_DONTCACHE issue")
---
 include/linux/fs.h | 4 ++--
 mm/filemap.c       | 2 +-
 2 files changed, 3 insertions(+), 3 deletions(-)
include/linux/fs.h +2 -2
index 2c3b2f8a621f..f12774291423 100644
@@ -2975,8 +2975,8 @@ static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count)
 	} else if (iocb->ki_flags & IOCB_DONTCACHE) {
 		struct address_space *mapping = iocb->ki_filp->f_mapping;
 
-		filemap_fdatawrite_range_kick(mapping, iocb->ki_pos,
-					      iocb->ki_pos + count);
+		filemap_fdatawrite_range_kick(mapping, iocb->ki_pos - count,
+					      iocb->ki_pos - 1);
 	}
 
 	return count;
mm/filemap.c +1 -1
index 804d7365680c..d4564a79eb35 100644
@@ -445,7 +445,7 @@ EXPORT_SYMBOL(filemap_fdatawrite_range);
  * filemap_fdatawrite_range_kick - start writeback on a range
  * @mapping:	target address_space
  * @start:	index to start writeback on
- * @end:	last (non-inclusive) index for writeback
+ * @end:	last (inclusive) index for writeback
  *
  * This is a non-integrity writeback helper, to start writing back folios
  * for the indicated range.
Signature · 1 line
2.19.1.6.gb485710b
Reviewed-by on v1 1/2
#2 lore raw

Re: [PATCH 1/2] mm/filemap: fix miscalculated file range for filemap_fdatawrite_range_kick()

Quoted 5 lines · Jingbo Xu

iocb->ki_pos has been updated with the number of written bytes since generic_perform_write().

Besides __filemap_fdatawrite_range() accepts the inclusive end of the data range.

Fix looks good, this got introduced because the logic got moved into generic_write_sync() rather than being a separate helper.

Signature · 1 line
Jens Axboe
author
#3 lore raw

PATCH 2/2 mm/truncate: don't skip dirty page in folio_unmap_invalidate()

In reply to Jingbo Xu

... otherwise this is a behavior change for the previous callers of invalidate_complete_folio2(), e.g. the page invalidation routine.

Fixes 4a9e23159fd3 ("mm/truncate: add folio_unmap_invalidate() helper")
---
 mm/truncate.c | 2 --
 1 file changed, 2 deletions(-)
mm/truncate.c +0 -2
index e2e115adfbc5..76d8fcd89bd0 100644
@@ -548,8 +548,6 @@ int folio_unmap_invalidate(struct address_space *mapping, struct folio *folio,
 
 	VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
 
-	if (folio_test_dirty(folio))
-		return 0;
 	if (folio_mapped(folio))
 		unmap_mapping_folio(folio);
 	BUG_ON(folio_mapped(folio));
Signature · 1 line
2.19.1.6.gb485710b
#4 lore raw

Re: [PATCH 2/2] mm/truncate: don't skip dirty page in folio_unmap_invalidate()

Quoted 2 lines · Jingbo Xu

... otherwise this is a behavior change for the previous callers of invalidate_complete_folio2(), e.g. the page invalidation routine.

Hm. Shouldn't the check be moved to caller of the helper in mm/filemap.c?

Otherwise we would drop pages without writing them back. And lose user's data.

Signature · 1 line
  Kiryl Shutsemau / Kirill A. Shutemov
author
#5 lore raw

Re: [PATCH 2/2] mm/truncate: don't skip dirty page in folio_unmap_invalidate()

Quoted 9 lines · Kirill A. Shutemov
Quoted 2 lines · Jingbo Xu

... otherwise this is a behavior change for the previous callers of invalidate_complete_folio2(), e.g. the page invalidation routine.

Hm. Shouldn't the check be moved to caller of the helper in mm/filemap.c?

Otherwise we would drop pages without writing them back. And lose user's data.

IMHO this check is not needed as the following folio_launder() called inside folio_unmap_invalidate() will write back the dirty page.

Hi Jens,

What do you think about it?

Signature · 2 lines
Thanks,
Jingbo
#6 lore raw

Re: [PATCH 2/2] mm/truncate: don't skip dirty page in folio_unmap_invalidate()

Quoted 13 lines · Jingbo Xu
Quoted 9 lines · Kirill A. Shutemov
Quoted 2 lines · Jingbo Xu

... otherwise this is a behavior change for the previous callers of invalidate_complete_folio2(), e.g. the page invalidation routine.

Hm. Shouldn't the check be moved to caller of the helper in mm/filemap.c?

Otherwise we would drop pages without writing them back. And lose user's data.

Quoted 6 lines

IMHO this check is not needed as the following folio_launder() called inside folio_unmap_invalidate() will write back the dirty page.

Hi Jens,

What do you think about it?

Yep agree on that.

Signature · 1 line
Jens Axboe
#7 lore raw

Re: [PATCH 2/2] mm/truncate: don't skip dirty page in folio_unmap_invalidate()

In reply to Jingbo Xu
Quoted 14 lines · Jingbo Xu

... otherwise this is a behavior change for the previous callers of invalidate_complete_folio2(), e.g. the page invalidation routine.

Fixes 4a9e23159fd3 ("mm/truncate: add folio_unmap_invalidate() helper")
---
 mm/truncate.c | 2 --
 1 file changed, 2 deletions(-)
mm/truncate.c +0 -0
index e2e115adfbc5..76d8fcd89bd0 100644

mm/truncate.c int folio_unmap_invalidate(struct address_space *mapping, struct folio *folio, · line 552

Quoted 5 lines
mm/truncate.c +0 -2
@@ -548,8 +548,6 @@ int folio_unmap_invalidate(struct address_space *mapping, struct folio *folio,
 
 	VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
 
-	if (folio_test_dirty(folio))
-		return 0;

Shouldn't that actually return -EBUSY because the folio could not be invalidated?

Indeed, further down the function the folio gets locked and the dirty test is repeated. If it fails there it returns -EBUSY....

-Dave.

Signature · 2 lines
Dave Chinner
[email protected]
author
#8 lore raw

Re: [PATCH 2/2] mm/truncate: don't skip dirty page in folio_unmap_invalidate()

Quoted 24 lines · Dave Chinner
Quoted 14 lines · Jingbo Xu

... otherwise this is a behavior change for the previous callers of invalidate_complete_folio2(), e.g. the page invalidation routine.

Fixes 4a9e23159fd3 ("mm/truncate: add folio_unmap_invalidate() helper")
---
 mm/truncate.c | 2 --
 1 file changed, 2 deletions(-)
mm/truncate.c +0 -0
index e2e115adfbc5..76d8fcd89bd0 100644

mm/truncate.c int folio_unmap_invalidate(struct address_space *mapping, struct folio *folio, · line 552

Quoted 5 lines
mm/truncate.c +0 -2
@@ -548,8 +548,6 @@ int folio_unmap_invalidate(struct address_space *mapping, struct folio *folio,
 
 	VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
 
-	if (folio_test_dirty(folio))
-		return 0;

Shouldn't that actually return -EBUSY because the folio could not be invalidated?

Quoted 4 lines

Indeed, further down the function the folio gets locked and the dirty test is repeated. If it fails there it returns -EBUSY....

-Dave.

Yeah, the original logic of invalidate_inode_pages2_range() is like

```
invalidate_inode_pages2_range
  # lock page
  # launder the page if it's dirty

  invalidate_complete_folio2
    # recheck if it's dirty, and skip the dirty page (no idea how page
could be redirtied after launder_page())
```

while after commit 4a9e23159fd3 ("mm/truncate: add folio_unmap_invalidate() helper"), this logic is changed to:

```
invalidate_inode_pages2_range
  # lock page
  folio_unmap_invalidate
    # check if it's dirty, and skip dirty page
    # launder the page if it's dirty (doubt if it's noops)

    # recheck if it's dirty, and skip the dirty page
```
Signature · 2 lines
Thanks,
Jingbo
author
#9 lore raw
In reply to Jingbo Xu

Hi Christian,

Quoted 10 lines · Jingbo Xu

Jingbo Xu (2):
mm/filemap: fix miscalculated file range for
filemap_fdatawrite_range_kick()
mm/truncate: don't skip dirty page in folio_unmap_invalidate()

 include/linux/fs.h | 4 ++--
 mm/filemap.c       | 2 +-
 mm/truncate.c      | 2 --
 3 files changed, 3 insertions(+), 5 deletions(-)

Would you consider these fixes?

Signature · 2 lines
Thanks,
Jingbo
maintainer Applied
#10 lore raw
In reply to Jingbo Xu
Quoted 11 lines · Jingbo Xu

Jingbo Xu (2):
mm/filemap: fix miscalculated file range for
filemap_fdatawrite_range_kick()
mm/truncate: don't skip dirty page in folio_unmap_invalidate()

include/linux/fs.h | 4 ++--

 mm/filemap.c       | 2 +-
 mm/truncate.c      | 2 --
 3 files changed, 3 insertions(+), 5 deletions(-)

[...]

Applied to the vfs.fixes branch of the vfs/vfs.git tree. Patches in the vfs.fixes branch should appear in linux-next soon.

Please report any outstanding bugs that were missed during review in a new review to the original patch series allowing us to drop it.

It's encouraged to provide Acked-bys and Reviewed-bys even though the patch has now been applied. If possible patch trailers will be updated.

Note that commit hashes shown below are subject to change due to rebase, trailer updates or similar. If in doubt, please check the listed branch.

tree:   https://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs.git
branch: vfs.fixes

[1/2] mm/filemap: fix miscalculated file range for filemap_fdatawrite_range_kick()

      https://git.kernel.org/vfs/vfs/c/8510edf191d2
[2/2] mm/truncate: don't skip dirty page in folio_unmap_invalidate()
      https://git.kernel.org/vfs/vfs/c/927289988068