Commits 41c4d25f78c0 and 41c4d25f78c0 introduced a regression into
3.6-rc1 for ext4 in nodealloc mode, such that mtime updates would not
take place for files modified via mmap if the page was already in the
page cache.
The problem was that ext4_page_mkwrite() had a shortcut which would
avoid calling __block_page_mkwrite() under some circumstances, and the
above two commit transferred the responsibility of calling
file_update_time() to __block_page_mkwrite --- which woudln't get
called in some circumstances.
Since __block_page_mkwrite() only has three years,
block_page_mkwrite(), ext4_page_mkwrite, and nilfs_page_mkwrite(), the
best way to solve this is to move the responsibility for calling
file_update_time() to its caller.
Signed-off-by: "Theodore Ts'o" <[email protected]>
Cc: Jan Kara <[email protected]>
Cc: KONISHI Ryusuke <[email protected]>
Cc: [email protected]
---
NOTE: Since this is a 3.6 regression, I may push this to Linus ahead of
the merge window, since it will also affect distibutions which use ext4
to mount ext3-formatted partitions.
fs/buffer.c | 13 +++++++------
fs/ext4/inode.c | 1 +
fs/nilfs2/file.c | 1 +
3 files changed, 9 insertions(+), 6 deletions(-)
diff --git a/fs/buffer.c b/fs/buffer.c
index 9f6d2e4..1fe3968 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -2318,12 +2318,6 @@ int __block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
loff_t size;
int ret;
- /*
- * Update file times before taking page lock. We may end up failing the
- * fault so this update may be superfluous but who really cares...
- */
- file_update_time(vma->vm_file);
-
lock_page(page);
size = i_size_read(inode);
if ((page->mapping != inode->i_mapping) ||
@@ -2361,6 +2355,13 @@ int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf,
struct super_block *sb = vma->vm_file->f_path.dentry->d_inode->i_sb;
sb_start_pagefault(sb);
+
+ /*
+ * Update file times before taking page lock. We may end up failing the
+ * fault so this update may be superfluous but who really cares...
+ */
+ file_update_time(vma->vm_file);
+
ret = __block_page_mkwrite(vma, vmf, get_block);
sb_end_pagefault(sb);
return block_page_mkwrite_return(ret);
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 09308ad..f18e786 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -4788,6 +4788,7 @@ int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
int retries = 0;
sb_start_pagefault(inode->i_sb);
+ file_update_time(vma->vm_file);
/* Delalloc case is easy... */
if (test_opt(inode->i_sb, DELALLOC) &&
!ext4_should_journal_data(inode) &&
diff --git a/fs/nilfs2/file.c b/fs/nilfs2/file.c
index a4d56ac..5b387a4 100644
--- a/fs/nilfs2/file.c
+++ b/fs/nilfs2/file.c
@@ -116,6 +116,7 @@ static int nilfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
if (unlikely(ret))
goto out;
+ file_update_time(vma->vm_file);
ret = __block_page_mkwrite(vma, vmf, nilfs_get_block);
if (ret) {
nilfs_transaction_abort(inode->i_sb);
--
1.7.12.rc0.22.gcdd159b
Theodore Ts'o wrote:
> Since __block_page_mkwrite() only has three years,
> block_page_mkwrite(), ext4_page_mkwrite, and nilfs_page_mkwrite(), the
years = callers?
Note: this fixes a failure in xfstests #215 for ext3 file systems
mounted using ext4 and ext4 file systems mounted with -o nodelalloc.
- Ted
On Sun, Sep 30, 2012 at 03:48:19PM -0400, Theodore Ts'o wrote:
> Commits 41c4d25f78c0 and 41c4d25f78c0 introduced a regression into
> 3.6-rc1 for ext4 in nodealloc mode, such that mtime updates would not
> take place for files modified via mmap if the page was already in the
> page cache.
>
> The problem was that ext4_page_mkwrite() had a shortcut which would
> avoid calling __block_page_mkwrite() under some circumstances, and the
> above two commit transferred the responsibility of calling
> file_update_time() to __block_page_mkwrite --- which woudln't get
> called in some circumstances.
>
> Since __block_page_mkwrite() only has three years,
> block_page_mkwrite(), ext4_page_mkwrite, and nilfs_page_mkwrite(), the
> best way to solve this is to move the responsibility for calling
> file_update_time() to its caller.
>
> Signed-off-by: "Theodore Ts'o" <[email protected]>
> Cc: Jan Kara <[email protected]>
> Cc: KONISHI Ryusuke <[email protected]>
> Cc: [email protected]
On Sun, Sep 30, 2012 at 03:48:19PM -0400, Theodore Ts'o wrote:
> Commits 41c4d25f78c0 and 41c4d25f78c0 introduced a regression
you mistakenly wrote the same commit ID twice.
Regards,
Willy
On Sun, Sep 30, 2012 at 12:50:02PM -0700, Jonathan Nieder wrote:
> Theodore Ts'o wrote:
>
> > Since __block_page_mkwrite() only has three years,
> > block_page_mkwrite(), ext4_page_mkwrite, and nilfs_page_mkwrite(), the
>
> years = callers?
Yes, thanks for pointing that out. I'll fix the commit description.
On Sun, Sep 30, 2012 at 09:51:24PM +0200, Willy Tarreau wrote:
> On Sun, Sep 30, 2012 at 03:48:19PM -0400, Theodore Ts'o wrote:
> > Commits 41c4d25f78c0 and 41c4d25f78c0 introduced a regression
>
> you mistakenly wrote the same commit ID twice.
Sorry, the two commit ID's are:
5e8830dc85d0 and 41c4d25f78c0
- Ted
The rewritten commit description:
ext4: fix mtime update in nodelalloc mode
Commits 5e8830dc85d0 and 41c4d25f78c0 introduced a regression into
3.6-rc1 for ext4 in nodealloc mode, such that mtime updates would not
take place for files modified via mmap if the page was already in the
page cache. This would also affect ext3 file systems mounted using
the ext4 file system driver.
The problem was that ext4_page_mkwrite() had a shortcut which would
avoid calling __block_page_mkwrite() under some circumstances, and the
above two commit transferred the responsibility of calling
file_update_time() to __block_page_mkwrite --- which woudln't get
called in some circumstances.
Since __block_page_mkwrite() only has three callers,
block_page_mkwrite(), ext4_page_mkwrite, and nilfs_page_mkwrite(), the
best way to solve this is to move the responsibility for calling
file_update_time() to its caller.
This problem was found via xfstests #215 with a file system mounted
with -o nodelalloc.
Signed-off-by: "Theodore Ts'o" <[email protected]>
Cc: Jan Kara <[email protected]>
Cc: KONISHI Ryusuke <[email protected]>
Cc: [email protected]
Note: If this gets pushed to Linus before the merge window opens, I'll
drop the cc of [email protected], since the regression was
introduced in 3.6-rc1.
- Ted
On Sun 30-09-12 16:30:26, Ted Tso wrote:
> The rewritten commit description:
>
> ext4: fix mtime update in nodelalloc mode
>
> Commits 5e8830dc85d0 and 41c4d25f78c0 introduced a regression into
> 3.6-rc1 for ext4 in nodealloc mode, such that mtime updates would not
> take place for files modified via mmap if the page was already in the
> page cache. This would also affect ext3 file systems mounted using
> the ext4 file system driver.
>
> The problem was that ext4_page_mkwrite() had a shortcut which would
> avoid calling __block_page_mkwrite() under some circumstances, and the
> above two commit transferred the responsibility of calling
> file_update_time() to __block_page_mkwrite --- which woudln't get
> called in some circumstances.
>
> Since __block_page_mkwrite() only has three callers,
> block_page_mkwrite(), ext4_page_mkwrite, and nilfs_page_mkwrite(), the
> best way to solve this is to move the responsibility for calling
> file_update_time() to its caller.
>
> This problem was found via xfstests #215 with a file system mounted
> with -o nodelalloc.
>
> Signed-off-by: "Theodore Ts'o" <[email protected]>
> Cc: Jan Kara <[email protected]>
> Cc: KONISHI Ryusuke <[email protected]>
> Cc: [email protected]
>
> Note: If this gets pushed to Linus before the merge window opens, I'll
> drop the cc of [email protected], since the regression was
> introduced in 3.6-rc1.
Ah, good catch. Thanks for fixing this. I agree with the fix so you can
add:
Reviewed-by: Jan Kara <[email protected]>
Honza