2011-07-18 15:59:34

by Josef Bacik

Subject: [PATCH] ext4: hold page lock when checking mapping in ext4_page_mkwrite

Pages can be evicted from cache outside of truncate, for example when
invalidating pages after writing with O_DIRECT. The only proper way to check
page->mapping is under the page lock, so fix ext4_page_mkwrite to do this.
Thanks,

Signed-off-by: Josef Bacik <[email protected]>
---
fs/ext4/inode.c | 3 ++-
1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index e3126c0..1ad7d10 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -5855,14 +5855,15 @@ int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
*/
down_read(&inode->i_alloc_sem);
size = i_size_read(inode);
+ lock_page(page);
if (page->mapping != mapping || size <= page_offset(page)
|| !PageUptodate(page)) {
+ unlock_page(page);
/* page got truncated from under us? */
goto out_unlock;
}
ret = 0;

- lock_page(page);
wait_on_page_writeback(page);
if (PageMappedToDisk(page)) {
up_read(&inode->i_alloc_sem);
--
1.7.5.2

2011-07-19 02:13:30

by Christoph Hellwig

Subject: Re: [PATCH] ext4: hold page lock when checking mapping in ext4_page_mkwrite

On Mon, Jul 18, 2011 at 11:59:32AM -0400, Josef Bacik wrote:
> Pages can be evicted from cache outside of truncate, for example when
> invalidating pages after writing with O_DIRECT. The only proper way to check
> page->mapping is under the page lock, so fix ext4_page_mkwrite to do this.
> Thanks,

Josef, please check the ext4_page_mkwrite implementation in the vfs
tree, which is just a wrapper around the generic code and thus shouldn't
have this problem any more.