f2fs: fix to update largest extent under lock

In order to avoid racing problem, make largest extent cache being updated
under lock.

Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This commit is contained in:
Chao Yu 2016-10-11 22:57:05 +08:00 committed by Jaegeuk Kim
parent 58736fa60f
commit b691d98fdd
1 changed files with 7 additions and 4 deletions

View File

@ -252,6 +252,7 @@ retry:
int update_inode(struct inode *inode, struct page *node_page)
{
struct f2fs_inode *ri;
struct extent_tree *et = F2FS_I(inode)->extent_tree;
f2fs_inode_synced(inode);
@ -267,11 +268,13 @@ int update_inode(struct inode *inode, struct page *node_page)
ri->i_size = cpu_to_le64(i_size_read(inode));
ri->i_blocks = cpu_to_le64(inode->i_blocks);
if (F2FS_I(inode)->extent_tree)
set_raw_extent(&F2FS_I(inode)->extent_tree->largest,
&ri->i_ext);
else
if (et) {
read_lock(&et->lock);
set_raw_extent(&et->largest, &ri->i_ext);
read_unlock(&et->lock);
} else {
memset(&ri->i_ext, 0, sizeof(ri->i_ext));
}
set_raw_inline(inode, ri);
ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);