f2fs: fix potential deadlock due to unpaired node_write lock use
Commit Message
If S_NOQUOTA is cleared from inode during data page writeback of quota
file, it may miss to unlock node_write lock, result in potential
deadlock, fix to use the lock in paired.
Kworker Thread
- writepage
if (IS_NOQUOTA())
f2fs_down_read(&sbi->node_write);
- vfs_cleanup_quota_inode
- inode->i_flags &= ~S_NOQUOTA;
if (IS_NOQUOTA())
f2fs_up_read(&sbi->node_write);
Fixes: 79963d967b49 ("f2fs: shrink node_write lock coverage")
Signed-off-by: Chao Yu <chao@kernel.org>
---
fs/f2fs/compress.c | 7 ++++---
fs/f2fs/data.c | 7 ++++---
2 files changed, 8 insertions(+), 6 deletions(-)
Comments
Hello:
This patch was applied to jaegeuk/f2fs.git (dev)
by Jaegeuk Kim <jaegeuk@kernel.org>:
On Sun, 14 May 2023 16:07:23 +0800 you wrote:
> If S_NOQUOTA is cleared from inode during data page writeback of quota
> file, it may miss to unlock node_write lock, result in potential
> deadlock, fix to use the lock in paired.
>
> Kworker Thread
> - writepage
> if (IS_NOQUOTA())
> f2fs_down_read(&sbi->node_write);
> - vfs_cleanup_quota_inode
> - inode->i_flags &= ~S_NOQUOTA;
> if (IS_NOQUOTA())
> f2fs_up_read(&sbi->node_write);
>
> [...]
Here is the summary with links:
- [f2fs-dev] f2fs: fix potential deadlock due to unpaired node_write lock use
https://git.kernel.org/jaegeuk/f2fs/c/a731b6aefbb4
You are awesome, thank you!
@@ -1215,6 +1215,7 @@ static int f2fs_write_compressed_pages(struct compress_ctx *cc,
unsigned int last_index = cc->cluster_size - 1;
loff_t psize;
int i, err;
+ bool quota_inode = IS_NOQUOTA(inode);
/* we should bypass data pages to proceed the kworker jobs */
if (unlikely(f2fs_cp_error(sbi))) {
@@ -1222,7 +1223,7 @@ static int f2fs_write_compressed_pages(struct compress_ctx *cc,
goto out_free;
}
- if (IS_NOQUOTA(inode)) {
+ if (quota_inode) {
/*
* We need to wait for node_write to avoid block allocation during
* checkpoint. This can only happen to quota writes which can cause
@@ -1344,7 +1345,7 @@ static int f2fs_write_compressed_pages(struct compress_ctx *cc,
set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
f2fs_put_dnode(&dn);
- if (IS_NOQUOTA(inode))
+ if (quota_inode)
f2fs_up_read(&sbi->node_write);
else
f2fs_unlock_op(sbi);
@@ -1370,7 +1371,7 @@ static int f2fs_write_compressed_pages(struct compress_ctx *cc,
out_put_dnode:
f2fs_put_dnode(&dn);
out_unlock_op:
- if (IS_NOQUOTA(inode))
+ if (quota_inode)
f2fs_up_read(&sbi->node_write);
else
f2fs_unlock_op(sbi);
@@ -2775,6 +2775,7 @@ int f2fs_write_single_data_page(struct page *page, int *submitted,
loff_t psize = (loff_t)(page->index + 1) << PAGE_SHIFT;
unsigned offset = 0;
bool need_balance_fs = false;
+ bool quota_inode = IS_NOQUOTA(inode);
int err = 0;
struct f2fs_io_info fio = {
.sbi = sbi,
@@ -2832,19 +2833,19 @@ int f2fs_write_single_data_page(struct page *page, int *submitted,
goto out;
/* Dentry/quota blocks are controlled by checkpoint */
- if (S_ISDIR(inode->i_mode) || IS_NOQUOTA(inode)) {
+ if (S_ISDIR(inode->i_mode) || quota_inode) {
/*
* We need to wait for node_write to avoid block allocation during
* checkpoint. This can only happen to quota writes which can cause
* the below discard race condition.
*/
- if (IS_NOQUOTA(inode))
+ if (quota_inode)
f2fs_down_read(&sbi->node_write);
fio.need_lock = LOCK_DONE;
err = f2fs_do_write_data_page(&fio);
- if (IS_NOQUOTA(inode))
+ if (quota_inode)
f2fs_up_read(&sbi->node_write);
goto done;