[PATCH V2 1/2] f2fs: compress: relocate some judgments in f2fs_reserve_compress_blocks

[PATCH V2 1/2] f2fs: compress: relocate some judgments in f2fs_reserve_compress_blocks

[Xiuhong Wang]

The following f2fs_io test will get a "0" result instead of -EINVAL,
unisoc # ./f2fs_io compress file
unisoc # ./f2fs_io reserve_cblocks file
 0
it's not reasonable, so the judgement of
atomic_read(&F2FS_I(inode)->i_compr_blocks) should be placed after
the judgement of is_inode_flag_set(inode, FI_COMPRESS_RELEASED).

Fixes: c75488fb4d82 ("f2fs: introduce F2FS_IOC_RESERVE_COMPRESS_BLOCKS")
Signed-off-by: Xiuhong Wang <xiuhong.wang@unisoc.com>
Signed-off-by: Zhiguo Niu <zhiguo.niu@unisoc.com>
---
 fs/f2fs/file.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 4ca6c693b33a..74c5e48fce22 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -3720,9 +3720,6 @@ static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
 	if (ret)
 		return ret;
 
-	if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
-		goto out;
-
 	f2fs_balance_fs(sbi, true);
 
 	inode_lock(inode);
@@ -3732,6 +3729,9 @@ static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
 		goto unlock_inode;
 	}
 
+	if (atomic_read(&F2FS_I(inode)->i_compr_blocks))
+		goto unlock_inode;
+
 	f2fs_down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
 	filemap_invalidate_lock(inode->i_mapping);
 
@@ -3778,7 +3778,6 @@ static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
 	}
 unlock_inode:
 	inode_unlock(inode);
-out:
 	mnt_drop_write_file(filp);
 
 	if (ret >= 0) {
-- 
2.25.1

[PATCH V2 2/2] f2fs: compress: fix reserve_cblocks counting error when out of space

[Xiuhong Wang]

When a file only needs one direct_node, performing the following
operations will cause the file to be unrepairable:

unisoc # ./f2fs_io compress test.apk
unisoc #df -h | grep dm-48
/dev/block/dm-48 112G 112G 1.2M 100% /data

unisoc # ./f2fs_io release_cblocks test.apk
924
unisoc # df -h | grep dm-48
/dev/block/dm-48 112G 112G 4.8M 100% /data

unisoc # dd if=/dev/random of=file4 bs=1M count=3
3145728 bytes (3.0 M) copied, 0.025 s, 120 M/s
unisoc # df -h | grep dm-48
/dev/block/dm-48 112G 112G 1.8M 100% /data

unisoc # ./f2fs_io reserve_cblocks test.apk
F2FS_IOC_RESERVE_COMPRESS_BLOCKS failed: No space left on device

adb reboot
unisoc # df -h  | grep dm-48
/dev/block/dm-48             112G 112G   11M 100% /data
unisoc # ./f2fs_io reserve_cblocks test.apk
0

This is because the file has only one direct_node. After returning
to -ENOSPC, reserved_blocks += ret will not be executed. As a result,
the reserved_blocks at this time is still 0, which is not the real
number of reserved blocks. Therefore, fsck cannot be set to repair
the file.

After this patch, the fsck flag will be set to fix this problem.

unisoc # df -h | grep dm-48
/dev/block/dm-48             112G 112G  1.8M 100% /data
unisoc # ./f2fs_io reserve_cblocks test.apk
F2FS_IOC_RESERVE_COMPRESS_BLOCKS failed: No space left on device

adb reboot then fsck will be executed
unisoc # df -h  | grep dm-48
/dev/block/dm-48             112G 112G   11M 100% /data
unisoc # ./f2fs_io reserve_cblocks test.apk
924

Fixes: c75488fb4d82 ("f2fs: introduce F2FS_IOC_RESERVE_COMPRESS_BLOCKS")
Signed-off-by: Xiuhong Wang <xiuhong.wang@unisoc.com>
Signed-off-by: Zhiguo Niu <zhiguo.niu@unisoc.com>
---
 fs/f2fs/file.c | 15 +++++++--------
 1 file changed, 7 insertions(+), 8 deletions(-)

diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 74c5e48fce22..dc9c6bac678d 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -3624,10 +3624,10 @@ static int f2fs_release_compress_blocks(struct file *filp, unsigned long arg)
 	return ret;
 }
 
-static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
+static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count,
+		unsigned int *reserved_blocks)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
-	unsigned int reserved_blocks = 0;
 	int cluster_size = F2FS_I(dn->inode)->i_cluster_size;
 	block_t blkaddr;
 	int i;
@@ -3691,12 +3691,12 @@ static int reserve_compress_blocks(struct dnode_of_data *dn, pgoff_t count)
 
 		f2fs_i_compr_blocks_update(dn->inode, compr_blocks, true);
 
-		reserved_blocks += reserved;
+		*reserved_blocks += reserved;
 next:
 		count -= cluster_size;
 	}
 
-	return reserved_blocks;
+	return 0;
 }
 
 static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
@@ -3757,7 +3757,7 @@ static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
 		count = min(end_offset - dn.ofs_in_node, last_idx - page_idx);
 		count = round_up(count, F2FS_I(inode)->i_cluster_size);
 
-		ret = reserve_compress_blocks(&dn, count);
+		ret = reserve_compress_blocks(&dn, count, &reserved_blocks);
 
 		f2fs_put_dnode(&dn);
 
@@ -3765,13 +3765,12 @@ static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
 			break;
 
 		page_idx += count;
-		reserved_blocks += ret;
 	}
 
 	filemap_invalidate_unlock(inode->i_mapping);
 	f2fs_up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
 
-	if (ret >= 0) {
+	if (!ret) {
 		clear_inode_flag(inode, FI_COMPRESS_RELEASED);
 		inode_set_ctime_current(inode);
 		f2fs_mark_inode_dirty_sync(inode, true);
@@ -3780,7 +3779,7 @@ static int f2fs_reserve_compress_blocks(struct file *filp, unsigned long arg)
 	inode_unlock(inode);
 	mnt_drop_write_file(filp);
 
-	if (ret >= 0) {
+	if (!ret) {
 		ret = put_user(reserved_blocks, (u64 __user *)arg);
 	} else if (reserved_blocks &&
 			atomic_read(&F2FS_I(inode)->i_compr_blocks)) {
-- 
2.25.1

Re: [PATCH V2 2/2] f2fs: compress: fix reserve_cblocks counting error when out of space

[Chao Yu]

On 2024/3/6 11:47, Xiuhong Wang wrote:
> When a file only needs one direct_node, performing the following
> operations will cause the file to be unrepairable:
> 
> unisoc # ./f2fs_io compress test.apk
> unisoc #df -h | grep dm-48
> /dev/block/dm-48 112G 112G 1.2M 100% /data
> 
> unisoc # ./f2fs_io release_cblocks test.apk
> 924
> unisoc # df -h | grep dm-48
> /dev/block/dm-48 112G 112G 4.8M 100% /data
> 
> unisoc # dd if=/dev/random of=file4 bs=1M count=3
> 3145728 bytes (3.0 M) copied, 0.025 s, 120 M/s
> unisoc # df -h | grep dm-48
> /dev/block/dm-48 112G 112G 1.8M 100% /data
> 
> unisoc # ./f2fs_io reserve_cblocks test.apk
> F2FS_IOC_RESERVE_COMPRESS_BLOCKS failed: No space left on device
> 
> adb reboot
> unisoc # df -h  | grep dm-48
> /dev/block/dm-48             112G 112G   11M 100% /data
> unisoc # ./f2fs_io reserve_cblocks test.apk
> 0
> 
> This is because the file has only one direct_node. After returning
> to -ENOSPC, reserved_blocks += ret will not be executed. As a result,
> the reserved_blocks at this time is still 0, which is not the real
> number of reserved blocks. Therefore, fsck cannot be set to repair
> the file.
> 
> After this patch, the fsck flag will be set to fix this problem.
> 
> unisoc # df -h | grep dm-48
> /dev/block/dm-48             112G 112G  1.8M 100% /data
> unisoc # ./f2fs_io reserve_cblocks test.apk
> F2FS_IOC_RESERVE_COMPRESS_BLOCKS failed: No space left on device
> 
> adb reboot then fsck will be executed
> unisoc # df -h  | grep dm-48
> /dev/block/dm-48             112G 112G   11M 100% /data
> unisoc # ./f2fs_io reserve_cblocks test.apk
> 924
> 
> Fixes: c75488fb4d82 ("f2fs: introduce F2FS_IOC_RESERVE_COMPRESS_BLOCKS")
> Signed-off-by: Xiuhong Wang <xiuhong.wang@unisoc.com>
> Signed-off-by: Zhiguo Niu <zhiguo.niu@unisoc.com>

Reviewed-by: Chao Yu <chao@kernel.org>

Thanks,

Re: [PATCH V2 1/2] f2fs: compress: relocate some judgments in f2fs_reserve_compress_blocks

[Chao Yu]

On 2024/3/6 11:47, Xiuhong Wang wrote:
> The following f2fs_io test will get a "0" result instead of -EINVAL,
> unisoc # ./f2fs_io compress file
> unisoc # ./f2fs_io reserve_cblocks file
>   0
> it's not reasonable, so the judgement of
> atomic_read(&F2FS_I(inode)->i_compr_blocks) should be placed after
> the judgement of is_inode_flag_set(inode, FI_COMPRESS_RELEASED).
> 
> Fixes: c75488fb4d82 ("f2fs: introduce F2FS_IOC_RESERVE_COMPRESS_BLOCKS")
> Signed-off-by: Xiuhong Wang <xiuhong.wang@unisoc.com>
> Signed-off-by: Zhiguo Niu <zhiguo.niu@unisoc.com>

Reviewed-by: Chao Yu <chao@kernel.org>

Thanks,

Re: [f2fs-dev] [PATCH V2 1/2] f2fs: compress: relocate some judgments in f2fs_reserve_compress_blocks

[patchwork-bot+f2fs]

Hello:

This series was applied to jaegeuk/f2fs.git (dev)
by Jaegeuk Kim <jaegeuk@kernel.org>:

On Wed, 6 Mar 2024 11:47:45 +0800 you wrote:
> The following f2fs_io test will get a "0" result instead of -EINVAL,
> unisoc # ./f2fs_io compress file
> unisoc # ./f2fs_io reserve_cblocks file
>  0
> it's not reasonable, so the judgement of
> atomic_read(&F2FS_I(inode)->i_compr_blocks) should be placed after
> the judgement of is_inode_flag_set(inode, FI_COMPRESS_RELEASED).
> 
> [...]

Here is the summary with links:
  - [f2fs-dev,V2,1/2] f2fs: compress: relocate some judgments in f2fs_reserve_compress_blocks
    https://git.kernel.org/jaegeuk/f2fs/c/b7d797d241c1
  - [f2fs-dev,V2,2/2] f2fs: compress: fix reserve_cblocks counting error when out of space
    https://git.kernel.org/jaegeuk/f2fs/c/2f6d721e14b6

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html