2020-10-20 08:56:28

by luomeng

[permalink] [raw]
Subject: [PATCH] ext4: fix invalid inode checksum

During the stability test, there are some errors:
ext4_lookup:1590: inode #6967: comm fsstress: iget: checksum invalid.

If the inode->i_iblocks too big and doesn't set huge file flag, checksum
will not be recalculated when update the inode information to it's buffer.
If other inode marks the buffer dirty, then the inconsistent inode will
be flushed to disk.

Fix this problem by checking i_blocks in advance.

Signed-off-by: Luo Meng <[email protected]>
---
fs/ext4/inode.c | 11 ++++++-----
1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index bf596467c234..fe53774b8b6c 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -4971,6 +4971,12 @@ static int ext4_do_update_inode(handle_t *handle,
if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);

+ err = ext4_inode_blocks_set(handle, raw_inode, ei);
+ if (err) {
+ spin_unlock(&ei->i_raw_lock);
+ goto out_brelse;
+ }
+
raw_inode->i_mode = cpu_to_le16(inode->i_mode);
i_uid = i_uid_read(inode);
i_gid = i_gid_read(inode);
@@ -5004,11 +5010,6 @@ static int ext4_do_update_inode(handle_t *handle,
EXT4_INODE_SET_XTIME(i_atime, inode, raw_inode);
EXT4_EINODE_SET_XTIME(i_crtime, ei, raw_inode);

- err = ext4_inode_blocks_set(handle, raw_inode, ei);
- if (err) {
- spin_unlock(&ei->i_raw_lock);
- goto out_brelse;
- }
raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
raw_inode->i_flags = cpu_to_le32(ei->i_flags & 0xFFFFFFFF);
if (likely(!test_opt2(inode->i_sb, HURD_COMPAT)))
--
2.25.4


2020-10-20 09:19:24

by Darrick J. Wong

[permalink] [raw]
Subject: Re: [PATCH] ext4: fix invalid inode checksum

On Tue, Oct 20, 2020 at 09:36:31AM +0800, Luo Meng wrote:
> During the stability test, there are some errors:
> ext4_lookup:1590: inode #6967: comm fsstress: iget: checksum invalid.
>
> If the inode->i_iblocks too big and doesn't set huge file flag, checksum
> will not be recalculated when update the inode information to it's buffer.
> If other inode marks the buffer dirty, then the inconsistent inode will
> be flushed to disk.
>
> Fix this problem by checking i_blocks in advance.
>
> Signed-off-by: Luo Meng <[email protected]>

Fun! Yikes!
Reviewed-by: Darrick J. Wong <[email protected]>

--D

> ---
> fs/ext4/inode.c | 11 ++++++-----
> 1 file changed, 6 insertions(+), 5 deletions(-)
>
> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> index bf596467c234..fe53774b8b6c 100644
> --- a/fs/ext4/inode.c
> +++ b/fs/ext4/inode.c
> @@ -4971,6 +4971,12 @@ static int ext4_do_update_inode(handle_t *handle,
> if (ext4_test_inode_state(inode, EXT4_STATE_NEW))
> memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
>
> + err = ext4_inode_blocks_set(handle, raw_inode, ei);
> + if (err) {
> + spin_unlock(&ei->i_raw_lock);
> + goto out_brelse;
> + }
> +
> raw_inode->i_mode = cpu_to_le16(inode->i_mode);
> i_uid = i_uid_read(inode);
> i_gid = i_gid_read(inode);
> @@ -5004,11 +5010,6 @@ static int ext4_do_update_inode(handle_t *handle,
> EXT4_INODE_SET_XTIME(i_atime, inode, raw_inode);
> EXT4_EINODE_SET_XTIME(i_crtime, ei, raw_inode);
>
> - err = ext4_inode_blocks_set(handle, raw_inode, ei);
> - if (err) {
> - spin_unlock(&ei->i_raw_lock);
> - goto out_brelse;
> - }
> raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
> raw_inode->i_flags = cpu_to_le32(ei->i_flags & 0xFFFFFFFF);
> if (likely(!test_opt2(inode->i_sb, HURD_COMPAT)))
> --
> 2.25.4
>

2020-10-28 21:41:54

by Theodore Ts'o

[permalink] [raw]
Subject: Re: [PATCH] ext4: fix invalid inode checksum

On Tue, Oct 20, 2020 at 09:36:31AM +0800, Luo Meng wrote:
> During the stability test, there are some errors:
> ext4_lookup:1590: inode #6967: comm fsstress: iget: checksum invalid.
>
> If the inode->i_iblocks too big and doesn't set huge file flag, checksum
> will not be recalculated when update the inode information to it's buffer.
> If other inode marks the buffer dirty, then the inconsistent inode will
> be flushed to disk.
>
> Fix this problem by checking i_blocks in advance.
>
> Signed-off-by: Luo Meng <[email protected]>

Thanks, this has been applied.

- Ted