On 2021/8/21 14:54, Zhang Yi wrote:
> In ext4_get_inode_loc(), we may skip IO and get an zero && uptodate
> inode buffer when the inode monopolize an inode block for performance
> reason. For most cases, ext4_mark_iloc_dirty() will fill the inode
> buffer to make it fine, but we could miss this call if something bad
> happened. Finally, __ext4_get_inode_loc_noinmem() may probably get an
> empty inode buffer and trigger ext4 error.
>
> For example, if we remove a nonexistent xattr on inode A,
> ext4_xattr_set_handle() will return ENODATA before invoking
> ext4_mark_iloc_dirty(), it will left an uptodate but zero buffer. We
> will get checksum error message in ext4_iget() when getting inode again.
>
> EXT4-fs error (device sda): ext4_lookup:1784: inode #131074: comm cat: iget: checksum invalid
>
> Even worse, if we allocate another inode B at the same inode block, it
> will corrupt the inode A on disk when write back inode B.
>
> So this patch postpone the initialization and mark buffer uptodate logic
> until shortly before we fill correct inode data in ext4_do_update_inode()
> if skip read I/O, ensure the buffer is really uptodate.
>
> Signed-off-by: Zhang Yi <yi.zhang@xxxxxxxxxx>
> Reviewed-by: Jan Kara <jack@xxxxxxx>
> ---
> fs/ext4/inode.c | 27 ++++++++++++++++++++++++---
> 1 file changed, 24 insertions(+), 3 deletions(-)
>
> diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
> index 8323d3e8f393..000abb5696b0 100644
> --- a/fs/ext4/inode.c
> +++ b/fs/ext4/inode.c
> @@ -4367,9 +4367,12 @@ static int __ext4_get_inode_loc(struct super_block *sb, unsigned long ino,
> }
> brelse(bitmap_bh);
> if (i == start + inodes_per_block) {
> - /* all other inodes are free, so skip I/O */
> - memset(bh->b_data, 0, bh->b_size);
> - set_buffer_uptodate(bh);
> + /*
> + * All other inodes are free, skip I/O. Return
> + * uninitialized buffer immediately, initialization
> + * is postponed until shortly before we fill inode
> + * contents.
> + */
> unlock_buffer(bh);
> goto has_buffer;
> }
> @@ -5028,6 +5031,24 @@ static int ext4_do_update_inode(handle_t *handle,
> gid_t i_gid;
> projid_t i_projid;
>
> + /*
> + * If the buffer is not uptodate, it means all information of the
> + * inode is in memory and we got this buffer without reading the
> + * block. We must be cautious that once we mark the buffer as
> + * uptodate, we rely on filling in the correct inode data later
> + * in this function. Otherwise if we left uptodate buffer without
> + * copying proper inode contents, we could corrupt the inode on
> + * disk after allocating another inode in the same block.
> + */
> + if (!buffer_uptodate(bh)) {
> + lock_buffer(bh);
> + if (!buffer_uptodate(bh)) {
> + memset(bh->b_data, 0, bh->b_size);
> + set_buffer_uptodate(bh);
> + }
> + unlock_buffer(bh);
> + }
Hi, Jan.
I notice that above solution is not correct. The problem is still in
ext4_xattr_set_handle(), if we set a new xattr entry in a pure inode,
the above hunk may zero out the ibody xattr entry we just set up in
ext4_xattr_ibody_set().
I guess we could not 'zero out buffer && mark buffer uptodate' here,
maybe __ext4_get_inode_loc() should return a really initialized buffer,
or else it's still fragile and hard to guarantee that the 'zero out'
and 'postponed set_buffer_uptodate()' will not zero out something we
just set or overwrite something we updated concurrently.
How about factor out the filling inode contents from ext4_do_update_inode()
into maybe ext4_fill_raw_inode(), and call it in __ext4_get_inode_loc() ?
Please see my v4 patchset.
Thanks,
Yi.
