On Sat, Jul 07, 2018 at 12:10:18AM -0400, Theodore Y. Ts'o wrote:
> On Fri, Jul 06, 2018 at 11:43:24AM -0600, dann frazier wrote:
> > Hi,
> > We're seeing a regression triggered by the stress-ng[*] "chdir" test
> > that I've bisected to:
> >
> > 044e6e3d74a3 ext4: don't update checksum of new initialized bitmaps
> >
> > So far we've only seen failures on servers based on HiSilicon's family
> > of ARM64 SoCs (D05/Hi1616 SoC, D06/Hi1620 SoC). On these systems it is
> > very reproducible.
>
> Thanks for the report. Can you verify whether or not this patch fixes
> things for you?
hey Ted,
Sorry for the delayed response - was afk for a long weekend.
Your patch does seem to fix the issue for me - after applying the
patch, I was able to survive 20 iterations (and counting), where
previously it would always fail the first time.
However, I've received a conflicting report from a colleague who
appears to still be seeing errors. I'll get back to you ASAP once I am
able to (in-?)validate that observation.
-dann
> - Ted
>
> diff --git a/fs/ext4/ialloc.c b/fs/ext4/ialloc.c
> index da6c10c1e37a..1cfb74bc4dca 100644
> --- a/fs/ext4/ialloc.c
> +++ b/fs/ext4/ialloc.c
> @@ -90,6 +90,8 @@ static int ext4_validate_inode_bitmap(struct super_block
> *sb,
> return -EFSCORRUPTED;
>
> ext4_lock_group(sb, block_group);
> + if (buffer_verified(bh))
> + goto verified;
> blk = ext4_inode_bitmap(sb, desc);
> if (!ext4_inode_bitmap_csum_verify(sb, block_group, desc, bh,
> EXT4_INODES_PER_GROUP(sb) / 8)) {
> @@ -101,6 +103,7 @@ static int ext4_validate_inode_bitmap(struct super_block
> *sb,
> return -EFSBADCRC;
> }
> set_buffer_verified(bh);
> +verified:
> ext4_unlock_group(sb, block_group);
> return 0;
> }
