Re: large file system & high object count testing

[Andreas Dilger <adilger@sun.com>]

On Aug 31, 2009  16:37 -0700, Justin Maggard wrote:
> On Mon, Aug 31, 2009 at 4:13 PM, Andreas Dilger<adilger@sun.com> wrote:
> > Justin, could you please replicate this corruption, collecting some
> > additional information before & after.  My recollection is that the
> > corruption appears in the first few groups, so 64kB should be plenty
> > to capture the group descriptor tables (where the checksum is kept).
> >
> > - mke2fs
> > - dd if=/dev/XXX bs=4k count=16 | gzip -9 > /tmp/gdt-new.gz
> > - mkdir ...
> > - sync
> > - dd if=/dev/XXX bs=4k count=16 | gzip -9 > /tmp/gdt-mkdir.gz
> > - umount
> > - dd if=/dev/XXX bs=4k count=16 | gzip -9 > /tmp/gdt-umount.gz
> > - e2fsck
> > - dd if=/dev/XXX bs=4k count=16 | gzip -9 > /tmp/gdt-e2fsck.gz
> >
> 
> No problem.  I just sent you an email with those four attached.  If
> anyone would like me to upload them somewhere else, just let me know.

Comparing the GDT dumps you gave makes it fairly clear what is wrong:

--- gdt-umount.od       2009-09-02 02:54:40.148704651 -0600
+++ gdt-e2fsck.od       2009-09-02 02:54:54.809699151 -0600
 001000 00000ee5 00000ef5 00000f05 07f568f0
 001010 00040002 00000000 00000000 339107f5
 001020 00000000 00000000 00000000 00000000
-*
+001030 00000000 00000000 00000000 000007fe

It seems that e2fsck isn't keeping one of the reserved fields zero,
so this is confusing the checksum.

truct ext4_group_desc
{
:
:
/*30*/  __le16  bg_used_dirs_count_hi;  /* Directories count MSB */
        __le16  bg_itable_unused_hi;    /* Unused inodes count MSB */
        __u32   bg_reserved2[3];
}

The bg_reserved[2] field is being changed incorrectly.

Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.

--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html