Re: [PATCH] fstests: ext4: regression test for fsync transaction ids initialization

[Eryu Guan <eguan@redhat.com>]

On Tue, Jun 14, 2016 at 10:37:16AM +0800, Wang Xiaoguang wrote:
> Hello,
> 
> Thanks for reviewing.
> 
> On 06/08/2016 04:55 PM, Eryu Guan wrote:
[snip]
> >>+
> >>+loop_dev1=`_create_loop_device $fs_img1`
> >>+loop_dev2=`_create_loop_device $fs_img2`
> >You're using two loop devices and one is used as external journal
> >device, I think test on SCRATCH_DEV should be fine, e.g.
> >
> >_scratch_mkfs_sized a 8M ext4 filesystem, then use a script to find the
> >journal superblock on disk, then modify the journal.s_sequence according
> >to the location of journal superblock and the offset of s_sequence to
> >it. I use the following script to do all this:
> >
> >#/bin/bash
> >dev=$1
> >mkfs -t ext4 $dev 8M
> >blocksize=`dumpe2fs -h $dev 2>/dev/null | grep "Block size" | awk '{print $3}'`
> >offset=0
> >found=0
> ># this is the jbd2 journal superblock magic number on disk, in big endian
> >magic="c0 3b 39 98"
> >
> ># 8M in bytes
> >filesize=$((8 * 1024 * 1024))
> >while [ $offset -lt $filesize ]; do
> >         if  | grep -i "$magic";then
> >                 echo "Found journal: $offset"
> >                 found=1
> >                 break
> >         fi
> >         offset=$((offset + blocksize))
> >done
> Thanks for your script, but I wonder whether there will be some other
> metadata which would have the same magic number "c0 3b 39 98", such
> as ext4 super_block or block_group_description.

I don't think so. If they have the same magic number, it looks like a
bug to me.

Thanks,
Eryu