Re: Possible bug in send/receive with simple reproducer

[Josef Bacik <jbacik@fusionio.com>]

On Mon, Aug 12, 2013 at 05:16:04PM +0200, Stefan Behrens wrote:
> On Mon, 12 Aug 2013 10:59:52 -0400, Josef Bacik wrote:
> > On Sun, Aug 11, 2013 at 09:53:01PM +0300, Emil Karlson wrote:
> >> Greetings
> >>
> >> Send fails for me unexpectedly:
> >>
> >> I get:
> >> ERROR: rename o262-5-0 -> snapshots failed. No such file or directory
> >>
> >> reproducer ( http://users.tkk.fi/~jkarlson/files/test4.txt  ):
> >>
> >> for i in 1 2; do
> >>   mkdir /mnt/$i
> >>   truncate -s 4G /mnt/$i.img
> >>   mkfs.btrfs /mnt/$i.img
> >>   mount -o loop /mnt/$i.img /mnt/$i
> >> done
> >>
> >> mkdir /mnt/1/testdir
> >> mkdir /mnt/1/testdir/1/
> >> mkdir /mnt/1/testdir/2/
> >> dd if=/dev/urandom  of=/mnt/1/testdir/aa count=16
> >> dd if=/dev/urandom  of=/mnt/1/testdir/bb count=16
> >>
> >> mkdir /mnt/1/snapshots
> >> btrfs sub snap -r /mnt/1/ /mnt/1/snapshots/backup2
> >> btrfs sub snap -r /mnt/1/ /mnt/1/snapshots/backup3
> >> btrfs send /mnt/1/snapshots/backup3/ | btrfs receive /mnt/2/
> >>
> >> umount /mnt/{1,2}; rm /mnt/{1,2}.img; rmdir /mnt/{1,2}
> > 
> > This was perfect, thank you for that.  I've posted a fix and I'll turn this into
> > an xfstest to make sure we don't ever regress.  Thanks again,
> >
> 
> If you create a subvolume below /mnt/[12] and run the test there, you
> have the problem reproduced again :)
> 

Yeah I noticed this problem when I converted it over to an xfstest, forgot that
I'm an idiot.  Trying to figure out a less braindead way to fix this.  Thanks,

Josef