[Btrfs-devel] Crash with reusing fs; on big volumes

Chris mason chris.mason at oracle.com
Sun Jan 20 16:26:15 PST 2008


On Saturday 19 January 2008, Jan Engelhardt wrote:
> Hi,
>
>
> so I just tried this...
>
> 17:08 localhost:~/btr/btrfs-progs-0.11 # mdadm -C /dev/md0 -fe 1.0 -l
> linear -n 59 /dev/sd[b-z] /dev/sda[a-z] /dev/sdb[a-h]
> mdadm: array /dev/md0 started.
>
> 17:09 localhost:~/btr/btrfs-progs-0.11 # fdisk -l /dev/md0
>
> Disk /dev/md0: 60183.2 GB, 60183225368576 bytes
> 2 heads, 4 sectors/track, -1 cylinders
> Units = cylinders of 8 * 512 = 4096 bytes
> Disk identifier: 0x00000000
>
> Disk /dev/md0 doesn't contain a valid partition table
>
> 16:54 localhost:~/btr/btrfs-progs-0.11 # ./mkfs.btrfs /dev/md0
> fs created on /dev/md0 nodesize 16384 leafsize 16384 sectorsize 4096
> bytes 60183225368576
>
> 17:09 localhost:~/btr/btrfs-progs-0.11 # mount -t btrfs /dev/md0 /mnt
>
> 17:09 localhost:~/btr/btrfs-progs-0.11 # df -Th
> Filesystem    Type    Size  Used Avail Use% Mounted on
> /dev/sda3      xfs    7.3G  1.1G  6.2G  16% /
> udev         tmpfs     62M  316K   62M   1% /dev
> ichi:/home     nfs    199G  163G   37G  82% /home
> /dev/md0     btrfs     55T  7.5M   55T   1% /mnt
>
> Looking good. So far.
>
> 17:16 localhost:~/btr/btrfs-progs-0.11 # mdadm -C /dev/md0 -fe 1.0 -l
> linear -n 16 /dev/sd[b-q]
> mdadm: array /dev/md0 started.
>
> 17:16 localhost:~/btr/btrfs-progs-0.11 # ./mkfs.btrfs /dev/md0
> fs created on /dev/md0 nodesize 16384 leafsize 16384 sectorsize 4096
> bytes 16320874676224
> 17:16 localhost:~/btr/btrfs-progs-0.11 # mount -t btrfs /dev/md0 /mnt
> Segmentation fault
>
>
> I have experienced this crash before (what I wrote yesterday in IRC).
> To reproduce, create a volume, mkfs.btrfs on it, mount, umount, shrink
> the volume, mkfs again, mount -> crash.
> (Just what I have presented here.)
>
> The actual size of the device does not seem to matter. Yesterday I
> had the md sized with 9 TB, then 21 TB, went back to 16 (crash). Went
> back to 8 (ok - huh?), went up to 9 (crash). Then tried 59 now (ok), 58
> (still ok), 16 (crash).

Sorry, I'm a little confused on the test case.  Since this is i386, could we 
try again with values that are all below 16TB?  Also, could I get an exact 
list of steps to reproduce?

Thanks for trying things out on these big arrays!

-chris




More information about the Btrfs-devel mailing list