l-m-c with --rootfs btrfs fails because of unsupported -U option

[Alexander Sack]

trying btrfs with 0.4.1 l-m-c fails like (using mkfs.btrfs from maverick):

Formating root partition

mkfs.btrfs: invalid option -- 'U'
usage: mkfs.btrfs [options] dev [ dev ... ]
options:
  -A --alloc-start the offset to start the FS
  -b --byte-count total number of bytes in the FS
  -d --data data profile, raid0, raid1, raid10 or single
  -l --leafsize size of btree leaves
  -L --label set a label
  -m --metadata metadata profile, values like data profile
  -n --nodesize size of btree nodes
  -s --sectorsize min block allocation
Btrfs Btrfs v0.19
Traceback (most recent call last):
  File "./linaro-media-create", line 99, in <module>
    args.should_format_bootfs, args.should_format_rootfs)
  File "/home/asac/Development/linaro/imaging/linaro-image-tools/linaro_media_create/partitions.py", line 77, in setup_partitions
    proc.wait()
  File "/home/asac/Development/linaro/imaging/linaro-image-tools/linaro_media_create/cmd_runner.py", line 46, in wait
    raise SubcommandNonZeroReturnValue(self._my_args, returncode)
linaro_media_create.cmd_runner.SubcommandNonZeroReturnValue: Sub process "['sudo', 'mkfs.btrfs', '-U', '7d9887a4-98af-4a5e-a69a-757d9da45a3c', '/dev/mmcblk0p2', '-L', 'rootfs']" returned a non-zero value: 1

[Guilherme Salgado (salgado)]

AIUI, l-m-c specifies the UUID when creating the filesystem so that it knows what to use on /etc/fstab later on. I can't seem to find a way to set the UUID of a btrfs; does anybody know how to do so?

If that's not possible maybe we'll have to let mkfs.btrfs generate the UUID itself (which I assume it does) and then query that later to use on /etc/fstab.

[James Westby (james-w)]

On Mon, 24 Jan 2011 13:04:18 -0000, Guilherme Salgado <email address hidden> wrote:
> AIUI, l-m-c specifies the UUID when creating the filesystem so that it
> knows what to use on /etc/fstab later on. I can't seem to find a way to
> set the UUID of a btrfs; does anybody know how to do so?
>
> If that's not possible maybe we'll have to let mkfs.btrfs generate the
> UUID itself (which I assume it does) and then query that later to use on
> /etc/fstab.

About how the Ubuntu installer does it:

<ogra> it likely calls blkid <device>
<cjwatson> yes, we just use blkid
<cjwatson> we have no need to force the UUID to a specific value in d-i
<cjwatson> partman-btrfs just lets mkfs.btrfs pick something

Thanks,

James

[Loïc Minier (lool)]

I think we should just generate the fs and read it's UUID, no matter what the fs

[Guilherme Salgado (salgado)]

FWIW, this was not a regression introduced during the port to python.

[Loïc Minier (lool)]

With latest l-i-t, it should be possible to create btrfs images; they didn't boot for me in qemu as I was using an older hwpack, and that caused btrfs to fail to insmod.