[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#950457: linux-image-5.4.0-0.bpo.2-amd64: Regression: mount option not correctly handled



Control: tag -1 upstream

On Sun, 2020-02-02 at 00:22 +0100, Vincent Danjean wrote:
> Package: src:linux
> Version: 5.4.8-1~bpo10+1
> Severity: important
> 
>   Hi,
> 
>   I'm using singularity on kvm Debian machines. After the last upgrade
> that installed the linux-image-5.4.0-0.bpo.2-amd64 kernel, I cannot
> start any singularity image. The error is:
> $ singularity -v -v shell /srv/scratch/atac-20180906-012322.simg
> [...]
> VERBOSE: Mounting squashfs image: /dev/loop0 -> /var/lib/singularity/mnt/container
> ERROR  : Failed to mount squashfs image in (read only): Invalid argument
> ABORT  : Retval = 255
> 
>   Using strace, I investiguate the problem, and find this:
> # mount -o ro,loop,offset=31,errors=remount-ro -t squashfs /srv/scratch/atac-20180906-012322.simg  /mnt/ 
> mount: /mnt: wrong fs type, bad option, bad superblock on /dev/loop0, missing codepage or helper program, or other error.
> # mount -o ro,loop,offset=31 -t squashfs /srv/scratch/atac-20180906-012322.simg  /mnt/

It seems that squashfs used to ignore unknown mount options, but
rejects them since:

commit 5a2be1288b514d74acdb3f0131d4d8fa3d689f06
Author: David Howells <dhowells@redhat.com>
Date:   Mon Mar 25 16:38:32 2019 +0000

    vfs: Convert squashfs to use the new mount API

Although ignoring mount options was a bug, this change is a user-space
regression and at least this option should still be ignored.

Ben.

> # ls /mnt
> [ all file of my singularity image ]
> 
>   With the previous installed kernel (5.3.0-0.bpo.2-amd64), the first mount
> (with the "errors=remount-ro" option) succeed. And, of course, strace told
> me that singularity is using the "errors=remount-ro" option...
> 
>   For now, I'm downgrading my kernel and using 5.3.0-0.bpo.2-amd64 as
> a workaround.
> 
>   Regards,
>     Vincent
> 
> PS: see https://github.com/sylabs/singularity/issues/4801 for
> the issue in singularity where it will be fixed (errors=remount-ro
> removed). But as I'm still using singularity from strech-backports,
> (singularity-container is not in buster and, in any case, I need
> to stick to 2.X version for singularity due to the use of datacenters
> where 3.X images are not yet supported)
[...]

-- 
Ben Hutchings
I haven't lost my mind; it's backed up on tape somewhere.


Reply to: