Re: [PATCH] iotests/108: probe loop device setup

[Eric Blake <eblake@redhat.com>]

On Tue, Oct 13, 2026 at 09:08:17PM +0200, Ilya Leoshkevich wrote:
> 
> 
> On 10/13/26 20:53, Eric Blake wrote:
> > On Tue, Oct 13, 2026 at 08:28:00PM +0200, Ilya Leoshkevich wrote:
> > > Running iotests/108 in a Docker container, that has passwordless sudo
> > > configured, fails with:
> > > 
> > >      losetup: /home/user/src/qemu/build/scratch/qcow2-file-108/t.qcow2: failed to set up loop device: No such file or directory
> > > 
> > > because it incorrectly assumes losetup is allowed to do loop device
> > > setup, when in reality it can only list devices.
> > > 
> > > Fix by probing loop device setup, using the test script as a backing
> > > file. With this, the test is correctly skipped.
> > > 
> > > Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
> > > ---
> > >   tests/qemu-iotests/108 | 3 ++-
> > >   1 file changed, 2 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/tests/qemu-iotests/108 b/tests/qemu-iotests/108
> > > index 1de718188aa..adbaaf87095 100755
> > > --- a/tests/qemu-iotests/108
> > > +++ b/tests/qemu-iotests/108
> > > @@ -53,7 +53,8 @@ _supported_os Linux
> > >   _unsupported_imgopts 'refcount_bits=\([^1]\|.\([^6]\|$\)\)' data_file
> > >   # This test either needs sudo -n losetup or FUSE exports to work
> > > -if sudo -n losetup &>/dev/null; then
> > > +if probe_dev=$(sudo -n losetup -r --show -f "$0" 2>/dev/null); then
> > > +    sudo -n losetup -d "$probe_dev"
> > 
> > Is 'sudo -n losetup -f' a simpler test?  It prints out the next
> > /dev/loopN value available only if a mount would succeed, which means
> > that it is a stronger test than the existing 'sudo -n losetup', but
> > also does not require a subsequent cleanup.
> 
> I don't think 'losetup -f' attempts a mount, since it works fine without
> privileges:
> 
> $ id -u
> 1000
> $ losetup -f
> /dev/loop0
> $ losetup -r -f /bin/true
> losetup: /bin/true: failed to set up loop device: Permission denied

Thanks for testing. In that case, I think your longer version is reasonable.

Reviewed-by: Eric Blake <eblake@redhat.com>

> 
> A quick glance at the source code [1] seems to confirm this: the
> iterator is immediately discarded after finding a free device name.
> The loop-control logic above looks similar.
> 
> [1]
> https://github.com/util-linux/util-linux/blob/v2.42.4/lib/loopdev.c#L1760
> 

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.
Virtualization:  qemu.org | libguestfs.org