On Fri, Jul 08, 2022 at 09:55:55AM -0600, Logan Gunthorpe wrote:
> I agree it's a mess, probably buggy and could use a cleanup with a
> free_disk method. But I'm not sure the all_mdevs lifetime issues are the
> problem here. If the entry in all_mdevs outlasts the disk, then
> md_alloc() will just fail earlier. Many test scripts rely on the fact
> that you can stop an mddev and recreate it immediately after. We need
> some way of ensuring any deleted disks are fully deleted before trying
> to make a new mddev, in case the new one has the same name as one being
> deleted.
I think those tests are broken. But fortunately that is just an
assumption in the tests, while device name reuse is a real problem.
I could not reproduce your problem, but on the for-5.20/block
tree I see a hang in 10ddf-geometry when running the tests.
The branch here:
git://git.infradead.org/users/hch/block.git md-lifetime-fixes
http://git.infradead.org/users/hch/block.git/shortlog/refs/heads/md-lifetime-fixes
fixes that for me and does not introduce new regressions. Can you
check if that helps your problem?
