Re: [PATCH] block: qcow2-bitmap: fix persistent bitmap corruption on ENOSPC

[Kevin Wolf <kwolf@redhat.com>]

Am 16.06.2026 um 11:06 hat Vladimir Sementsov-Ogievskiy geschrieben:
> 10.06.26 05:33, Christian Barry пишет:
> > From: Christian Barry <christian.barry@usp.br>
> > 
> > Changed qcow2_co_remove_persistent_dirty_bitmap() to treat free_bitmap_clusters()
> > failure as fatal and roll back the header update through a new helper
> > function rollback_ext_header_and_dir_helper(), which was also applied
> > to update_ext_header_and_dir(), which had internal rollback logic.
> > Changed bitmap_list_store() to zero the cluster tail after write,
> > guaranteeing that qemu-img won't parse into stale bytes.
> > 
> > Resolves: https://gitlab.com/qemu-project/qemu/-/work_items/3516
> 
> Hi!
> 
> First, seem like two different fixes merged into one commit, please if
> possible, make them separate two commits.
> 
> Next, please, make commit messages more descriptive, about what the
> issue we solve (link is good as additional information, but the commit
> message itself should be complete).
> 
> For example, I don't understand free_bitmap_clusters makes the image
> unreadable. And we we fix it in one place, leaving other calls to
> free_bitmap_clusters() unchanged. Could we fix free_bitmap_clusters()
> itself? See also my comments below

I haven't looked at the patch yet, but this sounds like apart from
splitting the patches into one logical change per patch and improving
the commit message, the patch series should also include a test case
(probably qemu-iotests) that demonstrates the problem and how the patch
changes behaviour.

Kevin