Re: [PATCH hci-dev 0/2] qcow2, block: fix reopen of a dirty read-only node and of a corrupt node #VSTOR-140073

[Andrey Drobyshev <andrey.drobyshev@virtuozzo.com>]

On 11/4/26 4:41 PM, Denis V. Lunev wrote:
> These are patches 1 and 2 of the hci-8.0 posting of September 16, merged
> upstream as af28b65fae6 and efe98c2a093. The repair of a dirty image
> that becomes writable is not part of this.
> 
> - A read-only to read-only reopen of a dirty qcow2 image fails, or
>   aborts on the BLK_PERM_WRITE assertion in bdrv_co_write_req_prepare()
>   when the file node below is writable.
> - A reopen of a node which qcow2_signal_corruption() left without a
>   driver crashes QEMU. commit_clean() reopens the base of a commit job
>   this way.
> 
> Patch 1 is a clean pick. Patch 2 differs only in 060.out: HMP qemu-io on
> this tree prints a failed read on the monitor instead of returning it.
> 
> Tested on dev-cl04. Unpatched hci-dev fails the new 039 and 060 cases
> with that assertion and with a segfault; with the series both pass. The
> other reopen iotests, qcow2 and raw, give the same result with and
> without it.
> 
> Denis V. Lunev (2):
>   qcow2: do not clear the dirty bit when reopening a read-only node
>     #VSTOR-140073
>   block: reject a reopen of an unusable node instead of crashing
>     #VSTOR-140073
> 
>  block.c                    | 14 ++++++++++-
>  block/qcow2.c              |  8 +++---
>  tests/qemu-iotests/039     | 50 ++++++++++++++++++++++++++++++++++++++
>  tests/qemu-iotests/039.out | 20 +++++++++++++++
>  tests/qemu-iotests/060     | 30 +++++++++++++++++++++++
>  tests/qemu-iotests/060.out | 13 ++++++++++
>  6 files changed, 131 insertions(+), 4 deletions(-)
> 
> 
> base-commit: ca7715cc68cdd379b9fbe32d794408bd87a6677e

Ack
Applied to hci-dev