'git bisect' confirms that the NBD iotests 94 and 119 have been broken
since commit effd60c8 in v9.0.0; but as Dan Berrange's efforts to
improve CI have proven, we haven't been reliably running them to
notice.  The change was good (moving coroutine commands to run in the
right context), but it meant that "execute":"quit" now waits to
complete until the coroutines tearing down NBD have first reported the
SHUTDOWN event, in the opposite order of what happened pre-patch.

Signed-off-by: Eric Blake <eblake@redhat.com>
Fixes: effd60c8 ("monitor: only run coroutine commands in qemu_aio_context", v9.0.0)
Reported-by: Daniel P. Berrangé <berrange@redhat.com>
---
 tests/qemu-iotests/094.out | 2 +-
 tests/qemu-iotests/119.out | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/tests/qemu-iotests/094.out b/tests/qemu-iotests/094.out
index 97f894cf8f0..9178474e79f 100644
--- a/tests/qemu-iotests/094.out
+++ b/tests/qemu-iotests/094.out
@@ -23,6 +23,6 @@ Formatting 'TEST_DIR/source.IMGFMT', fmt=IMGFMT size=67108864
 {'execute': 'quit'}
 {"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "concluded", "id": "src"}}
 {"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "JOB_STATUS_CHANGE", "data": {"status": "null", "id": "src"}}
-{"return": {}}
 {"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "SHUTDOWN", "data": {"guest": false, "reason": "host-qmp-quit"}}
+{"return": {}}
 *** done
diff --git a/tests/qemu-iotests/119.out b/tests/qemu-iotests/119.out
index 7b7f0f4bcc7..45f82a4faa9 100644
--- a/tests/qemu-iotests/119.out
+++ b/tests/qemu-iotests/119.out
@@ -5,7 +5,7 @@ QMP_VERSION
 read 65536/65536 bytes at offset 0
 64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
 {"return": ""}
-{"return": {}}
 {"timestamp": {"seconds":  TIMESTAMP, "microseconds":  TIMESTAMP}, "event": "SHUTDOWN", "data": {"guest": false, "reason": "host-qmp-quit"}}
+{"return": {}}

 *** done
-- 
2.51.0

Am 13.10.2025 um 23:36 hat Eric Blake geschrieben:
> 'git bisect' confirms that the NBD iotests 94 and 119 have been broken
> since commit effd60c8 in v9.0.0; but as Dan Berrange's efforts to
> improve CI have proven, we haven't been reliably running them to
> notice.  The change was good (moving coroutine commands to run in the
> right context), but it meant that "execute":"quit" now waits to
> complete until the coroutines tearing down NBD have first reported the
> SHUTDOWN event, in the opposite order of what happened pre-patch.
> 
> Signed-off-by: Eric Blake <eblake@redhat.com>
> Fixes: effd60c8 ("monitor: only run coroutine commands in qemu_aio_context", v9.0.0)
> Reported-by: Daniel P. Berrangé <berrange@redhat.com>

Thanks, applied to the block branch.

Kevin

On Tue, Oct 14, 2025 at 08:42:40AM +0200, Kevin Wolf wrote:
> Am 13.10.2025 um 23:36 hat Eric Blake geschrieben:
> > 'git bisect' confirms that the NBD iotests 94 and 119 have been broken
> > since commit effd60c8 in v9.0.0; but as Dan Berrange's efforts to
> > improve CI have proven, we haven't been reliably running them to
> > notice.  The change was good (moving coroutine commands to run in the
> > right context), but it meant that "execute":"quit" now waits to
> > complete until the coroutines tearing down NBD have first reported the
> > SHUTDOWN event, in the opposite order of what happened pre-patch.
> > 
> > Signed-off-by: Eric Blake <eblake@redhat.com>
> > Fixes: effd60c8 ("monitor: only run coroutine commands in qemu_aio_context", v9.0.0)
> > Reported-by: Daniel P. Berrangé <berrange@redhat.com>
> 
> Thanks, applied to the block branch.

Looks like iotest fuse-allow-other suffers from the same problem
(skipped on -nbd, passes for -raw and -qcow2 in v8.2.0, and fails with
reordered output in v9.0.0).  I'll submit a followup patch for that
one as well.

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