On 3/10/20 10:54 AM, Peter Krempa wrote:
oVirt used a quirk in the pre-blockdev semantics of drive-mirror which
opened the backing chain of the mirror destination only once
'block-job-complete' was called.
Our introduction of blockdev made qemu open the backing chain images
right at the start of the job. This broke oVirt's usage of this API
because they copy the data into the backing chain during the time the
block copy job is running.
Re-introduce late open of the backing chain if qemu allows to use
reminder: "allows to ${verb}" is not idiomatic; I'd suggest either
"allows us to use" or "allows the use of"
blockdev-snapshot on write-only nodes as it can be used to install the
backing chain even for an existing image now.
Signed-off-by: Peter Krempa <pkrempa@xxxxxxxxxx>
---
src/qemu/qemu_driver.c | 57 +++++++++++++++++++++++++++++++++++++++---
1 file changed, 53 insertions(+), 4 deletions(-)
@@ -17273,6 +17276,27 @@ qemuDomainBlockPivot(virQEMUDriverPtr driver,
blockNamedNodeData,
shallow, &actions) < 0)
return -1;
+
+ /* Open and install the backing chain of 'mirror' late if we can use
+ * blockdev-snapshot to do it. This is to appease oVirt that wants
+ * to copy data into the backing chain while the top image is being
+ * copied shallow */
+ if (reuse && shallow &&
+ virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_BLOCKDEV_SNAPSHOT_ALLOW_WRITE_ONLY) &&
+ virStorageSourceHasBacking(disk->mirror)) {
+
+ if (!(chainattachdata = qemuBuildStorageSourceChainAttachPrepareBlockdev(disk->mirror->backingStore,
+ priv->qemuCaps)))
+ return -1;
Long lines, but I'm not sure how to improve the situation.
The comments are good at explaining the situation - we have a window of
qemu releases that regress when using -blockdev, but as long as oVirt
can force either the old -drive behavior or insist on new-enough libvirt
coupled with new-enough qemu that restores the write-only-reopen feature
that we need, then oVirt won't hit the regression. I'm not sure how you
plan to advertise to oVirt if this is a new-enough libvirt + detection
of new-enough qemu to tell oVirt they don't need to cobble libvirt into
using -drive rather than -blockdev (they might solve that by minimum
required versions, rather than having to ask libvirt), but answering
that question doesn't interfere with the validity of this patch.
Reviewed-by: Eric Blake <eblake@xxxxxxxxxx>
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3226
Virtualization: qemu.org | libvirt.org
