Re: [libvirt] [PATCH 6/6] qemu: process: Don't redetect backing chain on reconnect
On 10/04/2017 07:59 AM, Peter Krempa wrote:
Skip purging the backing chain and redetecting it when it was not
going
to change during the time we were not present.
The decision is based on the new flag which records whether there were
blockjobs running to the status XML.
---
src/qemu/qemu_process.c | 19 ++++++++++++-------
1 file changed, 12 insertions(+), 7 deletions(-)
diff --git a/src/qemu/qemu_process.c b/src/qemu/qemu_process.c
index bde3ba462..80d204599 100644
--- a/src/qemu/qemu_process.c
+++ b/src/qemu/qemu_process.c
@@ -6892,13 +6892,18 @@ qemuProcessReconnect(void *opaque)
BTW: I'd support a
virDomainDiskDefPtr disk = obj->def->disks[i];
right about here too ;-)
if (virStorageTranslateDiskSourcePool(conn,
obj->def->disks[i]) < 0)
goto error;
- /* XXX we should be able to restore all data from XML in the future.
- * This should be the only place that calls qemuDomainDetermineDiskChain
- * with @report_broken == false to guarantee best-effort domain
- * reconnect */
- if (qemuDomainDetermineDiskChain(driver, obj, obj->def->disks[i],
- true, false) < 0)
- goto error;
+ /* backing chains need to be refreshed only if they could change */
+ if (!priv->reconnectNoActiveBlockjobs) {
...or from my previous patch suggestion
if (priv->NAME != VIR_TRISTATE_BOOL_NO)
Reviewed-by: John Ferlan <jferlan(a)redhat.com>
John
I can give an R-B on patch 5 depending on your thoughts there. Maybe I'm
missing something in future attractions...
+ /* This should be the only place that calls
+ * qemuDomainDetermineDiskChain with @report_broken == false
+ * to guarantee best-effort domain reconnect */
+ if (qemuDomainDetermineDiskChain(driver, obj, obj->def->disks[i],
+ true, false) < 0)
+ goto error;
+ } else {
+ VIR_DEBUG("skipping backing chain detection for '%s'",
+ obj->def->disks[i]->dst);
+ }
dev.type = VIR_DOMAIN_DEVICE_DISK;
dev.data.disk = obj->def->disks[i];