On 06/18/2013 04:36 AM, Osier Yang wrote:
The translation must be done before both of cgroup and security setting, otherwise since the disk source is not translated yet, it might be skipped on cgroup and security setting. --- src/qemu/qemu_process.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-)
I don't now about this one. What concerns me is the code that was run just before this movement, specifically: /* * Normally PCI addresses are assigned in the virDomainCreate * or virDomainDefine methods. We might still need to assign * some here to cope with the question of upgrades. Regardless * we also need to populate the PCi address set cache for later * use in hotplug */ if (virQEMUCapsGet(priv->qemuCaps, QEMU_CAPS_DEVICE)) { VIR_DEBUG("Assigning domain PCI addresses"); if ((qemuDomainAssignAddresses(vm->def, priv->qemuCaps, vm)) < 0) goto cleanup; } I didn't chase into the code, but I would think that PCI assignment could play a role in disk placements. I guess, I'm not comfortable with taking that leap of faith - perhaps someone else knows for certain. John
diff --git a/src/qemu/qemu_process.c b/src/qemu/qemu_process.c index c412ea2..351440b 100644 --- a/src/qemu/qemu_process.c +++ b/src/qemu/qemu_process.c @@ -3552,6 +3552,14 @@ int qemuProcessStart(virConnectPtr conn, } hookData.nodemask = nodemask;
+ /* "volume" type disk's source must be translated before + * cgroup and security setting. + */ + for (i = 0; i < vm->def->ndisks; i++) { + if (qemuTranslateDiskSourcePool(conn, vm->def->disks[i]) < 0) + goto cleanup; + } + VIR_DEBUG("Setting up domain cgroup (if required)"); if (qemuSetupCgroup(driver, vm, nodemask) < 0) goto cleanup; @@ -3598,11 +3606,6 @@ int qemuProcessStart(virConnectPtr conn, goto cleanup; }
- for (i = 0; i < vm->def->ndisks; i++) { - if (qemuTranslateDiskSourcePool(conn, vm->def->disks[i]) < 0) - goto cleanup; - } - VIR_DEBUG("Building emulator command line"); if (!(cmd = qemuBuildCommandLine(conn, driver, vm->def, priv->monConfig, priv->monJSON, priv->qemuCaps,