[libvirt] [PATCH] qemu_command: Count the cpus correctly when the <vcpus> are absent
qemuBuildSmpCommandLine - is ending up generating "-smp 0," after first/subsequent restarts of the guest and we see "error: internal error: qemu reported thread id for inactive vcpu 'X'" The PostParse logic which changes the hotplug = ABSENT to NO for each vcpus is not called with virDomainDefCopy(because of SKIP_POST_PARSE) after first/subsequent restarts of the guest. Skipping post-parse is fine for all devices as the changes done as part of previous post-parse are captured in virDomainDefFormat and effectively copied during virDomainDefCopy. With cpu-hotplug, the def->individualvcpus is what determines whether the <vcpus> are to be formatted or not in the xml. Since individualvcpus is not set, the copies will have hotpluggable as ABSENT. The fix here is qemuBuildSmpCommandLine to check if the individualvcpus are set OR check the ABSENT flag. We dont want to set the individualvcpus in post-parse as that would disallow migration backwards to qemus not supporting hotplug. So, set check the ABSENT flag too when counting the vcpus instead. Signed-off-by: Shivaprasad <sbhat@linux.vnet.ibm.com> --- src/qemu/qemu_command.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/src/qemu/qemu_command.c b/src/qemu/qemu_command.c index 529bcb1..0a3c9ed 100644 --- a/src/qemu/qemu_command.c +++ b/src/qemu/qemu_command.c @@ -7113,7 +7113,8 @@ qemuBuildSmpCommandLine(virCommandPtr cmd, * in a different way */ for (i = 0; i < maxvcpus; i++) { vcpu = virDomainDefGetVcpu(def, i); - if (vcpu->online && vcpu->hotpluggable == VIR_TRISTATE_BOOL_NO) + if (vcpu->online && ((vcpu->hotpluggable == VIR_TRISTATE_BOOL_NO) || + (vcpu->hotpluggable == VIR_TRISTATE_BOOL_ABSENT))) nvcpus++; }
On Tue, Oct 04, 2016 at 17:22:37 +0530, Shivaprasad wrote:
qemuBuildSmpCommandLine - is ending up generating "-smp 0," after first/subsequent restarts of the guest and we see "error: internal error: qemu reported thread id for inactive vcpu 'X'"
The PostParse logic which changes the hotplug = ABSENT to NO for each vcpus is not called with virDomainDefCopy(because of SKIP_POST_PARSE) after first/subsequent restarts of the guest. Skipping post-parse is fine for all devices as the changes done as part of previous post-parse are captured in virDomainDefFormat and effectively copied during virDomainDefCopy. With cpu-hotplug, the def->individualvcpus is what determines whether the <vcpus> are to be formatted or not in the xml. Since individualvcpus is not set, the copies will have hotpluggable as ABSENT.
The fix here is qemuBuildSmpCommandLine to check if the individualvcpus are set OR check the ABSENT flag. We dont want to set the individualvcpus in post-parse as that would disallow migration backwards to qemus not supporting hotplug. So, set check the ABSENT flag too when counting the vcpus instead.
Signed-off-by: Shivaprasad <sbhat@linux.vnet.ibm.com> --- src/qemu/qemu_command.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
Thanks for discovering this rather serious regression. Currently we've decided to revert the patches that skip the post parse callbacks since that's a more safe thing to do. After the release we'll revisit the post parse callback change to do the proper thing, so this patch might not be necessary then. Peter
participants (2)
-
Peter Krempa -
Shivaprasad