Re: [libvirt] [PATCH v2 2/3] qemu: hotplug: Fix mlock limit handling on memory hotplug

On 11/09/2015 07:50 AM, Peter Krempa wrote: > If mlock is required either due to use of VFIO hostdevs or due to the > fact that it's enabled it needs to be tweaked prior to adding new memory > or after removing a module. Add a helper to determine when it's > necessary and reuse it both on hotplug and hotunplug. > > Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1273491 > --- > src/qemu/qemu_domain.c | 27 +++++++++++++++++++++++++++ > src/qemu/qemu_domain.h | 1 + > src/qemu/qemu_hotplug.c | 32 +++++++++++++++++++++++++++++--- > 3 files changed, 57 insertions(+), 3 deletions(-) >

> @@ -1802,16 +1803,25 @@ qemuDomainAttachMemory(virQEMUDriverPtr driver, > goto cleanup; > } > > + mlock = qemuDomainRequiresMlock(vm->def); > + > + if (mlock && > + virProcessSetMaxMemLock(vm->pid, > + qemuDomainGetMlockLimitBytes(vm->def)) < 0) { > + mlock = false; Coverity found a need for: virJSONValueFree(props); ACK - with the adjustment John BTW: If you think it matters - if we got the LimitBytes before inserting vm->mem to the domain def, then we could know whether it changed anda

whether we needed to Set it here and the error path. One would like to believe an initial setting to hard_limit (such as 4G) wouldn't have a failure if someone passed 4G again...