Re: [libvirt] [PATCH] Don't return an error if fail to create blkio controller

On 03/06/2011 08:49 PM, Hu Tao wrote: s/kenels/kernels/ s/why check/why we special case/ But I don't like this hunk in the first place - it means that if you have a newer kernel that does support subdirectories, you ignore the write failure to add the process into the group. Rather, it's easier to just nuke the mount point once we detect the problem (by pretending that blkio is not mounted, we don't ever try the SetValueU64 in the first place on an old kernel, while a new kernel doesn't need any code change here). Here's what I squashed in before pushing just the first hunk: diff --git i/src/util/cgroup.c w/src/util/cgroup.c index d3855c4..afe8731 100644 --- i/src/util/cgroup.c +++ w/src/util/cgroup.c @@ -529,10 +529,11 @@ static int virCgroupMakeGroup(virCgroupPtr parent, virCgroupPtr group, mkdir(path, 0755) < 0) { /* With a kernel that doesn't support multi-level directory * for blkio controller, libvirt will fail and disable all - * other controllers even though they are available. So skip - * blkio here if mkdir fails. */ + * other controllers even though they are available. So + * treat blkio as unmounted if mkdir fails. */ if (i == VIR_CGROUP_CONTROLLER_BLKIO) { rc = 0; + VIR_FREE(group->controllers[i].mountPoint); VIR_FREE(path); continue; } else { -- Eric Blake eblake(a)redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org