>> >> The very last cases is the only one where this patch can potentially >> be beneficial. The problem is that because libvirt is in charge of >> enforcing CPU affinity, IIRC, we explicitly block QEMU from doing >> anything with CPU affinity. So AFAICT, this patch should result in >> an error from sched_setaffinity when run under libvirt. > > Yes, I had to disable capability dropping in qemu.conf. > > After all, I think maybe the right place to fix this is kernel? I mean, > why don't prealloc threads converge to the nodes they are working with?
Good question, I think the whole machinery doesn't support hugetlb, including recording remote faults and migrating the thread to a different node. -- Thanks, David / dhildenb