Re: [PATCH v4 for v7.6.0 00/14] Introduce virtio-mem <memory/> model

On 7/7/21 12:30 PM, David Hildenbrand wrote: > On 23.06.21 12:12, Michal Privoznik wrote: >> v4 of: >> >> https://listman.redhat.com/archives/libvir-list/2021-April/msg01138.html >> >> diff to v3: >> - Rebased code on the top of master >> - Tried to work in all Peter's review suggestions >> - Fixed a bug where adjusting <requested/> was viewed as hotplug of new >>    <memory/> by XML validator and thus if <maxMemory/> was close >> enough to >>    <currentMemory/> the validator reported an error (this was reported by >>    David). >> > > Hi Michal, > > 6. 4k source results in an error > >     <source> >       <pagesize unit='KiB'>4096</pagesize> >       <nodemask>0-1</nodemask> >     </source> > > "error: internal error: Unable to find any usable hugetlbfs mount for > 4096 KiB" > > This example is taken from https://libvirt.org/formatdomain.html for > DIMMs. Not sure what the expected behavior is. Just realized that this IS expected behavior. 4096KiB pages (=4MiB) are not regular system pages (4KiB). Thus libvirt is trying to find hugetlbfs mount point that's serving 4MiB pages, unsuccessfully. I'll post a patch that fixes the example though.