Hi,
I am using a 2.6.32
linux host kernel which is running qemu-kvm version 0.12.1. I have a guest running Ubuntu using linux
kernel 3.13. The host hardware is Xeon 2680.
I am able to set up and use1G
huge pages on the host with kernel bootline having the parameters default_hugepagesz=1G,
hugepages=32, hugepagesz=1G. /proc/meminfo shows the correct huge page size. There is sufficient memory on the system. When
I boot the guest kernel with default_hugepagesz=1G, hugepages=32,
hugepagesz=1G, the guest complains that this is “Unsupported page size of
1024m”. If I check guest /proc/cpuinfo
the pdpe1gb flag is not set though this
is part of the xml config file in virsh.
I am able to use 2M huge
pages on both host and guest and can mmap and access the huge pages on both host
and guest.
I am not sure why 1G huge pages do not come up
on the guest. Is this because of an old qemu-kvm version 0.12.1 being used? At this point I need to keep the host kernel version as is. Are there any workarounds to solve this problem? What probably is the root cause?
Thanks,
S Kumar