Hi,
the following patches add NUMA emulation to QEMU guests.
Although the ultimate goal is KVM with host side support, these
patches are pure QEMU with no host side binding.
This is a reworked version from end of last year, I adapted the command line
syntax to Anthony's wishes:
-numa node[,mem=<size>[MG]][,cpus=<from>[-<to>]][,nodeid=<nr>]
If we agree to this scheme (which drops mem=from-to and requires at least
one -numa node for each NUMA node), I will provide more detailed documentation.
Patch 1/4 adds the -numa command line parameter and sets a QEMU global
array with the parsed values. If no specific values for memory and CPUs are
given, all resources will be split equally across all nodes.
Patch 2/4 adds an "info numa" command to the monitor to output the current
topology. Since NUMA is advertised via static ACPI tables, no changes are
possible during runtime.
Patch 3/4 uses the QEMU firmware configuration interfacce to send the NUMA
topology to the BIOS, which has to setup the tables. Only one channel is used.
Patch 4/4 finally adds the BIOS support, which create the appropriate
SRAT table reflecting the given topology.
Looking forward to any comments.