It turns out PV EOI gets disabled after migration -
until next guest reset.
This is because we are missing code to actually migrate it.
This patch fixes it up: it applies cleanly to qemu.git
as well as qemu-kvm.git, so I think it's cleaner
to apply it in qemu.git to keep diff to minimum.
Note: there's talk about adding infrastructure for
CPUID whitelisting which thinkably could be used
for migration compat support. I am guessing this won't be
1.2 material - when it's ready we can easily replace
a simple flag that this patchset adds with something else.
So this just adds minimal code to avoid regressing
cross-version migration.
Note: there's a kernel bug in linux 3.6-rc3 - apply
my patch 'kvm: fix KVM_GET_MSR for PV EOI' in order to
use this patchset on it.
Needed for 1.2.
Changes from v2:
Fix HACKING rules violation (Blue Swirl)
Changes from v1:
Update all headers from 3.6-rc3 to keep them in sync (Jan)
Disable cpuid flag for qemu 1.2 and older (Orit)
Michael S. Tsirkin (4):
linux-headers: update to 3.6-rc3
pc: refactor compat code
cpuid: disable pv eoi for 1.1 and older compat types
kvm: get/set PV EOI MSR
hw/Makefile.objs | 2 +-
hw/cpu_flags.c | 32 +++++++++++++++++++++++++++
hw/cpu_flags.h | 9 ++++++++
hw/pc_piix.c | 46 ++++++++++++++++++++++++++++++++-------
linux-headers/asm-s390/kvm.h | 2 +-
linux-headers/asm-s390/kvm_para.h | 2 +-
linux-headers/asm-x86/kvm.h | 1 +
linux-headers/asm-x86/kvm_para.h | 7 ++++++
linux-headers/linux/kvm.h | 3 +++
target-i386/cpu.c | 8 +++++++
target-i386/cpu.h | 1 +
target-i386/kvm.c | 13 +++++++++++
target-i386/machine.c | 21 ++++++++++++++++++
13 files changed, 136 insertions(+), 11 deletions(-)
create mode 100644 hw/cpu_flags.c
create mode 100644 hw/cpu_flags.h
--
MST
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
