This fixes a regression introduced in commit 25fedfca94cf, "KVM: PPC:
Book3S HV: Move vcore preemption point up into kvmppc_run_vcpu", which
leads to a user-triggerable oops.
In the case where we try to run a vcore on a physical core that is
not in single-threaded mode, or the vcore has too many t
On 28/04/2015 16:10, Christian Borntraeger wrote:
> > Alternatively, the irq-disabled versions could be called
> > __kvm_guest_{enter,exit}. Then you can use those directly when it makes
> > sense.
>
> ..having a special __kvm_guest_{enter,exit} without the WARN_ON might be even
> the cheapest
Am 28.04.2015 um 13:37 schrieb Paolo Bonzini:
>> --- a/arch/powerpc/kvm/book3s_pr.c
>> +++ b/arch/powerpc/kvm/book3s_pr.c
>> @@ -891,7 +891,9 @@ int kvmppc_handle_exit_pr(struct kvm_run *run, struct
>> kvm_vcpu *vcpu,
>>
>> /* We get here with MSR.EE=1 */
>>
>> +local_irq_disable();
>
On 28/04/2015 12:32, Christian Borntraeger wrote:
> Some architectures already have irq disabled when calling
> kvm_guest_exit. Push down the disabling into the architectures
> to avoid double disabling. This also allows to replace
> irq_save with irq_disable which might be cheaper.
> arm and mip
Some architectures already have irq disabled when calling
kvm_guest_exit. Push down the disabling into the architectures
to avoid double disabling. This also allows to replace
irq_save with irq_disable which might be cheaper.
arm and mips already have interrupts disabled. s390/power/x86
need adopti
local_irq_disable can be cheaper than local_irq_save, especially
when done only once instead of twice. We can push down the
local_irq_save (and replace it with local_irq_disable) to
save some cycles.
x86, mips and arm already disable the interrupts before calling
kvm_guest_enter. Here we save one l
I was able to get rid of some nanoseconds for a guest exit loop
on s390. I did my best to not break other architectures but
review and comments on the general approach is welcome.
Downside is that the existing irq_save things will just work
no matter what the callers have done, the new code must do
On Tue, Apr 28, 2015 at 10:36:52AM +0530, Aneesh Kumar K.V wrote:
> Paul Mackerras writes:
>
> > The reference (R) and change (C) bits in a HPT entry can be set by
> > hardware at any time up until the HPTE is invalidated and the TLB
> > invalidation sequence has completed. This means that when