[PATCH v2 03/11] KVM: arm64: Make kvm_skip_instr() and co private to HYP

[Zenghui Yu]

On 2021/5/6 22:29, Marc Zyngier wrote:
> On Thu, 06 May 2021 12:43:26 +0100,
> Zenghui Yu <yuzenghui at huawei.com> wrote:
>>
>> On 2021/5/6 14:33, Marc Zyngier wrote:
>>> On Wed, 05 May 2021 17:46:51 +0100,
>>> Marc Zyngier <maz at kernel.org> wrote:
>>>>
>>>> Hi Zenghui,
>>>>
>>>> On Wed, 05 May 2021 15:23:02 +0100,
>>>> Zenghui Yu <yuzenghui at huawei.com> wrote:
>>>>>
>>>>> Hi Marc,
>>>>>
>>>>> On 2020/11/3 0:40, Marc Zyngier wrote:
>>>>>> In an effort to remove the vcpu PC manipulations from EL1 on nVHE
>>>>>> systems, move kvm_skip_instr() to be HYP-specific. EL1's intent
>>>>>> to increment PC post emulation is now signalled via a flag in the
>>>>>> vcpu structure.
>>>>>>
>>>>>> Signed-off-by: Marc Zyngier <maz at kernel.org>
>>>>>
>>>>> [...]
>>>>>
>>>>>> @@ -133,6 +134,8 @@ static int __kvm_vcpu_run_vhe(struct kvm_vcpu *vcpu)
>>>>>>  	__load_guest_stage2(vcpu->arch.hw_mmu);
>>>>>>  	__activate_traps(vcpu);
>>>>>> +	__adjust_pc(vcpu);
>>>>>
>>>>> If the INCREMENT_PC flag was set (e.g., for WFx emulation) while we're
>>>>> handling PSCI CPU_ON call targetting this VCPU, the *target_pc* (aka
>>>>> entry point address, normally provided by the primary VCPU) will be
>>>>> unexpectedly incremented here. That's pretty bad, I think.
>>>>
>>>> How can you online a CPU using PSCI if that CPU is currently spinning
>>>> on a WFI? Or is that we have transitioned via userspace to perform the
>>>> vcpu reset? I can imagine it happening in that case.
>>
>> I hadn't tried to reset VCPU from userspace. That would be a much easier
>> way to reproduce this problem.
> 
> Then I don't understand how you end-up there. If the vcpu was in WFI,
> it wasn't off and PSCI_CPU_ON doesn't have any effect.

I'm sorry for the misleading words.

The reported problem (secondary vcpu entry point corruption) was noticed
after a guest reboot. On rebooting, all vcpus will go back to userspace,
either because of a vcpu PSCI_SYSTEM_RESET call (with a
KVM_SYSTEM_EVENT_RESET system event in result), or because of a pending
signal targetting the vcpu thread. Userspace (I used QEMU) will then
perform the vcpu reset using the KVM_ARM_VCPU_INIT ioctl, of course!

WFI is the last instruction executed by the secondary vcpu before
rebooting. Emulating it results in a PC-altering flag.

What I was going to say is that maybe we can reproduce this problem with
a much simpler userspace program (not QEMU, no reboot) -- perform vcpu
reset while the vcpu is concurrently executing WFI, and see if the
result PC is set to 0 (per the KVM_ARM_VCPU_INIT doc). Maybe we can
achieve it with a kvm selftest case but "I hadn't tried", which turned
out to be misleading.

I'll have a look at your branch.


Thanks,
Zenghui