Re: [PATCH v3 07/26] KVM: VMX: Move preemption timer <=> hrtimer dance to common x86

[Tudor Ambarus <tudor.ambarus@linaro.org>]

On 3/29/23 14:47, Paolo Bonzini wrote:

Hi, Paolo!

> On 3/29/23 14:34, Tudor Ambarus wrote:
>> This patch fixes the bug reported at:
>> LINK:
>> https://syzkaller.appspot.com/bug?id=489beb3d76ef14cc6cd18125782dc6f86051a605
>>
>> One may find the strace at:
>> LINK:https://syzkaller.appspot.com/text?tag=CrashLog&x=1798b54ec80000
>> and the c reproducer at:
>> LINK:https://syzkaller.appspot.com/text?tag=ReproC&x=10365781c80000
>>
>> Since I've no experience with kvm, it would be helpful if one of you can
>> provide some guidance. Do you think it is worth to backport this patch
>> to stable (together with its prerequisite patches), or shall I try to
>> get familiar with the code and try to provide a less invasive fix?
> 
> I think it is enough to fix the conflicts in vmx_pre_block and
> vmx_post_block, there are no prerequisites:
> 
> diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
> index 0718658268fe..895069038856 100644
> --- a/arch/x86/kvm/vmx/vmx.c
> +++ b/arch/x86/kvm/vmx/vmx.c
> @@ -7577,17 +7577,11 @@ static int vmx_pre_block(struct kvm_vcpu *vcpu)
>      if (pi_pre_block(vcpu))
>          return 1;
>  
> -    if (kvm_lapic_hv_timer_in_use(vcpu))
> -        kvm_lapic_switch_to_sw_timer(vcpu);
> -
>      return 0;
>  }
>  
>  static void vmx_post_block(struct kvm_vcpu *vcpu)
>  {
> -    if (kvm_x86_ops.set_hv_timer)
> -        kvm_lapic_switch_to_hv_timer(vcpu);
> -
>      pi_post_block(vcpu);
>  }
>  
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index fcfa3fedf84f..4eca3ec38afd 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -10022,12 +10022,28 @@ static int vcpu_enter_guest(struct kvm_vcpu
> *vcpu)
>  
>  static inline int vcpu_block(struct kvm *kvm, struct kvm_vcpu *vcpu)
>  {
> +    bool hv_timer;
> +
>      if (!kvm_arch_vcpu_runnable(vcpu) &&
>          (!kvm_x86_ops.pre_block || static_call(kvm_x86_pre_block)(vcpu)
> == 0)) {
> +        /*
> +         * Switch to the software timer before halt-polling/blocking as
> +         * the guest's timer may be a break event for the vCPU, and the
> +         * hypervisor timer runs only when the CPU is in guest mode.
> +         * Switch before halt-polling so that KVM recognizes an expired
> +         * timer before blocking.
> +         */
> +        hv_timer = kvm_lapic_hv_timer_in_use(vcpu);
> +        if (hv_timer)
> +            kvm_lapic_switch_to_sw_timer(vcpu);
> +
>          srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx);
>          kvm_vcpu_block(vcpu);
>          vcpu->srcu_idx = srcu_read_lock(&kvm->srcu);
>  
> +        if (hv_timer)
> +            kvm_lapic_switch_to_hv_timer(vcpu);
> +
>          if (kvm_x86_ops.post_block)
>              static_call(kvm_x86_post_block)(vcpu);
>  
> @@ -10266,6 +10282,11 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
>              r = -EINTR;
>              goto out;
>          }
> +        /*
> +         * It should be impossible for the hypervisor timer to be in
> +         * use before KVM has ever run the vCPU.
> +         */
> +        WARN_ON_ONCE(kvm_lapic_hv_timer_in_use(vcpu));
>          kvm_vcpu_block(vcpu);
>          if (kvm_apic_accept_events(vcpu) < 0) {
>              r = 0;
> 
> The fix is due to the second "if" changing from
> kvm_x86_ops.set_hv_timer to hv_timer.
> 

Thanks for the prompt answer! I fixed the conflicts as per your
suggestion and tested the patch with the reproducer on top of
stable/linux-5.15.y and I confirm the reproducer is silenced. Sent the
patch proposal (with you in To:) at:
https://lore.kernel.org/all/20230329151747.2938509-1-tudor.ambarus@linaro.org/T/#u
Feel free to ACK/NACK it.

Cheers,
ta