mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* Re: [PATCH] KVM:Hyper-V reduce one kvm_write_guest operation
       [not found] <d1e55d03-1a21-e42b-8d10-af55e67eb567@gmail.com>
@ 2017-12-20  9:01 ` Paolo Bonzini
       [not found]   ` <CABvbLL00cQP4Xbgz7T-4WqRu6Ar8dWBLtuTqgvadeaahsUMowg@mail.gmail.com>
  0 siblings, 1 reply; 2+ messages in thread
From: Paolo Bonzini @ 2017-12-20  9:01 UTC (permalink / raw)
  To: rhett, rkrcmar, tglx, mingo, hpa, kvm, linux-kernel

On 20/12/2017 08:46, rhett wrote:
> in function kvm_hv_setup_tsc_page , the old code write the full tsc_ref
> struct firstly, and write a
> tsc_sequence field later, it can be wirten once.

No, it cannot and this comment says exactly why:

> -       /* Ensure sequence is zero before writing the rest of the struct.  */
> -       smp_wmb();
> -       if (kvm_write_guest(kvm, gfn_to_gpa(gfn), &hv->tsc_ref,
> sizeof(hv->tsc_ref)))
> -               goto out_unlock;
> -
>         /*
>          * Now switch to the TSC page mechanism by writing the sequence.
>          */

The sequence is: disable TSC page, write TSC parameters, enable TSC
page.  If the guest can read a partially-written TSC page, it can return
a wrong time.

Paolo

> @@ -922,7 +917,7 @@ void kvm_hv_setup_tsc_page(struct kvm *kvm,
>  
>         hv->tsc_ref.tsc_sequence = tsc_seq;
>         kvm_write_guest(kvm, gfn_to_gpa(gfn),
> -                       &hv->tsc_ref, sizeof(hv->tsc_ref.tsc_sequence));
> +                       &hv->tsc_ref, sizeof(hv->tsc_ref));
>  out_unlock:
>         mutex_unlock(&kvm->arch.hyperv.hv_lock);
>  }
> -- 
> 1.8.3.1
> 

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [PATCH] KVM:Hyper-V reduce one kvm_write_guest operation
       [not found]   ` <CABvbLL00cQP4Xbgz7T-4WqRu6Ar8dWBLtuTqgvadeaahsUMowg@mail.gmail.com>
@ 2017-12-20  9:45     ` Paolo Bonzini
  0 siblings, 0 replies; 2+ messages in thread
From: Paolo Bonzini @ 2017-12-20  9:45 UTC (permalink / raw)
  To: Rhett Rhett
  Cc: Radim Krčmář, tglx, mingo, hpa, kvm, linux-kernel

On 20/12/2017 10:33, Rhett Rhett wrote:
> tsc_sequence  had been set to 0 before,  so the tsc page won't function.
> guest has no chance to read tsc page before the last write.
> 

kvm_write_guest is not atomic.  The guest can see a non-zero sequence
before the other fields have been set.

Paolo

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2017-12-20  9:45 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
     [not found] <d1e55d03-1a21-e42b-8d10-af55e67eb567@gmail.com>
2017-12-20  9:01 ` [PATCH] KVM:Hyper-V reduce one kvm_write_guest operation Paolo Bonzini
     [not found]   ` <CABvbLL00cQP4Xbgz7T-4WqRu6Ar8dWBLtuTqgvadeaahsUMowg@mail.gmail.com>
2017-12-20  9:45     ` Paolo Bonzini

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®