Re: [PATCH] KVM: x86: revert the order of calls in kvm_fast_pio()

[Sean Christopherson]

On Tue, Apr 30, 2019 at 04:32:01PM +0200, Peter Zijlstra wrote:
> On Tue, Apr 30, 2019 at 10:24:23PM +0800, Perr Zhang wrote:
> > In commit 45def77ebf79, the order of function calls in kvm_fast_pio()
> > was changed. This causes that the vm(XP,and also XP's iso img) failed
> > to boot. This doesn't happen with win10 or ubuntu.
> > 
> > After revert the order, the vm(XP) succeedes to boot. In addition, the
> > change of calls's order of kvm_fast_pio() in commit 45def77ebf79 has no
> > obvious reason.

There are three reasons explicitly listed in the changelog:

    Updating %rip prior to executing to userspace has several drawbacks:
    
      - Userspace sees the wrong %rip on the exit, e.g. if PIO emulation
        fails it will likely yell about the wrong address.
      - Single step exits to userspace for are effectively dropped as
        KVM_EXIT_DEBUG is overwritten with KVM_EXIT_IO.
      - Behavior of PIO emulation is different depending on whether it
        goes down the fast path or the slow path.

> 
> This Changelog fails to explain why the order is important and equally
> fails to inform the future reader of that code. So this very same thing
> will happen again in 6 months time or thereabout.

There's a more precise fix submitted for this bug[1].  In theory v2
already went out, but I still don't see it posted to the KVM list.
Either the KVM list or my mail client is being weird.

[1] https://patchwork.kernel.org/patch/10919849/