public inbox for kvm@vger.kernel.org
 help / color / mirror / Atom feed
From: Marcelo Tosatti <mtosatti@redhat.com>
To: Avi Kivity <avi@redhat.com>
Cc: kvm@vger.kernel.org
Subject: Re: [PATCH v2 2/3] KVM: Optimize vcpu->requests slow path slightly
Date: Fri, 1 Jun 2012 21:23:03 -0300	[thread overview]
Message-ID: <20120602002303.GD5874@amt.cnet> (raw)
In-Reply-To: <4FC7396D.7050803@redhat.com>

On Thu, May 31, 2012 at 12:27:09PM +0300, Avi Kivity wrote:
> >> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> >> index 953e692..c0209eb 100644
> >> --- a/arch/x86/kvm/x86.c
> >> +++ b/arch/x86/kvm/x86.c
> >> @@ -5232,55 +5232,58 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu)
> >>  	bool req_int_win = !irqchip_in_kernel(vcpu->kvm) &&
> >>  		vcpu->run->request_interrupt_window;
> >>  	bool req_immediate_exit = 0;
> >> +	ulong reqs;
> >>  
> >>  	if (unlikely(req_int_win))
> >>  		kvm_make_request(KVM_REQ_EVENT, vcpu);
> >>  
> >>  	if (vcpu->requests) {
> >> -		if (kvm_check_request(KVM_REQ_MMU_RELOAD, vcpu))
> >> +		reqs = xchg(&vcpu->requests, 0UL);
> >> +
> >> +		if (test_bit(KVM_REQ_MMU_RELOAD, &reqs))
> >>  			kvm_mmu_unload(vcpu);
> >> -		if (kvm_check_request(KVM_REQ_MIGRATE_TIMER, vcpu))
> >> +		if (test_bit(KVM_REQ_MIGRATE_TIMER, &reqs))
> >>  			__kvm_migrate_timers(vcpu);
> >> -		if (kvm_check_request(KVM_REQ_CLOCK_UPDATE, vcpu)) {
> >> +		if (test_bit(KVM_REQ_CLOCK_UPDATE, &reqs)) {
> >>  			r = kvm_guest_time_update(vcpu);
> >>  			if (unlikely(r))
> >>  				goto out;
> >>  		}
> > 
> > Bailing out loses requests in "reqs". 
> 
> 
> Whoops, good catch.
> 
> > 
> > Caching the requests makes the following type of sequence behave strangely
> > 
> > req = xchg(&vcpu->requests);
> > if request is set
> >     request handler
> >         ...
> >         set REQ_EVENT
> >         ...
> > 
> > prepare for guest entry
> > vcpu->requests set
> >     bail
> 
> 
> I don't really mind that.  But I do want to reduce the overhead of a
> request, they're not that rare in normal workloads.
> 
> How about
> 
> 
>    for_each_set_bit(req, &vcpu->requests, BITS_PER_LONG) {
>        clear_bit(bit, &vcpu->requests);
>        r = request_handlers[bit](vcpu);
>        if (r)
>              goto out;
>    }
> 
> ? That makes for O(1) handling since usually we only have one request
> set (KVM_REQ_EVENT).  We'll make that the last one to avoid the scenario
> above.

That is better.


  reply	other threads:[~2012-06-02  0:37 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-05-20 13:49 [PATCH v2 0/3] Minor vcpu->requests improvements Avi Kivity
2012-05-20 13:49 ` [PATCH v2 1/3] KVM: Simplify KVM_REQ_EVENT/req_int_win handling Avi Kivity
2012-05-20 13:49 ` [PATCH v2 2/3] KVM: Optimize vcpu->requests slow path slightly Avi Kivity
2012-05-30 20:03   ` Marcelo Tosatti
2012-05-31  9:27     ` Avi Kivity
2012-06-02  0:23       ` Marcelo Tosatti [this message]
2012-05-20 13:49 ` [PATCH v2 3/3] KVM: Move mmu reload out of line Avi Kivity

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20120602002303.GD5874@amt.cnet \
    --to=mtosatti@redhat.com \
    --cc=avi@redhat.com \
    --cc=kvm@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox