kvm.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Gleb Natapov <gleb@redhat.com>
To: Marcelo Tosatti <mtosatti@redhat.com>
Cc: avi@redhat.com, kvm@vger.kernel.org
Subject: Re: [PATCH v3 1/8] Do not call ack notifiers on PIC reset.
Date: Thu, 13 Aug 2009 12:39:34 +0300	[thread overview]
Message-ID: <20090813093934.GA23736@redhat.com> (raw)
In-Reply-To: <20090813091105.GA17022@amt.cnet>

On Thu, Aug 13, 2009 at 06:11:05AM -0300, Marcelo Tosatti wrote:
> On Wed, Aug 12, 2009 at 03:17:15PM +0300, Gleb Natapov wrote:
> > For device assigned it may cause host hang since ack notifier callback
> > enables host interrupt and guest not necessary cleared interrupt
> > condition in an assigned device. For PIT we should not call ack notifier
> > here since interrupt was not acked by a guest and should be redelivered.
> > 
> > Signed-off-by: Gleb Natapov <gleb@redhat.com>
> > ---
> >  arch/x86/kvm/i8259.c |   16 ----------------
> >  1 files changed, 0 insertions(+), 16 deletions(-)
> > 
> > diff --git a/arch/x86/kvm/i8259.c b/arch/x86/kvm/i8259.c
> > index 01f1516..eb2b8b7 100644
> > --- a/arch/x86/kvm/i8259.c
> > +++ b/arch/x86/kvm/i8259.c
> > @@ -225,22 +225,6 @@ int kvm_pic_read_irq(struct kvm *kvm)
> >  
> >  void kvm_pic_reset(struct kvm_kpic_state *s)
> >  {
> > -	int irq, irqbase, n;
> > -	struct kvm *kvm = s->pics_state->irq_request_opaque;
> > -	struct kvm_vcpu *vcpu0 = kvm->bsp_vcpu;
> > -
> > -	if (s == &s->pics_state->pics[0])
> > -		irqbase = 0;
> > -	else
> > -		irqbase = 8;
> > -
> > -	for (irq = 0; irq < PIC_NUM_PINS/2; irq++) {
> > -		if (vcpu0 && kvm_apic_accept_pic_intr(vcpu0))
> > -			if (s->irr & (1 << irq) || s->isr & (1 << irq)) {
> > -				n = irq + irqbase;
> > -				kvm_notify_acked_irq(kvm, SELECT_PIC(n), n);
> > -			}
> > -	}
> >  	s->last_irr = 0;
> >  	s->irr = 0;
> >  	s->imr = 0;
> > -- 
> > 1.6.3.3
> 
> This used to be necessary to clear pending state from i8254.c
> "irq_acked" logic. I think it'll break it.
This is just a hack then and it does not exists in ioapic so if
it is really needed ioapic+pit combination is broken. But the problem
should be solved inside i8254.c not somewhere else. Setting irq_acked to 1 in
pit_load_count() seems like a right thing to do. Something like
the patch below. Ideally pending should be scaled instead of reset.
Also may be the problem exists because PIC doesn't call mask notifiers?


diff --git a/arch/x86/kvm/i8254.c b/arch/x86/kvm/i8254.c
index b857ca3..aa7f68e 100644
--- a/arch/x86/kvm/i8254.c
+++ b/arch/x86/kvm/i8254.c
@@ -325,6 +325,9 @@ static void pit_load_count(struct kvm *kvm, int channel, u32 val)
 		return;
 	}
 
+	atomic_set(&pt->pending, 0);	
+	ps->irq_ack = 1;
+
 	/* Two types of timer
 	 * mode 1 is one shot, mode 2 is period, otherwise del timer */
 	switch (ps->channels[0].mode) {

--
			Gleb.

  reply	other threads:[~2009-08-13  9:39 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2009-08-12 12:17 [PATCH v3 0/8] make interrupt injection lockless (almost) Gleb Natapov
2009-08-12 12:17 ` [PATCH v3 1/8] Do not call ack notifiers on PIC reset Gleb Natapov
2009-08-13  9:11   ` Marcelo Tosatti
2009-08-13  9:39     ` Gleb Natapov [this message]
2009-08-12 12:17 ` [PATCH v3 2/8] Change irq routing table to use gsi indexed array Gleb Natapov
2009-08-12 12:17 ` [PATCH v3 3/8] Maintain back mapping from irqchip/pin to gsi Gleb Natapov
2009-08-17  9:43   ` Avi Kivity
2009-08-12 12:17 ` [PATCH v3 4/8] Move irq routing data structure to rcu locking Gleb Natapov
2009-08-12 12:17 ` [PATCH v3 5/8] Move irq ack notifier list to arch independent code Gleb Natapov
2009-08-12 12:17 ` [PATCH v3 6/8] Convert irq notifiers lists to RCU locking Gleb Natapov
2009-08-12 12:17 ` [PATCH v3 7/8] Move IO APIC to its own lock Gleb Natapov
2009-08-13  9:13   ` Marcelo Tosatti
2009-08-13  9:48     ` Gleb Natapov
2009-08-13  9:49       ` Avi Kivity
2009-08-13 10:09         ` Gleb Natapov
2009-08-13 10:44           ` Avi Kivity
2009-08-13 15:15             ` Paul E. McKenney
2009-08-12 12:17 ` [PATCH v3 8/8] Drop kvm->irq_lock lock from irq injection path Gleb Natapov
2009-08-17  9:46 ` [PATCH v3 0/8] make interrupt injection lockless (almost) 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=20090813093934.GA23736@redhat.com \
    --to=gleb@redhat.com \
    --cc=avi@redhat.com \
    --cc=kvm@vger.kernel.org \
    --cc=mtosatti@redhat.com \
    /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;
as well as URLs for NNTP newsgroup(s).