Re: [Fedora-xen] [Xen-devel] [PATCH] Xen 4.1.2 HVM guest realtek nic problems (eth0 8139cp transmit queue timed out)

[Pasi Kärkkäinen]

On Mon, Nov 07, 2011 at 12:02:03PM +0000, Stefano Stabellini wrote:
>  On Thu, 3 Nov 2011, Keir Fraser wrote:
>  > On 03/11/2011 18:07, "Pasi KÃ?rkkÃ?inen" <pasik(a)iki.fi&gt; wrote:
>  > 
>  > > On Tue, Nov 01, 2011 at 10:56:06PM +0200, Pasi KÃ?rkkÃ?inen wrote:
>  > >> On Mon, Oct 31, 2011 at 09:29:24PM +0200, Pasi KÃ?rkkÃ?inen wrote:
>  > >>> On Mon, Oct 31, 2011 at 12:24:14PM -0700, Boris Derzhavets wrote:
>  > >>>>    Seems to related
>  > >>>> 
>  > >>>>    https://bugs.launchpad.net/ubuntu/+source/xen/+bug/854829
>  > >>>> 
>  > >>> 
>  > >>> Thanks, that seems to be the same bug.
>  > >>> 
>  > >>> Is the bugfix patch from xen-unstable going to backported to
>  > >>> xen-4.1-testing.hg ?
>  > >>> (4.1 backported patch available on ubuntu's launchpad above..)
>  > >>> 
>  > >> 
>  > >> So the Ubuntu backport from xen-unstable to Xen 4.1.1 is here:
>  > >> https://launchpadlibrarian.net/81948978/xen-pirq-resubmit-irq.patch
>  > >> 
>  > >> It seems to be shipping in Ubuntu 11.10 xen 4.1.1-2ubuntu4.1 packages.
>  > >> 
>  > >> Does that patch look suitable to be applied to xen-4.1-testing.hg ?
>  > >> This bug should be fixed for Xen 4.1.3.
>  > > 
>  > > Any comments? 
>  > 
>  > This looks like a backport of Stefano's xen-unstable c/s 24007. I would like
>  > him to submit/ack the backport, as it is not a trivial backport of the
>  > xen-unstable patch.
>  
>  I would rather use the following backport. Compared to the other one it
>  returns EINVAL in PHYSDEVOP_irq_status_query when the arguments are not
>  correct.
>   
Thanks! Mayoung added this patch in xen-4.1.2-1.1.fc14.src.rpm,
and binary rpms are built for Fedora 16 here:
http://koji.fedoraproject.org/koji/taskinfo?taskID=3495905

So people should test this patch (or the rpms above) and confirm 
that it fixes the emulated realtek/ne2k issues.

Thanks,

-- Pasi


>  ---
>  
>  diff -r 8c2d76193eaf xen/arch/x86/physdev.c
>  --- a/xen/arch/x86/physdev.c	Wed Nov 02 15:02:18 2011 +0000
>  +++ b/xen/arch/x86/physdev.c	Mon Nov 07 11:58:28 2011 +0000
>  @@ -261,6 +261,7 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_H
>           ret = -EINVAL;
>           if ( eoi.irq >= v->domain->nr_pirqs )
>               break;
>  +        spin_lock(&v->domain->event_lock);
>           if ( v->domain->arch.pirq_eoi_map )
>               evtchn_unmask(v->domain->pirq_to_evtchn[eoi.irq]);
>           if ( !is_hvm_domain(v->domain) ||
>  @@ -268,6 +269,19 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_H
>               ret = pirq_guest_eoi(v->domain, eoi.irq);
>           else
>               ret = 0;
>  +        if ( is_hvm_domain(v->domain) &&
>  +                domain_pirq_to_emuirq(v->domain, eoi.irq) > 0 )
>  +        {
>  +            struct hvm_irq *hvm_irq = &v->domain->arch.hvm_domain.irq;
>  +            int gsi = domain_pirq_to_emuirq(v->domain, eoi.irq);
>  +
>  +            /* if this is a level irq and count > 0, send another
>  +             * notification */ 
>  +            if ( gsi >= NR_ISAIRQS /* ISA irqs are edge triggered */
>  +                    && hvm_irq->gsi_assert_count[gsi] )
>  +                send_guest_pirq(v->domain, eoi.irq);
>  +        }
>  +        spin_unlock(&v->domain->event_lock);
>           break;
>       }
>   
>  @@ -323,9 +337,10 @@ ret_t do_physdev_op(int cmd, XEN_GUEST_H
>               break;
>           irq_status_query.flags = 0;
>           if ( is_hvm_domain(v->domain) &&
>  -             domain_pirq_to_irq(v->domain, irq) <= 0 )
>  +                domain_pirq_to_irq(v->domain, irq) <= 0 &&
>  +                domain_pirq_to_emuirq(v->domain, irq) == IRQ_UNBOUND )
>           {
>  -            ret = copy_to_guest(arg, &irq_status_query, 1) ? -EFAULT : 0;
>  +            ret = -EINVAL;
>               break;
>           }
>