Re: [PATCH 12/14] xen: events: remove use of nr_irqs as upper bound on number of pirqs

[Ian Campbell <Ian.Campbell@eu.citrix.com>]

On Thu, 2011-03-10 at 05:33 +0000, Konrad Rzeszutek Wilk wrote:
> >  int xen_irq_from_pirq(unsigned pirq)
> >  {
> > -	return pirq_to_irq[pirq];
> > +	int irq;
> > +
> > +	struct irq_info *info;
> > +
> > +	spin_lock(&irq_mapping_update_lock);
> > +
> > +	list_for_each_entry(info, &xen_irq_list_head, list) {
> > +		if (info == NULL || info->type != IRQT_PIRQ)
> > +			continue;
> > +		irq = info->irq;
> > +		if (info->u.pirq.pirq == pirq)
> > +			goto out;
> > +	}
> > +	irq = -1;
> > +out:
> > +	spin_lock(&irq_mapping_update_lock);
> > +
> > +	return -1;
> 
> Shouldn't this be:
> 
>   return irq

Yes. The impact of not doing so is that xen_hvm_setup_msi_irqs would
allocate a fresh PIRQ each time an MSI was reused, instead of resuing
the old one -- i.e. it's a leak. I retested PVHVM PCI hotplug with the
following patch.

> How come you are using the spin_lock here, but not
> in other places when iterating over the xen_irq_list_head?

Those other places already hold the lock in their caller (or are known
to be single threaded -- e.g. resume). Callers of xen_irq_from_pirq do
not hold the lock.

Ian.

8<------------------------

>From 9b1686b874c4893a3b014e400185940dcba43676 Mon Sep 17 00:00:00 2001
From: Ian Campbell <ian.campbell@citrix.com>
Date: Thu, 10 Mar 2011 08:54:06 +0000
Subject: [PATCH] xen: events: fix return value from xen_irq_from_pirq

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
---
 drivers/xen/events.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/xen/events.c b/drivers/xen/events.c
index 77c2b43..dc5b575 100644
--- a/drivers/xen/events.c
+++ b/drivers/xen/events.c
@@ -765,7 +765,7 @@ int xen_irq_from_pirq(unsigned pirq)
 out:
 	spin_lock(&irq_mapping_update_lock);
 
-	return -1;
+	return irq;
 }
 
 
-- 
1.5.6.5