Re: Re: [PATCH] xen: fix off-by-one error in find_unbound_irq

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

On Fri, 2010-02-26 at 20:13 +0000, Jeremy Fitzhardinge wrote: 
> On 02/26/2010 03:22 AM, Ian Campbell wrote:
> > BTW, this is against xen/master, the original patch isn't in xen/next.
> >    
> 
> Looks like it came from the pcifront branch which isn't in xen/next yet.

Makes sense.

I don't think my fix is right though, exiting that loop with irq ==
nr_irqs - 1 can be valid if the test in first iteration succeeds and we
break out

The error case is when start == nr_irqs which means we do no iterations
of the loop at all but still exit with irq == nr_irqs - 1.

Ian.

> 
>      J
> 
> > On Fri, 2010-02-26 at 10:59 +0000, Ian Campbell wrote:
> >    
> >> e459de95 "Find an unbound irq number in reverse order (high to low)" introduced
> >> an off by one error which would cause repeated allocations of the nr_irq'th IRQ
> >> if there are no spare interrupts (i.e. get_nr_hw_irqs() == nr_irqs).
> >>
> >> Signed-off-by: Ian Campbell<ian.campbell@citrix.com>
> >> Cc: Konrad Rzeszutek Wilk<konrad.wilk@oracle.com>
> >> Cc: Jeremy Fitzhardinge<jeremy.fitzhardinge@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 99f2b2a..5c64e1d 100644
> >> --- a/drivers/xen/events.c
> >> +++ b/drivers/xen/events.c
> >> @@ -377,7 +377,7 @@ static int find_unbound_irq(void)
> >>   		if (irq_info[irq].type == IRQT_UNBOUND)
> >>   			break;
> >>
> >> -	if (irq == start || irq == nr_irqs)
> >> +	if (irq == start || irq == nr_irqs - 1)
> >>   		panic("No available IRQ to bind to: increase nr_irqs!\n");
> >>
> >>   	desc = irq_to_desc_alloc_node(irq, 0);
> >>      
> >
> >
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.xensource.com
> > http://lists.xensource.com/xen-devel
> >
> >    
>