Re: [PATCH 2/2] xen: events: free irqs in error condition

[Roger Pau Monné]

On Mon, Feb 26, 2018 at 06:57:03PM +0000, Shah, Amit wrote:
> 
> On Mo, 2018-02-26 at 18:14 +0000, Roger Pau Monné wrote:
> > On Mon, Feb 26, 2018 at 05:36:35PM +0000, Amit Shah wrote:
> > > 
> > > In case of errors in irq setup for MSI, free up the allocated irqs.
> > > 
> > > Fixes: 4892c9b4ada9f9 ("xen: add support for MSI message groups")
> > > Reported-by: Hooman Mirhadi <mirhadih@xxxxxxxxxx>
> > > CC: <stable@xxxxxxxxxxxxxxx>
> > > CC: Roger Pau Monné <roger.pau@xxxxxxxxxx>
> > > CC: David Vrabel <david.vrabel@xxxxxxxxxx>
> > > CC: Boris Ostrovsky <boris.ostrovsky@xxxxxxxxxx>
> > > CC: Eduardo Valentin <eduval@xxxxxxxxxx>
> > > CC: Juergen Gross <jgross@xxxxxxxx>
> > > CC: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> > > CC: "K. Y. Srinivasan" <kys@xxxxxxxxxxxxx>
> > > CC: Liu Shuo <shuo.a.liu@xxxxxxxxx>
> > > CC: Anoob Soman <anoob.soman@xxxxxxxxxx>
> > > Signed-off-by: Amit Shah <aams@xxxxxxxxxx>
> > > ---
> > >  drivers/xen/events/events_base.c | 1 +
> > >  1 file changed, 1 insertion(+)
> > > 
> > > diff --git a/drivers/xen/events/events_base.c
> > > b/drivers/xen/events/events_base.c
> > > index b6b8b29..96aa575 100644
> > > --- a/drivers/xen/events/events_base.c
> > > +++ b/drivers/xen/events/events_base.c
> > > @@ -758,6 +758,7 @@ int xen_bind_pirq_msi_to_irq(struct pci_dev
> > > *dev, struct msi_desc *msidesc,
> > >  error_irq:
> > >  	for (; i >= 0; i--)
> > >  		__unbind_from_irq(irq + i);
> > > +	xen_free_irq(irq);
> > Hm, xen_free_irq calls irq_free_desc, which is irq_free_descs(irq,
> > 1),
> 
> Er...  right.
> 
> > I think you will have to introduce a new free function:
> > 
> > xen_free_irqs(unsigned irq, unsigned int nr)
> > 
> > That calls irq_free_descs(irq, nr)
> 
> Actually, xen_free_irq() is already done in __unbind_from_irq(), so
> this patch is actually wrong and not needed.

You still need to free unbound IRQs, AFAICT you could fix the issue
with a single patch, like:

while (nvec--) {
	if (nvec >= i)
		xen_free_irq(irq + i);
	else
		__unbind_from_irq(irq + i);
}

Roger.