| Message ID | 1457686822-39117-1-git-send-email-anna-maria@linutronix.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 11/03/16 09:00, Anna-Maria Gleixner wrote: > The evtchn_fifo_cpu_notification() hotplug callback lacks handling of > the CPU_UP_CANCELED case. That means, if CPU_UP_PREPARE fails, the > handle of the fifo events is not dropped. > > Add handling for CPU_UP_CANCELED transition to drop the fifo events > handle. __evtchn_fifo_handle_events() does not releases resources, it processes any pending events for this CPU. This patch would only be necessary if a CPU in CPU_UP_CANCELED state may have had unmasked interrupts. If so you would need: case CPU_UP_CANCELED: if (per_cpu(cpu_control_block, cpu)) __evtchn_fifo_handle_events(cpu, true); break; To handle the case where the control block allocation or initialization failed during CPU_UP_PREPARE. David > > Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> > Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com> > Cc: David Vrabel <david.vrabel@citrix.com> > Cc: xen-devel@lists.xenproject.org > Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> > --- > drivers/xen/events/events_fifo.c | 1 + > 1 file changed, 1 insertion(+) > > --- a/drivers/xen/events/events_fifo.c > +++ b/drivers/xen/events/events_fifo.c > @@ -432,6 +432,7 @@ static int evtchn_fifo_cpu_notification( > ret = evtchn_fifo_alloc_control_block(cpu); > break; > case CPU_DEAD: > + case CPU_UP_CANCELED: > __evtchn_fifo_handle_events(cpu, true); > break; > default: > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xen.org > http://lists.xen.org/xen-devel >
On Fri, 11 Mar 2016, David Vrabel wrote: > On 11/03/16 09:00, Anna-Maria Gleixner wrote: > > The evtchn_fifo_cpu_notification() hotplug callback lacks handling of > > the CPU_UP_CANCELED case. That means, if CPU_UP_PREPARE fails, the > > handle of the fifo events is not dropped. > > > > Add handling for CPU_UP_CANCELED transition to drop the fifo events > > handle. > > __evtchn_fifo_handle_events() does not releases resources, it processes > any pending events for this CPU. > > This patch would only be necessary if a CPU in CPU_UP_CANCELED state may > have had unmasked interrupts. If so you would need: Thanks for the explanation > case CPU_UP_CANCELED: > if (per_cpu(cpu_control_block, cpu)) > __evtchn_fifo_handle_events(cpu, true); > break; > > To handle the case where the control block allocation or initialization > failed during CPU_UP_PREPARE. This would not be necessary, because if the CPU_UP_PREPARE of this notifier fails, the CPU_UP_CANCELLED of this notifier will not be invoked. Anna-Maria
--- a/drivers/xen/events/events_fifo.c +++ b/drivers/xen/events/events_fifo.c @@ -432,6 +432,7 @@ static int evtchn_fifo_cpu_notification( ret = evtchn_fifo_alloc_control_block(cpu); break; case CPU_DEAD: + case CPU_UP_CANCELED: __evtchn_fifo_handle_events(cpu, true); break; default:
The evtchn_fifo_cpu_notification() hotplug callback lacks handling of the CPU_UP_CANCELED case. That means, if CPU_UP_PREPARE fails, the handle of the fifo events is not dropped. Add handling for CPU_UP_CANCELED transition to drop the fifo events handle. Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com> Cc: David Vrabel <david.vrabel@citrix.com> Cc: xen-devel@lists.xenproject.org Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de> --- drivers/xen/events/events_fifo.c | 1 + 1 file changed, 1 insertion(+)