On Fri, 14 Sep 2012, Marc Zyngier wrote:
On 14/09/12 12:13, Stefano Stabellini wrote:
Reset the IRQ_NOAUTOEN and IRQ_NOREQUEST flags that are enabled by default on ARM. If IRQ_NOAUTOEN is set, __setup_irq doesn't call irq_startup, that is responsible for calling irq_unmask at startup time. As a result event channels remain masked.
Signed-off-by: Stefano Stabellini stefano.stabellini@eu.citrix.com Acked-by: Konrad Rzeszutek Wilk konrad.wilk@oracle.com
drivers/xen/events.c | 1 + 1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/drivers/xen/events.c b/drivers/xen/events.c index 5ecb596..8ffb7b7 100644 --- a/drivers/xen/events.c +++ b/drivers/xen/events.c @@ -836,6 +836,7 @@ int bind_evtchn_to_irq(unsigned int evtchn) struct irq_info *info = info_for_irq(irq); WARN_ON(info == NULL || info->type != IRQT_EVTCHN); }
- irq_clear_status_flags(irq, IRQ_NOREQUEST|IRQ_NOAUTOEN);
This one just sent a shiver down my spine. Are you doing this for a PPI?
Not really: even though there is just one source of event notifications (that is a PPI), we have many event channels. When a domain receives a notification (via the PPI), it checks on a bitmask to which event channel it corresponds. From the Linux point of view every event channel is a Linux irq belonging to the xen_dynamic_chip (see drivers/xen/events.c:xen_dynamic_chip).
So here I am not doing this for the one PPI, but I am doing this for every Linux irq (of chip xen_dynamic_chip) that represents an event channel.