genirq: Avoid deadlock in spurious handling

commit 52553ddf(genirq: fix regression in irqfixup, irqpoll)
introduced a potential deadlock by calling the action handler with the
irq descriptor lock held.

Remove the call and let the handling code run even for an interrupt
where only a single action is registered. That matches the goal of
the above commit and avoids the deadlock.

Document the confusing action = desc->action reload in the handling
loop while at it.

Reported-and-tested-by: "Wang, Warner" <warner.wang@hp.com>
Tested-by: Edward Donovan <edward.donovan@numble.net>
Cc: "Wang, Song-Bo (Stoney)" <song-bo.wang@hp.com>
Cc: stable@vger.kernel.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

diff --git a/kernel/irq/spurious.c b/kernel/irq/spurious.c
index 611cd60..7b5f012 100644 (file)
--- a/kernel/irq/spurious.c
+++ b/kernel/irq/spurious.c
@@ -80,13 +80,11 @@ static int try_one_irq(int irq, struct irq_desc *desc, bool force)
 
        /*
         * All handlers must agree on IRQF_SHARED, so we test just the
-        * first. Check for action->next as well.
+        * first.
         */
        action = desc->action;
        if (!action || !(action->flags & IRQF_SHARED) ||
-           (action->flags & __IRQF_TIMER) ||
-           (action->handler(irq, action->dev_id) == IRQ_HANDLED) ||
-           !action->next)
+           (action->flags & __IRQF_TIMER))
                goto out;
 
        /* Already running on another processor */
@@ -104,6 +102,7 @@ static int try_one_irq(int irq, struct irq_desc *desc, bool force)
        do {
                if (handle_irq_event(desc) == IRQ_HANDLED)
                        ret = IRQ_HANDLED;
+               /* Make sure that there is still a valid action */
                action = desc->action;
        } while ((desc->istate & IRQS_PENDING) && action);
        desc->istate &= ~IRQS_POLL_INPROGRESS;