Re: [PATCH 1/1] Revert "genirq: Remove the second parameter from handle_irq_event_percpu()"

[Thomas Gleixner <tglx@linutronix.de>]

On Wed, 13 Jan 2016, zyjzyj2000@gmail.com wrote:

> After this commit 71f64340fc0e ("genirq: Remove the second parameter
> from handle_irq_event_percpu()") is applied, the variable action is
> not protected by raw_spin_lock. The following calltrace will pop up.

Thanks, for the report. I missed that detail when merging the patch!

Just for correctness sake: You miss to explain why this can happen.

It's not about the variable action, it's about desc->action not being
protected anymore. So the reason why this oopses is that the action is being
removed concurrently.

CPU 0			CPU 1

free_irq()		lock(desc)
lock(desc)		handle_edge_irq()
			  handle_irq_event(desc)
			    unlock(desc)
desc->action = NULL	    handle_irq_event_percpu(desc)
	       		      action = desc->action

While the original code did:

free_irq()		lock(desc)
lock(desc)		handle_edge_irq()
			  handle_irq_event()
	       		    action = desc->action
			    unlock(desc)
desc->action = NULL	    handle_irq_event_percpu(desc, action)
	       		    
So now the question is whether we revert that patch or simply change
handle_irq_event_percpu() to deal with that. Patch below.

That preserves us the code size reduction of commit 71f64340fc0e. This is safe
because we either see a valid desc->action or NULL. If the action is about to
be removed it is still valid as free_irq() is blocked on synchronize_irq().

free_irq()		lock(desc)
lock(desc)		handle_edge_irq()
			  handle_irq_event(desc)
			    set(INPROGRESS)
			    unlock(desc)
			      handle_irq_event_percpu(desc)
	       		        action = desc->action
desc->action = NULL
sychronize_irq()
  while(INPROGRESS);	   lock(desc)
			   clr(INPROGRESS)
free(action)

That's basically the same mechanism as we have for shared
interrupts. action->next can become NULL while handle_irq_event_percpu()
runs. Either it sees the action or NULL. It does not matter, because action
itself cannot go away.

Thanks,

	tglx

8<-------------

--- a/kernel/irq/handle.c
+++ b/kernel/irq/handle.c
@@ -136,9 +136,15 @@ irqreturn_t handle_irq_event_percpu(stru
 {
 	irqreturn_t retval = IRQ_NONE;
 	unsigned int flags = 0, irq = desc->irq_data.irq;
-	struct irqaction *action = desc->action;
+	struct irqaction *action;
 
-	do {
+	/*
+	 * READ_ONCE is not required here. The compiler cannot reload action
+	 * because it'll be action->next for the second iteration of the loop.
+	 */
+	action = desc->action;
+
+	while (action) {
 		irqreturn_t res;
 
 		trace_irq_handler_entry(irq, action);
@@ -173,7 +179,7 @@ irqreturn_t handle_irq_event_percpu(stru
 
 		retval |= res;
 		action = action->next;
-	} while (action);
+	}
 
 	add_interrupt_randomness(irq, flags);