[PATCH] irqpoll

Anyone reporting a stuck IRQ should try these options.  Its effectiveness
varies we've found in the Fedora case.  Quite a few systems with misdescribed
IRQ routing just work when you use irqpoll.  It also fixes up the VIA systems
although thats now fixed with the VIA quirk (which we could just make default
as its what Redmond OS does but Linus didn't like it historically).

A small number of systems have jammed IRQ sources or misdescribes that cause
an IRQ that we have no handler registered anywhere for.  In those cases it
doesn't help.

Signed-off-by: Alan Cox <number6@the-village.bc.nu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Documentation/kernel-parameters.txt

@@ -437,6 +437,10 @@ running once the system is up.
 			Format: {"of[f]" | "sk[ipmbr]"}
 			See comment in arch/i386/boot/edd.S
 
+	edd		[EDD]
+			Format: {"of[f]" | "sk[ipmbr]"}
+			See comment in arch/i386/boot/edd.S
+
 	eicon=		[HW,ISDN] 
 			Format: <id>,<membase>,<irq>
 
@@ -622,6 +626,17 @@ running once the system is up.
 	ips=		[HW,SCSI] Adaptec / IBM ServeRAID controller
 			See header of drivers/scsi/ips.c.
 
+	irqfixup	[HW]
+			When an interrupt is not handled search all handlers
+			for it. Intended to get systems with badly broken
+			firmware running.
+
+	irqpoll		[HW]
+			When an interrupt is not handled search all handlers
+			for it. Also check all handlers each timer
+			interrupt. Intended to get systems with badly broken
+			firmware running.
+
 	isapnp=		[ISAPNP]
 			Format: <RDP>, <reset>, <pci_scan>, <verbosity>
 

arch/ppc64/kernel/irq.c

@@ -245,7 +245,7 @@ void ppc_irq_dispatch_handler(struct pt_regs *regs, int irq)
 
 		spin_lock(&desc->lock);
 		if (!noirqdebug)
-			note_interrupt(irq, desc, action_ret);
+			note_interrupt(irq, desc, action_ret, regs);
 		if (likely(!(desc->status & IRQ_PENDING)))
 			break;
 		desc->status &= ~IRQ_PENDING;

include/linux/irq.h

@@ -85,9 +85,10 @@ extern int no_irq_affinity;
 extern int noirqdebug_setup(char *str);
 
 extern fastcall int handle_IRQ_event(unsigned int irq, struct pt_regs *regs,
-				       struct irqaction *action);
+					struct irqaction *action);
 extern fastcall unsigned int __do_IRQ(unsigned int irq, struct pt_regs *regs);
-extern void note_interrupt(unsigned int irq, irq_desc_t *desc, int action_ret);
+extern void note_interrupt(unsigned int irq, irq_desc_t *desc,
+					int action_ret, struct pt_regs *regs);
 extern int can_request_irq(unsigned int irq, unsigned long irqflags);
 
 extern void init_irq_proc(void);

kernel/irq/handle.c

@@ -172,7 +172,7 @@ fastcall unsigned int __do_IRQ(unsigned int irq, struct pt_regs *regs)
 
 		spin_lock(&desc->lock);
 		if (!noirqdebug)
-			note_interrupt(irq, desc, action_ret);
+			note_interrupt(irq, desc, action_ret, regs);
 		if (likely(!(desc->status & IRQ_PENDING)))
 			break;
 		desc->status &= ~IRQ_PENDING;

kernel/irq/spurious.c

@@ -11,6 +11,83 @@
 #include <linux/kallsyms.h>
 #include <linux/interrupt.h>
 
+static int irqfixup;
+
+/*
+ * Recovery handler for misrouted interrupts.
+ */
+
+static int misrouted_irq(int irq, struct pt_regs *regs)
+{
+	int i;
+	irq_desc_t *desc;
+	int ok = 0;
+	int work = 0;	/* Did we do work for a real IRQ */
+
+	for(i = 1; i < NR_IRQS; i++) {
+		struct irqaction *action;
+
+		if (i == irq)	/* Already tried */
+			continue;
+		desc = &irq_desc[i];
+		spin_lock(&desc->lock);
+		action = desc->action;
+		/* Already running on another processor */
+		if (desc->status & IRQ_INPROGRESS) {
+			/*
+			 * Already running: If it is shared get the other
+			 * CPU to go looking for our mystery interrupt too
+			 */
+			if (desc->action && (desc->action->flags & SA_SHIRQ))
+				desc->status |= IRQ_PENDING;
+			spin_unlock(&desc->lock);
+			continue;
+		}
+		/* Honour the normal IRQ locking */
+		desc->status |= IRQ_INPROGRESS;
+		spin_unlock(&desc->lock);
+		while (action) {
+			/* Only shared IRQ handlers are safe to call */
+			if (action->flags & SA_SHIRQ) {
+				if (action->handler(i, action->dev_id, regs) ==
+						IRQ_HANDLED)
+					ok = 1;
+			}
+			action = action->next;
+		}
+		local_irq_disable();
+		/* Now clean up the flags */
+		spin_lock(&desc->lock);
+		action = desc->action;
+
+		/*
+		 * While we were looking for a fixup someone queued a real
+		 * IRQ clashing with our walk
+		 */
+
+		while ((desc->status & IRQ_PENDING) && action) {
+			/*
+			 * Perform real IRQ processing for the IRQ we deferred
+			 */
+			work = 1;
+			spin_unlock(&desc->lock);
+			handle_IRQ_event(i, regs, action);
+			spin_lock(&desc->lock);
+			desc->status &= ~IRQ_PENDING;
+		}
+		desc->status &= ~IRQ_INPROGRESS;
+		/*
+		 * If we did actual work for the real IRQ line we must let the
+		 * IRQ controller clean up too
+		 */
+		if(work)
+			desc->handler->end(i);
+		spin_unlock(&desc->lock);
+	}
+	/* So the caller can adjust the irq error counts */
+	return ok;
+}
+
 /*
  * If 99,900 of the previous 100,000 interrupts have not been handled
  * then assume that the IRQ is stuck in some manner. Drop a diagnostic
@@ -31,7 +108,8 @@ __report_bad_irq(unsigned int irq, irq_desc_t *desc, irqreturn_t action_ret)
 		printk(KERN_ERR "irq event %d: bogus return value %x\n",
 				irq, action_ret);
 	} else {
-		printk(KERN_ERR "irq %d: nobody cared!\n", irq);
+		printk(KERN_ERR "irq %d: nobody cared (try booting with "
+				"the \"irqpoll\" option)\n", irq);
 	}
 	dump_stack();
 	printk(KERN_ERR "handlers:\n");
@@ -55,7 +133,8 @@ static void report_bad_irq(unsigned int irq, irq_desc_t *desc, irqreturn_t actio
 	}
 }
 
-void note_interrupt(unsigned int irq, irq_desc_t *desc, irqreturn_t action_ret)
+void note_interrupt(unsigned int irq, irq_desc_t *desc, irqreturn_t action_ret,
+			struct pt_regs *regs)
 {
 	if (action_ret != IRQ_HANDLED) {
 		desc->irqs_unhandled++;
@@ -63,6 +142,15 @@ void note_interrupt(unsigned int irq, irq_desc_t *desc, irqreturn_t action_ret)
 			report_bad_irq(irq, desc, action_ret);
 	}
 
+	if (unlikely(irqfixup)) {
+		/* Don't punish working computers */
+		if ((irqfixup == 2 && irq == 0) || action_ret == IRQ_NONE) {
+			int ok = misrouted_irq(irq, regs);
+			if (action_ret == IRQ_NONE)
+				desc->irqs_unhandled -= ok;
+		}
+	}
+
 	desc->irq_count++;
 	if (desc->irq_count < 100000)
 		return;
@@ -94,3 +182,24 @@ int __init noirqdebug_setup(char *str)
 
 __setup("noirqdebug", noirqdebug_setup);
 
+static int __init irqfixup_setup(char *str)
+{
+	irqfixup = 1;
+	printk(KERN_WARNING "Misrouted IRQ fixup support enabled.\n");
+	printk(KERN_WARNING "This may impact system performance.\n");
+	return 1;
+}
+
+__setup("irqfixup", irqfixup_setup);
+
+static int __init irqpoll_setup(char *str)
+{
+	irqfixup = 2;
+	printk(KERN_WARNING "Misrouted IRQ fixup and polling support "
+				"enabled\n");
+	printk(KERN_WARNING "This may significantly impact system "
+				"performance\n");
+	return 1;
+}
+
+__setup("irqpoll", irqpoll_setup);