[PATCH] x86/x86_64: deferred handling of writes to /proc/irqxx/smp_af…

…finity

When handling writes to /proc/irq, current code is re-programming rte
entries directly. This is not recommended and could potentially cause
chipset's to lockup, or cause missing interrupts.

CONFIG_IRQ_BALANCE does this correctly, where it re-programs only when the
interrupt is pending. The same needs to be done for /proc/irq handling as well.
Otherwise user space irq balancers are really not doing the right thing.

- Changed pending_irq_balance_cpumask to pending_irq_migrate_cpumask for
  lack of a generic name.
- added move_irq out of IRQ_BALANCE, and added this same to X86_64
- Added new proc handler for write, so we can do deferred write at irq
  handling time.
- Display of /proc/irq/XX/smp_affinity used to display CPU_MASKALL, instead
  it now shows only active cpu masks, or exactly what was set.
- Provided a common move_irq implementation, instead of duplicating
  when using generic irq framework.

Tested on i386/x86_64 and ia64 with CONFIG_PCI_MSI turned on and off.
Tested UP builds as well.

MSI testing: tbd: I have cards, need to look for a x-over cable, although I
did test an earlier version of this patch.  Will test in a couple days.

Signed-off-by: Ashok Raj <ashok.raj@intel.com>
Acked-by: Zwane Mwaikambo <zwane@holomorphy.com>
Grudgingly-acked-by: Andi Kleen <ak@muc.de>
Signed-off-by: Coywolf Qi Hunt <coywolf@lovecn.org>
Signed-off-by: Ashok Raj <ashok.raj@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

arch/i386/Kconfig

@@ -1318,6 +1318,11 @@ config GENERIC_IRQ_PROBE
 	bool
 	default y
 
+config GENERIC_PENDING_IRQ
+	bool
+	depends on GENERIC_HARDIRQS && SMP
+	default y
+
 config X86_SMP
 	bool
 	depends on SMP && !X86_VOYAGER

arch/i386/kernel/io_apic.c

@@ -33,6 +33,7 @@
 #include <linux/acpi.h>
 #include <linux/module.h>
 #include <linux/sysdev.h>
+
 #include <asm/io.h>
 #include <asm/smp.h>
 #include <asm/desc.h>
@@ -222,13 +223,21 @@ static void clear_IO_APIC (void)
 			clear_IO_APIC_pin(apic, pin);
 }
 
+#ifdef CONFIG_SMP
 static void set_ioapic_affinity_irq(unsigned int irq, cpumask_t cpumask)
 {
 	unsigned long flags;
 	int pin;
 	struct irq_pin_list *entry = irq_2_pin + irq;
 	unsigned int apicid_value;
+	cpumask_t tmp;
 
+	cpus_and(tmp, cpumask, cpu_online_map);
+	if (cpus_empty(tmp))
+		tmp = TARGET_CPUS;
+
+	cpus_and(cpumask, tmp, CPU_MASK_ALL);
+
 	apicid_value = cpu_mask_to_apicid(cpumask);
 	/* Prepare to do the io_apic_write */
 	apicid_value = apicid_value << 24;
@@ -242,6 +251,7 @@ static void set_ioapic_affinity_irq(unsigned int irq, cpumask_t cpumask)
 			break;
 		entry = irq_2_pin + entry->next;
 	}
+	set_irq_info(irq, cpumask);
 	spin_unlock_irqrestore(&ioapic_lock, flags);
 }
 
@@ -259,7 +269,6 @@ static void set_ioapic_affinity_irq(unsigned int irq, cpumask_t cpumask)
 #  define Dprintk(x...) 
 # endif
 
-cpumask_t __cacheline_aligned pending_irq_balance_cpumask[NR_IRQS];
 
 #define IRQBALANCE_CHECK_ARCH -999
 static int irqbalance_disabled = IRQBALANCE_CHECK_ARCH;
@@ -328,12 +337,7 @@ static inline void balance_irq(int cpu, int irq)
 	cpus_and(allowed_mask, cpu_online_map, irq_affinity[irq]);
 	new_cpu = move(cpu, allowed_mask, now, 1);
 	if (cpu != new_cpu) {
-		irq_desc_t *desc = irq_desc + irq;
-		unsigned long flags;
-
-		spin_lock_irqsave(&desc->lock, flags);
-		pending_irq_balance_cpumask[irq] = cpumask_of_cpu(new_cpu);
-		spin_unlock_irqrestore(&desc->lock, flags);
+		set_pending_irq(irq, cpumask_of_cpu(new_cpu));
 	}
 }
 
@@ -528,16 +532,12 @@ static void do_irq_balance(void)
 	cpus_and(tmp, target_cpu_mask, allowed_mask);
 
 	if (!cpus_empty(tmp)) {
-		irq_desc_t *desc = irq_desc + selected_irq;
-		unsigned long flags;
 
 		Dprintk("irq = %d moved to cpu = %d\n",
 				selected_irq, min_loaded);
 		/* mark for change destination */
-		spin_lock_irqsave(&desc->lock, flags);
-		pending_irq_balance_cpumask[selected_irq] =
-					cpumask_of_cpu(min_loaded);
-		spin_unlock_irqrestore(&desc->lock, flags);
+		set_pending_irq(selected_irq, cpumask_of_cpu(min_loaded));
+
 		/* Since we made a change, come back sooner to 
 		 * check for more variation.
 		 */
@@ -568,7 +568,8 @@ static int balanced_irq(void *unused)
 
 	/* push everything to CPU 0 to give us a starting point.  */
 	for (i = 0 ; i < NR_IRQS ; i++) {
-		pending_irq_balance_cpumask[i] = cpumask_of_cpu(0);
+		pending_irq_cpumask[i] = cpumask_of_cpu(0);
+		set_pending_irq(i, cpumask_of_cpu(0));
 	}
 
 	for ( ; ; ) {
@@ -647,20 +648,9 @@ int __init irqbalance_disable(char *str)
 
 __setup("noirqbalance", irqbalance_disable);
 
-static inline void move_irq(int irq)
-{
-	/* note - we hold the desc->lock */
-	if (unlikely(!cpus_empty(pending_irq_balance_cpumask[irq]))) {
-		set_ioapic_affinity_irq(irq, pending_irq_balance_cpumask[irq]);
-		cpus_clear(pending_irq_balance_cpumask[irq]);
-	}
-}
-
 late_initcall(balanced_irq_init);
-
-#else /* !CONFIG_IRQBALANCE */
-static inline void move_irq(int irq) { }
 #endif /* CONFIG_IRQBALANCE */
+#endif /* CONFIG_SMP */
 
 #ifndef CONFIG_SMP
 void fastcall send_IPI_self(int vector)
@@ -820,6 +810,7 @@ EXPORT_SYMBOL(IO_APIC_get_PCI_irq_vector);
  * we need to reprogram the ioredtbls to cater for the cpus which have come online
  * so mask in all cases should simply be TARGET_CPUS
  */
+#ifdef CONFIG_SMP
 void __init setup_ioapic_dest(void)
 {
 	int pin, ioapic, irq, irq_entry;
@@ -838,6 +829,7 @@ void __init setup_ioapic_dest(void)
 
 	}
 }
+#endif
 
 /*
  * EISA Edge/Level control register, ELCR
@@ -1249,6 +1241,7 @@ static void __init setup_IO_APIC_irqs(void)
 		spin_lock_irqsave(&ioapic_lock, flags);
 		io_apic_write(apic, 0x11+2*pin, *(((int *)&entry)+1));
 		io_apic_write(apic, 0x10+2*pin, *(((int *)&entry)+0));
+		set_native_irq_info(irq, TARGET_CPUS);
 		spin_unlock_irqrestore(&ioapic_lock, flags);
 	}
 	}
@@ -1944,6 +1937,7 @@ static void ack_edge_ioapic_vector(unsigned int vector)
 {
 	int irq = vector_to_irq(vector);
 
+	move_irq(vector);
 	ack_edge_ioapic_irq(irq);
 }
 
@@ -1958,6 +1952,7 @@ static void end_level_ioapic_vector (unsigned int vector)
 {
 	int irq = vector_to_irq(vector);
 
+	move_irq(vector);
 	end_level_ioapic_irq(irq);
 }
 
@@ -1975,14 +1970,17 @@ static void unmask_IO_APIC_vector (unsigned int vector)
 	unmask_IO_APIC_irq(irq);
 }
 
+#ifdef CONFIG_SMP
 static void set_ioapic_affinity_vector (unsigned int vector,
 					cpumask_t cpu_mask)
 {
 	int irq = vector_to_irq(vector);
 
+	set_native_irq_info(vector, cpu_mask);
 	set_ioapic_affinity_irq(irq, cpu_mask);
 }
 #endif
+#endif
 
 /*
  * Level and edge triggered IO-APIC interrupts need different handling,
@@ -2000,7 +1998,9 @@ static struct hw_interrupt_type ioapic_edge_type = {
 	.disable 	= disable_edge_ioapic,
 	.ack 		= ack_edge_ioapic,
 	.end 		= end_edge_ioapic,
+#ifdef CONFIG_SMP
 	.set_affinity 	= set_ioapic_affinity,
+#endif
 };
 
 static struct hw_interrupt_type ioapic_level_type = {
@@ -2011,7 +2011,9 @@ static struct hw_interrupt_type ioapic_level_type = {
 	.disable 	= disable_level_ioapic,
 	.ack 		= mask_and_ack_level_ioapic,
 	.end 		= end_level_ioapic,
+#ifdef CONFIG_SMP
 	.set_affinity 	= set_ioapic_affinity,
+#endif
 };
 
 static inline void init_IO_APIC_traps(void)
@@ -2569,6 +2571,7 @@ int io_apic_set_pci_routing (int ioapic, int pin, int irq, int edge_level, int a
 	spin_lock_irqsave(&ioapic_lock, flags);
 	io_apic_write(ioapic, 0x11+2*pin, *(((int *)&entry)+1));
 	io_apic_write(ioapic, 0x10+2*pin, *(((int *)&entry)+0));
+	set_native_irq_info(use_pci_vector() ? entry.vector : irq, TARGET_CPUS);
 	spin_unlock_irqrestore(&ioapic_lock, flags);
 
 	return 0;

arch/ia64/Kconfig

@@ -434,6 +434,11 @@ config GENERIC_IRQ_PROBE
 	bool
 	default y
 
+config GENERIC_PENDING_IRQ
+	bool
+	depends on GENERIC_HARDIRQS && SMP
+	default y
+
 source "arch/ia64/hp/sim/Kconfig"
 
 source "arch/ia64/oprofile/Kconfig"

arch/ia64/kernel/irq.c

@@ -91,23 +91,8 @@ int show_interrupts(struct seq_file *p, void *v)
 }
 
 #ifdef CONFIG_SMP
-/*
- * This is updated when the user sets irq affinity via /proc
- */
-static cpumask_t __cacheline_aligned pending_irq_cpumask[NR_IRQS];
-static unsigned long pending_irq_redir[BITS_TO_LONGS(NR_IRQS)];
-
 static char irq_redir [NR_IRQS]; // = { [0 ... NR_IRQS-1] = 1 };
 
-/*
- * Arch specific routine for deferred write to iosapic rte to reprogram
- * intr destination.
- */
-void proc_set_irq_affinity(unsigned int irq, cpumask_t mask_val)
-{
-	pending_irq_cpumask[irq] = mask_val;
-}
-
 void set_irq_affinity_info (unsigned int irq, int hwid, int redir)
 {
 	cpumask_t mask = CPU_MASK_NONE;
@@ -116,32 +101,10 @@ void set_irq_affinity_info (unsigned int irq, int hwid, int redir)
 
 	if (irq < NR_IRQS) {
 		irq_affinity[irq] = mask;
+		set_irq_info(irq, mask);
 		irq_redir[irq] = (char) (redir & 0xff);
 	}
 }
-
-
-void move_irq(int irq)
-{
-	/* note - we hold desc->lock */
-	cpumask_t tmp;
-	irq_desc_t *desc = irq_descp(irq);
-	int redir = test_bit(irq, pending_irq_redir);
-
-	if (unlikely(!desc->handler->set_affinity))
-		return;
-
-	if (!cpus_empty(pending_irq_cpumask[irq])) {
-		cpus_and(tmp, pending_irq_cpumask[irq], cpu_online_map);
-		if (unlikely(!cpus_empty(tmp))) {
-			desc->handler->set_affinity(irq | (redir ? IA64_IRQ_REDIRECTED : 0),
-						    pending_irq_cpumask[irq]);
-		}
-		cpus_clear(pending_irq_cpumask[irq]);
-	}
-}
-
-
 #endif /* CONFIG_SMP */
 
 #ifdef CONFIG_HOTPLUG_CPU

arch/x86_64/Kconfig

@@ -441,6 +441,11 @@ config ISA_DMA_API
 	bool
 	default y
 
+config GENERIC_PENDING_IRQ
+	bool
+	depends on GENERIC_HARDIRQS && SMP
+	default y
+
 menu "Power management options"
 
 source kernel/power/Kconfig