Merge branch 'tracing-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
[safe/jmp/linux-2.6] / kernel / irq / handle.c
index a2ee682..a600184 100644 (file)
  */
 
 #include <linux/irq.h>
+#include <linux/slab.h>
 #include <linux/module.h>
 #include <linux/random.h>
 #include <linux/interrupt.h>
 #include <linux/kernel_stat.h>
 #include <linux/rculist.h>
 #include <linux/hash.h>
+#include <linux/bootmem.h>
+#include <trace/events/irq.h>
 
 #include "internals.h"
 
@@ -69,6 +72,7 @@ int nr_irqs = NR_IRQS;
 EXPORT_SYMBOL_GPL(nr_irqs);
 
 #ifdef CONFIG_SPARSE_IRQ
+
 static struct irq_desc irq_desc_init = {
        .irq        = -1,
        .status     = IRQ_DISABLED,
@@ -76,44 +80,50 @@ static struct irq_desc irq_desc_init = {
        .handle_irq = handle_bad_irq,
        .depth      = 1,
        .lock       = __SPIN_LOCK_UNLOCKED(irq_desc_init.lock),
-#ifdef CONFIG_SMP
-       .affinity   = CPU_MASK_ALL
-#endif
 };
 
-void init_kstat_irqs(struct irq_desc *desc, int cpu, int nr)
+void __ref init_kstat_irqs(struct irq_desc *desc, int node, int nr)
 {
-       unsigned long bytes;
-       char *ptr;
-       int node;
+       void *ptr;
 
-       /* Compute how many bytes we need per irq and allocate them */
-       bytes = nr * sizeof(unsigned int);
+       if (slab_is_available())
+               ptr = kzalloc_node(nr * sizeof(*desc->kstat_irqs),
+                                  GFP_ATOMIC, node);
+       else
+               ptr = alloc_bootmem_node(NODE_DATA(node),
+                               nr * sizeof(*desc->kstat_irqs));
 
-       node = cpu_to_node(cpu);
-       ptr = kzalloc_node(bytes, GFP_ATOMIC, node);
-       printk(KERN_DEBUG "  alloc kstat_irqs on cpu %d node %d\n", cpu, node);
-
-       if (ptr)
-               desc->kstat_irqs = (unsigned int *)ptr;
+       /*
+        * don't overwite if can not get new one
+        * init_copy_kstat_irqs() could still use old one
+        */
+       if (ptr) {
+               printk(KERN_DEBUG "  alloc kstat_irqs on node %d\n", node);
+               desc->kstat_irqs = ptr;
+       }
 }
 
-static void init_one_irq_desc(int irq, struct irq_desc *desc, int cpu)
+static void init_one_irq_desc(int irq, struct irq_desc *desc, int node)
 {
        memcpy(desc, &irq_desc_init, sizeof(struct irq_desc));
 
        spin_lock_init(&desc->lock);
        desc->irq = irq;
 #ifdef CONFIG_SMP
-       desc->cpu = cpu;
+       desc->node = node;
 #endif
        lockdep_set_class(&desc->lock, &irq_desc_lock_class);
-       init_kstat_irqs(desc, cpu, nr_cpu_ids);
+       init_kstat_irqs(desc, node, nr_cpu_ids);
        if (!desc->kstat_irqs) {
                printk(KERN_ERR "can not alloc kstat_irqs\n");
                BUG_ON(1);
        }
-       arch_init_chip_data(desc, cpu);
+       if (!alloc_desc_masks(desc, node, false)) {
+               printk(KERN_ERR "can not alloc irq_desc cpumasks\n");
+               BUG_ON(1);
+       }
+       init_desc_masks(desc);
+       arch_init_chip_data(desc, node);
 }
 
 /*
@@ -121,7 +131,7 @@ static void init_one_irq_desc(int irq, struct irq_desc *desc, int cpu)
  */
 DEFINE_SPINLOCK(sparse_irq_lock);
 
-struct irq_desc *irq_desc_ptrs[NR_IRQS] __read_mostly;
+struct irq_desc **irq_desc_ptrs __read_mostly;
 
 static struct irq_desc irq_desc_legacy[NR_IRQS_LEGACY] __cacheline_aligned_in_smp = {
        [0 ... NR_IRQS_LEGACY-1] = {
@@ -131,14 +141,10 @@ static struct irq_desc irq_desc_legacy[NR_IRQS_LEGACY] __cacheline_aligned_in_sm
                .handle_irq = handle_bad_irq,
                .depth      = 1,
                .lock       = __SPIN_LOCK_UNLOCKED(irq_desc_init.lock),
-#ifdef CONFIG_SMP
-               .affinity   = CPU_MASK_ALL
-#endif
        }
 };
 
-/* FIXME: use bootmem alloc ...*/
-static unsigned int kstat_irqs_legacy[NR_IRQS_LEGACY][NR_CPUS];
+static unsigned int *kstat_irqs_legacy;
 
 int __init early_irq_init(void)
 {
@@ -148,18 +154,31 @@ int __init early_irq_init(void)
 
        init_irq_default_affinity();
 
+        /* initialize nr_irqs based on nr_cpu_ids */
+       arch_probe_nr_irqs();
+       printk(KERN_INFO "NR_IRQS:%d nr_irqs:%d\n", NR_IRQS, nr_irqs);
+
        desc = irq_desc_legacy;
        legacy_count = ARRAY_SIZE(irq_desc_legacy);
 
+       /* allocate irq_desc_ptrs array based on nr_irqs */
+       irq_desc_ptrs = alloc_bootmem(nr_irqs * sizeof(void *));
+
+       /* allocate based on nr_cpu_ids */
+       /* FIXME: invert kstat_irgs, and it'd be a per_cpu_alloc'd thing */
+       kstat_irqs_legacy = alloc_bootmem(NR_IRQS_LEGACY * nr_cpu_ids *
+                                         sizeof(int));
+
        for (i = 0; i < legacy_count; i++) {
                desc[i].irq = i;
-               desc[i].kstat_irqs = kstat_irqs_legacy[i];
+               desc[i].kstat_irqs = kstat_irqs_legacy + i * nr_cpu_ids;
                lockdep_set_class(&desc[i].lock, &irq_desc_lock_class);
-
+               alloc_desc_masks(&desc[i], 0, true);
+               init_desc_masks(&desc[i]);
                irq_desc_ptrs[i] = desc + i;
        }
 
-       for (i = legacy_count; i < NR_IRQS; i++)
+       for (i = legacy_count; i < nr_irqs; i++)
                irq_desc_ptrs[i] = NULL;
 
        return arch_early_irq_init();
@@ -167,19 +186,20 @@ int __init early_irq_init(void)
 
 struct irq_desc *irq_to_desc(unsigned int irq)
 {
-       return (irq < NR_IRQS) ? irq_desc_ptrs[irq] : NULL;
+       if (irq_desc_ptrs && irq < nr_irqs)
+               return irq_desc_ptrs[irq];
+
+       return NULL;
 }
 
-struct irq_desc *irq_to_desc_alloc_cpu(unsigned int irq, int cpu)
+struct irq_desc * __ref irq_to_desc_alloc_node(unsigned int irq, int node)
 {
        struct irq_desc *desc;
        unsigned long flags;
-       int node;
 
-       if (irq >= NR_IRQS) {
-               printk(KERN_WARNING "irq >= NR_IRQS in irq_to_desc_alloc: %d %d\n",
-                               irq, NR_IRQS);
-               WARN_ON(1);
+       if (irq >= nr_irqs) {
+               WARN(1, "irq (%d) >= nr_irqs (%d) in irq_to_desc_alloc\n",
+                       irq, nr_irqs);
                return NULL;
        }
 
@@ -194,15 +214,17 @@ struct irq_desc *irq_to_desc_alloc_cpu(unsigned int irq, int cpu)
        if (desc)
                goto out_unlock;
 
-       node = cpu_to_node(cpu);
-       desc = kzalloc_node(sizeof(*desc), GFP_ATOMIC, node);
-       printk(KERN_DEBUG "  alloc irq_desc for %d on cpu %d node %d\n",
-                irq, cpu, node);
+       if (slab_is_available())
+               desc = kzalloc_node(sizeof(*desc), GFP_ATOMIC, node);
+       else
+               desc = alloc_bootmem_node(NODE_DATA(node), sizeof(*desc));
+
+       printk(KERN_DEBUG "  alloc irq_desc for %d on node %d\n", irq, node);
        if (!desc) {
                printk(KERN_ERR "can not alloc irq_desc\n");
                BUG_ON(1);
        }
-       init_one_irq_desc(irq, desc, cpu);
+       init_one_irq_desc(irq, desc, node);
 
        irq_desc_ptrs[irq] = desc;
 
@@ -221,12 +243,10 @@ struct irq_desc irq_desc[NR_IRQS] __cacheline_aligned_in_smp = {
                .handle_irq = handle_bad_irq,
                .depth = 1,
                .lock = __SPIN_LOCK_UNLOCKED(irq_desc->lock),
-#ifdef CONFIG_SMP
-               .affinity = CPU_MASK_ALL
-#endif
        }
 };
 
+static unsigned int kstat_irqs_all[NR_IRQS][NR_CPUS];
 int __init early_irq_init(void)
 {
        struct irq_desc *desc;
@@ -235,12 +255,17 @@ int __init early_irq_init(void)
 
        init_irq_default_affinity();
 
+       printk(KERN_INFO "NR_IRQS:%d\n", NR_IRQS);
+
        desc = irq_desc;
        count = ARRAY_SIZE(irq_desc);
 
-       for (i = 0; i < count; i++)
+       for (i = 0; i < count; i++) {
                desc[i].irq = i;
-
+               alloc_desc_masks(&desc[i], 0, true);
+               init_desc_masks(&desc[i]);
+               desc[i].kstat_irqs = kstat_irqs_all[i];
+       }
        return arch_early_irq_init();
 }
 
@@ -249,12 +274,17 @@ struct irq_desc *irq_to_desc(unsigned int irq)
        return (irq < NR_IRQS) ? irq_desc + irq : NULL;
 }
 
-struct irq_desc *irq_to_desc_alloc_cpu(unsigned int irq, int cpu)
+struct irq_desc *irq_to_desc_alloc_node(unsigned int irq, int node)
 {
        return irq_to_desc(irq);
 }
 #endif /* !CONFIG_SPARSE_IRQ */
 
+void clear_kstat_irqs(struct irq_desc *desc)
+{
+       memset(desc->kstat_irqs, 0, nr_cpu_ids * sizeof(*(desc->kstat_irqs)));
+}
+
 /*
  * What should we do if we get a hw irq event on an illegal vector?
  * Each architecture has to answer this themself.
@@ -316,6 +346,15 @@ irqreturn_t no_action(int cpl, void *dev_id)
        return IRQ_NONE;
 }
 
+static void warn_no_thread(unsigned int irq, struct irqaction *action)
+{
+       if (test_and_set_bit(IRQTF_WARNED, &action->thread_flags))
+               return;
+
+       printk(KERN_WARNING "IRQ %d device %s returned IRQ_WAKE_THREAD "
+              "but no thread function available.", irq, action->name);
+}
+
 /**
  * handle_IRQ_event - irq action chain handler
  * @irq:       the interrupt number
@@ -328,15 +367,54 @@ irqreturn_t handle_IRQ_event(unsigned int irq, struct irqaction *action)
        irqreturn_t ret, retval = IRQ_NONE;
        unsigned int status = 0;
 
-       WARN_ONCE(!in_irq(), "BUG: IRQ handler called from non-hardirq context!");
-
        if (!(action->flags & IRQF_DISABLED))
                local_irq_enable_in_hardirq();
 
        do {
+               trace_irq_handler_entry(irq, action);
                ret = action->handler(irq, action->dev_id);
-               if (ret == IRQ_HANDLED)
+               trace_irq_handler_exit(irq, action, ret);
+
+               switch (ret) {
+               case IRQ_WAKE_THREAD:
+                       /*
+                        * Set result to handled so the spurious check
+                        * does not trigger.
+                        */
+                       ret = IRQ_HANDLED;
+
+                       /*
+                        * Catch drivers which return WAKE_THREAD but
+                        * did not set up a thread function
+                        */
+                       if (unlikely(!action->thread_fn)) {
+                               warn_no_thread(irq, action);
+                               break;
+                       }
+
+                       /*
+                        * Wake up the handler thread for this
+                        * action. In case the thread crashed and was
+                        * killed we just pretend that we handled the
+                        * interrupt. The hardirq handler above has
+                        * disabled the device interrupt, so no irq
+                        * storm is lurking.
+                        */
+                       if (likely(!test_bit(IRQTF_DIED,
+                                            &action->thread_flags))) {
+                               set_bit(IRQTF_RUNTHREAD, &action->thread_flags);
+                               wake_up_process(action->thread);
+                       }
+
+                       /* Fall through to add to randomness */
+               case IRQ_HANDLED:
                        status |= action->flags;
+                       break;
+
+               default:
+                       break;
+               }
+
                retval |= ret;
                action = action->next;
        } while (action);
@@ -349,6 +427,11 @@ irqreturn_t handle_IRQ_event(unsigned int irq, struct irqaction *action)
 }
 
 #ifndef CONFIG_GENERIC_HARDIRQS_NO__DO_IRQ
+
+#ifdef CONFIG_ENABLE_WARN_DEPRECATED
+# warning __do_IRQ is deprecated. Please convert to proper flow handlers
+#endif
+
 /**
  * __do_IRQ - original all in one highlevel IRQ handler
  * @irq:       the interrupt number
@@ -374,11 +457,8 @@ unsigned int __do_IRQ(unsigned int irq)
                /*
                 * No locking required for CPU-local interrupts:
                 */
-               if (desc->chip->ack) {
+               if (desc->chip->ack)
                        desc->chip->ack(irq);
-                       /* get new one */
-                       desc = irq_remap_to_desc(irq, desc);
-               }
                if (likely(!(desc->status & IRQ_DISABLED))) {
                        action_ret = handle_IRQ_event(irq, desc->action);
                        if (!noirqdebug)
@@ -389,10 +469,8 @@ unsigned int __do_IRQ(unsigned int irq)
        }
 
        spin_lock(&desc->lock);
-       if (desc->chip->ack) {
+       if (desc->chip->ack)
                desc->chip->ack(irq);
-               desc = irq_remap_to_desc(irq, desc);
-       }
        /*
         * REPLAY is when Linux resends an IRQ that was dropped earlier
         * WAITING is used by probe to mark irqs that are being tested
@@ -469,12 +547,10 @@ void early_init_irq_lock_class(void)
        }
 }
 
-#ifdef CONFIG_SPARSE_IRQ
 unsigned int kstat_irqs_cpu(unsigned int irq, int cpu)
 {
        struct irq_desc *desc = irq_to_desc(irq);
        return desc ? desc->kstat_irqs[cpu] : 0;
 }
-#endif
 EXPORT_SYMBOL(kstat_irqs_cpu);