genirq: Delegate irq affinity setting to the irq thread

irq_set_thread_affinity() calls set_cpus_allowed_ptr() which might
sleep, but irq_set_thread_affinity() is called with desc->lock held
and can be called from hard interrupt context as well. The code has
another bug as it does not hold a ref on the task struct as required
by set_cpus_allowed_ptr().

Just set the IRQTF_AFFINITY bit in action->thread_flags. The next time
the thread runs it migrates itself. Solves all of the above problems
nicely.

Add kerneldoc to irq_set_thread_affinity() while at it.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <new-submission>

diff --git a/include/linux/interrupt.h b/include/linux/interrupt.h
index 2721f07..88b056a 100644 (file)
--- a/include/linux/interrupt.h
+++ b/include/linux/interrupt.h
@@ -64,11 +64,13 @@
  * IRQTF_RUNTHREAD - signals that the interrupt handler thread should run
  * IRQTF_DIED      - handler thread died
  * IRQTF_WARNED    - warning "IRQ_WAKE_THREAD w/o thread_fn" has been printed
+ * IRQTF_AFFINITY  - irq thread is requested to adjust affinity
  */
 enum {
        IRQTF_RUNTHREAD,
        IRQTF_DIED,
        IRQTF_WARNED,
+       IRQTF_AFFINITY,
 };
 
 typedef irqreturn_t (*irq_handler_t)(int, void *);

diff --git a/kernel/irq/internals.h b/kernel/irq/internals.h
index 7346825..e70ed55 100644 (file)
--- a/kernel/irq/internals.h
+++ b/kernel/irq/internals.h
@@ -42,8 +42,7 @@ static inline void unregister_handler_proc(unsigned int irq,
 
 extern int irq_select_affinity_usr(unsigned int irq);
 
-extern void
-irq_set_thread_affinity(struct irq_desc *desc, const struct cpumask *cpumask);
+extern void irq_set_thread_affinity(struct irq_desc *desc);
 
 /*
  * Debugging printout:

diff --git a/kernel/irq/manage.c b/kernel/irq/manage.c
index 50da676..f0de36f 100644 (file)
--- a/kernel/irq/manage.c
+++ b/kernel/irq/manage.c
@@ -80,14 +80,22 @@ int irq_can_set_affinity(unsigned int irq)
        return 1;
 }
 
-void
-irq_set_thread_affinity(struct irq_desc *desc, const struct cpumask *cpumask)
+/**
+ *     irq_set_thread_affinity - Notify irq threads to adjust affinity
+ *     @desc:          irq descriptor which has affitnity changed
+ *
+ *     We just set IRQTF_AFFINITY and delegate the affinity setting
+ *     to the interrupt thread itself. We can not call
+ *     set_cpus_allowed_ptr() here as we hold desc->lock and this
+ *     code can be called from hard interrupt context.
+ */
+void irq_set_thread_affinity(struct irq_desc *desc)
 {
        struct irqaction *action = desc->action;
 
        while (action) {
                if (action->thread)
-                       set_cpus_allowed_ptr(action->thread, cpumask);
+                       set_bit(IRQTF_AFFINITY, &action->thread_flags);
                action = action->next;
        }
 }
@@ -112,7 +120,7 @@ int irq_set_affinity(unsigned int irq, const struct cpumask *cpumask)
        if (desc->status & IRQ_MOVE_PCNTXT) {
                if (!desc->chip->set_affinity(irq, cpumask)) {
                        cpumask_copy(desc->affinity, cpumask);
-                       irq_set_thread_affinity(desc, cpumask);
+                       irq_set_thread_affinity(desc);
                }
        }
        else {
@@ -122,7 +130,7 @@ int irq_set_affinity(unsigned int irq, const struct cpumask *cpumask)
 #else
        if (!desc->chip->set_affinity(irq, cpumask)) {
                cpumask_copy(desc->affinity, cpumask);
-               irq_set_thread_affinity(desc, cpumask);
+               irq_set_thread_affinity(desc);
        }
 #endif
        desc->status |= IRQ_AFFINITY_SET;
@@ -176,7 +184,7 @@ int irq_select_affinity_usr(unsigned int irq)
        spin_lock_irqsave(&desc->lock, flags);
        ret = setup_affinity(irq, desc);
        if (!ret)
-               irq_set_thread_affinity(desc, desc->affinity);
+               irq_set_thread_affinity(desc);
        spin_unlock_irqrestore(&desc->lock, flags);
 
        return ret;
@@ -444,6 +452,34 @@ static int irq_wait_for_interrupt(struct irqaction *action)
 }
 
 /*
+ * Check whether we need to change the affinity of the interrupt thread.
+ */
+static void
+irq_thread_check_affinity(struct irq_desc *desc, struct irqaction *action)
+{
+       cpumask_var_t mask;
+
+       if (!test_and_clear_bit(IRQTF_AFFINITY, &action->thread_flags))
+               return;
+
+       /*
+        * In case we are out of memory we set IRQTF_AFFINITY again and
+        * try again next time
+        */
+       if (!alloc_cpumask_var(&mask, GFP_KERNEL)) {
+               set_bit(IRQTF_AFFINITY, &action->thread_flags);
+               return;
+       }
+
+       spin_lock_irq(&desc->lock);
+       cpumask_copy(mask, desc->affinity);
+       spin_unlock_irq(&desc->lock);
+
+       set_cpus_allowed_ptr(current, mask);
+       free_cpumask_var(mask);
+}
+
+/*
  * Interrupt handler thread
  */
 static int irq_thread(void *data)
@@ -458,6 +494,8 @@ static int irq_thread(void *data)
 
        while (!irq_wait_for_interrupt(action)) {
 
+               irq_thread_check_affinity(desc, action);
+
                atomic_inc(&desc->threads_active);
 
                spin_lock_irq(&desc->lock);

diff --git a/kernel/irq/migration.c b/kernel/irq/migration.c
index cfe767c..fcb6c96 100644 (file)
--- a/kernel/irq/migration.c
+++ b/kernel/irq/migration.c
@@ -45,7 +45,7 @@ void move_masked_irq(int irq)
                   < nr_cpu_ids))
                if (!desc->chip->set_affinity(irq, desc->pending_mask)) {
                        cpumask_copy(desc->affinity, desc->pending_mask);
-                       irq_set_thread_affinity(desc, desc->pending_mask);
+                       irq_set_thread_affinity(desc);
                }
 
        cpumask_clear(desc->pending_mask);