genirq/cpuhotplug: Add support for cleaning up move in progress

In order to move x86 to the generic hotplug migration code, add support for
cleaning up move in progress bits.

On architectures which have this x86 specific (mis)feature not enabled,
this is optimized out by the compiler.

Signed-off-by: Thomas Gleixner <[email protected]>
Cc: Jens Axboe <[email protected]>
Cc: Marc Zyngier <[email protected]>
Cc: Michael Ellerman <[email protected]>
Cc: Keith Busch <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Christoph Hellwig <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]

Author: KAGA-KOKO

arch/x86/include/asm/irq.h

@@ -29,7 +29,6 @@ struct irq_desc;
 #include <linux/cpumask.h>
 extern int check_irq_vectors_for_cpu_disable(void);
 extern void fixup_irqs(void);
-extern void irq_force_complete_move(struct irq_desc *desc);
 #endif
 
 #ifdef CONFIG_HAVE_KVM

include/linux/irq.h

@@ -491,10 +491,12 @@ extern void irq_migrate_all_off_this_cpu(void);
 #if defined(CONFIG_SMP) && defined(CONFIG_GENERIC_PENDING_IRQ)
 void irq_move_irq(struct irq_data *data);
 void irq_move_masked_irq(struct irq_data *data);
+void irq_force_complete_move(struct irq_desc *desc);
 bool irq_fixup_move_pending(struct irq_desc *desc, bool force_clear);
 #else
 static inline void irq_move_irq(struct irq_data *data) { }
 static inline void irq_move_masked_irq(struct irq_data *data) { }
+static inline void irq_force_complete_move(struct irq_desc *desc) { }
 static inline bool irq_fixup_move_pending(struct irq_desc *desc, bool fclear)
 {
 	return false;

kernel/irq/cpuhotplug.c

@@ -18,7 +18,7 @@ static bool migrate_one_irq(struct irq_desc *desc)
 {
 	struct irq_data *d = irq_desc_get_irq_data(desc);
 	struct irq_chip *chip = irq_data_get_irq_chip(d);
-	const struct cpumask *affinity = d->common->affinity;
+	const struct cpumask *affinity;
 	bool brokeaff = false;
 	int err;
 
@@ -41,9 +41,33 @@ static bool migrate_one_irq(struct irq_desc *desc)
 	 * Note: Do not check desc->action as this might be a chained
 	 * interrupt.
 	 */
+	affinity = irq_data_get_affinity_mask(d);
 	if (irqd_is_per_cpu(d) || !irqd_is_started(d) ||
-	    !cpumask_test_cpu(smp_processor_id(), affinity))
+	    !cpumask_test_cpu(smp_processor_id(), affinity)) {
+		/*
+		 * If an irq move is pending, abort it if the dying CPU is
+		 * the sole target.
+		 */
+		irq_fixup_move_pending(desc, false);
 		return false;
+	}
+
+	/*
+	 * Complete an eventually pending irq move cleanup. If this
+	 * interrupt was moved in hard irq context, then the vectors need
+	 * to be cleaned up. It can't wait until this interrupt actually
+	 * happens and this CPU was involved.
+	 */
+	irq_force_complete_move(desc);
+
+	/*
+	 * If there is a setaffinity pending, then try to reuse the pending
+	 * mask, so the last change of the affinity does not get lost. If
+	 * there is no move pending or the pending mask does not contain
+	 * any online CPU, use the current affinity mask.
+	 */
+	if (irq_fixup_move_pending(desc, true))
+		affinity = irq_desc_get_pending_mask(desc);
 
 	if (cpumask_any_and(affinity, cpu_online_mask) >= nr_cpu_ids) {
 		affinity = cpu_online_mask;

kernel/irq/internals.h

@@ -268,6 +268,10 @@ irq_get_pending(struct cpumask *mask, struct irq_desc *desc)
 {
 	cpumask_copy(mask, desc->pending_mask);
 }
+static inline struct cpumask *irq_desc_get_pending_mask(struct irq_desc *desc)
+{
+	return desc->pending_mask;
+}
 #else /* CONFIG_GENERIC_PENDING_IRQ */
 static inline bool irq_can_move_pcntxt(struct irq_data *data)
 {
@@ -285,7 +289,11 @@ static inline void
 irq_get_pending(struct cpumask *mask, struct irq_desc *desc)
 {
 }
-#endif /* CONFIG_GENERIC_PENDING_IRQ */
+static inline struct cpumask *irq_desc_get_pending_mask(struct irq_desc *desc)
+{
+	return NULL;
+}
+#endif /* !CONFIG_GENERIC_PENDING_IRQ */
 
 #ifdef CONFIG_GENERIC_IRQ_DEBUGFS
 void irq_add_debugfs_entry(unsigned int irq, struct irq_desc *desc);