tracing/rb: Convert to hotplug state machine

Install the callbacks via the state machine. The notifier in struct
ring_buffer is replaced by the multi instance interface.  Upon
__ring_buffer_alloc() invocation, cpuhp_state_add_instance() will invoke
the trace_rb_cpu_prepare() on each CPU.

This callback may now fail. This means __ring_buffer_alloc() will fail and
cleanup (like previously) and during a CPU up event this failure will not
allow the CPU to come up.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: rt@linutronix.de
Link: http://lkml.kernel.org/r/20161126231350.10321-7-bigeasy@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

include/linux/cpuhotplug.h

@@ -62,6 +62,7 @@ enum cpuhp_state {
 	CPUHP_TOPOLOGY_PREPARE,
 	CPUHP_NET_IUCV_PREPARE,
 	CPUHP_ARM_BL_PREPARE,
+	CPUHP_TRACE_RB_PREPARE,
 	CPUHP_TIMERS_DEAD,
 	CPUHP_NOTF_ERR_INJ_PREPARE,
 	CPUHP_MIPS_SOC_PREPARE,

include/linux/ring_buffer.h

@@ -198,4 +198,10 @@ enum ring_buffer_flags {
 	RB_FL_OVERWRITE		= 1 << 0,
 };
 
+#ifdef CONFIG_RING_BUFFER
+int trace_rb_cpu_prepare(unsigned int cpu, struct hlist_node *node);
+#else
+#define trace_rb_cpu_prepare	NULL
+#endif
+
 #endif /* _LINUX_RING_BUFFER_H */

kernel/trace/ring_buffer.c

@@ -479,9 +479,7 @@ struct ring_buffer {
 
 	struct ring_buffer_per_cpu	**buffers;
 
-#ifdef CONFIG_HOTPLUG_CPU
-	struct notifier_block		cpu_notify;
-#endif
+	struct hlist_node		node;
 	u64				(*clock)(void);
 
 	struct rb_irq_work		irq_work;
@@ -1274,11 +1272,6 @@ static void rb_free_cpu_buffer(struct ring_buffer_per_cpu *cpu_buffer)
 	kfree(cpu_buffer);
 }
 
-#ifdef CONFIG_HOTPLUG_CPU
-static int rb_cpu_notify(struct notifier_block *self,
-			 unsigned long action, void *hcpu);
-#endif
-
 /**
  * __ring_buffer_alloc - allocate a new ring_buffer
  * @size: the size in bytes per cpu that is needed.
@@ -1296,6 +1289,7 @@ struct ring_buffer *__ring_buffer_alloc(unsigned long size, unsigned flags,
 	long nr_pages;
 	int bsize;
 	int cpu;
+	int ret;
 
 	/* keep it in its own cache line */
 	buffer = kzalloc(ALIGN(sizeof(*buffer), cache_line_size()),
@@ -1318,17 +1312,6 @@ struct ring_buffer *__ring_buffer_alloc(unsigned long size, unsigned flags,
 	if (nr_pages < 2)
 		nr_pages = 2;
 
-	/*
-	 * In case of non-hotplug cpu, if the ring-buffer is allocated
-	 * in early initcall, it will not be notified of secondary cpus.
-	 * In that off case, we need to allocate for all possible cpus.
-	 */
-#ifdef CONFIG_HOTPLUG_CPU
-	cpu_notifier_register_begin();
-	cpumask_copy(buffer->cpumask, cpu_online_mask);
-#else
-	cpumask_copy(buffer->cpumask, cpu_possible_mask);
-#endif
 	buffer->cpus = nr_cpu_ids;
 
 	bsize = sizeof(void *) * nr_cpu_ids;
@@ -1337,19 +1320,15 @@ struct ring_buffer *__ring_buffer_alloc(unsigned long size, unsigned flags,
 	if (!buffer->buffers)
 		goto fail_free_cpumask;
 
-	for_each_buffer_cpu(buffer, cpu) {
-		buffer->buffers[cpu] =
-			rb_allocate_cpu_buffer(buffer, nr_pages, cpu);
-		if (!buffer->buffers[cpu])
-			goto fail_free_buffers;
-	}
+	cpu = raw_smp_processor_id();
+	cpumask_set_cpu(cpu, buffer->cpumask);
+	buffer->buffers[cpu] = rb_allocate_cpu_buffer(buffer, nr_pages, cpu);
+	if (!buffer->buffers[cpu])
+		goto fail_free_buffers;
 
-#ifdef CONFIG_HOTPLUG_CPU
-	buffer->cpu_notify.notifier_call = rb_cpu_notify;
-	buffer->cpu_notify.priority = 0;
-	__register_cpu_notifier(&buffer->cpu_notify);
-	cpu_notifier_register_done();
-#endif
+	ret = cpuhp_state_add_instance(CPUHP_TRACE_RB_PREPARE, &buffer->node);
+	if (ret < 0)
+		goto fail_free_buffers;
 
 	mutex_init(&buffer->mutex);
 
@@ -1364,9 +1343,6 @@ struct ring_buffer *__ring_buffer_alloc(unsigned long size, unsigned flags,
 
  fail_free_cpumask:
 	free_cpumask_var(buffer->cpumask);
-#ifdef CONFIG_HOTPLUG_CPU
-	cpu_notifier_register_done();
-#endif
 
  fail_free_buffer:
 	kfree(buffer);
@@ -1383,18 +1359,11 @@ ring_buffer_free(struct ring_buffer *buffer)
 {
 	int cpu;
 
-#ifdef CONFIG_HOTPLUG_CPU
-	cpu_notifier_register_begin();
-	__unregister_cpu_notifier(&buffer->cpu_notify);
-#endif
+	cpuhp_state_remove_instance(CPUHP_TRACE_RB_PREPARE, &buffer->node);
 
 	for_each_buffer_cpu(buffer, cpu)
 		rb_free_cpu_buffer(buffer->buffers[cpu]);
 
-#ifdef CONFIG_HOTPLUG_CPU
-	cpu_notifier_register_done();
-#endif
-
 	kfree(buffer->buffers);
 	free_cpumask_var(buffer->cpumask);
 
@@ -4633,62 +4602,48 @@ int ring_buffer_read_page(struct ring_buffer *buffer,
 }
 EXPORT_SYMBOL_GPL(ring_buffer_read_page);
 
-#ifdef CONFIG_HOTPLUG_CPU
-static int rb_cpu_notify(struct notifier_block *self,
-			 unsigned long action, void *hcpu)
+/*
+ * We only allocate new buffers, never free them if the CPU goes down.
+ * If we were to free the buffer, then the user would lose any trace that was in
+ * the buffer.
+ */
+int trace_rb_cpu_prepare(unsigned int cpu, struct hlist_node *node)
 {
-	struct ring_buffer *buffer =
-		container_of(self, struct ring_buffer, cpu_notify);
-	long cpu = (long)hcpu;
+	struct ring_buffer *buffer;
 	long nr_pages_same;
 	int cpu_i;
 	unsigned long nr_pages;
 
-	switch (action) {
-	case CPU_UP_PREPARE:
-	case CPU_UP_PREPARE_FROZEN:
-		if (cpumask_test_cpu(cpu, buffer->cpumask))
-			return NOTIFY_OK;
-
-		nr_pages = 0;
-		nr_pages_same = 1;
-		/* check if all cpu sizes are same */
-		for_each_buffer_cpu(buffer, cpu_i) {
-			/* fill in the size from first enabled cpu */
-			if (nr_pages == 0)
-				nr_pages = buffer->buffers[cpu_i]->nr_pages;
-			if (nr_pages != buffer->buffers[cpu_i]->nr_pages) {
-				nr_pages_same = 0;
-				break;
-			}
-		}
-		/* allocate minimum pages, user can later expand it */
-		if (!nr_pages_same)
-			nr_pages = 2;
-		buffer->buffers[cpu] =
-			rb_allocate_cpu_buffer(buffer, nr_pages, cpu);
-		if (!buffer->buffers[cpu]) {
-			WARN(1, "failed to allocate ring buffer on CPU %ld\n",
-			     cpu);
-			return NOTIFY_OK;
+	buffer = container_of(node, struct ring_buffer, node);
+	if (cpumask_test_cpu(cpu, buffer->cpumask))
+		return 0;
+
+	nr_pages = 0;
+	nr_pages_same = 1;
+	/* check if all cpu sizes are same */
+	for_each_buffer_cpu(buffer, cpu_i) {
+		/* fill in the size from first enabled cpu */
+		if (nr_pages == 0)
+			nr_pages = buffer->buffers[cpu_i]->nr_pages;
+		if (nr_pages != buffer->buffers[cpu_i]->nr_pages) {
+			nr_pages_same = 0;
+			break;
 		}
-		smp_wmb();
-		cpumask_set_cpu(cpu, buffer->cpumask);
-		break;
-	case CPU_DOWN_PREPARE:
-	case CPU_DOWN_PREPARE_FROZEN:
-		/*
-		 * Do nothing.
-		 *  If we were to free the buffer, then the user would
-		 *  lose any trace that was in the buffer.
-		 */
-		break;
-	default:
-		break;
 	}
-	return NOTIFY_OK;
+	/* allocate minimum pages, user can later expand it */
+	if (!nr_pages_same)
+		nr_pages = 2;
+	buffer->buffers[cpu] =
+		rb_allocate_cpu_buffer(buffer, nr_pages, cpu);
+	if (!buffer->buffers[cpu]) {
+		WARN(1, "failed to allocate ring buffer on CPU %u\n",
+		     cpu);
+		return -ENOMEM;
+	}
+	smp_wmb();
+	cpumask_set_cpu(cpu, buffer->cpumask);
+	return 0;
 }
-#endif
 
 #ifdef CONFIG_RING_BUFFER_STARTUP_TEST
 /*

kernel/trace/trace.c

@@ -7659,10 +7659,21 @@ __init static int tracer_alloc_buffers(void)
 
 	raw_spin_lock_init(&global_trace.start_lock);
 
+	/*
+	 * The prepare callbacks allocates some memory for the ring buffer. We
+	 * don't free the buffer if the if the CPU goes down. If we were to free
+	 * the buffer, then the user would lose any trace that was in the
+	 * buffer. The memory will be removed once the "instance" is removed.
+	 */
+	ret = cpuhp_setup_state_multi(CPUHP_TRACE_RB_PREPARE,
+				      "trace/RB:preapre", trace_rb_cpu_prepare,
+				      NULL);
+	if (ret < 0)
+		goto out_free_cpumask;
 	/* Used for event triggers */
 	temp_buffer = ring_buffer_alloc(PAGE_SIZE, RB_FL_OVERWRITE);
 	if (!temp_buffer)
-		goto out_free_cpumask;
+		goto out_rm_hp_state;
 
 	if (trace_create_savedcmd() < 0)
 		goto out_free_temp_buffer;
@@ -7723,6 +7734,8 @@ out_free_savedcmd:
 	free_saved_cmdlines_buffer(savedcmd);
 out_free_temp_buffer:
 	ring_buffer_free(temp_buffer);
+out_rm_hp_state:
+	cpuhp_remove_multi_state(CPUHP_TRACE_RB_PREPARE);
 out_free_cpumask:
 	free_cpumask_var(global_trace.tracing_cpumask);
 out_free_buffer_mask: