ftrace: dynamic enabling/disabling of function calls

This patch adds a feature to dynamically replace the ftrace code
with the jmps to allow a kernel with ftrace configured to run
as fast as it can without it configured.

The way this works, is on bootup (if ftrace is enabled), a ftrace
function is registered to record the instruction pointer of all
places that call the function.

Later, if there's still any code to patch, a kthread is awoken
(rate limited to at most once a second) that performs a stop_machine,
and replaces all the code that was called with a jmp over the call
to ftrace. It only replaces what was found the previous time. Typically
the system reaches equilibrium quickly after bootup and there's no code
patching needed at all.

e.g.

  call ftrace  /* 5 bytes */

is replaced with

  jmp 3f  /* jmp is 2 bytes and we jump 3 forward */
3:

When we want to enable ftrace for function tracing, the IP recording
is removed, and stop_machine is called again to replace all the locations
of that were recorded back to the call of ftrace.  When it is disabled,
we replace the code back to the jmp.

Allocation is done by the kthread. If the ftrace recording function is
called, and we don't have any record slots available, then we simply
skip that call. Once a second a new page (if needed) is allocated for
recording new ftrace function calls.  A large batch is allocated at
boot up to get most of the calls there.

Because we do this via stop_machine, we don't have to worry about another
CPU executing a ftrace call as we modify it. But we do need to worry
about NMI's so all functions that might be called via nmi must be
annotated with notrace_nmi. When this code is configured in, the NMI code
will not call notrace.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

2 files changed, 341 insertions, 32 deletions

diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig
index 6430016b98e..cad9db1dee0 100644
--- a/kernel/trace/Kconfig
+++ b/kernel/trace/Kconfig
@@ -88,3 +88,20 @@ config CONTEXT_SWITCH_TRACER
 	  This tracer gets called from the context switch and records
 	  all switching of tasks.
 
+config DYNAMIC_FTRACE
+	bool "enable/disable ftrace tracepoints dynamically"
+	depends on FTRACE
+	default y
+	help
+         This option will modify all the calls to ftrace dynamically
+	 (will patch them out of the binary image and replaces them
+	 with a No-Op instruction) as they are called. A table is
+	 created to dynamically enable them again.
+
+	 This way a CONFIG_FTRACE kernel is slightly larger, but otherwise
+	 has native performance as long as no tracing is active.
+
+	 The changes to the code are done by a kernel thread that
+	 wakes up once a second and checks to see if any ftrace calls
+	 were made. If so, it runs stop_machine (stops all CPUS)
+	 and modifies the code to jump over the call to ftrace.
diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
index b6a80b98a3f..d1ae2ba2527 100644
--- a/kernel/trace/ftrace.c
+++ b/kernel/trace/ftrace.c
@@ -13,10 +13,19 @@
  *  Copyright (C) 2004 William Lee Irwin III
  */
 
-#include <linux/module.h>
+#include <linux/stop_machine.h>
+#include <linux/clocksource.h>
+#include <linux/kallsyms.h>
+#include <linux/kthread.h>
+#include <linux/hardirq.h>
 #include <linux/ftrace.h>
+#include <linux/module.h>
+#include <linux/hash.h>
+#include <linux/list.h>
+
+#include "trace.h"
 
-static DEFINE_SPINLOCK(ftrace_func_lock);
+static DEFINE_SPINLOCK(ftrace_lock);
 static struct ftrace_ops ftrace_list_end __read_mostly =
 {
 	.func = ftrace_stub,
@@ -44,21 +53,21 @@ notrace void ftrace_list_func(unsigned long ip, unsigned long parent_ip)
 }
 
 /**
- * register_ftrace_function - register a function for profiling
- * @ops - ops structure that holds the function for profiling.
- *
- * Register a function to be called by all functions in the
- * kernel.
+ * clear_ftrace_function - reset the ftrace function
  *
- * Note: @ops->func and all the functions it calls must be labeled
- *       with "notrace", otherwise it will go into a
- *       recursive loop.
+ * This NULLs the ftrace function and in essence stops
+ * tracing.  There may be lag
  */
-int register_ftrace_function(struct ftrace_ops *ops)
+void clear_ftrace_function(void)
 {
-	unsigned long flags;
+	ftrace_trace_function = ftrace_stub;
+}
+
+static int notrace __register_ftrace_function(struct ftrace_ops *ops)
+{
+	/* Should never be called by interrupts */
+	spin_lock(&ftrace_lock);
 
-	spin_lock_irqsave(&ftrace_func_lock, flags);
 	ops->next = ftrace_list;
 	/*
 	 * We are entering ops into the ftrace_list but another
@@ -68,6 +77,7 @@ int register_ftrace_function(struct ftrace_ops *ops)
 	 */
 	smp_wmb();
 	ftrace_list = ops;
+
 	/*
 	 * For one func, simply call it directly.
 	 * For more than one func, call the chain.
@@ -76,28 +86,22 @@ int register_ftrace_function(struct ftrace_ops *ops)
 		ftrace_trace_function = ops->func;
 	else
 		ftrace_trace_function = ftrace_list_func;
-	spin_unlock_irqrestore(&ftrace_func_lock, flags);
+
+	spin_unlock(&ftrace_lock);
 
 	return 0;
 }
 
-/**
- * unregister_ftrace_function - unresgister a function for profiling.
- * @ops - ops structure that holds the function to unregister
- *
- * Unregister a function that was added to be called by ftrace profiling.
- */
-int unregister_ftrace_function(struct ftrace_ops *ops)
+static int notrace __unregister_ftrace_function(struct ftrace_ops *ops)
 {
-	unsigned long flags;
 	struct ftrace_ops **p;
 	int ret = 0;
 
-	spin_lock_irqsave(&ftrace_func_lock, flags);
+	spin_lock(&ftrace_lock);
 
 	/*
-	 * If we are the only function, then the ftrace pointer is
-	 * pointing directly to that function.
+	 * If we are removing the last function, then simply point
+	 * to the ftrace_stub.
 	 */
 	if (ftrace_list == ops && ops->next == &ftrace_list_end) {
 		ftrace_trace_function = ftrace_stub;
@@ -117,22 +121,310 @@ int unregister_ftrace_function(struct ftrace_ops *ops)
 	*p = (*p)->next;
 
 	/* If we only have one func left, then call that directly */
-	if (ftrace_list->next == &ftrace_list_end)
+	if (ftrace_list == &ftrace_list_end ||
+	    ftrace_list->next == &ftrace_list_end)
 		ftrace_trace_function = ftrace_list->func;
 
  out:
-	spin_unlock_irqrestore(&ftrace_func_lock, flags);
+	spin_unlock(&ftrace_lock);
+
+	return ret;
+}
+
+#ifdef CONFIG_DYNAMIC_FTRACE
+
+static struct hlist_head ftrace_hash[FTRACE_HASHSIZE];
+
+static DEFINE_PER_CPU(int, ftrace_shutdown_disable_cpu);
+
+static DEFINE_SPINLOCK(ftrace_shutdown_lock);
+static DEFINE_MUTEX(ftraced_lock);
+
+static int ftraced_trigger;
+static int ftraced_suspend;
+
+static int ftrace_record_suspend;
+
+static inline int
+notrace ftrace_ip_in_hash(unsigned long ip, unsigned long key)
+{
+	struct dyn_ftrace *p;
+	struct hlist_node *t;
+	int found = 0;
+
+	hlist_for_each_entry(p, t, &ftrace_hash[key], node) {
+		if (p->ip == ip) {
+			found = 1;
+			break;
+		}
+	}
+
+	return found;
+}
+
+static inline void notrace
+ftrace_add_hash(struct dyn_ftrace *node, unsigned long key)
+{
+	hlist_add_head(&node->node, &ftrace_hash[key]);
+}
+
+static void notrace
+ftrace_record_ip(unsigned long ip, unsigned long parent_ip)
+{
+	struct dyn_ftrace *node;
+	unsigned long flags;
+	unsigned long key;
+	int resched;
+	int atomic;
+
+	resched = need_resched();
+	preempt_disable_notrace();
+
+	/* We simply need to protect against recursion */
+	__get_cpu_var(ftrace_shutdown_disable_cpu)++;
+	if (__get_cpu_var(ftrace_shutdown_disable_cpu) != 1)
+		goto out;
+
+	if (unlikely(ftrace_record_suspend))
+		goto out;
+
+	key = hash_long(ip, FTRACE_HASHBITS);
+
+	WARN_ON_ONCE(key >= FTRACE_HASHSIZE);
+
+	if (ftrace_ip_in_hash(ip, key))
+		goto out;
+
+	atomic = irqs_disabled();
+
+	spin_lock_irqsave(&ftrace_shutdown_lock, flags);
+
+	/* This ip may have hit the hash before the lock */
+	if (ftrace_ip_in_hash(ip, key))
+		goto out_unlock;
+
+	/*
+	 * There's a slight race that the ftraced will update the
+	 * hash and reset here. The arch alloc is responsible
+	 * for seeing if the IP has already changed, and if
+	 * it has, the alloc will fail.
+	 */
+	node = ftrace_alloc_shutdown_node(ip);
+	if (!node)
+		goto out_unlock;
+
+	node->ip = ip;
+
+	ftrace_add_hash(node, key);
+
+	ftraced_trigger = 1;
+
+ out_unlock:
+	spin_unlock_irqrestore(&ftrace_shutdown_lock, flags);
+ out:
+	__get_cpu_var(ftrace_shutdown_disable_cpu)--;
+
+	/* prevent recursion with scheduler */
+	if (resched)
+		preempt_enable_no_resched_notrace();
+	else
+		preempt_enable_notrace();
+}
+
+static struct ftrace_ops ftrace_shutdown_ops __read_mostly =
+{
+	.func = ftrace_record_ip,
+};
+
+
+static int notrace __ftrace_modify_code(void *data)
+{
+	void (*func)(void) = data;
+
+	func();
+	return 0;
+}
+
+static void notrace ftrace_run_startup_code(void)
+{
+	stop_machine_run(__ftrace_modify_code, ftrace_startup_code, NR_CPUS);
+}
+
+static void notrace ftrace_run_shutdown_code(void)
+{
+	stop_machine_run(__ftrace_modify_code, ftrace_shutdown_code, NR_CPUS);
+}
+
+static void notrace ftrace_startup(void)
+{
+	mutex_lock(&ftraced_lock);
+	ftraced_suspend++;
+	if (ftraced_suspend != 1)
+		goto out;
+	__unregister_ftrace_function(&ftrace_shutdown_ops);
+
+	ftrace_run_startup_code();
+ out:
+	mutex_unlock(&ftraced_lock);
+}
+
+static void notrace ftrace_shutdown(void)
+{
+	mutex_lock(&ftraced_lock);
+	ftraced_suspend--;
+	if (ftraced_suspend)
+		goto out;
+
+	ftrace_run_shutdown_code();
+
+	__register_ftrace_function(&ftrace_shutdown_ops);
+ out:
+	mutex_unlock(&ftraced_lock);
+}
+
+static cycle_t		ftrace_update_time;
+static unsigned long	ftrace_update_cnt;
+unsigned long		ftrace_update_tot_cnt;
+
+static int notrace __ftrace_update_code(void *ignore)
+{
+	struct dyn_ftrace *p;
+	struct hlist_head head;
+	struct hlist_node *t;
+	cycle_t start, stop;
+	int i;
+
+	/* Don't be calling ftrace ops now */
+	__unregister_ftrace_function(&ftrace_shutdown_ops);
+
+	start = now(raw_smp_processor_id());
+	ftrace_update_cnt = 0;
+
+	/* No locks needed, the machine is stopped! */
+	for (i = 0; i < FTRACE_HASHSIZE; i++) {
+		if (hlist_empty(&ftrace_hash[i]))
+			continue;
+
+		head = ftrace_hash[i];
+		INIT_HLIST_HEAD(&ftrace_hash[i]);
+
+		/* all CPUS are stopped, we are safe to modify code */
+		hlist_for_each_entry(p, t, &head, node) {
+			ftrace_code_disable(p);
+			ftrace_update_cnt++;
+		}
+
+	}
+
+	stop = now(raw_smp_processor_id());
+	ftrace_update_time = stop - start;
+	ftrace_update_tot_cnt += ftrace_update_cnt;
+
+	__register_ftrace_function(&ftrace_shutdown_ops);
 
 	return 0;
 }
 
+static void notrace ftrace_update_code(void)
+{
+	stop_machine_run(__ftrace_update_code, NULL, NR_CPUS);
+}
+
+static int notrace ftraced(void *ignore)
+{
+	unsigned long usecs;
+
+	set_current_state(TASK_INTERRUPTIBLE);
+
+	while (!kthread_should_stop()) {
+
+		/* check once a second */
+		schedule_timeout(HZ);
+
+		mutex_lock(&ftraced_lock);
+		if (ftraced_trigger && !ftraced_suspend) {
+			ftrace_record_suspend++;
+			ftrace_update_code();
+			usecs = nsecs_to_usecs(ftrace_update_time);
+			if (ftrace_update_tot_cnt > 100000) {
+				ftrace_update_tot_cnt = 0;
+				pr_info("hm, dftrace overflow: %lu change%s"
+					 " (%lu total) in %lu usec%s\n",
+					ftrace_update_cnt,
+					ftrace_update_cnt != 1 ? "s" : "",
+					ftrace_update_tot_cnt,
+					usecs, usecs != 1 ? "s" : "");
+				WARN_ON_ONCE(1);
+			}
+			ftraced_trigger = 0;
+			ftrace_record_suspend--;
+		}
+		mutex_unlock(&ftraced_lock);
+
+		ftrace_shutdown_replenish();
+
+		set_current_state(TASK_INTERRUPTIBLE);
+	}
+	__set_current_state(TASK_RUNNING);
+	return 0;
+}
+
+static int __init notrace ftrace_shutdown_init(void)
+{
+	struct task_struct *p;
+	int ret;
+
+	ret = ftrace_shutdown_arch_init();
+	if (ret)
+		return ret;
+
+	p = kthread_run(ftraced, NULL, "ftraced");
+	if (IS_ERR(p))
+		return -1;
+
+	__register_ftrace_function(&ftrace_shutdown_ops);
+
+	return 0;
+}
+
+core_initcall(ftrace_shutdown_init);
+#else
+# define ftrace_startup()	do { } while (0)
+# define ftrace_shutdown()	do { } while (0)
+#endif /* CONFIG_DYNAMIC_FTRACE */
+
 /**
- * clear_ftrace_function - reset the ftrace function
+ * register_ftrace_function - register a function for profiling
+ * @ops - ops structure that holds the function for profiling.
  *
- * This NULLs the ftrace function and in essence stops
- * tracing.  There may be lag
+ * Register a function to be called by all functions in the
+ * kernel.
+ *
+ * Note: @ops->func and all the functions it calls must be labeled
+ *       with "notrace", otherwise it will go into a
+ *       recursive loop.
  */
-void clear_ftrace_function(void)
+int register_ftrace_function(struct ftrace_ops *ops)
 {
-	ftrace_trace_function = ftrace_stub;
+	ftrace_startup();
+
+	return __register_ftrace_function(ops);
+}
+
+/**
+ * unregister_ftrace_function - unresgister a function for profiling.
+ * @ops - ops structure that holds the function to unregister
+ *
+ * Unregister a function that was added to be called by ftrace profiling.
+ */
+int unregister_ftrace_function(struct ftrace_ops *ops)
+{
+	int ret;
+
+	ret = __unregister_ftrace_function(ops);
+
+	if (ftrace_list == &ftrace_list_end)
+		ftrace_shutdown();
+
+	return ret;
 }