Merge tag 'trace-v5.13' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace

Pull tracing updates from Steven Rostedt:
 "New feature:

   - A new "func-no-repeats" option in tracefs/options directory.

     When set the function tracer will detect if the current function
     being traced is the same as the previous one, and instead of
     recording it, it will keep track of the number of times that the
     function is repeated in a row. And when another function is
     recorded, it will write a new event that shows the function that
     repeated, the number of times it repeated and the time stamp of
     when the last repeated function occurred.

  Enhancements:

   - In order to implement the above "func-no-repeats" option, the ring
     buffer timestamp can now give the accurate timestamp of the event
     as it is being recorded, instead of having to record an absolute
     timestamp for all events. This helps the histogram code which no
     longer needs to waste ring buffer space.

   - New validation logic to make sure all trace events that access
     dereferenced pointers do so in a safe way, and will warn otherwise.

  Fixes:

   - No longer limit the PIDs of tasks that are recorded for
     "saved_cmdlines" to PID_MAX_DEFAULT (32768), as systemd now allows
     for a much larger range. This caused the mapping of PIDs to the
     task names to be dropped for all tasks with a PID greater than
     32768.

   - Change trace_clock_global() to never block. This caused a deadlock.

  Clean ups:

   - Typos, prototype fixes, and removing of duplicate or unused code.

   - Better management of ftrace_page allocations"

* tag 'trace-v5.13' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace: (32 commits)
  tracing: Restructure trace_clock_global() to never block
  tracing: Map all PIDs to command lines
  ftrace: Reuse the output of the function tracer for func_repeats
  tracing: Add "func_no_repeats" option for function tracing
  tracing: Unify the logic for function tracing options
  tracing: Add method for recording "func_repeats" events
  tracing: Add "last_func_repeats" to struct trace_array
  tracing: Define new ftrace event "func_repeats"
  tracing: Define static void trace_print_time()
  ftrace: Simplify the calculation of page number for ftrace_page->records some more
  ftrace: Store the order of pages allocated in ftrace_page
  tracing: Remove unused argument from "ring_buffer_time_stamp()
  tracing: Remove duplicate struct declaration in trace_events.h
  tracing: Update create_system_filter() kernel-doc comment
  tracing: A minor cleanup for create_system_filter()
  kernel: trace: Mundane typo fixes in the file trace_events_filter.c
  tracing: Fix various typos in comments
  scripts/recordmcount.pl: Make vim and emacs indent the same
  scripts/recordmcount.pl: Make indent spacing consistent
  tracing: Add a verifier to check string pointers for trace events
  ...

1 files changed, 195 insertions, 28 deletions

diff --git a/kernel/trace/trace_functions.c b/kernel/trace/trace_functions.c
index f93723ca66bc..1f0e63f5d1f9 100644
--- a/kernel/trace/trace_functions.c
+++ b/kernel/trace/trace_functions.c
@@ -27,13 +27,28 @@ function_trace_call(unsigned long ip, unsigned long parent_ip,
 static void
 function_stack_trace_call(unsigned long ip, unsigned long parent_ip,
 			  struct ftrace_ops *op, struct ftrace_regs *fregs);
+static void
+function_no_repeats_trace_call(unsigned long ip, unsigned long parent_ip,
+			       struct ftrace_ops *op, struct ftrace_regs *fregs);
+static void
+function_stack_no_repeats_trace_call(unsigned long ip, unsigned long parent_ip,
+				     struct ftrace_ops *op,
+				     struct ftrace_regs *fregs);
 static struct tracer_flags func_flags;
 
 /* Our option */
 enum {
-	TRACE_FUNC_OPT_STACK	= 0x1,
+
+	TRACE_FUNC_NO_OPTS		= 0x0, /* No flags set. */
+	TRACE_FUNC_OPT_STACK		= 0x1,
+	TRACE_FUNC_OPT_NO_REPEATS	= 0x2,
+
+	/* Update this to next highest bit. */
+	TRACE_FUNC_OPT_HIGHEST_BIT	= 0x4
 };
 
+#define TRACE_FUNC_OPT_MASK	(TRACE_FUNC_OPT_HIGHEST_BIT - 1)
+
 int ftrace_allocate_ftrace_ops(struct trace_array *tr)
 {
 	struct ftrace_ops *ops;
@@ -86,6 +101,34 @@ void ftrace_destroy_function_files(struct trace_array *tr)
 	ftrace_free_ftrace_ops(tr);
 }
 
+static ftrace_func_t select_trace_function(u32 flags_val)
+{
+	switch (flags_val & TRACE_FUNC_OPT_MASK) {
+	case TRACE_FUNC_NO_OPTS:
+		return function_trace_call;
+	case TRACE_FUNC_OPT_STACK:
+		return function_stack_trace_call;
+	case TRACE_FUNC_OPT_NO_REPEATS:
+		return function_no_repeats_trace_call;
+	case TRACE_FUNC_OPT_STACK | TRACE_FUNC_OPT_NO_REPEATS:
+		return function_stack_no_repeats_trace_call;
+	default:
+		return NULL;
+	}
+}
+
+static bool handle_func_repeats(struct trace_array *tr, u32 flags_val)
+{
+	if (!tr->last_func_repeats &&
+	    (flags_val & TRACE_FUNC_OPT_NO_REPEATS)) {
+		tr->last_func_repeats = alloc_percpu(struct trace_func_repeats);
+		if (!tr->last_func_repeats)
+			return false;
+	}
+
+	return true;
+}
+
 static int function_trace_init(struct trace_array *tr)
 {
 	ftrace_func_t func;
@@ -97,12 +140,12 @@ static int function_trace_init(struct trace_array *tr)
 	if (!tr->ops)
 		return -ENOMEM;
 
-	/* Currently only the global instance can do stack tracing */
-	if (tr->flags & TRACE_ARRAY_FL_GLOBAL &&
-	    func_flags.val & TRACE_FUNC_OPT_STACK)
-		func = function_stack_trace_call;
-	else
-		func = function_trace_call;
+	func = select_trace_function(func_flags.val);
+	if (!func)
+		return -EINVAL;
+
+	if (!handle_func_repeats(tr, func_flags.val))
+		return -ENOMEM;
 
 	ftrace_init_array_ops(tr, func);
 
@@ -205,15 +248,137 @@ function_stack_trace_call(unsigned long ip, unsigned long parent_ip,
 	local_irq_restore(flags);
 }
 
+static inline bool is_repeat_check(struct trace_array *tr,
+				   struct trace_func_repeats *last_info,
+				   unsigned long ip, unsigned long parent_ip)
+{
+	if (last_info->ip == ip &&
+	    last_info->parent_ip == parent_ip &&
+	    last_info->count < U16_MAX) {
+		last_info->ts_last_call =
+			ring_buffer_time_stamp(tr->array_buffer.buffer);
+		last_info->count++;
+		return true;
+	}
+
+	return false;
+}
+
+static inline void process_repeats(struct trace_array *tr,
+				   unsigned long ip, unsigned long parent_ip,
+				   struct trace_func_repeats *last_info,
+				   unsigned int trace_ctx)
+{
+	if (last_info->count) {
+		trace_last_func_repeats(tr, last_info, trace_ctx);
+		last_info->count = 0;
+	}
+
+	last_info->ip = ip;
+	last_info->parent_ip = parent_ip;
+}
+
+static void
+function_no_repeats_trace_call(unsigned long ip, unsigned long parent_ip,
+			       struct ftrace_ops *op,
+			       struct ftrace_regs *fregs)
+{
+	struct trace_func_repeats *last_info;
+	struct trace_array *tr = op->private;
+	struct trace_array_cpu *data;
+	unsigned int trace_ctx;
+	unsigned long flags;
+	int bit;
+	int cpu;
+
+	if (unlikely(!tr->function_enabled))
+		return;
+
+	bit = ftrace_test_recursion_trylock(ip, parent_ip);
+	if (bit < 0)
+		return;
+
+	preempt_disable_notrace();
+
+	cpu = smp_processor_id();
+	data = per_cpu_ptr(tr->array_buffer.data, cpu);
+	if (atomic_read(&data->disabled))
+		goto out;
+
+	/*
+	 * An interrupt may happen at any place here. But as far as I can see,
+	 * the only damage that this can cause is to mess up the repetition
+	 * counter without valuable data being lost.
+	 * TODO: think about a solution that is better than just hoping to be
+	 * lucky.
+	 */
+	last_info = per_cpu_ptr(tr->last_func_repeats, cpu);
+	if (is_repeat_check(tr, last_info, ip, parent_ip))
+		goto out;
+
+	local_save_flags(flags);
+	trace_ctx = tracing_gen_ctx_flags(flags);
+	process_repeats(tr, ip, parent_ip, last_info, trace_ctx);
+
+	trace_function(tr, ip, parent_ip, trace_ctx);
+
+out:
+	ftrace_test_recursion_unlock(bit);
+	preempt_enable_notrace();
+}
+
+static void
+function_stack_no_repeats_trace_call(unsigned long ip, unsigned long parent_ip,
+				     struct ftrace_ops *op,
+				     struct ftrace_regs *fregs)
+{
+	struct trace_func_repeats *last_info;
+	struct trace_array *tr = op->private;
+	struct trace_array_cpu *data;
+	unsigned long flags;
+	long disabled;
+	int cpu;
+	unsigned int trace_ctx;
+
+	if (unlikely(!tr->function_enabled))
+		return;
+
+	/*
+	 * Need to use raw, since this must be called before the
+	 * recursive protection is performed.
+	 */
+	local_irq_save(flags);
+	cpu = raw_smp_processor_id();
+	data = per_cpu_ptr(tr->array_buffer.data, cpu);
+	disabled = atomic_inc_return(&data->disabled);
+
+	if (likely(disabled == 1)) {
+		last_info = per_cpu_ptr(tr->last_func_repeats, cpu);
+		if (is_repeat_check(tr, last_info, ip, parent_ip))
+			goto out;
+
+		trace_ctx = tracing_gen_ctx_flags(flags);
+		process_repeats(tr, ip, parent_ip, last_info, trace_ctx);
+
+		trace_function(tr, ip, parent_ip, trace_ctx);
+		__trace_stack(tr, trace_ctx, STACK_SKIP);
+	}
+
+ out:
+	atomic_dec(&data->disabled);
+	local_irq_restore(flags);
+}
+
 static struct tracer_opt func_opts[] = {
 #ifdef CONFIG_STACKTRACE
 	{ TRACER_OPT(func_stack_trace, TRACE_FUNC_OPT_STACK) },
 #endif
+	{ TRACER_OPT(func-no-repeats, TRACE_FUNC_OPT_NO_REPEATS) },
 	{ } /* Always set a last empty entry */
 };
 
 static struct tracer_flags func_flags = {
-	.val = 0, /* By default: all flags disabled */
+	.val = TRACE_FUNC_NO_OPTS, /* By default: all flags disabled */
 	.opts = func_opts
 };
 
@@ -235,30 +400,32 @@ static struct tracer function_trace;
 static int
 func_set_flag(struct trace_array *tr, u32 old_flags, u32 bit, int set)
 {
-	switch (bit) {
-	case TRACE_FUNC_OPT_STACK:
-		/* do nothing if already set */
-		if (!!set == !!(func_flags.val & TRACE_FUNC_OPT_STACK))
-			break;
-
-		/* We can change this flag when not running. */
-		if (tr->current_trace != &function_trace)
-			break;
+	ftrace_func_t func;
+	u32 new_flags;
 
-		unregister_ftrace_function(tr->ops);
+	/* Do nothing if already set. */
+	if (!!set == !!(func_flags.val & bit))
+		return 0;
 
-		if (set) {
-			tr->ops->func = function_stack_trace_call;
-			register_ftrace_function(tr->ops);
-		} else {
-			tr->ops->func = function_trace_call;
-			register_ftrace_function(tr->ops);
-		}
+	/* We can change this flag only when not running. */
+	if (tr->current_trace != &function_trace)
+		return 0;
 
-		break;
-	default:
+	new_flags = (func_flags.val & ~bit) | (set ? bit : 0);
+	func = select_trace_function(new_flags);
+	if (!func)
 		return -EINVAL;
-	}
+
+	/* Check if there's anything to change. */
+	if (tr->ops->func == func)
+		return 0;
+
+	if (!handle_func_repeats(tr, new_flags))
+		return -ENOMEM;
+
+	unregister_ftrace_function(tr->ops);
+	tr->ops->func = func;
+	register_ftrace_function(tr->ops);
 
 	return 0;
 }