tracing: fixes, clean ups and optimizations

---
 kernel/trace/trace.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

kernel/trace/trace.c +2 -2

index 0cd500b..25b91af 100644
@@ -3537,14 +3537,14 @@ static const char readme_msg[] =
 	"\n  snapshot\t\t- Like 'trace' but shows the content of the static snapshot buffer\n"
 	"\t\t\t  Read the contents for more information\n"
 #endif
-#ifdef CONFIG_STACKTRACE
+#ifdef CONFIG_STACK_TRACER
 	"  stack_trace\t\t- Shows the max stack trace when active\n"
 	"  stack_max_size\t- Shows current max stack size that was traced\n"
 	"\t\t\t  Write into this file to reset the max size (trigger a new trace)\n"
 #ifdef CONFIG_DYNAMIC_FTRACE
 	"  stack_trace_filter\t- Like set_ftrace_filter but limits what stack_trace traces\n"
 #endif
-#endif /* CONFIG_STACKTRACE */
+#endif /* CONFIG_STACK_TRACER */
 ;
 
 static ssize_t

Signature · 1 line

1.7.10.4

Steven Rostedt

author maintainer

2013-07-22 14:21 UTC #2 lore raw

PATCH 2/14 tracing: Use trace_seq_puts()/trace_seq_putc() where possible

From: "zhangwei(Jovi)" <[email protected]>

For string without format specifiers, use trace_seq_puts() or trace_seq_putc().

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by zhangwei(Jovi) <[email protected]>

[ fixed a trace_seq_putc(s, " ") to trace_seq_putc(s, ' ') ]

---
 kernel/trace/ring_buffer.c           |   10 +++----
 kernel/trace/trace_events_filter.c   |    4 +--
 kernel/trace/trace_functions_graph.c |   52 +++++++++++++++++-----------------
 kernel/trace/trace_mmiotrace.c       |    8 +++---
 kernel/trace/trace_output.c          |   14 ++++-----
 kernel/trace/trace_syscalls.c        |    2 +-
 6 files changed, 45 insertions(+), 45 deletions(-)

kernel/trace/ring_buffer.c +5 -5

index e444ff8..eef2e56 100644
@@ -36,11 +36,11 @@ int ring_buffer_print_entry_header(struct trace_seq *s)
 {
 	int ret;
 
-	ret = trace_seq_printf(s, "# compressed entry header\n");
-	ret = trace_seq_printf(s, "\ttype_len    :    5 bits\n");
-	ret = trace_seq_printf(s, "\ttime_delta  :   27 bits\n");
-	ret = trace_seq_printf(s, "\tarray       :   32 bits\n");
-	ret = trace_seq_printf(s, "\n");
+	ret = trace_seq_puts(s, "# compressed entry header\n");
+	ret = trace_seq_puts(s, "\ttype_len    :    5 bits\n");
+	ret = trace_seq_puts(s, "\ttime_delta  :   27 bits\n");
+	ret = trace_seq_puts(s, "\tarray       :   32 bits\n");
+	ret = trace_seq_putc(s, '\n');
 	ret = trace_seq_printf(s, "\tpadding     : type == %d\n",
 			       RINGBUF_TYPE_PADDING);
 	ret = trace_seq_printf(s, "\ttime_extend : type == %d\n",

kernel/trace/trace_events_filter.c +2 -2

index 0d883dc..0c7b75a 100644
@@ -646,7 +646,7 @@ void print_event_filter(struct ftrace_event_call *call, struct trace_seq *s)
 	if (filter && filter->filter_string)
 		trace_seq_printf(s, "%s\n", filter->filter_string);
 	else
-		trace_seq_printf(s, "none\n");
+		trace_seq_puts(s, "none\n");
 	mutex_unlock(&event_mutex);
 }

@@ -660,7 +660,7 @@ void print_subsystem_event_filter(struct event_subsystem *system,
 	if (filter && filter->filter_string)
 		trace_seq_printf(s, "%s\n", filter->filter_string);
 	else
-		trace_seq_printf(s, DEFAULT_SYS_FILTER_MESSAGE "\n");
+		trace_seq_puts(s, DEFAULT_SYS_FILTER_MESSAGE "\n");
 	mutex_unlock(&event_mutex);
 }

kernel/trace/trace_functions_graph.c +26 ~~-26~~

189 lines not on this page · raw message

kernel/trace/trace_mmiotrace.c +4 -4

index a5e8f48..b3dcfb2 100644
@@ -90,7 +90,7 @@ static int mmio_print_pcidev(struct trace_seq *s, const struct pci_dev *dev)
 	if (drv)
 		ret += trace_seq_printf(s, " %s\n", drv->name);
 	else
-		ret += trace_seq_printf(s, " \n");
+		ret += trace_seq_puts(s, " \n");
 	return ret;
 }

@@ -107,7 +107,7 @@ static void mmio_pipe_open(struct trace_iterator *iter)
 	struct header_iter *hiter;
 	struct trace_seq *s = &iter->seq;
 
-	trace_seq_printf(s, "VERSION 20070824\n");
+	trace_seq_puts(s, "VERSION 20070824\n");
 
 	hiter = kzalloc(sizeof(*hiter), GFP_KERNEL);
 	if (!hiter)

@@ -209,7 +209,7 @@ static enum print_line_t mmio_print_rw(struct trace_iterator *iter)
 			(rw->value >> 0) & 0xff, rw->pc, 0);
 		break;
 	default:
-		ret = trace_seq_printf(s, "rw what?\n");
+		ret = trace_seq_puts(s, "rw what?\n");
 		break;
 	}
 	if (ret)

@@ -245,7 +245,7 @@ static enum print_line_t mmio_print_map(struct trace_iterator *iter)
 			secs, usec_rem, m->map_id, 0UL, 0);
 		break;
 	default:
-		ret = trace_seq_printf(s, "map what?\n");
+		ret = trace_seq_puts(s, "map what?\n");
 		break;
 	}
 	if (ret)

kernel/trace/trace_output.c +7 -7

index bb922d9..34e7cba 100644
@@ -78,7 +78,7 @@ enum print_line_t trace_print_printk_msg_only(struct trace_iterator *iter)
 
 	trace_assign_type(field, entry);
 
-	ret = trace_seq_printf(s, "%s", field->buf);
+	ret = trace_seq_puts(s, field->buf);
 	if (!ret)
 		return TRACE_TYPE_PARTIAL_LINE;

@@ -558,14 +558,14 @@ seq_print_userip_objs(const struct userstack_entry *entry, struct trace_seq *s,
 			if (ret)
 				ret = trace_seq_puts(s, "??");
 			if (ret)
-				ret = trace_seq_puts(s, "\n");
+				ret = trace_seq_putc(s, '\n');
 			continue;
 		}
 		if (!ret)
 			break;
 		if (ret)
 			ret = seq_print_user_ip(s, mm, ip, sym_flags);
-		ret = trace_seq_puts(s, "\n");
+		ret = trace_seq_putc(s, '\n');
 	}
 
 	if (mm)

@@ -579,7 +579,7 @@ seq_print_ip_sym(struct trace_seq *s, unsigned long ip, unsigned long sym_flags)
 	int ret;
 
 	if (!ip)
-		return trace_seq_printf(s, "0");
+		return trace_seq_putc(s, '0');
 
 	if (sym_flags & TRACE_ITER_SYM_OFFSET)
 		ret = seq_print_sym_offset(s, "%s", ip);

@@ -964,14 +964,14 @@ static enum print_line_t trace_fn_trace(struct trace_iterator *iter, int flags,
 		goto partial;
 
 	if ((flags & TRACE_ITER_PRINT_PARENT) && field->parent_ip) {
-		if (!trace_seq_printf(s, " <-"))
+		if (!trace_seq_puts(s, " <-"))
 			goto partial;
 		if (!seq_print_ip_sym(s,
 				      field->parent_ip,
 				      flags))
 			goto partial;
 	}
-	if (!trace_seq_printf(s, "\n"))
+	if (!trace_seq_putc(s, '\n'))
 		goto partial;
 
 	return TRACE_TYPE_HANDLED;

@@ -1210,7 +1210,7 @@ static enum print_line_t trace_stack_print(struct trace_iterator *iter,
 
 		if (!seq_print_ip_sym(s, *p, flags))
 			goto partial;
-		if (!trace_seq_puts(s, "\n"))
+		if (!trace_seq_putc(s, '\n'))
 			goto partial;
 	}

kernel/trace/trace_syscalls.c +1 -1

index 322e164..0611562 100644
@@ -175,7 +175,7 @@ print_syscall_exit(struct trace_iterator *iter, int flags,
 	entry = syscall_nr_to_meta(syscall);
 
 	if (!entry) {
-		trace_seq_printf(s, "\n");
+		trace_seq_putc(s, '\n');
 		return TRACE_TYPE_HANDLED;
 	}

Signature · 1 line

1.7.10.4

Steven Rostedt

author maintainer

2013-07-22 14:21 UTC #3 lore raw

PATCH 3/14 tracing: Typo fix on ring buffer comments

From: "zhangwei(Jovi)" <[email protected]>

There have some mismatch between comments with
real function name, update it.

This patch also add some missed function arguments description.

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by zhangwei(Jovi) <[email protected]>

---
 kernel/trace/ring_buffer.c |   16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)

kernel/trace/ring_buffer.c +9 -7

index eef2e56..cc2f66f 100644
@@ -1066,7 +1066,7 @@ static int rb_check_list(struct ring_buffer_per_cpu *cpu_buffer,
 }
 
 /**
- * check_pages - integrity check of buffer pages
+ * rb_check_pages - integrity check of buffer pages
  * @cpu_buffer: CPU buffer with pages to test
  *
  * As a safety measure we check to make sure the data pages have not

@@ -1258,7 +1258,7 @@ static int rb_cpu_notify(struct notifier_block *self,
 #endif
 
 /**
- * ring_buffer_alloc - allocate a new ring_buffer
+ * __ring_buffer_alloc - allocate a new ring_buffer
  * @size: the size in bytes per cpu that is needed.
  * @flags: attributes to set for the ring buffer.
  *

@@ -1607,6 +1607,7 @@ static void update_pages_handler(struct work_struct *work)
  * ring_buffer_resize - resize the ring buffer
  * @buffer: the buffer to resize.
  * @size: the new size.
+ * @cpu_id: the cpu buffer to resize
  *
  * Minimum size is 2 * BUF_PAGE_SIZE.
  *

@@ -3956,11 +3957,11 @@ EXPORT_SYMBOL_GPL(ring_buffer_consume);
  * expected.
  *
  * After a sequence of ring_buffer_read_prepare calls, the user is
- * expected to make at least one call to ring_buffer_prepare_sync.
+ * expected to make at least one call to ring_buffer_read_prepare_sync.
  * Afterwards, ring_buffer_read_start is invoked to get things going
  * for real.
  *
- * This overall must be paired with ring_buffer_finish.
+ * This overall must be paired with ring_buffer_read_finish.
  */
 struct ring_buffer_iter *
 ring_buffer_read_prepare(struct ring_buffer *buffer, int cpu)

@@ -4009,7 +4010,7 @@ EXPORT_SYMBOL_GPL(ring_buffer_read_prepare_sync);
  * an intervening ring_buffer_read_prepare_sync must have been
  * performed.
  *
- * Must be paired with ring_buffer_finish.
+ * Must be paired with ring_buffer_read_finish.
  */
 void
 ring_buffer_read_start(struct ring_buffer_iter *iter)

@@ -4031,7 +4032,7 @@ ring_buffer_read_start(struct ring_buffer_iter *iter)
 EXPORT_SYMBOL_GPL(ring_buffer_read_start);
 
 /**
- * ring_buffer_finish - finish reading the iterator of the buffer
+ * ring_buffer_read_finish - finish reading the iterator of the buffer
  * @iter: The iterator retrieved by ring_buffer_start
  *
  * This re-enables the recording to the buffer, and frees the

@@ -4346,6 +4347,7 @@ EXPORT_SYMBOL_GPL(ring_buffer_swap_cpu);
 /**
  * ring_buffer_alloc_read_page - allocate a page to read from buffer
  * @buffer: the buffer to allocate for.
+ * @cpu: the cpu buffer to allocate.
  *
  * This function is used in conjunction with ring_buffer_read_page.
  * When reading a full page from the ring buffer, these functions

@@ -4403,7 +4405,7 @@ EXPORT_SYMBOL_GPL(ring_buffer_free_read_page);
  * to swap with a page in the ring buffer.
  *
  * for example:
- *	rpage = ring_buffer_alloc_read_page(buffer);
+ *	rpage = ring_buffer_alloc_read_page(buffer, cpu);
  *	if (!rpage)
  *		return error;
  *	ret = ring_buffer_read_page(buffer, &rpage, len, cpu, 0);

Signature · 1 line

1.7.10.4

Steven Rostedt

author maintainer

2013-07-22 14:21 UTC #4 lore raw

PATCH 4/14 tracing/function: Avoid perf_trace_buf_*() if event_function.perf_events is empty

From: Oleg Nesterov <[email protected]>

perf_trace_buf_prepare() + perf_trace_buf_submit(head, task => NULL) make no sense if hlist_empty(head). Change perf_ftrace_function_call() to check event_function.perf_events beforehand.

Link http://lkml.kernel.org/r/[email protected] open here

Acked-by Peter Zijlstra <[email protected]>

---
 kernel/trace/trace_event_perf.c |    6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

kernel/trace/trace_event_perf.c +4 -2

index 84b1e04..12df557 100644
@@ -266,6 +266,10 @@ perf_ftrace_function_call(unsigned long ip, unsigned long parent_ip,
 	struct pt_regs regs;
 	int rctx;
 
+	head = this_cpu_ptr(event_function.perf_events);
+	if (hlist_empty(head))
+		return;
+
 #define ENTRY_SIZE (ALIGN(sizeof(struct ftrace_entry) + sizeof(u32), \
 		    sizeof(u64)) - sizeof(u32))

@@ -279,8 +283,6 @@ perf_ftrace_function_call(unsigned long ip, unsigned long parent_ip,
 
 	entry->ip = ip;
 	entry->parent_ip = parent_ip;
-
-	head = this_cpu_ptr(event_function.perf_events);
 	perf_trace_buf_submit(entry, ENTRY_SIZE, rctx, 0,
 			      1, &regs, head, NULL);

Signature · 1 line

1.7.10.4

Steven Rostedt

author maintainer

2013-07-22 14:21 UTC #5 lore raw

PATCH 5/14 tracing/syscall: Avoid perf_trace_buf_*() if sys_data->perf_events is empty

From: Oleg Nesterov <[email protected]>

perf_trace_buf_prepare() + perf_trace_buf_submit(head, task => NULL) make no sense if hlist_empty(head). Change perf_syscall_enter/exit() to check sys_data->{enter,exit}_event->perf_events beforehand.

Link http://lkml.kernel.org/r/[email protected] open here

Acked-by Peter Zijlstra <[email protected]>

---
 kernel/trace/trace_syscalls.c |   12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)

kernel/trace/trace_syscalls.c +8 -4

index 0611562..ac00857 100644
@@ -566,6 +566,10 @@ static void perf_syscall_enter(void *ignore, struct pt_regs *regs, long id)
 	if (!sys_data)
 		return;
 
+	head = this_cpu_ptr(sys_data->enter_event->perf_events);
+	if (hlist_empty(head))
+		return;
+
 	/* get the size after alignment with the u32 buffer size field */
 	size = sizeof(unsigned long) * sys_data->nb_args + sizeof(*rec);
 	size = ALIGN(size + sizeof(u32), sizeof(u64));

@@ -583,8 +587,6 @@ static void perf_syscall_enter(void *ignore, struct pt_regs *regs, long id)
 	rec->nr = syscall_nr;
 	syscall_get_arguments(current, regs, 0, sys_data->nb_args,
 			       (unsigned long *)&rec->args);
-
-	head = this_cpu_ptr(sys_data->enter_event->perf_events);
 	perf_trace_buf_submit(rec, size, rctx, 0, 1, regs, head, NULL);
 }

@@ -642,6 +644,10 @@ static void perf_syscall_exit(void *ignore, struct pt_regs *regs, long ret)
 	if (!sys_data)
 		return;
 
+	head = this_cpu_ptr(sys_data->exit_event->perf_events);
+	if (hlist_empty(head))
+		return;
+
 	/* We can probably do that at build time */
 	size = ALIGN(sizeof(*rec) + sizeof(u32), sizeof(u64));
 	size -= sizeof(u32);

@@ -661,8 +667,6 @@ static void perf_syscall_exit(void *ignore, struct pt_regs *regs, long ret)
 
 	rec->nr = syscall_nr;
 	rec->ret = syscall_get_return_value(current, regs);
-
-	head = this_cpu_ptr(sys_data->exit_event->perf_events);
 	perf_trace_buf_submit(rec, size, rctx, 0, 1, regs, head, NULL);
 }

Signature · 1 line

1.7.10.4

Steven Rostedt

author maintainer

2013-07-22 14:21 UTC #6 lore raw

PATCH 6/14 tracing/perf: Move the PERF_MAX_TRACE_SIZE check into perf_trace_buf_prepare()

From: Oleg Nesterov <[email protected]>

Every perf_trace_buf_prepare() caller does WARN_ONCE(size > PERF_MAX_TRACE_SIZE, message) and "message" is almost the same.

Shift this WARN_ONCE() into perf_trace_buf_prepare(). This changes the meaning of _ONCE, but I think this is fine.

	- 4947014 2932448 10104832  17984294  1126b26 vmlinux
	+ 4948422 2932448 10104832  17985702  11270a6 vmlinux

on my build.

Link http://lkml.kernel.org/r/[email protected] open here

Acked-by Peter Zijlstra <[email protected]>

---
 include/trace/ftrace.h          |    4 ----
 kernel/trace/trace_event_perf.c |    4 ++++
 kernel/trace/trace_kprobe.c     |    6 ------
 kernel/trace/trace_syscalls.c   |   12 ------------
 kernel/trace/trace_uprobe.c     |    2 --
 5 files changed, 4 insertions(+), 24 deletions(-)

include/trace/ftrace.h +0 -4

index d615f78..41a6643 100644
@@ -670,10 +670,6 @@ perf_trace_##call(void *__data, proto)					\
 			     sizeof(u64));				\
 	__entry_size -= sizeof(u32);					\
 									\
-	if (WARN_ONCE(__entry_size > PERF_MAX_TRACE_SIZE,		\
-		      "profile buffer not large enough"))		\
-		return;							\
-									\
 	entry = (struct ftrace_raw_##call *)perf_trace_buf_prepare(	\
 		__entry_size, event_call->event.type, &__regs, &rctx);	\
 	if (!entry)							\

kernel/trace/trace_event_perf.c +4 -0

index 12df557..80c36bc 100644
@@ -236,6 +236,10 @@ __kprobes void *perf_trace_buf_prepare(int size, unsigned short type,
 
 	BUILD_BUG_ON(PERF_MAX_TRACE_SIZE % sizeof(unsigned long));
 
+	if (WARN_ONCE(size > PERF_MAX_TRACE_SIZE,
+			"perf buffer not large enough"))
+		return NULL;
+
 	pc = preempt_count();
 
 	*rctxp = perf_swevent_get_recursion_context();

kernel/trace/trace_kprobe.c +0 -6

index 7ed6976..ae6ce83 100644
@@ -1087,9 +1087,6 @@ kprobe_perf_func(struct trace_probe *tp, struct pt_regs *regs)
 	__size = sizeof(*entry) + tp->size + dsize;
 	size = ALIGN(__size + sizeof(u32), sizeof(u64));
 	size -= sizeof(u32);
-	if (WARN_ONCE(size > PERF_MAX_TRACE_SIZE,
-		     "profile buffer not large enough"))
-		return;
 
 	entry = perf_trace_buf_prepare(size, call->event.type, regs, &rctx);
 	if (!entry)

@@ -1120,9 +1117,6 @@ kretprobe_perf_func(struct trace_probe *tp, struct kretprobe_instance *ri,
 	__size = sizeof(*entry) + tp->size + dsize;
 	size = ALIGN(__size + sizeof(u32), sizeof(u64));
 	size -= sizeof(u32);
-	if (WARN_ONCE(size > PERF_MAX_TRACE_SIZE,
-		     "profile buffer not large enough"))
-		return;
 
 	entry = perf_trace_buf_prepare(size, call->event.type, regs, &rctx);
 	if (!entry)

kernel/trace/trace_syscalls.c +0 ~~-12~~

index ac00857..8fd0365 100644
@@ -575,10 +575,6 @@ static void perf_syscall_enter(void *ignore, struct pt_regs *regs, long id)
 	size = ALIGN(size + sizeof(u32), sizeof(u64));
 	size -= sizeof(u32);
 
-	if (WARN_ONCE(size > PERF_MAX_TRACE_SIZE,
-		      "perf buffer not large enough"))
-		return;
-
 	rec = (struct syscall_trace_enter *)perf_trace_buf_prepare(size,
 				sys_data->enter_event->event.type, regs, &rctx);
 	if (!rec)

@@ -652,14 +648,6 @@ static void perf_syscall_exit(void *ignore, struct pt_regs *regs, long ret)
 	size = ALIGN(sizeof(*rec) + sizeof(u32), sizeof(u64));
 	size -= sizeof(u32);
 
-	/*
-	 * Impossible, but be paranoid with the future
-	 * How to put this check outside runtime?
-	 */
-	if (WARN_ONCE(size > PERF_MAX_TRACE_SIZE,
-		"exit event has grown above perf buffer size"))
-		return;
-
 	rec = (struct syscall_trace_exit *)perf_trace_buf_prepare(size,
 				sys_data->exit_event->event.type, regs, &rctx);
 	if (!rec)

kernel/trace/trace_uprobe.c +0 -2

index d5d0cd3..a23d2d7 100644
@@ -818,8 +818,6 @@ static void uprobe_perf_print(struct trace_uprobe *tu,
 
 	size = SIZEOF_TRACE_ENTRY(is_ret_probe(tu));
 	size = ALIGN(size + tu->size + sizeof(u32), sizeof(u64)) - sizeof(u32);
-	if (WARN_ONCE(size > PERF_MAX_TRACE_SIZE, "profile buffer not large enough"))
-		return;
 
 	preempt_disable();
 	head = this_cpu_ptr(call->perf_events);

Signature · 1 line

1.7.10.4

Steven Rostedt

author maintainer

2013-07-22 14:21 UTC #7 lore raw

PATCH 7/14 tracing/kprobe: Wait for disabling all running kprobe handlers

Link http://lkml.kernel.org/r/20130709093526.20138.93100.stgit@mhiramat-M0-7522 open here

From: Masami Hiramatsu <[email protected]>

Wait for disabling all running kprobe handlers when a kprobe event is disabled, since the caller, trace_remove_event_call() supposes that a removing event is disabled completely by disabling the event.
With this change, ftrace can ensure that there is no running event handlers after disabling it.

Signed-off-by Masami Hiramatsu <[email protected]>

---
 kernel/trace/trace_kprobe.c |   23 +++++++++++++++++------
 1 file changed, 17 insertions(+), 6 deletions(-)

kernel/trace/trace_kprobe.c +17 -6

index ae6ce83..3811487 100644
@@ -243,11 +243,11 @@ find_event_file_link(struct trace_probe *tp, struct ftrace_event_file *file)
 static int
 disable_trace_probe(struct trace_probe *tp, struct ftrace_event_file *file)
 {
+	struct event_file_link *link = NULL;
+	int wait = 0;
 	int ret = 0;
 
 	if (file) {
-		struct event_file_link *link;
-
 		link = find_event_file_link(tp, file);
 		if (!link) {
 			ret = -EINVAL;

@@ -255,10 +255,7 @@ disable_trace_probe(struct trace_probe *tp, struct ftrace_event_file *file)
 		}
 
 		list_del_rcu(&link->list);
-		/* synchronize with kprobe_trace_func/kretprobe_trace_func */
-		synchronize_sched();
-		kfree(link);
-
+		wait = 1;
 		if (!list_empty(&tp->files))
 			goto out;

@@ -271,8 +268,22 @@ disable_trace_probe(struct trace_probe *tp, struct ftrace_event_file *file)
 			disable_kretprobe(&tp->rp);
 		else
 			disable_kprobe(&tp->rp.kp);
+		wait = 1;
 	}
  out:
+	if (wait) {
+		/*
+		 * Synchronize with kprobe_trace_func/kretprobe_trace_func
+		 * to ensure disabled (all running handlers are finished).
+		 * This is not only for kfree(), but also the caller,
+		 * trace_remove_event_call() supposes it for releasing
+		 * event_call related objects, which will be accessed in
+		 * the kprobe_trace_func/kretprobe_trace_func.
+		 */
+		synchronize_sched();
+		kfree(link);	/* Ignored if link == NULL */
+	}
+
 	return ret;
 }

Signature · 1 line

1.7.10.4

Steven Rostedt

author maintainer

2013-07-22 14:19 UTC #8 lore raw

PATCH 8/14 tracing: Fix error handling to ensure instances can always be removed

From: Alexander Z Lam <[email protected]>

Remove debugfs directories for tracing instances during creation if an error occurs causing the trace_array for that instance to not be added to ftrace_trace_arrays. If the directory continues to exist after the error, it cannot be removed because the respective trace_array is not in ftrace_trace_arrays.

Link http://lkml.kernel.org/r/[email protected] open here

Cc [email protected] # 3.10

Cc Vaibhav Nagarnaik <[email protected]>

Cc David Sharp <[email protected]>

Cc Alexander Z Lam <[email protected]>

Signed-off-by Alexander Z Lam <[email protected]>

---
 kernel/trace/trace.c |    4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

kernel/trace/trace.c +3 -1

index 25b91af..7c3da7b 100644
@@ -5973,8 +5973,10 @@ static int new_instance_create(const char *name)
 		goto out_free_tr;
 
 	ret = event_trace_add_tracer(tr->dir, tr);
-	if (ret)
+	if (ret) {
+		debugfs_remove_recursive(tr->dir);
 		goto out_free_tr;
+	}
 
 	init_tracer_debugfs(tr, tr->dir);

Signature · 1 line

1.7.10.4

Steven Rostedt

author maintainer

2013-07-22 14:19 UTC #9 lore raw

PATCH 9/14 tracing: Miscellaneous fixes for trace_array ref counting

From: Alexander Z Lam <[email protected]>

Some error paths did not handle ref counting properly, and some trace files need ref counting.

Link http://lkml.kernel.org/r/[email protected] open here

Cc [email protected] # 3.10

Cc Vaibhav Nagarnaik <[email protected]>

Cc David Sharp <[email protected]>

Cc Alexander Z Lam <[email protected]>

Signed-off-by Alexander Z Lam <[email protected]>

---
 kernel/trace/trace.c        |   24 ++++++++++++++++++------
 kernel/trace/trace_events.c |   21 +++++++++++++++++++--
 2 files changed, 37 insertions(+), 8 deletions(-)

kernel/trace/trace.c +18 -6

index 7c3da7b..7d9ceab 100644
@@ -3008,7 +3008,6 @@ static int tracing_release(struct inode *inode, struct file *file)
 
 	iter = m->private;
 	tr = iter->tr;
-	trace_array_put(tr);
 
 	mutex_lock(&trace_types_lock);

@@ -3023,6 +3022,9 @@ static int tracing_release(struct inode *inode, struct file *file)
 	if (!iter->snapshot)
 		/* reenable tracing if it was previously enabled */
 		tracing_start_tr(tr);
+
+	__trace_array_put(tr);
+
 	mutex_unlock(&trace_types_lock);
 
 	mutex_destroy(&iter->mutex);

@@ -3447,6 +3449,7 @@ tracing_trace_options_write(struct file *filp, const char __user *ubuf,
 static int tracing_trace_options_open(struct inode *inode, struct file *file)
 {
 	struct trace_array *tr = inode->i_private;
+	int ret;
 
 	if (tracing_disabled)
 		return -ENODEV;

@@ -3454,7 +3457,11 @@ static int tracing_trace_options_open(struct inode *inode, struct file *file)
 	if (trace_array_get(tr) < 0)
 		return -ENODEV;
 
-	return single_open(file, tracing_trace_options_show, inode->i_private);
+	ret = single_open(file, tracing_trace_options_show, inode->i_private);
+	if (ret < 0)
+		trace_array_put(tr);
+
+	return ret;
 }
 
 static const struct file_operations tracing_iter_fops = {

@@ -3958,6 +3965,7 @@ static int tracing_open_pipe(struct inode *inode, struct file *filp)
 	iter = kzalloc(sizeof(*iter), GFP_KERNEL);
 	if (!iter) {
 		ret = -ENOMEM;
+		__trace_array_put(tr);
 		goto out;
 	}

@@ -4704,21 +4712,24 @@ static int tracing_snapshot_open(struct inode *inode, struct file *file)
 			ret = PTR_ERR(iter);
 	} else {
 		/* Writes still need the seq_file to hold the private data */
+		ret = -ENOMEM;
 		m = kzalloc(sizeof(*m), GFP_KERNEL);
 		if (!m)
-			return -ENOMEM;
+			goto out;
 		iter = kzalloc(sizeof(*iter), GFP_KERNEL);
 		if (!iter) {
 			kfree(m);
-			return -ENOMEM;
+			goto out;
 		}
+		ret = 0;
+
 		iter->tr = tr;
 		iter->trace_buffer = &tc->tr->max_buffer;
 		iter->cpu_file = tc->cpu;
 		m->private = iter;
 		file->private_data = m;
 	}
-
+out:
 	if (ret < 0)
 		trace_array_put(tr);

@@ -5328,9 +5339,10 @@ tracing_stats_read(struct file *filp, char __user *ubuf,
 }
 
 static const struct file_operations tracing_stats_fops = {
-	.open		= tracing_open_generic,
+	.open		= tracing_open_generic_tc,
 	.read		= tracing_stats_read,
 	.llseek		= generic_file_llseek,
+	.release	= tracing_release_generic_tc,
 };
 
 #ifdef CONFIG_DYNAMIC_FTRACE

kernel/trace/trace_events.c +19 -2

index 7d85429..7a75cb2 100644
@@ -1218,6 +1218,7 @@ show_header(struct file *filp, char __user *ubuf, size_t cnt, loff_t *ppos)
 
 static int ftrace_event_avail_open(struct inode *inode, struct file *file);
 static int ftrace_event_set_open(struct inode *inode, struct file *file);
+static int ftrace_event_release(struct inode *inode, struct file *file);
 
 static const struct seq_operations show_event_seq_ops = {
 	.start = t_start,

@@ -1245,7 +1246,7 @@ static const struct file_operations ftrace_set_event_fops = {
 	.read = seq_read,
 	.write = ftrace_event_write,
 	.llseek = seq_lseek,
-	.release = seq_release,
+	.release = ftrace_event_release,
 };
 
 static const struct file_operations ftrace_enable_fops = {

@@ -1323,6 +1324,15 @@ ftrace_event_open(struct inode *inode, struct file *file,
 	return ret;
 }
 
+static int ftrace_event_release(struct inode *inode, struct file *file)
+{
+	struct trace_array *tr = inode->i_private;
+
+	trace_array_put(tr);
+
+	return seq_release(inode, file);
+}
+
 static int
 ftrace_event_avail_open(struct inode *inode, struct file *file)
 {

@@ -1336,12 +1346,19 @@ ftrace_event_set_open(struct inode *inode, struct file *file)
 {
 	const struct seq_operations *seq_ops = &show_set_event_seq_ops;
 	struct trace_array *tr = inode->i_private;
+	int ret;
+
+	if (trace_array_get(tr) < 0)
+		return -ENODEV;
 
 	if ((file->f_mode & FMODE_WRITE) &&
 	    (file->f_flags & O_TRUNC))
 		ftrace_clear_events(tr);
 
-	return ftrace_event_open(inode, file, seq_ops);
+	ret = ftrace_event_open(inode, file, seq_ops);
+	if (ret < 0)
+		trace_array_put(tr);
+	return ret;
 }
 
 static struct event_subsystem *

Signature · 1 line

1.7.10.4

Steven Rostedt

author maintainer

2013-07-22 14:22 UTC #10 lore raw

PATCH 10/14 tracing: Add ref_data to function and fgraph tracer structs

From: "Steven Rostedt (Red Hat)" <[email protected]>

The selftest for function and function graph tracers are defined as __init, as they are only executed at boot up. The "tracer" structs that are associated to those tracers are not setup as __init as they are used after boot. To stop mismatch warnings, those structures need to be annotated with __ref_data.

Currently, the tracer structures are defined to __read_mostly, as they do not really change. But in the future they should be converted to consts, but that will take a little work because they have a "next" pointer that gets updated when they are registered. That will have to wait till the next major release.

Link http://lkml.kernel.org/r/[email protected] open here

Reported-by kbuild test robot <[email protected]>

Reported-by Chen Gang <[email protected]>

---
 kernel/trace/trace.h                 |    9 +++++++++
 kernel/trace/trace_functions.c       |    2 +-
 kernel/trace/trace_functions_graph.c |    2 +-
 3 files changed, 11 insertions(+), 2 deletions(-)

kernel/trace/trace.h +9 -0

index 4a4f6e1..57b7bb0 100644
@@ -680,6 +680,15 @@ extern int trace_selftest_startup_sched_switch(struct tracer *trace,
 					       struct trace_array *tr);
 extern int trace_selftest_startup_branch(struct tracer *trace,
 					 struct trace_array *tr);
+/*
+ * Tracer data references selftest functions that only occur
+ * on boot up. These can be __init functions. Thus, when selftests
+ * are enabled, then the tracers need to reference __init functions.
+ */
+#define __tracer_data		__refdata
+#else
+/* Tracers are seldom changed. Optimize when selftests are disabled. */
+#define __tracer_data		__read_mostly
 #endif /* CONFIG_FTRACE_STARTUP_TEST */
 
 extern void *head_page(struct trace_array_cpu *data);

kernel/trace/trace_functions.c +1 -1

index b863f93..38fe148 100644
@@ -199,7 +199,7 @@ static int func_set_flag(u32 old_flags, u32 bit, int set)
 	return 0;
 }
 
-static struct tracer function_trace __read_mostly =
+static struct tracer function_trace __tracer_data =
 {
 	.name		= "function",
 	.init		= function_trace_init,

kernel/trace/trace_functions_graph.c +1 -1

index d56ae9b..b5c0924 100644
@@ -1448,7 +1448,7 @@ static struct trace_event graph_trace_ret_event = {
 	.funcs		= &graph_functions
 };
 
-static struct tracer graph_trace __read_mostly = {
+static struct tracer graph_trace __tracer_data = {
 	.name		= "function_graph",
 	.open		= graph_trace_open,
 	.pipe_open	= graph_trace_open,

Signature · 1 line

1.7.10.4

Steven Rostedt

author maintainer

2013-07-22 14:21 UTC #11 lore raw

PATCH 11/14 tracing: Simplify the iteration logic in f_start/f_next

From: Oleg Nesterov <[email protected]>

f_next() looks overcomplicated, and it is not strictly correct even if this doesn't matter.

Say, FORMAT_FIELD_SEPERATOR should not return NULL (means EOF) if trace_get_fields() returns an empty list, we should simply advance to FORMAT_PRINTFMT as we do when we find the end of list.

1. Change f_next() to return "struct list_head *" rather than "ftrace_event_field *", and change f_show() to do list_entry().

This simplifies the code a bit, only f_show() needs to know about ftrace_event_field, and f_next() can play with ->prev directly

2. Change f_next() to not play with ->prev / return inside the switch() statement. It can simply set node = head/common_head, the prev-or-advance-to-the-next-magic below does all work.

While at it. f_start() looks overcomplicated too. I don't think *pos == 0 makes sense as a separate case, just change this code to do "while" instead of "do/while".

The patch also moves f_start() down, close to f_stop(). This is purely cosmetic, just to make the locking added by the next patch more clear/visible.

Link http://lkml.kernel.org/r/[email protected] open here

---
 kernel/trace/trace_events.c |   60 ++++++++++++++++---------------------------
 1 file changed, 22 insertions(+), 38 deletions(-)

kernel/trace/trace_events.c +22 ~~-38~~

index 7a75cb2..76defd9 100644
@@ -826,59 +826,33 @@ enum {
 static void *f_next(struct seq_file *m, void *v, loff_t *pos)
 {
 	struct ftrace_event_call *call = m->private;
-	struct ftrace_event_field *field;
 	struct list_head *common_head = &ftrace_common_fields;
 	struct list_head *head = trace_get_fields(call);
+	struct list_head *node = v;
 
 	(*pos)++;
 
 	switch ((unsigned long)v) {
 	case FORMAT_HEADER:
-		if (unlikely(list_empty(common_head)))
-			return NULL;
-
-		field = list_entry(common_head->prev,
-				   struct ftrace_event_field, link);
-		return field;
+		node = common_head;
+		break;
 
 	case FORMAT_FIELD_SEPERATOR:
-		if (unlikely(list_empty(head)))
-			return NULL;
-
-		field = list_entry(head->prev, struct ftrace_event_field, link);
-		return field;
+		node = head;
+		break;
 
 	case FORMAT_PRINTFMT:
 		/* all done */
 		return NULL;
 	}
 
-	field = v;
-	if (field->link.prev == common_head)
+	node = node->prev;
+	if (node == common_head)
 		return (void *)FORMAT_FIELD_SEPERATOR;
-	else if (field->link.prev == head)
+	else if (node == head)
 		return (void *)FORMAT_PRINTFMT;
-
-	field = list_entry(field->link.prev, struct ftrace_event_field, link);
-
-	return field;
-}
-
-static void *f_start(struct seq_file *m, loff_t *pos)
-{
-	loff_t l = 0;
-	void *p;
-
-	/* Start by showing the header */
-	if (!*pos)
-		return (void *)FORMAT_HEADER;
-
-	p = (void *)FORMAT_HEADER;
-	do {
-		p = f_next(m, p, &l);
-	} while (p && l < *pos);
-
-	return p;
+	else
+		return node;
 }
 
 static int f_show(struct seq_file *m, void *v)

@@ -904,8 +878,7 @@ static int f_show(struct seq_file *m, void *v)
 		return 0;
 	}
 
-	field = v;
-
+	field = list_entry(v, struct ftrace_event_field, link);
 	/*
 	 * Smartly shows the array type(except dynamic array).
 	 * Normal:

@@ -932,6 +905,17 @@ static int f_show(struct seq_file *m, void *v)
 	return 0;
 }
 
+static void *f_start(struct seq_file *m, loff_t *pos)
+{
+	void *p = (void *)FORMAT_HEADER;
+	loff_t l = 0;
+
+	while (l < *pos && p)
+		p = f_next(m, p, &l);
+
+	return p;
+}
+
 static void f_stop(struct seq_file *m, void *p)
 {
 }

Signature · 1 line

1.7.10.4

Steven Rostedt

author maintainer

2013-07-22 14:22 UTC #12 lore raw

PATCH 12/14 tracing: Do not (ab)use trace_seq in event_id_read()

From: Oleg Nesterov <[email protected]>

event_id_read() has no reason to kmalloc "struct trace_seq" (more than PAGE_SIZE!), it can use a small buffer instead.

Note: "if (*ppos) return 0" looks strange and even wrong, simple_read_from_buffer() handles ppos != 0 case corrrectly.

And it seems that almost every user of trace_seq in this file should be converted too. Unless you use seq_open(), trace_seq buys nothing compared to the raw buffer, but it needs a bit more memory and code.

Link http://lkml.kernel.org/r/[email protected] open here

---
 kernel/trace/trace_events.c |   17 ++++-------------
 1 file changed, 4 insertions(+), 13 deletions(-)

kernel/trace/trace_events.c +4 ~~-13~~

index 76defd9..898f868 100644
@@ -947,23 +947,14 @@ static ssize_t
 event_id_read(struct file *filp, char __user *ubuf, size_t cnt, loff_t *ppos)
 {
 	struct ftrace_event_call *call = filp->private_data;
-	struct trace_seq *s;
-	int r;
+	char buf[32];
+	int len;
 
 	if (*ppos)
 		return 0;
 
-	s = kmalloc(sizeof(*s), GFP_KERNEL);
-	if (!s)
-		return -ENOMEM;
-
-	trace_seq_init(s);
-	trace_seq_printf(s, "%d\n", call->event.type);
-
-	r = simple_read_from_buffer(ubuf, cnt, ppos,
-				    s->buffer, s->len);
-	kfree(s);
-	return r;
+	len = sprintf(buf, "%d\n", call->event.type);
+	return simple_read_from_buffer(ubuf, cnt, ppos, buf, len);
 }
 
 static ssize_t

Signature · 1 line

1.7.10.4

Steven Rostedt

author maintainer

2013-07-22 14:22 UTC #13 lore raw

PATCH 13/14 tracing: Kill trace_array->waiter

From: Oleg Nesterov <[email protected]>

Trivial. trace_array->waiter has no users since 6eaaa5d5 "tracing/core: use appropriate waiting on trace_pipe".

Link http://lkml.kernel.org/r/[email protected] open here

---
 kernel/trace/trace.h |    1 -
 1 file changed, 1 deletion(-)

kernel/trace/trace.h +0 -1

index 57b7bb0..e7d643b 100644
@@ -214,7 +214,6 @@ struct trace_array {
 	struct dentry		*event_dir;
 	struct list_head	systems;
 	struct list_head	events;
-	struct task_struct	*waiter;
 	int			ref;
 };

Signature · 1 line

1.7.10.4

Steven Rostedt

author maintainer

2013-07-22 14:20 UTC #14 lore raw

PATCH 14/14 tracing: Kill the unbalanced tr->ref++ in tracing_buffers_open()

From: Oleg Nesterov <[email protected]>

tracing_buffers_open() does trace_array_get() and then it wrongly inrcements tr->ref again under trace_types_lock. This means that every caller leaks trace_array:

	# cd /sys/kernel/debug/tracing/
	# mkdir instances/X
	# true < instances/X/per_cpu/cpu0/trace_pipe_raw
	# rmdir instances/X
	rmdir: failed to remove `instances/X': Device or resource busy

Link http://lkml.kernel.org/r/[email protected] open here

Cc Ingo Molnar <[email protected]>

Cc Frederic Weisbecker <[email protected]>

Cc Masami Hiramatsu <[email protected]>

Cc [email protected] # 3.10