Skip to contentSkip to conversation

tracing: seqfile fixes

PATCH In review Li Zefan 2 participants 19 messages Jun 19, 2009 – Jun 23, 2009

Subsystems (2)
lore
Share & tools
lore

Li Zefan

author
#0 lore raw

cover tracing: seqfile fixes

While testing syscall tracepoints proposed by Jason, I found some entries were missing when reading available_events.

It turned out there's bug in seqfile handling. The bug is, it's wrong to increment @pos in seq start().

I fixed similar bugs in some other places. (the last patch fixes a different seqfile bug)

[PATCH 1/5] tracing/events: don't increment @pos in s_start() [PATCH 2/5] tracing_bprintk: don't increment @pos in t_start() [PATCH 3/5] trace_stat: don't increment @pos in stat_seq_start() [PATCH 4/5] ftrace: don't increment @pos in g_start() [PATCH 5/5] tracing: reset iterator in t_start()

---
 kernel/trace/ftrace.c       |   28 +++++++++++++---------------
 kernel/trace/trace.c        |   18 ++++--------------
 kernel/trace/trace_events.c |   28 ++++++++++++++++++++++------
 kernel/trace/trace_printk.c |   26 ++++++--------------------
 kernel/trace/trace_stat.c   |    6 +-----
 5 files changed, 46 insertions(+), 60 deletions(-)
author
#1 lore raw

PATCH 1/5 tracing/events: don't increment @pos in s_start()

While testing syscall tracepoints posted by Jason, I found 3 entries were missing when reading available_events. The output size of available_events is < 4 pages, which means we lost 1 item per page.

The cause is, it's wrong to increment @pos in s_start().

Actually there's another bug here -- reading avaiable_events/set_events can race with module unload:

  # cat available_events               |
      s_start()                        |
      s_stop()                         |
                                       | # rmmod foo.ko
      s_start()                        |
        call = list_entry(m->private)  |

@call might be freed and accessing it will lead to crash.

[ Impact fix missing entries when reading available_events/set_events ]

---
 kernel/trace/trace_events.c |   28 ++++++++++++++++++++++------
 1 files changed, 22 insertions(+), 6 deletions(-)
kernel/trace/trace_events.c +22 -6
index aa341ff..4ab596e 100644
@@ -300,10 +300,18 @@ t_next(struct seq_file *m, void *v, loff_t *pos)
 
 static void *t_start(struct seq_file *m, loff_t *pos)
 {
+	struct ftrace_event_call *call = NULL;
+	loff_t l;
+
 	mutex_lock(&event_mutex);
-	if (*pos == 0)
-		m->private = ftrace_events.next;
-	return t_next(m, NULL, pos);
+
+	m->private = ftrace_events.next;
+	for (l = 0; l <= *pos; ) {
+		call = t_next(m, NULL, &l);
+		if (!call)
+			break;
+	}
+	return call;
 }
 
 static void *
@@ -332,10 +340,18 @@ s_next(struct seq_file *m, void *v, loff_t *pos)
 
 static void *s_start(struct seq_file *m, loff_t *pos)
 {
+	struct ftrace_event_call *call = NULL;
+	loff_t l;
+
 	mutex_lock(&event_mutex);
-	if (*pos == 0)
-		m->private = ftrace_events.next;
-	return s_next(m, NULL, pos);
+
+	m->private = ftrace_events.next;
+	for (l = 0; l <= *pos; ) {
+		call = s_next(m, NULL, &l);
+		if (!call)
+			break;
+	}
+	return call;
 }
 
 static int t_show(struct seq_file *m, void *v)
Signature · 1 line
1.5.4.rc3
author
#2 lore raw

PATCH 2/5 tracing_bprintk: don't increment @pos in t_start()

In reply to Li Zefan

It's wrong to increment @pos in t_start(), otherwise we'll lose some entries when reading printk_formats, if the output is large than PAGE_SIZE.

[ Impact: fix missing entries when reading printk_formats ]

---
 kernel/trace/trace_printk.c |   26 ++++++--------------------
 1 files changed, 6 insertions(+), 20 deletions(-)
kernel/trace/trace_printk.c +6 -20
index 9bece96..7b62781 100644
@@ -155,25 +155,19 @@ int __ftrace_vprintk(unsigned long ip, const char *fmt, va_list ap)
 EXPORT_SYMBOL_GPL(__ftrace_vprintk);
 
 static void *
-t_next(struct seq_file *m, void *v, loff_t *pos)
+t_start(struct seq_file *m, loff_t *pos)
 {
-	const char **fmt = m->private;
-	const char **next = fmt;
-
-	(*pos)++;
+	const char **fmt = __start___trace_bprintk_fmt + *pos;
 
 	if ((unsigned long)fmt >= (unsigned long)__stop___trace_bprintk_fmt)
 		return NULL;
-
-	next = fmt;
-	m->private = ++next;
-
 	return fmt;
 }
 
-static void *t_start(struct seq_file *m, loff_t *pos)
+static void *t_next(struct seq_file *m, void * v, loff_t *pos)
 {
-	return t_next(m, NULL, pos);
+	(*pos)++;
+	return t_start(m, pos);
 }
 
 static int t_show(struct seq_file *m, void *v)
@@ -224,15 +218,7 @@ static const struct seq_operations show_format_seq_ops = {
 static int
 ftrace_formats_open(struct inode *inode, struct file *file)
 {
-	int ret;
-
-	ret = seq_open(file, &show_format_seq_ops);
-	if (!ret) {
-		struct seq_file *m = file->private_data;
-
-		m->private = __start___trace_bprintk_fmt;
-	}
-	return ret;
+	return seq_open(file, &show_format_seq_ops);
 }
 
 static const struct file_operations ftrace_formats_fops = {
Signature · 1 line
1.5.4.rc3
author
#7 lore raw

PATCH 3/5 trace_stat: don't increment @pos in stat_seq_start()

In reply to Li Zefan

It's wrong to increment @pos in stat_seq_start(). It causes some stat entries lost when reading stat file, if the output of the file is large than PAGE_SIZE.

[ Impact: fix missing entries when reading a stat file ]

---
 kernel/trace/trace_stat.c |    6 +-----
 1 files changed, 1 insertions(+), 5 deletions(-)
kernel/trace/trace_stat.c +1 -5
index c006437..e66f5e4 100644
@@ -199,17 +199,13 @@ static void *stat_seq_start(struct seq_file *s, loff_t *pos)
 	mutex_lock(&session->stat_mutex);
 
 	/* If we are in the beginning of the file, print the headers */
-	if (!*pos && session->ts->stat_headers) {
-		(*pos)++;
+	if (!*pos && session->ts->stat_headers)
 		return SEQ_START_TOKEN;
-	}
 
 	node = rb_first(&session->stat_root);
 	for (i = 0; node && i < *pos; i++)
 		node = rb_next(node);
 
-	(*pos)++;
-
 	return node;
 }
 
Signature · 1 line
1.5.4.rc3
author
#8 lore raw

PATCH 4/5 ftrace: don't increment @pos in g_start()

In reply to Li Zefan

It's wrong to increment @pos in g_start(). It causes some entries lost when reading set_graph_function, if the output of the file is large than PAGE_SIZE.

[ Impact: fix missing entries when reading set_graph_function ]

---
 kernel/trace/ftrace.c |   28 +++++++++++++---------------
 1 files changed, 13 insertions(+), 15 deletions(-)
kernel/trace/ftrace.c +13 -15
index ec18bb4..2c1c761 100644
@@ -2492,32 +2492,30 @@ int ftrace_graph_count;
 unsigned long ftrace_graph_funcs[FTRACE_GRAPH_MAX_FUNCS] __read_mostly;
 
 static void *
-g_next(struct seq_file *m, void *v, loff_t *pos)
+__g_next(struct seq_file *m, loff_t *pos)
 {
 	unsigned long *array = m->private;
-	int index = *pos;
 
-	(*pos)++;
+	/* Nothing, tell g_show to print all functions are enabled */
+	if (!ftrace_graph_count && !*pos)
+		return (void *)1;
 
-	if (index >= ftrace_graph_count)
+	if (*pos >= ftrace_graph_count)
 		return NULL;
+	return &array[*pos];
+}
 
-	return &array[index];
+static void *
+g_next(struct seq_file *m, void *v, loff_t *pos)
+{
+	(*pos)++;
+	return __g_next(m, pos);
 }
 
 static void *g_start(struct seq_file *m, loff_t *pos)
 {
-	void *p = NULL;
-
 	mutex_lock(&graph_lock);
-
-	/* Nothing, tell g_show to print all functions are enabled */
-	if (!ftrace_graph_count && !*pos)
-		return (void *)1;
-
-	p = g_next(m, p, pos);
-
-	return p;
+	return __g_next(m, pos);
 }
 
 static void g_stop(struct seq_file *m, void *p)
Signature · 1 line
1.5.4.rc3
author
#15 lore raw

PATCH 5/5 tracing: reset iterator in t_start()

In reply to Li Zefan

The iterator is m->private, but it's not reset to trace_types in t_start(). If the output is larger than PAGE_SIZE and t_start() is called the 2nd time, things will go wrong.

[ Impact: fix output of current_tracer when it's larger than PAGE_SIZE ]

---
 kernel/trace/trace.c |   18 ++++--------------
 1 files changed, 4 insertions(+), 14 deletions(-)
kernel/trace/trace.c +4 -14
index 076fa6f..3bb3100 100644
@@ -2053,25 +2053,23 @@ static int tracing_open(struct inode *inode, struct file *file)
 static void *
 t_next(struct seq_file *m, void *v, loff_t *pos)
 {
-	struct tracer *t = m->private;
+	struct tracer *t = v;
 
 	(*pos)++;
 
 	if (t)
 		t = t->next;
 
-	m->private = t;
-
 	return t;
 }
 
 static void *t_start(struct seq_file *m, loff_t *pos)
 {
-	struct tracer *t = m->private;
+	struct tracer *t;
 	loff_t l = 0;
 
 	mutex_lock(&trace_types_lock);
-	for (; t && l < *pos; t = t_next(m, t, &l))
+	for (t = trace_types; t && l < *pos; t = t_next(m, t, &l))
 		;
 
 	return t;
@@ -2107,18 +2105,10 @@ static struct seq_operations show_traces_seq_ops = {
 
 static int show_traces_open(struct inode *inode, struct file *file)
 {
-	int ret;
-
 	if (tracing_disabled)
 		return -ENODEV;
 
-	ret = seq_open(file, &show_traces_seq_ops);
-	if (!ret) {
-		struct seq_file *m = file->private_data;
-		m->private = trace_types;
-	}
-
-	return ret;
+	return seq_open(file, &show_traces_seq_ops);
 }
 
 static ssize_t
Signature · 1 line
1.5.4.rc3
#9 lore raw

PATCH 1/1 ftrace: don't increment @pos in g_start()

In reply to Li Zefan

how about this one?

It's wrong to increment @pos in g_start(). It causes some entries lost when reading set_graph_function, if the output of the file is large than PAGE_SIZE.

[ Impact: fix missing entries when reading set_graph_function ]

---
 kernel/trace/ftrace.c |   14 +++++---------
 1 files changed, 5 insertions(+), 9 deletions(-)
kernel/trace/ftrace.c +5 -9
index 134e580..1beaac6 100644
@@ -2495,29 +2495,25 @@ static void *
 g_next(struct seq_file *m, void *v, loff_t *pos)
 {
 	unsigned long *array = m->private;
-	int index = *pos;
 
-	(*pos)++;
+	if (v)
+		(*pos)++;
 
-	if (index >= ftrace_graph_count)
+	if (*pos >= ftrace_graph_count)
 		return NULL;
 
-	return &array[index];
+	return &array[*pos];
 }
 
 static void *g_start(struct seq_file *m, loff_t *pos)
 {
-	void *p = NULL;
-
 	mutex_lock(&graph_lock);
 
 	/* Nothing, tell g_show to print all functions are enabled */
 	if (!ftrace_graph_count && !*pos)
 		return (void *)1;
 
-	p = g_next(m, p, pos);
-
-	return p;
+	return g_next(m, NULL, pos);
 }
 
 static void g_stop(struct seq_file *m, void *p)
Signature · 1 line
1.6.0.3
author
#10 lore raw

Re: [PATCH] ftrace: don't increment @pos in g_start()

Quoted 2 lines · Liming Wang

how about this one?

Yeah, this should work, and cleaner than my version.

Quoted 51 lines

It's wrong to increment @pos in g_start(). It causes some entries lost when reading set_graph_function, if the output of the file is large than PAGE_SIZE.

[ Impact: fix missing entries when reading set_graph_function ]

---
 kernel/trace/ftrace.c |   14 +++++---------
 1 files changed, 5 insertions(+), 9 deletions(-)
kernel/trace/ftrace.c +5 -9
index 134e580..1beaac6 100644
@@ -2495,29 +2495,25 @@ static void *
 g_next(struct seq_file *m, void *v, loff_t *pos)
 {
 	unsigned long *array = m->private;
-	int index = *pos;
 
-	(*pos)++;
+	if (v)
+		(*pos)++;
 
-	if (index >= ftrace_graph_count)
+	if (*pos >= ftrace_graph_count)
 		return NULL;
 
-	return &array[index];
+	return &array[*pos];
 }
 
 static void *g_start(struct seq_file *m, loff_t *pos)
 {
-	void *p = NULL;
-
 	mutex_lock(&graph_lock);
 
 	/* Nothing, tell g_show to print all functions are enabled */
 	if (!ftrace_graph_count && !*pos)
 		return (void *)1;
 
-	p = g_next(m, p, pos);
-
-	return p;
+	return g_next(m, NULL, pos);
 }
 
 static void g_stop(struct seq_file *m, void *p)
#3 lore raw

Re: [PATCH 2/5] tracing_bprintk: don't increment @pos in t_start()

In reply to Li Zefan

Hi,

Quoted 36 lines · Li Zefan

It's wrong to increment @pos in t_start(), otherwise we'll lose some entries when reading printk_formats, if the output is large than PAGE_SIZE.

[ Impact: fix missing entries when reading printk_formats ]

---
 kernel/trace/trace_printk.c |   26 ++++++--------------------
 1 files changed, 6 insertions(+), 20 deletions(-)
kernel/trace/trace_printk.c +2 -9
index 9bece96..7b62781 100644
@@ -155,25 +155,19 @@ int __ftrace_vprintk(unsigned long ip, const char *fmt, va_list ap)
 EXPORT_SYMBOL_GPL(__ftrace_vprintk);
 
 static void *
-t_next(struct seq_file *m, void *v, loff_t *pos)
+t_start(struct seq_file *m, loff_t *pos)
 {
-	const char **fmt = m->private;
-	const char **next = fmt;
-
-	(*pos)++;
+	const char **fmt = __start___trace_bprintk_fmt + *pos;
 
 	if ((unsigned long)fmt >= (unsigned long)__stop___trace_bprintk_fmt)
 		return NULL;
-
-	next = fmt;
-	m->private = ++next;
-
 	return fmt;

kernel/trace/trace_printk.c int __ftrace_vprintk(unsigned long ip, const char *fmt, va_list ap) · line 172

Quoted 10 lines
kernel/trace/trace_printk.c +3 -2
 }
 
-static void *t_start(struct seq_file *m, loff_t *pos)
+static void *t_next(struct seq_file *m, void * v, loff_t *pos)
 {
-	return t_next(m, NULL, pos);
+	(*pos)++;
+	return t_start(m, pos);
 }
 

I prefer to .start to call .next, so I rewrite it to following:

@@ -157,17 +157,15 @@ EXPORT_SYMBOL_GPL(__ftrace_vprintk);
static void *
t_next(struct seq_file *m, void *v, loff_t *pos) {

-       const char **fmt = m->private;
-       const char **next = fmt;
+       const char **fmt;

- (*pos)++;
+ if (v)
+ (*pos)++;
+ fmt = __start___trace_bprintk_fmt + *pos;

if ((unsigned long)fmt >= (unsigned long)__stop___trace_bprintk_fmt)
return NULL;

- next = fmt;
- m->private = ++next;
-

        return fmt;

}

but I have not tested it, if you have no objection, then:

Liming Wang

Quoted 18 lines
kernel/trace/trace_printk.c +1 -9
 static int t_show(struct seq_file *m, void *v)
@@ -224,15 +218,7 @@ static const struct seq_operations show_format_seq_ops = {
 static int
 ftrace_formats_open(struct inode *inode, struct file *file)
 {
-	int ret;
-
-	ret = seq_open(file, &show_format_seq_ops);
-	if (!ret) {
-		struct seq_file *m = file->private_data;
-
-		m->private = __start___trace_bprintk_fmt;
-	}
-	return ret;
+	return seq_open(file, &show_format_seq_ops);
 }
 
 static const struct file_operations ftrace_formats_fops = {
author
#4 lore raw

Re: [PATCH 2/5] tracing_bprintk: don't increment @pos in t_start()

Quoted 42 lines · Wang Liming

Hi,

Quoted 39 lines · Li Zefan

It's wrong to increment @pos in t_start(), otherwise we'll lose some entries when reading printk_formats, if the output is large than PAGE_SIZE.

[ Impact: fix missing entries when reading printk_formats ]

---
 kernel/trace/trace_printk.c |   26 ++++++--------------------
 1 files changed, 6 insertions(+), 20 deletions(-)
kernel/trace/trace_printk.c +2 -9
index 9bece96..7b62781 100644
@@ -155,25 +155,19 @@ int __ftrace_vprintk(unsigned long ip, const
 char *fmt, va_list ap)
 EXPORT_SYMBOL_GPL(__ftrace_vprintk);
 
 static void *
-t_next(struct seq_file *m, void *v, loff_t *pos)
+t_start(struct seq_file *m, loff_t *pos)
 {
-    const char **fmt = m->private;
-    const char **next = fmt;
-
-    (*pos)++;
+    const char **fmt = __start___trace_bprintk_fmt + *pos;
 
     if ((unsigned long)fmt >= (unsigned long)__stop___trace_bprintk_fmt)
         return NULL;
-
-    next = fmt;
-    m->private = ++next;
-
     return fmt;
 }
 
Quoted 10 lines

kernel/trace/trace_printk.c int __ftrace_vprintk(unsigned long ip, const char *fmt, va_list ap) · line 172

Quoted 8 lines
kernel/trace/trace_printk.c +3 -2
-static void *t_start(struct seq_file *m, loff_t *pos)
+static void *t_next(struct seq_file *m, void * v, loff_t *pos)
 {
-    return t_next(m, NULL, pos);
+    (*pos)++;
+    return t_start(m, pos);
 }
 

I prefer to .start to call .next, so I rewrite it to following:

Thanks for the comment, but I don't think .next calls .start is bad, and I'm not the only one doing this. Grep c_start() to see some of them.

author
#11 lore raw

Re: [PATCH] ftrace: don't increment @pos in g_start()

In reply to Li Zefan
Quoted 6 lines · Li Zefan
Quoted 2 lines · Liming Wang

how about this one?

Yeah, this should work, and cleaner than my version.

Hmmm, the patch is cleaner in diffstat but the resulted code isn't..

After yours:
text data bss dec hex filename
14879 5480 4240 24599 6017 kernel/trace/ftrace.o

After mine:
text data bss dec hex filename
14873 5480 4240 24593 6011 kernel/trace/ftrace.o

Quoted 51 lines
Quoted 51 lines

It's wrong to increment @pos in g_start(). It causes some entries lost when reading set_graph_function, if the output of the file is large than PAGE_SIZE.

[ Impact: fix missing entries when reading set_graph_function ]

---
 kernel/trace/ftrace.c |   14 +++++---------
 1 files changed, 5 insertions(+), 9 deletions(-)
kernel/trace/ftrace.c +5 -9
index 134e580..1beaac6 100644
@@ -2495,29 +2495,25 @@ static void *
 g_next(struct seq_file *m, void *v, loff_t *pos)
 {
 	unsigned long *array = m->private;
-	int index = *pos;
 
-	(*pos)++;
+	if (v)
+		(*pos)++;
 
-	if (index >= ftrace_graph_count)
+	if (*pos >= ftrace_graph_count)
 		return NULL;
 
-	return &array[index];
+	return &array[*pos];
 }
 
 static void *g_start(struct seq_file *m, loff_t *pos)
 {
-	void *p = NULL;
-
 	mutex_lock(&graph_lock);
 
 	/* Nothing, tell g_show to print all functions are enabled */
 	if (!ftrace_graph_count && !*pos)
 		return (void *)1;
 
-	p = g_next(m, p, pos);
-
-	return p;
+	return g_next(m, NULL, pos);
 }
 
 static void g_stop(struct seq_file *m, void *p)
#5 lore raw

Re: [PATCH 2/5] tracing_bprintk: don't increment @pos in t_start()

In reply to Li Zefan
Quoted 54 lines · Li Zefan
Quoted 42 lines · Wang Liming

Hi,

Quoted 39 lines · Li Zefan

It's wrong to increment @pos in t_start(), otherwise we'll lose some entries when reading printk_formats, if the output is large than PAGE_SIZE.

[ Impact: fix missing entries when reading printk_formats ]

---
 kernel/trace/trace_printk.c |   26 ++++++--------------------
 1 files changed, 6 insertions(+), 20 deletions(-)
kernel/trace/trace_printk.c +2 -9
index 9bece96..7b62781 100644
@@ -155,25 +155,19 @@ int __ftrace_vprintk(unsigned long ip, const
 char *fmt, va_list ap)
 EXPORT_SYMBOL_GPL(__ftrace_vprintk);
 
 static void *
-t_next(struct seq_file *m, void *v, loff_t *pos)
+t_start(struct seq_file *m, loff_t *pos)
 {
-    const char **fmt = m->private;
-    const char **next = fmt;
-
-    (*pos)++;
+    const char **fmt = __start___trace_bprintk_fmt + *pos;
 
     if ((unsigned long)fmt >= (unsigned long)__stop___trace_bprintk_fmt)
         return NULL;
-
-    next = fmt;
-    m->private = ++next;
-
     return fmt;
 }
 
Quoted 10 lines

kernel/trace/trace_printk.c int __ftrace_vprintk(unsigned long ip, const char *fmt, va_list ap) · line 172

Quoted 8 lines
kernel/trace/trace_printk.c +3 -2
-static void *t_start(struct seq_file *m, loff_t *pos)
+static void *t_next(struct seq_file *m, void * v, loff_t *pos)
 {
-    return t_next(m, NULL, pos);
+    (*pos)++;
+    return t_start(m, pos);
 }
 

I prefer to .start to call .next, so I rewrite it to following:

Quoted 3 lines

Thanks for the comment, but I don't think .next calls .start is bad, and I'm not the only one doing this. Grep c_start() to see some of them.

Yes, you are not the only one, but it's the only one in the tracing code. :) I just think we should make the seq_* uniform so that we can understand them more clearly.

Liming Wang

Quoted 2 lines
#12 lore raw

Re: [PATCH] ftrace: don't increment @pos in g_start()

In reply to Li Zefan
Quoted 10 lines · Li Zefan
Quoted 5 lines · Li Zefan
Quoted 2 lines · Liming Wang

how about this one?

Yeah, this should work, and cleaner than my version.

Hmmm, the patch is cleaner in diffstat but the resulted code isn't..

Quoted 7 lines

After yours:
text data bss dec hex filename
14879 5480 4240 24599 6017 kernel/trace/ftrace.o

After mine:
text data bss dec hex filename
14873 5480 4240 24593 6011 kernel/trace/ftrace.o

Hmmm, if you prefer to smaller target size, I don't care. But in my system, I got the same size:

text data bss dec hex filename
14330 5019 104 19453 4bfd kernel/trace/ftrace.o

I use objdump to compute the actual size of all modified functions:

After mine:
func size
g_start 0x50
g_next 0x70

After yours:
func size
__g_next 0x70
g_next 0x20
g_start 0x30

I used Steve git tree and commit e482f8395f215e0ad6557b2722cd9b9b308035c4. My gcc version is :
gcc version 4.2.4

I don't know where the difference.

Liming Wang

Quoted 54 lines
Quoted 53 lines
Quoted 53 lines

It's wrong to increment @pos in g_start(). It causes some entries lost when reading set_graph_function, if the output of the file is large than PAGE_SIZE.

[ Impact: fix missing entries when reading set_graph_function ]

---
 kernel/trace/ftrace.c |   14 +++++---------
 1 files changed, 5 insertions(+), 9 deletions(-)
kernel/trace/ftrace.c +5 -9
index 134e580..1beaac6 100644
@@ -2495,29 +2495,25 @@ static void *
 g_next(struct seq_file *m, void *v, loff_t *pos)
 {
 	unsigned long *array = m->private;
-	int index = *pos;
 
-	(*pos)++;
+	if (v)
+		(*pos)++;
 
-	if (index >= ftrace_graph_count)
+	if (*pos >= ftrace_graph_count)
 		return NULL;
 
-	return &array[index];
+	return &array[*pos];
 }
 
 static void *g_start(struct seq_file *m, loff_t *pos)
 {
-	void *p = NULL;
-
 	mutex_lock(&graph_lock);
 
 	/* Nothing, tell g_show to print all functions are enabled */
 	if (!ftrace_graph_count && !*pos)
 		return (void *)1;
 
-	p = g_next(m, p, pos);
-
-	return p;
+	return g_next(m, NULL, pos);
 }
 
 static void g_stop(struct seq_file *m, void *p)
author
#6 lore raw

Re: [PATCH 2/5] tracing_bprintk: don't increment @pos in t_start()

In reply to Wang Liming
Quoted 56 lines · Wang Liming
Quoted 55 lines · Li Zefan
Quoted 53 lines · Wang Liming

Hi,

Quoted 38 lines · Li Zefan

It's wrong to increment @pos in t_start(), otherwise we'll lose some entries when reading printk_formats, if the output is large than PAGE_SIZE.

[ Impact: fix missing entries when reading printk_formats ]

---
 kernel/trace/trace_printk.c |   26 ++++++--------------------
 1 files changed, 6 insertions(+), 20 deletions(-)
kernel/trace/trace_printk.c +2 -9
index 9bece96..7b62781 100644
@@ -155,25 +155,19 @@ int __ftrace_vprintk(unsigned long ip, const
 char *fmt, va_list ap)
 EXPORT_SYMBOL_GPL(__ftrace_vprintk);
 
 static void *
-t_next(struct seq_file *m, void *v, loff_t *pos)
+t_start(struct seq_file *m, loff_t *pos)
 {
-    const char **fmt = m->private;
-    const char **next = fmt;
-
-    (*pos)++;
+    const char **fmt = __start___trace_bprintk_fmt + *pos;
 
     if ((unsigned long)fmt >= (unsigned
 long)__stop___trace_bprintk_fmt)
         return NULL;
-
-    next = fmt;
-    m->private = ++next;
-

kernel/trace/trace_printk.c int __ftrace_vprintk(unsigned long ip, const · line 173

Quoted 10 lines
kernel/trace/trace_printk.c +3 -2
     return fmt;
 }
 
-static void *t_start(struct seq_file *m, loff_t *pos)
+static void *t_next(struct seq_file *m, void * v, loff_t *pos)
 {
-    return t_next(m, NULL, pos);
+    (*pos)++;
+    return t_start(m, pos);
 }

I prefer to .start to call .next, so I rewrite it to following:

Quoted 8 lines
Quoted 3 lines

Thanks for the comment, but I don't think .next calls .start is bad, and I'm not the only one doing this. Grep c_start() to see some of them.

Yes, you are not the only one, but it's the only one in the tracing code. :)
I just think we should make the seq_* uniform so that we can understand them more clearly.

I don't see how this make seq_* un-uniform..

And I don't want to add extra checking for this kind of uniform.

And if some next()s check (v == NULL) while others don't, do you think it's uniform or not?

author
#13 lore raw

Re: [PATCH] ftrace: don't increment @pos in g_start()

In reply to Wang Liming
Quoted 37 lines · Wang Liming
Quoted 10 lines · Li Zefan
Quoted 5 lines · Li Zefan
Quoted 2 lines · Liming Wang

how about this one?

Yeah, this should work, and cleaner than my version.

Hmmm, the patch is cleaner in diffstat but the resulted code isn't..

Quoted 7 lines

After yours:
text data bss dec hex filename
14879 5480 4240 24599 6017 kernel/trace/ftrace.o

After mine:
text data bss dec hex filename
14873 5480 4240 24593 6011 kernel/trace/ftrace.o

Hmmm, if you prefer to smaller target size, I don't care. But in my system, I got the same size:

text data bss dec hex filename
14330 5019 104 19453 4bfd kernel/trace/ftrace.o

I use objdump to compute the actual size of all modified functions:

After mine:
func size
g_start 0x50
g_next 0x70

After yours:
func size
__g_next 0x70
g_next 0x20
g_start 0x30

Quoted 6 lines

I used Steve git tree and commit e482f8395f215e0ad6557b2722cd9b9b308035c4. My gcc version is :
gcc version 4.2.4

I don't know where the difference.

Maybe because of different gcc versions:

# gcc --version
gcc (GCC) 4.1.2 20070925 (Red Hat 4.1.2-33)

The point is, I don't see how the patch you posted is better than mine. And it's fine for me to pick up yours if it's indeed better.

#14 lore raw

Re: [PATCH] ftrace: don't increment @pos in g_start()

Quoted 41 lines · Li Zefan
Quoted 40 lines · Wang Liming
Quoted 9 lines · Li Zefan
Quoted 5 lines · Li Zefan
Quoted 2 lines · Liming Wang

how about this one?

Yeah, this should work, and cleaner than my version.

Hmmm, the patch is cleaner in diffstat but the resulted code isn't..

Quoted 7 lines

After yours:
text data bss dec hex filename
14879 5480 4240 24599 6017 kernel/trace/ftrace.o

After mine:
text data bss dec hex filename
14873 5480 4240 24593 6011 kernel/trace/ftrace.o

Hmmm, if you prefer to smaller target size, I don't care. But in my system, I got the same size:

text data bss dec hex filename
14330 5019 104 19453 4bfd kernel/trace/ftrace.o

I use objdump to compute the actual size of all modified functions:

After mine:
func size
g_start 0x50
g_next 0x70

After yours:
func size
__g_next 0x70
g_next 0x20
g_start 0x30

I used Steve git tree and commit e482f8395f215e0ad6557b2722cd9b9b308035c4. My gcc version is :
gcc version 4.2.4

Quoted 7 lines
Quoted 2 lines

I don't know where the difference.

Maybe because of different gcc versions:

# gcc --version
gcc (GCC) 4.1.2 20070925 (Red Hat 4.1.2-33)

Maybe.

Quoted 3 lines

The point is, I don't see how the patch you posted is better than mine. And it's fine for me to pick up yours if it's indeed better.

OK, it's fine to me to pick up yours. Nothing different. Back to your patch:

Quoted 8 lines
kernel/trace/ftrace.c +0 -0
index ec18bb4..2c1c761 100644
@@ -2492,32 +2492,30 @@ int ftrace_graph_count;
 unsigned long ftrace_graph_funcs[FTRACE_GRAPH_MAX_FUNCS] __read_mostly;
 
 static void *

kernel/trace/ftrace.c int ftrace_graph_count; · line 2501

Quoted 10 lines
kernel/trace/ftrace.c +4 -3
-g_next(struct seq_file *m, void *v, loff_t *pos)
+__g_next(struct seq_file *m, loff_t *pos)
 {
 	unsigned long *array = m->private;
-	int index = *pos;
 
-	(*pos)++;
+	/* Nothing, tell g_show to print all functions are enabled */
+	if (!ftrace_graph_count && !*pos)
+		return (void *)1;

I think this checking should be put back to g_start, because it's only necessary for g_start.

Liming Wang

Quoted 1 line
#16 lore raw

Re: [PATCH 5/5] tracing: reset iterator in t_start()

In reply to Li Zefan
Quoted 54 lines · Li Zefan

The iterator is m->private, but it's not reset to trace_types in t_start(). If the output is larger than PAGE_SIZE and t_start() is called the 2nd time, things will go wrong.

[ Impact: fix output of current_tracer when it's larger than PAGE_SIZE ]

---
 kernel/trace/trace.c |   18 ++++--------------
 1 files changed, 4 insertions(+), 14 deletions(-)
kernel/trace/trace.c +3 -8
index 076fa6f..3bb3100 100644
@@ -2053,25 +2053,23 @@ static int tracing_open(struct inode *inode, struct file *file)
 static void *
 t_next(struct seq_file *m, void *v, loff_t *pos)
 {
-	struct tracer *t = m->private;
+	struct tracer *t = v;
 
 	(*pos)++;
 
 	if (t)
 		t = t->next;
 
-	m->private = t;
-
 	return t;
 }
 
 static void *t_start(struct seq_file *m, loff_t *pos)
 {
-	struct tracer *t = m->private;
+	struct tracer *t;
 	loff_t l = 0;
 
 	mutex_lock(&trace_types_lock);
-	for (; t && l < *pos; t = t_next(m, t, &l))
+	for (t = trace_types; t && l < *pos; t = t_next(m, t, &l))
 		;
 
 	return t;
@@ -2107,18 +2105,10 @@ static struct seq_operations show_traces_seq_ops = {
 
 static int show_traces_open(struct inode *inode, struct file *file)
 {
-	int ret;
-
 	if (tracing_disabled)
 		return -ENODEV;
 
-	ret = seq_open(file, &show_traces_seq_ops);

kernel/trace/trace.c static struct seq_operations show_traces_seq_ops = { · line 2114

Quoted 10 lines
kernel/trace/trace.c +1 -6
-	if (!ret) {
-		struct seq_file *m = file->private_data;
-		m->private = trace_types;
-	}
-
-	return ret;
+	return seq_open(file, &show_traces_seq_ops);
 }
 
 static ssize_t

Another version:
Since we have saved current (struct tracer *) in m->private in .next, in .start, we don't need to call .next to find the one that should be printed in 2nd or nth time.

kernel/trace/trace.c +0 -5
index cae34c6..02cdccc 100644
@@ -2055,8 +2055,6 @@ t_next(struct seq_file *m, void *v, loff_t *pos)
 {
        struct tracer *t = m->private;
 
-       (*pos)++;
-
        if (t)
                t = t->next;
 
@@ -2068,11 +2066,8 @@ t_next(struct seq_file *m, void *v, loff_t *pos)
 static void *t_start(struct seq_file *m, loff_t *pos)
 {
        struct tracer *t = m->private;
-       loff_t l = 0;
 
        mutex_lock(&trace_types_lock);
-       for (; t && l < *pos; t = t_next(m, t, &l))
-               ;
 
        return t;
 }
author
#17 lore raw

Re: [PATCH 5/5] tracing: reset iterator in t_start()

Quoted 5 lines

Another version:
Since we have saved current (struct tracer *) in m->private in .next, in .start, we don't need to call .next to find the one that should be printed in 2nd or nth time.

I don't like this for 2 reasons.

1. It's strange that @pos is not incremented in next().

2.
t_stop()
mutex_unlock()
unregister_tracer(t)
t_start()

     mutex_lock()
       t = m->private
       ...
       t = t-next.

We access t->next though @t was unregistered. This is not good, though it does no harm here.

Quoted 27 lines
kernel/trace/trace.c +0 -5
index cae34c6..02cdccc 100644
@@ -2055,8 +2055,6 @@ t_next(struct seq_file *m, void *v, loff_t *pos)
 {
        struct tracer *t = m->private;
 
-       (*pos)++;
-
        if (t)
                t = t->next;
 
@@ -2068,11 +2066,8 @@ t_next(struct seq_file *m, void *v, loff_t *pos)
 static void *t_start(struct seq_file *m, loff_t *pos)
 {
        struct tracer *t = m->private;
-       loff_t l = 0;
 
        mutex_lock(&trace_types_lock);
-       for (; t && l < *pos; t = t_next(m, t, &l))
-               ;
 
        return t;
 }
#18 lore raw

Re: [PATCH 5/5] tracing: reset iterator in t_start()

Quoted 9 lines · Li Zefan
Quoted 5 lines

Another version:
Since we have saved current (struct tracer *) in m->private in .next, in .start, we don't need to call .next to find the one that should be printed in 2nd or nth time.

I don't like this for 2 reasons.

1. It's strange that @pos is not incremented in next().

Yes, it's strang, but we know that @pos sometimes is not necessary, such in this position.

Quoted 6 lines

2.
t_stop()
mutex_unlock()
unregister_tracer(t)
t_start()

Quoted 7 lines
     mutex_lock()
       t = m->private
       ...
       t = t-next.

We access t->next though @t was unregistered. This is not good, though it does no harm here.

OK, it's a realy race problem if we call unregister_tracer.
btw: who realy calls this function? :)

Liming Wang

Quoted 29 lines
Quoted 28 lines
kernel/trace/trace.c +0 -5
index cae34c6..02cdccc 100644
@@ -2055,8 +2055,6 @@ t_next(struct seq_file *m, void *v, loff_t *pos)
 {
        struct tracer *t = m->private;
 
-       (*pos)++;
-
        if (t)
                t = t->next;
 
@@ -2068,11 +2066,8 @@ t_next(struct seq_file *m, void *v, loff_t *pos)
 static void *t_start(struct seq_file *m, loff_t *pos)
 {
        struct tracer *t = m->private;
-       loff_t l = 0;
 
        mutex_lock(&trace_types_lock);
-       for (; t && l < *pos; t = t_next(m, t, &l))
-               ;
 
        return t;
 }