[PATCH -tip 1/4] [BUGFIX] perf tools: Initialize dso->node member in dso__new

[PATCH -tip 1/4] [BUGFIX] perf tools: Initialize dso->node member in dso__new

[Masami Hiramatsu]

If dso->node member is not initilized, it causes a segmentation
fault when adding to other lists. It should be initilized in
dso__new().

Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
---

 tools/perf/util/symbol.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
index e782e7d..e77c33a 100644
--- a/tools/perf/util/symbol.c
+++ b/tools/perf/util/symbol.c
@@ -189,6 +189,7 @@ struct dso *dso__new(const char *name)
 		self->sorted_by_name = 0;
 		self->has_build_id = 0;
 		self->kernel = DSO_TYPE_USER;
+		INIT_LIST_HEAD(&self->node);
 	}
 
 	return self;


-- 
Masami Hiramatsu
e-mail: mhiramat@redhat.com

[PATCH -tip 3/4] perf probe: Fix to exit callback soon after finding too many probe points

[Masami Hiramatsu]

Fix to exit callback soon after finding too many probe points.
Don't try to continue searching because it already failed.

Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
---

 tools/perf/util/probe-finder.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c
index 3e79775..57b51ca 100644
--- a/tools/perf/util/probe-finder.c
+++ b/tools/perf/util/probe-finder.c
@@ -922,6 +922,8 @@ static int probe_point_inline_cb(Dwarf_Die *in_die, void *data)
 			 (uintmax_t)pf->addr);
 
 		param->retval = convert_probe_point(in_die, pf);
+		if (param->retval < 0)
+			return DWARF_CB_ABORT;
 	}
 
 	return DWARF_CB_OK;
@@ -1157,6 +1159,8 @@ static int line_range_funcdecl_cb(Dwarf_Die *sp_die, void *data)
 		return DWARF_CB_OK;
 
 	param->retval = line_range_add_line(src, lineno, lf->lr);
+	if (param->retval < 0)
+		return DWARF_CB_ABORT;
 	return DWARF_CB_OK;
 }
 


-- 
Masami Hiramatsu
e-mail: mhiramat@redhat.com

[PATCH -tip 4/4] perf probe: Add --max-probes option

[Masami Hiramatsu]

Add --max-probes option to change the maximum limit of
findable probe points per event, since inlined function can be
expanded into thousands of probe points. Default value is 128.

Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Suggested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
---

 tools/perf/Documentation/perf-probe.txt |    3 +++
 tools/perf/builtin-probe.c              |    9 ++++++++-
 tools/perf/util/probe-event.c           |   17 ++++++++++-------
 tools/perf/util/probe-event.h           |    4 ++--
 tools/perf/util/probe-finder.c          |   11 ++++++-----
 tools/perf/util/probe-finder.h          |    6 ++++--
 6 files changed, 33 insertions(+), 17 deletions(-)

diff --git a/tools/perf/Documentation/perf-probe.txt b/tools/perf/Documentation/perf-probe.txt
index 63c25d3..94a258c 100644
--- a/tools/perf/Documentation/perf-probe.txt
+++ b/tools/perf/Documentation/perf-probe.txt
@@ -62,6 +62,9 @@ OPTIONS
 	Dry run. With this option, --add and --del doesn't execute actual
 	adding and removal operations.
 
+--max-probes::
+	Set the maximum number of probe points for an event. Default is 128.
+
 PROBE SYNTAX
 ------------
 Probe points are defined by following syntax.
diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c
index c1e5403..61c6d70 100644
--- a/tools/perf/builtin-probe.c
+++ b/tools/perf/builtin-probe.c
@@ -54,6 +54,7 @@ static struct {
 	struct perf_probe_event events[MAX_PROBES];
 	struct strlist *dellist;
 	struct line_range line_range;
+	int max_probe_points;
 } params;
 
 
@@ -179,6 +180,8 @@ static const struct option options[] = {
 		   "file", "vmlinux pathname"),
 #endif
 	OPT__DRY_RUN(&probe_event_dry_run),
+	OPT_INTEGER('\0', "max-probes", &params.max_probe_points,
+		 "Set how many probe points can be found for a probe."),
 	OPT_END()
 };
 
@@ -200,6 +203,9 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used)
 		}
 	}
 
+	if (params.max_probe_points == 0)
+		params.max_probe_points = MAX_PROBES;
+
 	if ((!params.nevents && !params.dellist && !params.list_events &&
 	     !params.show_lines))
 		usage_with_options(probe_usage, options);
@@ -246,7 +252,8 @@ int cmd_probe(int argc, const char **argv, const char *prefix __used)
 
 	if (params.nevents) {
 		ret = add_perf_probe_events(params.events, params.nevents,
-					    params.force_add);
+					    params.force_add,
+					    params.max_probe_points);
 		if (ret < 0) {
 			pr_err("  Error: Failed to add events. (%d)\n", ret);
 			return ret;
diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c
index 5d3baec..9ded38c 100644
--- a/tools/perf/util/probe-event.c
+++ b/tools/perf/util/probe-event.c
@@ -150,7 +150,8 @@ static int convert_to_perf_probe_point(struct kprobe_trace_point *tp,
 
 /* Try to find perf_probe_event with debuginfo */
 static int try_to_find_kprobe_trace_events(struct perf_probe_event *pev,
-					   struct kprobe_trace_event **tevs)
+					   struct kprobe_trace_event **tevs,
+					   int max_tevs)
 {
 	bool need_dwarf = perf_probe_event_need_dwarf(pev);
 	int fd, ntevs;
@@ -166,7 +167,7 @@ static int try_to_find_kprobe_trace_events(struct perf_probe_event *pev,
 	}
 
 	/* Searching trace events corresponding to probe event */
-	ntevs = find_kprobe_trace_events(fd, pev, tevs);
+	ntevs = find_kprobe_trace_events(fd, pev, tevs, max_tevs);
 	close(fd);
 
 	if (ntevs > 0) {	/* Succeeded to find trace events */
@@ -318,7 +319,8 @@ static int convert_to_perf_probe_point(struct kprobe_trace_point *tp,
 }
 
 static int try_to_find_kprobe_trace_events(struct perf_probe_event *pev,
-				struct kprobe_trace_event **tevs __unused)
+				struct kprobe_trace_event **tevs __unused,
+				int max_tevs __unused)
 {
 	if (perf_probe_event_need_dwarf(pev)) {
 		pr_warning("Debuginfo-analysis is not supported.\n");
@@ -1408,14 +1410,15 @@ static int __add_kprobe_trace_events(struct perf_probe_event *pev,
 }
 
 static int convert_to_kprobe_trace_events(struct perf_probe_event *pev,
-					  struct kprobe_trace_event **tevs)
+					  struct kprobe_trace_event **tevs,
+					  int max_tevs)
 {
 	struct symbol *sym;
 	int ret = 0, i;
 	struct kprobe_trace_event *tev;
 
 	/* Convert perf_probe_event with debuginfo */
-	ret = try_to_find_kprobe_trace_events(pev, tevs);
+	ret = try_to_find_kprobe_trace_events(pev, tevs, max_tevs);
 	if (ret != 0)
 		return ret;
 
@@ -1487,7 +1490,7 @@ struct __event_package {
 };
 
 int add_perf_probe_events(struct perf_probe_event *pevs, int npevs,
-			  bool force_add)
+			  bool force_add, int max_tevs)
 {
 	int i, j, ret;
 	struct __event_package *pkgs;
@@ -1506,7 +1509,7 @@ int add_perf_probe_events(struct perf_probe_event *pevs, int npevs,
 		pkgs[i].pev = &pevs[i];
 		/* Convert with or without debuginfo */
 		ret  = convert_to_kprobe_trace_events(pkgs[i].pev,
-						      &pkgs[i].tevs);
+						      &pkgs[i].tevs, max_tevs);
 		if (ret < 0)
 			goto end;
 		pkgs[i].ntevs = ret;
diff --git a/tools/perf/util/probe-event.h b/tools/perf/util/probe-event.h
index e7ff0d0..e9db1a2 100644
--- a/tools/perf/util/probe-event.h
+++ b/tools/perf/util/probe-event.h
@@ -115,8 +115,8 @@ extern void clear_kprobe_trace_event(struct kprobe_trace_event *tev);
 extern int parse_line_range_desc(const char *cmd, struct line_range *lr);
 
 
-extern int add_perf_probe_events(struct perf_probe_event *pevs, int ntevs,
-				 bool force_add);
+extern int add_perf_probe_events(struct perf_probe_event *pevs, int npevs,
+				 bool force_add, int max_probe_points);
 extern int del_perf_probe_events(struct strlist *dellist);
 extern int show_perf_probe_events(void);
 extern int show_line_range(struct line_range *lr);
diff --git a/tools/perf/util/probe-finder.c b/tools/perf/util/probe-finder.c
index 57b51ca..e3e3c1e 100644
--- a/tools/perf/util/probe-finder.c
+++ b/tools/perf/util/probe-finder.c
@@ -677,8 +677,9 @@ static int convert_probe_point(Dwarf_Die *sp_die, struct probe_finder *pf)
 	Dwarf_Attribute fb_attr;
 	size_t nops;
 
-	if (pf->ntevs == MAX_PROBES) {
-		pr_warning("Too many( > %d) probe point found.\n", MAX_PROBES);
+	if (pf->ntevs == pf->max_tevs) {
+		pr_warning("Too many( > %d) probe point found.\n",
+			   pf->max_tevs);
 		return -ERANGE;
 	}
 	tev = &pf->tevs[pf->ntevs++];
@@ -983,9 +984,9 @@ static int find_probe_point_by_func(struct probe_finder *pf)
 
 /* Find kprobe_trace_events specified by perf_probe_event from debuginfo */
 int find_kprobe_trace_events(int fd, struct perf_probe_event *pev,
-			     struct kprobe_trace_event **tevs)
+			     struct kprobe_trace_event **tevs, int max_tevs)
 {
-	struct probe_finder pf = {.pev = pev};
+	struct probe_finder pf = {.pev = pev, .max_tevs = max_tevs};
 	struct perf_probe_point *pp = &pev->point;
 	Dwarf_Off off, noff;
 	size_t cuhl;
@@ -993,7 +994,7 @@ int find_kprobe_trace_events(int fd, struct perf_probe_event *pev,
 	Dwarf *dbg;
 	int ret = 0;
 
-	pf.tevs = zalloc(sizeof(struct kprobe_trace_event) * MAX_PROBES);
+	pf.tevs = zalloc(sizeof(struct kprobe_trace_event) * max_tevs);
 	if (pf.tevs == NULL)
 		return -ENOMEM;
 	*tevs = pf.tevs;
diff --git a/tools/perf/util/probe-finder.h b/tools/perf/util/probe-finder.h
index 310ce89..66f1980 100644
--- a/tools/perf/util/probe-finder.h
+++ b/tools/perf/util/probe-finder.h
@@ -18,7 +18,8 @@ static inline int is_c_varname(const char *name)
 #ifdef DWARF_SUPPORT
 /* Find kprobe_trace_events specified by perf_probe_event from debuginfo */
 extern int find_kprobe_trace_events(int fd, struct perf_probe_event *pev,
-				    struct kprobe_trace_event **tevs);
+				    struct kprobe_trace_event **tevs,
+				    int max_tevs);
 
 /* Find a perf_probe_point from debuginfo */
 extern int find_perf_probe_point(int fd, unsigned long addr,
@@ -32,7 +33,8 @@ extern int find_line_range(int fd, struct line_range *lr);
 struct probe_finder {
 	struct perf_probe_event	*pev;		/* Target probe event */
 	struct kprobe_trace_event *tevs;	/* Result trace events */
-	int			ntevs;		/* number of trace events */
+	int			ntevs;		/* Number of trace events */
+	int			max_tevs;	/* Max number of trace events */
 
 	/* For function searching */
 	int			lno;		/* Line number */


-- 
Masami Hiramatsu
e-mail: mhiramat@redhat.com

[PATCH -tip 2/4] perf probe: Fix to use symtab only if no debuginfo

[Masami Hiramatsu]

Fix perf probe to use symtab only if there is no debuginfo,
because debuginfo has more precise information than symtab.
If we can't find a function in debuginfo, we never find it
in symtab.

Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
---

 tools/perf/util/probe-event.c |   17 +++++++++--------
 1 files changed, 9 insertions(+), 8 deletions(-)

diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c
index 4fb4803..5d3baec 100644
--- a/tools/perf/util/probe-event.c
+++ b/tools/perf/util/probe-event.c
@@ -180,15 +180,16 @@ static int try_to_find_kprobe_trace_events(struct perf_probe_event *pev,
 		return -ENOENT;
 	}
 	/* Error path : ntevs < 0 */
-	if (need_dwarf) {
-		if (ntevs == -EBADF)
-			pr_warning("No dwarf info found in the vmlinux - "
-				"please rebuild with CONFIG_DEBUG_INFO=y.\n");
-		return ntevs;
+	pr_debug("An error occurred in debuginfo analysis (%d).\n", ntevs);
+	if (ntevs == -EBADF) {
+		pr_warning("Warning: No dwarf info found in the vmlinux - "
+			"please rebuild kernel with CONFIG_DEBUG_INFO=y.\n");
+		if (!need_dwarf) {
+			pr_debug("Trying to use symbols.\nn");
+			return 0;
+		}
 	}
-	pr_debug("An error occurred in debuginfo analysis."
-		 " Try to use symbols.\n");
-	return 0;
+	return ntevs;
 }
 
 #define LINEBUF_SIZE 256


-- 
Masami Hiramatsu
e-mail: mhiramat@redhat.com

Re: [PATCH -tip 2/4] perf probe: Fix to use symtab only if no debuginfo

[Arnaldo Carvalho de Melo]

Em Wed, Apr 21, 2010 at 03:56:24PM -0400, Masami Hiramatsu escreveu:
> Fix perf probe to use symtab only if there is no debuginfo,
> because debuginfo has more precise information than symtab.

"more precise"? Wouldn't it be "more expressive", i.e. with more
information (parameter list with types, return type, location of
parameters, etc), because what is common to both have the same value,
i.e. the function address and size.

> If we can't find a function in debuginfo, we never find it
> in symtab.
> 
> Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
> Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
> Cc: Paul Mackerras <paulus@samba.org>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Mike Galbraith <efault@gmx.de>
> Cc: Frederic Weisbecker <fweisbec@gmail.com>
> Cc: Ingo Molnar <mingo@elte.hu>
> ---
> 
>  tools/perf/util/probe-event.c |   17 +++++++++--------
>  1 files changed, 9 insertions(+), 8 deletions(-)
> 
> diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c
> index 4fb4803..5d3baec 100644
> --- a/tools/perf/util/probe-event.c
> +++ b/tools/perf/util/probe-event.c
> @@ -180,15 +180,16 @@ static int try_to_find_kprobe_trace_events(struct perf_probe_event *pev,
>  		return -ENOENT;
>  	}
>  	/* Error path : ntevs < 0 */
> -	if (need_dwarf) {
> -		if (ntevs == -EBADF)
> -			pr_warning("No dwarf info found in the vmlinux - "
> -				"please rebuild with CONFIG_DEBUG_INFO=y.\n");
> -		return ntevs;
> +	pr_debug("An error occurred in debuginfo analysis (%d).\n", ntevs);
> +	if (ntevs == -EBADF) {
> +		pr_warning("Warning: No dwarf info found in the vmlinux - "
> +			"please rebuild kernel with CONFIG_DEBUG_INFO=y.\n");
> +		if (!need_dwarf) {
> +			pr_debug("Trying to use symbols.\nn");
> +			return 0;
> +		}
>  	}
> -	pr_debug("An error occurred in debuginfo analysis."
> -		 " Try to use symbols.\n");
> -	return 0;
> +	return ntevs;
>  }
>  
>  #define LINEBUF_SIZE 256
> 
> 
> -- 
> Masami Hiramatsu
> e-mail: mhiramat@redhat.com

Re: [PATCH -tip 1/4] [BUGFIX] perf tools: Initialize dso->node member in dso__new

[Arnaldo Carvalho de Melo]

Em Wed, Apr 21, 2010 at 03:56:16PM -0400, Masami Hiramatsu escreveu:
> If dso->node member is not initilized, it causes a segmentation
> fault when adding to other lists. It should be initilized in
> dso__new().
> 
> Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
> Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Paul Mackerras <paulus@samba.org>
> Cc: Mike Galbraith <efault@gmx.de>
> Cc: Frederic Weisbecker <fweisbec@gmail.com>
> Cc: Ingo Molnar <mingo@elte.hu>
> ---
> 
>  tools/perf/util/symbol.c |    1 +
>  1 files changed, 1 insertions(+), 0 deletions(-)
> 
> diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
> index e782e7d..e77c33a 100644
> --- a/tools/perf/util/symbol.c
> +++ b/tools/perf/util/symbol.c
> @@ -189,6 +189,7 @@ struct dso *dso__new(const char *name)
>  		self->sorted_by_name = 0;
>  		self->has_build_id = 0;
>  		self->kernel = DSO_TYPE_USER;
> +		INIT_LIST_HEAD(&self->node);
>  	}
>  
>  	return self;

Trying to understand how this would be a problem, as:

static inline void __list_add(struct list_head *new,
                              struct list_head *prev,
                              struct list_head *next)
{
        next->prev = new;
        new->next = next;
        new->prev = prev;
        prev->next = new;
}

This is not a list head, just a node.

Looking at other messages where you described the problem to try to
understand why this would help.

- Arnaldo

Re: [PATCH -tip 2/4] perf probe: Fix to use symtab only if no debuginfo

[Masami Hiramatsu]

Arnaldo Carvalho de Melo wrote:
> Em Wed, Apr 21, 2010 at 03:56:24PM -0400, Masami Hiramatsu escreveu:
>> Fix perf probe to use symtab only if there is no debuginfo,
>> because debuginfo has more precise information than symtab.
> 
> "more precise"? Wouldn't it be "more expressive", i.e. with more
> information (parameter list with types, return type, location of
> parameters, etc), because what is common to both have the same value,
> i.e. the function address and size.

Indeed. debuginfo has just more information than symtab, but
symtab certainly provides precise information for symbol-address.

Thank you,


-- 
Masami Hiramatsu
e-mail: mhiramat@redhat.com

Re: [PATCH -tip 2/4] perf probe: Fix to use symtab only if no debuginfo

[Arnaldo Carvalho de Melo]

Em Mon, Apr 26, 2010 at 02:59:36PM -0400, Masami Hiramatsu escreveu:
> Arnaldo Carvalho de Melo wrote:
> > Em Wed, Apr 21, 2010 at 03:56:24PM -0400, Masami Hiramatsu escreveu:
> >> Fix perf probe to use symtab only if there is no debuginfo,
> >> because debuginfo has more precise information than symtab.
> > 
> > "more precise"? Wouldn't it be "more expressive", i.e. with more
> > information (parameter list with types, return type, location of
> > parameters, etc), because what is common to both have the same value,
> > i.e. the function address and size.
> 
> Indeed. debuginfo has just more information than symtab, but
> symtab certainly provides precise information for symbol-address.

I applied the patch, just replaced "more precise information" with "more
information".

- Arnaldo

Re: [PATCH -tip 2/4] perf probe: Fix to use symtab only if no debuginfo

[Masami Hiramatsu]

Arnaldo Carvalho de Melo wrote:
> Em Mon, Apr 26, 2010 at 02:59:36PM -0400, Masami Hiramatsu escreveu:
>> Arnaldo Carvalho de Melo wrote:
>>> Em Wed, Apr 21, 2010 at 03:56:24PM -0400, Masami Hiramatsu escreveu:
>>>> Fix perf probe to use symtab only if there is no debuginfo,
>>>> because debuginfo has more precise information than symtab.
>>>
>>> "more precise"? Wouldn't it be "more expressive", i.e. with more
>>> information (parameter list with types, return type, location of
>>> parameters, etc), because what is common to both have the same value,
>>> i.e. the function address and size.
>>
>> Indeed. debuginfo has just more information than symtab, but
>> symtab certainly provides precise information for symbol-address.
> 
> I applied the patch, just replaced "more precise information" with "more
> information".

Thank you very much! :)

> 
> - Arnaldo
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

-- 
Masami Hiramatsu
e-mail: mhiramat@redhat.com