public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
* [PATCH] perf: fix uninitialized variable in thread__set_comm()
@ 2011-09-06 15:00 Stephane Eranian
  2011-10-07 10:14 ` Stephane Eranian
  0 siblings, 1 reply; 4+ messages in thread
From: Stephane Eranian @ 2011-09-06 15:00 UTC (permalink / raw)
  To: linux-kernel; +Cc: acme, peterz, mingo, dsahern


In thread__set_comm(), a command string could be freed and then
reused to point to a new command. Yet, the length was never reset
nor recomputed. This could lead to bogus column formatting.

This patch below fixes the problem by having thread__set_comm()
immediately compute the string length in thread->comm_len. It
gets rid of thread__comm_len() which is not needed anymore.
It was doing the lazy eval of strlen() which could cause the
issue.

Signed-off-by: Stephane Eranian <eranian@google.com>
---

diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c
index 677e1da..b883821 100644
--- a/tools/perf/util/hist.c
+++ b/tools/perf/util/hist.c
@@ -61,7 +61,7 @@ static void hists__calc_col_len(struct hists *self, struct hist_entry *h)
 					   unresolved_col_width);
 	}
 
-	len = thread__comm_len(h->thread);
+	len = h->thread->comm_len;
 	if (hists__new_col_len(self, HISTC_COMM, len))
 		hists__set_col_len(self, HISTC_THREAD, len + 6);
 
@@ -731,7 +731,7 @@ static size_t hist_entry__fprintf_callchain(struct hist_entry *self,
 		struct sort_entry *se = list_first_entry(&hist_entry__sort_list,
 							 typeof(*se), list);
 		left_margin = hists__col_len(hists, se->se_width_idx);
-		left_margin -= thread__comm_len(self->thread);
+		left_margin -= self->thread->comm_len;
 	}
 
 	return hist_entry_callchain__fprintf(fp, self, session_total,
diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c
index d5d3b22..3ce2a80 100644
--- a/tools/perf/util/thread.c
+++ b/tools/perf/util/thread.c
@@ -35,26 +35,18 @@ int thread__set_comm(struct thread *self, const char *comm)
 
 	if (self->comm)
 		free(self->comm);
+
 	self->comm = strdup(comm);
+
 	err = self->comm == NULL ? -ENOMEM : 0;
 	if (!err) {
 		self->comm_set = true;
+		self->comm_len = strlen(self->comm);
 		map_groups__flush(&self->mg);
 	}
 	return err;
 }
 
-int thread__comm_len(struct thread *self)
-{
-	if (!self->comm_len) {
-		if (!self->comm)
-			return 0;
-		self->comm_len = strlen(self->comm);
-	}
-
-	return self->comm_len;
-}
-
 static size_t thread__fprintf(struct thread *self, FILE *fp)
 {
 	return fprintf(fp, "Thread %d %s\n", self->pid, self->comm) +
diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h
index e5f2401..701ea09 100644
--- a/tools/perf/util/thread.h
+++ b/tools/perf/util/thread.h
@@ -23,7 +23,6 @@ struct perf_session;
 void thread__delete(struct thread *self);
 
 int thread__set_comm(struct thread *self, const char *comm);
-int thread__comm_len(struct thread *self);
 struct thread *perf_session__findnew(struct perf_session *self, pid_t pid);
 void thread__insert_map(struct thread *self, struct map *map);
 int thread__fork(struct thread *self, struct thread *parent);

^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [PATCH] perf: fix uninitialized variable in thread__set_comm()
  2011-09-06 15:00 [PATCH] perf: fix uninitialized variable in thread__set_comm() Stephane Eranian
@ 2011-10-07 10:14 ` Stephane Eranian
  2011-10-07 20:45   ` Arnaldo Carvalho de Melo
  0 siblings, 1 reply; 4+ messages in thread
From: Stephane Eranian @ 2011-10-07 10:14 UTC (permalink / raw)
  To: linux-kernel; +Cc: acme, peterz, mingo, dsahern

Arnaldo,

Did you also integrate this patch?


On Tue, Sep 6, 2011 at 5:00 PM, Stephane Eranian <eranian@google.com> wrote:
>
> In thread__set_comm(), a command string could be freed and then
> reused to point to a new command. Yet, the length was never reset
> nor recomputed. This could lead to bogus column formatting.
>
> This patch below fixes the problem by having thread__set_comm()
> immediately compute the string length in thread->comm_len. It
> gets rid of thread__comm_len() which is not needed anymore.
> It was doing the lazy eval of strlen() which could cause the
> issue.
>
> Signed-off-by: Stephane Eranian <eranian@google.com>
> ---
>
> diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c
> index 677e1da..b883821 100644
> --- a/tools/perf/util/hist.c
> +++ b/tools/perf/util/hist.c
> @@ -61,7 +61,7 @@ static void hists__calc_col_len(struct hists *self, struct hist_entry *h)
>                                           unresolved_col_width);
>        }
>
> -       len = thread__comm_len(h->thread);
> +       len = h->thread->comm_len;
>        if (hists__new_col_len(self, HISTC_COMM, len))
>                hists__set_col_len(self, HISTC_THREAD, len + 6);
>
> @@ -731,7 +731,7 @@ static size_t hist_entry__fprintf_callchain(struct hist_entry *self,
>                struct sort_entry *se = list_first_entry(&hist_entry__sort_list,
>                                                         typeof(*se), list);
>                left_margin = hists__col_len(hists, se->se_width_idx);
> -               left_margin -= thread__comm_len(self->thread);
> +               left_margin -= self->thread->comm_len;
>        }
>
>        return hist_entry_callchain__fprintf(fp, self, session_total,
> diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c
> index d5d3b22..3ce2a80 100644
> --- a/tools/perf/util/thread.c
> +++ b/tools/perf/util/thread.c
> @@ -35,26 +35,18 @@ int thread__set_comm(struct thread *self, const char *comm)
>
>        if (self->comm)
>                free(self->comm);
> +
>        self->comm = strdup(comm);
> +
>        err = self->comm == NULL ? -ENOMEM : 0;
>        if (!err) {
>                self->comm_set = true;
> +               self->comm_len = strlen(self->comm);
>                map_groups__flush(&self->mg);
>        }
>        return err;
>  }
>
> -int thread__comm_len(struct thread *self)
> -{
> -       if (!self->comm_len) {
> -               if (!self->comm)
> -                       return 0;
> -               self->comm_len = strlen(self->comm);
> -       }
> -
> -       return self->comm_len;
> -}
> -
>  static size_t thread__fprintf(struct thread *self, FILE *fp)
>  {
>        return fprintf(fp, "Thread %d %s\n", self->pid, self->comm) +
> diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h
> index e5f2401..701ea09 100644
> --- a/tools/perf/util/thread.h
> +++ b/tools/perf/util/thread.h
> @@ -23,7 +23,6 @@ struct perf_session;
>  void thread__delete(struct thread *self);
>
>  int thread__set_comm(struct thread *self, const char *comm);
> -int thread__comm_len(struct thread *self);
>  struct thread *perf_session__findnew(struct perf_session *self, pid_t pid);
>  void thread__insert_map(struct thread *self, struct map *map);
>  int thread__fork(struct thread *self, struct thread *parent);
>

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH] perf: fix uninitialized variable in thread__set_comm()
  2011-10-07 10:14 ` Stephane Eranian
@ 2011-10-07 20:45   ` Arnaldo Carvalho de Melo
  2011-10-07 21:53     ` Stephane Eranian
  0 siblings, 1 reply; 4+ messages in thread
From: Arnaldo Carvalho de Melo @ 2011-10-07 20:45 UTC (permalink / raw)
  To: Stephane Eranian; +Cc: linux-kernel, peterz, mingo, dsahern

Em Fri, Oct 07, 2011 at 12:14:18PM +0200, Stephane Eranian escreveu:
> Arnaldo,
> 
> Did you also integrate this patch?
> 
> 
> On Tue, Sep 6, 2011 at 5:00 PM, Stephane Eranian <eranian@google.com> wrote:
> >
> > In thread__set_comm(), a command string could be freed and then
> > reused to point to a new command. Yet, the length was never reset
> > nor recomputed. This could lead to bogus column formatting.

Perhaps I'm out of coffee right now, but how can it be reused in such a
way?

I don't have a problem in doing the strlen in thread__set_comm, just
want to understand the sequence of events that lead to this
inconsistency.

- Arnaldo

> > This patch below fixes the problem by having thread__set_comm()
> > immediately compute the string length in thread->comm_len. It
> > gets rid of thread__comm_len() which is not needed anymore.
> > It was doing the lazy eval of strlen() which could cause the
> > issue.
> >
> > Signed-off-by: Stephane Eranian <eranian@google.com>
> > ---
> >
> > diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c
> > index 677e1da..b883821 100644
> > --- a/tools/perf/util/hist.c
> > +++ b/tools/perf/util/hist.c
> > @@ -61,7 +61,7 @@ static void hists__calc_col_len(struct hists *self, struct hist_entry *h)
> >                                           unresolved_col_width);
> >        }
> >
> > -       len = thread__comm_len(h->thread);
> > +       len = h->thread->comm_len;
> >        if (hists__new_col_len(self, HISTC_COMM, len))
> >                hists__set_col_len(self, HISTC_THREAD, len + 6);
> >
> > @@ -731,7 +731,7 @@ static size_t hist_entry__fprintf_callchain(struct hist_entry *self,
> >                struct sort_entry *se = list_first_entry(&hist_entry__sort_list,
> >                                                         typeof(*se), list);
> >                left_margin = hists__col_len(hists, se->se_width_idx);
> > -               left_margin -= thread__comm_len(self->thread);
> > +               left_margin -= self->thread->comm_len;
> >        }
> >
> >        return hist_entry_callchain__fprintf(fp, self, session_total,
> > diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c
> > index d5d3b22..3ce2a80 100644
> > --- a/tools/perf/util/thread.c
> > +++ b/tools/perf/util/thread.c
> > @@ -35,26 +35,18 @@ int thread__set_comm(struct thread *self, const char *comm)
> >
> >        if (self->comm)
> >                free(self->comm);
> > +
> >        self->comm = strdup(comm);
> > +
> >        err = self->comm == NULL ? -ENOMEM : 0;
> >        if (!err) {
> >                self->comm_set = true;
> > +               self->comm_len = strlen(self->comm);
> >                map_groups__flush(&self->mg);
> >        }
> >        return err;
> >  }
> >
> > -int thread__comm_len(struct thread *self)
> > -{
> > -       if (!self->comm_len) {
> > -               if (!self->comm)
> > -                       return 0;
> > -               self->comm_len = strlen(self->comm);
> > -       }
> > -
> > -       return self->comm_len;
> > -}
> > -
> >  static size_t thread__fprintf(struct thread *self, FILE *fp)
> >  {
> >        return fprintf(fp, "Thread %d %s\n", self->pid, self->comm) +
> > diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h
> > index e5f2401..701ea09 100644
> > --- a/tools/perf/util/thread.h
> > +++ b/tools/perf/util/thread.h
> > @@ -23,7 +23,6 @@ struct perf_session;
> >  void thread__delete(struct thread *self);
> >
> >  int thread__set_comm(struct thread *self, const char *comm);
> > -int thread__comm_len(struct thread *self);
> >  struct thread *perf_session__findnew(struct perf_session *self, pid_t pid);
> >  void thread__insert_map(struct thread *self, struct map *map);
> >  int thread__fork(struct thread *self, struct thread *parent);
> >

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH] perf: fix uninitialized variable in thread__set_comm()
  2011-10-07 20:45   ` Arnaldo Carvalho de Melo
@ 2011-10-07 21:53     ` Stephane Eranian
  0 siblings, 0 replies; 4+ messages in thread
From: Stephane Eranian @ 2011-10-07 21:53 UTC (permalink / raw)
  To: Arnaldo Carvalho de Melo; +Cc: linux-kernel, peterz, mingo, dsahern

On Fri, Oct 7, 2011 at 10:45 PM, Arnaldo Carvalho de Melo
<acme@redhat.com> wrote:
> Em Fri, Oct 07, 2011 at 12:14:18PM +0200, Stephane Eranian escreveu:
>> Arnaldo,
>>
>> Did you also integrate this patch?
>>
>>
>> On Tue, Sep 6, 2011 at 5:00 PM, Stephane Eranian <eranian@google.com> wrote:
>> >
>> > In thread__set_comm(), a command string could be freed and then
>> > reused to point to a new command. Yet, the length was never reset
>> > nor recomputed. This could lead to bogus column formatting.
>
> Perhaps I'm out of coffee right now, but how can it be reused in such a
> way?
>
you come into thread__set_comm() with self->com != NULL,
you free() self->comm, you strdup() the new command, as far as I know it
may have a different length from what you just freed, err = 0, set comm_set
to skip strlen() later on. And you have a stale self->comm_len.
At least that's my understanding of the code. I got there because I noticed
there was something wrong with column formatting.


> I don't have a problem in doing the strlen in thread__set_comm, just
> want to understand the sequence of events that lead to this
> inconsistency.
>
> - Arnaldo
>
>> > This patch below fixes the problem by having thread__set_comm()
>> > immediately compute the string length in thread->comm_len. It
>> > gets rid of thread__comm_len() which is not needed anymore.
>> > It was doing the lazy eval of strlen() which could cause the
>> > issue.
>> >
>> > Signed-off-by: Stephane Eranian <eranian@google.com>
>> > ---
>> >
>> > diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c
>> > index 677e1da..b883821 100644
>> > --- a/tools/perf/util/hist.c
>> > +++ b/tools/perf/util/hist.c
>> > @@ -61,7 +61,7 @@ static void hists__calc_col_len(struct hists *self, struct hist_entry *h)
>> >                                           unresolved_col_width);
>> >        }
>> >
>> > -       len = thread__comm_len(h->thread);
>> > +       len = h->thread->comm_len;
>> >        if (hists__new_col_len(self, HISTC_COMM, len))
>> >                hists__set_col_len(self, HISTC_THREAD, len + 6);
>> >
>> > @@ -731,7 +731,7 @@ static size_t hist_entry__fprintf_callchain(struct hist_entry *self,
>> >                struct sort_entry *se = list_first_entry(&hist_entry__sort_list,
>> >                                                         typeof(*se), list);
>> >                left_margin = hists__col_len(hists, se->se_width_idx);
>> > -               left_margin -= thread__comm_len(self->thread);
>> > +               left_margin -= self->thread->comm_len;
>> >        }
>> >
>> >        return hist_entry_callchain__fprintf(fp, self, session_total,
>> > diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c
>> > index d5d3b22..3ce2a80 100644
>> > --- a/tools/perf/util/thread.c
>> > +++ b/tools/perf/util/thread.c
>> > @@ -35,26 +35,18 @@ int thread__set_comm(struct thread *self, const char *comm)
>> >
>> >        if (self->comm)
>> >                free(self->comm);
>> > +
>> >        self->comm = strdup(comm);
>> > +
>> >        err = self->comm == NULL ? -ENOMEM : 0;
>> >        if (!err) {
>> >                self->comm_set = true;
>> > +               self->comm_len = strlen(self->comm);
>> >                map_groups__flush(&self->mg);
>> >        }
>> >        return err;
>> >  }
>> >
>> > -int thread__comm_len(struct thread *self)
>> > -{
>> > -       if (!self->comm_len) {
>> > -               if (!self->comm)
>> > -                       return 0;
>> > -               self->comm_len = strlen(self->comm);
>> > -       }
>> > -
>> > -       return self->comm_len;
>> > -}
>> > -
>> >  static size_t thread__fprintf(struct thread *self, FILE *fp)
>> >  {
>> >        return fprintf(fp, "Thread %d %s\n", self->pid, self->comm) +
>> > diff --git a/tools/perf/util/thread.h b/tools/perf/util/thread.h
>> > index e5f2401..701ea09 100644
>> > --- a/tools/perf/util/thread.h
>> > +++ b/tools/perf/util/thread.h
>> > @@ -23,7 +23,6 @@ struct perf_session;
>> >  void thread__delete(struct thread *self);
>> >
>> >  int thread__set_comm(struct thread *self, const char *comm);
>> > -int thread__comm_len(struct thread *self);
>> >  struct thread *perf_session__findnew(struct perf_session *self, pid_t pid);
>> >  void thread__insert_map(struct thread *self, struct map *map);
>> >  int thread__fork(struct thread *self, struct thread *parent);
>> >
>

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2011-10-07 21:53 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2011-09-06 15:00 [PATCH] perf: fix uninitialized variable in thread__set_comm() Stephane Eranian
2011-10-07 10:14 ` Stephane Eranian
2011-10-07 20:45   ` Arnaldo Carvalho de Melo
2011-10-07 21:53     ` Stephane Eranian

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox