Skip to contentSkip to conversation

does gcc gives a false warning in kernel/trace/trace_events_filter.c ?

DISCUSSION Toralf Förster 2 participants 5 messages Sep 2, 2012 – Sep 6, 2012

lore
Share & tools
lore
#0 lore raw

The current git tree of linux gave with gcc-4.6.3 :

kernel/trace/trace_events_filter.c: In function ‘ftrace_function_set_filter_cb’:
kernel/trace/trace_events_filter.c:2074:8: warning: ‘ret’ may be used uninitialized in this function [-Wuninitialized]

which refers to this piece of code:

  2061  static int ftrace_function_set_filter_cb(enum move_type move,
  2062                                           struct filter_pred *pred,
  2063                                           int *err, void *data)
  2064  {
  2065          /* Checking the node is valid for function trace. */
  2066          if ((move != MOVE_DOWN) ||
  2067              (pred->left != FILTER_PRED_INVALID)) {
  2068                  *err = ftrace_function_check_pred(pred, 0);
  2069          } else {
  2070                  *err = ftrace_function_check_pred(pred, 1);
  2071                  if (*err)
  2072                          return WALK_PRED_ABORT;
  2073 
  2074                  *err = __ftrace_function_set_filter(pred->op == OP_EQ,
  2075                                                      pred->regex.pattern,
  2076                                                      pred->regex.len,
  2077                                                      data);
  2078          }
  2079 
  2080          return (*err) ? WALK_PRED_ABORT : WALK_PRED_DEFAULT;
  2081  }
  2082  
Quoted 1 line

From a Gentoo forum user I got a hint :

"Maybe it's some kind of a weird inlining issue? I think it's referring to the ret in __ftrace_function_set_filter(), which would be uninitialized if the for-loop does not run (re_cnt ≤ 0)"

Now I'm wondering if re_cnt can become zero or if gcc is wrong here ?

Signature · 3 lines
MfG/Sincerely
Toralf Förster
pgp finger print: 7B1A 07F4 EC82 0F90 D4C2 8936 872A E508 7DB6 9DA3
#1 lore raw
Quoted 33 lines · Toralf Förster

The current git tree of linux gave with gcc-4.6.3 :

kernel/trace/trace_events_filter.c: In function ‘ftrace_function_set_filter_cb’:
kernel/trace/trace_events_filter.c:2074:8: warning: ‘ret’ may be used uninitialized in this function [-Wuninitialized]

which refers to this piece of code:

  2061  static int ftrace_function_set_filter_cb(enum move_type move,
  2062                                           struct filter_pred *pred,
  2063                                           int *err, void *data)
  2064  {
  2065          /* Checking the node is valid for function trace. */
  2066          if ((move != MOVE_DOWN) ||
  2067              (pred->left != FILTER_PRED_INVALID)) {
  2068                  *err = ftrace_function_check_pred(pred, 0);
  2069          } else {
  2070                  *err = ftrace_function_check_pred(pred, 1);
  2071                  if (*err)
  2072                          return WALK_PRED_ABORT;
  2073 
  2074                  *err = __ftrace_function_set_filter(pred->op == OP_EQ,
  2075                                                      pred->regex.pattern,
  2076                                                      pred->regex.len,
  2077                                                      data);
  2078          }
  2079 
  2080          return (*err) ? WALK_PRED_ABORT : WALK_PRED_DEFAULT;
  2081  }
  2082  
Quoted 8 lines
Quoted 1 line

From a Gentoo forum user I got a hint :

"Maybe it's some kind of a weird inlining issue? I think it's referring to the ret in __ftrace_function_set_filter(), which would be uninitialized if the for-loop does not run (re_cnt ≤ 0)"

Now I'm wondering if re_cnt can become zero or if gcc is wrong here ?

Strange, as ret is initialized to 'ret = -EINVAL;' in __ftrace_function_set_filter(). I'm thinking that gcc got confused here. Maybe report it to the gcc maintainers?

-- Steve

#2 lore raw
Quoted 40 lines · Steven Rostedt
Quoted 39 lines · Toralf Förster

The current git tree of linux gave with gcc-4.6.3 :

kernel/trace/trace_events_filter.c: In function ‘ftrace_function_set_filter_cb’:
kernel/trace/trace_events_filter.c:2074:8: warning: ‘ret’ may be used uninitialized in this function [-Wuninitialized]

which refers to this piece of code:

  2061  static int ftrace_function_set_filter_cb(enum move_type move,
  2062                                           struct filter_pred *pred,
  2063                                           int *err, void *data)
  2064  {
  2065          /* Checking the node is valid for function trace. */
  2066          if ((move != MOVE_DOWN) ||
  2067              (pred->left != FILTER_PRED_INVALID)) {
  2068                  *err = ftrace_function_check_pred(pred, 0);
  2069          } else {
  2070                  *err = ftrace_function_check_pred(pred, 1);
  2071                  if (*err)
  2072                          return WALK_PRED_ABORT;
  2073 
  2074                  *err = __ftrace_function_set_filter(pred->op == OP_EQ,
  2075                                                      pred->regex.pattern,
  2076                                                      pred->regex.len,
  2077                                                      data);
  2078          }
  2079 
  2080          return (*err) ? WALK_PRED_ABORT : WALK_PRED_DEFAULT;
  2081  }
  2082  
Quoted 1 line

From a Gentoo forum user I got a hint :

"Maybe it's some kind of a weird inlining issue? I think it's referring to the ret in __ftrace_function_set_filter(), which would be uninitialized if the for-loop does not run (re_cnt ≤ 0)"

Quoted 9 lines
Quoted 2 lines

Now I'm wondering if re_cnt can become zero or if gcc is wrong here ?

Strange, as ret is initialized to 'ret = -EINVAL;' in __ftrace_function_set_filter(). I'm thinking that gcc got confused here. Maybe report it to the gcc maintainers?

-- Steve

I filed a bug report

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54495
and got this answer :

--- Comment #1 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-09-05 22:14:00 UTC --- But if the call to ftrace_function_filter_re sets re_cnt to 0, then ret indeed will be used uninitialized AFAICT. What am I missing?

Signature · 3 lines
MfG/Sincerely
Toralf Förster
pgp finger print: 7B1A 07F4 EC82 0F90 D4C2 8936 872A E508 7DB6 9DA3
#3 lore raw
Quoted 9 lines · Toralf Förster

I filed a bug report

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54495
and got this answer :

--- Comment #1 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-09-05 22:14:00 UTC --- But if the call to ftrace_function_filter_re sets re_cnt to 0, then ret indeed will be used uninitialized AFAICT. What am I missing?

That I think we are looking at two different code bases ;-)

I've been looking at what's been queued for 3.7 and not what's in mainline. If you look at tip/master, or even linux-next, you'll find:

commit 92d8d4a8b0f "tracing/filter: Add missing initialization"

Which does:

static int __ftrace_function_set_filter(int filter, char *buf, int len,
struct function_filter_data *data)
{
- int i, re_cnt, ret;

+       int i, re_cnt, ret = -EINVAL;
        int *reset;
        char **re;

Thus, you were correct. This could have been marked urgent, but as it isn't that big of a deal I just queued it for the next merge window.

-- Steve

#4 lore raw
Quoted 30 lines · Steven Rostedt
Quoted 9 lines · Toralf Förster

I filed a bug report

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54495
and got this answer :

--- Comment #1 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2012-09-05 22:14:00 UTC --- But if the call to ftrace_function_filter_re sets re_cnt to 0, then ret indeed will be used uninitialized AFAICT. What am I missing?

That I think we are looking at two different code bases ;-)

I've been looking at what's been queued for 3.7 and not what's in mainline. If you look at tip/master, or even linux-next, you'll find:

commit 92d8d4a8b0f "tracing/filter: Add missing initialization"

Which does:

static int __ftrace_function_set_filter(int filter, char *buf, int len,
struct function_filter_data *data)
{

(quoted diff) +1 -1
-       int i, re_cnt, ret;
+       int i, re_cnt, ret = -EINVAL;
        int *reset;
        char **re;
Quoted 7 lines

Thus, you were correct. This could have been marked urgent, but as it isn't that big of a deal I just queued it for the next merge window.

-- Steve

ah - thx
:-)

Signature · 3 lines
MfG/Sincerely
Toralf Förster
pgp finger print: 7B1A 07F4 EC82 0F90 D4C2 8936 872A E508 7DB6 9DA3