[patch] trace: find the correct ftrace event

[patch] trace: find the correct ftrace event

[Dan Carpenter]

The original code doesn't work because "call" is never NULL there.

Signed-off-by: Dan Carpenter <error27@gmail.com>
---
My code should work, but it seems like there should but it seems like
there should be a more elegant way to do this?

diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c
index 4615f62..6070c70 100644
--- a/kernel/trace/trace_events_filter.c
+++ b/kernel/trace/trace_events_filter.c
@@ -1388,16 +1388,19 @@ int ftrace_profile_set_filter(struct perf_event *event, int event_id,
 	struct event_filter *filter;
 	struct filter_parse_state *ps;
 	struct ftrace_event_call *call = NULL;
+	int found = 0;
 
 	mutex_lock(&event_mutex);
 
 	list_for_each_entry(call, &ftrace_events, list) {
-		if (call->id == event_id)
+		if (call->id == event_id) {
+			found = 1;
 			break;
+		}
 	}
 
 	err = -EINVAL;
-	if (!call)
+	if (!found)
 		goto out_unlock;
 
 	err = -EEXIST;

Re: [patch] trace: find the correct ftrace event

[Steven Rostedt]

On Sat, 2010-03-20 at 16:19 +0300, Dan Carpenter wrote:
> The original code doesn't work because "call" is never NULL there.

A one line change would have done the same ;-)

> Signed-off-by: Dan Carpenter <error27@gmail.com>
> ---
> My code should work, but it seems like there should but it seems like
> there should be a more elegant way to do this?
> 
> diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c
> index 4615f62..6070c70 100644
> --- a/kernel/trace/trace_events_filter.c
> +++ b/kernel/trace/trace_events_filter.c
> @@ -1388,16 +1388,19 @@ int ftrace_profile_set_filter(struct perf_event *event, int event_id,
>  	struct event_filter *filter;
>  	struct filter_parse_state *ps;
>  	struct ftrace_event_call *call = NULL;
> +	int found = 0;
>  
>  	mutex_lock(&event_mutex);
>  
>  	list_for_each_entry(call, &ftrace_events, list) {
> -		if (call->id == event_id)
> +		if (call->id == event_id) {
> +			found = 1;
>  			break;
> +		}
>  	}
>  
>  	err = -EINVAL;
> -	if (!call)

	if (call == &ftrace_events)

-- Steve

> +	if (!found)
>  		goto out_unlock;
>  
>  	err = -EEXIST;

Re: [patch] trace: find the correct ftrace event

[Dan Carpenter]

On Sat, Mar 20, 2010 at 09:37:13AM -0400, Steven Rostedt wrote:
> On Sat, 2010-03-20 at 16:19 +0300, Dan Carpenter wrote:
> > The original code doesn't work because "call" is never NULL there.
> 
> A one line change would have done the same ;-)
> 
> > Signed-off-by: Dan Carpenter <error27@gmail.com>
> > ---
> > My code should work, but it seems like there should but it seems like
> > there should be a more elegant way to do this?
> > 
> > diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c
> > index 4615f62..6070c70 100644
> > --- a/kernel/trace/trace_events_filter.c
> > +++ b/kernel/trace/trace_events_filter.c
> > @@ -1388,16 +1388,19 @@ int ftrace_profile_set_filter(struct perf_event *event, int event_id,
> >  	struct event_filter *filter;
> >  	struct filter_parse_state *ps;
> >  	struct ftrace_event_call *call = NULL;
> > +	int found = 0;
> >  
> >  	mutex_lock(&event_mutex);
> >  
> >  	list_for_each_entry(call, &ftrace_events, list) {
> > -		if (call->id == event_id)
> > +		if (call->id == event_id) {
> > +			found = 1;
> >  			break;
> > +		}
> >  	}
> >  
> >  	err = -EINVAL;
> > -	if (!call)
> 
> 	if (call == &ftrace_events)
> 

Man...  That makes my head hurt.  It only works because ->list is the 
first element in the struct.

We'd need a cast.

Is that really cleaner?

regards,
dan carpenter

> -- Steve
> 
> > +	if (!found)
> >  		goto out_unlock;
> >  
> >  	err = -EEXIST;
> 
> 
> --
> 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/

Re: [patch] trace: find the correct ftrace event

[Steven Rostedt]

On Sat, 2010-03-20 at 16:56 +0300, Dan Carpenter wrote:

> > > diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c
> > > index 4615f62..6070c70 100644
> > > --- a/kernel/trace/trace_events_filter.c
> > > +++ b/kernel/trace/trace_events_filter.c
> > > @@ -1388,16 +1388,19 @@ int ftrace_profile_set_filter(struct perf_event *event, int event_id,
> > >  	struct event_filter *filter;
> > >  	struct filter_parse_state *ps;
> > >  	struct ftrace_event_call *call = NULL;
> > > +	int found = 0;
> > >  
> > >  	mutex_lock(&event_mutex);
> > >  
> > >  	list_for_each_entry(call, &ftrace_events, list) {
> > > -		if (call->id == event_id)
> > > +		if (call->id == event_id) {
> > > +			found = 1;
> > >  			break;
> > > +		}
> > >  	}
> > >  
> > >  	err = -EINVAL;
> > > -	if (!call)
> > 
> > 	if (call == &ftrace_events)
> > 
> 
> Man...  That makes my head hurt.  It only works because ->list is the 
> first element in the struct.
> 
> We'd need a cast.

Oops, I missed the fact that it was list_*_entry(). That should be:

	if (call->list != &ftrace_events)

No cast needed.

> 
> Is that really cleaner?

Perhaps not, but I think it is more elegant ;-)

-- Steve

Re: [patch] trace: find the correct ftrace event

[Dan Carpenter]

On Sat, Mar 20, 2010 at 10:07:16AM -0400, Steven Rostedt wrote:
> On Sat, 2010-03-20 at 16:56 +0300, Dan Carpenter wrote:
> 
> > > > diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c
> > > > index 4615f62..6070c70 100644
> > > > --- a/kernel/trace/trace_events_filter.c
> > > > +++ b/kernel/trace/trace_events_filter.c
> > > > @@ -1388,16 +1388,19 @@ int ftrace_profile_set_filter(struct perf_event *event, int event_id,
> > > >  	struct event_filter *filter;
> > > >  	struct filter_parse_state *ps;
> > > >  	struct ftrace_event_call *call = NULL;
> > > > +	int found = 0;
> > > >  
> > > >  	mutex_lock(&event_mutex);
> > > >  
> > > >  	list_for_each_entry(call, &ftrace_events, list) {
> > > > -		if (call->id == event_id)
> > > > +		if (call->id == event_id) {
> > > > +			found = 1;
> > > >  			break;
> > > > +		}
> > > >  	}
> > > >  
> > > >  	err = -EINVAL;
> > > > -	if (!call)
> > > 
> > > 	if (call == &ftrace_events)
> > > 
> > 
> > Man...  That makes my head hurt.  It only works because ->list is the 
> > first element in the struct.
> > 
> > We'd need a cast.
> 
> Oops, I missed the fact that it was list_*_entry(). That should be:
> 
> 	if (call->list != &ftrace_events)
> 
> No cast needed.
> 

Hm...  "list" is a struct not a pointer to a struct.

So it would have to be:

        if (&call->list == &ftrace_events)
                goto out_unlock;

That's how the list_for_each_entry() determines the end of the list as
well.

I'll send a patch to do that.

regards,
dan carpenter

> > 
> > Is that really cleaner?
> 
> Perhaps not, but I think it is more elegant ;-)
> 
> -- Steve

[patch v2] trace: find the correct ftrace event

[Dan Carpenter]

The original code doesn't work because "call" is never NULL there.                                                           
                                                                                                                             
Signed-off-by: Dan Carpenter <error27@gmail.com>                                                                             
---
The original patch was fine, but this version is more tasteful.  ;)

diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c
index 4615f62..c1a263a 100644
--- a/kernel/trace/trace_events_filter.c
+++ b/kernel/trace/trace_events_filter.c
@@ -1397,7 +1397,7 @@ int ftrace_profile_set_filter(struct perf_event *event, int event_id,
 	}
 
 	err = -EINVAL;
-	if (!call)
+	if (&call->list == &ftrace_events)
 		goto out_unlock;
 
 	err = -EEXIST;

Re: [patch v2] trace: find the correct ftrace event

[Steven Rostedt]

On Sat, 2010-03-20 at 17:39 +0300, Dan Carpenter wrote:
> The original code doesn't work because "call" is never NULL there.                                                           
>                                                                                                                              
> Signed-off-by: Dan Carpenter <error27@gmail.com>     

Thanks, I'll apply and test this on Monday.

-- Steve

>                                                                         
> ---
> The original patch was fine, but this version is more tasteful.  ;)
> 
> diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c
> index 4615f62..c1a263a 100644
> --- a/kernel/trace/trace_events_filter.c
> +++ b/kernel/trace/trace_events_filter.c
> @@ -1397,7 +1397,7 @@ int ftrace_profile_set_filter(struct perf_event *event, int event_id,
>  	}
>  
>  	err = -EINVAL;
> -	if (!call)
> +	if (&call->list == &ftrace_events)
>  		goto out_unlock;
>  
>  	err = -EEXIST;

[tip:perf/core] perf: Fix check at end of event search

[tip-bot for Dan Carpenter]

Commit-ID:  d9f599e1e6d019968b35d2dc63074b9e8964fa69
Gitweb:     http://git.kernel.org/tip/d9f599e1e6d019968b35d2dc63074b9e8964fa69
Author:     Dan Carpenter <error27@gmail.com>
AuthorDate: Sat, 20 Mar 2010 17:39:11 +0300
Committer:  Steven Rostedt <rostedt@goodmis.org>
CommitDate: Thu, 6 May 2010 19:49:52 -0400

perf: Fix check at end of event search

The original code doesn't work because "call" is never NULL there.

Signed-off-by: Dan Carpenter <error27@gmail.com>
LKML-Reference: <20100320143911.GF5331@bicker>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
---
 kernel/trace/trace_events_filter.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/kernel/trace/trace_events_filter.c b/kernel/trace/trace_events_filter.c
index 88c0b6d..58092d8 100644
--- a/kernel/trace/trace_events_filter.c
+++ b/kernel/trace/trace_events_filter.c
@@ -1398,7 +1398,7 @@ int ftrace_profile_set_filter(struct perf_event *event, int event_id,
 	}
 
 	err = -EINVAL;
-	if (!call)
+	if (&call->list == &ftrace_events)
 		goto out_unlock;
 
 	err = -EEXIST;