[PATCH 1/2] tracing/kprobes: move free_trace_probe into unregister_trace_probe

[PATCH 1/2] tracing/kprobes: move free_trace_probe into unregister_trace_probe

[zhangwei(Jovi)]

There have no good reason to call free_trace_probe
every time when unregister_trace_probe return 0.

Move free_trace_probe into unregister_trace_probe,
make code simpler.

Signed-off-by: zhangwei(Jovi) <jovi.zhangwei@huawei.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Oleg Nesterov <oleg@redhat.com>
---
 kernel/trace/trace_kprobe.c |    5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c
index 9f46e98..f193c38 100644
--- a/kernel/trace/trace_kprobe.c
+++ b/kernel/trace/trace_kprobe.c
@@ -401,6 +401,7 @@ static int unregister_trace_probe(struct trace_probe *tp)
 	__unregister_trace_probe(tp);
 	list_del(&tp->list);
 	unregister_probe_event(tp);
+	free_trace_probe(tp);

 	return 0;
 }
@@ -419,7 +420,6 @@ static int register_trace_probe(struct trace_probe *tp)
 		ret = unregister_trace_probe(old_tp);
 		if (ret < 0)
 			goto end;
-		free_trace_probe(old_tp);
 	}

 	/* Register new event */
@@ -550,8 +550,6 @@ static int create_trace_probe(int argc, char **argv)
 		}
 		/* delete an event */
 		ret = unregister_trace_probe(tp);
-		if (ret == 0)
-			free_trace_probe(tp);
 		mutex_unlock(&probe_lock);
 		return ret;
 	}
@@ -680,7 +678,6 @@ static int release_all_trace_probes(void)
 	while (!list_empty(&probe_list)) {
 		tp = list_entry(probe_list.next, struct trace_probe, list);
 		unregister_trace_probe(tp);
-		free_trace_probe(tp);
 	}

 end:
-- 
1.7.9.7

Re: [PATCH 1/2] tracing/kprobes: move free_trace_probe into unregister_trace_probe

[Masami Hiramatsu]

(2013/06/25 17:15), zhangwei(Jovi) wrote:
> There have no good reason to call free_trace_probe
> every time when unregister_trace_probe return 0.
> 
> Move free_trace_probe into unregister_trace_probe,
> make code simpler.

Sorry, nack. For the symmetrical coding reason, I don't like
involving "free" and "alloc" into "unregister"/"register"
functions. I think those should be just another actions.

Thank you,

> 
> Signed-off-by: zhangwei(Jovi) <jovi.zhangwei@huawei.com>
> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
> Cc: Oleg Nesterov <oleg@redhat.com>
> ---
>  kernel/trace/trace_kprobe.c |    5 +----
>  1 file changed, 1 insertion(+), 4 deletions(-)
> 
> diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c
> index 9f46e98..f193c38 100644
> --- a/kernel/trace/trace_kprobe.c
> +++ b/kernel/trace/trace_kprobe.c
> @@ -401,6 +401,7 @@ static int unregister_trace_probe(struct trace_probe *tp)
>  	__unregister_trace_probe(tp);
>  	list_del(&tp->list);
>  	unregister_probe_event(tp);
> +	free_trace_probe(tp);
> 
>  	return 0;
>  }
> @@ -419,7 +420,6 @@ static int register_trace_probe(struct trace_probe *tp)
>  		ret = unregister_trace_probe(old_tp);
>  		if (ret < 0)
>  			goto end;
> -		free_trace_probe(old_tp);
>  	}
> 
>  	/* Register new event */
> @@ -550,8 +550,6 @@ static int create_trace_probe(int argc, char **argv)
>  		}
>  		/* delete an event */
>  		ret = unregister_trace_probe(tp);
> -		if (ret == 0)
> -			free_trace_probe(tp);
>  		mutex_unlock(&probe_lock);
>  		return ret;
>  	}
> @@ -680,7 +678,6 @@ static int release_all_trace_probes(void)
>  	while (!list_empty(&probe_list)) {
>  		tp = list_entry(probe_list.next, struct trace_probe, list);
>  		unregister_trace_probe(tp);
> -		free_trace_probe(tp);
>  	}
> 
>  end:
> 


-- 
Masami HIRAMATSU
IT Management Research Dept. Linux Technology Center
Hitachi, Ltd., Yokohama Research Laboratory
E-mail: masami.hiramatsu.pt@hitachi.com

Re: [PATCH 1/2] tracing/kprobes: move free_trace_probe into unregister_trace_probe

[zhangwei(Jovi)]

On 2013/6/25 18:10, Masami Hiramatsu wrote:
> (2013/06/25 17:15), zhangwei(Jovi) wrote:
>> There have no good reason to call free_trace_probe
>> every time when unregister_trace_probe return 0.
>>
>> Move free_trace_probe into unregister_trace_probe,
>> make code simpler.
> 
> Sorry, nack. For the symmetrical coding reason, I don't like
> involving "free" and "alloc" into "unregister"/"register"
> functions. I think those should be just another actions.
> 
> Thank you,

That's fine, I just saw there have a little inconsistent between
trace_kprobe.c and trace_uprobe.c.

Please ignore this patch if you don't like. :)
Thanks.

> 
>>
>> Signed-off-by: zhangwei(Jovi) <jovi.zhangwei@huawei.com>
>> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
>> Cc: Oleg Nesterov <oleg@redhat.com>
>> ---
>>  kernel/trace/trace_kprobe.c |    5 +----
>>  1 file changed, 1 insertion(+), 4 deletions(-)
>>
>> diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c
>> index 9f46e98..f193c38 100644
>> --- a/kernel/trace/trace_kprobe.c
>> +++ b/kernel/trace/trace_kprobe.c
>> @@ -401,6 +401,7 @@ static int unregister_trace_probe(struct trace_probe *tp)
>>  	__unregister_trace_probe(tp);
>>  	list_del(&tp->list);
>>  	unregister_probe_event(tp);
>> +	free_trace_probe(tp);
>>
>>  	return 0;
>>  }
>> @@ -419,7 +420,6 @@ static int register_trace_probe(struct trace_probe *tp)
>>  		ret = unregister_trace_probe(old_tp);
>>  		if (ret < 0)
>>  			goto end;
>> -		free_trace_probe(old_tp);
>>  	}
>>
>>  	/* Register new event */
>> @@ -550,8 +550,6 @@ static int create_trace_probe(int argc, char **argv)
>>  		}
>>  		/* delete an event */
>>  		ret = unregister_trace_probe(tp);
>> -		if (ret == 0)
>> -			free_trace_probe(tp);
>>  		mutex_unlock(&probe_lock);
>>  		return ret;
>>  	}
>> @@ -680,7 +678,6 @@ static int release_all_trace_probes(void)
>>  	while (!list_empty(&probe_list)) {
>>  		tp = list_entry(probe_list.next, struct trace_probe, list);
>>  		unregister_trace_probe(tp);
>> -		free_trace_probe(tp);
>>  	}
>>
>>  end:
>>
> 
> 

Re: [PATCH 1/2] tracing/kprobes: move free_trace_probe into unregister_trace_probe

[Steven Rostedt]

On Tue, 2013-06-25 at 18:37 +0800, zhangwei(Jovi) wrote:
> On 2013/6/25 18:10, Masami Hiramatsu wrote:
> > (2013/06/25 17:15), zhangwei(Jovi) wrote:
> >> There have no good reason to call free_trace_probe
> >> every time when unregister_trace_probe return 0.
> >>
> >> Move free_trace_probe into unregister_trace_probe,
> >> make code simpler.
> > 
> > Sorry, nack. For the symmetrical coding reason, I don't like
> > involving "free" and "alloc" into "unregister"/"register"
> > functions. I think those should be just another actions.
> > 
> > Thank you,
> 
> That's fine, I just saw there have a little inconsistent between
> trace_kprobe.c and trace_uprobe.c.
> 

Is there a place that trace_kprobe.c frees the tp structure in
unregister?

-- Steve

Re: [PATCH 1/2] tracing/kprobes: move free_trace_probe into unregister_trace_probe

[Jovi Zhang]

On Tue, Jun 25, 2013 at 9:34 PM, Steven Rostedt <rostedt@goodmis.org> wrote:
> On Tue, 2013-06-25 at 18:37 +0800, zhangwei(Jovi) wrote:
>> On 2013/6/25 18:10, Masami Hiramatsu wrote:
>> > (2013/06/25 17:15), zhangwei(Jovi) wrote:
>> >> There have no good reason to call free_trace_probe
>> >> every time when unregister_trace_probe return 0.
>> >>
>> >> Move free_trace_probe into unregister_trace_probe,
>> >> make code simpler.
>> >
>> > Sorry, nack. For the symmetrical coding reason, I don't like
>> > involving "free" and "alloc" into "unregister"/"register"
>> > functions. I think those should be just another actions.
>> >
>> > Thank you,
>>
>> That's fine, I just saw there have a little inconsistent between
>> trace_kprobe.c and trace_uprobe.c.
>>
>
> Is there a place that trace_kprobe.c frees the tp structure in
> unregister?
>

I won't argue put the free operation into unregister function in
trace_kprobe.c, as I said, one minor problem is the code
pattern between trace_kprobe.c and trace_uprobe.c is so
similar on unregister, but with little inconsistent, maybe
we can unify those code in trace_probe.c someday.

Anyway, this is not big functionality issue,  free to leave in there
if authors don't argue.

jovi

Re: [PATCH 1/2] tracing/kprobes: move free_trace_probe into unregister_trace_probe

[Steven Rostedt]

On Tue, 2013-06-25 at 19:10 +0900, Masami Hiramatsu wrote:
> (2013/06/25 17:15), zhangwei(Jovi) wrote:
> > There have no good reason to call free_trace_probe
> > every time when unregister_trace_probe return 0.
> > 
> > Move free_trace_probe into unregister_trace_probe,
> > make code simpler.
> 
> Sorry, nack. For the symmetrical coding reason, I don't like
> involving "free" and "alloc" into "unregister"/"register"
> functions. I think those should be just another actions.

I totally agree with Masami here. Please do not add "side effects" to
the unregister function.

Thanks,

-- Steve