[PATCH] perf probe: Return errno when does not hit any event

[PATCH] perf probe: Return errno when does not hit any event

[Kefeng Wang]

On old perf, when using perf probe -d to delete an inexistent event,
it return errno, eg,

-bash-4.3# perf probe -d xxx  || echo $?
Info: Event "*:xxx" does not exist.
  Error: Failed to delete events.
255

But now perf_del_probe_events() will always set ret = 0, different
from previous del_perf_probe_events(). After this, it return errno
again, eg,

-bash-4.3# ./perf probe -d xxx  || echo $?
  Error: Failed to delete events.
254

And it is more appropriate to return -ENOENT instead of -EPERM.

Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
---
 tools/perf/builtin-probe.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c
index 1fcebc3..c46b41c 100644
--- a/tools/perf/builtin-probe.c
+++ b/tools/perf/builtin-probe.c
@@ -444,7 +444,8 @@ static int perf_del_probe_events(struct strfilter *filter)
 	if (ret == -ENOENT && ret2 == -ENOENT)
 		pr_debug("\"%s\" does not hit any event.\n", str);
 		/* Note that this is silently ignored */
-	ret = 0;
+	else
+		ret = 0;
 
 error:
 	if (kfd >= 0)
-- 
1.7.12.4

Re: [PATCH] perf probe: Return errno when does not hit any event

[Kefeng Wang]

+ Arnaldo Carvalho de Melo <acme@kernel.org>

On 2017/3/6 17:34, Kefeng Wang wrote:
> On old perf, when using perf probe -d to delete an inexistent event,
> it return errno, eg,
> 
> -bash-4.3# perf probe -d xxx  || echo $?
> Info: Event "*:xxx" does not exist.
>   Error: Failed to delete events.
> 255
> 
> But now perf_del_probe_events() will always set ret = 0, different
> from previous del_perf_probe_events(). After this, it return errno
> again, eg,
> 
> -bash-4.3# ./perf probe -d xxx  || echo $?
>   Error: Failed to delete events.
> 254
> 
> And it is more appropriate to return -ENOENT instead of -EPERM.
> 
> Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
> ---
>  tools/perf/builtin-probe.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c
> index 1fcebc3..c46b41c 100644
> --- a/tools/perf/builtin-probe.c
> +++ b/tools/perf/builtin-probe.c
> @@ -444,7 +444,8 @@ static int perf_del_probe_events(struct strfilter *filter)
>  	if (ret == -ENOENT && ret2 == -ENOENT)
>  		pr_debug("\"%s\" does not hit any event.\n", str);
>  		/* Note that this is silently ignored */
> -	ret = 0;
> +	else
> +		ret = 0;
>  
>  error:
>  	if (kfd >= 0)
> 

Re: [PATCH] perf probe: Return errno when does not hit any event

[Kefeng Wang]

Hi all, any comments, thanks.

On 2017/3/7 15:33, Kefeng Wang wrote:
> + Arnaldo Carvalho de Melo <acme@kernel.org>
> 
> On 2017/3/6 17:34, Kefeng Wang wrote:
>> On old perf, when using perf probe -d to delete an inexistent event,
>> it return errno, eg,
>>
>> -bash-4.3# perf probe -d xxx  || echo $?
>> Info: Event "*:xxx" does not exist.
>>   Error: Failed to delete events.
>> 255
>>
>> But now perf_del_probe_events() will always set ret = 0, different
>> from previous del_perf_probe_events(). After this, it return errno
>> again, eg,
>>
>> -bash-4.3# ./perf probe -d xxx  || echo $?
>>   Error: Failed to delete events.
>> 254
>>
>> And it is more appropriate to return -ENOENT instead of -EPERM.
>>
>> Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
>> ---
>>  tools/perf/builtin-probe.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c
>> index 1fcebc3..c46b41c 100644
>> --- a/tools/perf/builtin-probe.c
>> +++ b/tools/perf/builtin-probe.c
>> @@ -444,7 +444,8 @@ static int perf_del_probe_events(struct strfilter *filter)
>>  	if (ret == -ENOENT && ret2 == -ENOENT)
>>  		pr_debug("\"%s\" does not hit any event.\n", str);
>>  		/* Note that this is silently ignored */
>> -	ret = 0;
>> +	else
>> +		ret = 0;
>>  
>>  error:
>>  	if (kfd >= 0)
>>
> 
> 
> .
> 

Re: [PATCH] perf probe: Return errno when does not hit any event

[Arnaldo Carvalho de Melo]

Em Tue, Mar 14, 2017 at 09:19:47PM +0800, Kefeng Wang escreveu:
> Hi all, any comments, thanks.

For 'perf probe' make sure Masami is in the CC list, adding him, Masami?

- Arnaldo
 
> On 2017/3/7 15:33, Kefeng Wang wrote:
> > + Arnaldo Carvalho de Melo <acme@kernel.org>
> > 
> > On 2017/3/6 17:34, Kefeng Wang wrote:
> >> On old perf, when using perf probe -d to delete an inexistent event,
> >> it return errno, eg,
> >>
> >> -bash-4.3# perf probe -d xxx  || echo $?
> >> Info: Event "*:xxx" does not exist.
> >>   Error: Failed to delete events.
> >> 255
> >>
> >> But now perf_del_probe_events() will always set ret = 0, different
> >> from previous del_perf_probe_events(). After this, it return errno
> >> again, eg,
> >>
> >> -bash-4.3# ./perf probe -d xxx  || echo $?
> >>   Error: Failed to delete events.
> >> 254
> >>
> >> And it is more appropriate to return -ENOENT instead of -EPERM.
> >>
> >> Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
> >> ---
> >>  tools/perf/builtin-probe.c | 3 ++-
> >>  1 file changed, 2 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c
> >> index 1fcebc3..c46b41c 100644
> >> --- a/tools/perf/builtin-probe.c
> >> +++ b/tools/perf/builtin-probe.c
> >> @@ -444,7 +444,8 @@ static int perf_del_probe_events(struct strfilter *filter)
> >>  	if (ret == -ENOENT && ret2 == -ENOENT)
> >>  		pr_debug("\"%s\" does not hit any event.\n", str);
> >>  		/* Note that this is silently ignored */
> >> -	ret = 0;
> >> +	else
> >> +		ret = 0;
> >>  
> >>  error:
> >>  	if (kfd >= 0)
> >>
> > 
> > 
> > .
> > 

Re: [PATCH] perf probe: Return errno when does not hit any event

[Masami Hiramatsu]

On Tue, 14 Mar 2017 10:30:56 -0300
Arnaldo Carvalho de Melo <acme@kernel.org> wrote:

> Em Tue, Mar 14, 2017 at 09:19:47PM +0800, Kefeng Wang escreveu:
> > Hi all, any comments, thanks.
> 
> For 'perf probe' make sure Masami is in the CC list, adding him, Masami?

Thanks Arnaldo,

Hi Kefeng, I've made the error ignored by design, since user might pass
the wildcard pattern to perf probe -d just for making sure.

You can check the pattern hits any event by using perf-probe --list PATTERN.
Would you have any use case to check the result?

Thank you,

> 
> - Arnaldo
>  
> > On 2017/3/7 15:33, Kefeng Wang wrote:
> > > + Arnaldo Carvalho de Melo <acme@kernel.org>
> > > 
> > > On 2017/3/6 17:34, Kefeng Wang wrote:
> > >> On old perf, when using perf probe -d to delete an inexistent event,
> > >> it return errno, eg,
> > >>
> > >> -bash-4.3# perf probe -d xxx  || echo $?
> > >> Info: Event "*:xxx" does not exist.
> > >>   Error: Failed to delete events.
> > >> 255
> > >>
> > >> But now perf_del_probe_events() will always set ret = 0, different
> > >> from previous del_perf_probe_events(). After this, it return errno
> > >> again, eg,
> > >>
> > >> -bash-4.3# ./perf probe -d xxx  || echo $?
> > >>   Error: Failed to delete events.
> > >> 254
> > >>
> > >> And it is more appropriate to return -ENOENT instead of -EPERM.
> > >>
> > >> Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
> > >> ---
> > >>  tools/perf/builtin-probe.c | 3 ++-
> > >>  1 file changed, 2 insertions(+), 1 deletion(-)
> > >>
> > >> diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c
> > >> index 1fcebc3..c46b41c 100644
> > >> --- a/tools/perf/builtin-probe.c
> > >> +++ b/tools/perf/builtin-probe.c
> > >> @@ -444,7 +444,8 @@ static int perf_del_probe_events(struct strfilter *filter)
> > >>  	if (ret == -ENOENT && ret2 == -ENOENT)
> > >>  		pr_debug("\"%s\" does not hit any event.\n", str);
> > >>  		/* Note that this is silently ignored */
> > >> -	ret = 0;
> > >> +	else
> > >> +		ret = 0;
> > >>  
> > >>  error:
> > >>  	if (kfd >= 0)
> > >>
> > > 
> > > 
> > > .
> > > 


-- 
Masami Hiramatsu <mhiramat@kernel.org>

Re: [PATCH] perf probe: Return errno when does not hit any event

[Kefeng Wang]

On 2017/3/16 17:39, Masami Hiramatsu wrote:
> On Tue, 14 Mar 2017 10:30:56 -0300
> Arnaldo Carvalho de Melo <acme@kernel.org> wrote:
> 
>> Em Tue, Mar 14, 2017 at 09:19:47PM +0800, Kefeng Wang escreveu:
>>> Hi all, any comments, thanks.
>>
>> For 'perf probe' make sure Masami is in the CC list, adding him, Masami?
> 
> Thanks Arnaldo,
> 
> Hi Kefeng, I've made the error ignored by design, since user might pass
> the wildcard pattern to perf probe -d just for making sure.
> 
> You can check the pattern hits any event by using perf-probe --list PATTERN.
> Would you have any use case to check the result?

There are some perf test cases from our inner tester, they will check
the returned value for each execution, and they think return errno is
better to cater for users when delete inexistent events, as it was before.
And the patch only returns error when does not hit any event.

If such behavior, error ignored is more reasonableo<\f I will push them to
modify the test cases.

Thanks,
Kefeng

> 
> Thank you,
> 
>>
>> - Arnaldo
>>  
>>> On 2017/3/7 15:33, Kefeng Wang wrote:
>>>> + Arnaldo Carvalho de Melo <acme@kernel.org>
>>>>
>>>> On 2017/3/6 17:34, Kefeng Wang wrote:
>>>>> On old perf, when using perf probe -d to delete an inexistent event,
>>>>> it return errno, eg,
>>>>>
>>>>> -bash-4.3# perf probe -d xxx  || echo $?
>>>>> Info: Event "*:xxx" does not exist.
>>>>>   Error: Failed to delete events.
>>>>> 255
>>>>>
>>>>> But now perf_del_probe_events() will always set ret = 0, different
>>>>> from previous del_perf_probe_events(). After this, it return errno
>>>>> again, eg,
>>>>>
>>>>> -bash-4.3# ./perf probe -d xxx  || echo $?
>>>>>   Error: Failed to delete events.
>>>>> 254
>>>>>
>>>>> And it is more appropriate to return -ENOENT instead of -EPERM.
>>>>>
>>>>> Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
>>>>> ---
>>>>>  tools/perf/builtin-probe.c | 3 ++-
>>>>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c
>>>>> index 1fcebc3..c46b41c 100644
>>>>> --- a/tools/perf/builtin-probe.c
>>>>> +++ b/tools/perf/builtin-probe.c
>>>>> @@ -444,7 +444,8 @@ static int perf_del_probe_events(struct strfilter *filter)
>>>>>  	if (ret == -ENOENT && ret2 == -ENOENT)
>>>>>  		pr_debug("\"%s\" does not hit any event.\n", str);
>>>>>  		/* Note that this is silently ignored */
>>>>> -	ret = 0;
>>>>> +	else
>>>>> +		ret = 0;
>>>>>  
>>>>>  error:
>>>>>  	if (kfd >= 0)
>>>>>
>>>>
>>>>
>>>> .
>>>>
> 
> 

Re: [PATCH] perf probe: Return errno when does not hit any event

[Masami Hiramatsu]

On Thu, 16 Mar 2017 20:07:08 +0800
Kefeng Wang <wangkefeng.wang@huawei.com> wrote:

> 
> 
> On 2017/3/16 17:39, Masami Hiramatsu wrote:
> > On Tue, 14 Mar 2017 10:30:56 -0300
> > Arnaldo Carvalho de Melo <acme@kernel.org> wrote:
> > 
> >> Em Tue, Mar 14, 2017 at 09:19:47PM +0800, Kefeng Wang escreveu:
> >>> Hi all, any comments, thanks.
> >>
> >> For 'perf probe' make sure Masami is in the CC list, adding him, Masami?
> > 
> > Thanks Arnaldo,
> > 
> > Hi Kefeng, I've made the error ignored by design, since user might pass
> > the wildcard pattern to perf probe -d just for making sure.
> > 
> > You can check the pattern hits any event by using perf-probe --list PATTERN.
> > Would you have any use case to check the result?
> 
> There are some perf test cases from our inner tester, they will check
> the returned value for each execution, and they think return errno is
> better to cater for users when delete inexistent events, as it was before.
> And the patch only returns error when does not hit any event.
> 
> If such behavior, error ignored is more reasonableo<\f I will push them to
> modify the test cases.

Hmm, the commit dddc7ee32fa1 ("perf probe: Fix an error when deleting probes successfully") introduced this behavior. But I think this fix will not be
the best solution.

OK, so could you update your patch according my comment and resend it?


> >>>>> diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c
> >>>>> index 1fcebc3..c46b41c 100644
> >>>>> --- a/tools/perf/builtin-probe.c
> >>>>> +++ b/tools/perf/builtin-probe.c
> >>>>> @@ -444,7 +444,8 @@ static int perf_del_probe_events(struct strfilter *filter)
> >>>>>  	if (ret == -ENOENT && ret2 == -ENOENT)
> >>>>>  		pr_debug("\"%s\" does not hit any event.\n", str);

This should be pr_warning if this case has an error.

> >>>>>  		/* Note that this is silently ignored */

This comment should be removed too, since this error is not ignored anymore.


> >>>>> -	ret = 0;
> >>>>> +	else
> >>>>> +		ret = 0;
> >>>>>  

With these patches, perf probe shows an error if there is no event hit
for --delete.

$ sudo ./perf probe -d hoge
"hoge" does not hit any event.
  Error: Failed to delete events.


Thanks,


> >>>>>  error:
> >>>>>  	if (kfd >= 0)
> >>>>>
> >>>>
> >>>>
> >>>> .
> >>>>
> > 
> > 
> 


-- 
Masami Hiramatsu <mhiramat@kernel.org>

[PATCH v2] perf probe: Return errno when does not hit any event

[Kefeng Wang]

On old perf, when using perf probe -d to delete an inexistent event,
it return errno, eg,

-bash-4.3# perf probe -d xxx  || echo $?
Info: Event "*:xxx" does not exist.
  Error: Failed to delete events.
255

But now perf_del_probe_events() will always set ret = 0, different
from previous del_perf_probe_events(). After this, it return errno
again, eg,

-bash-4.3# ./perf probe -d xxx  || echo $?
"xxx" does not hit any event.
  Error: Failed to delete events.
254

And it is more appropriate to return -ENOENT instead of -EPERM.

Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Wang Nan <wangnan0@huawei.com>

Fixes: dddc7ee32fa1 ("perf probe: Fix an error when deleting probes successfully")
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
---

v1->v2:
- Using pr_wanring to show warning infos and drop inappropriate comment
  suggested by Masami Hiramatsu.

 tools/perf/builtin-probe.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c
index 1fcebc3..51cdc23 100644
--- a/tools/perf/builtin-probe.c
+++ b/tools/perf/builtin-probe.c
@@ -442,9 +442,9 @@ static int perf_del_probe_events(struct strfilter *filter)
 	}
 
 	if (ret == -ENOENT && ret2 == -ENOENT)
-		pr_debug("\"%s\" does not hit any event.\n", str);
-		/* Note that this is silently ignored */
-	ret = 0;
+		pr_warning("\"%s\" does not hit any event.\n", str);
+	else
+		ret = 0;
 
 error:
 	if (kfd >= 0)
-- 
1.7.12.4

Re: [PATCH v2] perf probe: Return errno when does not hit any event

[Masami Hiramatsu]

On Fri, 17 Mar 2017 16:16:32 +0800
Kefeng Wang <wangkefeng.wang@huawei.com> wrote:

> On old perf, when using perf probe -d to delete an inexistent event,
> it return errno, eg,
> 
> -bash-4.3# perf probe -d xxx  || echo $?
> Info: Event "*:xxx" does not exist.
>   Error: Failed to delete events.
> 255
> 
> But now perf_del_probe_events() will always set ret = 0, different
> from previous del_perf_probe_events(). After this, it return errno
> again, eg,
> 
> -bash-4.3# ./perf probe -d xxx  || echo $?
> "xxx" does not hit any event.
>   Error: Failed to delete events.
> 254
> 
> And it is more appropriate to return -ENOENT instead of -EPERM.

Looks good to me.

Acked-by: Masami Hiramatsu <mhiramat@kernel.org>

Thanks!

> 
> Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
> Cc: Masami Hiramatsu <mhiramat@kernel.org>
> Cc: Jiri Olsa <jolsa@kernel.org>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Wang Nan <wangnan0@huawei.com>
> 
> Fixes: dddc7ee32fa1 ("perf probe: Fix an error when deleting probes successfully")
> Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
> ---
> 
> v1->v2:
> - Using pr_wanring to show warning infos and drop inappropriate comment
>   suggested by Masami Hiramatsu.
> 
>  tools/perf/builtin-probe.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c
> index 1fcebc3..51cdc23 100644
> --- a/tools/perf/builtin-probe.c
> +++ b/tools/perf/builtin-probe.c
> @@ -442,9 +442,9 @@ static int perf_del_probe_events(struct strfilter *filter)
>  	}
>  
>  	if (ret == -ENOENT && ret2 == -ENOENT)
> -		pr_debug("\"%s\" does not hit any event.\n", str);
> -		/* Note that this is silently ignored */
> -	ret = 0;
> +		pr_warning("\"%s\" does not hit any event.\n", str);
> +	else
> +		ret = 0;
>  
>  error:
>  	if (kfd >= 0)
> -- 
> 1.7.12.4
> 


-- 
Masami Hiramatsu <mhiramat@kernel.org>

Re: [PATCH v2] perf probe: Return errno when does not hit any event

[Arnaldo Carvalho de Melo]

Em Sat, Mar 18, 2017 at 08:23:02AM +0900, Masami Hiramatsu escreveu:
> On Fri, 17 Mar 2017 16:16:32 +0800 Kefeng Wang <wangkefeng.wang@huawei.com> wrote:
> > And it is more appropriate to return -ENOENT instead of -EPERM.
 
> Looks good to me.
 
> Acked-by: Masami Hiramatsu <mhiramat@kernel.org>

Thanks, applied.

- Arnaldo

[tip:perf/core] perf probe: Return errno when not hitting any event

[tip-bot for Kefeng Wang]

Commit-ID:  70946723eeb859466f026274b29c6196e39149c4
Gitweb:     http://git.kernel.org/tip/70946723eeb859466f026274b29c6196e39149c4
Author:     Kefeng Wang <wangkefeng.wang@huawei.com>
AuthorDate: Fri, 17 Mar 2017 16:16:32 +0800
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Tue, 21 Mar 2017 10:45:02 -0300

perf probe: Return errno when not hitting any event

On old perf, when using 'perf probe -d' to delete an inexistent event,
it returns errno, eg,

  -bash-4.3# perf probe -d xxx  || echo $?
  Info: Event "*:xxx" does not exist.
    Error: Failed to delete events.
  255

But now perf_del_probe_events() will always set ret = 0, different from
previous del_perf_probe_events(). After this, it returns errno again,
eg,

  -bash-4.3# ./perf probe -d xxx  || echo $?
  "xxx" does not hit any event.
    Error: Failed to delete events.
  254

And it is more appropriate to return -ENOENT instead of -EPERM.

Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Hanjun Guo <guohanjun@huawei.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Wang Nan <wangnan0@huawei.com>
Fixes: dddc7ee32fa1 ("perf probe: Fix an error when deleting probes successfully")
Link: http://lkml.kernel.org/r/1489738592-61011-1-git-send-email-wangkefeng.wang@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-probe.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c
index 1fcebc3..51cdc23 100644
--- a/tools/perf/builtin-probe.c
+++ b/tools/perf/builtin-probe.c
@@ -442,9 +442,9 @@ static int perf_del_probe_events(struct strfilter *filter)
 	}
 
 	if (ret == -ENOENT && ret2 == -ENOENT)
-		pr_debug("\"%s\" does not hit any event.\n", str);
-		/* Note that this is silently ignored */
-	ret = 0;
+		pr_warning("\"%s\" does not hit any event.\n", str);
+	else
+		ret = 0;
 
 error:
 	if (kfd >= 0)