[v1] libbpf: remove sample_period init in perf_buffer

[RFC PATCH bpf-next] libbpf: remove sample_period init in perf_buffer

Posted by Tao Chen 9 months, 3 weeks ago

It seems that sample_period no used in perf buffer, actually only
wakeup_events valid about events aggregation for wakeup. So remove
it to avoid causing confusion.

Fixes: fb84b8224655 ("libbpf: add perf buffer API")
Signed-off-by: Tao Chen <chen.dylane@linux.dev>
---
 tools/lib/bpf/libbpf.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index 194809da5172..1830e3c011a5 100644
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c
@@ -13306,7 +13306,6 @@ struct perf_buffer *perf_buffer__new(int map_fd, size_t page_cnt,
 	attr.config = PERF_COUNT_SW_BPF_OUTPUT;
 	attr.type = PERF_TYPE_SOFTWARE;
 	attr.sample_type = PERF_SAMPLE_RAW;
-	attr.sample_period = sample_period;
 	attr.wakeup_events = sample_period;
 
 	p.attr = &attr;
-- 
2.43.0

Re: [RFC PATCH bpf-next] libbpf: remove sample_period init in perf_buffer

Posted by Jiri Olsa 9 months, 3 weeks ago

On Tue, Apr 22, 2025 at 05:15:58PM +0800, Tao Chen wrote:
> It seems that sample_period no used in perf buffer, actually only
> wakeup_events valid about events aggregation for wakeup. So remove
> it to avoid causing confusion.

I don't see too much confusion in keeping it, but I think it
should be safe to remove it

PERF_COUNT_SW_BPF_OUTPUT is "trigered" by bpf_perf_event_output,
AFAICS there's no path checking on sample_period for this event
used in context of perf_buffer__new, Namhyung, thoughts?

thanks,
jirka


> 
> Fixes: fb84b8224655 ("libbpf: add perf buffer API")
> Signed-off-by: Tao Chen <chen.dylane@linux.dev>
> ---
>  tools/lib/bpf/libbpf.c | 1 -
>  1 file changed, 1 deletion(-)
> 
> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> index 194809da5172..1830e3c011a5 100644
> --- a/tools/lib/bpf/libbpf.c
> +++ b/tools/lib/bpf/libbpf.c
> @@ -13306,7 +13306,6 @@ struct perf_buffer *perf_buffer__new(int map_fd, size_t page_cnt,
>  	attr.config = PERF_COUNT_SW_BPF_OUTPUT;
>  	attr.type = PERF_TYPE_SOFTWARE;
>  	attr.sample_type = PERF_SAMPLE_RAW;
> -	attr.sample_period = sample_period;
>  	attr.wakeup_events = sample_period;
>  
>  	p.attr = &attr;
> -- 
> 2.43.0
>

Re: [RFC PATCH bpf-next] libbpf: remove sample_period init in perf_buffer

Posted by Namhyung Kim 9 months, 2 weeks ago

Hello,

On Tue, Apr 22, 2025 at 03:43:43PM +0200, Jiri Olsa wrote:
> On Tue, Apr 22, 2025 at 05:15:58PM +0800, Tao Chen wrote:
> > It seems that sample_period no used in perf buffer, actually only
> > wakeup_events valid about events aggregation for wakeup. So remove
> > it to avoid causing confusion.
> 
> I don't see too much confusion in keeping it, but I think it
> should be safe to remove it
> 
> PERF_COUNT_SW_BPF_OUTPUT is "trigered" by bpf_perf_event_output,
> AFAICS there's no path checking on sample_period for this event
> used in context of perf_buffer__new, Namhyung, thoughts?

It seems to be ok to call mmap(2) for non-sampling events.

Acked-by: Namhyung Kim <namhyung@kernel.org>

Thanks,
Namhyung

> 
> > 
> > Fixes: fb84b8224655 ("libbpf: add perf buffer API")
> > Signed-off-by: Tao Chen <chen.dylane@linux.dev>
> > ---
> >  tools/lib/bpf/libbpf.c | 1 -
> >  1 file changed, 1 deletion(-)
> > 
> > diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> > index 194809da5172..1830e3c011a5 100644
> > --- a/tools/lib/bpf/libbpf.c
> > +++ b/tools/lib/bpf/libbpf.c
> > @@ -13306,7 +13306,6 @@ struct perf_buffer *perf_buffer__new(int map_fd, size_t page_cnt,
> >  	attr.config = PERF_COUNT_SW_BPF_OUTPUT;
> >  	attr.type = PERF_TYPE_SOFTWARE;
> >  	attr.sample_type = PERF_SAMPLE_RAW;
> > -	attr.sample_period = sample_period;
> >  	attr.wakeup_events = sample_period;
> >  
> >  	p.attr = &attr;
> > -- 
> > 2.43.0
> >

Re: [RFC PATCH bpf-next] libbpf: remove sample_period init in perf_buffer

Posted by Jiri Olsa 9 months, 2 weeks ago

On Tue, Apr 22, 2025 at 12:05:55PM -0700, Namhyung Kim wrote:
> Hello,
> 
> On Tue, Apr 22, 2025 at 03:43:43PM +0200, Jiri Olsa wrote:
> > On Tue, Apr 22, 2025 at 05:15:58PM +0800, Tao Chen wrote:
> > > It seems that sample_period no used in perf buffer, actually only
> > > wakeup_events valid about events aggregation for wakeup. So remove
> > > it to avoid causing confusion.
> > 
> > I don't see too much confusion in keeping it, but I think it
> > should be safe to remove it
> > 
> > PERF_COUNT_SW_BPF_OUTPUT is "trigered" by bpf_perf_event_output,
> > AFAICS there's no path checking on sample_period for this event
> > used in context of perf_buffer__new, Namhyung, thoughts?
> 
> It seems to be ok to call mmap(2) for non-sampling events.
> 
> Acked-by: Namhyung Kim <namhyung@kernel.org>

Tao Chen,
could you please resend without rfc tag? plz keeps acks

Acked-by: Jiri Olsa <jolsa@kernel.org>

thanks,
jirka


> 
> Thanks,
> Namhyung
> 
> > 
> > > 
> > > Fixes: fb84b8224655 ("libbpf: add perf buffer API")
> > > Signed-off-by: Tao Chen <chen.dylane@linux.dev>
> > > ---
> > >  tools/lib/bpf/libbpf.c | 1 -
> > >  1 file changed, 1 deletion(-)
> > > 
> > > diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> > > index 194809da5172..1830e3c011a5 100644
> > > --- a/tools/lib/bpf/libbpf.c
> > > +++ b/tools/lib/bpf/libbpf.c
> > > @@ -13306,7 +13306,6 @@ struct perf_buffer *perf_buffer__new(int map_fd, size_t page_cnt,
> > >  	attr.config = PERF_COUNT_SW_BPF_OUTPUT;
> > >  	attr.type = PERF_TYPE_SOFTWARE;
> > >  	attr.sample_type = PERF_SAMPLE_RAW;
> > > -	attr.sample_period = sample_period;
> > >  	attr.wakeup_events = sample_period;
> > >  
> > >  	p.attr = &attr;
> > > -- 
> > > 2.43.0
> > >

Re: [RFC PATCH bpf-next] libbpf: remove sample_period init in perf_buffer

Posted by Tao Chen 9 months, 2 weeks ago

在 2025/4/23 23:51, Jiri Olsa 写道:
> On Tue, Apr 22, 2025 at 12:05:55PM -0700, Namhyung Kim wrote:
>> Hello,
>>
>> On Tue, Apr 22, 2025 at 03:43:43PM +0200, Jiri Olsa wrote:
>>> On Tue, Apr 22, 2025 at 05:15:58PM +0800, Tao Chen wrote:
>>>> It seems that sample_period no used in perf buffer, actually only
>>>> wakeup_events valid about events aggregation for wakeup. So remove
>>>> it to avoid causing confusion.
>>>
>>> I don't see too much confusion in keeping it, but I think it
>>> should be safe to remove it
>>>
>>> PERF_COUNT_SW_BPF_OUTPUT is "trigered" by bpf_perf_event_output,
>>> AFAICS there's no path checking on sample_period for this event
>>> used in context of perf_buffer__new, Namhyung, thoughts?
>>
>> It seems to be ok to call mmap(2) for non-sampling events.
>>
>> Acked-by: Namhyung Kim <namhyung@kernel.org>
> 
> Tao Chen,
> could you please resend without rfc tag? plz keeps acks
> 

sure, will resend it, thanks.

> Acked-by: Jiri Olsa <jolsa@kernel.org>
> 
> thanks,
> jirka
> 
> 
>>
>> Thanks,
>> Namhyung
>>
>>>
>>>>
>>>> Fixes: fb84b8224655 ("libbpf: add perf buffer API")
>>>> Signed-off-by: Tao Chen <chen.dylane@linux.dev>
>>>> ---
>>>>   tools/lib/bpf/libbpf.c | 1 -
>>>>   1 file changed, 1 deletion(-)
>>>>
>>>> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
>>>> index 194809da5172..1830e3c011a5 100644
>>>> --- a/tools/lib/bpf/libbpf.c
>>>> +++ b/tools/lib/bpf/libbpf.c
>>>> @@ -13306,7 +13306,6 @@ struct perf_buffer *perf_buffer__new(int map_fd, size_t page_cnt,
>>>>   	attr.config = PERF_COUNT_SW_BPF_OUTPUT;
>>>>   	attr.type = PERF_TYPE_SOFTWARE;
>>>>   	attr.sample_type = PERF_SAMPLE_RAW;
>>>> -	attr.sample_period = sample_period;
>>>>   	attr.wakeup_events = sample_period;
>>>>   
>>>>   	p.attr = &attr;
>>>> -- 
>>>> 2.43.0
>>>>


-- 
Best Regards
Tao Chen