Re: [PATCH perf/core 0/6] perf-probe: Bugfix and add new options for cache

[Namhyung Kim <namhyung@kernel.org>]

On Wed, 05 Nov 2014 01:22:46 +0900, Masami Hiramatsu wrote:
> (2014/11/04 23:38), Arnaldo Carvalho de Melo wrote:
>> Em Tue, Nov 04, 2014 at 01:36:31PM +0900, Masami Hiramatsu escreveu:
>>> OK, I agree using .debug/.buildid/ to store caches.
>>> Here is what I'm thinking,
>>  
>>>   # this makes caches for given pattern instead of adding probes.
>>>   perf probe --cache '* $params'
>> 
>>>   # the cache is stored in .debug/.buildid/<buildid>.probe
>>>   # the cache entry can be queried by buildid and eventname
>> 
>> To follow the existing standard this would instead go to:
>> 
>>>   # the cache is stored in .debug/probes/path/to/dso/name/buildid
>>>   # And can be found via its buildid link .debug/.buildid/bu/ildid -> ../../probes/path/to/dso/name/buildid
>
> Ah, I see. so you meant adding a top-level .debug/probes/ dir.
> But in that case, shouldn't we change .debug/.buildid/bu/ildid to
> .debug/probes/.buildid/bu/ildid ?

Either is fine to me.  But my concern is that it might be bloated as
system/package update is going on, so we need to control it somehow.
That's why I suggested the probe-cache command.

>
>> 
>>>   perf probe --query ${remote_buildid}:do_fork
>>>   p:probe/do_fork _text+298722 clone_flags=%di:u64 stack_start=%si:u64 stack_size=%dx:u64 parent_tidptr=%cx:u64 child_tidptr=%r8:u64
>> 
>>>   # or perf can set it up directly to local
>>>   perf probe --query-add do_fork
>> 
>> You missed the build id above, no? I.e. it would be:
>> 
>>>   # or perf can set it up directly to local
>>>   perf probe --query-add ${remote_buildid}:do_fork
>
> No, since this command set the event to local machine, perf-probe
> should check the local build-id and query the appropriate event
> from the cache.
> # BTW, maybe we'd better use perf probe --add '$do_fork' (calls
> # "cache of do_fork")  instead of long --query-add. :)

It should take care of uprobe case too.  So simple do_fork should have
group/event or provider/marker form instead so that it can help to find
which binary defines the cached event.  Maybe we also need to keep a
event-to-binary table and then check (current?) build-id somehow to
identify correct event to be used.

Also this function entry level event cache can be used with uprobes..

Thanks,
Namhyung