Re: perf: PERF_FLAG_FD_OUTPUT has been broken since 2.6.35

[Arnaldo Carvalho de Melo <acme@kernel.org>]

Em Fri, Jan 09, 2015 at 11:26:54AM -0500, Vince Weaver escreveu:
> 
> On Fri, 9 Jan 2015, Peter Zijlstra wrote:
> > > So is this worth fixing seeing as apparently no one uses this feature?
> > 
> > I think there's a fair argument for removing it, Ingo, Acme?
> 
> could the functionality be replaced with a subsequent call to
> 	ioctl(PERF_EVENT_IOC_SET_OUTPUT)
> ?

That is the only thing tools/perf uses:

[acme@zoo linux]$ find tools/perf -name "*.[chly]" | xargs grep PERF_EVENT_IOC_SET_OUTPUT 
tools/perf/util/evlist.h: * @refcnt - e.g. code using PERF_EVENT_IOC_SET_OUTPUT to share this
tools/perf/util/evlist.c:			if (ioctl(fd, PERF_EVENT_IOC_SET_OUTPUT, *output) != 0)
tools/perf/tests/perf-record.c:	 * (using ioctl(PERF_EVENT_IOC_SET_OUTPUT)).
[acme@zoo linux]$ find tools/perf -name "*.[chly]" | xargs grep PERF_FLAG_FD_OUTPUT
[acme@zoo linux]$
 
> Although I suppose there's a possibility for losing a small amount of data 
> or some other reason that PERF_FLAG_FD_OUTPUT was introduced in the first 
> place.

Humm, IIRC tools/perf starts with the event disabled and then asks for
enable_on_exec when starting workloads but yes, when you're attaching to
something that is already running you'd take a bit longer to start getting
samples.
 
> In addition, if we remove PERF_FLAG_FD_OUTPUT would there then be any 
> reason to keep PERF_FLAG_FD_NO_GROUP around?
> 
> Vince