Messages in this thread | | | Date | Wed, 22 Apr 2015 09:36:20 -0600 | From | David Ahern <> | Subject | Re: [PATCH] tracing: Export key trace event symbols |
| |
On 4/22/15 8:47 AM, Arnaldo Carvalho de Melo wrote: > Em Wed, Apr 22, 2015 at 08:53:14AM -0400, Steven Rostedt escreveu: >> >On Tue, 21 Apr 2015 21:24:51 -0500 >> >Ron Rechenmacher<ron@fnal.gov> wrote: >>> > >I've looked at the above reference briefly and it appears that user-space >>> > >would be mmapping the buffer read-only. Is that correct? >> > >> >Correct, but I'm sure we could still add something (if it doesn't >> >already exist) to have userspace write into the buffer. Ftrace has that >> >with the trace_marker file. > There is something in the works, I guess Pawell Moll (sp) was working on it, and > David Ahern (CCed) should know, David? >
I played around with generating perf events in userspace with the intention of having the userspace events get merged with kernel events during the processing stage, but I did not take it to the point of integrating into perf. This was around October 2013. I got distracted with other topics and have not come back to it.
Pawel has a patch that allows userspace to inject events into the stream via ioctl calls.
Stephane also injects events for JIT.
One of the key requirements is a common time basis (e.g., CLOCK_MONOTONIC or PERF_CLOCK) to be able to merge the events properly. I have a kernel module that exports perf_clock to userspace via clock_gettime; the 4.1 kernel should have the code that allows the clock id to be specified providing a solution to this problem.
David
| |