linux-kernel - Re: [PATCH 4/4] perf tools: Add perf trace

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20090817163635.GA30281@elte.hu>
Date:	Mon, 17 Aug 2009 18:36:35 +0200
From:	Ingo Molnar <mingo@...e.hu>
To:	Frederic Weisbecker <fweisbec@...il.com>
Cc:	LKML <linux-kernel@...r.kernel.org>,
	Peter Zijlstra <peterz@...radead.org>,
	Arnaldo Carvalho de Melo <acme@...hat.com>,
	Mike Galbraith <efault@....de>,
	Thomas Gleixner <tglx@...utronix.de>,
	"Luis Claudio R. Goncalves" <lclaudio@...g.org>,
	Clark Williams <williams@...hat.com>,
	Jon Masters <jonathan@...masters.org>,
	Mathieu Desnoyers <mathieu.desnoyers@...ymtl.ca>,
	Christoph Hellwig <hch@...radead.org>,
	Xiao Guangrong <xiaoguangrong@...fujitsu.com>,
	Zhaolei <zhaolei@...fujitsu.com>, Li Zefan <lizf@...fujitsu.com>,
	Lai Jiangshan <laijs@...fujitsu.com>,
	Masami Hiramatsu <mhiramat@...hat.com>,
	Tom Zanussi <tzanussi@...il.com>,
	"Frank Ch. Eigler" <fche@...hat.com>,
	Roland McGrath <roland@...hat.com>,
	Jason Baron <jbaron@...hat.com>,
	Steven Rostedt <rostedt@...dmis.org>,
	Paul Mackerras <paulus@...ba.org>,
	Jiaying Zhang <jiayingz@...gle.com>,
	Anton Blanchard <anton@...ba.org>
Subject: Re: [PATCH 4/4] perf tools: Add perf trace


* Frederic Weisbecker <fweisbec@...il.com> wrote:

> On Mon, Aug 17, 2009 at 04:40:07PM +0200, Ingo Molnar wrote:
> > 
> > * Frederic Weisbecker <fweisbec@...il.com> wrote:
> > 
> > > This adds perf trace into the set of perf tools. It is written to 
> > > fetch the tracepoints samples from perf events and display them, 
> > > according to the events informations given by the debugfs files 
> > > through the util/trace* tools.
> > > 
> > > It is a rough first shot and doesn't yet handle the cpu, 
> > > timestamps fields and some other things.
> > > 
> > > Example:
> > > 
> > > perf record -f -e workqueue:workqueue_execution:record -F 1 -a
> > > perf trace
> > > 
> > >        kblockd/0-236   [000]     0.000000: workqueue_execution: thread=:236 func=cfq_kick_queue+0x0
> > >      kondemand/0-360   [000]     0.000000: workqueue_execution: thread=:360 func=do_dbs_timer+0x0
> > >      kondemand/0-360   [000]     0.000000: workqueue_execution: thread=:360 func=do_dbs_timer+0x0
> > >      kondemand/1-361   [000]     0.000000: workqueue_execution: thread=:361 func=do_dbs_timer+0x0
> > >      kondemand/1-361   [000]     0.000000: workqueue_execution: thread=:361 func=do_dbs_timer+0x0
> > >      kondemand/1-361   [000]     0.000000: workqueue_execution: thread=:361 func=do_dbs_timer+0x0
> > >      kondemand/1-361   [000]     0.000000: workqueue_execution: thread=:361 func=do_dbs_timer+0x0
> > >      kondemand/1-361   [000]     0.000000: workqueue_execution: thread=:361 func=do_dbs_timer+0x0
> > >      kondemand/1-361   [000]     0.000000: workqueue_execution: thread=:361 func=do_dbs_timer+0x0
> > >      kondemand/1-361   [000]     0.000000: workqueue_execution: thread=:361 func=do_dbs_timer+0x0
> > >      kondemand/1-361   [000]     0.000000: workqueue_execution: thread=:361 func=do_dbs_timer+0x0
> > >      kondemand/1-361   [000]     0.000000: workqueue_execution: thread=:361 func=do_dbs_timer+0x0
> > >      kondemand/1-361   [000]     0.000000: workqueue_execution: thread=:361 func=do_dbs_timer+0x0
> > >      kondemand/1-361   [000]     0.000000: workqueue_execution: thread=:361 func=do_dbs_timer+0x0
> > >      kondemand/1-361   [000]     0.000000: workqueue_execution: thread=:361 func=do_dbs_timer+0x0
> > >      kondemand/1-361   [000]     0.000000: workqueue_execution: thread=:361 func=do_dbs_timer+0x0
> > 
> > Nice!
> > 
> > > Todo:
> > > 
> > > - A lot of things!
> > 
> > heh :-)
> > 
> > To help this move forward i've put the patches into 
> > tip:perfcounters/tracing. We might still rebase that branch, should 
> > it become necessary, before merging it into perfcounters/core - but 
> > it should give a first glimpse to people who want to try this.
> > 
> > A first bugreport:
> > 
> > took me some time to figure out that i need to pass in -R to create 
> > a trace.info. I think we want to share the namespace anyway - 
> > there's no real difference between perf.data and trace.info - both 
> > contain trace records.
> 
> 
> 
> trace.info only contains the events descriptions and other ftrace things:
> 
> - formats, ftrace printk strings, ...
> 
> I plan to integrate that into perf.trace so that we have only one file
> to move to perform offline analysis.
> 
> Concerning the -R thing (or ::record suffix), indeed I have yet to
> warn the user from perf trace about that. Todo listed!
> 
> > 
> > once i added -R i ran into this problem:
> > 
> > # perf record -R -f -e workqueue:workqueue_execution:record -F 1 -a
> > ^C
> > [ perf record: Captured and wrote 0.250 MB perf.data (~10928 samples) ]
> > 
> > aldebaran:/home/mingo/linux/linux/tools/perf> perf trace
> >   Fatal: bad op token )
> 
> 
> Oh!
>
> Hmm, also I forgot to explain a detail, there is a little bug in 
> ftrace syscalls formats which reports the format is too big when a 
> syscall takes no parameter. It needs a trivial fix, I'll send it 
> soon.
> 
> Anyway, that made me a very similar Fatal thing but not exactly 
> the same. Could you send me your config? May be you have 
> tracepoint/ftrace plugins I haven't built (well I should build all 
> of them actually).

sure, config attached.

	Ingo

View attachment "config" of type "text/plain" (65300 bytes)