Re: [PATCH 0/4] workqueue_trace: Add max execution time mesurement for per worklet

From: Frederic Weisbecker <fweisbec@gmail.com>
To: Andrew Morton <akpm@linux-foundation.org>
Cc: Zhaolei <zhaolei@cn.fujitsu.com>,
	Steven Rostedt <rostedt@goodmis.org>, Ingo Molnar <mingo@elte.hu>,
	Tom Zanussi <tzanussi@gmail.com>,
	KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>,
	Oleg Nesterov <oleg@redhat.com>,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH 0/4] workqueue_trace: Add max execution time mesurement for per worklet
Date: Wed, 29 Apr 2009 15:33:44 +0200	[thread overview]
Message-ID: <20090429133342.GA5987@nowhere> (raw)
In-Reply-To: <20090429000616.f7e72b73.akpm@linux-foundation.org>

On Wed, Apr 29, 2009 at 12:06:16AM -0700, Andrew Morton wrote:
> What is the meaning of M_EXECus?  Mean execution time in microseconds,
> I assume?  I wonder if that is sufficiently high resolution nowadays.

It's the maximum execution time encountered for a given worklet.
I also have a small patch, about ready, to get the average.

> You'll note that all the functions are reported as "foo+0x0/0xN". 
> There is a way of suppressing the unneeded and unuseful "+0x0/0xN",
> but I forget what it is and the comments over lib/vsprintf.c:pointer()
> didn't help.  I think it's %pS.

I don't know any way to perform this.
%pF and %pS will act pretty much the same, except:

Extract from lib/vsprintf.c:

 * Note: The difference between 'S' and 'F' is that on ia64 and ppc64
 * function pointers are really function descriptors, which contain a
 * pointer to the real address.

But the output formatting is the same: address, offset, size, modname.

I found it a bit annoying, so I proposed the following patch recently:

http://lkml.org/lkml/2009/4/15/310

And someone came with a better idea, IMO:

http://lkml.org/lkml/2009/4/17/105

> 
> The patchset adds no user documentation for the feature and should not
> (IMO) be merged in such a state.  Please.
>

Ok, I will try something today about the documentation.

Thanks!