Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S933174AbZKXO4I (ORCPT ); Tue, 24 Nov 2009 09:56:08 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1756848AbZKXO4H (ORCPT ); Tue, 24 Nov 2009 09:56:07 -0500 Received: from casper.infradead.org ([85.118.1.10]:47844 "EHLO casper.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751363AbZKXO4G (ORCPT ); Tue, 24 Nov 2009 09:56:06 -0500 Date: Tue, 24 Nov 2009 06:57:53 -0800 From: Arjan van de Ven To: Ingo Molnar Cc: Li Zefan , Pekka Enberg , Eduard - Gabriel Munteanu , Peter Zijlstra , Frederic Weisbecker , LKML , "linux-mm@kvack.org" Subject: Re: [PATCH 0/5] perf kmem: Add more functions and show more statistics Message-ID: <20091124065753.6a08435b@infradead.org> In-Reply-To: <20091124083423.GE21991@elte.hu> References: <4B0B6E44.6090106@cn.fujitsu.com> <84144f020911232315h7c8b7348u9ad97f585f54a014@mail.gmail.com> <20091124073426.GA21991@elte.hu> <4B0B937F.4080906@cn.fujitsu.com> <20091124083423.GE21991@elte.hu> Organization: Intel X-Mailer: Claws Mail 3.7.3 (GTK+ 2.16.6; i586-redhat-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit X-SRS-Rewrite: SMTP reverse-path rewritten from by casper.infradead.org See http://www.infradead.org/rpr.html Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1226 Lines: 28 On Tue, 24 Nov 2009 09:34:23 +0100 Ingo Molnar wrote: > > It would be great if perf can be used for boot time tracing. This > > needs pretty big work on kernel side. > > What would be needed is to open per cpu events right after perf > events initializes, and allocate memory for output buffers to them. > > They would round-robin after that point, and we could use > perf_event_open() (with a special flag) to 'attach' to them and > mmap() them - at which point they'd turn into regular objects with a > lot of boot time data in them. I'm not too worried about this btw; we can start the userland trace early enough in the boot (the kernel is done after 0.6 seconds after all) to capture the relevant stuff. The actual kernel mostly gets captured with scripts/bootgraph.pl already. Yes it would be nice to do a timechart earlier, but if it's extremely hard... Also unless it starts before the drivers (eg the normal driver initcall level), it is not useful. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/