From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753003Ab2IZGzY (ORCPT ); Wed, 26 Sep 2012 02:55:24 -0400 Received: from mail-pb0-f46.google.com ([209.85.160.46]:61406 "EHLO mail-pb0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752861Ab2IZGzX (ORCPT ); Wed, 26 Sep 2012 02:55:23 -0400 MIME-Version: 1.0 In-Reply-To: <5062938F.2080703@hitachi.com> References: <5062938F.2080703@hitachi.com> Date: Wed, 26 Sep 2012 15:55:23 +0900 Message-ID: Subject: Re: [QUESTION] Can uprobe_event support @ADDR, $retval, offs(FETCHARG)? From: Hyeoncheol Lee To: Masami Hiramatsu Cc: srikar@linux.vnet.ibm.com, linux-kernel@vger.kernel.org, yrl.pp-manager.tt@hitachi.com Content-Type: text/plain; charset=ISO-8859-1 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi, 2012/9/26 Masami Hiramatsu : > (2012/09/26 11:52), Hyeoncheol Lee wrote: >> Hi, >> >> uprobe_event only supports %REG arguments. I think that memory fetch, >> return value fetch, memory dereference functions in >> kernel/trace/trace_probe.c are good for uprobe_event. So with a little >> modification of parse_probe_arg(), uprobe_event can support @ADDR, >> $retval, offs(FETCHARGS) except @SYM, $stack, $stackN. Is it right? > > Hi Hyeoncheol, > > Perhaps, it is not so small things, but at least, we can try. > In the userspace, memories(pages) can be paged out on swap or > files. In that case, memory dereference function needs to track > down the data on the disk and it causes I/O. This means we will > see the visible performance degradation with tracing. > And also, sometime a pointer value (address) is broken, in that > case we have to ensure the address is actually valid before > accessing it. > > Of cause, without tracking paged-out data, it is easy > to support, because that is already done in kprobe event. > I'm not sure how it is useful, because sometimes it will > fail to access gather the data. > However it is good for the first step, I think. Yes, we should consider paged-out data. I forgot this. > > Srikar, what would you think? > > BTW, if we can support offs(FETCHARGS), $stack and $stackN > are also available. ;) Yes, but we should check whether the given address is in an user stack or not. and these require CPU architecture dependent codes. > > Thank you, > > -- > Masami HIRAMATSU > Software Platform Research Dept. Linux Technology Center > Hitachi, Ltd., Yokohama Research Laboratory > E-mail: masami.hiramatsu.pt@hitachi.com > > Thanks for your answer. It is very help to me. ;) Hyeoncheol