From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752817AbbANL6m (ORCPT ); Wed, 14 Jan 2015 06:58:42 -0500 Received: from mx1.redhat.com ([209.132.183.28]:56138 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751706AbbANL6j (ORCPT ); Wed, 14 Jan 2015 06:58:39 -0500 Date: Wed, 14 Jan 2015 12:57:49 +0100 From: Jiri Olsa To: Wang Nan Cc: jolsa@kernel.org, linux-kernel@vger.kernel.org, lizefan@huawei.com, namhyung@kernel.org, a.p.zijlstra@chello.nl, paulus@samba.org, mingo@redhat.com, acme@kernel.org, masami.hiramatsu.pt@hitachi.com Subject: Re: [PATCH v2] perf: fix dwarf unwind using libunwind. Message-ID: <20150114115749.GB6369@krava.redhat.com> References: <20150113091928.GA2160@krava.brq.redhat.com> <1421203007-75799-1-git-send-email-wangnan0@huawei.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1421203007-75799-1-git-send-email-wangnan0@huawei.com> User-Agent: Mutt/1.5.23 (2014-03-12) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Jan 14, 2015 at 10:36:47AM +0800, Wang Nan wrote: > Perf tool fails to unwind user stack if the event raises in a shared SNIP > + > __attribute__ ((noinline)) > static int krava_3(struct thread *thread) > { > - return unwind_thread(thread); > + struct thread *array[2] = {thread, thread}; > + void *fp = &bsearch; > + /* > + * make _bsearch a volatile function pointer to > + * prevent potential optimization, which may expand > + * bsearch and call compare directly from this function, > + * instead of libc shared object. > + */ > + void *(*volatile _bsearch)(void *, void *, size_t, > + size_t, int (*)(void *, void *)); > + > + _bsearch = fp; > + _bsearch(array, &thread, 2, sizeof(struct thread **), compare); > + return global_unwind_retval; > } ah, I've got confused with the NO_LIBUNWIND_DEBUG_FRAME name and got the impression that we could use it in the tests/make as another compile option test.. but your change is even better ;-) thanks for updating this test Acked-by: Jiri Olsa thanks, jirka