From mboxrd@z Thu Jan 1 00:00:00 1970 From: mathieu.poirier@linaro.org (Mathieu Poirier) Date: Fri, 16 Nov 2018 16:05:11 -0700 Subject: [PATCH v1 2/5] perf cs-etm: Avoid stale branch samples when flush packet In-Reply-To: <1541912383-19915-3-git-send-email-leo.yan@linaro.org> References: <1541912383-19915-1-git-send-email-leo.yan@linaro.org> <1541912383-19915-3-git-send-email-leo.yan@linaro.org> Message-ID: <20181116230511.GB25258@xps15> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On Sun, Nov 11, 2018 at 12:59:40PM +0800, Leo Yan wrote: > At the end of trace buffer handling, function cs_etm__flush() is invoked > to flush any remaining branch stack entries. As a side effect, it also > generates branch sample, because the 'etmq->packet' doesn't contains any > new coming packet but point to one stale packet after packets swapping, > so it wrongly makes synthesize branch samples with stale packet info. > > We could review below detailed flow which causes issue: > > Packet1: start_addr=0xffff000008b1fbf0 end_addr=0xffff000008b1fbfc > Packet2: start_addr=0xffff000008b1fb5c end_addr=0xffff000008b1fb6c > > step 1: cs_etm__sample(): > sample: ip=(0xffff000008b1fbfc-4) addr=0xffff000008b1fb5c > > step 2: flush packet in cs_etm__run_decoder(): > cs_etm__run_decoder() > `-> err = cs_etm__flush(etmq, false); > sample: ip=(0xffff000008b1fb6c-4) addr=0xffff000008b1fbf0 > > Packet1 and packet2 are two continuous packets, when packet2 is the new > coming packet, cs_etm__sample() generates branch sample for these two > packets and use [packet1::end_addr - 4 => packet2::start_addr] as branch > jump flow, thus we can see the first generated branch sample in step 1. > At the end of cs_etm__sample() it swaps packets so 'etm->prev_packet'= > packet2 and 'etm->packet'=packet1, so far it's okay for branch sample. > > If packet2 is the last one packet in trace buffer, even there have no > any new coming packet, cs_etm__run_decoder() invokes cs_etm__flush() to > flush branch stack entries as expected, but it also generates branch > samples by taking 'etm->packet' as a new coming packet, thus the branch > jump flow is as [packet2::end_addr - 4 => packet1::start_addr]; this > is the second sample which is generated in step 2. So actually the > second sample is a stale sample and we should not generate it. > > This patch is to add new argument 'new_packet' for cs_etm__flush(), we > can pass 'true' for this argument if there have a new packet, otherwise > it will pass 'false' for the purpose of only flushing branch stack > entries and avoid to generate sample for stale packet. Very good explanation, thanks for taking the time to write this. > > Signed-off-by: Leo Yan > --- > tools/perf/util/cs-etm.c | 20 +++++++++++++++++--- > 1 file changed, 17 insertions(+), 3 deletions(-) > > diff --git a/tools/perf/util/cs-etm.c b/tools/perf/util/cs-etm.c > index fe18d7b..f4fa877 100644 > --- a/tools/perf/util/cs-etm.c > +++ b/tools/perf/util/cs-etm.c > @@ -955,7 +955,7 @@ static int cs_etm__sample(struct cs_etm_queue *etmq) > return 0; > } > > -static int cs_etm__flush(struct cs_etm_queue *etmq) > +static int cs_etm__flush(struct cs_etm_queue *etmq, bool new_packet) > { > int err = 0; > struct cs_etm_auxtrace *etm = etmq->etm; > @@ -989,6 +989,20 @@ static int cs_etm__flush(struct cs_etm_queue *etmq) > > } > > + /* > + * If 'new_packet' is false, this time call has no a new packet > + * coming and 'etmq->packet' contains the stale packet which is > + * set at the previous time with packets swapping. In this case > + * this function is invoked only for flushing branch stack at > + * the end of buffer handling. > + * > + * Simply to say, branch samples should be generated when every > + * time receive one new packet; otherwise, directly bail out to > + * avoid generate branch sample with stale packet. > + */ > + if (!new_packet) > + return 0; > + > if (etm->sample_branches && > etmq->prev_packet->sample_type == CS_ETM_RANGE) { > err = cs_etm__synth_branch_sample(etmq); > @@ -1075,7 +1089,7 @@ static int cs_etm__run_decoder(struct cs_etm_queue *etmq) > * Discontinuity in trace, flush > * previous branch stack > */ > - cs_etm__flush(etmq); > + cs_etm__flush(etmq, true); > break; > case CS_ETM_EMPTY: > /* > @@ -1092,7 +1106,7 @@ static int cs_etm__run_decoder(struct cs_etm_queue *etmq) > > if (err == 0) > /* Flush any remaining branch stack entries */ > - err = cs_etm__flush(etmq); > + err = cs_etm__flush(etmq, false); I understand what you're doing and it will yield the correct results. What I'm not sure about is if we wouldn't be better off splitting cs_etm__flush() in order to reduce the complexity of the main decoding loop. That is rename cs_etm__flush() to something like cs_etm__trace_on() and spin off a new cs_etm__end_block(). It does introduce a little bit of code duplication but I think we'd win in terms of readability and flexibility. Thanks, Mathieu > } > > return err; > -- > 2.7.4 > From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-8.6 required=3.0 tests=DKIM_SIGNED,DKIM_VALID, DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_PATCH,MAILING_LIST_MULTI, SIGNED_OFF_BY,SPF_PASS,URIBL_BLOCKED,USER_AGENT_MUTT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 30EFEC43441 for ; Fri, 16 Nov 2018 23:05:19 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id D71482086A for ; Fri, 16 Nov 2018 23:05:18 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (1024-bit key) header.d=linaro.org header.i=@linaro.org header.b="LAH+U5vf" DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org D71482086A Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=linaro.org Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=linux-kernel-owner@vger.kernel.org Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1730375AbeKQJTd (ORCPT ); Sat, 17 Nov 2018 04:19:33 -0500 Received: from mail-pg1-f194.google.com ([209.85.215.194]:46031 "EHLO mail-pg1-f194.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1725819AbeKQJTd (ORCPT ); Sat, 17 Nov 2018 04:19:33 -0500 Received: by mail-pg1-f194.google.com with SMTP id y4so11206945pgc.12 for ; Fri, 16 Nov 2018 15:05:15 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linaro.org; s=google; h=date:from:to:cc:subject:message-id:references:mime-version :content-disposition:in-reply-to:user-agent; bh=CmU0ypN2uBNgVuQePI33oHrQC1mVLrfBCLqLPyWeIP0=; b=LAH+U5vfSQvoEjYUOCEfCIK5GQRvvSpIrmhtTViTlXY4d7+ZyOYmtM9OJw1dpkGdEz /t08vyd/CVEw1y/d5RvIyZ0peuS+oSKucEjQiI9csaYwWhKSm28epo29UPj9w7yWCMb2 fhWt9XREPQd/KZbIKuah31Ob04dX89TQ3knWQ= X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:date:from:to:cc:subject:message-id:references :mime-version:content-disposition:in-reply-to:user-agent; bh=CmU0ypN2uBNgVuQePI33oHrQC1mVLrfBCLqLPyWeIP0=; b=k8SYXiXppd2yJnP/m0yOmRnL3ZBSvcLjClrCKwYgZCwEDU1Ur8+FAgTEAUJtIE5reG 62D1+rXaEWpoT51pMeGvfdAEi6i/W//aNsefMaMRmO+3UVXZgBr4aseH1PszXer+hJfj H/SIWh3sS0DvJubfYQ/8CTj5GJYKd+VFzWLuSD97MvObvN9HL4bL/ZHzM+8hbTZaTiFH LysfzUACs7O0CkNnuGOqmvcmmU2RPXqnQlvGVp/gMsrL+657Er1ICRlZINwt74j46OF6 nqsF8rz8al91NdP/W07kZYEKNn86ag98AXsZnsDNRCsgs29NA5IAOWWXxeHORjNHsz1U EJJA== X-Gm-Message-State: AGRZ1gK89HbFwDrHQzYqzdePOMBsxzCICxpKKIvO4IPDP90s/TUlHfG2 7V5wltV/+Nk1Hbo+j7KwW5k6dw== X-Google-Smtp-Source: AJdET5chqBLmMk3b90pRr2e2NGuvaoZjlGzefkVnJyx/IvA3blemMwRjquog2+46ex1o7j8hJaQyKA== X-Received: by 2002:a63:1258:: with SMTP id 24mr11572361pgs.114.1542409514846; Fri, 16 Nov 2018 15:05:14 -0800 (PST) Received: from xps15 (S0106002369de4dac.cg.shawcable.net. [68.147.8.254]) by smtp.gmail.com with ESMTPSA id c127-v6sm34822429pfa.31.2018.11.16.15.05.13 (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Fri, 16 Nov 2018 15:05:14 -0800 (PST) Date: Fri, 16 Nov 2018 16:05:11 -0700 From: Mathieu Poirier To: Leo Yan Cc: Arnaldo Carvalho de Melo , Alexander Shishkin , Jiri Olsa , Namhyung Kim , linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org, Mike Leach , Robert Walker , Al Grant , Coresight ML Subject: Re: [PATCH v1 2/5] perf cs-etm: Avoid stale branch samples when flush packet Message-ID: <20181116230511.GB25258@xps15> References: <1541912383-19915-1-git-send-email-leo.yan@linaro.org> <1541912383-19915-3-git-send-email-leo.yan@linaro.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1541912383-19915-3-git-send-email-leo.yan@linaro.org> User-Agent: Mutt/1.5.24 (2015-08-30) Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Sun, Nov 11, 2018 at 12:59:40PM +0800, Leo Yan wrote: > At the end of trace buffer handling, function cs_etm__flush() is invoked > to flush any remaining branch stack entries. As a side effect, it also > generates branch sample, because the 'etmq->packet' doesn't contains any > new coming packet but point to one stale packet after packets swapping, > so it wrongly makes synthesize branch samples with stale packet info. > > We could review below detailed flow which causes issue: > > Packet1: start_addr=0xffff000008b1fbf0 end_addr=0xffff000008b1fbfc > Packet2: start_addr=0xffff000008b1fb5c end_addr=0xffff000008b1fb6c > > step 1: cs_etm__sample(): > sample: ip=(0xffff000008b1fbfc-4) addr=0xffff000008b1fb5c > > step 2: flush packet in cs_etm__run_decoder(): > cs_etm__run_decoder() > `-> err = cs_etm__flush(etmq, false); > sample: ip=(0xffff000008b1fb6c-4) addr=0xffff000008b1fbf0 > > Packet1 and packet2 are two continuous packets, when packet2 is the new > coming packet, cs_etm__sample() generates branch sample for these two > packets and use [packet1::end_addr - 4 => packet2::start_addr] as branch > jump flow, thus we can see the first generated branch sample in step 1. > At the end of cs_etm__sample() it swaps packets so 'etm->prev_packet'= > packet2 and 'etm->packet'=packet1, so far it's okay for branch sample. > > If packet2 is the last one packet in trace buffer, even there have no > any new coming packet, cs_etm__run_decoder() invokes cs_etm__flush() to > flush branch stack entries as expected, but it also generates branch > samples by taking 'etm->packet' as a new coming packet, thus the branch > jump flow is as [packet2::end_addr - 4 => packet1::start_addr]; this > is the second sample which is generated in step 2. So actually the > second sample is a stale sample and we should not generate it. > > This patch is to add new argument 'new_packet' for cs_etm__flush(), we > can pass 'true' for this argument if there have a new packet, otherwise > it will pass 'false' for the purpose of only flushing branch stack > entries and avoid to generate sample for stale packet. Very good explanation, thanks for taking the time to write this. > > Signed-off-by: Leo Yan > --- > tools/perf/util/cs-etm.c | 20 +++++++++++++++++--- > 1 file changed, 17 insertions(+), 3 deletions(-) > > diff --git a/tools/perf/util/cs-etm.c b/tools/perf/util/cs-etm.c > index fe18d7b..f4fa877 100644 > --- a/tools/perf/util/cs-etm.c > +++ b/tools/perf/util/cs-etm.c > @@ -955,7 +955,7 @@ static int cs_etm__sample(struct cs_etm_queue *etmq) > return 0; > } > > -static int cs_etm__flush(struct cs_etm_queue *etmq) > +static int cs_etm__flush(struct cs_etm_queue *etmq, bool new_packet) > { > int err = 0; > struct cs_etm_auxtrace *etm = etmq->etm; > @@ -989,6 +989,20 @@ static int cs_etm__flush(struct cs_etm_queue *etmq) > > } > > + /* > + * If 'new_packet' is false, this time call has no a new packet > + * coming and 'etmq->packet' contains the stale packet which is > + * set at the previous time with packets swapping. In this case > + * this function is invoked only for flushing branch stack at > + * the end of buffer handling. > + * > + * Simply to say, branch samples should be generated when every > + * time receive one new packet; otherwise, directly bail out to > + * avoid generate branch sample with stale packet. > + */ > + if (!new_packet) > + return 0; > + > if (etm->sample_branches && > etmq->prev_packet->sample_type == CS_ETM_RANGE) { > err = cs_etm__synth_branch_sample(etmq); > @@ -1075,7 +1089,7 @@ static int cs_etm__run_decoder(struct cs_etm_queue *etmq) > * Discontinuity in trace, flush > * previous branch stack > */ > - cs_etm__flush(etmq); > + cs_etm__flush(etmq, true); > break; > case CS_ETM_EMPTY: > /* > @@ -1092,7 +1106,7 @@ static int cs_etm__run_decoder(struct cs_etm_queue *etmq) > > if (err == 0) > /* Flush any remaining branch stack entries */ > - err = cs_etm__flush(etmq); > + err = cs_etm__flush(etmq, false); I understand what you're doing and it will yield the correct results. What I'm not sure about is if we wouldn't be better off splitting cs_etm__flush() in order to reduce the complexity of the main decoding loop. That is rename cs_etm__flush() to something like cs_etm__trace_on() and spin off a new cs_etm__end_block(). It does introduce a little bit of code duplication but I think we'd win in terms of readability and flexibility. Thanks, Mathieu > } > > return err; > -- > 2.7.4 >