From: Milian Wolff <milian.wolff@kdab.com>
To: Linux-kernel@vger.kernel.org
Cc: linux-perf-users@vger.kernel.org,
Milian Wolff <milian.wolff@kdab.com>,
Arnaldo Carvalho de Melo <acme@redhat.com>,
Namhyung Kim <namhyung@kernel.org>, Jiri Olsa <jolsa@redhat.com>
Subject: [PATCH 2/2] perf report: report module before querying isactivation in dwfl unwind
Date: Fri, 2 Jun 2017 16:37:53 +0200
Message-ID: <20170602143753.16907-2-milian.wolff@kdab.com> (raw)
In-Reply-To: <20170602143753.16907-1-milian.wolff@kdab.com>
The PC returned by dwfl_frame_pc may map into a not-yet-reported
module. We have to report it before we continue unwinding. But when
we query for the isactivation flag in dwfl_frame_pc, libdw will
actually do one more unwinding step internally which can then break
and lead to missed frames or broken stacks.
With libunwind we get e.g.:
~~~~~
heaptrack_gui 2228 135073.400474: 613969 cycles:
108c8e [unknown] (/usr/lib/libQt5Core.so.5.8.0)
1093bc [unknown] (/usr/lib/libQt5Core.so.5.8.0)
109e7b QLocale::QLocale (/usr/lib/libQt5Core.so.5.8.0)
1470ff [unknown] (/usr/lib/libQt5Core.so.5.8.0)
147f67 QSystemLocale::query (/usr/lib/libQt5Core.so.5.8.0)
109fbf QLocalePrivate::updateSystemPrivate (/usr/lib/libQt5Core.so.5.8.0)
10aa27 QLocale::QLocale (/usr/lib/libQt5Core.so.5.8.0)
1e02c3 [unknown] (/usr/lib/libQt5Core.so.5.8.0)
2113bb [unknown] (/usr/lib/libQt5Core.so.5.8.0)
211505 [unknown] (/usr/lib/libQt5Core.so.5.8.0)
1b5df0 QFileInfo::exists (/usr/lib/libQt5Core.so.5.8.0)
92eb2 [unknown] (/usr/lib/libQt5Core.so.5.8.0)
93423 [unknown] (/usr/lib/libQt5Core.so.5.8.0)
93d2a QLibraryInfo::location (/usr/lib/libQt5Core.so.5.8.0)
2170af [unknown] (/usr/lib/libQt5Core.so.5.8.0)
297c53 QCoreApplicationPrivate::init (/usr/lib/libQt5Core.so.5.8.0)
f7cde QGuiApplicationPrivate::init (/usr/lib/libQt5Gui.so.5.8.0)
1589e8 QApplicationPrivate::init (/usr/lib/libQt5Widgets.so.5.8.0)
78622 main (/home/milian/projects/compiled/other/bin/heaptrack_gui)
20439 __libc_start_main (/usr/lib/libc-2.25.so)
78299 _start (/home/milian/projects/compiled/other/bin/heaptrack_gui)
heaptrack_gui 2228 135073.401156: 569521 cycles:
131633 QString::endsWith (/usr/lib/libQt5Core.so.5.8.0)
1a0701 QDir::cleanPath (/usr/lib/libQt5Core.so.5.8.0)
21b82d [unknown] (/usr/lib/libQt5Core.so.5.8.0)
1b3727 QFileInfo::canonicalFilePath (/usr/lib/libQt5Core.so.5.8.0)
2780c7 QFactoryLoader::update (/usr/lib/libQt5Core.so.5.8.0)
279525 QFactoryLoader::QFactoryLoader (/usr/lib/libQt5Core.so.5.8.0)
e5bd0 QPlatformIntegrationFactory::create (/usr/lib/libQt5Gui.so.5.8.0)
f5a1c QGuiApplicationPrivate::createPlatformIntegration (/usr/lib/libQt5Gui.so.5.8.0)
f650c QGuiApplicationPrivate::createEventDispatcher (/usr/lib/libQt5Gui.so.5.8.0)
298524 QCoreApplicationPrivate::init (/usr/lib/libQt5Core.so.5.8.0)
f7cde QGuiApplicationPrivate::init (/usr/lib/libQt5Gui.so.5.8.0)
1589e8 QApplicationPrivate::init (/usr/lib/libQt5Widgets.so.5.8.0)
78622 main (/home/milian/projects/compiled/other/bin/heaptrack_gui)
20439 __libc_start_main (/usr/lib/libc-2.25.so)
78299 _start (/home/milian/projects/compiled/other/bin/heaptrack_gui)
~~~~~
Note the two frames 1589e8 and 78622 in the first sample. These are
missing when unwinding with libdw. The second sample's breakage is
more obvious:
~~~~~
heaptrack_gui 2228 135073.400474: 613969 cycles:
108c8e [unknown] (/usr/lib/libQt5Core.so.5.8.0)
1093bc [unknown] (/usr/lib/libQt5Core.so.5.8.0)
109e7b QLocale::QLocale (/usr/lib/libQt5Core.so.5.8.0)
1470ff [unknown] (/usr/lib/libQt5Core.so.5.8.0)
147f67 QSystemLocale::query (/usr/lib/libQt5Core.so.5.8.0)
109fbf QLocalePrivate::updateSystemPrivate (/usr/lib/libQt5Core.so.5.8.0)
10aa27 QLocale::QLocale (/usr/lib/libQt5Core.so.5.8.0)
1e02c3 [unknown] (/usr/lib/libQt5Core.so.5.8.0)
2113bb [unknown] (/usr/lib/libQt5Core.so.5.8.0)
211505 [unknown] (/usr/lib/libQt5Core.so.5.8.0)
1b5df0 QFileInfo::exists (/usr/lib/libQt5Core.so.5.8.0)
92eb2 [unknown] (/usr/lib/libQt5Core.so.5.8.0)
93423 [unknown] (/usr/lib/libQt5Core.so.5.8.0)
93d2a QLibraryInfo::location (/usr/lib/libQt5Core.so.5.8.0)
2170af [unknown] (/usr/lib/libQt5Core.so.5.8.0)
297c53 QCoreApplicationPrivate::init (/usr/lib/libQt5Core.so.5.8.0)
f7cde QGuiApplicationPrivate::init (/usr/lib/libQt5Gui.so.5.8.0)
20439 __libc_start_main (/usr/lib/libc-2.25.so)
78299 _start (/home/milian/projects/compiled/other/bin/heaptrack_gui)
heaptrack_gui 2228 135073.401156: 569521 cycles:
131633 QString::endsWith (/usr/lib/libQt5Core.so.5.8.0)
1a0701 QDir::cleanPath (/usr/lib/libQt5Core.so.5.8.0)
21b82d [unknown] (/usr/lib/libQt5Core.so.5.8.0)
1b3727 QFileInfo::canonicalFilePath (/usr/lib/libQt5Core.so.5.8.0)
2780c7 QFactoryLoader::update (/usr/lib/libQt5Core.so.5.8.0)
279525 QFactoryLoader::QFactoryLoader (/usr/lib/libQt5Core.so.5.8.0)
e5bd0 QPlatformIntegrationFactory::create (/usr/lib/libQt5Gui.so.5.8.0)
723dbf [unknown] ([unknown])
~~~~~
This patch fixes this issue and the libdw unwinder mimicks the
libunwind behavior more closely.
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Milian Wolff <milian.wolff@kdab.com>
---
tools/perf/util/unwind-libdw.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/tools/perf/util/unwind-libdw.c b/tools/perf/util/unwind-libdw.c
index da45c4be5fb3..7755a5e0fe5e 100644
--- a/tools/perf/util/unwind-libdw.c
+++ b/tools/perf/util/unwind-libdw.c
@@ -178,6 +178,14 @@ frame_callback(Dwfl_Frame *state, void *arg)
Dwarf_Addr pc;
bool isactivation;
+ if (!dwfl_frame_pc(state, &pc, NULL)) {
+ pr_err("%s", dwfl_errmsg(-1));
+ return DWARF_CB_ABORT;
+ }
+
+ // report the module before we query for isactivation
+ report_module(pc, ui);
+
if (!dwfl_frame_pc(state, &pc, &isactivation)) {
pr_err("%s", dwfl_errmsg(-1));
return DWARF_CB_ABORT;
--
2.13.0
next prev parent reply index
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-06-02 14:37 [PATCH 1/2] perf report: ensure the perf DSO mapping matches what libdw sees Milian Wolff
2017-06-02 14:37 ` Milian Wolff [this message]
2017-06-02 15:25 ` [PATCH 2/2] perf report: report module before querying isactivation in dwfl unwind Arnaldo Carvalho de Melo
2017-06-05 10:47 ` Jiri Olsa
2017-06-15 19:20 ` Jan Kratochvil
2017-06-16 16:06 ` Arnaldo Carvalho de Melo
2017-06-16 17:36 ` Arnaldo Carvalho de Melo
2017-06-02 15:23 ` [PATCH 1/2] perf report: ensure the perf DSO mapping matches what libdw sees Arnaldo Carvalho de Melo
2017-06-02 16:21 ` Milian Wolff
2017-06-02 19:49 ` Arnaldo Carvalho de Melo
2017-06-03 11:36 ` Milian Wolff
2017-06-05 10:41 ` Jiri Olsa
2017-06-15 20:38 ` Jan Kratochvil
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20170602143753.16907-2-milian.wolff@kdab.com \
--to=milian.wolff@kdab.com \
--cc=Linux-kernel@vger.kernel.org \
--cc=acme@redhat.com \
--cc=jolsa@redhat.com \
--cc=linux-perf-users@vger.kernel.org \
--cc=namhyung@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Linux-perf-users Archive on lore.kernel.org
Archives are clonable:
git clone --mirror https://lore.kernel.org/linux-perf-users/0 linux-perf-users/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 linux-perf-users linux-perf-users/ https://lore.kernel.org/linux-perf-users \
linux-perf-users@vger.kernel.org
public-inbox-index linux-perf-users
Example config snippet for mirrors
Newsgroup available over NNTP:
nntp://nntp.lore.kernel.org/org.kernel.vger.linux-perf-users
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git