linux-kernel - Re: [PATCH 4/9] perf session: Create kernel maps in the constructor

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20100128182951.GB2499@ghostprotocols.net>
Date:	Thu, 28 Jan 2010 16:29:51 -0200
From:	Arnaldo Carvalho de Melo <acme@...radead.org>
To:	Masami Hiramatsu <mhiramat@...hat.com>
Cc:	Ingo Molnar <mingo@...e.hu>, linux-kernel@...r.kernel.org,
	Frédéric Weisbecker <fweisbec@...il.com>,
	Mike Galbraith <efault@....de>,
	Peter Zijlstra <a.p.zijlstra@...llo.nl>,
	Paul Mackerras <paulus@...ba.org>
Subject: Re: [PATCH 4/9] perf session: Create kernel maps in the constructor

Em Thu, Jan 28, 2010 at 11:28:29AM -0500, Masami Hiramatsu escreveu:

> Yeah, and if you can separate an interface only for handling symbols
> from debug/elf binaries, it will be helpful for me too.

It was done on this same series, now you can do as the patch at the end
of this message, that will be on my next series, does.

Testing with this patch:

[root@...pio ~]# perf probe --list
[root@...pio ~]# perf probe schedule
Added new event:
  probe:schedule                           (on schedule+0)

You can now use it on all perf tools, such as:

	perf record -e probe:schedule -a sleep 1

[root@...pio ~]# perf record -e probe:schedule -a sleep 1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.158 MB perf.data (~6923 samples) ]
[root@...pio ~]# perf report | head -9
# Samples: 4
#
# Overhead          Command          Shared Object  Symbol
# ........  ...............  .....................  ......
#
    25.00%             perf  libc-2.10.2.so         [.] __GI___libc_read
    25.00%      kondemand/1  [kernel.kallsyms]      [k] kernel_thread_helper
    25.00%      kondemand/0  [kernel.kallsyms]      [k] kernel_thread_helper
[root@...pio ~]#

And what about support for modules? You have everything you need there
as well :-)

If you do:

	struct map *mod;
	struct symbol *sym;

	mod = map_groups__find_by_name(&session.kmap_groups,
				       MAP__FUNCTION, "[ipv6]");

	sym = map__find_symbol_by_name(module, "ip6_xmit");

you'll get ipv6.ko located and loaded, mod->dso->long_name thus will
have the file to use with libdwarf and sym->start will have the
unrelocated address, and mod->start will have where ipv6 is loaded in
kernel space.

Also you should be able to find variables by name or address too, just
use MAP__VARIABLE where above you see MAP__FUNCTION.

- Arnaldo

>From 387ab05301cce32c8b091ba9feb94c906124e8a8 Mon Sep 17 00:00:00 2001
From: Arnaldo Carvalho de Melo <acme@...hat.com>
Date: Thu, 28 Jan 2010 16:12:31 -0200
Subject: [PATCH 1/1] perf probe: Don't use a perf_session instance just to resolve symbols
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8bit

With the recent modifications done to untie the session and symbol
layers, 'perf probe' now can use just the symbols layer.

Cc: Frédéric Weisbecker <fweisbec@...il.com>
Cc: Masami Hiramatsu <mhiramat@...hat.com>
Cc: Mike Galbraith <efault@....de>
Cc: Peter Zijlstra <a.p.zijlstra@...llo.nl>
Cc: Paul Mackerras <paulus@...ba.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@...hat.com>
---
 tools/perf/builtin-probe.c |   15 +++++++++------
 1 files changed, 9 insertions(+), 6 deletions(-)

diff --git a/tools/perf/builtin-probe.c b/tools/perf/builtin-probe.c
index 4fa73ec..cd45e02 100644
--- a/tools/perf/builtin-probe.c
+++ b/tools/perf/builtin-probe.c
@@ -41,7 +41,6 @@
 #include "util/debugfs.h"
 #include "util/symbol.h"
 #include "util/thread.h"
-#include "util/session.h"
 #include "util/parse-options.h"
 #include "util/parse-events.h"	/* For debugfs_path */
 #include "util/probe-finder.h"
@@ -59,7 +58,8 @@ static struct {
 	int nr_probe;
 	struct probe_point probes[MAX_PROBES];
 	struct strlist *dellist;
-	struct perf_session *psession;
+	struct map_groups kmap_groups;
+	struct map *kmaps[MAP__NR_TYPES];
 	struct map *kmap;
 	struct line_range line_range;
 } session;
@@ -212,10 +212,13 @@ static void init_vmlinux(void)
 		pr_debug("Use vmlinux: %s\n", symbol_conf.vmlinux_name);
 	if (symbol__init() < 0)
 		die("Failed to init symbol map.");
-	session.psession = perf_session__new(NULL, O_WRONLY, false);
-	if (session.psession == NULL)
-		die("Failed to init perf_session.");
-	session.kmap = session.psession->vmlinux_maps[MAP__FUNCTION];
+
+	map_groups__init(&session.kmap_groups);
+	if (map_groups__create_kernel_maps(&session.kmap_groups,
+					   session.kmaps) < 0)
+		die("Failed to create kernel maps.");
+
+	session.kmap = session.kmaps[MAP__FUNCTION];
 	if (!session.kmap)
 		die("Could not find kernel map.\n");
 }
-- 
1.6.2.5

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/