linux-kernel - [PATCH] perf tools: Speed up report for perf compiled with linwunwind

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-next>] [day] [month] [year] [list]

Message-Id: <20190426073804.17238-1-jolsa@kernel.org>
Date:   Fri, 26 Apr 2019 09:38:04 +0200
From:   Jiri Olsa <jolsa@...nel.org>
To:     Arnaldo Carvalho de Melo <acme@...nel.org>
Cc:     lkml <linux-kernel@...r.kernel.org>,
        Ingo Molnar <mingo@...nel.org>,
        Namhyung Kim <namhyung@...nel.org>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Peter Zijlstra <a.p.zijlstra@...llo.nl>
Subject: [PATCH] perf tools: Speed up report for perf compiled with linwunwind

When compiled with libunwind, perf does some preparatory work
when processing side-band events. This is not needed when report
actually don't unwind dwarf callchains, so it's disabled with
dwarf_callchain_users bool.

However we could move that check to higher level and shield more
unwanted code for normal report processing, giving us following
speed up on kernel build profile:

Before:
  $ perf record make -j40
  ...
  $ ll ../../perf.data
  -rw-------. 1 jolsa jolsa 461783932 Apr 26 09:11 perf.data
  $ perf stat -e cycles:u,instructions:u perf report -i perf.data > out

   Performance counter stats for 'perf report -i perf.data':

    78,669,920,155      cycles:u
    99,076,431,951      instructions:u            #    1.26  insn per cycle

      55.382823668 seconds time elapsed

      27.512341000 seconds user
      27.712871000 seconds sys

After:
  $ perf stat -e cycles:u,instructions:u perf report -i perf.data > out

   Performance counter stats for 'perf report -i perf.data':

    59,626,798,904      cycles:u
    88,583,575,849      instructions:u            #    1.49  insn per cycle

      21.296935559 seconds time elapsed

      20.010191000 seconds user
       1.202935000 seconds sys

The speed is higher with profile having many side-band events,
because these trigger libunwind preparatory code.

This does not apply for perf compiled with libdw for dwarf unwind,
only for build with libunwind.

Link: http://lkml.kernel.org/n/tip-rqm3489ghqq07fdrcpniynfs@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@...nel.org>
---
 tools/perf/util/thread.c                 |  3 ++-
 tools/perf/util/unwind-libunwind-local.c |  6 ------
 tools/perf/util/unwind-libunwind.c       | 10 ++++++++++
 3 files changed, 12 insertions(+), 7 deletions(-)

diff --git a/tools/perf/util/thread.c b/tools/perf/util/thread.c
index 50678d318185..403045a2bbea 100644
--- a/tools/perf/util/thread.c
+++ b/tools/perf/util/thread.c
@@ -15,6 +15,7 @@
 #include "map.h"
 #include "symbol.h"
 #include "unwind.h"
+#include "callchain.h"
 
 #include <api/fs/fs.h>
 
@@ -327,7 +328,7 @@ static int thread__prepare_access(struct thread *thread)
 {
 	int err = 0;
 
-	if (symbol_conf.use_callchain)
+	if (dwarf_callchain_users)
 		err = __thread__prepare_access(thread);
 
 	return err;
diff --git a/tools/perf/util/unwind-libunwind-local.c b/tools/perf/util/unwind-libunwind-local.c
index f3c666a84e4d..25e1406b1f8b 100644
--- a/tools/perf/util/unwind-libunwind-local.c
+++ b/tools/perf/util/unwind-libunwind-local.c
@@ -617,8 +617,6 @@ static unw_accessors_t accessors = {
 
 static int _unwind__prepare_access(struct thread *thread)
 {
-	if (!dwarf_callchain_users)
-		return 0;
 	thread->addr_space = unw_create_addr_space(&accessors, 0);
 	if (!thread->addr_space) {
 		pr_err("unwind: Can't create unwind address space.\n");
@@ -631,15 +629,11 @@ static int _unwind__prepare_access(struct thread *thread)
 
 static void _unwind__flush_access(struct thread *thread)
 {
-	if (!dwarf_callchain_users)
-		return;
 	unw_flush_cache(thread->addr_space, 0, 0);
 }
 
 static void _unwind__finish_access(struct thread *thread)
 {
-	if (!dwarf_callchain_users)
-		return;
 	unw_destroy_addr_space(thread->addr_space);
 }
 
diff --git a/tools/perf/util/unwind-libunwind.c b/tools/perf/util/unwind-libunwind.c
index 9778b3133b77..c0811977d7d5 100644
--- a/tools/perf/util/unwind-libunwind.c
+++ b/tools/perf/util/unwind-libunwind.c
@@ -5,6 +5,7 @@
 #include "session.h"
 #include "debug.h"
 #include "env.h"
+#include "callchain.h"
 
 struct unwind_libunwind_ops __weak *local_unwind_libunwind_ops;
 struct unwind_libunwind_ops __weak *x86_32_unwind_libunwind_ops;
@@ -24,6 +25,9 @@ int unwind__prepare_access(struct thread *thread, struct map *map,
 	struct unwind_libunwind_ops *ops = local_unwind_libunwind_ops;
 	int err;
 
+	if (!dwarf_callchain_users)
+		return 0;
+
 	if (thread->addr_space) {
 		pr_debug("unwind: thread map already set, dso=%s\n",
 			 map->dso->name);
@@ -65,12 +69,18 @@ int unwind__prepare_access(struct thread *thread, struct map *map,
 
 void unwind__flush_access(struct thread *thread)
 {
+	if (!dwarf_callchain_users)
+		return;
+
 	if (thread->unwind_libunwind_ops)
 		thread->unwind_libunwind_ops->flush_access(thread);
 }
 
 void unwind__finish_access(struct thread *thread)
 {
+	if (!dwarf_callchain_users)
+		return;
+
 	if (thread->unwind_libunwind_ops)
 		thread->unwind_libunwind_ops->finish_access(thread);
 }
-- 
2.20.1