Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1334248 > unrolled thread

[PATCH 00/23] perf tools: Several memory events updates

Started byJiri Olsa <jolsa@kernel.org>
First post2016-02-15 09:40 +0100
Last post2016-02-15 09:50 +0100
Articles 20 on this page of 25 — 5 participants

Back to article view | Back to linux.kernel

Contents

  [PATCH 00/23] perf tools: Several memory events updates Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
    [PATCH 10/23] perf x86 intel: Add DATALA events into sysfs Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
    [PATCH 11/23] perf mem: Add Intel DATALA memory events Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
    [PATCH 18/23] perf tools: Change perf_mem__lvl_scnprintf to return nb of displayed bytes Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
    [PATCH 13/23] perf tools: Introduce perf_mem__tlb_scnprintf function Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
    [PATCH 09/23] perf mem: Add -u/-k options Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
    [PATCH 03/23] perf tools: Introduce cl_offset function Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
    [PATCH 06/23] perf mem: Check for memory events support Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
    [PATCH 15/23] perf tools: Introduce perf_mem__snp_scnprintf function Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
    [PATCH 01/23] perf record: Add --all-user/--all-kernel options Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
      Re: [PATCH 01/23] perf record: Add --all-user/--all-kernel options Arnaldo Carvalho de Melo <acme@kernel.org> - 2016-02-16 18:00 +0100
        Re: [PATCH 01/23] perf record: Add --all-user/--all-kernel options Andi Kleen <ak@linux.intel.com> - 2016-02-17 04:20 +0100
          Re: [PATCH 01/23] perf record: Add --all-user/--all-kernel options Arnaldo Carvalho de Melo <acme@kernel.org> - 2016-02-17 15:30 +0100
            Re: [PATCH 01/23] perf record: Add --all-user/--all-kernel options Andi Kleen <ak@linux.intel.com> - 2016-02-17 16:40 +0100
        Re: [PATCH 01/23] perf record: Add --all-user/--all-kernel options Jiri Olsa <jolsa@redhat.com> - 2016-02-17 15:40 +0100
          Re: [PATCH 01/23] perf record: Add --all-user/--all-kernel options Arnaldo Carvalho de Melo <acme@kernel.org> - 2016-02-17 16:00 +0100
      [tip:perf/core] perf record: Add --all-user/--all-kernel options tip-bot for Jiri Olsa <tipbot@zytor.com> - 2016-02-20 12:40 +0100
    [PATCH 17/23] perf tools: Change perf_mem__tlb_scnprintf to return nb of displayed bytes Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
    [PATCH 23/23] perf script: Display data_src values Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
    [PATCH 05/23] perf mem: Add -e record option Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
    [PATCH 14/23] perf tools: Introduce perf_mem__lvl_scnprintf function Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
    [PATCH 21/23] perf script: Add data_src and weight column definitions Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:40 +0100
    [PATCH 07/23] perf mem: Introduce perf_mem_events__name function Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:50 +0100
    [PATCH 04/23] perf tools: Add monitored events array Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:50 +0100
    [PATCH 02/23] perf tools: Make cl_address global Jiri Olsa <jolsa@kernel.org> - 2016-02-15 09:50 +0100

Page 1 of 2  [1] 2  Next page →

#1334248 — [PATCH 00/23] perf tools: Several memory events updates

FromJiri Olsa <jolsa@kernel.org>
Date2016-02-15 09:40 +0100
Subject[PATCH 00/23] perf tools: Several memory events updates
Message-ID<r2mRk-6Rc-3@gated-at.bofh.it>
hi,
this is generic part of the c2c patchset. It contains
several enhancements for perf mem and script commands:

  - display memory events properties in perf script
  - perf mem DATALA events support
  - few other assorted fixes/options

Also available in here:
  git://git.kernel.org/pub/scm/linux/kernel/git/jolsa/perf.git
  perf/mem

thanks,
jirka


---
Jiri Olsa (23):
      perf record: Add --all-user/--all-kernel options
      perf tools: Make cl_address global
      perf tools: Introduce cl_offset function
      perf tools: Add monitored events array
      perf mem: Add -e record option
      perf mem: Check for memory events support
      perf mem: Introduce perf_mem_events__name function
      perf mem: Add -l/--ldlat option
      perf mem: Add -u/-k options
      perf x86 intel: Add DATALA events into sysfs
      perf mem: Add Intel DATALA memory events
      perf tools: Use ARRAY_SIZE in mem sort display functions
      perf tools: Introduce perf_mem__tlb_scnprintf function
      perf tools: Introduce perf_mem__lvl_scnprintf function
      perf tools: Introduce perf_mem__snp_scnprintf function
      perf tools: Introduce perf_mem__lck_scnprintf function
      perf tools: Change perf_mem__tlb_scnprintf to return nb of displayed bytes
      perf tools: Change perf_mem__lvl_scnprintf to return nb of displayed bytes
      perf tools: Change perf_mem__snp_scnprintf to return nb of displayed bytes
      perf tools: Change perf_mem__lck_scnprintf to return nb of displayed bytes
      perf script: Add data_src and weight column definitions
      perf script: Display addr/data_src/weight columns for raw events
      perf script: Display data_src values

 arch/x86/kernel/cpu/perf_event_intel.c   |  39 +++++++++++++
 tools/perf/Documentation/perf-record.txt |   6 ++
 tools/perf/builtin-mem.c                 |  96 +++++++++++++++++++++++++++---
 tools/perf/builtin-record.c              |   4 ++
 tools/perf/builtin-script.c              |  41 ++++++++++++-
 tools/perf/perf.h                        |   2 +
 tools/perf/util/Build                    |   1 +
 tools/perf/util/evsel.c                  |  10 ++++
 tools/perf/util/mem-events.c             | 289 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/mem-events.h             |  56 ++++++++++++++++++
 tools/perf/util/sort.c                   | 148 ++--------------------------------------------
 tools/perf/util/sort.h                   |  11 ++++
 12 files changed, 551 insertions(+), 152 deletions(-)
 create mode 100644 tools/perf/util/mem-events.c
 create mode 100644 tools/perf/util/mem-events.h

[toc] | [next] | [standalone]

#1334250 — [PATCH 10/23] perf x86 intel: Add DATALA events into sysfs

FromJiri Olsa <jolsa@kernel.org>
Date2016-02-15 09:40 +0100
Subject[PATCH 10/23] perf x86 intel: Add DATALA events into sysfs
Message-ID<r2mRl-6Rc-19@gated-at.bofh.it>
In reply to#1334248
Adding DATALA events into Haaswell events sysfs
directory, so it's easier to use them.

Adding all DATALA events as specified by Intel SDM
manual 18.10.3 PEBS Data Address Profiling.

Link: http://lkml.kernel.org/n/tip-dvpk5ys80v41oh5xgqvloviw@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 arch/x86/kernel/cpu/perf_event_intel.c | 39 ++++++++++++++++++++++++++++++++++
 1 file changed, 39 insertions(+)

diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index fed2ab1f1065..aadccb782e78 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -3253,6 +3253,26 @@ static __init void intel_ht_bug(void)
 EVENT_ATTR_STR(mem-loads,	mem_ld_hsw,	"event=0xcd,umask=0x1,ldlat=3");
 EVENT_ATTR_STR(mem-stores,	mem_st_hsw,	"event=0xd0,umask=0x82")
 
+EVENT_ATTR_STR(mem-stlb-miss-loads,	hsw_stlb_miss_loads,	"event=0xd0,umask=0x11");
+EVENT_ATTR_STR(mem-stlb-miss-stores,	hsw_stlb_miss_stores,	"event=0xd0,umask=0x12");
+EVENT_ATTR_STR(mem-lock-loads,		hsw_lock_loads,		"event=0xd0,umask=0x21");
+EVENT_ATTR_STR(mem-split-loads,		hsw_split_loads,	"event=0xd0,umask=0x41");
+EVENT_ATTR_STR(mem-split-stores,	hsw_split_stores,	"event=0xd0,umask=0x42");
+EVENT_ATTR_STR(mem-all-loads,		hsw_all_loads,		"event=0xd0,umask=0x81");
+EVENT_ATTR_STR(mem-all-stores,		hsw_all_stores,		"event=0xd0,umask=0x82");
+EVENT_ATTR_STR(mem-load-l1-hit,		hsw_load_l1_hit,	"event=0xd1,umask=0x01");
+EVENT_ATTR_STR(mem-load-l2-hit,		hsw_load_l2_hit,	"event=0xd1,umask=0x02");
+EVENT_ATTR_STR(mem-load-l3-hit,		hsw_load_l3_hit,	"event=0xd1,umask=0x04");
+EVENT_ATTR_STR(mem-load-l1-miss,	hsw_load_l1_miss,	"event=0xd1,umask=0x08");
+EVENT_ATTR_STR(mem-load-l2-miss,	hsw_load_l2_miss,	"event=0xd1,umask=0x10");
+EVENT_ATTR_STR(mem-load-l3-miss,	hsw_load_l3_miss,	"event=0xd1,umask=0x20");
+EVENT_ATTR_STR(mem-load-hit-lfb,	hsw_load_hit_lfb,	"event=0xd1,umask=0x40");
+EVENT_ATTR_STR(mem-snp-miss,		hsw_snp_miss,		"event=0xd2,umask=0x01");
+EVENT_ATTR_STR(mem-snp-hit,		hsw_snp_hit,		"event=0xd2,umask=0x02");
+EVENT_ATTR_STR(mem-snp-hitm,		hsw_snp_hitm,		"event=0xd2,umask=0x04");
+EVENT_ATTR_STR(mem-snp-none,		hsw_snp_none,		"event=0xd2,umask=0x08");
+EVENT_ATTR_STR(mem-local-dram,		hsw_local_dram,		"event=0xd3,umask=0x01");
+
 /* Haswell special events */
 EVENT_ATTR_STR(tx-start,	tx_start,	"event=0xc9,umask=0x1");
 EVENT_ATTR_STR(tx-commit,	tx_commit,	"event=0xc9,umask=0x2");
@@ -3282,6 +3302,25 @@ static struct attribute *hsw_events_attrs[] = {
 	EVENT_PTR(cycles_ct),
 	EVENT_PTR(mem_ld_hsw),
 	EVENT_PTR(mem_st_hsw),
+	EVENT_PTR(hsw_stlb_miss_loads),
+	EVENT_PTR(hsw_stlb_miss_stores),
+	EVENT_PTR(hsw_lock_loads),
+	EVENT_PTR(hsw_split_loads),
+	EVENT_PTR(hsw_split_stores),
+	EVENT_PTR(hsw_all_loads),
+	EVENT_PTR(hsw_all_stores),
+	EVENT_PTR(hsw_load_l1_hit),
+	EVENT_PTR(hsw_load_l2_hit),
+	EVENT_PTR(hsw_load_l3_hit),
+	EVENT_PTR(hsw_load_l1_miss),
+	EVENT_PTR(hsw_load_l2_miss),
+	EVENT_PTR(hsw_load_l3_miss),
+	EVENT_PTR(hsw_load_hit_lfb),
+	EVENT_PTR(hsw_snp_miss),
+	EVENT_PTR(hsw_snp_hit),
+	EVENT_PTR(hsw_snp_hitm),
+	EVENT_PTR(hsw_snp_none),
+	EVENT_PTR(hsw_local_dram),
 	NULL
 };
 
-- 
2.4.3

[toc] | [prev] | [next] | [standalone]

#1334253 — [PATCH 11/23] perf mem: Add Intel DATALA memory events

FromJiri Olsa <jolsa@kernel.org>
Date2016-02-15 09:40 +0100
Subject[PATCH 11/23] perf mem: Add Intel DATALA memory events
Message-ID<r2mRl-6Rc-35@gated-at.bofh.it>
In reply to#1334248
Adding Intel DATALA memory events into
perf mem record command:

  $ perf mem record -e list
  [ok] ldlat-loads
  [ok] ldlat-stores
       stlb-miss-loads
       stlb-miss-stores
       lock-loads
       split-loads
       split-stores
       all-loads
       all-stores
       l1-hit
       l2-hit
       l3-hit
       l1-miss
       l2-miss
       l3-miss
       lfb
       snp-miss
       snp-hit
       snp-hitm
       snp-none
       local-dram

All events from above lists are compiled in, but only
those with [ok] are supported by kernel/HW. You need
previous DATALA patch to support all of them.

Link: http://lkml.kernel.org/n/tip-cximteb8r1a6750wz8sgeajh@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/util/mem-events.c | 19 +++++++++++++++++++
 tools/perf/util/mem-events.h | 19 +++++++++++++++++++
 2 files changed, 38 insertions(+)

diff --git a/tools/perf/util/mem-events.c b/tools/perf/util/mem-events.c
index 3772a3a8a6ee..92d3bc9914e6 100644
--- a/tools/perf/util/mem-events.c
+++ b/tools/perf/util/mem-events.c
@@ -16,6 +16,25 @@ unsigned int perf_mem_events__loads_ldlat = 30;
 struct perf_mem_event perf_mem_events[PERF_MEM_EVENTS__MAX] = {
 	E("ldlat-loads",	"cpu/mem-loads,ldlat=%u/P",	"mem-loads"),
 	E("ldlat-stores",	"cpu/mem-stores/P",		"mem-stores"),
+	E("stlb-miss-loads",	"cpu/mem-stlb-miss-loads/P",	"mem-stlb-miss-loads"),
+	E("stlb-miss-stores",	"cpu/mem-stlb-miss-stores/P",	"mem-stlb-miss-stores"),
+	E("lock-loads",		"cpu/mem-lock-loads/P",		"mem-lock-loads"),
+	E("split-loads",	"cpu/mem-split-loads/P",	"mem-split-loads"),
+	E("split-stores",	"cpu/mem-split-stores/P",	"mem-split-stores"),
+	E("all-loads",		"cpu/mem-all-loads/P",		"mem-all-loads"),
+	E("all-stores",		"cpu/mem-all-stores/P",		"mem-all-stores"),
+	E("l1-hit",		"cpu/mem-load-l1-hit/P",	"mem-load-l1-hit"),
+	E("l2-hit",		"cpu/mem-load-l2-hit/P",	"mem-load-l2-hit"),
+	E("l3-hit",		"cpu/mem-load-l3-hit/P",	"mem-load-l3-hit"),
+	E("l1-miss",		"cpu/mem-load-l1-miss/P",	"mem-load-l1-miss"),
+	E("l2-miss",		"cpu/mem-load-l2-miss/P",	"mem-load-l2-miss"),
+	E("l3-miss",		"cpu/mem-load-l3-miss/P",	"mem-load-l3-miss"),
+	E("lfb",		"cpu/mem-load-hit-lfb/P",	"mem-load-hit-lfb"),
+	E("snp-miss",		"cpu/mem-snp-miss/P",		"mem-snp-miss"),
+	E("snp-hit",		"cpu/mem-snp-hit/P",		"mem-snp-hit"),
+	E("snp-hitm",		"cpu/mem-snp-hitm/P",		"mem-snp-hitm"),
+	E("snp-none",		"cpu/mem-snp-none/P",		"mem-snp-none"),
+	E("local-dram",		"cpu/mem-local-dram/P",		"mem-local-dram"),
 };
 #undef E
 
diff --git a/tools/perf/util/mem-events.h b/tools/perf/util/mem-events.h
index d690907f7e50..4abf606b256b 100644
--- a/tools/perf/util/mem-events.h
+++ b/tools/perf/util/mem-events.h
@@ -14,6 +14,25 @@ struct perf_mem_event {
 enum {
 	PERF_MEM_EVENTS__LOAD,
 	PERF_MEM_EVENTS__STORE,
+	PERF_MEM_EVENTS__STLB_MISS_LOADS,
+	PERF_MEM_EVENTS__STLB_MISS_STORES,
+	PERF_MEM_EVENTS__LOCK_LOADS,
+	PERF_MEM_EVENTS__SPLIT_LOADS,
+	PERF_MEM_EVENTS__SPLIT_STORES,
+	PERF_MEM_EVENTS__ALL_LOADS,
+	PERF_MEM_EVENTS__ALL_STORES,
+	PERF_MEM_EVENTS__L1_HIT,
+	PERF_MEM_EVENTS__L2_HIT,
+	PERF_MEM_EVENTS__L3_HIT,
+	PERF_MEM_EVENTS__L1_MISS,
+	PERF_MEM_EVENTS__L2_MISS,
+	PERF_MEM_EVENTS__L3_MISS,
+	PERF_MEM_EVENTS__LFB,
+	PERF_MEM_EVENTS__SNP_MISS,
+	PERF_MEM_EVENTS__SNP_HIT,
+	PERF_MEM_EVENTS__SNP_HITM,
+	PERF_MEM_EVENTS__SNP_NONE,
+	PERF_MEM_EVENTS__LOCAL_DRAM,
 	PERF_MEM_EVENTS__MAX,
 };
 
-- 
2.4.3

[toc] | [prev] | [next] | [standalone]

#1334254 — [PATCH 18/23] perf tools: Change perf_mem__lvl_scnprintf to return nb of displayed bytes

FromJiri Olsa <jolsa@kernel.org>
Date2016-02-15 09:40 +0100
Subject[PATCH 18/23] perf tools: Change perf_mem__lvl_scnprintf to return nb of displayed bytes
Message-ID<r2mRm-6Rc-37@gated-at.bofh.it>
In reply to#1334248
Moving strncat/strcpy calls into scnprintf to easily track
number of displayed bytes. It will be used in following patch.

Link: http://lkml.kernel.org/n/tip-9q9h9trua9zljnw26c56memh@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/util/mem-events.c | 13 +++++++------
 tools/perf/util/mem-events.h |  2 +-
 2 files changed, 8 insertions(+), 7 deletions(-)

diff --git a/tools/perf/util/mem-events.c b/tools/perf/util/mem-events.c
index 440221ca4f8a..5f713d77ac8a 100644
--- a/tools/perf/util/mem-events.c
+++ b/tools/perf/util/mem-events.c
@@ -183,7 +183,7 @@ static const char * const mem_lvl[] = {
 	"Uncached",
 };
 
-void perf_mem__lvl_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
+int perf_mem__lvl_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
 {
 	size_t i, l = 0;
 	u64 m =  PERF_MEM_LVL_NA;
@@ -208,15 +208,16 @@ void perf_mem__lvl_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
 			strcat(out, " or ");
 			l += 4;
 		}
-		strncat(out, mem_lvl[i], sz - l);
-		l += strlen(mem_lvl[i]);
+		l += scnprintf(out + l, sz - l, mem_lvl[i]);
 	}
 	if (*out == '\0')
-		strcpy(out, "N/A");
+		l += scnprintf(out, sz - l, "N/A");
 	if (hit)
-		strncat(out, " hit", sz - l);
+		l += scnprintf(out + l, sz - l, " hit");
 	if (miss)
-		strncat(out, " miss", sz - l);
+		l += scnprintf(out + l, sz - l, " miss");
+
+	return l;
 }
 
 static const char * const snoop_access[] = {
diff --git a/tools/perf/util/mem-events.h b/tools/perf/util/mem-events.h
index ce9ed0540014..a06efff4ef09 100644
--- a/tools/perf/util/mem-events.h
+++ b/tools/perf/util/mem-events.h
@@ -46,7 +46,7 @@ char *perf_mem_events__name(int i);
 
 struct mem_info;
 int perf_mem__tlb_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
-void perf_mem__lvl_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
+int perf_mem__lvl_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
 void perf_mem__snp_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
 void perf_mem__lck_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
 
-- 
2.4.3

[toc] | [prev] | [next] | [standalone]

#1334255 — [PATCH 13/23] perf tools: Introduce perf_mem__tlb_scnprintf function

FromJiri Olsa <jolsa@kernel.org>
Date2016-02-15 09:40 +0100
Subject[PATCH 13/23] perf tools: Introduce perf_mem__tlb_scnprintf function
Message-ID<r2mRl-6Rc-29@gated-at.bofh.it>
In reply to#1334248
Move meminfo's tlb display function into mem-events.c
object, so it could be reused later from script code.

Link: http://lkml.kernel.org/n/tip-kdsvxdm3ucwknyvkluwavydh@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/util/mem-events.c | 47 ++++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/mem-events.h |  3 +++
 tools/perf/util/sort.c       | 44 ++---------------------------------------
 3 files changed, 52 insertions(+), 42 deletions(-)

diff --git a/tools/perf/util/mem-events.c b/tools/perf/util/mem-events.c
index 92d3bc9914e6..68f6f8e4922a 100644
--- a/tools/perf/util/mem-events.c
+++ b/tools/perf/util/mem-events.c
@@ -8,6 +8,7 @@
 #include <api/fs/fs.h>
 #include "mem-events.h"
 #include "debug.h"
+#include "symbol.h"
 
 unsigned int perf_mem_events__loads_ldlat = 30;
 
@@ -117,3 +118,49 @@ int perf_mem_events__init(void)
 
 	return found ? 0 : -ENOENT;
 }
+
+static const char * const tlb_access[] = {
+	"N/A",
+	"HIT",
+	"MISS",
+	"L1",
+	"L2",
+	"Walker",
+	"Fault",
+};
+
+void perf_mem__tlb_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
+{
+	size_t l = 0, i;
+	u64 m = PERF_MEM_TLB_NA;
+	u64 hit, miss;
+
+	sz -= 1; /* -1 for null termination */
+	out[0] = '\0';
+
+	if (mem_info)
+		m = mem_info->data_src.mem_dtlb;
+
+	hit = m & PERF_MEM_TLB_HIT;
+	miss = m & PERF_MEM_TLB_MISS;
+
+	/* already taken care of */
+	m &= ~(PERF_MEM_TLB_HIT|PERF_MEM_TLB_MISS);
+
+	for (i = 0; m && i < ARRAY_SIZE(tlb_access); i++, m >>= 1) {
+		if (!(m & 0x1))
+			continue;
+		if (l) {
+			strcat(out, " or ");
+			l += 4;
+		}
+		strncat(out, tlb_access[i], sz - l);
+		l += strlen(tlb_access[i]);
+	}
+	if (*out == '\0')
+		strcpy(out, "N/A");
+	if (hit)
+		strncat(out, " hit", sz - l);
+	if (miss)
+		strncat(out, " miss", sz - l);
+}
diff --git a/tools/perf/util/mem-events.h b/tools/perf/util/mem-events.h
index 4abf606b256b..1af23966e3fc 100644
--- a/tools/perf/util/mem-events.h
+++ b/tools/perf/util/mem-events.h
@@ -43,5 +43,8 @@ int perf_mem_events__parse(const char *str);
 int perf_mem_events__init(void);
 
 char *perf_mem_events__name(int i);
+
+struct mem_info;
+void perf_mem__tlb_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
 #endif /* __PERF_MEM_EVENTS_H */
 
diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
index 5cb3e13b8f2e..b7834a5a002d 100644
--- a/tools/perf/util/sort.c
+++ b/tools/perf/util/sort.c
@@ -6,6 +6,7 @@
 #include "evsel.h"
 #include "evlist.h"
 #include <traceevent/event-parse.h>
+#include "mem-events.h"
 
 regex_t		parent_regex;
 const char	default_parent_pattern[] = "^sys_|^do_page_fault";
@@ -834,53 +835,12 @@ sort__tlb_cmp(struct hist_entry *left, struct hist_entry *right)
 	return (int64_t)(data_src_r.mem_dtlb - data_src_l.mem_dtlb);
 }
 
-static const char * const tlb_access[] = {
-	"N/A",
-	"HIT",
-	"MISS",
-	"L1",
-	"L2",
-	"Walker",
-	"Fault",
-};
-
 static int hist_entry__tlb_snprintf(struct hist_entry *he, char *bf,
 				    size_t size, unsigned int width)
 {
 	char out[64];
-	size_t sz = sizeof(out) - 1; /* -1 for null termination */
-	size_t l = 0, i;
-	u64 m = PERF_MEM_TLB_NA;
-	u64 hit, miss;
-
-	out[0] = '\0';
-
-	if (he->mem_info)
-		m = he->mem_info->data_src.mem_dtlb;
-
-	hit = m & PERF_MEM_TLB_HIT;
-	miss = m & PERF_MEM_TLB_MISS;
-
-	/* already taken care of */
-	m &= ~(PERF_MEM_TLB_HIT|PERF_MEM_TLB_MISS);
-
-	for (i = 0; m && i < ARRAY_SIZE(tlb_access); i++, m >>= 1) {
-		if (!(m & 0x1))
-			continue;
-		if (l) {
-			strcat(out, " or ");
-			l += 4;
-		}
-		strncat(out, tlb_access[i], sz - l);
-		l += strlen(tlb_access[i]);
-	}
-	if (*out == '\0')
-		strcpy(out, "N/A");
-	if (hit)
-		strncat(out, " hit", sz - l);
-	if (miss)
-		strncat(out, " miss", sz - l);
 
+	perf_mem__tlb_scnprintf(out, sizeof(out), he->mem_info);
 	return repsep_snprintf(bf, size, "%-*s", width, out);
 }
 
-- 
2.4.3

[toc] | [prev] | [next] | [standalone]

#1334256 — [PATCH 09/23] perf mem: Add -u/-k options

FromJiri Olsa <jolsa@kernel.org>
Date2016-02-15 09:40 +0100
Subject[PATCH 09/23] perf mem: Add -u/-k options
Message-ID<r2mRm-6Rc-39@gated-at.bofh.it>
In reply to#1334248
Add -u/-k (--all-user/--all-kernel) options to use
the perf record --all-user/--all-kernel options.

Link: http://lkml.kernel.org/n/tip-adxn5c48oe0gmjrjmq6we27q@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/builtin-mem.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/tools/perf/builtin-mem.c b/tools/perf/builtin-mem.c
index 94243429bb6e..cb4aa83162a6 100644
--- a/tools/perf/builtin-mem.c
+++ b/tools/perf/builtin-mem.c
@@ -62,6 +62,7 @@ static int __cmd_record(int argc, const char **argv, struct perf_mem *mem)
 	int rec_argc, i = 0, j;
 	const char **rec_argv;
 	int ret;
+	bool all_user = false, all_kernel = false;
 	struct option options[] = {
 	OPT_CALLBACK('e', "event", &mem, "event",
 		     "event selector. use 'perf mem record -e list' to list available events",
@@ -69,13 +70,15 @@ static int __cmd_record(int argc, const char **argv, struct perf_mem *mem)
 	OPT_UINTEGER('l', "ldlat", &perf_mem_events__loads_ldlat, "mem-loads latency"),
 	OPT_INCR('v', "verbose", &verbose,
 		 "be more verbose (show counter open errors, etc)"),
+	OPT_BOOLEAN('u', "--all-user", &all_user, "collect only user level data"),
+	OPT_BOOLEAN('k', "--all-kernel", &all_kernel, "collect only kernel level data"),
 	OPT_END()
 	};
 
 	argc = parse_options(argc, argv, options, record_mem_usage,
 			     PARSE_OPT_STOP_AT_NON_OPTION);
 
-	rec_argc = argc + 7; /* max number of arguments */
+	rec_argc = argc + 9; /* max number of arguments */
 	rec_argv = calloc(rec_argc + 1, sizeof(char *));
 	if (!rec_argv)
 		return -1;
@@ -104,6 +107,12 @@ static int __cmd_record(int argc, const char **argv, struct perf_mem *mem)
 		rec_argv[i++] = perf_mem_events__name(j);
 	};
 
+	if (all_user)
+		rec_argv[i++] = "--all-user";
+
+	if (all_kernel)
+		rec_argv[i++] = "--all-kernel";
+
 	for (j = 0; j < argc; j++, i++)
 		rec_argv[i] = argv[j];
 
-- 
2.4.3

[toc] | [prev] | [next] | [standalone]

#1334258 — [PATCH 03/23] perf tools: Introduce cl_offset function

FromJiri Olsa <jolsa@kernel.org>
Date2016-02-15 09:40 +0100
Subject[PATCH 03/23] perf tools: Introduce cl_offset function
Message-ID<r2mRm-6Rc-41@gated-at.bofh.it>
In reply to#1334248
It'll be used in following patches.

Link: http://lkml.kernel.org/n/tip-m7dfmhya60y30hzhboeaczqk@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/util/sort.h | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/tools/perf/util/sort.h b/tools/perf/util/sort.h
index 46f159f9d947..5b9c6246de6d 100644
--- a/tools/perf/util/sort.h
+++ b/tools/perf/util/sort.h
@@ -168,6 +168,12 @@ static inline u64 cl_address(u64 address)
 	return (address & ~(cacheline_size - 1));
 }
 
+static inline u64 cl_offset(u64 address)
+{
+	/* return the cacheline of the address */
+	return (address & (cacheline_size - 1));
+}
+
 enum sort_mode {
 	SORT_MODE__NORMAL,
 	SORT_MODE__BRANCH,
-- 
2.4.3

[toc] | [prev] | [next] | [standalone]

#1334259 — [PATCH 06/23] perf mem: Check for memory events support

FromJiri Olsa <jolsa@kernel.org>
Date2016-02-15 09:40 +0100
Subject[PATCH 06/23] perf mem: Check for memory events support
Message-ID<r2mRm-6Rc-45@gated-at.bofh.it>
In reply to#1334248
Check if current kernel support available memory events
and display the status within -e  list option:

  $ perf mem record -e list
  [ok] ldlat-loads
  [ok] ldlat-stores

Link: http://lkml.kernel.org/n/tip-r8smhxjfe69q67jwt7mzs0e9@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/builtin-mem.c     | 14 +++++++++++++-
 tools/perf/util/mem-events.c | 35 ++++++++++++++++++++++++++++++++---
 tools/perf/util/mem-events.h |  3 +++
 3 files changed, 48 insertions(+), 4 deletions(-)

diff --git a/tools/perf/builtin-mem.c b/tools/perf/builtin-mem.c
index b3f8a89ede40..e64e0aa7ba0e 100644
--- a/tools/perf/builtin-mem.c
+++ b/tools/perf/builtin-mem.c
@@ -40,7 +40,8 @@ static int parse_record_events(const struct option *opt,
 	for (j = 0; j < PERF_MEM_EVENTS__MAX; j++) {
 		struct perf_mem_event *e = &perf_mem_events[j];
 
-		fprintf(stderr, "%-20s%s",
+		fprintf(stderr, "%s%-20s%s",
+			e->supported ? "[ok] " : "     ",
 			e->tag, verbose ? "" : "\n");
 		if (verbose)
 			fprintf(stderr, " [%s]\n", e->name);
@@ -92,6 +93,12 @@ static int __cmd_record(int argc, const char **argv, struct perf_mem *mem)
 		if (!perf_mem_events[j].record)
 			continue;
 
+		if (!perf_mem_events[j].supported) {
+			pr_err("failed: event '%s' not supported\n",
+			       perf_mem_events[j].name);
+			return -1;
+		}
+
 		rec_argv[i++] = "-e";
 		rec_argv[i++] = perf_mem_events[j].name;
 	};
@@ -355,6 +362,11 @@ int cmd_mem(int argc, const char **argv, const char *prefix __maybe_unused)
 		NULL
 	};
 
+	if (perf_mem_events__init()) {
+		pr_err("failed: memory events not supported\n");
+		return -1;
+	}
+
 	argc = parse_options_subcommand(argc, argv, mem_options, mem_subcommands,
 					mem_usage, PARSE_OPT_STOP_AT_NON_OPTION);
 
diff --git a/tools/perf/util/mem-events.c b/tools/perf/util/mem-events.c
index b1507c04b257..e21853fe1312 100644
--- a/tools/perf/util/mem-events.c
+++ b/tools/perf/util/mem-events.c
@@ -2,15 +2,20 @@
 #include <stdlib.h>
 #include <string.h>
 #include <errno.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <unistd.h>
+#include <api/fs/fs.h>
 #include "mem-events.h"
 #include "debug.h"
 
-#define E(t, n) { .tag = t, .name = n }
+#define E(t, n, s) { .tag = t, .name = n, .sysfs_name = s }
 
 struct perf_mem_event perf_mem_events[PERF_MEM_EVENTS__MAX] = {
-	E("ldlat-loads",	"cpu/mem-loads,ldlat=30/P"),
-	E("ldlat-stores",	"cpu/mem-stores/P"),
+	E("ldlat-loads",	"cpu/mem-loads,ldlat=30/P",	"mem-loads"),
+	E("ldlat-stores",	"cpu/mem-stores/P",		"mem-stores"),
 };
+#undef E
 
 #undef E
 
@@ -49,3 +54,27 @@ int perf_mem_events__parse(const char *str)
 	pr_err("failed: event '%s' not found, use '-e list' to get list of available events\n", str);
 	return -1;
 }
+
+int perf_mem_events__init(void)
+{
+	const char *mnt = sysfs__mount();
+	bool found = false;
+	int j;
+
+	if (!mnt)
+		return -ENOENT;
+
+	for (j = 0; j < PERF_MEM_EVENTS__MAX; j++) {
+		char path[PATH_MAX];
+		struct perf_mem_event *e = &perf_mem_events[j];
+		struct stat st;
+
+		scnprintf(path, PATH_MAX, "%s/devices/cpu/events/%s",
+			  mnt, e->sysfs_name);
+
+		if (!stat(path, &st))
+			e->supported = found = true;
+	}
+
+	return found ? 0 : -ENOENT;
+}
diff --git a/tools/perf/util/mem-events.h b/tools/perf/util/mem-events.h
index 0b5247468462..b9a786346c03 100644
--- a/tools/perf/util/mem-events.h
+++ b/tools/perf/util/mem-events.h
@@ -5,8 +5,10 @@
 
 struct perf_mem_event {
 	bool		record;
+	bool		supported;
 	const char	*tag;
 	const char	*name;
+	const char	*sysfs_name;
 };
 
 enum {
@@ -18,6 +20,7 @@ enum {
 extern struct perf_mem_event perf_mem_events[PERF_MEM_EVENTS__MAX];
 
 int perf_mem_events__parse(const char *str);
+int perf_mem_events__init(void);
 
 #endif /* __PERF_MEM_EVENTS_H */
 
-- 
2.4.3

[toc] | [prev] | [next] | [standalone]

#1334260 — [PATCH 15/23] perf tools: Introduce perf_mem__snp_scnprintf function

FromJiri Olsa <jolsa@kernel.org>
Date2016-02-15 09:40 +0100
Subject[PATCH 15/23] perf tools: Introduce perf_mem__snp_scnprintf function
Message-ID<r2mRm-6Rc-43@gated-at.bofh.it>
In reply to#1334248
Move meminfo's snp display function into mem-events.c
object, so it could be reused later from script code.

Link: http://lkml.kernel.org/n/tip-zhd9q561o49xzhrtg8cz5jui@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/util/mem-events.c | 34 ++++++++++++++++++++++++++++++++++
 tools/perf/util/mem-events.h |  1 +
 tools/perf/util/sort.c       | 31 +------------------------------
 3 files changed, 36 insertions(+), 30 deletions(-)

diff --git a/tools/perf/util/mem-events.c b/tools/perf/util/mem-events.c
index f291f93ac11d..1d57defe6ea7 100644
--- a/tools/perf/util/mem-events.c
+++ b/tools/perf/util/mem-events.c
@@ -217,3 +217,37 @@ void perf_mem__lvl_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
 	if (miss)
 		strncat(out, " miss", sz - l);
 }
+
+static const char * const snoop_access[] = {
+	"N/A",
+	"None",
+	"Miss",
+	"Hit",
+	"HitM",
+};
+
+void perf_mem__snp_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
+{
+	size_t i, l = 0;
+	u64 m = PERF_MEM_SNOOP_NA;
+
+	sz -= 1; /* -1 for null termination */
+	out[0] = '\0';
+
+	if (mem_info)
+		m = mem_info->data_src.mem_snoop;
+
+	for (i = 0; m && i < ARRAY_SIZE(snoop_access); i++, m >>= 1) {
+		if (!(m & 0x1))
+			continue;
+		if (l) {
+			strcat(out, " or ");
+			l += 4;
+		}
+		strncat(out, snoop_access[i], sz - l);
+		l += strlen(snoop_access[i]);
+	}
+
+	if (*out == '\0')
+		strcpy(out, "N/A");
+}
diff --git a/tools/perf/util/mem-events.h b/tools/perf/util/mem-events.h
index fc6d94fc96e8..7b9e172cfa72 100644
--- a/tools/perf/util/mem-events.h
+++ b/tools/perf/util/mem-events.h
@@ -47,5 +47,6 @@ char *perf_mem_events__name(int i);
 struct mem_info;
 void perf_mem__tlb_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
 void perf_mem__lvl_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
+void perf_mem__snp_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
 #endif /* __PERF_MEM_EVENTS_H */
 
diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
index c7b52f4987d8..de936ce324b7 100644
--- a/tools/perf/util/sort.c
+++ b/tools/perf/util/sort.c
@@ -891,41 +891,12 @@ sort__snoop_cmp(struct hist_entry *left, struct hist_entry *right)
 	return (int64_t)(data_src_r.mem_snoop - data_src_l.mem_snoop);
 }
 
-static const char * const snoop_access[] = {
-	"N/A",
-	"None",
-	"Miss",
-	"Hit",
-	"HitM",
-};
-
 static int hist_entry__snoop_snprintf(struct hist_entry *he, char *bf,
 				    size_t size, unsigned int width)
 {
 	char out[64];
-	size_t sz = sizeof(out) - 1; /* -1 for null termination */
-	size_t i, l = 0;
-	u64 m = PERF_MEM_SNOOP_NA;
-
-	out[0] = '\0';
-
-	if (he->mem_info)
-		m = he->mem_info->data_src.mem_snoop;
-
-	for (i = 0; m && i < ARRAY_SIZE(snoop_access); i++, m >>= 1) {
-		if (!(m & 0x1))
-			continue;
-		if (l) {
-			strcat(out, " or ");
-			l += 4;
-		}
-		strncat(out, snoop_access[i], sz - l);
-		l += strlen(snoop_access[i]);
-	}
-
-	if (*out == '\0')
-		strcpy(out, "N/A");
 
+	perf_mem__snp_scnprintf(out, sizeof(out), he->mem_info);
 	return repsep_snprintf(bf, size, "%-*s", width, out);
 }
 
-- 
2.4.3

[toc] | [prev] | [next] | [standalone]

#1334261 — [PATCH 01/23] perf record: Add --all-user/--all-kernel options

FromJiri Olsa <jolsa@kernel.org>
Date2016-02-15 09:40 +0100
Subject[PATCH 01/23] perf record: Add --all-user/--all-kernel options
Message-ID<r2mRm-6Rc-49@gated-at.bofh.it>
In reply to#1334248
Allow user to easily switch all events to user or
kernel space with simple --all-user or --all-kernel
options.

This will be handy within perf mem/c2c wrappers to
switch easily monitoring modes.

Link: http://lkml.kernel.org/n/tip-hdd2u0y1o8cwfpplrpzc1iqd@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/Documentation/perf-record.txt |  6 ++++++
 tools/perf/builtin-record.c              |  4 ++++
 tools/perf/perf.h                        |  2 ++
 tools/perf/util/evsel.c                  | 10 ++++++++++
 4 files changed, 22 insertions(+)

diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt
index fbceb631387c..19aa17532a16 100644
--- a/tools/perf/Documentation/perf-record.txt
+++ b/tools/perf/Documentation/perf-record.txt
@@ -341,6 +341,12 @@ Specify vmlinux path which has debuginfo.
 --buildid-all::
 Record build-id of all DSOs regardless whether it's actually hit or not.
 
+--all-kernel::
+Configure all used events to run in kernel space.
+
+--all-user::
+Configure all used events to run in user space.
+
 SEE ALSO
 --------
 linkperf:perf-stat[1], linkperf:perf-list[1]
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index caa8235ed027..b64f31a4c993 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -1151,6 +1151,10 @@ struct option __record_options[] = {
 			"per thread proc mmap processing timeout in ms"),
 	OPT_BOOLEAN(0, "switch-events", &record.opts.record_switch_events,
 		    "Record context switch events"),
+	OPT_BOOLEAN(0, "all-kernel", &record.opts.all_kernel,
+		    "Configure all used events to run in kernel space."),
+	OPT_BOOLEAN(0, "all-user", &record.opts.all_user,
+		    "Configure all used events to run in user space."),
 	OPT_STRING(0, "clang-path", &llvm_param.clang_path, "clang path",
 		   "clang binary to use for compiling BPF scriptlets"),
 	OPT_STRING(0, "clang-opt", &llvm_param.clang_opt, "clang options",
diff --git a/tools/perf/perf.h b/tools/perf/perf.h
index 90129accffbe..5381a01c0610 100644
--- a/tools/perf/perf.h
+++ b/tools/perf/perf.h
@@ -58,6 +58,8 @@ struct record_opts {
 	bool	     full_auxtrace;
 	bool	     auxtrace_snapshot_mode;
 	bool	     record_switch_events;
+	bool	     all_kernel;
+	bool	     all_user;
 	unsigned int freq;
 	unsigned int mmap_pages;
 	unsigned int auxtrace_mmap_pages;
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index 467808680ee4..6ae20d0056de 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -898,6 +898,16 @@ void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts)
 	if (evsel->precise_max)
 		perf_event_attr__set_max_precise_ip(attr);
 
+	if (opts->all_user) {
+		attr->exclude_kernel = 1;
+		attr->exclude_user   = 0;
+	}
+
+	if (opts->all_kernel) {
+		attr->exclude_kernel = 0;
+		attr->exclude_user   = 1;
+	}
+
 	/*
 	 * Apply event specific term settings,
 	 * it overloads any global configuration.
-- 
2.4.3

[toc] | [prev] | [next] | [standalone]

#1335633 — Re: [PATCH 01/23] perf record: Add --all-user/--all-kernel options

FromArnaldo Carvalho de Melo <acme@kernel.org>
Date2016-02-16 18:00 +0100
SubjectRe: [PATCH 01/23] perf record: Add --all-user/--all-kernel options
Message-ID<r2R8K-21y-3@gated-at.bofh.it>
In reply to#1334261
Em Mon, Feb 15, 2016 at 09:34:31AM +0100, Jiri Olsa escreveu:
> Allow user to easily switch all events to user or
> kernel space with simple --all-user or --all-kernel
> options.
> 
> This will be handy within perf mem/c2c wrappers to
> switch easily monitoring modes.

Humm, some oddities:

  # perf record --all-user -e cycles -a
  # perf report --tui       # And then notice there are some kernel samples, zoom into kernel DSO
  Samples: 7  of event 'cycles', Event count (approx.): 3158810, DSO: [kernel.vmlinux]
  Overhead  Comman  Symbol
    84.05%  chrome  [k] page_fault
    15.73%  chrome  [k] entry_SYSCALL_64
     0.22%  chrome  [k] apic_timer_interrupt

Ditto when using --all-kernel, some userspace samples are there, also perhaps we
should show "cycles:u" or "cycles:k" when this --all-user or --all-kernel features
are used.

Need to investigate why there are kernel samples when --all-user is used and
the other way around as well.

Also if I use both it quietly accepts and shows just one of them, I guess we
should bail out in case someone tries both.

- Arnaldo 

> Link: http://lkml.kernel.org/n/tip-hdd2u0y1o8cwfpplrpzc1iqd@git.kernel.org
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  tools/perf/Documentation/perf-record.txt |  6 ++++++
>  tools/perf/builtin-record.c              |  4 ++++
>  tools/perf/perf.h                        |  2 ++
>  tools/perf/util/evsel.c                  | 10 ++++++++++
>  4 files changed, 22 insertions(+)
> 
> diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt
> index fbceb631387c..19aa17532a16 100644
> --- a/tools/perf/Documentation/perf-record.txt
> +++ b/tools/perf/Documentation/perf-record.txt
> @@ -341,6 +341,12 @@ Specify vmlinux path which has debuginfo.
>  --buildid-all::
>  Record build-id of all DSOs regardless whether it's actually hit or not.
>  
> +--all-kernel::
> +Configure all used events to run in kernel space.
> +
> +--all-user::
> +Configure all used events to run in user space.
> +
>  SEE ALSO
>  --------
>  linkperf:perf-stat[1], linkperf:perf-list[1]
> diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
> index caa8235ed027..b64f31a4c993 100644
> --- a/tools/perf/builtin-record.c
> +++ b/tools/perf/builtin-record.c
> @@ -1151,6 +1151,10 @@ struct option __record_options[] = {
>  			"per thread proc mmap processing timeout in ms"),
>  	OPT_BOOLEAN(0, "switch-events", &record.opts.record_switch_events,
>  		    "Record context switch events"),
> +	OPT_BOOLEAN(0, "all-kernel", &record.opts.all_kernel,
> +		    "Configure all used events to run in kernel space."),
> +	OPT_BOOLEAN(0, "all-user", &record.opts.all_user,
> +		    "Configure all used events to run in user space."),
>  	OPT_STRING(0, "clang-path", &llvm_param.clang_path, "clang path",
>  		   "clang binary to use for compiling BPF scriptlets"),
>  	OPT_STRING(0, "clang-opt", &llvm_param.clang_opt, "clang options",
> diff --git a/tools/perf/perf.h b/tools/perf/perf.h
> index 90129accffbe..5381a01c0610 100644
> --- a/tools/perf/perf.h
> +++ b/tools/perf/perf.h
> @@ -58,6 +58,8 @@ struct record_opts {
>  	bool	     full_auxtrace;
>  	bool	     auxtrace_snapshot_mode;
>  	bool	     record_switch_events;
> +	bool	     all_kernel;
> +	bool	     all_user;
>  	unsigned int freq;
>  	unsigned int mmap_pages;
>  	unsigned int auxtrace_mmap_pages;
> diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
> index 467808680ee4..6ae20d0056de 100644
> --- a/tools/perf/util/evsel.c
> +++ b/tools/perf/util/evsel.c
> @@ -898,6 +898,16 @@ void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts)
>  	if (evsel->precise_max)
>  		perf_event_attr__set_max_precise_ip(attr);
>  
> +	if (opts->all_user) {
> +		attr->exclude_kernel = 1;
> +		attr->exclude_user   = 0;
> +	}
> +
> +	if (opts->all_kernel) {
> +		attr->exclude_kernel = 0;
> +		attr->exclude_user   = 1;
> +	}
> +
>  	/*
>  	 * Apply event specific term settings,
>  	 * it overloads any global configuration.
> -- 
> 2.4.3

[toc] | [prev] | [next] | [standalone]

#1336029 — Re: [PATCH 01/23] perf record: Add --all-user/--all-kernel options

FromAndi Kleen <ak@linux.intel.com>
Date2016-02-17 04:20 +0100
SubjectRe: [PATCH 01/23] perf record: Add --all-user/--all-kernel options
Message-ID<r30OJ-mV-1@gated-at.bofh.it>
In reply to#1335633
> Need to investigate why there are kernel samples when --all-user is used and
> the other way around as well.

This is a known issue due to skid. PEBS usually avoids it.

-Andi

[toc] | [prev] | [next] | [standalone]

#1336472 — Re: [PATCH 01/23] perf record: Add --all-user/--all-kernel options

FromArnaldo Carvalho de Melo <acme@kernel.org>
Date2016-02-17 15:30 +0100
SubjectRe: [PATCH 01/23] perf record: Add --all-user/--all-kernel options
Message-ID<r3bh7-7Ch-1@gated-at.bofh.it>
In reply to#1336029
Em Tue, Feb 16, 2016 at 07:10:37PM -0800, Andi Kleen escreveu:
> > Need to investigate why there are kernel samples when --all-user is used and
> > the other way around as well.

> This is a known issue due to skid. PEBS usually avoids it.

Right, thanks for pointing this out, we have to detect that :p is not
being used and warn the user about this possibility, i.e. user asks for
just kernel samples, tool finds user samples, tell that using :p+ may
help.

Now, using plain:

 # perf record --all-user -a sleep 2

will make it use the default which is 'cycles' with the highest
precision available on the machine, and that leaves just a few samples
that supposedly have the MISC_KERNEL bit set in the perf_event_attr
header, those are not mapping to any mmap area known to the tool, I'm
investigating what this is.

I'll apply the patch as-is, we'll check later OPT_ magic to tell that
those options are mutually exclusive.

- Arnaldo

[toc] | [prev] | [next] | [standalone]

#1336523 — Re: [PATCH 01/23] perf record: Add --all-user/--all-kernel options

FromAndi Kleen <ak@linux.intel.com>
Date2016-02-17 16:40 +0100
SubjectRe: [PATCH 01/23] perf record: Add --all-user/--all-kernel options
Message-ID<r3cmS-8kt-17@gated-at.bofh.it>
In reply to#1336472
On Wed, Feb 17, 2016 at 11:26:16AM -0300, Arnaldo Carvalho de Melo wrote:
> Em Tue, Feb 16, 2016 at 07:10:37PM -0800, Andi Kleen escreveu:
> > > Need to investigate why there are kernel samples when --all-user is used and
> > > the other way around as well.
> 
> > This is a known issue due to skid. PEBS usually avoids it.
> 
> Right, thanks for pointing this out, we have to detect that :p is not
> being used and warn the user about this possibility, i.e. user asks for
> just kernel samples, tool finds user samples, tell that using :p+ may
> help.

If you're really so worried about it the kernel driver could filter it.

-Andi

[toc] | [prev] | [next] | [standalone]

#1336486 — Re: [PATCH 01/23] perf record: Add --all-user/--all-kernel options

FromJiri Olsa <jolsa@redhat.com>
Date2016-02-17 15:40 +0100
SubjectRe: [PATCH 01/23] perf record: Add --all-user/--all-kernel options
Message-ID<r3bqO-7FR-15@gated-at.bofh.it>
In reply to#1335633
On Tue, Feb 16, 2016 at 01:56:01PM -0300, Arnaldo Carvalho de Melo wrote:
> Em Mon, Feb 15, 2016 at 09:34:31AM +0100, Jiri Olsa escreveu:
> > Allow user to easily switch all events to user or
> > kernel space with simple --all-user or --all-kernel
> > options.
> > 
> > This will be handy within perf mem/c2c wrappers to
> > switch easily monitoring modes.
> 
> Humm, some oddities:
> 
>   # perf record --all-user -e cycles -a
>   # perf report --tui       # And then notice there are some kernel samples, zoom into kernel DSO
>   Samples: 7  of event 'cycles', Event count (approx.): 3158810, DSO: [kernel.vmlinux]
>   Overhead  Comman  Symbol
>     84.05%  chrome  [k] page_fault
>     15.73%  chrome  [k] entry_SYSCALL_64
>      0.22%  chrome  [k] apic_timer_interrupt
> 
> Ditto when using --all-kernel, some userspace samples are there, also perhaps we
> should show "cycles:u" or "cycles:k" when this --all-user or --all-kernel features
> are used.
> 
> Need to investigate why there are kernel samples when --all-user is used and
> the other way around as well.
> 
> Also if I use both it quietly accepts and shows just one of them, I guess we
> should bail out in case someone tries both.

right, that should be handled with warning and exit
I'll do it in next version

thanks,
jirka

[toc] | [prev] | [next] | [standalone]

#1336497 — Re: [PATCH 01/23] perf record: Add --all-user/--all-kernel options

FromArnaldo Carvalho de Melo <acme@kernel.org>
Date2016-02-17 16:00 +0100
SubjectRe: [PATCH 01/23] perf record: Add --all-user/--all-kernel options
Message-ID<r3bKa-7NK-15@gated-at.bofh.it>
In reply to#1336486
Em Wed, Feb 17, 2016 at 03:39:38PM +0100, Jiri Olsa escreveu:
> On Tue, Feb 16, 2016 at 01:56:01PM -0300, Arnaldo Carvalho de Melo wrote:
> > Em Mon, Feb 15, 2016 at 09:34:31AM +0100, Jiri Olsa escreveu:
> > > Allow user to easily switch all events to user or
> > > kernel space with simple --all-user or --all-kernel
> > > options.
> > > 
> > > This will be handy within perf mem/c2c wrappers to
> > > switch easily monitoring modes.
> > 
> > Humm, some oddities:
> > 
> >   # perf record --all-user -e cycles -a
> >   # perf report --tui       # And then notice there are some kernel samples, zoom into kernel DSO
> >   Samples: 7  of event 'cycles', Event count (approx.): 3158810, DSO: [kernel.vmlinux]
> >   Overhead  Comman  Symbol
> >     84.05%  chrome  [k] page_fault
> >     15.73%  chrome  [k] entry_SYSCALL_64
> >      0.22%  chrome  [k] apic_timer_interrupt
> > 
> > Ditto when using --all-kernel, some userspace samples are there, also perhaps we
> > should show "cycles:u" or "cycles:k" when this --all-user or --all-kernel features
> > are used.
> > 
> > Need to investigate why there are kernel samples when --all-user is used and
> > the other way around as well.
> > 
> > Also if I use both it quietly accepts and shows just one of them, I guess we
> > should bail out in case someone tries both.
> 
> right, that should be handled with warning and exit
> I'll do it in next version

Ok, so I'll wait for it then,

builtin-probe.c uses things like:

        set_option_flag(options, 'a', "add", PARSE_OPT_EXCLUSIVE); 
        set_option_flag(options, 'd', "del", PARSE_OPT_EXCLUSIVE);

We have:

#define OPT_BOOLEAN_FLAG(s, l, v, h, f)     { .type = OPTION_BOOLEAN,
.short_name = (s), .long_name = (l), .value = check_vtype(v, bool *),
.help = (h), .flags = (f) }


That allows setting that PARSE_OPT_EXCLUSIVE flag, but we don't have
_FLAG variants for all opt types, for instance there is no
OPT_CALLBACK_FLAG() that could be used in builtin-probe.c to avoid those
unconditional set_option_flag(EXCLUSIVE) :-\

- Arnaldo

[toc] | [prev] | [next] | [standalone]

#1338690 — [tip:perf/core] perf record: Add --all-user/--all-kernel options

Fromtip-bot for Jiri Olsa <tipbot@zytor.com>
Date2016-02-20 12:40 +0100
Subject[tip:perf/core] perf record: Add --all-user/--all-kernel options
Message-ID<r4e3g-4EY-15@gated-at.bofh.it>
In reply to#1334261
Commit-ID:  85723885feb823b4fc352b727ece0b6d00306c4d
Gitweb:     http://git.kernel.org/tip/85723885feb823b4fc352b727ece0b6d00306c4d
Author:     Jiri Olsa <jolsa@kernel.org>
AuthorDate: Mon, 15 Feb 2016 09:34:31 +0100
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Thu, 18 Feb 2016 10:48:44 -0300

perf record: Add --all-user/--all-kernel options

Allow user to easily switch all events to user or kernel space with simple
--all-user or --all-kernel options.

This will be handy within perf mem/c2c wrappers to switch easily monitoring
modes.

Committer note:

Testing it:

  # perf record --all-kernel --all-user -a sleep 2
   Error: option `all-user' cannot be used with all-kernel
   Usage: perf record [<options>] [<command>]
      or: perf record [<options>] -- <command> [<options>]

        --all-user        Configure all used events to run in user space.
        --all-kernel      Configure all used events to run in kernel space.
  # perf record --all-user --all-kernel -a sleep 2
   Error: option `all-kernel' cannot be used with all-user
   Usage: perf record [<options>] [<command>]
      or: perf record [<options>] -- <command> [<options>]

        --all-kernel      Configure all used events to run in kernel space.
        --all-user        Configure all used events to run in user space.
  # perf record --all-user -a sleep 1
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 1.416 MB perf.data (162 samples) ]
  # perf report | grep '\[k\]'
  # perf record --all-kernel -a sleep 1
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 1.423 MB perf.data (296 samples) ]
  # perf report | grep '\[\.\]'
  #

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1455525293-8671-2-git-send-email-jolsa@kernel.org
[ Made those options to be mutually exclusive ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/Documentation/perf-record.txt |  6 ++++++
 tools/perf/builtin-record.c              |  6 ++++++
 tools/perf/perf.h                        |  2 ++
 tools/perf/util/evsel.c                  | 10 ++++++++++
 4 files changed, 24 insertions(+)

diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt
index fbceb63..19aa175 100644
--- a/tools/perf/Documentation/perf-record.txt
+++ b/tools/perf/Documentation/perf-record.txt
@@ -341,6 +341,12 @@ Specify vmlinux path which has debuginfo.
 --buildid-all::
 Record build-id of all DSOs regardless whether it's actually hit or not.
 
+--all-kernel::
+Configure all used events to run in kernel space.
+
+--all-user::
+Configure all used events to run in user space.
+
 SEE ALSO
 --------
 linkperf:perf-stat[1], linkperf:perf-list[1]
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 0ee0d5c..cf3a28d 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -1140,6 +1140,12 @@ struct option __record_options[] = {
 			"per thread proc mmap processing timeout in ms"),
 	OPT_BOOLEAN(0, "switch-events", &record.opts.record_switch_events,
 		    "Record context switch events"),
+	OPT_BOOLEAN_FLAG(0, "all-kernel", &record.opts.all_kernel,
+			 "Configure all used events to run in kernel space.",
+			 PARSE_OPT_EXCLUSIVE),
+	OPT_BOOLEAN_FLAG(0, "all-user", &record.opts.all_user,
+			 "Configure all used events to run in user space.",
+			 PARSE_OPT_EXCLUSIVE),
 	OPT_STRING(0, "clang-path", &llvm_param.clang_path, "clang path",
 		   "clang binary to use for compiling BPF scriptlets"),
 	OPT_STRING(0, "clang-opt", &llvm_param.clang_opt, "clang options",
diff --git a/tools/perf/perf.h b/tools/perf/perf.h
index 90129ac..5381a01 100644
--- a/tools/perf/perf.h
+++ b/tools/perf/perf.h
@@ -58,6 +58,8 @@ struct record_opts {
 	bool	     full_auxtrace;
 	bool	     auxtrace_snapshot_mode;
 	bool	     record_switch_events;
+	bool	     all_kernel;
+	bool	     all_user;
 	unsigned int freq;
 	unsigned int mmap_pages;
 	unsigned int auxtrace_mmap_pages;
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index 4678086..6ae20d0 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -898,6 +898,16 @@ void perf_evsel__config(struct perf_evsel *evsel, struct record_opts *opts)
 	if (evsel->precise_max)
 		perf_event_attr__set_max_precise_ip(attr);
 
+	if (opts->all_user) {
+		attr->exclude_kernel = 1;
+		attr->exclude_user   = 0;
+	}
+
+	if (opts->all_kernel) {
+		attr->exclude_kernel = 0;
+		attr->exclude_user   = 1;
+	}
+
 	/*
 	 * Apply event specific term settings,
 	 * it overloads any global configuration.

[toc] | [prev] | [next] | [standalone]

#1334264 — [PATCH 17/23] perf tools: Change perf_mem__tlb_scnprintf to return nb of displayed bytes

FromJiri Olsa <jolsa@kernel.org>
Date2016-02-15 09:40 +0100
Subject[PATCH 17/23] perf tools: Change perf_mem__tlb_scnprintf to return nb of displayed bytes
Message-ID<r2mRm-6Rc-53@gated-at.bofh.it>
In reply to#1334248
Moving strncat/strcpy calls into scnprintf to easily track
number of displayed bytes. It will be used in following patch.

Link: http://lkml.kernel.org/n/tip-9q9h9trua9zljnw26c56memh@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/util/mem-events.c | 13 +++++++------
 tools/perf/util/mem-events.h |  2 +-
 2 files changed, 8 insertions(+), 7 deletions(-)

diff --git a/tools/perf/util/mem-events.c b/tools/perf/util/mem-events.c
index 915fae6ad3ee..440221ca4f8a 100644
--- a/tools/perf/util/mem-events.c
+++ b/tools/perf/util/mem-events.c
@@ -129,7 +129,7 @@ static const char * const tlb_access[] = {
 	"Fault",
 };
 
-void perf_mem__tlb_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
+int perf_mem__tlb_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
 {
 	size_t l = 0, i;
 	u64 m = PERF_MEM_TLB_NA;
@@ -154,15 +154,16 @@ void perf_mem__tlb_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
 			strcat(out, " or ");
 			l += 4;
 		}
-		strncat(out, tlb_access[i], sz - l);
-		l += strlen(tlb_access[i]);
+		l += scnprintf(out + l, sz - l, tlb_access[i]);
 	}
 	if (*out == '\0')
-		strcpy(out, "N/A");
+		l += scnprintf(out, sz - l, "N/A");
 	if (hit)
-		strncat(out, " hit", sz - l);
+		l += scnprintf(out + l, sz - l, " hit");
 	if (miss)
-		strncat(out, " miss", sz - l);
+		l += scnprintf(out + l, sz - l, " miss");
+
+	return l;
 }
 
 static const char * const mem_lvl[] = {
diff --git a/tools/perf/util/mem-events.h b/tools/perf/util/mem-events.h
index 0dd2efb37c44..ce9ed0540014 100644
--- a/tools/perf/util/mem-events.h
+++ b/tools/perf/util/mem-events.h
@@ -45,7 +45,7 @@ int perf_mem_events__init(void);
 char *perf_mem_events__name(int i);
 
 struct mem_info;
-void perf_mem__tlb_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
+int perf_mem__tlb_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
 void perf_mem__lvl_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
 void perf_mem__snp_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
 void perf_mem__lck_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
-- 
2.4.3

[toc] | [prev] | [next] | [standalone]

#1334266 — [PATCH 23/23] perf script: Display data_src values

FromJiri Olsa <jolsa@kernel.org>
Date2016-02-15 09:40 +0100
Subject[PATCH 23/23] perf script: Display data_src values
Message-ID<r2mRm-6Rc-59@gated-at.bofh.it>
In reply to#1334248
Adding support to display data_src values,
for events with data_src data in sample.

Example:
  $ perf script
  ...
           rcuos/3    32 [002] ... 68501042 Local RAM hit|SNP None or Hit|TLB L1 or L2 hit|LCK No   ...
           rcuos/3    32 [002] ... 68100142 L1 hit|SNP None|TLB L1 or L2 hit|LCK No                 ...
           swapper     0 [002] ... 68100242 LFB hit|SNP None|TLB L1 or L2 hit|LCK No                ...
           swapper     0 [000] ... 68100142 L1 hit|SNP None|TLB L1 or L2 hit|LCK No                 ...
           swapper     0 [000] ... 50100142 L1 hit|SNP None|TLB L2 miss|LCK No                      ...
           rcuos/3    32 [002] ... 68100142 L1 hit|SNP None|TLB L1 or L2 hit|LCK No                 ...
   plugin-containe 16538 [000] ... 6a100142 L1 hit|SNP None|TLB L1 or L2 hit|LCK Yes                ...
           gkrellm  1736 [000] ... 68100242 LFB hit|SNP None|TLB L1 or L2 hit|LCK No                ...
           gkrellm  1736 [000] ... 6a100142 L1 hit|SNP None|TLB L1 or L2 hit|LCK Yes                ...

                                   ^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                             data_src value                     data_src translation

Link: http://lkml.kernel.org/n/tip-48x7lm7d765bzo3oghap376m@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/builtin-script.c  | 20 +++++++++++++++++++-
 tools/perf/util/mem-events.c | 15 +++++++++++++++
 tools/perf/util/mem-events.h |  2 ++
 3 files changed, 36 insertions(+), 1 deletion(-)

diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
index f4caf4898245..8ff5ff0fe38c 100644
--- a/tools/perf/builtin-script.c
+++ b/tools/perf/builtin-script.c
@@ -23,6 +23,7 @@
 #include "util/stat.h"
 #include <linux/bitmap.h>
 #include "asm/bug.h"
+#include "util/mem-events.h"
 
 static char const		*script_name;
 static char const		*generate_script_lang;
@@ -649,6 +650,23 @@ static int perf_evlist__max_name_len(struct perf_evlist *evlist)
 	return max;
 }
 
+static size_t data_src__printf(u64 data_src)
+{
+	struct mem_info mi = { .data_src.val = data_src };
+	char decode[100];
+	char out[100];
+	static int maxlen;
+	int len;
+
+	perf_script__meminfo_scnprintf(decode, 100, &mi);
+
+	len = scnprintf(out, 100, "%16" PRIx64 " %s", data_src, decode);
+	if (maxlen < len)
+		maxlen = len;
+
+	return printf("%-*s", maxlen, out);
+}
+
 static void process_event(struct perf_script *script, union perf_event *event,
 			  struct perf_sample *sample, struct perf_evsel *evsel,
 			  struct addr_location *al)
@@ -689,7 +707,7 @@ static void process_event(struct perf_script *script, union perf_event *event,
 		print_sample_addr(event, sample, thread, attr);
 
 	if (PRINT_FIELD(DATA_SRC))
-		printf("%16" PRIx64, sample->data_src);
+		data_src__printf(sample->data_src);
 
 	if (PRINT_FIELD(WEIGHT))
 		printf("%16" PRIu64, sample->weight);
diff --git a/tools/perf/util/mem-events.c b/tools/perf/util/mem-events.c
index 0da707da127c..31e079a1ab7b 100644
--- a/tools/perf/util/mem-events.c
+++ b/tools/perf/util/mem-events.c
@@ -272,3 +272,18 @@ int perf_mem__lck_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
 
 	return l;
 }
+
+int perf_script__meminfo_scnprintf(char *out, size_t sz, struct mem_info *mem_info)
+{
+	int i = 0;
+
+	i += perf_mem__lvl_scnprintf(out, sz, mem_info);
+	i += scnprintf(out + i, sz - i, "|SNP ");
+	i += perf_mem__snp_scnprintf(out + i, sz - i, mem_info);
+	i += scnprintf(out + i, sz - i, "|TLB ");
+	i += perf_mem__tlb_scnprintf(out + i, sz - i, mem_info);
+	i += scnprintf(out + i, sz - i, "|LCK ");
+	i += perf_mem__lck_scnprintf(out + i, sz - i, mem_info);
+
+	return i;
+}
diff --git a/tools/perf/util/mem-events.h b/tools/perf/util/mem-events.h
index 759d35da2d6a..0f63d6a1c53a 100644
--- a/tools/perf/util/mem-events.h
+++ b/tools/perf/util/mem-events.h
@@ -50,5 +50,7 @@ int perf_mem__lvl_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
 int perf_mem__snp_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
 int perf_mem__lck_scnprintf(char *out, size_t sz, struct mem_info *mem_info);
 
+int perf_script__meminfo_scnprintf(char *bf, size_t size, struct mem_info *mem_info);
+
 #endif /* __PERF_MEM_EVENTS_H */
 
-- 
2.4.3

[toc] | [prev] | [next] | [standalone]

#1334267 — [PATCH 05/23] perf mem: Add -e record option

FromJiri Olsa <jolsa@kernel.org>
Date2016-02-15 09:40 +0100
Subject[PATCH 05/23] perf mem: Add -e record option
Message-ID<r2mRn-6Rc-61@gated-at.bofh.it>
In reply to#1334248
Adding -e option for perf mem record command,
to be able to specify memory event directly.

Get list of available events:

  $ perf mem record -e list
  ldlat-loads
  ldlat-stores

Monitor ldlat-loads:
  $ perf mem record -e ldlat-loads true

Link: http://lkml.kernel.org/n/tip-x5famenpxjfmhojzo95sf26s@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 tools/perf/builtin-mem.c     | 63 +++++++++++++++++++++++++++++++++++++++++---
 tools/perf/util/mem-events.c | 47 ++++++++++++++++++++++++++++++---
 tools/perf/util/mem-events.h |  3 +++
 3 files changed, 106 insertions(+), 7 deletions(-)

diff --git a/tools/perf/builtin-mem.c b/tools/perf/builtin-mem.c
index 36c56a4b778b..b3f8a89ede40 100644
--- a/tools/perf/builtin-mem.c
+++ b/tools/perf/builtin-mem.c
@@ -7,6 +7,7 @@
 #include "util/session.h"
 #include "util/data.h"
 #include "util/mem-events.h"
+#include "util/debug.h"
 
 #define MEM_OPERATION_LOAD	0x1
 #define MEM_OPERATION_STORE	0x2
@@ -22,11 +23,55 @@ struct perf_mem {
 	DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
 };
 
+static int parse_record_events(const struct option *opt,
+			       const char *str, int unset __maybe_unused)
+{
+	struct perf_mem *mem = *(struct perf_mem **)opt->value;
+	int j;
+
+	if (strcmp(str, "list")) {
+		if (!perf_mem_events__parse(str)) {
+			mem->operation = 0;
+			return 0;
+		}
+		exit(-1);
+	}
+
+	for (j = 0; j < PERF_MEM_EVENTS__MAX; j++) {
+		struct perf_mem_event *e = &perf_mem_events[j];
+
+		fprintf(stderr, "%-20s%s",
+			e->tag, verbose ? "" : "\n");
+		if (verbose)
+			fprintf(stderr, " [%s]\n", e->name);
+	}
+	exit(0);
+}
+
+static const char * const __usage[] = {
+	"perf mem record [<options>] [<command>]",
+	"perf mem record [<options>] -- <command> [<options>]",
+	NULL
+};
+
+static const char * const *record_mem_usage = __usage;
+
 static int __cmd_record(int argc, const char **argv, struct perf_mem *mem)
 {
 	int rec_argc, i = 0, j;
 	const char **rec_argv;
 	int ret;
+	struct option options[] = {
+	OPT_CALLBACK('e', "event", &mem, "event",
+		     "event selector. use 'perf mem record -e list' to list available events",
+		     parse_record_events),
+	OPT_INCR('v', "verbose", &verbose,
+		 "be more verbose (show counter open errors, etc)"),
+	OPT_END()
+	};
+
+	argc = parse_options(argc, argv, options, record_mem_usage,
+			     PARSE_OPT_STOP_AT_NON_OPTION);
 
 	rec_argc = argc + 7; /* max number of arguments */
 	rec_argv = calloc(rec_argc + 1, sizeof(char *));
@@ -35,10 +80,11 @@ static int __cmd_record(int argc, const char **argv, struct perf_mem *mem)
 
 	rec_argv[i++] = "record";
 
-	if (mem->operation & MEM_OPERATION_LOAD) {
+	if (mem->operation & MEM_OPERATION_LOAD)
 		perf_mem_events[PERF_MEM_EVENTS__LOAD].record = true;
+
+	if (perf_mem_events[PERF_MEM_EVENTS__LOAD].record)
 		rec_argv[i++] = "-W";
-	}
 
 	rec_argv[i++] = "-d";
 
@@ -50,9 +96,19 @@ static int __cmd_record(int argc, const char **argv, struct perf_mem *mem)
 		rec_argv[i++] = perf_mem_events[j].name;
 	};
 
-	for (j = 1; j < argc; j++, i++)
+	for (j = 0; j < argc; j++, i++)
 		rec_argv[i] = argv[j];
 
+	if (verbose > 0) {
+		pr_debug("calling: record ");
+
+		while (rec_argv[j]) {
+			pr_debug("%s ", rec_argv[j]);
+			j++;
+		}
+		pr_debug("\n");
+	}
+
 	ret = cmd_record(i, rec_argv, NULL);
 	free(rec_argv);
 	return ret;
@@ -299,7 +355,6 @@ int cmd_mem(int argc, const char **argv, const char *prefix __maybe_unused)
 		NULL
 	};
 
-
 	argc = parse_options_subcommand(argc, argv, mem_options, mem_subcommands,
 					mem_usage, PARSE_OPT_STOP_AT_NON_OPTION);
 
diff --git a/tools/perf/util/mem-events.c b/tools/perf/util/mem-events.c
index c6ba0a19ec9a..b1507c04b257 100644
--- a/tools/perf/util/mem-events.c
+++ b/tools/perf/util/mem-events.c
@@ -1,10 +1,51 @@
+#include <stddef.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
 #include "mem-events.h"
+#include "debug.h"
 
-#define E(n) { .name = n }
+#define E(t, n) { .tag = t, .name = n }
 
 struct perf_mem_event perf_mem_events[PERF_MEM_EVENTS__MAX] = {
-	E("cpu/mem-loads,ldlat=30/P"),
-	E("cpu/mem-stores/P"),
+	E("ldlat-loads",	"cpu/mem-loads,ldlat=30/P"),
+	E("ldlat-stores",	"cpu/mem-stores/P"),
 };
 
 #undef E
+
+int perf_mem_events__parse(const char *str)
+{
+	char *tok, *saveptr = NULL;
+	bool found = false;
+	char *buf;
+	int j;
+
+	/* We need buffer that we know we can write to. */
+	buf = malloc(strlen(str) + 1);
+	if (!buf)
+		return -ENOMEM;
+
+	strcpy(buf, str);
+
+	tok = strtok_r((char *)buf, ",", &saveptr);
+
+	while (tok) {
+		for (j = 0; j < PERF_MEM_EVENTS__MAX; j++) {
+			struct perf_mem_event *e = &perf_mem_events[j];
+
+			if (strstr(e->tag, tok))
+				e->record = found = true;
+		}
+
+		tok = strtok_r(NULL, ",", &saveptr);
+	}
+
+	free(buf);
+
+	if (found)
+		return 0;
+
+	pr_err("failed: event '%s' not found, use '-e list' to get list of available events\n", str);
+	return -1;
+}
diff --git a/tools/perf/util/mem-events.h b/tools/perf/util/mem-events.h
index ee5fde222389..0b5247468462 100644
--- a/tools/perf/util/mem-events.h
+++ b/tools/perf/util/mem-events.h
@@ -5,6 +5,7 @@
 
 struct perf_mem_event {
 	bool		record;
+	const char	*tag;
 	const char	*name;
 };
 
@@ -16,5 +17,7 @@ enum {
 
 extern struct perf_mem_event perf_mem_events[PERF_MEM_EVENTS__MAX];
 
+int perf_mem_events__parse(const char *str);
+
 #endif /* __PERF_MEM_EVENTS_H */
 
-- 
2.4.3

[toc] | [prev] | [next] | [standalone]

Page 1 of 2  [1] 2  Next page →

Back to top | Article view | linux.kernel

csiph-web