* [PATCH v2 01/11] perf tools script: Support insn output for normal samples
2019-02-26 3:04 Support sample context in perf report Andi Kleen
@ 2019-02-26 3:04 ` Andi Kleen
2019-02-26 3:04 ` [PATCH v2 02/11] perf tools report: Support nano seconds Andi Kleen
` (11 subsequent siblings)
12 siblings, 0 replies; 27+ messages in thread
From: Andi Kleen @ 2019-02-26 3:04 UTC (permalink / raw)
To: acme; +Cc: jolsa, linux-perf-users, linux-kernel, eranian, Andi Kleen
From: Andi Kleen <ak@linux.intel.com>
perf script -F +insn was only working for PT traces because
the PT instruction decoder was filling in the insn/insn_len
sample attributes. Support it for non PT samples too on x86
using the existing x86 instruction decoder.
% perf record -a sleep 1
% perf script -F ip,sym,insn --xed
ffffffff811704c9 remote_function movl %eax, 0x18(%rbx)
ffffffff8100bb50 intel_bts_enable_local retq
ffffffff81048612 native_apic_mem_write movl %esi, -0xa04000(%rdi)
ffffffff81048612 native_apic_mem_write movl %esi, -0xa04000(%rdi)
ffffffff81048612 native_apic_mem_write movl %esi, -0xa04000(%rdi)
ffffffff810f1f79 generic_exec_single xor %eax, %eax
ffffffff811704c9 remote_function movl %eax, 0x18(%rbx)
ffffffff8100bb34 intel_bts_enable_local movl 0x2000(%rax), %edx
ffffffff81048610 native_apic_mem_write mov %edi, %edi
...
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
v2:
Avoid printing instruction when empty
Only decode when perf.data file was collected on same architecture
---
tools/perf/arch/x86/util/Build | 1 +
tools/perf/arch/x86/util/archinsn.c | 28 ++++++++++++++++++++++++++++
tools/perf/builtin-script.c | 21 ++++++++++++++++++++-
tools/perf/util/archinsn.h | 12 ++++++++++++
4 files changed, 61 insertions(+), 1 deletion(-)
create mode 100644 tools/perf/arch/x86/util/archinsn.c
create mode 100644 tools/perf/util/archinsn.h
diff --git a/tools/perf/arch/x86/util/Build b/tools/perf/arch/x86/util/Build
index 7aab0be5fc5f..7b8e69bbbdfe 100644
--- a/tools/perf/arch/x86/util/Build
+++ b/tools/perf/arch/x86/util/Build
@@ -6,6 +6,7 @@ perf-y += perf_regs.o
perf-y += group.o
perf-y += machine.o
perf-y += event.o
+perf-y += archinsn.o
perf-$(CONFIG_DWARF) += dwarf-regs.o
perf-$(CONFIG_BPF_PROLOGUE) += dwarf-regs.o
diff --git a/tools/perf/arch/x86/util/archinsn.c b/tools/perf/arch/x86/util/archinsn.c
new file mode 100644
index 000000000000..10b3c2a08b8f
--- /dev/null
+++ b/tools/perf/arch/x86/util/archinsn.c
@@ -0,0 +1,28 @@
+// SPDX-License-Identifier: GPL-2.0
+#include "perf.h"
+#include "archinsn.h"
+#include "fetch.h"
+#include "util/intel-pt-decoder/insn.h"
+#include "machine.h"
+#include "thread.h"
+#include "symbol.h"
+
+void arch_fetch_insn(struct perf_sample *sample,
+ struct thread *thread,
+ struct machine *machine)
+{
+ struct insn insn;
+ int len;
+ bool is64bit = false;
+
+ if (!sample->ip)
+ return;
+ len = fetch_exe(sample->ip, thread, machine, sample->insn,
+ sizeof(sample->insn), &is64bit);
+ if (len <= 0)
+ return;
+ insn_init(&insn, sample->insn, len, is64bit);
+ insn_get_length(&insn);
+ if (insn_complete(&insn) && insn.length <= len)
+ sample->insn_len = insn.length;
+}
diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
index 29e95a25c6e6..1d651a9b110e 100644
--- a/tools/perf/builtin-script.c
+++ b/tools/perf/builtin-script.c
@@ -29,10 +29,12 @@
#include "util/time-utils.h"
#include "util/path.h"
#include "print_binary.h"
+#include "archinsn.h"
#include <linux/bitmap.h>
#include <linux/kernel.h>
#include <linux/stringify.h>
#include <linux/time64.h>
+#include <sys/utsname.h>
#include "asm/bug.h"
#include "util/mem-events.h"
#include "util/dump-insn.h"
@@ -63,6 +65,7 @@ static const char *cpu_list;
static DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
static struct perf_stat_config stat_config;
static int max_blocks;
+static bool native_arch;
unsigned int scripting_max_stack = PERF_MAX_STACK_DEPTH;
@@ -1227,6 +1230,12 @@ static int perf_sample__fprintf_callindent(struct perf_sample *sample,
return len + dlen;
}
+__weak void arch_fetch_insn(struct perf_sample *sample __maybe_unused,
+ struct thread *thread __maybe_unused,
+ struct machine *machine __maybe_unused)
+{
+}
+
static int perf_sample__fprintf_insn(struct perf_sample *sample,
struct perf_event_attr *attr,
struct thread *thread,
@@ -1234,9 +1243,12 @@ static int perf_sample__fprintf_insn(struct perf_sample *sample,
{
int printed = 0;
+ if (sample->insn_len == 0 && native_arch)
+ arch_fetch_insn(sample, thread, machine);
+
if (PRINT_FIELD(INSNLEN))
printed += fprintf(fp, " ilen: %d", sample->insn_len);
- if (PRINT_FIELD(INSN)) {
+ if (PRINT_FIELD(INSN) && sample->insn_len) {
int i;
printed += fprintf(fp, " insn:");
@@ -3271,6 +3283,7 @@ int cmd_script(int argc, const char **argv)
.set = false,
.default_no_sample = true,
};
+ struct utsname uts;
char *script_path = NULL;
const char **__argv;
int i, j, err = 0;
@@ -3609,6 +3622,12 @@ int cmd_script(int argc, const char **argv)
if (symbol__init(&session->header.env) < 0)
goto out_delete;
+ uname(&uts);
+ if (!strcmp(uts.machine, session->header.env.arch) ||
+ (!strcmp(uts.machine, "x86_64") &&
+ !strcmp(session->header.env.arch, "i386")))
+ native_arch = true;
+
script.session = session;
script__setup_sample_type(&script);
diff --git a/tools/perf/util/archinsn.h b/tools/perf/util/archinsn.h
new file mode 100644
index 000000000000..448cbb6b8d7e
--- /dev/null
+++ b/tools/perf/util/archinsn.h
@@ -0,0 +1,12 @@
+#ifndef INSN_H
+#define INSN_H 1
+
+struct perf_sample;
+struct machine;
+struct thread;
+
+void arch_fetch_insn(struct perf_sample *sample,
+ struct thread *thread,
+ struct machine *machine);
+
+#endif
--
2.17.2
^ permalink raw reply related [flat|nested] 27+ messages in thread* [PATCH v2 02/11] perf tools report: Support nano seconds
2019-02-26 3:04 Support sample context in perf report Andi Kleen
2019-02-26 3:04 ` [PATCH v2 01/11] perf tools script: Support insn output for normal samples Andi Kleen
@ 2019-02-26 3:04 ` Andi Kleen
2019-02-26 3:04 ` [PATCH v2 03/11] perf tools report: Parse time quantum Andi Kleen
` (10 subsequent siblings)
12 siblings, 0 replies; 27+ messages in thread
From: Andi Kleen @ 2019-02-26 3:04 UTC (permalink / raw)
To: acme; +Cc: jolsa, linux-perf-users, linux-kernel, eranian, Andi Kleen
From: Andi Kleen <ak@linux.intel.com>
Upcoming changes add timestamp output in perf report. Add a --ns
argument similar to perf script to support nanoseconds resolution
when needed.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
v2:
Move flag into symbol_conf and change all users
---
tools/perf/Documentation/perf-report.txt | 3 +++
tools/perf/builtin-report.c | 1 +
tools/perf/builtin-script.c | 11 +++++------
tools/perf/util/symbol.c | 1 +
tools/perf/util/symbol_conf.h | 1 +
5 files changed, 11 insertions(+), 6 deletions(-)
diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
index 1a27bfe05039..51dbc519dbce 100644
--- a/tools/perf/Documentation/perf-report.txt
+++ b/tools/perf/Documentation/perf-report.txt
@@ -477,6 +477,9 @@ include::itrace.txt[]
Please note that not all mmaps are stored, options affecting which ones
are include 'perf record --data', for instance.
+--ns::
+ Show time stamps in nanoseconds.
+
--stats::
Display overall events statistics without any further processing.
(like the one at the end of the perf report -D command)
diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index 2e8c74d6430c..6b2cc91ee26a 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -1147,6 +1147,7 @@ int cmd_report(int argc, const char **argv)
OPT_CALLBACK(0, "percent-type", &report.annotation_opts, "local-period",
"Set percent type local/global-period/hits",
annotate_parse_percent_type),
+ OPT_BOOLEAN(0, "ns", &symbol_conf.nanosecs, "Show times in nanosecs"),
OPT_END()
};
struct perf_data data = {
diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
index 1d651a9b110e..1e6f2f095152 100644
--- a/tools/perf/builtin-script.c
+++ b/tools/perf/builtin-script.c
@@ -60,7 +60,6 @@ static bool no_callchain;
static bool latency_format;
static bool system_wide;
static bool print_flags;
-static bool nanosecs;
static const char *cpu_list;
static DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
static struct perf_stat_config stat_config;
@@ -691,7 +690,7 @@ static int perf_sample__fprintf_start(struct perf_sample *sample,
secs = nsecs / NSEC_PER_SEC;
nsecs -= secs * NSEC_PER_SEC;
- if (nanosecs)
+ if (symbol_conf.nanosecs)
printed += fprintf(fp, "%5lu.%09llu: ", secs, nsecs);
else {
char sample_time[32];
@@ -3238,7 +3237,7 @@ static int parse_insn_trace(const struct option *opt __maybe_unused,
{
parse_output_fields(NULL, "+insn,-event,-period", 0);
itrace_parse_synth_opts(opt, "i0ns", 0);
- nanosecs = true;
+ symbol_conf.nanosecs = true;
return 0;
}
@@ -3256,7 +3255,7 @@ static int parse_call_trace(const struct option *opt __maybe_unused,
{
parse_output_fields(NULL, "-ip,-addr,-event,-period,+callindent", 0);
itrace_parse_synth_opts(opt, "cewp", 0);
- nanosecs = true;
+ symbol_conf.nanosecs = true;
return 0;
}
@@ -3266,7 +3265,7 @@ static int parse_callret_trace(const struct option *opt __maybe_unused,
{
parse_output_fields(NULL, "-ip,-addr,-event,-period,+callindent,+flags", 0);
itrace_parse_synth_opts(opt, "crewp", 0);
- nanosecs = true;
+ symbol_conf.nanosecs = true;
return 0;
}
@@ -3402,7 +3401,7 @@ int cmd_script(int argc, const char **argv)
OPT_BOOLEAN('f', "force", &symbol_conf.force, "don't complain, do it"),
OPT_INTEGER(0, "max-blocks", &max_blocks,
"Maximum number of code blocks to dump with brstackinsn"),
- OPT_BOOLEAN(0, "ns", &nanosecs,
+ OPT_BOOLEAN(0, "ns", &symbol_conf.nanosecs,
"Use 9 decimal places when displaying time"),
OPT_CALLBACK_OPTARG(0, "itrace", &itrace_synth_opts, NULL, "opts",
"Instruction Tracing options\n" ITRACE_HELP,
diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
index 758bf5f74e6e..eb873ea1c405 100644
--- a/tools/perf/util/symbol.c
+++ b/tools/perf/util/symbol.c
@@ -39,6 +39,7 @@ int vmlinux_path__nr_entries;
char **vmlinux_path;
struct symbol_conf symbol_conf = {
+ .nanosecs = false,
.use_modules = true,
.try_vmlinux_path = true,
.demangle = true,
diff --git a/tools/perf/util/symbol_conf.h b/tools/perf/util/symbol_conf.h
index fffea68c1203..095a297c8b47 100644
--- a/tools/perf/util/symbol_conf.h
+++ b/tools/perf/util/symbol_conf.h
@@ -8,6 +8,7 @@ struct strlist;
struct intlist;
struct symbol_conf {
+ bool nanosecs;
unsigned short priv_size;
bool try_vmlinux_path,
init_annotation,
--
2.17.2
^ permalink raw reply related [flat|nested] 27+ messages in thread* [PATCH v2 03/11] perf tools report: Parse time quantum
2019-02-26 3:04 Support sample context in perf report Andi Kleen
2019-02-26 3:04 ` [PATCH v2 01/11] perf tools script: Support insn output for normal samples Andi Kleen
2019-02-26 3:04 ` [PATCH v2 02/11] perf tools report: Support nano seconds Andi Kleen
@ 2019-02-26 3:04 ` Andi Kleen
2019-02-26 3:04 ` [PATCH v2 04/11] perf tools report: Support time sort key Andi Kleen
` (9 subsequent siblings)
12 siblings, 0 replies; 27+ messages in thread
From: Andi Kleen @ 2019-02-26 3:04 UTC (permalink / raw)
To: acme; +Cc: jolsa, linux-perf-users, linux-kernel, eranian, Andi Kleen
From: Andi Kleen <ak@linux.intel.com>
Many workloads change over time. perf report currently aggregates
the whole time range reported in perf.data.
This patch adds an option for a time quantum to quantisize the
perf.data over time.
This just adds the option, will be used in follow on patches
for a time sort key.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
v2:
Move time_quantum to symbol_conf. check for zero time quantum
---
tools/perf/Documentation/perf-report.txt | 4 +++
tools/perf/builtin-report.c | 41 ++++++++++++++++++++++++
tools/perf/util/symbol.c | 1 +
tools/perf/util/symbol_conf.h | 1 +
4 files changed, 47 insertions(+)
diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
index 51dbc519dbce..9ec1702bccdd 100644
--- a/tools/perf/Documentation/perf-report.txt
+++ b/tools/perf/Documentation/perf-report.txt
@@ -497,6 +497,10 @@ include::itrace.txt[]
The period/hits keywords set the base the percentage is computed
on - the samples period or the number of samples (hits).
+--time-quantum::
+ Configure time quantum for time sort key. Default 100ms.
+ Accepts s, us, ms, ns units.
+
include::callchain-overhead-calculation.txt[]
SEE ALSO
diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index 6b2cc91ee26a..f290c19304c8 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -47,6 +47,7 @@
#include <errno.h>
#include <inttypes.h>
#include <regex.h>
+#include "sane_ctype.h"
#include <signal.h>
#include <linux/bitmap.h>
#include <linux/stringify.h>
@@ -926,6 +927,43 @@ report_parse_callchain_opt(const struct option *opt, const char *arg, int unset)
return parse_callchain_report_opt(arg);
}
+static int
+parse_time_quantum(const struct option *opt, const char *arg,
+ int unset __maybe_unused)
+{
+ unsigned long *time_q = opt->value;
+ char *end;
+
+ *time_q = strtoul(arg, &end, 0);
+ if (end == arg)
+ goto parse_err;
+ if (*time_q == 0) {
+ pr_err("time quantum cannot be 0");
+ return -1;
+ }
+ while (isspace(*end))
+ end++;
+ if (*end == 0)
+ return 0;
+ if (!strcmp(end, "s")) {
+ *time_q *= 1000000000;
+ return 0;
+ }
+ if (!strcmp(end, "ms")) {
+ *time_q *= 1000000;
+ return 0;
+ }
+ if (!strcmp(end, "us")) {
+ *time_q *= 1000;
+ return 0;
+ }
+ if (!strcmp(end, "ns"))
+ return 0;
+parse_err:
+ pr_err("Cannot parse time quantum `%s'\n", arg);
+ return -1;
+}
+
int
report_parse_ignore_callees_opt(const struct option *opt __maybe_unused,
const char *arg, int unset __maybe_unused)
@@ -1148,6 +1186,9 @@ int cmd_report(int argc, const char **argv)
"Set percent type local/global-period/hits",
annotate_parse_percent_type),
OPT_BOOLEAN(0, "ns", &symbol_conf.nanosecs, "Show times in nanosecs"),
+ OPT_CALLBACK(0, "time-quantum", &symbol_conf.time_quantum, "time (ms|us|ns)",
+ "Set time quantum for time sort key (default 100ms)",
+ parse_time_quantum),
OPT_END()
};
struct perf_data data = {
diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
index eb873ea1c405..0f80743a1c25 100644
--- a/tools/perf/util/symbol.c
+++ b/tools/perf/util/symbol.c
@@ -45,6 +45,7 @@ struct symbol_conf symbol_conf = {
.demangle = true,
.demangle_kernel = false,
.cumulate_callchain = true,
+ .time_quantum = 100000000, /* 100ms */
.show_hist_headers = true,
.symfs = "",
.event_group = true,
diff --git a/tools/perf/util/symbol_conf.h b/tools/perf/util/symbol_conf.h
index 095a297c8b47..a5684a71b78e 100644
--- a/tools/perf/util/symbol_conf.h
+++ b/tools/perf/util/symbol_conf.h
@@ -56,6 +56,7 @@ struct symbol_conf {
*sym_list_str,
*col_width_list_str,
*bt_stop_list_str;
+ unsigned long time_quantum;
struct strlist *dso_list,
*comm_list,
*sym_list,
--
2.17.2
^ permalink raw reply related [flat|nested] 27+ messages in thread* [PATCH v2 04/11] perf tools report: Support time sort key
2019-02-26 3:04 Support sample context in perf report Andi Kleen
` (2 preceding siblings ...)
2019-02-26 3:04 ` [PATCH v2 03/11] perf tools report: Parse time quantum Andi Kleen
@ 2019-02-26 3:04 ` Andi Kleen
2019-02-26 3:04 ` [PATCH v2 05/11] perf tools report: Use less for scripts output Andi Kleen
` (8 subsequent siblings)
12 siblings, 0 replies; 27+ messages in thread
From: Andi Kleen @ 2019-02-26 3:04 UTC (permalink / raw)
To: acme; +Cc: jolsa, linux-perf-users, linux-kernel, eranian, Andi Kleen
From: Andi Kleen <ak@linux.intel.com>
Add a time sort key to perf report to display samples for
different time quantums separately. This allows easier
analysis of workloads that change over time, and also
will allow looking at the context of samples.
% perf record ...
% perf report --sort time,overhead,symbol --time-quantum 1ms --stdio
...
0.67% 277061.87300 [.] _dl_start
0.50% 277061.87300 [.] f1
0.50% 277061.87300 [.] f2
0.33% 277061.87300 [.] main
0.29% 277061.87300 [.] _dl_lookup_symbol_x
0.29% 277061.87300 [.] dl_main
0.29% 277061.87300 [.] do_lookup_x
0.17% 277061.87300 [.] _dl_debug_initialize
0.17% 277061.87300 [.] _dl_init_paths
0.08% 277061.87300 [.] check_match
0.04% 277061.87300 [.] _dl_count_modids
1.33% 277061.87400 [.] f1
1.33% 277061.87400 [.] f2
1.33% 277061.87400 [.] main
1.17% 277061.87500 [.] main
1.08% 277061.87500 [.] f1
1.08% 277061.87500 [.] f2
1.00% 277061.87600 [.] main
0.83% 277061.87600 [.] f1
0.83% 277061.87600 [.] f2
1.00% 277061.87700 [.] main
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
v2:
Use symbol_conf.time_quantum
---
tools/perf/Documentation/perf-report.txt | 2 ++
tools/perf/util/hist.c | 10 +++++++
tools/perf/util/hist.h | 1 +
tools/perf/util/sort.c | 38 ++++++++++++++++++++++++
tools/perf/util/sort.h | 2 ++
5 files changed, 53 insertions(+)
diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
index 9ec1702bccdd..546d87221ad8 100644
--- a/tools/perf/Documentation/perf-report.txt
+++ b/tools/perf/Documentation/perf-report.txt
@@ -105,6 +105,8 @@ OPTIONS
guest machine
- sample: Number of sample
- period: Raw number of event count of sample
+ - time: Separate the samples by time stamp with the resolution specified by
+ --time-quantum (default 100ms). Specify with overhead and before it.
By default, comm, dso and symbol keys are used.
(i.e. --sort comm,dso,symbol)
diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c
index 669f961316f0..6040eb49ea23 100644
--- a/tools/perf/util/hist.c
+++ b/tools/perf/util/hist.c
@@ -192,6 +192,7 @@ void hists__calc_col_len(struct hists *hists, struct hist_entry *h)
hists__new_col_len(hists, HISTC_MEM_LVL, 21 + 3);
hists__new_col_len(hists, HISTC_LOCAL_WEIGHT, 12);
hists__new_col_len(hists, HISTC_GLOBAL_WEIGHT, 12);
+ hists__new_col_len(hists, HISTC_TIME, 12);
if (h->srcline) {
len = MAX(strlen(h->srcline), strlen(sort_srcline.se_header));
@@ -246,6 +247,14 @@ static void he_stat__add_cpumode_period(struct he_stat *he_stat,
}
}
+static long hist_time(unsigned long time)
+{
+ unsigned long time_quantum = symbol_conf.time_quantum;
+ if (time_quantum)
+ return (time / time_quantum) * time_quantum;
+ return (time / 1000000) * 1000000;
+}
+
static void he_stat__add_period(struct he_stat *he_stat, u64 period,
u64 weight)
{
@@ -626,6 +635,7 @@ __hists__add_entry(struct hists *hists,
.raw_data = sample->raw_data,
.raw_size = sample->raw_size,
.ops = ops,
+ .time = hist_time(sample->time),
}, *he = hists__findnew_entry(hists, &entry, al, sample_self);
if (!hists->has_callchains && he && he->callchain_size != 0)
diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h
index 4af27fbab24f..6279eca56409 100644
--- a/tools/perf/util/hist.h
+++ b/tools/perf/util/hist.h
@@ -31,6 +31,7 @@ enum hist_filter {
enum hist_column {
HISTC_SYMBOL,
+ HISTC_TIME,
HISTC_DSO,
HISTC_THREAD,
HISTC_COMM,
diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
index 2b6c1ccb878c..414c8704cced 100644
--- a/tools/perf/util/sort.c
+++ b/tools/perf/util/sort.c
@@ -15,6 +15,7 @@
#include <traceevent/event-parse.h>
#include "mem-events.h"
#include "annotate.h"
+#include "time-utils.h"
#include <linux/kernel.h>
regex_t parent_regex;
@@ -648,6 +649,42 @@ struct sort_entry sort_socket = {
.se_width_idx = HISTC_SOCKET,
};
+/* --sort time */
+
+static int64_t
+sort__time_cmp(struct hist_entry *left, struct hist_entry *right)
+{
+ return right->time - left->time;
+}
+
+static int hist_entry__time_snprintf(struct hist_entry *he, char *bf,
+ size_t size, unsigned int width)
+{
+ unsigned long secs;
+ unsigned long long nsecs;
+ char he_time[32];
+
+ nsecs = he->time;
+ secs = nsecs / 1000000000;
+ nsecs -= secs * 1000000000;
+
+ if (symbol_conf.nanosecs)
+ snprintf(he_time, sizeof he_time, "%5lu.%09llu: ",
+ secs, nsecs);
+ else
+ timestamp__scnprintf_usec(he->time, he_time,
+ sizeof(he_time));
+
+ return repsep_snprintf(bf, size, "%-.*s", width, he_time);
+}
+
+struct sort_entry sort_time = {
+ .se_header = "Time",
+ .se_cmp = sort__time_cmp,
+ .se_snprintf = hist_entry__time_snprintf,
+ .se_width_idx = HISTC_TIME,
+};
+
/* --sort trace */
static char *get_trace_output(struct hist_entry *he)
@@ -1628,6 +1665,7 @@ static struct sort_dimension common_sort_dimensions[] = {
DIM(SORT_DSO_SIZE, "dso_size", sort_dso_size),
DIM(SORT_CGROUP_ID, "cgroup_id", sort_cgroup_id),
DIM(SORT_SYM_IPC_NULL, "ipc_null", sort_sym_ipc_null),
+ DIM(SORT_TIME, "time", sort_time),
};
#undef DIM
diff --git a/tools/perf/util/sort.h b/tools/perf/util/sort.h
index 2fbee0b1011c..19dceb7f6145 100644
--- a/tools/perf/util/sort.h
+++ b/tools/perf/util/sort.h
@@ -135,6 +135,7 @@ struct hist_entry {
char *srcfile;
struct symbol *parent;
struct branch_info *branch_info;
+ long time;
struct hists *hists;
struct mem_info *mem_info;
void *raw_data;
@@ -231,6 +232,7 @@ enum sort_type {
SORT_DSO_SIZE,
SORT_CGROUP_ID,
SORT_SYM_IPC_NULL,
+ SORT_TIME,
/* branch stack specific sort keys */
__SORT_BRANCH_STACK,
--
2.17.2
^ permalink raw reply related [flat|nested] 27+ messages in thread* [PATCH v2 05/11] perf tools report: Use less for scripts output
2019-02-26 3:04 Support sample context in perf report Andi Kleen
` (3 preceding siblings ...)
2019-02-26 3:04 ` [PATCH v2 04/11] perf tools report: Support time sort key Andi Kleen
@ 2019-02-26 3:04 ` Andi Kleen
2019-02-26 3:04 ` [PATCH v2 06/11] perf tools report: Support running scripts for current time range Andi Kleen
` (7 subsequent siblings)
12 siblings, 0 replies; 27+ messages in thread
From: Andi Kleen @ 2019-02-26 3:04 UTC (permalink / raw)
To: acme; +Cc: jolsa, linux-perf-users, linux-kernel, eranian, Andi Kleen
From: Andi Kleen <ak@linux.intel.com>
The UI viewer for scripts output has a lot of limitations: limited size,
no search or save function, slow, and various other issues.
Just use 'less' to display directly on the terminal instead.
This won't work in gtk mode, but gtk doesn't support these
context menus anyways. If that is ever done could use an terminal
for the output.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
v2:
Remove some unneeded headers
---
tools/perf/ui/browsers/scripts.c | 130 ++++---------------------------
1 file changed, 17 insertions(+), 113 deletions(-)
diff --git a/tools/perf/ui/browsers/scripts.c b/tools/perf/ui/browsers/scripts.c
index 90a32ac69e76..7f36630694bf 100644
--- a/tools/perf/ui/browsers/scripts.c
+++ b/tools/perf/ui/browsers/scripts.c
@@ -1,35 +1,12 @@
// SPDX-License-Identifier: GPL-2.0
-#include <elf.h>
-#include <inttypes.h>
-#include <sys/ttydefaults.h>
-#include <string.h>
#include "../../util/sort.h"
#include "../../util/util.h"
#include "../../util/hist.h"
#include "../../util/debug.h"
#include "../../util/symbol.h"
#include "../browser.h"
-#include "../helpline.h"
#include "../libslang.h"
-/* 2048 lines should be enough for a script output */
-#define MAX_LINES 2048
-
-/* 160 bytes for one output line */
-#define AVERAGE_LINE_LEN 160
-
-struct script_line {
- struct list_head node;
- char line[AVERAGE_LINE_LEN];
-};
-
-struct perf_script_browser {
- struct ui_browser b;
- struct list_head entries;
- const char *script_name;
- int nr_lines;
-};
-
#define SCRIPT_NAMELEN 128
#define SCRIPT_MAX_NO 64
/*
@@ -73,69 +50,29 @@ static int list_scripts(char *script_name)
return ret;
}
-static void script_browser__write(struct ui_browser *browser,
- void *entry, int row)
+static void run_script(char *cmd)
{
- struct script_line *sline = list_entry(entry, struct script_line, node);
- bool current_entry = ui_browser__is_current_entry(browser, row);
-
- ui_browser__set_color(browser, current_entry ? HE_COLORSET_SELECTED :
- HE_COLORSET_NORMAL);
-
- ui_browser__write_nstring(browser, sline->line, browser->width);
+ pr_debug("Running %s\n", cmd);
+ SLang_reset_tty();
+ if (system(cmd) < 0)
+ pr_warning("Cannot run %s\n", cmd);
+ /*
+ * SLang doesn't seem to reset the whole terminal, so be more
+ * forceful to get back to the original state.
+ */
+ printf("\033[c\033[H\033[J");
+ fflush(stdout);
+ SLang_init_tty(0, 0, 0);
+ SLsmg_refresh();
}
-static int script_browser__run(struct perf_script_browser *browser)
-{
- int key;
-
- if (ui_browser__show(&browser->b, browser->script_name,
- "Press ESC to exit") < 0)
- return -1;
-
- while (1) {
- key = ui_browser__run(&browser->b, 0);
-
- /* We can add some special key handling here if needed */
- break;
- }
-
- ui_browser__hide(&browser->b);
- return key;
-}
-
-
int script_browse(const char *script_opt)
{
char cmd[SCRIPT_FULLPATH_LEN*2], script_name[SCRIPT_FULLPATH_LEN];
- char *line = NULL;
- size_t len = 0;
- ssize_t retlen;
- int ret = -1, nr_entries = 0;
- FILE *fp;
- void *buf;
- struct script_line *sline;
-
- struct perf_script_browser script = {
- .b = {
- .refresh = ui_browser__list_head_refresh,
- .seek = ui_browser__list_head_seek,
- .write = script_browser__write,
- },
- .script_name = script_name,
- };
-
- INIT_LIST_HEAD(&script.entries);
-
- /* Save each line of the output in one struct script_line object. */
- buf = zalloc((sizeof(*sline)) * MAX_LINES);
- if (!buf)
- return -1;
- sline = buf;
memset(script_name, 0, SCRIPT_FULLPATH_LEN);
if (list_scripts(script_name))
- goto exit;
+ return -1;
sprintf(cmd, "perf script -s %s ", script_name);
@@ -147,42 +84,9 @@ int script_browse(const char *script_opt)
strcat(cmd, input_name);
}
- strcat(cmd, " 2>&1");
-
- fp = popen(cmd, "r");
- if (!fp)
- goto exit;
-
- while ((retlen = getline(&line, &len, fp)) != -1) {
- strncpy(sline->line, line, AVERAGE_LINE_LEN);
-
- /* If one output line is very large, just cut it short */
- if (retlen >= AVERAGE_LINE_LEN) {
- sline->line[AVERAGE_LINE_LEN - 1] = '\0';
- sline->line[AVERAGE_LINE_LEN - 2] = '\n';
- }
- list_add_tail(&sline->node, &script.entries);
-
- if (script.b.width < retlen)
- script.b.width = retlen;
-
- if (nr_entries++ >= MAX_LINES - 1)
- break;
- sline++;
- }
-
- if (script.b.width > AVERAGE_LINE_LEN)
- script.b.width = AVERAGE_LINE_LEN;
-
- free(line);
- pclose(fp);
+ strcat(cmd, " 2>&1 | less");
- script.nr_lines = nr_entries;
- script.b.nr_entries = nr_entries;
- script.b.entries = &script.entries;
+ run_script(cmd);
- ret = script_browser__run(&script);
-exit:
- free(buf);
- return ret;
+ return 0;
}
--
2.17.2
^ permalink raw reply related [flat|nested] 27+ messages in thread* [PATCH v2 06/11] perf tools report: Support running scripts for current time range
2019-02-26 3:04 Support sample context in perf report Andi Kleen
` (4 preceding siblings ...)
2019-02-26 3:04 ` [PATCH v2 05/11] perf tools report: Use less for scripts output Andi Kleen
@ 2019-02-26 3:04 ` Andi Kleen
2019-02-26 3:04 ` [PATCH v2 07/11] perf tools: Add perf_exe() helper to find perf binary Andi Kleen
` (6 subsequent siblings)
12 siblings, 0 replies; 27+ messages in thread
From: Andi Kleen @ 2019-02-26 3:04 UTC (permalink / raw)
To: acme; +Cc: jolsa, linux-perf-users, linux-kernel, eranian, Andi Kleen
From: Andi Kleen <ak@linux.intel.com>
When using the time sort key, add new context menus to run
scripts for only the currently selected time range. Compute
the correct range for the selection add pass it as the --time option to
perf script.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
v2:
Use symbol_conf.time_quantum
---
tools/perf/ui/browsers/hists.c | 82 +++++++++++++++++++++++++++++-----
1 file changed, 71 insertions(+), 11 deletions(-)
diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c
index aef800d97ea1..663970f766e1 100644
--- a/tools/perf/ui/browsers/hists.c
+++ b/tools/perf/ui/browsers/hists.c
@@ -30,6 +30,7 @@
#include "srcline.h"
#include "string2.h"
#include "units.h"
+#include "time-utils.h"
#include "sane_ctype.h"
@@ -2338,6 +2339,7 @@ static int switch_data_file(void)
}
struct popup_action {
+ unsigned long time;
struct thread *thread;
struct map_symbol ms;
int socket;
@@ -2527,36 +2529,64 @@ static int
do_run_script(struct hist_browser *browser __maybe_unused,
struct popup_action *act)
{
- char script_opt[64];
- memset(script_opt, 0, sizeof(script_opt));
+ char *script_opt;
+ int len;
+ int n = 0;
+
+ len = 100;
+ if (act->thread)
+ len += strlen(thread__comm_str(act->thread));
+ else if (act->ms.sym)
+ len += strlen(act->ms.sym->name);
+ script_opt = malloc(len);
+ if (!script_opt)
+ return -1;
+ script_opt[0] = 0;
if (act->thread) {
- scnprintf(script_opt, sizeof(script_opt), " -c %s ",
+ n = scnprintf(script_opt, len, " -c %s ",
thread__comm_str(act->thread));
} else if (act->ms.sym) {
- scnprintf(script_opt, sizeof(script_opt), " -S %s ",
+ n = scnprintf(script_opt, len, " -S %s ",
act->ms.sym->name);
}
+ if (act->time) {
+ char start[64], end[64];
+ unsigned long starttime = act->time;
+ unsigned long endtime = act->time + symbol_conf.time_quantum;
+
+ if (starttime == endtime) { /* Display 1ms as fallback */
+ starttime -= 1*1000000;
+ endtime += 1*1000000;
+ }
+ timestamp__scnprintf_usec(starttime, start, sizeof start);
+ timestamp__scnprintf_usec(endtime, end, sizeof end);
+ n += snprintf(script_opt + n, len - n, " --time %s,%s", start, end);
+ }
+
script_browse(script_opt);
+ free(script_opt);
return 0;
}
static int
-add_script_opt(struct hist_browser *browser __maybe_unused,
+add_script_opt_2(struct hist_browser *browser __maybe_unused,
struct popup_action *act, char **optstr,
- struct thread *thread, struct symbol *sym)
+ struct thread *thread, struct symbol *sym,
+ const char *tstr)
{
+
if (thread) {
- if (asprintf(optstr, "Run scripts for samples of thread [%s]",
- thread__comm_str(thread)) < 0)
+ if (asprintf(optstr, "Run scripts for samples of thread [%s]%s",
+ thread__comm_str(thread), tstr) < 0)
return 0;
} else if (sym) {
- if (asprintf(optstr, "Run scripts for samples of symbol [%s]",
- sym->name) < 0)
+ if (asprintf(optstr, "Run scripts for samples of symbol [%s]%s",
+ sym->name, tstr) < 0)
return 0;
} else {
- if (asprintf(optstr, "Run scripts for all samples") < 0)
+ if (asprintf(optstr, "Run scripts for all samples%s", tstr) < 0)
return 0;
}
@@ -2566,6 +2596,36 @@ add_script_opt(struct hist_browser *browser __maybe_unused,
return 1;
}
+static int
+add_script_opt(struct hist_browser *browser,
+ struct popup_action *act, char **optstr,
+ struct thread *thread, struct symbol *sym)
+{
+ int n, j;
+ struct hist_entry *he;
+
+ n = add_script_opt_2(browser, act, optstr, thread, sym, "");
+
+ he = hist_browser__selected_entry(browser);
+ if (sort_order && strstr(sort_order, "time")) {
+ char tstr[128];
+
+ optstr++;
+ act++;
+ j = sprintf(tstr, " in ");
+ j += timestamp__scnprintf_usec(he->time, tstr + j,
+ sizeof tstr - j);
+ j += sprintf(tstr + j, "-");
+ timestamp__scnprintf_usec(he->time + symbol_conf.time_quantum,
+ tstr + j,
+ sizeof tstr - j);
+ n += add_script_opt_2(browser, act, optstr, thread, sym,
+ tstr);
+ act->time = he->time;
+ }
+ return n;
+}
+
static int
do_switch_data(struct hist_browser *browser __maybe_unused,
struct popup_action *act __maybe_unused)
--
2.17.2
^ permalink raw reply related [flat|nested] 27+ messages in thread* [PATCH v2 07/11] perf tools: Add perf_exe() helper to find perf binary
2019-02-26 3:04 Support sample context in perf report Andi Kleen
` (5 preceding siblings ...)
2019-02-26 3:04 ` [PATCH v2 06/11] perf tools report: Support running scripts for current time range Andi Kleen
@ 2019-02-26 3:04 ` Andi Kleen
2019-02-26 3:04 ` [PATCH v2 08/11] perf tools report: Support builtin perf script in scripts menu Andi Kleen
` (5 subsequent siblings)
12 siblings, 0 replies; 27+ messages in thread
From: Andi Kleen @ 2019-02-26 3:04 UTC (permalink / raw)
To: acme; +Cc: jolsa, linux-perf-users, linux-kernel, eranian, Andi Kleen
From: Andi Kleen <ak@linux.intel.com>
Also convert one existing user.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
tools/perf/util/header.c | 12 +++---------
tools/perf/util/util.c | 10 ++++++++++
tools/perf/util/util.h | 2 ++
3 files changed, 15 insertions(+), 9 deletions(-)
diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
index 61ce197c5362..2f290a0343d4 100644
--- a/tools/perf/util/header.c
+++ b/tools/perf/util/header.c
@@ -526,17 +526,11 @@ static int write_event_desc(struct feat_fd *ff,
static int write_cmdline(struct feat_fd *ff,
struct perf_evlist *evlist __maybe_unused)
{
- char buf[MAXPATHLEN];
- u32 n;
- int i, ret;
+ char pbuf[MAXPATHLEN], *buf;
+ int i, ret, n;
/* actual path to perf binary */
- ret = readlink("/proc/self/exe", buf, sizeof(buf) - 1);
- if (ret <= 0)
- return -1;
-
- /* readlink() does not add null termination */
- buf[ret] = '\0';
+ buf = perf_exe(pbuf, MAXPATHLEN);
/* account for binary path */
n = perf_env.nr_cmdline + 1;
diff --git a/tools/perf/util/util.c b/tools/perf/util/util.c
index 320b0fef249a..dd693e018aef 100644
--- a/tools/perf/util/util.c
+++ b/tools/perf/util/util.c
@@ -507,3 +507,13 @@ const char *perf_tip(const char *dirpath)
return tip;
}
+
+char *perf_exe(char *buf, int len)
+{
+ int n = readlink("/proc/self/exe", buf, len);
+ if (n > 0) {
+ buf[n] = 0;
+ return buf;
+ }
+ return strcpy(buf, "perf");
+}
diff --git a/tools/perf/util/util.h b/tools/perf/util/util.h
index ece040b799f6..dc2a99ad5c96 100644
--- a/tools/perf/util/util.h
+++ b/tools/perf/util/util.h
@@ -76,6 +76,8 @@ extern bool perf_singlethreaded;
void perf_set_singlethreaded(void);
void perf_set_multithreaded(void);
+char *perf_exe(char *buf, int len);
+
#ifndef O_CLOEXEC
#ifdef __sparc__
#define O_CLOEXEC 0x400000
--
2.17.2
^ permalink raw reply related [flat|nested] 27+ messages in thread* [PATCH v2 08/11] perf tools report: Support builtin perf script in scripts menu
2019-02-26 3:04 Support sample context in perf report Andi Kleen
` (6 preceding siblings ...)
2019-02-26 3:04 ` [PATCH v2 07/11] perf tools: Add perf_exe() helper to find perf binary Andi Kleen
@ 2019-02-26 3:04 ` Andi Kleen
2019-02-26 3:04 ` [PATCH v2 09/11] perf tools: Add utility function to print ns time stamps Andi Kleen
` (4 subsequent siblings)
12 siblings, 0 replies; 27+ messages in thread
From: Andi Kleen @ 2019-02-26 3:04 UTC (permalink / raw)
To: acme; +Cc: jolsa, linux-perf-users, linux-kernel, eranian, Andi Kleen
From: Andi Kleen <ak@linux.intel.com>
The scripts menu traditionally only showed custom perf scripts.
Allow to run standard perf script with useful default options too.
- Normal perf script
- perf script with assembler (needs xed installed)
- perf script with source code output (needs debuginfo)
- perf script with custom arguments
Then we automatically select the right options to
display the information in the perf.data file.
For example with -b display branch contexts.
It's not easily possible to check for xed's existence
in advance. perf script usually gives sensible error messages when
it's not available.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
v2:
Pass --inline if needed
squash! perf tools report: Support builtin perf script in scripts menu
Avoid compiler warnings. Allocate cmd buffer dynamically.
---
tools/perf/ui/browsers/annotate.c | 2 +-
tools/perf/ui/browsers/hists.c | 20 +++--
tools/perf/ui/browsers/scripts.c | 127 +++++++++++++++++++++++-------
tools/perf/util/hist.h | 8 +-
4 files changed, 119 insertions(+), 38 deletions(-)
diff --git a/tools/perf/ui/browsers/annotate.c b/tools/perf/ui/browsers/annotate.c
index 35bdfd8b1e71..98d934a36d86 100644
--- a/tools/perf/ui/browsers/annotate.c
+++ b/tools/perf/ui/browsers/annotate.c
@@ -750,7 +750,7 @@ static int annotate_browser__run(struct annotate_browser *browser,
continue;
case 'r':
{
- script_browse(NULL);
+ script_browse(NULL, NULL);
continue;
}
case 'k':
diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c
index 663970f766e1..10c5aa13ca0e 100644
--- a/tools/perf/ui/browsers/hists.c
+++ b/tools/perf/ui/browsers/hists.c
@@ -2343,6 +2343,7 @@ struct popup_action {
struct thread *thread;
struct map_symbol ms;
int socket;
+ struct perf_evsel *evsel;
int (*fn)(struct hist_browser *browser, struct popup_action *act);
};
@@ -2565,7 +2566,7 @@ do_run_script(struct hist_browser *browser __maybe_unused,
n += snprintf(script_opt + n, len - n, " --time %s,%s", start, end);
}
- script_browse(script_opt);
+ script_browse(script_opt, act->evsel);
free(script_opt);
return 0;
}
@@ -2574,7 +2575,7 @@ static int
add_script_opt_2(struct hist_browser *browser __maybe_unused,
struct popup_action *act, char **optstr,
struct thread *thread, struct symbol *sym,
- const char *tstr)
+ struct perf_evsel *evsel, const char *tstr)
{
if (thread) {
@@ -2592,6 +2593,7 @@ add_script_opt_2(struct hist_browser *browser __maybe_unused,
act->thread = thread;
act->ms.sym = sym;
+ act->evsel = evsel;
act->fn = do_run_script;
return 1;
}
@@ -2599,12 +2601,13 @@ add_script_opt_2(struct hist_browser *browser __maybe_unused,
static int
add_script_opt(struct hist_browser *browser,
struct popup_action *act, char **optstr,
- struct thread *thread, struct symbol *sym)
+ struct thread *thread, struct symbol *sym,
+ struct perf_evsel *evsel)
{
int n, j;
struct hist_entry *he;
- n = add_script_opt_2(browser, act, optstr, thread, sym, "");
+ n = add_script_opt_2(browser, act, optstr, thread, sym, evsel, "");
he = hist_browser__selected_entry(browser);
if (sort_order && strstr(sort_order, "time")) {
@@ -2620,7 +2623,7 @@ add_script_opt(struct hist_browser *browser,
tstr + j,
sizeof tstr - j);
n += add_script_opt_2(browser, act, optstr, thread, sym,
- tstr);
+ evsel, tstr);
act->time = he->time;
}
return n;
@@ -3091,7 +3094,7 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
nr_options += add_script_opt(browser,
&actions[nr_options],
&options[nr_options],
- thread, NULL);
+ thread, NULL, evsel);
}
/*
* Note that browser->selection != NULL
@@ -3106,11 +3109,12 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
nr_options += add_script_opt(browser,
&actions[nr_options],
&options[nr_options],
- NULL, browser->selection->sym);
+ NULL, browser->selection->sym,
+ evsel);
}
}
nr_options += add_script_opt(browser, &actions[nr_options],
- &options[nr_options], NULL, NULL);
+ &options[nr_options], NULL, NULL, evsel);
nr_options += add_switch_opt(browser, &actions[nr_options],
&options[nr_options]);
skip_scripting:
diff --git a/tools/perf/ui/browsers/scripts.c b/tools/perf/ui/browsers/scripts.c
index 7f36630694bf..9e5f87558af6 100644
--- a/tools/perf/ui/browsers/scripts.c
+++ b/tools/perf/ui/browsers/scripts.c
@@ -17,36 +17,111 @@
*/
#define SCRIPT_FULLPATH_LEN 256
+struct script_config {
+ const char **names;
+ char **paths;
+ int index;
+ const char *perf;
+ char extra_format[256];
+};
+
+void attr_to_script(char *extra_format, struct perf_event_attr *attr)
+{
+ extra_format[0] = 0;
+ if (attr->read_format & PERF_FORMAT_GROUP)
+ strcat(extra_format, " -F +metric");
+ if (attr->sample_type & PERF_SAMPLE_BRANCH_STACK)
+ strcat(extra_format, " -F +brstackinsn --xed");
+ if (attr->sample_type & PERF_SAMPLE_REGS_INTR)
+ strcat(extra_format, " -F +iregs");
+ if (attr->sample_type & PERF_SAMPLE_REGS_USER)
+ strcat(extra_format, " -F +uregs");
+ if (attr->sample_type & PERF_SAMPLE_PHYS_ADDR)
+ strcat(extra_format, " -F +phys_addr");
+}
+
+static int add_script_option(const char *name, const char *opt,
+ struct script_config *c)
+{
+ c->names[c->index] = name;
+ if (asprintf(&c->paths[c->index],
+ "%s script %s -F +metric %s %s",
+ c->perf, opt, symbol_conf.inline_name ? " --inline" : "",
+ c->extra_format) < 0)
+ return -1;
+ c->index++;
+ return 0;
+}
+
/*
* When success, will copy the full path of the selected script
* into the buffer pointed by script_name, and return 0.
* Return -1 on failure.
*/
-static int list_scripts(char *script_name)
+static int list_scripts(char *script_name, bool *custom,
+ struct perf_evsel *evsel)
{
- char *buf, *names[SCRIPT_MAX_NO], *paths[SCRIPT_MAX_NO];
- int i, num, choice, ret = -1;
+ char *buf, *paths[SCRIPT_MAX_NO], *names[SCRIPT_MAX_NO];
+ int i, num, choice;
+ int ret = 0;
+ int max_std, custom_perf;
+ char pbuf[256];
+ const char *perf = perf_exe(pbuf, sizeof pbuf);
+ struct script_config scriptc = {
+ .names = (const char **)names,
+ .paths = paths,
+ .perf = perf
+ };
+
+ script_name[0] = 0;
/* Preset the script name to SCRIPT_NAMELEN */
buf = malloc(SCRIPT_MAX_NO * (SCRIPT_NAMELEN + SCRIPT_FULLPATH_LEN));
if (!buf)
- return ret;
+ return -1;
+
+ if (evsel)
+ attr_to_script(scriptc.extra_format, &evsel->attr);
+ add_script_option("Show individual samples", "", &scriptc);
+ add_script_option("Show individual samples with assembler", "-F +insn --xed",
+ &scriptc);
+ add_script_option("Show individual samples with source", "-F +srcline,+srccode",
+ &scriptc);
+ custom_perf = scriptc.index;
+ add_script_option("Show samples with custom perf script arguments", "", &scriptc);
+ i = scriptc.index;
+ max_std = i;
- for (i = 0; i < SCRIPT_MAX_NO; i++) {
- names[i] = buf + i * (SCRIPT_NAMELEN + SCRIPT_FULLPATH_LEN);
+ for (; i < SCRIPT_MAX_NO; i++) {
+ names[i] = buf + (i - max_std) * (SCRIPT_NAMELEN + SCRIPT_FULLPATH_LEN);
paths[i] = names[i] + SCRIPT_NAMELEN;
}
- num = find_scripts(names, paths);
- if (num > 0) {
- choice = ui__popup_menu(num, names);
- if (choice < num && choice >= 0) {
- strcpy(script_name, paths[choice]);
- ret = 0;
- }
+ num = find_scripts(names + max_std, paths + max_std);
+ if (num < 0)
+ num = 0;
+ choice = ui__popup_menu(num + max_std, (char * const *)names);
+ if (choice < 0) {
+ ret = -1;
+ goto out;
}
+ if (choice == custom_perf) {
+ char script_args[50];
+ int key = ui_browser__input_window("perf script command",
+ "Enter perf script command line (without perf script prefix)",
+ script_args, "", 0);
+ if (key != K_ENTER)
+ return -1;
+ sprintf(script_name, "%s script %s", perf, script_args);
+ } else if (choice < num + max_std) {
+ strcpy(script_name, paths[choice]);
+ }
+ *custom = choice >= max_std;
+out:
free(buf);
+ for (i = 0; i < max_std; i++)
+ free(paths[i]);
return ret;
}
@@ -66,27 +141,25 @@ static void run_script(char *cmd)
SLsmg_refresh();
}
-int script_browse(const char *script_opt)
+int script_browse(const char *script_opt, struct perf_evsel *evsel)
{
- char cmd[SCRIPT_FULLPATH_LEN*2], script_name[SCRIPT_FULLPATH_LEN];
+ char *cmd, script_name[SCRIPT_FULLPATH_LEN];
+ bool custom = false;
memset(script_name, 0, SCRIPT_FULLPATH_LEN);
- if (list_scripts(script_name))
+ if (list_scripts(script_name, &custom, evsel))
return -1;
- sprintf(cmd, "perf script -s %s ", script_name);
-
- if (script_opt)
- strcat(cmd, script_opt);
-
- if (input_name) {
- strcat(cmd, " -i ");
- strcat(cmd, input_name);
- }
-
- strcat(cmd, " 2>&1 | less");
+ if (asprintf(&cmd, "%s%s %s %s%s 2>&1 | less",
+ custom ? "perf script -s " : "",
+ script_name,
+ script_opt ? script_opt : "",
+ input_name ? "-i " : "",
+ input_name ? input_name : "") < 0)
+ return -1;
run_script(cmd);
+ free(cmd);
return 0;
}
diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h
index 6279eca56409..2113a6639cea 100644
--- a/tools/perf/util/hist.h
+++ b/tools/perf/util/hist.h
@@ -436,6 +436,8 @@ struct annotation_options;
#ifdef HAVE_SLANG_SUPPORT
#include "../ui/keysyms.h"
+void attr_to_script(char *buf, struct perf_event_attr *attr);
+
int map_symbol__tui_annotate(struct map_symbol *ms, struct perf_evsel *evsel,
struct hist_browser_timer *hbt,
struct annotation_options *annotation_opts);
@@ -450,7 +452,8 @@ int perf_evlist__tui_browse_hists(struct perf_evlist *evlist, const char *help,
struct perf_env *env,
bool warn_lost_event,
struct annotation_options *annotation_options);
-int script_browse(const char *script_opt);
+
+int script_browse(const char *script_opt, struct perf_evsel *evsel);
#else
static inline
int perf_evlist__tui_browse_hists(struct perf_evlist *evlist __maybe_unused,
@@ -479,7 +482,8 @@ static inline int hist_entry__tui_annotate(struct hist_entry *he __maybe_unused,
return 0;
}
-static inline int script_browse(const char *script_opt __maybe_unused)
+static inline int script_browse(const char *script_opt __maybe_unused,
+ struct perf_evsel *evsel __maybe_unused)
{
return 0;
}
--
2.17.2
^ permalink raw reply related [flat|nested] 27+ messages in thread* [PATCH v2 09/11] perf tools: Add utility function to print ns time stamps
2019-02-26 3:04 Support sample context in perf report Andi Kleen
` (7 preceding siblings ...)
2019-02-26 3:04 ` [PATCH v2 08/11] perf tools report: Support builtin perf script in scripts menu Andi Kleen
@ 2019-02-26 3:04 ` Andi Kleen
2019-02-26 3:04 ` [PATCH v2 10/11] perf tools report: Implement browsing of individual samples Andi Kleen
` (3 subsequent siblings)
12 siblings, 0 replies; 27+ messages in thread
From: Andi Kleen @ 2019-02-26 3:04 UTC (permalink / raw)
To: acme; +Cc: jolsa, linux-perf-users, linux-kernel, eranian, Andi Kleen
From: Andi Kleen <ak@linux.intel.com>
Add a utility function to print nanosecond timestamps.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
tools/perf/util/time-utils.c | 8 ++++++++
tools/perf/util/time-utils.h | 1 +
2 files changed, 9 insertions(+)
diff --git a/tools/perf/util/time-utils.c b/tools/perf/util/time-utils.c
index 6193b46050a5..a63bdf4cfd1b 100644
--- a/tools/perf/util/time-utils.c
+++ b/tools/perf/util/time-utils.c
@@ -404,6 +404,14 @@ int timestamp__scnprintf_usec(u64 timestamp, char *buf, size_t sz)
return scnprintf(buf, sz, "%"PRIu64".%06"PRIu64, sec, usec);
}
+int timestamp__scnprintf_nsec(u64 timestamp, char *buf, size_t sz)
+{
+ u64 sec = timestamp / NSEC_PER_SEC;
+ u64 nsec = timestamp % NSEC_PER_SEC;
+
+ return scnprintf(buf, sz, "%"PRIu64".%09"PRIu64, sec, nsec);
+}
+
int fetch_current_timestamp(char *buf, size_t sz)
{
struct timeval tv;
diff --git a/tools/perf/util/time-utils.h b/tools/perf/util/time-utils.h
index 70b177d2b98c..9266cf4a8e58 100644
--- a/tools/perf/util/time-utils.h
+++ b/tools/perf/util/time-utils.h
@@ -24,6 +24,7 @@ bool perf_time__ranges_skip_sample(struct perf_time_interval *ptime_buf,
int num, u64 timestamp);
int timestamp__scnprintf_usec(u64 timestamp, char *buf, size_t sz);
+int timestamp__scnprintf_nsec(u64 timestamp, char *buf, size_t sz);
int fetch_current_timestamp(char *buf, size_t sz);
--
2.17.2
^ permalink raw reply related [flat|nested] 27+ messages in thread* [PATCH v2 10/11] perf tools report: Implement browsing of individual samples
2019-02-26 3:04 Support sample context in perf report Andi Kleen
` (8 preceding siblings ...)
2019-02-26 3:04 ` [PATCH v2 09/11] perf tools: Add utility function to print ns time stamps Andi Kleen
@ 2019-02-26 3:04 ` Andi Kleen
2019-02-26 3:04 ` [PATCH v2 11/11] perf tools: Add some new tips describing the new options Andi Kleen
` (2 subsequent siblings)
12 siblings, 0 replies; 27+ messages in thread
From: Andi Kleen @ 2019-02-26 3:04 UTC (permalink / raw)
To: acme; +Cc: jolsa, linux-perf-users, linux-kernel, eranian, Andi Kleen
From: Andi Kleen <ak@linux.intel.com>
Now report can show whole time periods with perf script,
but the user still has to find individual samples of interest
manually.
It would be expensive and complicated to search for the
right samples in the whole perf file. Typically users
only need to look at a small number of samples for useful
analysis.
Also the full scripts tend to show samples of all CPUs and all
threads mixed up, which can be very confusing on larger systems.
Add a new --samples option to save a small random number of samples
per hist entry
Use a reservoir sample technique to select a representatve
number of samples.
Then allow browsing the samples using perf script
as part of the hist entry context menu. This automatically
adds the right filters, so only the thread or cpu of the sample
is displayed. Then we use less' search functionality
to directly jump the to the time stamp of the selected
sample.
It uses different menus for assembler and source display.
Assembler needs xed installed and source needs debuginfo.
Currently it only supports as many samples as fit on
the screen due to some limitations in the slang ui code.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
v2:
Free names on error path
Pass --inline and --show-*-event to child perf as needed.
---
tools/perf/Documentation/perf-report.txt | 4 ++
tools/perf/builtin-report.c | 2 +
tools/perf/ui/browsers/Build | 1 +
tools/perf/ui/browsers/hists.c | 47 ++++++++++++++
tools/perf/ui/browsers/res_sample.c | 80 ++++++++++++++++++++++++
tools/perf/ui/browsers/scripts.c | 2 +-
tools/perf/util/hist.c | 36 +++++++++++
tools/perf/util/hist.h | 19 ++++++
tools/perf/util/sort.h | 8 +++
tools/perf/util/symbol.c | 1 +
tools/perf/util/symbol_conf.h | 1 +
11 files changed, 200 insertions(+), 1 deletion(-)
create mode 100644 tools/perf/ui/browsers/res_sample.c
diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt
index 546d87221ad8..f441baa794ce 100644
--- a/tools/perf/Documentation/perf-report.txt
+++ b/tools/perf/Documentation/perf-report.txt
@@ -461,6 +461,10 @@ include::itrace.txt[]
--socket-filter::
Only report the samples on the processor socket that match with this filter
+--samples=N::
+ Save N individual samples for each histogram entry to show context in perf
+ report tui browser.
+
--raw-trace::
When displaying traceevent output, do not use print fmt or plugins.
diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index f290c19304c8..f16a9a6f436a 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -1158,6 +1158,8 @@ int cmd_report(int argc, const char **argv)
OPT_BOOLEAN(0, "demangle-kernel", &symbol_conf.demangle_kernel,
"Enable kernel symbol demangling"),
OPT_BOOLEAN(0, "mem-mode", &report.mem_mode, "mem access profile"),
+ OPT_INTEGER(0, "samples", &symbol_conf.res_sample,
+ "Number of samples to save per histogram entry for individual browsing"),
OPT_CALLBACK(0, "percent-limit", &report, "percent",
"Don't show entries under that percent", parse_percent_limit),
OPT_CALLBACK(0, "percentage", NULL, "relative|absolute",
diff --git a/tools/perf/ui/browsers/Build b/tools/perf/ui/browsers/Build
index 8fee56b46502..fdf86f7981ca 100644
--- a/tools/perf/ui/browsers/Build
+++ b/tools/perf/ui/browsers/Build
@@ -3,6 +3,7 @@ perf-y += hists.o
perf-y += map.o
perf-y += scripts.o
perf-y += header.o
+perf-y += res_sample.o
CFLAGS_annotate.o += -DENABLE_SLFUTURE_CONST
CFLAGS_hists.o += -DENABLE_SLFUTURE_CONST
diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c
index 10c5aa13ca0e..e35b274ee863 100644
--- a/tools/perf/ui/browsers/hists.c
+++ b/tools/perf/ui/browsers/hists.c
@@ -2344,6 +2344,7 @@ struct popup_action {
struct map_symbol ms;
int socket;
struct perf_evsel *evsel;
+ enum rstype rstype;
int (*fn)(struct hist_browser *browser, struct popup_action *act);
};
@@ -2571,6 +2572,19 @@ do_run_script(struct hist_browser *browser __maybe_unused,
return 0;
}
+static int
+do_res_sample_script(struct hist_browser *browser __maybe_unused,
+ struct popup_action *act)
+{
+ struct hist_entry *he;
+ int num_res;
+
+ he = hist_browser__selected_entry(browser);
+ num_res = min(he->num_res, symbol_conf.res_sample);
+ res_sample_browse(he->res_samples, num_res, act->evsel, act->rstype);
+ return 0;
+}
+
static int
add_script_opt_2(struct hist_browser *browser __maybe_unused,
struct popup_action *act, char **optstr,
@@ -2629,6 +2643,27 @@ add_script_opt(struct hist_browser *browser,
return n;
}
+static int
+add_res_sample_opt(struct hist_browser *browser __maybe_unused,
+ struct popup_action *act, char **optstr,
+ struct res_sample *res_sample,
+ struct perf_evsel *evsel,
+ enum rstype type)
+{
+ if (!res_sample)
+ return 0;
+
+ if (asprintf(optstr, "Show context for individual samples %s",
+ type == A_ASM ? "with assembler" :
+ type == A_SOURCE ? "with source" : "") < 0)
+ return 0;
+
+ act->fn = do_res_sample_script;
+ act->evsel = evsel;
+ act->rstype = type;
+ return 1;
+}
+
static int
do_switch_data(struct hist_browser *browser __maybe_unused,
struct popup_action *act __maybe_unused)
@@ -3115,6 +3150,18 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
}
nr_options += add_script_opt(browser, &actions[nr_options],
&options[nr_options], NULL, NULL, evsel);
+ nr_options += add_res_sample_opt(browser, &actions[nr_options],
+ &options[nr_options],
+ hist_browser__selected_entry(browser)->res_samples,
+ evsel, A_NORMAL);
+ nr_options += add_res_sample_opt(browser, &actions[nr_options],
+ &options[nr_options],
+ hist_browser__selected_entry(browser)->res_samples,
+ evsel, A_ASM);
+ nr_options += add_res_sample_opt(browser, &actions[nr_options],
+ &options[nr_options],
+ hist_browser__selected_entry(browser)->res_samples,
+ evsel, A_SOURCE);
nr_options += add_switch_opt(browser, &actions[nr_options],
&options[nr_options]);
skip_scripting:
diff --git a/tools/perf/ui/browsers/res_sample.c b/tools/perf/ui/browsers/res_sample.c
new file mode 100644
index 000000000000..2f10c40cc7ab
--- /dev/null
+++ b/tools/perf/ui/browsers/res_sample.c
@@ -0,0 +1,80 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Display a menu with individual samples to browse with perf script */
+#include "util.h"
+#include "hist.h"
+#include "evsel.h"
+#include "hists.h"
+#include "sort.h"
+#include "time-utils.h"
+
+/* In ns. Could make configurable. */
+#define CONTEXT_LEN 100000000
+
+int res_sample_browse(struct res_sample *res_samples, int num_res,
+ struct perf_evsel *evsel, enum rstype rstype)
+{
+ char **names;
+ int i, n;
+ int choice;
+ char *cmd;
+ char pbuf[256], tidbuf[32], cpubuf[32];
+ const char *perf = perf_exe(pbuf, sizeof pbuf);
+ char trange[128], tsample[64];
+ struct res_sample *r;
+ char extra_format[256];
+
+ /* For now since ui__popup_menu doesn't like lists that don't fit */
+ num_res = max(min(SLtt_Screen_Rows - 4, num_res), 0);
+
+ names = calloc(num_res, sizeof(char *));
+ if (!names)
+ return -1;
+ for (i = 0; i < num_res; i++) {
+ char tbuf[64];
+
+ timestamp__scnprintf_nsec(res_samples[i].time, tbuf, sizeof tbuf);
+ if (asprintf(&names[i], "%s: CPU %d tid %d", tbuf,
+ res_samples[i].cpu, res_samples[i].tid) < 0) {
+ while (--i >= 0)
+ free(names[i]);
+ free(names);
+ return -1;
+ }
+ }
+ choice = ui__popup_menu(num_res, names);
+ for (i = 0; i < num_res; i++)
+ free(names[i]);
+ free(names);
+
+ if (choice < 0 || choice >= num_res)
+ return -1;
+ r = &res_samples[choice];
+
+ n = timestamp__scnprintf_nsec(r->time - CONTEXT_LEN, trange, sizeof trange);
+ trange[n++] = ',';
+ timestamp__scnprintf_nsec(r->time + CONTEXT_LEN, trange + n, sizeof trange - n);
+
+ timestamp__scnprintf_nsec(r->time, tsample, sizeof tsample);
+
+ attr_to_script(extra_format, &evsel->attr);
+
+ if (asprintf(&cmd, "%s script %s%s --time %s %s%s %s%s --ns %s %s %s %s | less +/%s",
+ perf,
+ input_name ? "-i " : "",
+ input_name ? input_name : "",
+ trange,
+ r->cpu ? "--cpu " : "",
+ r->cpu ? (sprintf(cpubuf, "%d", r->cpu), cpubuf) : "",
+ r->tid ? "--tid " : "",
+ r->tid ? (sprintf(tidbuf, "%d", r->tid), tidbuf) : "",
+ extra_format,
+ rstype == A_ASM ? "-F +insn --xed" :
+ rstype == A_SOURCE ? "-F +srcline,+srccode" : "",
+ symbol_conf.inline_name ? "--inline" : "",
+ "--show-lost-events --show-switch-events --show-task-events",
+ tsample) < 0)
+ return -1;
+ run_script(cmd);
+ free(cmd);
+ return 0;
+}
diff --git a/tools/perf/ui/browsers/scripts.c b/tools/perf/ui/browsers/scripts.c
index 9e5f87558af6..cdba58447b85 100644
--- a/tools/perf/ui/browsers/scripts.c
+++ b/tools/perf/ui/browsers/scripts.c
@@ -125,7 +125,7 @@ static int list_scripts(char *script_name, bool *custom,
return ret;
}
-static void run_script(char *cmd)
+void run_script(char *cmd)
{
pr_debug("Running %s\n", cmd);
SLang_reset_tty();
diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c
index 6040eb49ea23..ba8a58f2cdd0 100644
--- a/tools/perf/util/hist.c
+++ b/tools/perf/util/hist.c
@@ -593,6 +593,40 @@ static struct hist_entry *hists__findnew_entry(struct hists *hists,
return he;
}
+static unsigned random_max(unsigned high)
+{
+ unsigned thresh = -high % high;
+ for (;;) {
+ unsigned r = random();
+ if (r >= thresh)
+ return r % high;
+ }
+}
+
+static void hists__res_sample(struct hist_entry *he, struct perf_sample *sample)
+{
+ struct res_sample *r;
+ int j;
+
+ if (!he->res_samples) {
+ he->res_samples = calloc(sizeof(struct res_sample),
+ symbol_conf.res_sample);
+ if (!he->res_samples)
+ return;
+ }
+ if (he->num_res < symbol_conf.res_sample) {
+ j = he->num_res++;
+ } else {
+ j = random_max(++he->num_res + 1);
+ if (he->num_res > symbol_conf.res_sample)
+ return;
+ }
+ r = &he->res_samples[j];
+ r->time = sample->time;
+ r->cpu = sample->cpu;
+ r->tid = sample->tid;
+}
+
static struct hist_entry*
__hists__add_entry(struct hists *hists,
struct addr_location *al,
@@ -640,6 +674,8 @@ __hists__add_entry(struct hists *hists,
if (!hists->has_callchains && he && he->callchain_size != 0)
hists->has_callchains = true;
+ if (he && symbol_conf.res_sample)
+ hists__res_sample(he, sample);
return he;
}
diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h
index 2113a6639cea..67b9b91e92df 100644
--- a/tools/perf/util/hist.h
+++ b/tools/perf/util/hist.h
@@ -433,6 +433,13 @@ struct hist_browser_timer {
};
struct annotation_options;
+struct res_sample;
+
+enum rstype {
+ A_NORMAL,
+ A_ASM,
+ A_SOURCE
+};
#ifdef HAVE_SLANG_SUPPORT
#include "../ui/keysyms.h"
@@ -454,6 +461,10 @@ int perf_evlist__tui_browse_hists(struct perf_evlist *evlist, const char *help,
struct annotation_options *annotation_options);
int script_browse(const char *script_opt, struct perf_evsel *evsel);
+
+void run_script(char *cmd);
+int res_sample_browse(struct res_sample *res_samples, int num_res,
+ struct perf_evsel *evsel, enum rstype rstype);
#else
static inline
int perf_evlist__tui_browse_hists(struct perf_evlist *evlist __maybe_unused,
@@ -488,6 +499,14 @@ static inline int script_browse(const char *script_opt __maybe_unused,
return 0;
}
+static inline int res_sample_browse(struct res_sample *res_samples __maybe_unused,
+ int num_res __maybe_unused,
+ struct perf_evsel *evsel __maybe_unused,
+ enum rstype rstype __maybe_unused)
+{
+ return 0;
+}
+
#define K_LEFT -1000
#define K_RIGHT -2000
#define K_SWITCH_INPUT_DATA -3000
diff --git a/tools/perf/util/sort.h b/tools/perf/util/sort.h
index 19dceb7f6145..bb9442ab7a0c 100644
--- a/tools/perf/util/sort.h
+++ b/tools/perf/util/sort.h
@@ -47,6 +47,12 @@ extern struct sort_entry sort_srcline;
extern enum sort_type sort__first_dimension;
extern const char default_mem_sort_order[];
+struct res_sample {
+ u64 time;
+ int cpu;
+ int tid;
+};
+
struct he_stat {
u64 period;
u64 period_sys;
@@ -140,6 +146,8 @@ struct hist_entry {
struct mem_info *mem_info;
void *raw_data;
u32 raw_size;
+ int num_res;
+ struct res_sample *res_samples;
void *trace_output;
struct perf_hpp_list *hpp_list;
struct hist_entry *parent_he;
diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
index 0f80743a1c25..754f7f2c1650 100644
--- a/tools/perf/util/symbol.c
+++ b/tools/perf/util/symbol.c
@@ -50,6 +50,7 @@ struct symbol_conf symbol_conf = {
.symfs = "",
.event_group = true,
.inline_name = true,
+ .res_sample = 0,
};
static enum dso_binary_type binary_type_symtab[] = {
diff --git a/tools/perf/util/symbol_conf.h b/tools/perf/util/symbol_conf.h
index a5684a71b78e..6c55fa6fccec 100644
--- a/tools/perf/util/symbol_conf.h
+++ b/tools/perf/util/symbol_conf.h
@@ -68,6 +68,7 @@ struct symbol_conf {
struct intlist *pid_list,
*tid_list;
const char *symfs;
+ int res_sample;
};
extern struct symbol_conf symbol_conf;
--
2.17.2
^ permalink raw reply related [flat|nested] 27+ messages in thread* [PATCH v2 11/11] perf tools: Add some new tips describing the new options
2019-02-26 3:04 Support sample context in perf report Andi Kleen
` (9 preceding siblings ...)
2019-02-26 3:04 ` [PATCH v2 10/11] perf tools report: Implement browsing of individual samples Andi Kleen
@ 2019-02-26 3:04 ` Andi Kleen
2019-02-26 22:33 ` Support sample context in perf report Jiri Olsa
2019-02-27 11:18 ` Jiri Olsa
12 siblings, 0 replies; 27+ messages in thread
From: Andi Kleen @ 2019-02-26 3:04 UTC (permalink / raw)
To: acme; +Cc: jolsa, linux-perf-users, linux-kernel, eranian, Andi Kleen
From: Andi Kleen <ak@linux.intel.com>
And one old option.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
tools/perf/Documentation/tips.txt | 3 +++
1 file changed, 3 insertions(+)
diff --git a/tools/perf/Documentation/tips.txt b/tools/perf/Documentation/tips.txt
index 849599f39c5e..4ec8107ed512 100644
--- a/tools/perf/Documentation/tips.txt
+++ b/tools/perf/Documentation/tips.txt
@@ -15,6 +15,7 @@ To see callchains in a more compact form: perf report -g folded
Show individual samples with: perf script
Limit to show entries above 5% only: perf report --percent-limit 5
Profiling branch (mis)predictions with: perf record -b / perf report
+To show assembler sample contexts use perf record -b / perf script -F +brstackinsn --xed
Treat branches as callchains: perf report --branch-history
To count events in every 1000 msec: perf stat -I 1000
Print event counts in CSV format with: perf stat -x,
@@ -34,3 +35,5 @@ Show current config key-value pairs: perf config --list
Show user configuration overrides: perf config --user --list
To add Node.js USDT(User-Level Statically Defined Tracing): perf buildid-cache --add `which node`
To report cacheline events from previous recording: perf c2c report
+To browse sample contexts use perf report --sample 10 and select in context menu
+To separate samples by time use perf report --sort time,overhead,sym
--
2.17.2
^ permalink raw reply related [flat|nested] 27+ messages in thread* Re: Support sample context in perf report
2019-02-26 3:04 Support sample context in perf report Andi Kleen
` (10 preceding siblings ...)
2019-02-26 3:04 ` [PATCH v2 11/11] perf tools: Add some new tips describing the new options Andi Kleen
@ 2019-02-26 22:33 ` Jiri Olsa
2019-02-26 22:55 ` Andi Kleen
2019-02-27 11:18 ` Jiri Olsa
12 siblings, 1 reply; 27+ messages in thread
From: Jiri Olsa @ 2019-02-26 22:33 UTC (permalink / raw)
To: Andi Kleen; +Cc: acme, jolsa, linux-perf-users, linux-kernel, eranian
On Mon, Feb 25, 2019 at 07:04:01PM -0800, Andi Kleen wrote:
> [Changes:
> Removed already merged patches.
> Address review feedback, see individual patches.
> Now compiles with gcc 8.
> Some minor bug fixes and improvements.]
>
> We currently have two ways to look at sample data in perf:
> either use perf report to aggregate everything, or use
> perf script to look at all individual samples.
>
> Both ways are useful. Of course aggregation is useful
> to quickly find the most expensive part of the code.
>
> But sometimes a single sample is not good enough to
> determine the problem and we need to look at context, either
> through branch contexts, or other previous samples (e.g. for
> correlating different micro architecture events or computing
> metrics)
>
> This can be done through perf script today, but it can
> be rather cumbersome to find the right samples to look
> at.
>
> Another problem with perf report is that it aggregates
> the whole measurement period. But many real workloads
> have phases where they behave quite differently, and it is
> often not useful to combine them into a single histogram.
>
> While this can be worked around with the --time option
> to report, it can be quite cumbersome.
>
> This patch kit attempts to address some of these
> problems in perf report by making it time aware.
>
> - It adds a new time sort key that allows perf report
> to separate samples from different regions. The time
> regions can be defined with a new --time-quantum option.
>
> - Then it extends the perf script support in the
> tui record browser to allow browsing samples for
> different time regions from within a perf report
> session.
>
> - Extends the report browser script display
> to automatically select sensible defaults
> based on what was recorded. For example it will
> automatically show branch contexts with -b.
>
> - Support browsing the context of individual samples.
> perf report can save a limited number of random samples
> per histogram entry with the new --samples option.
> Then the browser allows directly jumping to any
> of the saved samples and browsing the context on the current
> thread or CPU.
>
> There could be probably be done more to make
> perf report even better for such use cases (e.g. a real
> time line display), but this basic support is good
> enough for many useful usages.
>
> Also available in
> git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc.git perf/streams-2
im still getting compile error the new branch:
CC ui/browsers/hists.o
ui/browsers/hists.c: In function ‘perf_evsel__hists_browse’:
ui/browsers/hists.c:2567:8: error: ‘%s’ directive output may be truncated writing up to 63 bytes into a region of size between 28 and 91 [-Werror=format-truncation=]
n += snprintf(script_opt + n, len - n, " --time %s,%s", start, end);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from /usr/include/stdio.h:862,
from ui/browsers/hists.c:5:
/usr/include/bits/stdio2.h:64:10: note: ‘__builtin___snprintf_chk’ output between 10 and 136 bytes into a destination of size 100
return __builtin___snprintf_chk (__s, __n, __USE_FORTIFY_LEVEL - 1,
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
__bos (__s), __fmt, __va_arg_pack ());
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
mv: cannot stat 'ui/browsers/.hists.o.tmp': No such file or directory
jirka
^ permalink raw reply [flat|nested] 27+ messages in thread* Re: Support sample context in perf report
2019-02-26 22:33 ` Support sample context in perf report Jiri Olsa
@ 2019-02-26 22:55 ` Andi Kleen
0 siblings, 0 replies; 27+ messages in thread
From: Andi Kleen @ 2019-02-26 22:55 UTC (permalink / raw)
To: Jiri Olsa
Cc: Andi Kleen, acme, jolsa, linux-perf-users, linux-kernel, eranian
Jiri Olsa <jolsa@redhat.com> writes:
>
> im still getting compile error the new branch:
>
> CC ui/browsers/hists.o
> ui/browsers/hists.c: In function ‘perf_evsel__hists_browse’:
> ui/browsers/hists.c:2567:8: error: ‘%s’ directive output may be truncated writing up to 63 bytes into a region of size between 28 and 91 [-Werror=format-truncation=]
> n += snprintf(script_opt + n, len - n, " --time %s,%s", start, end);
> ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> In file included from /usr/include/stdio.h:862,
> from ui/browsers/hists.c:5:
> /usr/include/bits/stdio2.h:64:10: note: ‘__builtin___snprintf_chk’ output between 10 and 136 bytes into a destination of size 100
> return __builtin___snprintf_chk (__s, __n, __USE_FORTIFY_LEVEL - 1,
> ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> __bos (__s), __fmt, __va_arg_pack ());
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> cc1: all warnings being treated as errors
> mv: cannot stat 'ui/browsers/.hists.o.tmp': No such file or directory
I tested with gcc 8 and it built on a opensuse leap system.
Of course you never know where you end up with the gcc -Werror
russian roulette. I don't think any of those can really overflow,
it's all false positives. This one is particularly annoying
because the compiler seems to assume that every char[] variable
is filled up to the maximum, which is flat out wrong.
Anyways this patch should help.
-Andi
diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c
index e35b274ee863..8ca988506388 100644
--- a/tools/perf/ui/browsers/hists.c
+++ b/tools/perf/ui/browsers/hists.c
@@ -2554,7 +2554,7 @@ do_run_script(struct hist_browser *browser __maybe_unused,
}
if (act->time) {
- char start[64], end[64];
+ char start[22], end[22];
unsigned long starttime = act->time;
unsigned long endtime = act->time + symbol_conf.time_quantum;
^ permalink raw reply related [flat|nested] 27+ messages in thread
* Re: Support sample context in perf report
2019-02-26 3:04 Support sample context in perf report Andi Kleen
` (11 preceding siblings ...)
2019-02-26 22:33 ` Support sample context in perf report Jiri Olsa
@ 2019-02-27 11:18 ` Jiri Olsa
2019-02-27 16:01 ` Andi Kleen
12 siblings, 1 reply; 27+ messages in thread
From: Jiri Olsa @ 2019-02-27 11:18 UTC (permalink / raw)
To: Andi Kleen; +Cc: acme, jolsa, linux-perf-users, linux-kernel, eranian
On Mon, Feb 25, 2019 at 07:04:01PM -0800, Andi Kleen wrote:
> [Changes:
> Removed already merged patches.
> Address review feedback, see individual patches.
> Now compiles with gcc 8.
> Some minor bug fixes and improvements.]
>
> We currently have two ways to look at sample data in perf:
> either use perf report to aggregate everything, or use
> perf script to look at all individual samples.
>
> Both ways are useful. Of course aggregation is useful
> to quickly find the most expensive part of the code.
>
> But sometimes a single sample is not good enough to
> determine the problem and we need to look at context, either
> through branch contexts, or other previous samples (e.g. for
> correlating different micro architecture events or computing
> metrics)
>
> This can be done through perf script today, but it can
> be rather cumbersome to find the right samples to look
> at.
>
> Another problem with perf report is that it aggregates
> the whole measurement period. But many real workloads
> have phases where they behave quite differently, and it is
> often not useful to combine them into a single histogram.
>
> While this can be worked around with the --time option
> to report, it can be quite cumbersome.
>
> This patch kit attempts to address some of these
> problems in perf report by making it time aware.
>
> - It adds a new time sort key that allows perf report
> to separate samples from different regions. The time
> regions can be defined with a new --time-quantum option.
>
> - Then it extends the perf script support in the
> tui record browser to allow browsing samples for
> different time regions from within a perf report
> session.
>
> - Extends the report browser script display
> to automatically select sensible defaults
> based on what was recorded. For example it will
> automatically show branch contexts with -b.
>
> - Support browsing the context of individual samples.
> perf report can save a limited number of random samples
> per histogram entry with the new --samples option.
> Then the browser allows directly jumping to any
> of the saved samples and browsing the context on the current
> thread or CPU.
>
> There could be probably be done more to make
> perf report even better for such use cases (e.g. a real
> time line display), but this basic support is good
> enough for many useful usages.
>
> Also available in
> git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc.git perf/streams-2
your post is missing this patch, it's only in the branch:
perf tools: Add utility function to fetch executable
jirka
^ permalink raw reply [flat|nested] 27+ messages in thread* Re: Support sample context in perf report
2019-02-27 11:18 ` Jiri Olsa
@ 2019-02-27 16:01 ` Andi Kleen
2019-02-27 16:16 ` Jiri Olsa
0 siblings, 1 reply; 27+ messages in thread
From: Andi Kleen @ 2019-02-27 16:01 UTC (permalink / raw)
To: Jiri Olsa
Cc: Andi Kleen, acme, jolsa, linux-perf-users, linux-kernel, eranian
> > Also available in
> > git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc.git perf/streams-2
>
> your post is missing this patch, it's only in the branch:
> perf tools: Add utility function to fetch executable
Because Arnaldo already merged it. But the branch is still based
on tip/perf/core, which doesn't have it yet.
-Andi
^ permalink raw reply [flat|nested] 27+ messages in thread
* Re: Support sample context in perf report
2019-02-27 16:01 ` Andi Kleen
@ 2019-02-27 16:16 ` Jiri Olsa
2019-02-27 17:29 ` Andi Kleen
0 siblings, 1 reply; 27+ messages in thread
From: Jiri Olsa @ 2019-02-27 16:16 UTC (permalink / raw)
To: Andi Kleen; +Cc: acme, jolsa, linux-perf-users, linux-kernel, eranian
On Wed, Feb 27, 2019 at 08:01:35AM -0800, Andi Kleen wrote:
> > > Also available in
> > > git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc.git perf/streams-2
> >
> > your post is missing this patch, it's only in the branch:
> > perf tools: Add utility function to fetch executable
>
> Because Arnaldo already merged it. But the branch is still based
can't see it on his perf/core branch,
please paste me commit id
jirka
> on tip/perf/core, which doesn't have it yet.
>
> -Andi
^ permalink raw reply [flat|nested] 27+ messages in thread
* Re: Support sample context in perf report
2019-02-27 16:16 ` Jiri Olsa
@ 2019-02-27 17:29 ` Andi Kleen
2019-02-27 17:41 ` Arnaldo Carvalho de Melo
0 siblings, 1 reply; 27+ messages in thread
From: Andi Kleen @ 2019-02-27 17:29 UTC (permalink / raw)
To: Jiri Olsa
Cc: Andi Kleen, acme, jolsa, linux-perf-users, linux-kernel, eranian
On Wed, Feb 27, 2019 at 05:16:59PM +0100, Jiri Olsa wrote:
> On Wed, Feb 27, 2019 at 08:01:35AM -0800, Andi Kleen wrote:
> > > > Also available in
> > > > git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc.git perf/streams-2
> > >
> > > your post is missing this patch, it's only in the branch:
> > > perf tools: Add utility function to fetch executable
> >
> > Because Arnaldo already merged it. But the branch is still based
>
> can't see it on his perf/core branch,
> please paste me commit id
I don't know the commit id, but there was an email from him earlier
that said "applied". Maybe he hasn't pushed yet.
-Andi
^ permalink raw reply [flat|nested] 27+ messages in thread
* Re: Support sample context in perf report
2019-02-27 17:29 ` Andi Kleen
@ 2019-02-27 17:41 ` Arnaldo Carvalho de Melo
2019-02-27 17:56 ` Jiri Olsa
0 siblings, 1 reply; 27+ messages in thread
From: Arnaldo Carvalho de Melo @ 2019-02-27 17:41 UTC (permalink / raw)
To: Andi Kleen; +Cc: Jiri Olsa, jolsa, linux-perf-users, linux-kernel, eranian
Em Wed, Feb 27, 2019 at 09:29:54AM -0800, Andi Kleen escreveu:
> On Wed, Feb 27, 2019 at 05:16:59PM +0100, Jiri Olsa wrote:
> > On Wed, Feb 27, 2019 at 08:01:35AM -0800, Andi Kleen wrote:
> > > > > Also available in
> > > > > git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc.git perf/streams-2
> > > >
> > > > your post is missing this patch, it's only in the branch:
> > > > perf tools: Add utility function to fetch executable
> > >
> > > Because Arnaldo already merged it. But the branch is still based
> >
> > can't see it on his perf/core branch,
> > please paste me commit id
>
> I don't know the commit id, but there was an email from him earlier
> that said "applied". Maybe he hasn't pushed yet.
Perhaps you're referring to this one:
commit 94816add0005595ea33fc8456519be582330401e
Author: Andi Kleen <ak@linux.intel.com>
Date: Sun Feb 24 07:37:19 2019 -0800
perf tools: Add perf_exe() helper to find perf binary
Also convert one existing user.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
?
- Arnaldo
^ permalink raw reply [flat|nested] 27+ messages in thread* Re: Support sample context in perf report
2019-02-27 17:41 ` Arnaldo Carvalho de Melo
@ 2019-02-27 17:56 ` Jiri Olsa
0 siblings, 0 replies; 27+ messages in thread
From: Jiri Olsa @ 2019-02-27 17:56 UTC (permalink / raw)
To: Arnaldo Carvalho de Melo
Cc: Andi Kleen, jolsa, linux-perf-users, linux-kernel, eranian
On Wed, Feb 27, 2019 at 02:41:20PM -0300, Arnaldo Carvalho de Melo wrote:
> Em Wed, Feb 27, 2019 at 09:29:54AM -0800, Andi Kleen escreveu:
> > On Wed, Feb 27, 2019 at 05:16:59PM +0100, Jiri Olsa wrote:
> > > On Wed, Feb 27, 2019 at 08:01:35AM -0800, Andi Kleen wrote:
> > > > > > Also available in
> > > > > > git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc.git perf/streams-2
> > > > >
> > > > > your post is missing this patch, it's only in the branch:
> > > > > perf tools: Add utility function to fetch executable
> > > >
> > > > Because Arnaldo already merged it. But the branch is still based
> > >
> > > can't see it on his perf/core branch,
> > > please paste me commit id
> >
> > I don't know the commit id, but there was an email from him earlier
> > that said "applied". Maybe he hasn't pushed yet.
>
> Perhaps you're referring to this one:
>
> commit 94816add0005595ea33fc8456519be582330401e
> Author: Andi Kleen <ak@linux.intel.com>
> Date: Sun Feb 24 07:37:19 2019 -0800
>
> perf tools: Add perf_exe() helper to find perf binary
>
> Also convert one existing user.
>
> Signed-off-by: Andi Kleen <ak@linux.intel.com>
> Acked-by: Jiri Olsa <jolsa@kernel.org>
>
this one:
perf tools: Add utility function to fetch executable
https://kernel.googlesource.com/pub/scm/linux/kernel/git/ak/linux-misc/+/32e6167f6881c5cfa43a88a9b7fea666d1dbc375%5E%21/
jirka
^ permalink raw reply [flat|nested] 27+ messages in thread