Skip to contentSkip to conversation

Support sample context in perf report

PATCH In review Andi Kleen 3 participants 20 messages Mar 9, 2019 – Mar 22, 2019 perf events

lore
Share & tools
lore

Andi Kleen

author
#0 lore raw

[Changes:
v5:
Address review comments.
Fix perf script --cpu filtering
Use _NSEC defines.
Fix DEBUG=0 build again
Make sample context size configurable.
Some minor 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-5

author
#1 lore raw

PATCH v5 1/15 perf tools: Add utility function to fetch executable

From: Andi Kleen <[email protected]>

Add a utility function to fetch executable code. Convert one user over to it. There are more places doing that, but they do significantly different actions, so they are not easy to fit into a single library function.

Signed-off-by Andi Kleen <[email protected]>
---
 tools/perf/util/Build       |  1 +
 tools/perf/util/fetch.c     | 28 ++++++++++++++++++++++++++++
 tools/perf/util/fetch.h     |  7 +++++++
 tools/perf/util/intel-bts.c | 21 +++------------------
 4 files changed, 39 insertions(+), 18 deletions(-)
 create mode 100644 tools/perf/util/fetch.c
 create mode 100644 tools/perf/util/fetch.h
tools/perf/util/Build +1 -0
index 8dd3102301ea..649321fc3fb9 100644
@@ -65,6 +65,7 @@ perf-y += trace-event-scripting.o
 perf-y += trace-event.o
 perf-y += svghelper.o
 perf-y += sort.o
+perf-y += fetch.o
 perf-y += hist.o
 perf-y += util.o
 perf-y += xyarray.o
tools/perf/util/fetch.c new file +28 -0
new file mode 100644
index 000000000000..1430083e7eca
@@ -0,0 +1,28 @@
+#include "perf.h"
+#include "machine.h"
+#include "thread.h"
+#include "symbol.h"
+#include "map.h"
+#include "fetch.h"
+
+int fetch_exe(u64 ip, struct thread *thread, struct machine *machine,
+	      char *buf, int len, bool *is64bit)
+{
+	struct addr_location al;
+	u8 cpumode;
+	long offset;
+
+	if (machine__kernel_ip(machine, ip))
+		cpumode = PERF_RECORD_MISC_KERNEL;
+	else
+		cpumode = PERF_RECORD_MISC_USER;
+	if (!thread__find_map(thread, cpumode, ip, &al) || !al.map->dso)
+		return -1;
+	if (al.map->dso->data.status == DSO_DATA_STATUS_ERROR)
+		return -1;
+	map__load(al.map);
+	offset = al.map->map_ip(al.map, ip);
+	if (is64bit)
+		*is64bit = al.map->dso->is_64_bit;
+	return dso__data_read_offset(al.map->dso, machine, offset, (u8 *)buf, len);
+}
tools/perf/util/fetch.h new file +7 -0
new file mode 100644
index 000000000000..7b77b8cee55a
@@ -0,0 +1,7 @@
+#ifndef FETCH_H
+#define FETCH_H 1
+
+int fetch_exe(u64 ip, struct thread *thread, struct machine *machine,
+	      char *buf, int len, bool *is64bit);
+
+#endif
tools/perf/util/intel-bts.c +3 -18
index 0c0180c67574..915f4662e52e 100644
@@ -38,6 +38,7 @@
 #include "auxtrace.h"
 #include "intel-pt-decoder/intel-pt-insn-decoder.h"
 #include "intel-bts.h"
+#include "fetch.h"
 
 #define MAX_TIMESTAMP (~0ULL)
 

@@ -328,35 +329,19 @@ static int intel_bts_get_next_insn(struct intel_bts_queue *btsq, u64 ip)
 {
 	struct machine *machine = btsq->bts->machine;
 	struct thread *thread;
-	struct addr_location al;
 	unsigned char buf[INTEL_PT_INSN_BUF_SZ];
 	ssize_t len;
-	int x86_64;
-	uint8_t cpumode;
+	bool x86_64;
 	int err = -1;
 
-	if (machine__kernel_ip(machine, ip))
-		cpumode = PERF_RECORD_MISC_KERNEL;
-	else
-		cpumode = PERF_RECORD_MISC_USER;
-
 	thread = machine__find_thread(machine, -1, btsq->tid);
 	if (!thread)
 		return -1;
 
-	if (!thread__find_map(thread, cpumode, ip, &al) || !al.map->dso)
-		goto out_put;
-
-	len = dso__data_read_addr(al.map->dso, al.map, machine, ip, buf,
-				  INTEL_PT_INSN_BUF_SZ);
+	len = fetch_exe(ip, thread, machine, (char*)buf, INTEL_PT_INSN_BUF_SZ, &x86_64);
 	if (len <= 0)
 		goto out_put;
 
-	/* Load maps to ensure dso->is_64_bit has been updated */
-	map__load(al.map);
-
-	x86_64 = al.map->dso->is_64_bit;
-
 	if (intel_pt_get_insn(buf, len, x86_64, &btsq->intel_pt_insn))
 		goto out_put;
Signature · 1 line
2.20.1
#2 lore raw

Re: [PATCH v5 01/15] perf tools: Add utility function to fetch executable

Quoted 6 lines · Andi Kleen

From: Andi Kleen <[email protected]>

Add a utility function to fetch executable code. Convert one user over to it. There are more places doing that, but they do significantly different actions, so they are not easy to fit into a single library function.

fyi, Arnaldo added thread__memcpy based on your patch:

153259382633 perf thread: Generalize function to copy from thread addr space from intel-bts code

jirka

author
#3 lore raw

PATCH v5 2/15 perf tools script: Support insn output for normal samples

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

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.

This adds some extra checking to ensure that we don't try to decode instructions when using perf.data from a different architecture.

% 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 <[email protected]>

---
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
tools/perf/arch/x86/util/Build +1 -0
index 7aab0be5fc5f..7b8e69bbbdfe 100644
@@ -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
tools/perf/arch/x86/util/archinsn.c new file +28 -0
new file mode 100644
index 000000000000..10b3c2a08b8f
@@ -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;
+}
tools/perf/builtin-script.c +20 -1
index 2d8cb1d1682c..fbc440bdf880 100644
@@ -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:");

@@ -3277,6 +3289,7 @@ 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;

@@ -3615,6 +3628,12 @@ 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);
tools/perf/util/archinsn.h new file +12 -0
new file mode 100644
index 000000000000..448cbb6b8d7e
@@ -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
Signature · 1 line
2.20.1
author
#4 lore raw

PATCH v5 3/15 perf tools script: Filter COMM/FORK/.. events by CPU

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

The --cpu option only filtered samples. Filter other perf events, such as COMM, FORK, SWITCH by the CPU too.

Reported-by Jiri Olsa <[email protected]>
Signed-off-by Andi Kleen <[email protected]>

---
v2: Only filter printf output

---
 tools/perf/builtin-script.c | 62 +++++++++++++++++++++++--------------
 1 file changed, 39 insertions(+), 23 deletions(-)
tools/perf/builtin-script.c +39 -23
index fbc440bdf880..f6b5b19b3307 100644
@@ -2053,9 +2053,11 @@ static int process_comm_event(struct perf_tool *tool,
 		sample->tid = event->comm.tid;
 		sample->pid = event->comm.pid;
 	}
-	perf_sample__fprintf_start(sample, thread, evsel,
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
 				   PERF_RECORD_COMM, stdout);
-	perf_event__fprintf(event, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 	ret = 0;
 out:
 	thread__put(thread);

@@ -2089,9 +2091,11 @@ static int process_namespaces_event(struct perf_tool *tool,
 		sample->tid = event->namespaces.tid;
 		sample->pid = event->namespaces.pid;
 	}
-	perf_sample__fprintf_start(sample, thread, evsel,
-				   PERF_RECORD_NAMESPACES, stdout);
-	perf_event__fprintf(event, stdout);
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   PERF_RECORD_NAMESPACES, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 	ret = 0;
 out:
 	thread__put(thread);

@@ -2123,9 +2127,11 @@ static int process_fork_event(struct perf_tool *tool,
 		sample->tid = event->fork.tid;
 		sample->pid = event->fork.pid;
 	}
-	perf_sample__fprintf_start(sample, thread, evsel,
-				   PERF_RECORD_FORK, stdout);
-	perf_event__fprintf(event, stdout);
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   PERF_RECORD_FORK, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 	thread__put(thread);
 
 	return 0;

@@ -2153,9 +2159,11 @@ static int process_exit_event(struct perf_tool *tool,
 		sample->tid = event->fork.tid;
 		sample->pid = event->fork.pid;
 	}
-	perf_sample__fprintf_start(sample, thread, evsel,
-				   PERF_RECORD_EXIT, stdout);
-	perf_event__fprintf(event, stdout);
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   PERF_RECORD_EXIT, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 
 	if (perf_event__process_exit(tool, event, sample, machine) < 0)
 		err = -1;

@@ -2189,9 +2197,11 @@ static int process_mmap_event(struct perf_tool *tool,
 		sample->tid = event->mmap.tid;
 		sample->pid = event->mmap.pid;
 	}
-	perf_sample__fprintf_start(sample, thread, evsel,
-				   PERF_RECORD_MMAP, stdout);
-	perf_event__fprintf(event, stdout);
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   PERF_RECORD_MMAP, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 	thread__put(thread);
 	return 0;
 }

@@ -2221,9 +2231,11 @@ static int process_mmap2_event(struct perf_tool *tool,
 		sample->tid = event->mmap2.tid;
 		sample->pid = event->mmap2.pid;
 	}
-	perf_sample__fprintf_start(sample, thread, evsel,
-				   PERF_RECORD_MMAP2, stdout);
-	perf_event__fprintf(event, stdout);
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   PERF_RECORD_MMAP2, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 	thread__put(thread);
 	return 0;
 }

@@ -2248,9 +2260,11 @@ static int process_switch_event(struct perf_tool *tool,
 		return -1;
 	}
 
-	perf_sample__fprintf_start(sample, thread, evsel,
-				   PERF_RECORD_SWITCH, stdout);
-	perf_event__fprintf(event, stdout);
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   PERF_RECORD_SWITCH, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 	thread__put(thread);
 	return 0;
 }

@@ -2271,9 +2285,11 @@ process_lost_event(struct perf_tool *tool,
 	if (thread == NULL)
 		return -1;
 
-	perf_sample__fprintf_start(sample, thread, evsel,
-				   PERF_RECORD_LOST, stdout);
-	perf_event__fprintf(event, stdout);
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   PERF_RECORD_LOST, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 	thread__put(thread);
 	return 0;
 }
Signature · 1 line
2.20.1
#5 lore raw

Re: [PATCH v5 03/15] perf tools script: Filter COMM/FORK/.. events by CPU

Quoted 19 lines · Andi Kleen

From: Andi Kleen <[email protected]>

The --cpu option only filtered samples. Filter other perf events, such as COMM, FORK, SWITCH by the CPU too.

Reported-by Jiri Olsa <[email protected]>
Signed-off-by Andi Kleen <[email protected]>

---
v2: Only filter printf output

---
 tools/perf/builtin-script.c | 62 +++++++++++++++++++++++--------------
 1 file changed, 39 insertions(+), 23 deletions(-)
tools/perf/builtin-script.c +0 -0
index fbc440bdf880..f6b5b19b3307 100644

tools/perf/builtin-script.c static int process_comm_event(struct perf_tool *tool, · line 2060

Quoted 10 lines
tools/perf/builtin-script.c +4 -2
@@ -2053,9 +2053,11 @@ static int process_comm_event(struct perf_tool *tool,
 		sample->tid = event->comm.tid;
 		sample->pid = event->comm.pid;
 	}
-	perf_sample__fprintf_start(sample, thread, evsel,
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
 				   PERF_RECORD_COMM, stdout);
-	perf_event__fprintf(event, stdout);
+		perf_event__fprintf(event, stdout);
+	}

looks like this is all same code, and could be put into a function, but it can be done later

thanks,
jirka

Quoted 111 lines
tools/perf/builtin-script.c +35 -21
 	ret = 0;
 out:
 	thread__put(thread);

@@ -2089,9 +2091,11 @@ static int process_namespaces_event(struct perf_tool *tool,
 		sample->tid = event->namespaces.tid;
 		sample->pid = event->namespaces.pid;
 	}
-	perf_sample__fprintf_start(sample, thread, evsel,
-				   PERF_RECORD_NAMESPACES, stdout);
-	perf_event__fprintf(event, stdout);
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   PERF_RECORD_NAMESPACES, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 	ret = 0;
 out:
 	thread__put(thread);

@@ -2123,9 +2127,11 @@ static int process_fork_event(struct perf_tool *tool,
 		sample->tid = event->fork.tid;
 		sample->pid = event->fork.pid;
 	}
-	perf_sample__fprintf_start(sample, thread, evsel,
-				   PERF_RECORD_FORK, stdout);
-	perf_event__fprintf(event, stdout);
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   PERF_RECORD_FORK, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 	thread__put(thread);
 
 	return 0;

@@ -2153,9 +2159,11 @@ static int process_exit_event(struct perf_tool *tool,
 		sample->tid = event->fork.tid;
 		sample->pid = event->fork.pid;
 	}
-	perf_sample__fprintf_start(sample, thread, evsel,
-				   PERF_RECORD_EXIT, stdout);
-	perf_event__fprintf(event, stdout);
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   PERF_RECORD_EXIT, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 
 	if (perf_event__process_exit(tool, event, sample, machine) < 0)
 		err = -1;

@@ -2189,9 +2197,11 @@ static int process_mmap_event(struct perf_tool *tool,
 		sample->tid = event->mmap.tid;
 		sample->pid = event->mmap.pid;
 	}
-	perf_sample__fprintf_start(sample, thread, evsel,
-				   PERF_RECORD_MMAP, stdout);
-	perf_event__fprintf(event, stdout);
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   PERF_RECORD_MMAP, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 	thread__put(thread);
 	return 0;
 }

@@ -2221,9 +2231,11 @@ static int process_mmap2_event(struct perf_tool *tool,
 		sample->tid = event->mmap2.tid;
 		sample->pid = event->mmap2.pid;
 	}
-	perf_sample__fprintf_start(sample, thread, evsel,
-				   PERF_RECORD_MMAP2, stdout);
-	perf_event__fprintf(event, stdout);
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   PERF_RECORD_MMAP2, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 	thread__put(thread);
 	return 0;
 }

@@ -2248,9 +2260,11 @@ static int process_switch_event(struct perf_tool *tool,
 		return -1;
 	}
 
-	perf_sample__fprintf_start(sample, thread, evsel,
-				   PERF_RECORD_SWITCH, stdout);
-	perf_event__fprintf(event, stdout);
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   PERF_RECORD_SWITCH, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 	thread__put(thread);
 	return 0;
 }

@@ -2271,9 +2285,11 @@ process_lost_event(struct perf_tool *tool,
 	if (thread == NULL)
 		return -1;
 
-	perf_sample__fprintf_start(sample, thread, evsel,
-				   PERF_RECORD_LOST, stdout);
-	perf_event__fprintf(event, stdout);
+	if (!cpu_list || test_bit(sample->cpu, cpu_bitmap)) {
+		perf_sample__fprintf_start(sample, thread, evsel,
+					   PERF_RECORD_LOST, stdout);
+		perf_event__fprintf(event, stdout);
+	}
 	thread__put(thread);
 	return 0;
 }
Signature · 1 line
2.20.1
author
#6 lore raw

PATCH v5 4/15 perf tools report: Support nano seconds

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

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 <[email protected]>

---
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(-)
tools/perf/Documentation/perf-report.txt +3 -0
index 1a27bfe05039..51dbc519dbce 100644
@@ -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)
tools/perf/builtin-report.c +1 -0
index 1532ebde6c4b..09180e559ad6 100644
@@ -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 = {
tools/perf/builtin-script.c +5 -6
index f6b5b19b3307..77cc014ff8d0 100644
@@ -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];

@@ -3260,7 +3259,7 @@ 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;
 }
 

@@ -3278,7 +3277,7 @@ 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;
 }
 

@@ -3288,7 +3287,7 @@ 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;
 }
 

@@ -3424,7 +3423,7 @@ 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,
tools/perf/util/symbol.c +1 -0
index 758bf5f74e6e..eb873ea1c405 100644
@@ -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,
tools/perf/util/symbol_conf.h +1 -0
index fffea68c1203..095a297c8b47 100644
@@ -8,6 +8,7 @@ struct strlist;
 struct intlist;
 
 struct symbol_conf {
+	bool		nanosecs;
 	unsigned short	priv_size;
 	bool		try_vmlinux_path,
 			init_annotation,
Signature · 1 line
2.20.1
author
#7 lore raw

PATCH v5 5/15 perf tools report: Parse time quantum

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

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 <[email protected]>

---
v2:
Move time_quantum to symbol_conf. check for zero time quantum v3:
Document s unit
v4:
Use _NSEC defines

---
 tools/perf/Documentation/perf-report.txt |  4 +++
 tools/perf/builtin-report.c              | 42 ++++++++++++++++++++++++
 tools/perf/util/symbol.c                 |  2 ++
 tools/perf/util/symbol_conf.h            |  1 +
 4 files changed, 49 insertions(+)
tools/perf/Documentation/perf-report.txt +4 -0
index 51dbc519dbce..9ec1702bccdd 100644
@@ -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
tools/perf/builtin-report.c +42 -0

63 lines not on this page · raw message

tools/perf/util/symbol.c +2 -0
index eb873ea1c405..913d24340d6b 100644
@@ -29,6 +29,7 @@
 #include <elf.h>
 #include <limits.h>
 #include <symbol/kallsyms.h>
+#include <linux/time64.h>
 #include <sys/utsname.h>
 
 static int dso__load_kernel_sym(struct dso *dso, struct map *map);

@@ -45,6 +46,7 @@ struct symbol_conf symbol_conf = {
 	.demangle		= true,
 	.demangle_kernel	= false,
 	.cumulate_callchain	= true,
+	.time_quantum		= 100 * NSEC_PER_MSEC, /* 100ms */
 	.show_hist_headers	= true,
 	.symfs			= "",
 	.event_group		= true,
tools/perf/util/symbol_conf.h +1 -0
index 095a297c8b47..a5684a71b78e 100644
@@ -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,
Signature · 1 line
2.20.1
author
#8 lore raw

PATCH v5 6/15 perf tools report: Support time sort key

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

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 <[email protected]>

---
v2:
Use symbol_conf.time_quantum
v3:
Use _PER_NSEC defines
Fix fallback path for zero quantum

---
 tools/perf/Documentation/perf-report.txt |  2 ++
 tools/perf/util/hist.c                   | 11 +++++++
 tools/perf/util/hist.h                   |  1 +
 tools/perf/util/sort.c                   | 39 ++++++++++++++++++++++++
 tools/perf/util/sort.h                   |  2 ++
 5 files changed, 55 insertions(+)
tools/perf/Documentation/perf-report.txt +2 -0
index 9ec1702bccdd..546d87221ad8 100644
@@ -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)
tools/perf/util/hist.c +11 -0
index 669f961316f0..08e2b3be4c1b 100644
@@ -19,6 +19,7 @@
 #include <math.h>
 #include <inttypes.h>
 #include <sys/param.h>
+#include <linux/time64.h>
 
 static bool hists__filter_entry_by_dso(struct hists *hists,
 				       struct hist_entry *he);

@@ -192,6 +193,7 @@ 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 +248,14 @@ 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;
+}
+
 static void he_stat__add_period(struct he_stat *he_stat, u64 period,
 				u64 weight)
 {

@@ -626,6 +636,7 @@ __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)
tools/perf/util/hist.h +1 -0
index 4af27fbab24f..6279eca56409 100644
@@ -31,6 +31,7 @@ enum hist_filter {
 
 enum hist_column {
 	HISTC_SYMBOL,
+	HISTC_TIME,
 	HISTC_DSO,
 	HISTC_THREAD,
 	HISTC_COMM,
tools/perf/util/sort.c +39 -0
index d2299e912e59..bdd30cab51cb 100644
@@ -3,6 +3,7 @@
 #include <inttypes.h>
 #include <regex.h>
 #include <linux/mman.h>
+#include <linux/time64.h>
 #include "sort.h"
 #include "hist.h"
 #include "comm.h"

@@ -15,6 +16,7 @@
 #include <traceevent/event-parse.h>
 #include "mem-events.h"
 #include "annotate.h"
+#include "time-utils.h"
 #include <linux/kernel.h>
 
 regex_t		parent_regex;

@@ -654,6 +656,42 @@ 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 / NSEC_PER_SEC;
+	nsecs -= secs * NSEC_PER_SEC;
+
+	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)

@@ -1634,6 +1672,7 @@ 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
tools/perf/util/sort.h +2 -0
index 2fbee0b1011c..19dceb7f6145 100644
@@ -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,
Signature · 1 line
2.20.1
author
#9 lore raw

PATCH v5 7/15 perf tools report: Use less for scripts output

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

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 <[email protected]>

---

v2:
Remove some unneeded headers

---
 tools/perf/ui/browsers/scripts.c | 130 ++++---------------------------
 1 file changed, 17 insertions(+), 113 deletions(-)
tools/perf/ui/browsers/scripts.c +17 -113
index 90a32ac69e76..7f36630694bf 100644
@@ -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;
 }
Signature · 1 line
2.20.1
merge bot
#10 lore raw

[tip:perf/urgent] perf report: Use less for scripts output

perf report: Use less for scripts output
Commit-ID:  75065a85a9705ad4c0135f07fd4467d46ff342a3
Gitweb:     https://git.kernel.org/tip/75065a85a9705ad4c0135f07fd4467d46ff342a3
Author:     Andi Kleen <[email protected]>
AuthorDate: Fri, 8 Mar 2019 21:56:20 -0800
Committer:  Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Mon, 11 Mar 2019 14:03:43 -0300

perf report: Use less for scripts output

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 <[email protected]>
Acked-by Feng Tang <[email protected]>
Cc Jiri Olsa <[email protected]>
Link http://lkml.kernel.org/r/[email protected] open here
Signed-off-by Arnaldo Carvalho de Melo <[email protected]>
---
 tools/perf/ui/browsers/scripts.c | 130 +++++----------------------------------
 1 file changed, 17 insertions(+), 113 deletions(-)
tools/perf/ui/browsers/scripts.c +17 -113
index 90a32ac69e76..7f36630694bf 100644
@@ -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;
 }
author
#11 lore raw

PATCH v5 8/15 perf tools report: Support running scripts for current time range

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

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 <[email protected]>

---
v2:
Use symbol_conf.time_quantum
v3:
Work around broken gcc fortify code warning
Use _NSEC defines

---
 tools/perf/ui/browsers/hists.c | 83 +++++++++++++++++++++++++++++-----
 1 file changed, 72 insertions(+), 11 deletions(-)
tools/perf/ui/browsers/hists.c +72 -11
index aef800d97ea1..f98aeac607dd 100644
@@ -7,6 +7,7 @@
 #include <string.h>
 #include <linux/rbtree.h>
 #include <sys/ttydefaults.h>
+#include <linux/time64.h>
 
 #include "../../util/callchain.h"
 #include "../../util/evsel.h"

@@ -30,6 +31,7 @@
 #include "srcline.h"
 #include "string2.h"
 #include "units.h"
+#include "time-utils.h"
 
 #include "sane_ctype.h"
 

@@ -2338,6 +2340,7 @@ static int switch_data_file(void)
 }
 
 struct popup_action {
+	unsigned long		time;
 	struct thread 		*thread;
 	struct map_symbol 	ms;
 	int			socket;

@@ -2527,36 +2530,64 @@ 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[32], end[32];
+		unsigned long starttime = act->time;
+		unsigned long endtime = act->time + symbol_conf.time_quantum;
+
+		if (starttime == endtime) { /* Display 1ms as fallback */
+			starttime -= 1*NSEC_PER_MSEC;
+			endtime += 1*NSEC_PER_MSEC;
+		}
+		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 +2597,36 @@ 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)
Signature · 1 line
2.20.1
author
#12 lore raw

PATCH v5 9/15 perf tools report: Support builtin perf script in scripts menu

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

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 <[email protected]>

---
v2:
Pass --inline if needed
v3:
Avoid compiler warnings. Allocate cmd buffer dynamically. Change formatting

---
 tools/perf/ui/browsers/annotate.c |   2 +-
 tools/perf/ui/browsers/hists.c    |  23 +++---
 tools/perf/ui/browsers/scripts.c  | 127 +++++++++++++++++++++++-------
 tools/perf/util/hist.h            |   8 +-
 4 files changed, 120 insertions(+), 40 deletions(-)
tools/perf/ui/browsers/annotate.c +1 -1
index 35bdfd8b1e71..98d934a36d86 100644
@@ -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':
tools/perf/ui/browsers/hists.c +13 -10
index f98aeac607dd..fb4430f8982c 100644
@@ -2344,6 +2344,7 @@ 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);
 };

@@ -2566,7 +2567,7 @@ 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;
 }

@@ -2575,7 +2576,7 @@ 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) {

@@ -2593,6 +2594,7 @@ 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;
 }

@@ -2600,12 +2602,13 @@ 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")) {

@@ -2618,10 +2621,9 @@ add_script_opt(struct hist_browser *browser,
 					       sizeof tstr - j);
 		j += sprintf(tstr + j, "-");
 		timestamp__scnprintf_usec(he->time + symbol_conf.time_quantum,
-				          tstr + j,
-				          sizeof tstr - j);
+				          tstr + j, sizeof tstr - j);
 		n += add_script_opt_2(browser, act, optstr, thread, sym,
-					  tstr);
+					  evsel, tstr);
 		act->time = he->time;
 	}
 	return n;

@@ -3092,7 +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

@@ -3107,11 +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:
tools/perf/ui/browsers/scripts.c +100 -27

163 lines not on this page · raw message

tools/perf/util/hist.h +6 -2
index 6279eca56409..2113a6639cea 100644
@@ -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;
 }
Signature · 1 line
2.20.1
author
#13 lore raw

PATCH v5 10/15 perf tools: Add utility function to print ns time stamps

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

Add a utility function to print nanosecond timestamps.

Signed-off-by Andi Kleen <[email protected]>
---
 tools/perf/util/time-utils.c | 8 ++++++++
 tools/perf/util/time-utils.h | 1 +
 2 files changed, 9 insertions(+)
tools/perf/util/time-utils.c +8 -0
index 6193b46050a5..a63bdf4cfd1b 100644
@@ -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;
tools/perf/util/time-utils.h +1 -0
index 70b177d2b98c..9266cf4a8e58 100644
@@ -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);
Signature · 1 line
2.20.1
author
#14 lore raw

PATCH v5 11/15 perf tools report: Implement browsing of individual samples

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

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 <[email protected]>

---
v2:
Free names on error path
Pass --inline and --show-*-event to child perf as needed. v3:
Don't use num_res to keep track of total samples Don't pass -1 cpus to perf script.
Don't pass extra event option with tid filtering. Fix sampling bug.
v4:
Add perfconfig option for context
Use _PER_NSEC defines

---
 tools/perf/Documentation/perf-config.txt |  6 ++
 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      | 95 ++++++++++++++++++++++++
 tools/perf/ui/browsers/scripts.c         |  2 +-
 tools/perf/util/hist.c                   | 36 +++++++++
 tools/perf/util/hist.h                   | 22 ++++++
 tools/perf/util/sort.h                   |  8 ++
 tools/perf/util/symbol.c                 |  1 +
 tools/perf/util/symbol_conf.h            |  1 +
 12 files changed, 224 insertions(+), 1 deletion(-)
 create mode 100644 tools/perf/ui/browsers/res_sample.c
tools/perf/Documentation/perf-config.txt +6 -0
index 86f3dcc15f83..2d0fb7613134 100644
@@ -584,6 +584,12 @@ llvm.*::
 	llvm.opts::
 		Options passed to llc.
 
+samples.*::
+
+	samples.context::
+		Define how many ns worth of time to show
+		around samples in perf report sample context browser.
+
 SEE ALSO
 --------
 linkperf:perf[1]
tools/perf/Documentation/perf-report.txt +4 -0
index 546d87221ad8..f441baa794ce 100644
@@ -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.
tools/perf/builtin-report.c +2 -0
index 46bd80363c90..9c57c0ccd673 100644
@@ -1159,6 +1159,8 @@ 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",
tools/perf/ui/browsers/Build +1 -0
index 8fee56b46502..fdf86f7981ca 100644
@@ -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
tools/perf/ui/browsers/hists.c +47 -0

77 lines not on this page · raw message

tools/perf/ui/browsers/res_sample.c new file +95 -0

95 lines not on this page · raw message

tools/perf/ui/browsers/scripts.c +1 -1
index 9e5f87558af6..cdba58447b85 100644
@@ -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();
tools/perf/util/hist.c +36 -0
index 08e2b3be4c1b..da3a944f42aa 100644
@@ -445,6 +445,14 @@ static int hist_entry__init(struct hist_entry *he,
 			return -ENOMEM;
 		}
 	}
+
+	if (symbol_conf.res_sample) {
+		he->res_samples = calloc(sizeof(struct res_sample),
+					symbol_conf.res_sample);
+		if (!he->res_samples)
+			return -ENOMEM;
+	}
+
 	INIT_LIST_HEAD(&he->pairs.node);
 	thread__get(he->thread);
 	he->hroot_in  = RB_ROOT_CACHED;

@@ -594,6 +602,32 @@ 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->num_res < symbol_conf.res_sample) {
+		j = he->num_res++;
+	} else {
+		j = random_max(symbol_conf.res_sample);
+	}
+	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,

@@ -641,6 +675,8 @@ __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;
 }
tools/perf/util/hist.h +22 -0
index 2113a6639cea..76ff6c6d03b8 100644
@@ -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 +461,11 @@ 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);
+void res_sample_init(void);
 #else
 static inline
 int perf_evlist__tui_browse_hists(struct perf_evlist *evlist __maybe_unused,

@@ -488,6 +500,16 @@ 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;
+}
+
+static inline void res_sample_init(void) {}
+
 #define K_LEFT  -1000
 #define K_RIGHT -2000
 #define K_SWITCH_INPUT_DATA -3000
tools/perf/util/sort.h +8 -0
index 19dceb7f6145..bb9442ab7a0c 100644
@@ -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;
tools/perf/util/symbol.c +1 -0
index 913d24340d6b..fb7bb43786d7 100644
@@ -51,6 +51,7 @@ struct symbol_conf symbol_conf = {
 	.symfs			= "",
 	.event_group		= true,
 	.inline_name		= true,
+	.res_sample		= 0,
 };
 
 static enum dso_binary_type binary_type_symtab[] = {
tools/perf/util/symbol_conf.h +1 -0
index a5684a71b78e..6c55fa6fccec 100644
@@ -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;
Signature · 1 line
2.20.1
#15 lore raw

Re: [PATCH v5 11/15] perf tools report: Implement browsing of individual samples

On Fri, Mar 08, 2019 at 09:56:24PM -0800, Andi Kleen wrote:

SNIP

tools/perf/util/hist.c static int hist_entry__init(struct hist_entry *he, · line 452

Quoted 13 lines
tools/perf/util/hist.c +5 -0
index 08e2b3be4c1b..da3a944f42aa 100644
@@ -445,6 +445,14 @@ static int hist_entry__init(struct hist_entry *he,
 			return -ENOMEM;
 		}
 	}
+
+	if (symbol_conf.res_sample) {
+		he->res_samples = calloc(sizeof(struct res_sample),
+					symbol_conf.res_sample);
+		if (!he->res_samples)

you need to free the memory in here, anyway this function just changed, so please rebase on latest Arnaldo's perf/core

jirka

Quoted 3 lines
tools/perf/util/hist.c +3 -0
+			return -ENOMEM;
+	}
+

SNIP

author
#16 lore raw

PATCH v5 12/15 perf tools: Add some new tips describing the new options

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

Signed-off-by Andi Kleen <[email protected]>

---
v2:
Even more tips.
v3:
Even more tips

---
 tools/perf/Documentation/tips.txt | 7 +++++++
 1 file changed, 7 insertions(+)
tools/perf/Documentation/tips.txt +7 -0
index 849599f39c5e..869965d629ce 100644
@@ -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 +35,9 @@ 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
+To set sample time separation other than 100ms with --sort time use --time-quantum
+Add -I to perf report to sample register values visible in perf report context.
+To show IPC for sampling periods use perf record -e '{cycles,instructions}:S' and then browse context
+To show context switches in perf report sample context add --switch-events to perf record.
Signature · 1 line
2.20.1
author
#17 lore raw

PATCH v5 13/15 perf tools report: Add custom scripts to script menu

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

Add a way to define custom scripts through ~/.perfconfig, which are then added to the scripts menu. The scripts get the same arguments as perf script, in particular -i, --cpu, --tid.

Signed-off-by Andi Kleen <[email protected]>
---
 tools/perf/Documentation/perf-config.txt |  8 ++++++++
 tools/perf/ui/browsers/scripts.c         | 20 ++++++++++++++++++++
 2 files changed, 28 insertions(+)
tools/perf/Documentation/perf-config.txt +8 -0
index 2d0fb7613134..c3323228d3d0 100644
@@ -590,6 +590,14 @@ samples.*::
 		Define how many ns worth of time to show
 		around samples in perf report sample context browser.
 
+script.*::
+
+	Any option defines a script that is added to the scripts menu
+	in the interactive perf browser and whose output is displayed.
+	The name of the option is the name, the value is a script command line.
+	The script gets the same options passed as a full perf script,
+	in particular -i perfdata file, --cpu, --tid
+
 SEE ALSO
 --------
 linkperf:perf[1]
tools/perf/ui/browsers/scripts.c +20 -0
index cdba58447b85..5d407ab834b5 100644
@@ -6,6 +6,7 @@
 #include "../../util/symbol.h"
 #include "../browser.h"
 #include "../libslang.h"
+#include "config.h"
 
 #define SCRIPT_NAMELEN	128
 #define SCRIPT_MAX_NO	64

@@ -53,6 +54,24 @@ static int add_script_option(const char *name, const char *opt,
 	return 0;
 }
 
+static int scripts_config(const char *var, const char *value, void *data)
+{
+	struct script_config *c = data;
+
+	if (!strstarts(var, "script."))
+		return -1;
+	if (c->index >= SCRIPT_MAX_NO)
+		return -1;
+	c->names[c->index] = strdup(var + 7);
+	if (!c->names[c->index])
+		return -1;
+	if (asprintf(&c->paths[c->index], "%s %s", value,
+		     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.

@@ -87,6 +106,7 @@ static int list_scripts(char *script_name, bool *custom,
 			  &scriptc);
 	add_script_option("Show individual samples with source", "-F +srcline,+srccode",
 			  &scriptc);
+	perf_config(scripts_config, &scriptc);
 	custom_perf = scriptc.index;
 	add_script_option("Show samples with custom perf script arguments", "", &scriptc);
 	i = scriptc.index;
Signature · 1 line
2.20.1
author
#18 lore raw

PATCH v5 14/15 perf tools script: Add array bound checking to list_scripts

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

Don't overflow array when the scripts directory is too large, or the script file name is too long.

Signed-off-by Andi Kleen <[email protected]>
---
 tools/perf/builtin-script.c      | 8 ++++++--
 tools/perf/builtin.h             | 3 ++-
 tools/perf/ui/browsers/scripts.c | 3 ++-
 3 files changed, 10 insertions(+), 4 deletions(-)
tools/perf/builtin-script.c +6 -2
index 77cc014ff8d0..389118500694 100644
@@ -2975,7 +2975,8 @@ static int check_ev_match(char *dir_name, char *scriptname,
  * will list all statically runnable scripts, select one, execute it and
  * show the output in a perf browser.
  */
-int find_scripts(char **scripts_array, char **scripts_path_array)
+int find_scripts(char **scripts_array, char **scripts_path_array, int num,
+		 int pathlen)
 {
 	struct dirent *script_dirent, *lang_dirent;
 	char scripts_path[MAXPATHLEN], lang_path[MAXPATHLEN];

@@ -3020,7 +3021,10 @@ int find_scripts(char **scripts_array, char **scripts_path_array)
 			/* Skip those real time scripts: xxxtop.p[yl] */
 			if (strstr(script_dirent->d_name, "top."))
 				continue;
-			sprintf(scripts_path_array[i], "%s/%s", lang_path,
+			if (i >= num)
+				break;
+			snprintf(scripts_path_array[i], pathlen, "%s/%s",
+				lang_path,
 				script_dirent->d_name);
 			temp = strchr(script_dirent->d_name, '.');
 			snprintf(scripts_array[i],
tools/perf/builtin.h +2 -1
index 05745f3ce912..999fe9170122 100644
@@ -40,5 +40,6 @@ int cmd_mem(int argc, const char **argv);
 int cmd_data(int argc, const char **argv);
 int cmd_ftrace(int argc, const char **argv);
 
-int find_scripts(char **scripts_array, char **scripts_path_array);
+int find_scripts(char **scripts_array, char **scripts_path_array, int num,
+		 int pathlen);
 #endif
tools/perf/ui/browsers/scripts.c +2 -1
index 5d407ab834b5..6759d6657e1a 100644
@@ -117,7 +117,8 @@ static int list_scripts(char *script_name, bool *custom,
 		paths[i] = names[i] + SCRIPT_NAMELEN;
 	}
 
-	num = find_scripts(names + max_std, paths + max_std);
+	num = find_scripts(names + max_std, paths + max_std, SCRIPT_MAX_NO - max_std,
+			SCRIPT_FULLPATH_LEN);
 	if (num < 0)
 		num = 0;
 	choice = ui__popup_menu(num + max_std, (char * const *)names);
Signature · 1 line
2.20.1
author
#19 lore raw

PATCH v5 15/15 perf tools ui: Fix ui popup browser for many entries

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

Fix the argv ui browser code to correctly display more entries than fit on the screen without crashing. The problem was some type confusion with pointer types in the ->seek function. Do the argv arithmetic correctly with char ** pointers. Also add some asserts to find overruns and limit the display function correctly.

Then finally remove a workaround for this in the res sample browser.

Signed-off-by Andi Kleen <[email protected]>
---
 tools/perf/ui/browser.c             | 10 +++++++---
 tools/perf/ui/browsers/res_sample.c |  3 ---
 2 files changed, 7 insertions(+), 6 deletions(-)
tools/perf/ui/browser.c +7 -3
index 4f75561424ed..4ad37d8c7d6a 100644
@@ -611,14 +611,16 @@ void ui_browser__argv_seek(struct ui_browser *browser, off_t offset, int whence)
 		browser->top = browser->entries;
 		break;
 	case SEEK_CUR:
-		browser->top = browser->top + browser->top_idx + offset;
+		browser->top = (char **)browser->top + offset;
 		break;
 	case SEEK_END:
-		browser->top = browser->top + browser->nr_entries - 1 + offset;
+		browser->top = (char **)browser->entries + browser->nr_entries - 1 + offset;
 		break;
 	default:
 		return;
 	}
+	assert((char **)browser->top < (char **)browser->entries + browser->nr_entries);
+	assert((char **)browser->top >= (char **)browser->entries);
 }
 
 unsigned int ui_browser__argv_refresh(struct ui_browser *browser)

@@ -630,7 +632,9 @@ unsigned int ui_browser__argv_refresh(struct ui_browser *browser)
 		browser->top = browser->entries;
 
 	pos = (char **)browser->top;
-	while (idx < browser->nr_entries) {
+	while (idx < browser->nr_entries &&
+	       row < (unsigned)SLtt_Screen_Rows - 1) {
+		assert(pos < (char **)browser->entries + browser->nr_entries);
 		if (!browser->filter || !browser->filter(browser, *pos)) {
 			ui_browser__gotorc(browser, row, 0);
 			browser->write(browser, pos, row);
tools/perf/ui/browsers/res_sample.c +0 -3
index c2a8536904bc..8befd8e984ab 100644
@@ -37,9 +37,6 @@ int res_sample_browse(struct res_sample *res_samples, int num_res,
 	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;
Signature · 1 line
2.20.1