perf trace: New tool

Initially should look loosely like the venerable 'strace' tool, but
using the infrastructure in the perf tools to allow tracing extra
targets:

  [acme@sandy linux]$ perf trace --hell
  Error: unknown option `hell'

   usage: perf trace <PID>

      -p, --pid <pid>       trace events on existing process id
          --tid <tid>       trace events on existing thread id
          --all-cpus        system-wide collection from all CPUs
          --cpu <cpu>       list of cpus to monitor
          --no-inherit      child tasks do not inherit counters
          --mmap-pages <n>  number of mmap data pages
          --uid <user>      user to profile

  [acme@sandy linux]$

Those should have the same semantics as when using with 'perf record'.

It gets stuck sometimes, but hey, it works sometimes too!

In time it should support perf.data based workloads, i.e. it should have
a:
	-o filename

Command line option that will produce a perf.data file that can then be
used with 'perf trace' or any of the other perf tools (script, report,
etc).

It will also eventually have the set of functionalities described in the
previous 'trace' prototype by Thomas Gleixner:

   "Announcing a new utility: 'trace'"
   http://lwn.net/Articles/415728/

Also planned is to have some of the features suggested in the comments
of that LWN article.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/n/tip-v9x3q9rv4caxtox7wtjpchq5@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

6 files changed, 358 insertions, 1 deletions

diff --git a/tools/perf/Documentation/perf-trace.txt b/tools/perf/Documentation/perf-trace.txt
new file mode 100644
index 00000000000..3a2ae37310a
--- /dev/null
+++ b/tools/perf/Documentation/perf-trace.txt
@@ -0,0 +1,53 @@
+perf-trace(1)
+=============
+
+NAME
+----
+perf-trace - strace inspired tool
+
+SYNOPSIS
+--------
+[verse]
+'perf trace'
+
+DESCRIPTION
+-----------
+This command will show the events associated with the target, initially
+syscalls, but other system events like pagefaults, task lifetime events,
+scheduling events, etc.
+
+Initially this is a live mode only tool, but eventually will work with
+perf.data files like the other tools, allowing a detached 'record' from
+analysis phases.
+
+OPTIONS
+-------
+
+--all-cpus::
+        System-wide collection from all CPUs.
+
+-p::
+--pid=::
+	Record events on existing process ID (comma separated list).
+
+--tid=::
+        Record events on existing thread ID (comma separated list).
+
+--uid=::
+        Record events in threads owned by uid. Name or number.
+
+--no-inherit::
+	Child tasks do not inherit counters.
+
+--mmap-pages=::
+	Number of mmap data pages. Must be a power of two.
+
+--cpu::
+Collect samples only on the list of CPUs provided. Multiple CPUs can be provided as a
+comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2.
+In per-thread mode with inheritance mode on (default), Events are captured only when
+the thread executes on the designated CPUs. Default is to monitor all CPUs.
+
+SEE ALSO
+--------
+linkperf:perf-record[1], linkperf:perf-script[1]
diff --git a/tools/perf/Makefile b/tools/perf/Makefile
index 251dcd7fb5a..6958ba4f5dc 100644
--- a/tools/perf/Makefile
+++ b/tools/perf/Makefile
@@ -102,7 +102,7 @@ ifdef PARSER_DEBUG
 endif
 
 CFLAGS = -fno-omit-frame-pointer -ggdb3 -funwind-tables -Wall -Wextra -std=gnu99 $(CFLAGS_WERROR) $(CFLAGS_OPTIMIZE) $(EXTRA_WARNINGS) $(EXTRA_CFLAGS) $(PARSER_DEBUG_CFLAGS)
-EXTLIBS = -lpthread -lrt -lelf -lm
+EXTLIBS = -lpthread -lrt -lelf -lm -laudit
 ALL_CFLAGS = $(CFLAGS) -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE
 ALL_LDFLAGS = $(LDFLAGS)
 STRIP ?= strip
@@ -442,6 +442,7 @@ BUILTIN_OBJS += $(OUTPUT)builtin-kmem.o
 BUILTIN_OBJS += $(OUTPUT)builtin-lock.o
 BUILTIN_OBJS += $(OUTPUT)builtin-kvm.o
 BUILTIN_OBJS += $(OUTPUT)builtin-test.o
+BUILTIN_OBJS += $(OUTPUT)builtin-trace.o
 BUILTIN_OBJS += $(OUTPUT)builtin-inject.o
 
 PERFLIBS = $(LIB_FILE) $(LIBTRACEEVENT)
diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c
new file mode 100644
index 00000000000..5fa1820cc96
--- /dev/null
+++ b/tools/perf/builtin-trace.c
@@ -0,0 +1,300 @@
+#include "builtin.h"
+#include "util/evlist.h"
+#include "util/parse-options.h"
+#include "util/thread_map.h"
+#include "event-parse.h"
+
+#include <libaudit.h>
+#include <stdlib.h>
+
+static struct syscall_fmt {
+	const char *name;
+	bool	   errmsg;
+	bool	   timeout;
+} syscall_fmts[] = {
+	{ .name	    = "futex",	  .errmsg = true, },
+	{ .name	    = "poll",	  .errmsg = true, .timeout = true, },
+	{ .name	    = "ppoll",	  .errmsg = true, .timeout = true, },
+	{ .name	    = "read",	  .errmsg = true, },
+	{ .name	    = "recvfrom", .errmsg = true, },
+	{ .name	    = "select",	  .errmsg = true, .timeout = true, },
+};
+
+static int syscall_fmt__cmp(const void *name, const void *fmtp)
+{
+	const struct syscall_fmt *fmt = fmtp;
+	return strcmp(name, fmt->name);
+}
+
+static struct syscall_fmt *syscall_fmt__find(const char *name)
+{
+	const int nmemb = ARRAY_SIZE(syscall_fmts);
+	return bsearch(name, syscall_fmts, nmemb, sizeof(struct syscall_fmt), syscall_fmt__cmp);
+}
+
+struct syscall {
+	struct event_format *tp_format;
+	const char	    *name;
+	struct syscall_fmt  *fmt;
+};
+
+struct trace {
+	int			audit_machine;
+	struct {
+		int		max;
+		struct syscall  *table;
+	} syscalls;
+	struct perf_record_opts opts;
+};
+
+static int trace__read_syscall_info(struct trace *trace, int id)
+{
+	char tp_name[128];
+	struct syscall *sc;
+
+	if (id > trace->syscalls.max) {
+		struct syscall *nsyscalls = realloc(trace->syscalls.table, (id + 1) * sizeof(*sc));
+
+		if (nsyscalls == NULL)
+			return -1;
+
+		if (trace->syscalls.max != -1) {
+			memset(nsyscalls + trace->syscalls.max + 1, 0,
+			       (id - trace->syscalls.max) * sizeof(*sc));
+		} else {
+			memset(nsyscalls, 0, (id + 1) * sizeof(*sc));
+		}
+
+		trace->syscalls.table = nsyscalls;
+		trace->syscalls.max   = id;
+	}
+
+	sc = trace->syscalls.table + id;
+	sc->name = audit_syscall_to_name(id, trace->audit_machine);
+	if (sc->name == NULL)
+		return -1;
+
+	snprintf(tp_name, sizeof(tp_name), "sys_enter_%s", sc->name);
+
+	sc->tp_format = event_format__new("syscalls", tp_name);
+	sc->fmt = syscall_fmt__find(sc->name);
+
+	return sc->tp_format != NULL ? 0 : -1;
+}
+
+static size_t syscall__fprintf_args(struct syscall *sc, unsigned long *args, FILE *fp)
+{
+	int i = 0;
+	size_t printed = 0;
+
+	if (sc->tp_format != NULL) {
+		struct format_field *field;
+
+		for (field = sc->tp_format->format.fields->next; field; field = field->next) {
+			printed += fprintf(fp, "%s%s: %ld", printed ? ", " : "",
+					   field->name, args[i++]);
+		}
+	} else {
+		while (i < 6) {
+			printed += fprintf(fp, "%sarg%d: %ld", printed ? ", " : "", i, args[i]);
+			++i;
+		}
+	}
+
+	return printed;
+}
+
+static int trace__run(struct trace *trace)
+{
+	struct perf_evlist *evlist = perf_evlist__new(NULL, NULL);
+	struct perf_evsel *evsel, *evsel_enter, *evsel_exit;
+	int err = -1, i, nr_events = 0, before;
+
+	if (evlist == NULL) {
+		printf("Not enough memory to run!\n");
+		goto out;
+	}
+
+	evsel_enter = perf_evsel__newtp("raw_syscalls", "sys_enter", 0);
+	if (evsel_enter == NULL) {
+		printf("Couldn't read the raw_syscalls:sys_enter tracepoint information!\n");
+		goto out_delete_evlist;
+	}
+
+	perf_evlist__add(evlist, evsel_enter);
+
+	evsel_exit = perf_evsel__newtp("raw_syscalls", "sys_exit", 1);
+	if (evsel_exit == NULL) {
+		printf("Couldn't read the raw_syscalls:sys_exit tracepoint information!\n");
+		goto out_delete_evlist;
+	}
+
+	perf_evlist__add(evlist, evsel_exit);
+
+	err = perf_evlist__create_maps(evlist, &trace->opts.target);
+	if (err < 0) {
+		printf("Problems parsing the target to trace, check your options!\n");
+		goto out_delete_evlist;
+	}
+
+	perf_evlist__config_attrs(evlist, &trace->opts);
+
+	err = perf_evlist__open(evlist);
+	if (err < 0) {
+		printf("Couldn't create the events: %s\n", strerror(errno));
+		goto out_delete_evlist;
+	}
+
+	err = perf_evlist__mmap(evlist, UINT_MAX, false);
+	if (err < 0) {
+		printf("Couldn't mmap the events: %s\n", strerror(errno));
+		goto out_delete_evlist;
+	}
+
+	perf_evlist__enable(evlist);
+again:
+	before = nr_events;
+
+	for (i = 0; i < evlist->nr_mmaps; i++) {
+		union perf_event *event;
+
+		while ((event = perf_evlist__mmap_read(evlist, i)) != NULL) {
+			const u32 type = event->header.type;
+			struct syscall *sc;
+			struct perf_sample sample;
+			int id;
+
+			++nr_events;
+
+			switch (type) {
+			case PERF_RECORD_SAMPLE:
+				break;
+			case PERF_RECORD_LOST:
+				printf("LOST %" PRIu64 " events!\n", event->lost.lost);
+				continue;
+			default:
+				printf("Unexpected %s event, skipping...\n",
+					perf_event__name(type));
+				continue;
+			}
+
+			err = perf_evlist__parse_sample(evlist, event, &sample);
+			if (err) {
+				printf("Can't parse sample, err = %d, skipping...\n", err);
+				continue;
+			}
+
+			evsel = perf_evlist__id2evsel(evlist, sample.id);
+			if (evsel == NULL) {
+				printf("Unknown tp ID %" PRIu64 ", skipping...\n", sample.id);
+				continue;
+			}
+
+			id = perf_evsel__intval(evsel, &sample, "id");
+			if (id < 0) {
+				printf("Invalid syscall %d id, skipping...\n", id);
+				continue;
+			}
+
+			if ((id > trace->syscalls.max || trace->syscalls.table[id].name == NULL) &&
+			    trace__read_syscall_info(trace, id))
+				continue;
+
+			if ((id > trace->syscalls.max || trace->syscalls.table[id].name == NULL))
+				continue;
+
+			sc = &trace->syscalls.table[id];
+
+			if (evlist->threads->map[0] == -1 || evlist->threads->nr > 1)
+				printf("%d ", sample.tid);
+
+			if (evsel == evsel_enter) {
+				void *args = perf_evsel__rawptr(evsel, &sample, "args");
+
+				printf("%s(", sc->name);
+				syscall__fprintf_args(sc, args, stdout);
+			} else if (evsel == evsel_exit) {
+				int ret = perf_evsel__intval(evsel, &sample, "ret");
+
+				if (ret < 0 && sc->fmt && sc->fmt->errmsg) {
+					char bf[256];
+					const char *emsg = strerror_r(-ret, bf, sizeof(bf)),
+						   *e = audit_errno_to_name(-ret);
+
+					printf(") = -1 %s %s", e, emsg);
+				} else if (ret == 0 && sc->fmt && sc->fmt->timeout)
+					printf(") = 0 Timeout");
+				else
+					printf(") = %d", ret);
+
+				putchar('\n');
+			}
+		}
+	}
+
+	if (nr_events == before)
+		poll(evlist->pollfd, evlist->nr_fds, -1);
+
+	goto again;
+
+out_delete_evlist:
+	perf_evlist__delete(evlist);
+out:
+	return err;
+}
+
+int cmd_trace(int argc, const char **argv, const char *prefix __maybe_unused)
+{
+	const char * const trace_usage[] = {
+		"perf trace [<options>]",
+		NULL
+	};
+	struct trace trace = {
+		.audit_machine = audit_detect_machine(),
+		.syscalls = {
+			. max = -1,
+		},
+		.opts = {
+			.target = {
+				.uid	   = UINT_MAX,
+				.uses_mmap = true,
+			},
+			.user_freq     = UINT_MAX,
+			.user_interval = ULLONG_MAX,
+			.no_delay      = true,
+			.mmap_pages    = 1024,
+		},
+	};
+	const struct option trace_options[] = {
+	OPT_STRING('p', "pid", &trace.opts.target.pid, "pid",
+		    "trace events on existing process id"),
+	OPT_STRING(0, "tid", &trace.opts.target.tid, "tid",
+		    "trace events on existing thread id"),
+	OPT_BOOLEAN(0, "all-cpus", &trace.opts.target.system_wide,
+		    "system-wide collection from all CPUs"),
+	OPT_STRING(0, "cpu", &trace.opts.target.cpu_list, "cpu",
+		    "list of cpus to monitor"),
+	OPT_BOOLEAN(0, "no-inherit", &trace.opts.no_inherit,
+		    "child tasks do not inherit counters"),
+	OPT_UINTEGER(0, "mmap-pages", &trace.opts.mmap_pages,
+		     "number of mmap data pages"),
+	OPT_STRING(0, "uid", &trace.opts.target.uid_str, "user",
+		   "user to profile"),
+	OPT_END()
+	};
+	int err;
+
+	argc = parse_options(argc, argv, trace_options, trace_usage, 0);
+	if (argc)
+		usage_with_options(trace_usage, trace_options);
+
+	err = perf_target__parse_uid(&trace.opts.target);
+	if (err) {
+		char bf[BUFSIZ];
+		perf_target__strerror(&trace.opts.target, err, bf, sizeof(bf));
+		printf("%s", bf);
+		return err;
+	}
+
+	return trace__run(&trace);
+}
diff --git a/tools/perf/builtin.h b/tools/perf/builtin.h
index 3ea74ed1b26..08143bd854c 100644
--- a/tools/perf/builtin.h
+++ b/tools/perf/builtin.h
@@ -34,6 +34,7 @@ extern int cmd_kmem(int argc, const char **argv, const char *prefix);
 extern int cmd_lock(int argc, const char **argv, const char *prefix);
 extern int cmd_kvm(int argc, const char **argv, const char *prefix);
 extern int cmd_test(int argc, const char **argv, const char *prefix);
+extern int cmd_trace(int argc, const char **argv, const char *prefix);
 extern int cmd_inject(int argc, const char **argv, const char *prefix);
 
 extern int find_scripts(char **scripts_array, char **scripts_path_array);
diff --git a/tools/perf/command-list.txt b/tools/perf/command-list.txt
index 0303ec69227..3e86bbd8c2d 100644
--- a/tools/perf/command-list.txt
+++ b/tools/perf/command-list.txt
@@ -17,6 +17,7 @@ perf-report			mainporcelain common
 perf-stat			mainporcelain common
 perf-timechart			mainporcelain common
 perf-top			mainporcelain common
+perf-trace			mainporcelain common
 perf-script			mainporcelain common
 perf-probe			mainporcelain full
 perf-kmem			mainporcelain common
diff --git a/tools/perf/perf.c b/tools/perf/perf.c
index fb8578cfa03..3fb052c9a27 100644
--- a/tools/perf/perf.c
+++ b/tools/perf/perf.c
@@ -55,6 +55,7 @@ static struct cmd_struct commands[] = {
 	{ "lock",	cmd_lock,	0 },
 	{ "kvm",	cmd_kvm,	0 },
 	{ "test",	cmd_test,	0 },
+	{ "trace",	cmd_trace,	0 },
 	{ "inject",	cmd_inject,	0 },
 };