Skip to content

Commit

Permalink
perf trace: New tool
Browse files Browse the repository at this point in the history
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>
  • Loading branch information
acmel committed Sep 26, 2012
1 parent 201b733 commit 514f1c6
Show file tree
Hide file tree
Showing 6 changed files with 358 additions and 1 deletion.
53 changes: 53 additions & 0 deletions 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]
3 changes: 2 additions & 1 deletion tools/perf/Makefile
Expand Up @@ -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
Expand Down Expand Up @@ -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)
Expand Down
300 changes: 300 additions & 0 deletions 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);
}
1 change: 1 addition & 0 deletions tools/perf/builtin.h
Expand Up @@ -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);
Expand Down
1 change: 1 addition & 0 deletions tools/perf/command-list.txt
Expand Up @@ -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
Expand Down
1 change: 1 addition & 0 deletions tools/perf/perf.c
Expand Up @@ -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 },
};

Expand Down

0 comments on commit 514f1c6

Please sign in to comment.