linux_dsm_epyc7002/tools/perf/util/evsel.c

778 lines
17 KiB
C
Raw Normal View History

/*
* Copyright (C) 2011, Red Hat Inc, Arnaldo Carvalho de Melo <acme@redhat.com>
*
* Parts came from builtin-{top,stat,record}.c, see those files for further
* copyright notes.
*
* Released under the GPL v2. (and only v2, not any later version)
*/
perf tool: Fix endianness handling of u32 data in samples Currently, analyzing PPC data files on x86 the cpu field is always 0 and the tid and pid are backwards. For example, analyzing a PPC file on PPC the pid/tid fields show: rsyslogd 1210/1212 and analyzing the same PPC file using an x86 perf binary shows: rsyslogd 1212/1210 The problem is that the swap_op method for samples is perf_event__all64_swap which assumes all elements in the sample_data struct are u64s. cpu, tid and pid are u32s and need to be handled individually. Given that the swap is done before the sample is parsed, the simplest solution is to undo the 64-bit swap of those elements when the sample is parsed and do the proper swap. The RAW data field is generic and perf cannot have programmatic knowledge of how to treat that data. Instead a warning is given to the user. Thanks to Anton Blanchard for providing a data file for a mult-CPU PPC system so I could verify the fix for the CPU fields. v3 -> v4: - fixed use of WARN_ONCE v2 -> v3: - used WARN_ONCE for message regarding raw data - removed struct wrapper around union - fixed whitespace issues v1 -> v2: - added a union for undoing the byte-swap on u64 and redoing swap on u32's to address compiler errors (see git commit 65014ab3) Cc: Anton Blanchard <anton@samba.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/1315321946-16993-1-git-send-email-dsahern@gmail.com Signed-off-by: David Ahern <dsahern@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-09-06 22:12:26 +07:00
#include <byteswap.h>
#include "asm/bug.h"
#include "evsel.h"
#include "evlist.h"
#include "util.h"
#include "cpumap.h"
#include "thread_map.h"
#include "target.h"
#include "../../include/linux/perf_event.h"
#define FD(e, x, y) (*(int *)xyarray__entry(e->fd, x, y))
#define GROUP_FD(group_fd, cpu) (*(int *)xyarray__entry(group_fd, cpu, 0))
int __perf_evsel__sample_size(u64 sample_type)
{
u64 mask = sample_type & PERF_SAMPLE_MASK;
int size = 0;
int i;
for (i = 0; i < 64; i++) {
if (mask & (1ULL << i))
size++;
}
size *= sizeof(u64);
return size;
}
void hists__init(struct hists *hists)
{
memset(hists, 0, sizeof(*hists));
hists->entries_in_array[0] = hists->entries_in_array[1] = RB_ROOT;
hists->entries_in = &hists->entries_in_array[0];
hists->entries_collapsed = RB_ROOT;
hists->entries = RB_ROOT;
pthread_mutex_init(&hists->lock, NULL);
}
void perf_evsel__init(struct perf_evsel *evsel,
struct perf_event_attr *attr, int idx)
{
evsel->idx = idx;
evsel->attr = *attr;
INIT_LIST_HEAD(&evsel->node);
hists__init(&evsel->hists);
}
struct perf_evsel *perf_evsel__new(struct perf_event_attr *attr, int idx)
{
struct perf_evsel *evsel = zalloc(sizeof(*evsel));
if (evsel != NULL)
perf_evsel__init(evsel, attr, idx);
return evsel;
}
static const char *perf_evsel__hw_names[PERF_COUNT_HW_MAX] = {
"cycles",
"instructions",
"cache-references",
"cache-misses",
"branches",
"branch-misses",
"bus-cycles",
"stalled-cycles-frontend",
"stalled-cycles-backend",
"ref-cycles",
};
const char *__perf_evsel__hw_name(u64 config)
{
if (config < PERF_COUNT_HW_MAX && perf_evsel__hw_names[config])
return perf_evsel__hw_names[config];
return "unknown-hardware";
}
static int perf_evsel__hw_name(struct perf_evsel *evsel, char *bf, size_t size)
{
int colon = 0;
struct perf_event_attr *attr = &evsel->attr;
int r = scnprintf(bf, size, "%s", __perf_evsel__hw_name(attr->config));
bool exclude_guest_default = false;
#define MOD_PRINT(context, mod) do { \
if (!attr->exclude_##context) { \
if (!colon) colon = r++; \
r += scnprintf(bf + r, size - r, "%c", mod); \
} } while(0)
if (attr->exclude_kernel || attr->exclude_user || attr->exclude_hv) {
MOD_PRINT(kernel, 'k');
MOD_PRINT(user, 'u');
MOD_PRINT(hv, 'h');
exclude_guest_default = true;
}
if (attr->precise_ip) {
if (!colon)
colon = r++;
r += scnprintf(bf + r, size - r, "%.*s", attr->precise_ip, "ppp");
exclude_guest_default = true;
}
if (attr->exclude_host || attr->exclude_guest == exclude_guest_default) {
MOD_PRINT(host, 'H');
MOD_PRINT(guest, 'G');
}
#undef MOD_PRINT
if (colon)
bf[colon] = ':';
return r;
}
int perf_evsel__name(struct perf_evsel *evsel, char *bf, size_t size)
{
int ret;
switch (evsel->attr.type) {
case PERF_TYPE_RAW:
ret = scnprintf(bf, size, "raw 0x%" PRIx64, evsel->attr.config);
break;
case PERF_TYPE_HARDWARE:
ret = perf_evsel__hw_name(evsel, bf, size);
break;
default:
/*
* FIXME
*
* This is the minimal perf_evsel__name so that we can
* reconstruct event names taking into account event modifiers.
*
* The old event_name uses it now for raw anr hw events, so that
* we don't drag all the parsing stuff into the python binding.
*
* On the next devel cycle the rest of the event naming will be
* brought here.
*/
return 0;
}
return ret;
}
void perf_evsel__config(struct perf_evsel *evsel, struct perf_record_opts *opts,
struct perf_evsel *first)
{
struct perf_event_attr *attr = &evsel->attr;
int track = !evsel->idx; /* only the first counter needs these */
attr->disabled = 1;
attr->sample_id_all = opts->sample_id_all_missing ? 0 : 1;
attr->inherit = !opts->no_inherit;
attr->read_format = PERF_FORMAT_TOTAL_TIME_ENABLED |
PERF_FORMAT_TOTAL_TIME_RUNNING |
PERF_FORMAT_ID;
attr->sample_type |= PERF_SAMPLE_IP | PERF_SAMPLE_TID;
/*
* We default some events to a 1 default interval. But keep
* it a weak assumption overridable by the user.
*/
if (!attr->sample_period || (opts->user_freq != UINT_MAX &&
opts->user_interval != ULLONG_MAX)) {
if (opts->freq) {
attr->sample_type |= PERF_SAMPLE_PERIOD;
attr->freq = 1;
attr->sample_freq = opts->freq;
} else {
attr->sample_period = opts->default_interval;
}
}
if (opts->no_samples)
attr->sample_freq = 0;
if (opts->inherit_stat)
attr->inherit_stat = 1;
if (opts->sample_address) {
attr->sample_type |= PERF_SAMPLE_ADDR;
attr->mmap_data = track;
}
if (opts->call_graph)
attr->sample_type |= PERF_SAMPLE_CALLCHAIN;
if (perf_target__has_cpu(&opts->target))
attr->sample_type |= PERF_SAMPLE_CPU;
if (opts->period)
attr->sample_type |= PERF_SAMPLE_PERIOD;
if (!opts->sample_id_all_missing &&
(opts->sample_time || !opts->no_inherit ||
perf_target__has_cpu(&opts->target)))
attr->sample_type |= PERF_SAMPLE_TIME;
if (opts->raw_samples) {
attr->sample_type |= PERF_SAMPLE_TIME;
attr->sample_type |= PERF_SAMPLE_RAW;
attr->sample_type |= PERF_SAMPLE_CPU;
}
if (opts->no_delay) {
attr->watermark = 0;
attr->wakeup_events = 1;
}
if (opts->branch_stack) {
attr->sample_type |= PERF_SAMPLE_BRANCH_STACK;
attr->branch_sample_type = opts->branch_stack;
}
attr->mmap = track;
attr->comm = track;
if (perf_target__none(&opts->target) &&
(!opts->group || evsel == first)) {
attr->enable_on_exec = 1;
}
}
int perf_evsel__alloc_fd(struct perf_evsel *evsel, int ncpus, int nthreads)
{
int cpu, thread;
evsel->fd = xyarray__new(ncpus, nthreads, sizeof(int));
if (evsel->fd) {
for (cpu = 0; cpu < ncpus; cpu++) {
for (thread = 0; thread < nthreads; thread++) {
FD(evsel, cpu, thread) = -1;
}
}
}
return evsel->fd != NULL ? 0 : -ENOMEM;
}
int perf_evsel__alloc_id(struct perf_evsel *evsel, int ncpus, int nthreads)
{
evsel->sample_id = xyarray__new(ncpus, nthreads, sizeof(struct perf_sample_id));
if (evsel->sample_id == NULL)
return -ENOMEM;
evsel->id = zalloc(ncpus * nthreads * sizeof(u64));
if (evsel->id == NULL) {
xyarray__delete(evsel->sample_id);
evsel->sample_id = NULL;
return -ENOMEM;
}
return 0;
}
int perf_evsel__alloc_counts(struct perf_evsel *evsel, int ncpus)
{
evsel->counts = zalloc((sizeof(*evsel->counts) +
(ncpus * sizeof(struct perf_counts_values))));
return evsel->counts != NULL ? 0 : -ENOMEM;
}
void perf_evsel__free_fd(struct perf_evsel *evsel)
{
xyarray__delete(evsel->fd);
evsel->fd = NULL;
}
void perf_evsel__free_id(struct perf_evsel *evsel)
{
xyarray__delete(evsel->sample_id);
evsel->sample_id = NULL;
free(evsel->id);
evsel->id = NULL;
}
void perf_evsel__close_fd(struct perf_evsel *evsel, int ncpus, int nthreads)
{
int cpu, thread;
for (cpu = 0; cpu < ncpus; cpu++)
for (thread = 0; thread < nthreads; ++thread) {
close(FD(evsel, cpu, thread));
FD(evsel, cpu, thread) = -1;
}
}
void perf_evsel__exit(struct perf_evsel *evsel)
{
assert(list_empty(&evsel->node));
xyarray__delete(evsel->fd);
xyarray__delete(evsel->sample_id);
free(evsel->id);
}
void perf_evsel__delete(struct perf_evsel *evsel)
{
perf_evsel__exit(evsel);
close_cgroup(evsel->cgrp);
free(evsel->name);
free(evsel);
}
int __perf_evsel__read_on_cpu(struct perf_evsel *evsel,
int cpu, int thread, bool scale)
{
struct perf_counts_values count;
size_t nv = scale ? 3 : 1;
if (FD(evsel, cpu, thread) < 0)
return -EINVAL;
if (evsel->counts == NULL && perf_evsel__alloc_counts(evsel, cpu + 1) < 0)
return -ENOMEM;
if (readn(FD(evsel, cpu, thread), &count, nv * sizeof(u64)) < 0)
return -errno;
if (scale) {
if (count.run == 0)
count.val = 0;
else if (count.run < count.ena)
count.val = (u64)((double)count.val * count.ena / count.run + 0.5);
} else
count.ena = count.run = 0;
evsel->counts->cpu[cpu] = count;
return 0;
}
int __perf_evsel__read(struct perf_evsel *evsel,
int ncpus, int nthreads, bool scale)
{
size_t nv = scale ? 3 : 1;
int cpu, thread;
struct perf_counts_values *aggr = &evsel->counts->aggr, count;
aggr->val = aggr->ena = aggr->run = 0;
for (cpu = 0; cpu < ncpus; cpu++) {
for (thread = 0; thread < nthreads; thread++) {
if (FD(evsel, cpu, thread) < 0)
continue;
if (readn(FD(evsel, cpu, thread),
&count, nv * sizeof(u64)) < 0)
return -errno;
aggr->val += count.val;
if (scale) {
aggr->ena += count.ena;
aggr->run += count.run;
}
}
}
evsel->counts->scaled = 0;
if (scale) {
if (aggr->run == 0) {
evsel->counts->scaled = -1;
aggr->val = 0;
return 0;
}
if (aggr->run < aggr->ena) {
evsel->counts->scaled = 1;
aggr->val = (u64)((double)aggr->val * aggr->ena / aggr->run + 0.5);
}
} else
aggr->ena = aggr->run = 0;
return 0;
}
static int __perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus,
struct thread_map *threads, bool group,
struct xyarray *group_fds)
{
int cpu, thread;
unsigned long flags = 0;
int pid = -1, err;
if (evsel->fd == NULL &&
perf_evsel__alloc_fd(evsel, cpus->nr, threads->nr) < 0)
return -ENOMEM;
if (evsel->cgrp) {
flags = PERF_FLAG_PID_CGROUP;
pid = evsel->cgrp->fd;
}
for (cpu = 0; cpu < cpus->nr; cpu++) {
int group_fd = group_fds ? GROUP_FD(group_fds, cpu) : -1;
for (thread = 0; thread < threads->nr; thread++) {
if (!evsel->cgrp)
pid = threads->map[thread];
FD(evsel, cpu, thread) = sys_perf_event_open(&evsel->attr,
pid,
cpus->map[cpu],
group_fd, flags);
if (FD(evsel, cpu, thread) < 0) {
err = -errno;
goto out_close;
}
if (group && group_fd == -1)
group_fd = FD(evsel, cpu, thread);
}
}
return 0;
out_close:
do {
while (--thread >= 0) {
close(FD(evsel, cpu, thread));
FD(evsel, cpu, thread) = -1;
}
thread = threads->nr;
} while (--cpu >= 0);
return err;
}
void perf_evsel__close(struct perf_evsel *evsel, int ncpus, int nthreads)
{
if (evsel->fd == NULL)
return;
perf_evsel__close_fd(evsel, ncpus, nthreads);
perf_evsel__free_fd(evsel);
evsel->fd = NULL;
}
static struct {
struct cpu_map map;
int cpus[1];
} empty_cpu_map = {
.map.nr = 1,
.cpus = { -1, },
};
static struct {
struct thread_map map;
int threads[1];
} empty_thread_map = {
.map.nr = 1,
.threads = { -1, },
};
int perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus,
struct thread_map *threads, bool group,
struct xyarray *group_fd)
{
if (cpus == NULL) {
/* Work around old compiler warnings about strict aliasing */
cpus = &empty_cpu_map.map;
}
if (threads == NULL)
threads = &empty_thread_map.map;
return __perf_evsel__open(evsel, cpus, threads, group, group_fd);
}
int perf_evsel__open_per_cpu(struct perf_evsel *evsel,
struct cpu_map *cpus, bool group,
struct xyarray *group_fd)
{
return __perf_evsel__open(evsel, cpus, &empty_thread_map.map, group,
group_fd);
}
int perf_evsel__open_per_thread(struct perf_evsel *evsel,
struct thread_map *threads, bool group,
struct xyarray *group_fd)
{
return __perf_evsel__open(evsel, &empty_cpu_map.map, threads, group,
group_fd);
}
static int perf_event__parse_id_sample(const union perf_event *event, u64 type,
perf evsel: Fix 32 bit values endianity swap for sample_id_all header We swap the sample_id_all header by u64 pointers. Some members of the header happen to be 32 bit values. We need to handle them separatelly. Together with other endianity patches, this change fixies perf report discrepancies on origin and target systems as described in test 1 below, e.g. following perf report diff: ... 0.12% ps [kernel.kallsyms] [k] clear_page - 0.12% awk bash [.] alloc_word_desc + 0.12% awk bash [.] yyparse 0.11% beah-rhts-task libpython2.6.so.1.0 [.] 0x5560e 0.10% perf libc-2.12.so [.] __ctype_toupper_loc - 0.09% rhts-test-runne bash [.] maybe_make_export_env + 0.09% rhts-test-runne bash [.] 0x385a0 0.09% ps [kernel.kallsyms] [k] page_fault ... Note, running following to test perf endianity handling: test 1) - origin system: # perf record -a -- sleep 10 (any perf record will do) # perf report > report.origin # perf archive perf.data - copy the perf.data, report.origin and perf.data.tar.bz2 to a target system and run: # tar xjvf perf.data.tar.bz2 -C ~/.debug # perf report > report.target # diff -u report.origin report.target - the diff should produce no output (besides some white space stuff and possibly different date/TZ output) test 2) - origin system: # perf record -ag -fo /tmp/perf.data -- sleep 1 - mount origin system root to the target system on /mnt/origin - target system: # perf script --symfs /mnt/origin -I -i /mnt/origin/tmp/perf.data \ --kallsyms /mnt/origin/proc/kallsyms - complete perf.data header is displayed Signed-off-by: Jiri Olsa <jolsa@redhat.com> Reviewed-by: David Ahern <dsahern@gmail.com> Tested-by: David Ahern <dsahern@gmail.com> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1338380624-7443-4-git-send-email-jolsa@redhat.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-05-30 19:23:44 +07:00
struct perf_sample *sample,
bool swapped)
{
const u64 *array = event->sample.array;
perf evsel: Fix 32 bit values endianity swap for sample_id_all header We swap the sample_id_all header by u64 pointers. Some members of the header happen to be 32 bit values. We need to handle them separatelly. Together with other endianity patches, this change fixies perf report discrepancies on origin and target systems as described in test 1 below, e.g. following perf report diff: ... 0.12% ps [kernel.kallsyms] [k] clear_page - 0.12% awk bash [.] alloc_word_desc + 0.12% awk bash [.] yyparse 0.11% beah-rhts-task libpython2.6.so.1.0 [.] 0x5560e 0.10% perf libc-2.12.so [.] __ctype_toupper_loc - 0.09% rhts-test-runne bash [.] maybe_make_export_env + 0.09% rhts-test-runne bash [.] 0x385a0 0.09% ps [kernel.kallsyms] [k] page_fault ... Note, running following to test perf endianity handling: test 1) - origin system: # perf record -a -- sleep 10 (any perf record will do) # perf report > report.origin # perf archive perf.data - copy the perf.data, report.origin and perf.data.tar.bz2 to a target system and run: # tar xjvf perf.data.tar.bz2 -C ~/.debug # perf report > report.target # diff -u report.origin report.target - the diff should produce no output (besides some white space stuff and possibly different date/TZ output) test 2) - origin system: # perf record -ag -fo /tmp/perf.data -- sleep 1 - mount origin system root to the target system on /mnt/origin - target system: # perf script --symfs /mnt/origin -I -i /mnt/origin/tmp/perf.data \ --kallsyms /mnt/origin/proc/kallsyms - complete perf.data header is displayed Signed-off-by: Jiri Olsa <jolsa@redhat.com> Reviewed-by: David Ahern <dsahern@gmail.com> Tested-by: David Ahern <dsahern@gmail.com> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1338380624-7443-4-git-send-email-jolsa@redhat.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-05-30 19:23:44 +07:00
union u64_swap u;
array += ((event->header.size -
sizeof(event->header)) / sizeof(u64)) - 1;
if (type & PERF_SAMPLE_CPU) {
perf evsel: Fix 32 bit values endianity swap for sample_id_all header We swap the sample_id_all header by u64 pointers. Some members of the header happen to be 32 bit values. We need to handle them separatelly. Together with other endianity patches, this change fixies perf report discrepancies on origin and target systems as described in test 1 below, e.g. following perf report diff: ... 0.12% ps [kernel.kallsyms] [k] clear_page - 0.12% awk bash [.] alloc_word_desc + 0.12% awk bash [.] yyparse 0.11% beah-rhts-task libpython2.6.so.1.0 [.] 0x5560e 0.10% perf libc-2.12.so [.] __ctype_toupper_loc - 0.09% rhts-test-runne bash [.] maybe_make_export_env + 0.09% rhts-test-runne bash [.] 0x385a0 0.09% ps [kernel.kallsyms] [k] page_fault ... Note, running following to test perf endianity handling: test 1) - origin system: # perf record -a -- sleep 10 (any perf record will do) # perf report > report.origin # perf archive perf.data - copy the perf.data, report.origin and perf.data.tar.bz2 to a target system and run: # tar xjvf perf.data.tar.bz2 -C ~/.debug # perf report > report.target # diff -u report.origin report.target - the diff should produce no output (besides some white space stuff and possibly different date/TZ output) test 2) - origin system: # perf record -ag -fo /tmp/perf.data -- sleep 1 - mount origin system root to the target system on /mnt/origin - target system: # perf script --symfs /mnt/origin -I -i /mnt/origin/tmp/perf.data \ --kallsyms /mnt/origin/proc/kallsyms - complete perf.data header is displayed Signed-off-by: Jiri Olsa <jolsa@redhat.com> Reviewed-by: David Ahern <dsahern@gmail.com> Tested-by: David Ahern <dsahern@gmail.com> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1338380624-7443-4-git-send-email-jolsa@redhat.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-05-30 19:23:44 +07:00
u.val64 = *array;
if (swapped) {
/* undo swap of u64, then swap on individual u32s */
u.val64 = bswap_64(u.val64);
u.val32[0] = bswap_32(u.val32[0]);
}
sample->cpu = u.val32[0];
array--;
}
if (type & PERF_SAMPLE_STREAM_ID) {
sample->stream_id = *array;
array--;
}
if (type & PERF_SAMPLE_ID) {
sample->id = *array;
array--;
}
if (type & PERF_SAMPLE_TIME) {
sample->time = *array;
array--;
}
if (type & PERF_SAMPLE_TID) {
perf evsel: Fix 32 bit values endianity swap for sample_id_all header We swap the sample_id_all header by u64 pointers. Some members of the header happen to be 32 bit values. We need to handle them separatelly. Together with other endianity patches, this change fixies perf report discrepancies on origin and target systems as described in test 1 below, e.g. following perf report diff: ... 0.12% ps [kernel.kallsyms] [k] clear_page - 0.12% awk bash [.] alloc_word_desc + 0.12% awk bash [.] yyparse 0.11% beah-rhts-task libpython2.6.so.1.0 [.] 0x5560e 0.10% perf libc-2.12.so [.] __ctype_toupper_loc - 0.09% rhts-test-runne bash [.] maybe_make_export_env + 0.09% rhts-test-runne bash [.] 0x385a0 0.09% ps [kernel.kallsyms] [k] page_fault ... Note, running following to test perf endianity handling: test 1) - origin system: # perf record -a -- sleep 10 (any perf record will do) # perf report > report.origin # perf archive perf.data - copy the perf.data, report.origin and perf.data.tar.bz2 to a target system and run: # tar xjvf perf.data.tar.bz2 -C ~/.debug # perf report > report.target # diff -u report.origin report.target - the diff should produce no output (besides some white space stuff and possibly different date/TZ output) test 2) - origin system: # perf record -ag -fo /tmp/perf.data -- sleep 1 - mount origin system root to the target system on /mnt/origin - target system: # perf script --symfs /mnt/origin -I -i /mnt/origin/tmp/perf.data \ --kallsyms /mnt/origin/proc/kallsyms - complete perf.data header is displayed Signed-off-by: Jiri Olsa <jolsa@redhat.com> Reviewed-by: David Ahern <dsahern@gmail.com> Tested-by: David Ahern <dsahern@gmail.com> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1338380624-7443-4-git-send-email-jolsa@redhat.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-05-30 19:23:44 +07:00
u.val64 = *array;
if (swapped) {
/* undo swap of u64, then swap on individual u32s */
u.val64 = bswap_64(u.val64);
u.val32[0] = bswap_32(u.val32[0]);
u.val32[1] = bswap_32(u.val32[1]);
}
sample->pid = u.val32[0];
sample->tid = u.val32[1];
}
return 0;
}
static bool sample_overlap(const union perf_event *event,
const void *offset, u64 size)
{
const void *base = event;
if (offset + size > base + event->header.size)
return true;
return false;
}
int perf_event__parse_sample(const union perf_event *event, u64 type,
int sample_size, bool sample_id_all,
perf tool: Fix endianness handling of u32 data in samples Currently, analyzing PPC data files on x86 the cpu field is always 0 and the tid and pid are backwards. For example, analyzing a PPC file on PPC the pid/tid fields show: rsyslogd 1210/1212 and analyzing the same PPC file using an x86 perf binary shows: rsyslogd 1212/1210 The problem is that the swap_op method for samples is perf_event__all64_swap which assumes all elements in the sample_data struct are u64s. cpu, tid and pid are u32s and need to be handled individually. Given that the swap is done before the sample is parsed, the simplest solution is to undo the 64-bit swap of those elements when the sample is parsed and do the proper swap. The RAW data field is generic and perf cannot have programmatic knowledge of how to treat that data. Instead a warning is given to the user. Thanks to Anton Blanchard for providing a data file for a mult-CPU PPC system so I could verify the fix for the CPU fields. v3 -> v4: - fixed use of WARN_ONCE v2 -> v3: - used WARN_ONCE for message regarding raw data - removed struct wrapper around union - fixed whitespace issues v1 -> v2: - added a union for undoing the byte-swap on u64 and redoing swap on u32's to address compiler errors (see git commit 65014ab3) Cc: Anton Blanchard <anton@samba.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/1315321946-16993-1-git-send-email-dsahern@gmail.com Signed-off-by: David Ahern <dsahern@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-09-06 22:12:26 +07:00
struct perf_sample *data, bool swapped)
{
const u64 *array;
perf tool: Fix endianness handling of u32 data in samples Currently, analyzing PPC data files on x86 the cpu field is always 0 and the tid and pid are backwards. For example, analyzing a PPC file on PPC the pid/tid fields show: rsyslogd 1210/1212 and analyzing the same PPC file using an x86 perf binary shows: rsyslogd 1212/1210 The problem is that the swap_op method for samples is perf_event__all64_swap which assumes all elements in the sample_data struct are u64s. cpu, tid and pid are u32s and need to be handled individually. Given that the swap is done before the sample is parsed, the simplest solution is to undo the 64-bit swap of those elements when the sample is parsed and do the proper swap. The RAW data field is generic and perf cannot have programmatic knowledge of how to treat that data. Instead a warning is given to the user. Thanks to Anton Blanchard for providing a data file for a mult-CPU PPC system so I could verify the fix for the CPU fields. v3 -> v4: - fixed use of WARN_ONCE v2 -> v3: - used WARN_ONCE for message regarding raw data - removed struct wrapper around union - fixed whitespace issues v1 -> v2: - added a union for undoing the byte-swap on u64 and redoing swap on u32's to address compiler errors (see git commit 65014ab3) Cc: Anton Blanchard <anton@samba.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/1315321946-16993-1-git-send-email-dsahern@gmail.com Signed-off-by: David Ahern <dsahern@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-09-06 22:12:26 +07:00
/*
* used for cross-endian analysis. See git commit 65014ab3
* for why this goofiness is needed.
*/
union u64_swap u;
perf tool: Fix endianness handling of u32 data in samples Currently, analyzing PPC data files on x86 the cpu field is always 0 and the tid and pid are backwards. For example, analyzing a PPC file on PPC the pid/tid fields show: rsyslogd 1210/1212 and analyzing the same PPC file using an x86 perf binary shows: rsyslogd 1212/1210 The problem is that the swap_op method for samples is perf_event__all64_swap which assumes all elements in the sample_data struct are u64s. cpu, tid and pid are u32s and need to be handled individually. Given that the swap is done before the sample is parsed, the simplest solution is to undo the 64-bit swap of those elements when the sample is parsed and do the proper swap. The RAW data field is generic and perf cannot have programmatic knowledge of how to treat that data. Instead a warning is given to the user. Thanks to Anton Blanchard for providing a data file for a mult-CPU PPC system so I could verify the fix for the CPU fields. v3 -> v4: - fixed use of WARN_ONCE v2 -> v3: - used WARN_ONCE for message regarding raw data - removed struct wrapper around union - fixed whitespace issues v1 -> v2: - added a union for undoing the byte-swap on u64 and redoing swap on u32's to address compiler errors (see git commit 65014ab3) Cc: Anton Blanchard <anton@samba.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/1315321946-16993-1-git-send-email-dsahern@gmail.com Signed-off-by: David Ahern <dsahern@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-09-06 22:12:26 +07:00
memset(data, 0, sizeof(*data));
data->cpu = data->pid = data->tid = -1;
data->stream_id = data->id = data->time = -1ULL;
perf evsel: Fix an issue where perf report fails to show the proper percentage This patch fixes an issue where perf report shows nan% for certain perf.data files. The below is from a report for a do_fork probe: -nan% sshd [kernel.kallsyms] [k] do_fork -nan% packagekitd [kernel.kallsyms] [k] do_fork -nan% dbus-daemon [kernel.kallsyms] [k] do_fork -nan% bash [kernel.kallsyms] [k] do_fork A git bisect shows commit f3bda2c as the cause. However, looking back through the git history, I saw commit 640c03c which seems to have removed the required initialization for perf_sample->period. The problem only started showing after commit f3bda2c. The below patch re-introduces the initialization and it fixes the problem for me. With the below patch, for the same perf.data: 73.08% bash [kernel.kallsyms] [k] do_fork 8.97% 11-dhclient [kernel.kallsyms] [k] do_fork 6.41% sshd [kernel.kallsyms] [k] do_fork 3.85% 20-chrony [kernel.kallsyms] [k] do_fork 2.56% sendmail [kernel.kallsyms] [k] do_fork This patch applies over current linux-tip commit 9949284. Problem introduced in: $ git describe 640c03c v2.6.37-rc3-83-g640c03c Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Robert Richter <robert.richter@amd.com> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: stable@kernel.org Link: http://lkml.kernel.org/r/20120203170113.5190.25558.stgit@localhost6.localdomain6 Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-02-04 00:01:13 +07:00
data->period = 1;
if (event->header.type != PERF_RECORD_SAMPLE) {
if (!sample_id_all)
return 0;
perf evsel: Fix 32 bit values endianity swap for sample_id_all header We swap the sample_id_all header by u64 pointers. Some members of the header happen to be 32 bit values. We need to handle them separatelly. Together with other endianity patches, this change fixies perf report discrepancies on origin and target systems as described in test 1 below, e.g. following perf report diff: ... 0.12% ps [kernel.kallsyms] [k] clear_page - 0.12% awk bash [.] alloc_word_desc + 0.12% awk bash [.] yyparse 0.11% beah-rhts-task libpython2.6.so.1.0 [.] 0x5560e 0.10% perf libc-2.12.so [.] __ctype_toupper_loc - 0.09% rhts-test-runne bash [.] maybe_make_export_env + 0.09% rhts-test-runne bash [.] 0x385a0 0.09% ps [kernel.kallsyms] [k] page_fault ... Note, running following to test perf endianity handling: test 1) - origin system: # perf record -a -- sleep 10 (any perf record will do) # perf report > report.origin # perf archive perf.data - copy the perf.data, report.origin and perf.data.tar.bz2 to a target system and run: # tar xjvf perf.data.tar.bz2 -C ~/.debug # perf report > report.target # diff -u report.origin report.target - the diff should produce no output (besides some white space stuff and possibly different date/TZ output) test 2) - origin system: # perf record -ag -fo /tmp/perf.data -- sleep 1 - mount origin system root to the target system on /mnt/origin - target system: # perf script --symfs /mnt/origin -I -i /mnt/origin/tmp/perf.data \ --kallsyms /mnt/origin/proc/kallsyms - complete perf.data header is displayed Signed-off-by: Jiri Olsa <jolsa@redhat.com> Reviewed-by: David Ahern <dsahern@gmail.com> Tested-by: David Ahern <dsahern@gmail.com> Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com> Cc: David Ahern <dsahern@gmail.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1338380624-7443-4-git-send-email-jolsa@redhat.com Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2012-05-30 19:23:44 +07:00
return perf_event__parse_id_sample(event, type, data, swapped);
}
array = event->sample.array;
if (sample_size + sizeof(event->header) > event->header.size)
return -EFAULT;
if (type & PERF_SAMPLE_IP) {
data->ip = event->ip.ip;
array++;
}
if (type & PERF_SAMPLE_TID) {
perf tool: Fix endianness handling of u32 data in samples Currently, analyzing PPC data files on x86 the cpu field is always 0 and the tid and pid are backwards. For example, analyzing a PPC file on PPC the pid/tid fields show: rsyslogd 1210/1212 and analyzing the same PPC file using an x86 perf binary shows: rsyslogd 1212/1210 The problem is that the swap_op method for samples is perf_event__all64_swap which assumes all elements in the sample_data struct are u64s. cpu, tid and pid are u32s and need to be handled individually. Given that the swap is done before the sample is parsed, the simplest solution is to undo the 64-bit swap of those elements when the sample is parsed and do the proper swap. The RAW data field is generic and perf cannot have programmatic knowledge of how to treat that data. Instead a warning is given to the user. Thanks to Anton Blanchard for providing a data file for a mult-CPU PPC system so I could verify the fix for the CPU fields. v3 -> v4: - fixed use of WARN_ONCE v2 -> v3: - used WARN_ONCE for message regarding raw data - removed struct wrapper around union - fixed whitespace issues v1 -> v2: - added a union for undoing the byte-swap on u64 and redoing swap on u32's to address compiler errors (see git commit 65014ab3) Cc: Anton Blanchard <anton@samba.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/1315321946-16993-1-git-send-email-dsahern@gmail.com Signed-off-by: David Ahern <dsahern@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-09-06 22:12:26 +07:00
u.val64 = *array;
if (swapped) {
/* undo swap of u64, then swap on individual u32s */
u.val64 = bswap_64(u.val64);
u.val32[0] = bswap_32(u.val32[0]);
u.val32[1] = bswap_32(u.val32[1]);
}
data->pid = u.val32[0];
data->tid = u.val32[1];
array++;
}
if (type & PERF_SAMPLE_TIME) {
data->time = *array;
array++;
}
data->addr = 0;
if (type & PERF_SAMPLE_ADDR) {
data->addr = *array;
array++;
}
data->id = -1ULL;
if (type & PERF_SAMPLE_ID) {
data->id = *array;
array++;
}
if (type & PERF_SAMPLE_STREAM_ID) {
data->stream_id = *array;
array++;
}
if (type & PERF_SAMPLE_CPU) {
perf tool: Fix endianness handling of u32 data in samples Currently, analyzing PPC data files on x86 the cpu field is always 0 and the tid and pid are backwards. For example, analyzing a PPC file on PPC the pid/tid fields show: rsyslogd 1210/1212 and analyzing the same PPC file using an x86 perf binary shows: rsyslogd 1212/1210 The problem is that the swap_op method for samples is perf_event__all64_swap which assumes all elements in the sample_data struct are u64s. cpu, tid and pid are u32s and need to be handled individually. Given that the swap is done before the sample is parsed, the simplest solution is to undo the 64-bit swap of those elements when the sample is parsed and do the proper swap. The RAW data field is generic and perf cannot have programmatic knowledge of how to treat that data. Instead a warning is given to the user. Thanks to Anton Blanchard for providing a data file for a mult-CPU PPC system so I could verify the fix for the CPU fields. v3 -> v4: - fixed use of WARN_ONCE v2 -> v3: - used WARN_ONCE for message regarding raw data - removed struct wrapper around union - fixed whitespace issues v1 -> v2: - added a union for undoing the byte-swap on u64 and redoing swap on u32's to address compiler errors (see git commit 65014ab3) Cc: Anton Blanchard <anton@samba.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/1315321946-16993-1-git-send-email-dsahern@gmail.com Signed-off-by: David Ahern <dsahern@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-09-06 22:12:26 +07:00
u.val64 = *array;
if (swapped) {
/* undo swap of u64, then swap on individual u32s */
u.val64 = bswap_64(u.val64);
u.val32[0] = bswap_32(u.val32[0]);
}
data->cpu = u.val32[0];
array++;
}
if (type & PERF_SAMPLE_PERIOD) {
data->period = *array;
array++;
}
if (type & PERF_SAMPLE_READ) {
fprintf(stderr, "PERF_SAMPLE_READ is unsupported for now\n");
return -1;
}
if (type & PERF_SAMPLE_CALLCHAIN) {
if (sample_overlap(event, array, sizeof(data->callchain->nr)))
return -EFAULT;
data->callchain = (struct ip_callchain *)array;
if (sample_overlap(event, array, data->callchain->nr))
return -EFAULT;
array += 1 + data->callchain->nr;
}
if (type & PERF_SAMPLE_RAW) {
const u64 *pdata;
perf tool: Fix endianness handling of u32 data in samples Currently, analyzing PPC data files on x86 the cpu field is always 0 and the tid and pid are backwards. For example, analyzing a PPC file on PPC the pid/tid fields show: rsyslogd 1210/1212 and analyzing the same PPC file using an x86 perf binary shows: rsyslogd 1212/1210 The problem is that the swap_op method for samples is perf_event__all64_swap which assumes all elements in the sample_data struct are u64s. cpu, tid and pid are u32s and need to be handled individually. Given that the swap is done before the sample is parsed, the simplest solution is to undo the 64-bit swap of those elements when the sample is parsed and do the proper swap. The RAW data field is generic and perf cannot have programmatic knowledge of how to treat that data. Instead a warning is given to the user. Thanks to Anton Blanchard for providing a data file for a mult-CPU PPC system so I could verify the fix for the CPU fields. v3 -> v4: - fixed use of WARN_ONCE v2 -> v3: - used WARN_ONCE for message regarding raw data - removed struct wrapper around union - fixed whitespace issues v1 -> v2: - added a union for undoing the byte-swap on u64 and redoing swap on u32's to address compiler errors (see git commit 65014ab3) Cc: Anton Blanchard <anton@samba.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/1315321946-16993-1-git-send-email-dsahern@gmail.com Signed-off-by: David Ahern <dsahern@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-09-06 22:12:26 +07:00
u.val64 = *array;
if (WARN_ONCE(swapped,
"Endianness of raw data not corrected!\n")) {
/* undo swap of u64, then swap on individual u32s */
u.val64 = bswap_64(u.val64);
u.val32[0] = bswap_32(u.val32[0]);
u.val32[1] = bswap_32(u.val32[1]);
}
if (sample_overlap(event, array, sizeof(u32)))
return -EFAULT;
perf tool: Fix endianness handling of u32 data in samples Currently, analyzing PPC data files on x86 the cpu field is always 0 and the tid and pid are backwards. For example, analyzing a PPC file on PPC the pid/tid fields show: rsyslogd 1210/1212 and analyzing the same PPC file using an x86 perf binary shows: rsyslogd 1212/1210 The problem is that the swap_op method for samples is perf_event__all64_swap which assumes all elements in the sample_data struct are u64s. cpu, tid and pid are u32s and need to be handled individually. Given that the swap is done before the sample is parsed, the simplest solution is to undo the 64-bit swap of those elements when the sample is parsed and do the proper swap. The RAW data field is generic and perf cannot have programmatic knowledge of how to treat that data. Instead a warning is given to the user. Thanks to Anton Blanchard for providing a data file for a mult-CPU PPC system so I could verify the fix for the CPU fields. v3 -> v4: - fixed use of WARN_ONCE v2 -> v3: - used WARN_ONCE for message regarding raw data - removed struct wrapper around union - fixed whitespace issues v1 -> v2: - added a union for undoing the byte-swap on u64 and redoing swap on u32's to address compiler errors (see git commit 65014ab3) Cc: Anton Blanchard <anton@samba.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/1315321946-16993-1-git-send-email-dsahern@gmail.com Signed-off-by: David Ahern <dsahern@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-09-06 22:12:26 +07:00
data->raw_size = u.val32[0];
pdata = (void *) array + sizeof(u32);
if (sample_overlap(event, pdata, data->raw_size))
return -EFAULT;
data->raw_data = (void *) pdata;
array = (void *)array + data->raw_size + sizeof(u32);
}
if (type & PERF_SAMPLE_BRANCH_STACK) {
u64 sz;
data->branch_stack = (struct branch_stack *)array;
array++; /* nr */
sz = data->branch_stack->nr * sizeof(struct branch_entry);
sz /= sizeof(u64);
array += sz;
}
return 0;
}
int perf_event__synthesize_sample(union perf_event *event, u64 type,
const struct perf_sample *sample,
bool swapped)
{
u64 *array;
/*
* used for cross-endian analysis. See git commit 65014ab3
* for why this goofiness is needed.
*/
union u64_swap u;
array = event->sample.array;
if (type & PERF_SAMPLE_IP) {
event->ip.ip = sample->ip;
array++;
}
if (type & PERF_SAMPLE_TID) {
u.val32[0] = sample->pid;
u.val32[1] = sample->tid;
if (swapped) {
/*
* Inverse of what is done in perf_event__parse_sample
*/
u.val32[0] = bswap_32(u.val32[0]);
u.val32[1] = bswap_32(u.val32[1]);
u.val64 = bswap_64(u.val64);
}
*array = u.val64;
array++;
}
if (type & PERF_SAMPLE_TIME) {
*array = sample->time;
array++;
}
if (type & PERF_SAMPLE_ADDR) {
*array = sample->addr;
array++;
}
if (type & PERF_SAMPLE_ID) {
*array = sample->id;
array++;
}
if (type & PERF_SAMPLE_STREAM_ID) {
*array = sample->stream_id;
array++;
}
if (type & PERF_SAMPLE_CPU) {
u.val32[0] = sample->cpu;
if (swapped) {
/*
* Inverse of what is done in perf_event__parse_sample
*/
u.val32[0] = bswap_32(u.val32[0]);
u.val64 = bswap_64(u.val64);
}
*array = u.val64;
array++;
}
if (type & PERF_SAMPLE_PERIOD) {
*array = sample->period;
array++;
}
return 0;
}