linux/tools/perf/util/session.c

928 lines
23 KiB
C
Raw Normal View History

#define _FILE_OFFSET_BITS 64
#include <linux/kernel.h>
perf tools: Cross platform perf.data analysis support There are still some problems related to loading vmlinux files, but those are unrelated to the feature implemented in this patch, so will get fixed in the next patches, but here are some results: 1. collect perf.data file on a Fedora 12 machine, x86_64, 64-bit userland 2. transfer it to a Debian Testing machine, PARISC64, 32-bit userland acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | head -5 74f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms] 55fdd56670453ea66c011158c4b9d30179c1d049 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/ipt_MASQUERADE.ko 41adff63c730890480980d5d8ba513f1c216a858 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/iptable_nat.ko 90a33def1077bb8e97b8a78546dc96c2de62df46 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/nf_nat.ko 984c7bea90ce1376d5c8e7ef43a781801286e62d /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/tun.ko acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | tail -5 22492f3753c6a67de5c7ccbd6b863390c92c0723 /usr/lib64/libXt.so.6.0.0 353802bb7e1b895ba43507cc678f951e778e4c6f /usr/lib64/libMagickCore.so.2.0.0 d10c2897558595efe7be8b0584cf7e6398bc776c /usr/lib64/libfprint.so.0.0.0 a83ecfb519a788774a84d5ddde633c9ba56c03ab /home/acme/bin/perf d3ca765a8ecf257d263801d7ad8c49c189082317 /usr/lib64/libdwarf.so.0.0 acme@parisc:~/git/linux-2.6-tip$ acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm The file [kernel.kallsyms] cannot be used, trying to use /proc/kallsyms... ^^^^ The problem related to vmlinux handling, it shouldn't be trying this ^^^^ rather alien /proc/kallsyms at all... /lib64/libpthread-2.10.2.so with build id 5c68f7afeb33309c78037e374b0deee84dd441f6 not found, continuing without symbols /lib64/libc-2.10.2.so with build id eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 not found, continuing without symbols /home/acme/bin/perf with build id a83ecfb519a788774a84d5ddde633c9ba56c03ab not found, continuing without symbols /usr/sbin/openvpn with build id f2037a091ef36b591187a858d75e203690ea9409 not found, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwlcore.ko, continuing without symbols <SNIP more complaints about not finding the right build-ids, those will have to wait for 'perf archive' or plain copying what was collected by 'perf record' on the x86_64, source machine, see further below for an example of this > # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 acme@parisc:~/git/linux-2.6-tip$ Which matches what we get when running the same command for the same perf.data file on the F12, x86_64, source machine: [root@doppio linux-2.6-tip]# perf report --sort comm # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 [root@doppio linux-2.6-tip]# The other modes work as well, modulo the problem with vmlinux: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object # ........ ............... ................................. # 35.11% find ffffffff81002b5a 18.25% perf ffffffff8102235f 16.17% find libc-2.10.2.so 9.07% find find 5.80% swapper ffffffff8102235f 3.95% perf libc-2.10.2.so 2.33% init ffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k 1.35% find [e1000e] 0.68% sleep libc-2.10.2.so acme@parisc:~/git/linux-2.6-tip$ And the lack of the right buildids: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso,symbol 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object Symbol # ........ ............... ................................. ...... # 35.11% find ffffffff81002b5a [k] 0xffffffff81002b5a 18.25% perf ffffffff8102235f [k] 0xffffffff8102235f 16.17% find libc-2.10.2.so [.] 0x00000000045782 9.07% find find [.] 0x0000000000fb0e 5.80% swapper ffffffff8102235f [k] 0xffffffff8102235f 3.95% perf libc-2.10.2.so [.] 0x0000000007f398 2.33% init ffffffff810091b9 [k] 0xffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k [.] 0x00000000105440 1.35% find [e1000e] [k] 0x00000000010948 0.68% sleep libc-2.10.2.so [.] 0x0000000011ad5b acme@parisc:~/git/linux-2.6-tip$ But if we: acme@parisc:~/git/linux-2.6-tip$ ls ~/.debug ls: cannot access /home/acme/.debug: No such file or directory acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/lib64/libc-2.10.2.so/ acme@parisc:~/git/linux-2.6-tip$ scp doppio:.debug/lib64/libc-2.10.2.so/* ~/.debug/lib64/libc-2.10.2.so/ acme@doppio's password: eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 100% 1783KB 714.7KB/s 00:02 acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/.build-id/eb acme@parisc:~/git/linux-2.6-tip$ ln -s ../../lib64/libc-2.10.2.so/eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 ~/.debug/.build-id/eb/4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 acme@parisc:~/git/linux-2.6-tip$ perf report --dsos libc-2.10.2.so 2> /dev/null # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # acme@parisc:~/git/linux-2.6-tip$ Which matches what we get on the source, F12, x86_64 machine: [root@doppio linux-2.6-tip]# perf report --dsos libc-2.10.2.so # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # [root@doppio linux-2.6-tip]# So I think this is really, really nice in that it demonstrates the portability of perf.data files and the use of build-ids accross such aliens worlds :-) There are some things to fix tho, like the bitmap on the header, but things are looking good. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1263478990-8200-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-14 22:23:10 +08:00
#include <byteswap.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/mman.h>
#include "session.h"
#include "sort.h"
#include "util.h"
static int perf_session__open(struct perf_session *self, bool force)
{
struct stat input_stat;
if (!strcmp(self->filename, "-")) {
self->fd_pipe = true;
self->fd = STDIN_FILENO;
if (perf_header__read(self, self->fd) < 0)
pr_err("incompatible file format");
return 0;
}
self->fd = open(self->filename, O_RDONLY);
if (self->fd < 0) {
int err = errno;
pr_err("failed to open %s: %s", self->filename, strerror(err));
if (err == ENOENT && !strcmp(self->filename, "perf.data"))
pr_err(" (try 'perf record' first)");
pr_err("\n");
return -errno;
}
if (fstat(self->fd, &input_stat) < 0)
goto out_close;
if (!force && input_stat.st_uid && (input_stat.st_uid != geteuid())) {
pr_err("file %s not owned by current user or root\n",
self->filename);
goto out_close;
}
if (!input_stat.st_size) {
pr_info("zero-sized file (%s), nothing to do!\n",
self->filename);
goto out_close;
}
if (perf_header__read(self, self->fd) < 0) {
pr_err("incompatible file format");
goto out_close;
}
self->size = input_stat.st_size;
return 0;
out_close:
close(self->fd);
self->fd = -1;
return -1;
}
void perf_session__update_sample_type(struct perf_session *self)
{
self->sample_type = perf_header__sample_type(&self->header);
}
int perf_session__create_kernel_maps(struct perf_session *self)
{
int ret = machine__create_kernel_maps(&self->host_machine);
if (ret >= 0)
ret = machines__create_guest_kernel_maps(&self->machines);
return ret;
}
static void perf_session__destroy_kernel_maps(struct perf_session *self)
{
machine__destroy_kernel_maps(&self->host_machine);
machines__destroy_guest_kernel_maps(&self->machines);
}
perf: add perf-inject builtin Currently, perf 'live mode' writes build-ids at the end of the session, which isn't actually useful for processing live mode events. What would be better would be to have the build-ids sent before any of the samples that reference them, which can be done by processing the event stream and retrieving the build-ids on the first hit. Doing that in perf-record itself, however, is off-limits. This patch introduces perf-inject, which does the same job while leaving perf-record untouched. Normal mode perf still records the build-ids at the end of the session as it should, but for live mode, perf-inject can be injected in between the record and report steps e.g.: perf record -o - ./hackbench 10 | perf inject -v -b | perf report -v -i - perf-inject reads a perf-record event stream and repipes it to stdout. At any point the processing code can inject other events into the event stream - in this case build-ids (-b option) are read and injected as needed into the event stream. Build-ids are just the first user of perf-inject - potentially anything that needs userspace processing to augment the trace stream with additional information could make use of this facility. Cc: Ingo Molnar <mingo@elte.hu> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Frédéric Weisbecker <fweisbec@gmail.com> LKML-Reference: <1272696080-16435-3-git-send-email-tzanussi@gmail.com> Signed-off-by: Tom Zanussi <tzanussi@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-05-01 14:41:20 +08:00
struct perf_session *perf_session__new(const char *filename, int mode, bool force, bool repipe)
{
size_t len = filename ? strlen(filename) + 1 : 0;
struct perf_session *self = zalloc(sizeof(*self) + len);
if (self == NULL)
goto out;
if (perf_header__init(&self->header) < 0)
goto out_free;
memcpy(self->filename, filename, len);
self->threads = RB_ROOT;
INIT_LIST_HEAD(&self->dead_threads);
2010-05-11 00:04:11 +08:00
self->hists_tree = RB_ROOT;
self->last_match = NULL;
self->mmap_window = 32;
self->machines = RB_ROOT;
perf: add perf-inject builtin Currently, perf 'live mode' writes build-ids at the end of the session, which isn't actually useful for processing live mode events. What would be better would be to have the build-ids sent before any of the samples that reference them, which can be done by processing the event stream and retrieving the build-ids on the first hit. Doing that in perf-record itself, however, is off-limits. This patch introduces perf-inject, which does the same job while leaving perf-record untouched. Normal mode perf still records the build-ids at the end of the session as it should, but for live mode, perf-inject can be injected in between the record and report steps e.g.: perf record -o - ./hackbench 10 | perf inject -v -b | perf report -v -i - perf-inject reads a perf-record event stream and repipes it to stdout. At any point the processing code can inject other events into the event stream - in this case build-ids (-b option) are read and injected as needed into the event stream. Build-ids are just the first user of perf-inject - potentially anything that needs userspace processing to augment the trace stream with additional information could make use of this facility. Cc: Ingo Molnar <mingo@elte.hu> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Frédéric Weisbecker <fweisbec@gmail.com> LKML-Reference: <1272696080-16435-3-git-send-email-tzanussi@gmail.com> Signed-off-by: Tom Zanussi <tzanussi@gmail.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-05-01 14:41:20 +08:00
self->repipe = repipe;
INIT_LIST_HEAD(&self->ordered_samples.samples);
machine__init(&self->host_machine, "", HOST_KERNEL_ID);
if (mode == O_RDONLY) {
if (perf_session__open(self, force) < 0)
goto out_delete;
} else if (mode == O_WRONLY) {
/*
* In O_RDONLY mode this will be performed when reading the
* kernel MMAP event, in event__process_mmap().
*/
if (perf_session__create_kernel_maps(self) < 0)
goto out_delete;
}
perf_session__update_sample_type(self);
out:
return self;
out_free:
free(self);
return NULL;
out_delete:
perf_session__delete(self);
return NULL;
}
static void perf_session__delete_dead_threads(struct perf_session *self)
{
struct thread *n, *t;
list_for_each_entry_safe(t, n, &self->dead_threads, node) {
list_del(&t->node);
thread__delete(t);
}
}
static void perf_session__delete_threads(struct perf_session *self)
{
struct rb_node *nd = rb_first(&self->threads);
while (nd) {
struct thread *t = rb_entry(nd, struct thread, rb_node);
rb_erase(&t->rb_node, &self->threads);
nd = rb_next(nd);
thread__delete(t);
}
}
void perf_session__delete(struct perf_session *self)
{
perf_header__exit(&self->header);
perf_session__destroy_kernel_maps(self);
perf_session__delete_dead_threads(self);
perf_session__delete_threads(self);
machine__exit(&self->host_machine);
close(self->fd);
free(self);
}
void perf_session__remove_thread(struct perf_session *self, struct thread *th)
{
self->last_match = NULL;
rb_erase(&th->rb_node, &self->threads);
/*
* We may have references to this thread, for instance in some hist_entry
* instances, so just move them to a separate list.
*/
list_add_tail(&th->node, &self->dead_threads);
}
static bool symbol__match_parent_regex(struct symbol *sym)
{
if (sym->name && !regexec(&parent_regex, sym->name, 0, NULL, 0))
return 1;
return 0;
}
struct map_symbol *perf_session__resolve_callchain(struct perf_session *self,
struct thread *thread,
struct ip_callchain *chain,
struct symbol **parent)
{
u8 cpumode = PERF_RECORD_MISC_USER;
unsigned int i;
struct map_symbol *syms = calloc(chain->nr, sizeof(*syms));
if (!syms)
return NULL;
for (i = 0; i < chain->nr; i++) {
u64 ip = chain->ips[i];
struct addr_location al;
if (ip >= PERF_CONTEXT_MAX) {
switch (ip) {
case PERF_CONTEXT_HV:
cpumode = PERF_RECORD_MISC_HYPERVISOR; break;
case PERF_CONTEXT_KERNEL:
cpumode = PERF_RECORD_MISC_KERNEL; break;
case PERF_CONTEXT_USER:
cpumode = PERF_RECORD_MISC_USER; break;
default:
break;
}
continue;
}
al.filtered = false;
thread__find_addr_location(thread, self, cpumode,
MAP__FUNCTION, thread->pid, ip, &al, NULL);
if (al.sym != NULL) {
if (sort__has_parent && !*parent &&
symbol__match_parent_regex(al.sym))
*parent = al.sym;
if (!symbol_conf.use_callchain)
break;
syms[i].map = al.map;
syms[i].sym = al.sym;
}
}
return syms;
}
static int process_event_stub(event_t *event __used,
struct perf_session *session __used)
{
dump_printf(": unhandled!\n");
return 0;
}
perf: Provide a new deterministic events reordering algorithm The current events reordering algorithm is based on a heuristic that gets broken once we deal with a very fast flow of events. Indeed the time period based flushing is not suitable anymore in the following case, assuming we have a flush period of two seconds. CPU 0 | CPU 1 | cnt1 timestamps | cnt1 timestamps | 0 | 0 1 | 1 2 | 2 3 | 3 [...] | [...] 4 seconds later If we spend too much time to read the buffers (case of a lot of events to record in each buffers or when we have a lot of CPU buffers to read), in the next pass the CPU 0 buffer could contain a slice of several seconds of events. We'll read them all and notice we've reached the period to flush. In the above example we flush the first half of the CPU 0 buffer, then we read the CPU 1 buffer where we have events that were on the flush slice and then the reordering fails. It's simple to reproduce with: perf lock record perf bench sched messaging To solve this, we use a new solution that doesn't rely on an heuristical time slice period anymore but on a deterministic basis based on how perf record does its job. perf record saves the buffers through passes. A pass is a tour on every buffers from every CPUs. This is made in order: for each CPU we read the buffers of every counters. So the more buffers we visit, the later will be the timstamps of their events. When perf record finishes a pass it records a PERF_RECORD_FINISHED_ROUND pseudo event. We record the max timestamp t found in the pass n. Assuming these timestamps are monotonic across cpus, we know that if a buffer still has events with timestamps below t, they will be all available and then read in the pass n + 1. Hence when we start to read the pass n + 2, we can safely flush every events with timestamps below t. ============ PASS n ================= CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 1 | 2 2 | 3 - | 4 <--- max recorded ============ PASS n + 1 ============== CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 3 | 5 4 | 6 5 | 7 <---- max recorded Flush every events below timestamp 4 ============ PASS n + 2 ============== CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 6 | 8 7 | 9 - | 10 Flush every events below timestamp 7 etc... It also works on perf.data versions that don't have PERF_RECORD_FINISHED_ROUND pseudo events. The difference is that the events will be only flushed in the end of the perf.data processing. It will then consume more memory and scale less with large perf.data files. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Tom Zanussi <tzanussi@gmail.com> Cc: Masami Hiramatsu <mhiramat@redhat.com>
2010-05-03 21:14:33 +08:00
static int process_finished_round_stub(event_t *event __used,
struct perf_session *session __used,
struct perf_event_ops *ops __used)
{
dump_printf(": unhandled!\n");
return 0;
}
static int process_finished_round(event_t *event,
struct perf_session *session,
struct perf_event_ops *ops);
static void perf_event_ops__fill_defaults(struct perf_event_ops *handler)
{
if (handler->sample == NULL)
handler->sample = process_event_stub;
if (handler->mmap == NULL)
handler->mmap = process_event_stub;
if (handler->comm == NULL)
handler->comm = process_event_stub;
if (handler->fork == NULL)
handler->fork = process_event_stub;
if (handler->exit == NULL)
handler->exit = process_event_stub;
if (handler->lost == NULL)
handler->lost = event__process_lost;
if (handler->read == NULL)
handler->read = process_event_stub;
if (handler->throttle == NULL)
handler->throttle = process_event_stub;
if (handler->unthrottle == NULL)
handler->unthrottle = process_event_stub;
if (handler->attr == NULL)
handler->attr = process_event_stub;
if (handler->event_type == NULL)
handler->event_type = process_event_stub;
if (handler->tracing_data == NULL)
handler->tracing_data = process_event_stub;
if (handler->build_id == NULL)
handler->build_id = process_event_stub;
perf: Provide a new deterministic events reordering algorithm The current events reordering algorithm is based on a heuristic that gets broken once we deal with a very fast flow of events. Indeed the time period based flushing is not suitable anymore in the following case, assuming we have a flush period of two seconds. CPU 0 | CPU 1 | cnt1 timestamps | cnt1 timestamps | 0 | 0 1 | 1 2 | 2 3 | 3 [...] | [...] 4 seconds later If we spend too much time to read the buffers (case of a lot of events to record in each buffers or when we have a lot of CPU buffers to read), in the next pass the CPU 0 buffer could contain a slice of several seconds of events. We'll read them all and notice we've reached the period to flush. In the above example we flush the first half of the CPU 0 buffer, then we read the CPU 1 buffer where we have events that were on the flush slice and then the reordering fails. It's simple to reproduce with: perf lock record perf bench sched messaging To solve this, we use a new solution that doesn't rely on an heuristical time slice period anymore but on a deterministic basis based on how perf record does its job. perf record saves the buffers through passes. A pass is a tour on every buffers from every CPUs. This is made in order: for each CPU we read the buffers of every counters. So the more buffers we visit, the later will be the timstamps of their events. When perf record finishes a pass it records a PERF_RECORD_FINISHED_ROUND pseudo event. We record the max timestamp t found in the pass n. Assuming these timestamps are monotonic across cpus, we know that if a buffer still has events with timestamps below t, they will be all available and then read in the pass n + 1. Hence when we start to read the pass n + 2, we can safely flush every events with timestamps below t. ============ PASS n ================= CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 1 | 2 2 | 3 - | 4 <--- max recorded ============ PASS n + 1 ============== CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 3 | 5 4 | 6 5 | 7 <---- max recorded Flush every events below timestamp 4 ============ PASS n + 2 ============== CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 6 | 8 7 | 9 - | 10 Flush every events below timestamp 7 etc... It also works on perf.data versions that don't have PERF_RECORD_FINISHED_ROUND pseudo events. The difference is that the events will be only flushed in the end of the perf.data processing. It will then consume more memory and scale less with large perf.data files. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Tom Zanussi <tzanussi@gmail.com> Cc: Masami Hiramatsu <mhiramat@redhat.com>
2010-05-03 21:14:33 +08:00
if (handler->finished_round == NULL) {
if (handler->ordered_samples)
handler->finished_round = process_finished_round;
else
handler->finished_round = process_finished_round_stub;
}
}
perf tools: Cross platform perf.data analysis support There are still some problems related to loading vmlinux files, but those are unrelated to the feature implemented in this patch, so will get fixed in the next patches, but here are some results: 1. collect perf.data file on a Fedora 12 machine, x86_64, 64-bit userland 2. transfer it to a Debian Testing machine, PARISC64, 32-bit userland acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | head -5 74f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms] 55fdd56670453ea66c011158c4b9d30179c1d049 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/ipt_MASQUERADE.ko 41adff63c730890480980d5d8ba513f1c216a858 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/iptable_nat.ko 90a33def1077bb8e97b8a78546dc96c2de62df46 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/nf_nat.ko 984c7bea90ce1376d5c8e7ef43a781801286e62d /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/tun.ko acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | tail -5 22492f3753c6a67de5c7ccbd6b863390c92c0723 /usr/lib64/libXt.so.6.0.0 353802bb7e1b895ba43507cc678f951e778e4c6f /usr/lib64/libMagickCore.so.2.0.0 d10c2897558595efe7be8b0584cf7e6398bc776c /usr/lib64/libfprint.so.0.0.0 a83ecfb519a788774a84d5ddde633c9ba56c03ab /home/acme/bin/perf d3ca765a8ecf257d263801d7ad8c49c189082317 /usr/lib64/libdwarf.so.0.0 acme@parisc:~/git/linux-2.6-tip$ acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm The file [kernel.kallsyms] cannot be used, trying to use /proc/kallsyms... ^^^^ The problem related to vmlinux handling, it shouldn't be trying this ^^^^ rather alien /proc/kallsyms at all... /lib64/libpthread-2.10.2.so with build id 5c68f7afeb33309c78037e374b0deee84dd441f6 not found, continuing without symbols /lib64/libc-2.10.2.so with build id eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 not found, continuing without symbols /home/acme/bin/perf with build id a83ecfb519a788774a84d5ddde633c9ba56c03ab not found, continuing without symbols /usr/sbin/openvpn with build id f2037a091ef36b591187a858d75e203690ea9409 not found, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwlcore.ko, continuing without symbols <SNIP more complaints about not finding the right build-ids, those will have to wait for 'perf archive' or plain copying what was collected by 'perf record' on the x86_64, source machine, see further below for an example of this > # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 acme@parisc:~/git/linux-2.6-tip$ Which matches what we get when running the same command for the same perf.data file on the F12, x86_64, source machine: [root@doppio linux-2.6-tip]# perf report --sort comm # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 [root@doppio linux-2.6-tip]# The other modes work as well, modulo the problem with vmlinux: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object # ........ ............... ................................. # 35.11% find ffffffff81002b5a 18.25% perf ffffffff8102235f 16.17% find libc-2.10.2.so 9.07% find find 5.80% swapper ffffffff8102235f 3.95% perf libc-2.10.2.so 2.33% init ffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k 1.35% find [e1000e] 0.68% sleep libc-2.10.2.so acme@parisc:~/git/linux-2.6-tip$ And the lack of the right buildids: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso,symbol 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object Symbol # ........ ............... ................................. ...... # 35.11% find ffffffff81002b5a [k] 0xffffffff81002b5a 18.25% perf ffffffff8102235f [k] 0xffffffff8102235f 16.17% find libc-2.10.2.so [.] 0x00000000045782 9.07% find find [.] 0x0000000000fb0e 5.80% swapper ffffffff8102235f [k] 0xffffffff8102235f 3.95% perf libc-2.10.2.so [.] 0x0000000007f398 2.33% init ffffffff810091b9 [k] 0xffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k [.] 0x00000000105440 1.35% find [e1000e] [k] 0x00000000010948 0.68% sleep libc-2.10.2.so [.] 0x0000000011ad5b acme@parisc:~/git/linux-2.6-tip$ But if we: acme@parisc:~/git/linux-2.6-tip$ ls ~/.debug ls: cannot access /home/acme/.debug: No such file or directory acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/lib64/libc-2.10.2.so/ acme@parisc:~/git/linux-2.6-tip$ scp doppio:.debug/lib64/libc-2.10.2.so/* ~/.debug/lib64/libc-2.10.2.so/ acme@doppio's password: eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 100% 1783KB 714.7KB/s 00:02 acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/.build-id/eb acme@parisc:~/git/linux-2.6-tip$ ln -s ../../lib64/libc-2.10.2.so/eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 ~/.debug/.build-id/eb/4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 acme@parisc:~/git/linux-2.6-tip$ perf report --dsos libc-2.10.2.so 2> /dev/null # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # acme@parisc:~/git/linux-2.6-tip$ Which matches what we get on the source, F12, x86_64 machine: [root@doppio linux-2.6-tip]# perf report --dsos libc-2.10.2.so # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # [root@doppio linux-2.6-tip]# So I think this is really, really nice in that it demonstrates the portability of perf.data files and the use of build-ids accross such aliens worlds :-) There are some things to fix tho, like the bitmap on the header, but things are looking good. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1263478990-8200-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-14 22:23:10 +08:00
void mem_bswap_64(void *src, int byte_size)
{
u64 *m = src;
while (byte_size > 0) {
*m = bswap_64(*m);
byte_size -= sizeof(u64);
++m;
}
}
static void event__all64_swap(event_t *self)
{
struct perf_event_header *hdr = &self->header;
mem_bswap_64(hdr + 1, self->header.size - sizeof(*hdr));
}
static void event__comm_swap(event_t *self)
{
self->comm.pid = bswap_32(self->comm.pid);
self->comm.tid = bswap_32(self->comm.tid);
}
static void event__mmap_swap(event_t *self)
{
self->mmap.pid = bswap_32(self->mmap.pid);
self->mmap.tid = bswap_32(self->mmap.tid);
self->mmap.start = bswap_64(self->mmap.start);
self->mmap.len = bswap_64(self->mmap.len);
self->mmap.pgoff = bswap_64(self->mmap.pgoff);
}
static void event__task_swap(event_t *self)
{
self->fork.pid = bswap_32(self->fork.pid);
self->fork.tid = bswap_32(self->fork.tid);
self->fork.ppid = bswap_32(self->fork.ppid);
self->fork.ptid = bswap_32(self->fork.ptid);
self->fork.time = bswap_64(self->fork.time);
}
static void event__read_swap(event_t *self)
{
self->read.pid = bswap_32(self->read.pid);
self->read.tid = bswap_32(self->read.tid);
self->read.value = bswap_64(self->read.value);
self->read.time_enabled = bswap_64(self->read.time_enabled);
self->read.time_running = bswap_64(self->read.time_running);
self->read.id = bswap_64(self->read.id);
}
static void event__attr_swap(event_t *self)
{
size_t size;
self->attr.attr.type = bswap_32(self->attr.attr.type);
self->attr.attr.size = bswap_32(self->attr.attr.size);
self->attr.attr.config = bswap_64(self->attr.attr.config);
self->attr.attr.sample_period = bswap_64(self->attr.attr.sample_period);
self->attr.attr.sample_type = bswap_64(self->attr.attr.sample_type);
self->attr.attr.read_format = bswap_64(self->attr.attr.read_format);
self->attr.attr.wakeup_events = bswap_32(self->attr.attr.wakeup_events);
self->attr.attr.bp_type = bswap_32(self->attr.attr.bp_type);
self->attr.attr.bp_addr = bswap_64(self->attr.attr.bp_addr);
self->attr.attr.bp_len = bswap_64(self->attr.attr.bp_len);
size = self->header.size;
size -= (void *)&self->attr.id - (void *)self;
mem_bswap_64(self->attr.id, size);
}
static void event__event_type_swap(event_t *self)
{
self->event_type.event_type.event_id =
bswap_64(self->event_type.event_type.event_id);
}
static void event__tracing_data_swap(event_t *self)
{
self->tracing_data.size = bswap_32(self->tracing_data.size);
}
perf tools: Cross platform perf.data analysis support There are still some problems related to loading vmlinux files, but those are unrelated to the feature implemented in this patch, so will get fixed in the next patches, but here are some results: 1. collect perf.data file on a Fedora 12 machine, x86_64, 64-bit userland 2. transfer it to a Debian Testing machine, PARISC64, 32-bit userland acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | head -5 74f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms] 55fdd56670453ea66c011158c4b9d30179c1d049 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/ipt_MASQUERADE.ko 41adff63c730890480980d5d8ba513f1c216a858 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/iptable_nat.ko 90a33def1077bb8e97b8a78546dc96c2de62df46 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/nf_nat.ko 984c7bea90ce1376d5c8e7ef43a781801286e62d /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/tun.ko acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | tail -5 22492f3753c6a67de5c7ccbd6b863390c92c0723 /usr/lib64/libXt.so.6.0.0 353802bb7e1b895ba43507cc678f951e778e4c6f /usr/lib64/libMagickCore.so.2.0.0 d10c2897558595efe7be8b0584cf7e6398bc776c /usr/lib64/libfprint.so.0.0.0 a83ecfb519a788774a84d5ddde633c9ba56c03ab /home/acme/bin/perf d3ca765a8ecf257d263801d7ad8c49c189082317 /usr/lib64/libdwarf.so.0.0 acme@parisc:~/git/linux-2.6-tip$ acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm The file [kernel.kallsyms] cannot be used, trying to use /proc/kallsyms... ^^^^ The problem related to vmlinux handling, it shouldn't be trying this ^^^^ rather alien /proc/kallsyms at all... /lib64/libpthread-2.10.2.so with build id 5c68f7afeb33309c78037e374b0deee84dd441f6 not found, continuing without symbols /lib64/libc-2.10.2.so with build id eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 not found, continuing without symbols /home/acme/bin/perf with build id a83ecfb519a788774a84d5ddde633c9ba56c03ab not found, continuing without symbols /usr/sbin/openvpn with build id f2037a091ef36b591187a858d75e203690ea9409 not found, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwlcore.ko, continuing without symbols <SNIP more complaints about not finding the right build-ids, those will have to wait for 'perf archive' or plain copying what was collected by 'perf record' on the x86_64, source machine, see further below for an example of this > # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 acme@parisc:~/git/linux-2.6-tip$ Which matches what we get when running the same command for the same perf.data file on the F12, x86_64, source machine: [root@doppio linux-2.6-tip]# perf report --sort comm # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 [root@doppio linux-2.6-tip]# The other modes work as well, modulo the problem with vmlinux: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object # ........ ............... ................................. # 35.11% find ffffffff81002b5a 18.25% perf ffffffff8102235f 16.17% find libc-2.10.2.so 9.07% find find 5.80% swapper ffffffff8102235f 3.95% perf libc-2.10.2.so 2.33% init ffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k 1.35% find [e1000e] 0.68% sleep libc-2.10.2.so acme@parisc:~/git/linux-2.6-tip$ And the lack of the right buildids: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso,symbol 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object Symbol # ........ ............... ................................. ...... # 35.11% find ffffffff81002b5a [k] 0xffffffff81002b5a 18.25% perf ffffffff8102235f [k] 0xffffffff8102235f 16.17% find libc-2.10.2.so [.] 0x00000000045782 9.07% find find [.] 0x0000000000fb0e 5.80% swapper ffffffff8102235f [k] 0xffffffff8102235f 3.95% perf libc-2.10.2.so [.] 0x0000000007f398 2.33% init ffffffff810091b9 [k] 0xffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k [.] 0x00000000105440 1.35% find [e1000e] [k] 0x00000000010948 0.68% sleep libc-2.10.2.so [.] 0x0000000011ad5b acme@parisc:~/git/linux-2.6-tip$ But if we: acme@parisc:~/git/linux-2.6-tip$ ls ~/.debug ls: cannot access /home/acme/.debug: No such file or directory acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/lib64/libc-2.10.2.so/ acme@parisc:~/git/linux-2.6-tip$ scp doppio:.debug/lib64/libc-2.10.2.so/* ~/.debug/lib64/libc-2.10.2.so/ acme@doppio's password: eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 100% 1783KB 714.7KB/s 00:02 acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/.build-id/eb acme@parisc:~/git/linux-2.6-tip$ ln -s ../../lib64/libc-2.10.2.so/eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 ~/.debug/.build-id/eb/4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 acme@parisc:~/git/linux-2.6-tip$ perf report --dsos libc-2.10.2.so 2> /dev/null # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # acme@parisc:~/git/linux-2.6-tip$ Which matches what we get on the source, F12, x86_64 machine: [root@doppio linux-2.6-tip]# perf report --dsos libc-2.10.2.so # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # [root@doppio linux-2.6-tip]# So I think this is really, really nice in that it demonstrates the portability of perf.data files and the use of build-ids accross such aliens worlds :-) There are some things to fix tho, like the bitmap on the header, but things are looking good. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1263478990-8200-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-14 22:23:10 +08:00
typedef void (*event__swap_op)(event_t *self);
static event__swap_op event__swap_ops[] = {
[PERF_RECORD_MMAP] = event__mmap_swap,
[PERF_RECORD_COMM] = event__comm_swap,
[PERF_RECORD_FORK] = event__task_swap,
[PERF_RECORD_EXIT] = event__task_swap,
[PERF_RECORD_LOST] = event__all64_swap,
[PERF_RECORD_READ] = event__read_swap,
[PERF_RECORD_SAMPLE] = event__all64_swap,
[PERF_RECORD_HEADER_ATTR] = event__attr_swap,
[PERF_RECORD_HEADER_EVENT_TYPE] = event__event_type_swap,
[PERF_RECORD_HEADER_TRACING_DATA] = event__tracing_data_swap,
[PERF_RECORD_HEADER_BUILD_ID] = NULL,
[PERF_RECORD_HEADER_MAX] = NULL,
perf tools: Cross platform perf.data analysis support There are still some problems related to loading vmlinux files, but those are unrelated to the feature implemented in this patch, so will get fixed in the next patches, but here are some results: 1. collect perf.data file on a Fedora 12 machine, x86_64, 64-bit userland 2. transfer it to a Debian Testing machine, PARISC64, 32-bit userland acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | head -5 74f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms] 55fdd56670453ea66c011158c4b9d30179c1d049 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/ipt_MASQUERADE.ko 41adff63c730890480980d5d8ba513f1c216a858 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/iptable_nat.ko 90a33def1077bb8e97b8a78546dc96c2de62df46 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/nf_nat.ko 984c7bea90ce1376d5c8e7ef43a781801286e62d /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/tun.ko acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | tail -5 22492f3753c6a67de5c7ccbd6b863390c92c0723 /usr/lib64/libXt.so.6.0.0 353802bb7e1b895ba43507cc678f951e778e4c6f /usr/lib64/libMagickCore.so.2.0.0 d10c2897558595efe7be8b0584cf7e6398bc776c /usr/lib64/libfprint.so.0.0.0 a83ecfb519a788774a84d5ddde633c9ba56c03ab /home/acme/bin/perf d3ca765a8ecf257d263801d7ad8c49c189082317 /usr/lib64/libdwarf.so.0.0 acme@parisc:~/git/linux-2.6-tip$ acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm The file [kernel.kallsyms] cannot be used, trying to use /proc/kallsyms... ^^^^ The problem related to vmlinux handling, it shouldn't be trying this ^^^^ rather alien /proc/kallsyms at all... /lib64/libpthread-2.10.2.so with build id 5c68f7afeb33309c78037e374b0deee84dd441f6 not found, continuing without symbols /lib64/libc-2.10.2.so with build id eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 not found, continuing without symbols /home/acme/bin/perf with build id a83ecfb519a788774a84d5ddde633c9ba56c03ab not found, continuing without symbols /usr/sbin/openvpn with build id f2037a091ef36b591187a858d75e203690ea9409 not found, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwlcore.ko, continuing without symbols <SNIP more complaints about not finding the right build-ids, those will have to wait for 'perf archive' or plain copying what was collected by 'perf record' on the x86_64, source machine, see further below for an example of this > # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 acme@parisc:~/git/linux-2.6-tip$ Which matches what we get when running the same command for the same perf.data file on the F12, x86_64, source machine: [root@doppio linux-2.6-tip]# perf report --sort comm # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 [root@doppio linux-2.6-tip]# The other modes work as well, modulo the problem with vmlinux: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object # ........ ............... ................................. # 35.11% find ffffffff81002b5a 18.25% perf ffffffff8102235f 16.17% find libc-2.10.2.so 9.07% find find 5.80% swapper ffffffff8102235f 3.95% perf libc-2.10.2.so 2.33% init ffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k 1.35% find [e1000e] 0.68% sleep libc-2.10.2.so acme@parisc:~/git/linux-2.6-tip$ And the lack of the right buildids: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso,symbol 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object Symbol # ........ ............... ................................. ...... # 35.11% find ffffffff81002b5a [k] 0xffffffff81002b5a 18.25% perf ffffffff8102235f [k] 0xffffffff8102235f 16.17% find libc-2.10.2.so [.] 0x00000000045782 9.07% find find [.] 0x0000000000fb0e 5.80% swapper ffffffff8102235f [k] 0xffffffff8102235f 3.95% perf libc-2.10.2.so [.] 0x0000000007f398 2.33% init ffffffff810091b9 [k] 0xffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k [.] 0x00000000105440 1.35% find [e1000e] [k] 0x00000000010948 0.68% sleep libc-2.10.2.so [.] 0x0000000011ad5b acme@parisc:~/git/linux-2.6-tip$ But if we: acme@parisc:~/git/linux-2.6-tip$ ls ~/.debug ls: cannot access /home/acme/.debug: No such file or directory acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/lib64/libc-2.10.2.so/ acme@parisc:~/git/linux-2.6-tip$ scp doppio:.debug/lib64/libc-2.10.2.so/* ~/.debug/lib64/libc-2.10.2.so/ acme@doppio's password: eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 100% 1783KB 714.7KB/s 00:02 acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/.build-id/eb acme@parisc:~/git/linux-2.6-tip$ ln -s ../../lib64/libc-2.10.2.so/eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 ~/.debug/.build-id/eb/4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 acme@parisc:~/git/linux-2.6-tip$ perf report --dsos libc-2.10.2.so 2> /dev/null # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # acme@parisc:~/git/linux-2.6-tip$ Which matches what we get on the source, F12, x86_64 machine: [root@doppio linux-2.6-tip]# perf report --dsos libc-2.10.2.so # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # [root@doppio linux-2.6-tip]# So I think this is really, really nice in that it demonstrates the portability of perf.data files and the use of build-ids accross such aliens worlds :-) There are some things to fix tho, like the bitmap on the header, but things are looking good. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1263478990-8200-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-14 22:23:10 +08:00
};
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
struct sample_queue {
u64 timestamp;
event_t *event;
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
struct list_head list;
};
static void flush_sample_queue(struct perf_session *s,
struct perf_event_ops *ops)
{
struct ordered_samples *os = &s->ordered_samples;
struct list_head *head = &os->samples;
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
struct sample_queue *tmp, *iter;
u64 limit = os->next_flush;
u64 last_ts = os->last_sample ? os->last_sample->timestamp : 0ULL;
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
perf: Provide a new deterministic events reordering algorithm The current events reordering algorithm is based on a heuristic that gets broken once we deal with a very fast flow of events. Indeed the time period based flushing is not suitable anymore in the following case, assuming we have a flush period of two seconds. CPU 0 | CPU 1 | cnt1 timestamps | cnt1 timestamps | 0 | 0 1 | 1 2 | 2 3 | 3 [...] | [...] 4 seconds later If we spend too much time to read the buffers (case of a lot of events to record in each buffers or when we have a lot of CPU buffers to read), in the next pass the CPU 0 buffer could contain a slice of several seconds of events. We'll read them all and notice we've reached the period to flush. In the above example we flush the first half of the CPU 0 buffer, then we read the CPU 1 buffer where we have events that were on the flush slice and then the reordering fails. It's simple to reproduce with: perf lock record perf bench sched messaging To solve this, we use a new solution that doesn't rely on an heuristical time slice period anymore but on a deterministic basis based on how perf record does its job. perf record saves the buffers through passes. A pass is a tour on every buffers from every CPUs. This is made in order: for each CPU we read the buffers of every counters. So the more buffers we visit, the later will be the timstamps of their events. When perf record finishes a pass it records a PERF_RECORD_FINISHED_ROUND pseudo event. We record the max timestamp t found in the pass n. Assuming these timestamps are monotonic across cpus, we know that if a buffer still has events with timestamps below t, they will be all available and then read in the pass n + 1. Hence when we start to read the pass n + 2, we can safely flush every events with timestamps below t. ============ PASS n ================= CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 1 | 2 2 | 3 - | 4 <--- max recorded ============ PASS n + 1 ============== CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 3 | 5 4 | 6 5 | 7 <---- max recorded Flush every events below timestamp 4 ============ PASS n + 2 ============== CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 6 | 8 7 | 9 - | 10 Flush every events below timestamp 7 etc... It also works on perf.data versions that don't have PERF_RECORD_FINISHED_ROUND pseudo events. The difference is that the events will be only flushed in the end of the perf.data processing. It will then consume more memory and scale less with large perf.data files. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Tom Zanussi <tzanussi@gmail.com> Cc: Masami Hiramatsu <mhiramat@redhat.com>
2010-05-03 21:14:33 +08:00
if (!ops->ordered_samples || !limit)
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
return;
list_for_each_entry_safe(iter, tmp, head, list) {
if (iter->timestamp > limit)
break;
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
ops->sample(iter->event, s);
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
os->last_flush = iter->timestamp;
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
list_del(&iter->list);
free(iter->event);
free(iter);
}
if (list_empty(head)) {
os->last_sample = NULL;
} else if (last_ts <= limit) {
os->last_sample =
list_entry(head->prev, struct sample_queue, list);
}
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
}
perf: Provide a new deterministic events reordering algorithm The current events reordering algorithm is based on a heuristic that gets broken once we deal with a very fast flow of events. Indeed the time period based flushing is not suitable anymore in the following case, assuming we have a flush period of two seconds. CPU 0 | CPU 1 | cnt1 timestamps | cnt1 timestamps | 0 | 0 1 | 1 2 | 2 3 | 3 [...] | [...] 4 seconds later If we spend too much time to read the buffers (case of a lot of events to record in each buffers or when we have a lot of CPU buffers to read), in the next pass the CPU 0 buffer could contain a slice of several seconds of events. We'll read them all and notice we've reached the period to flush. In the above example we flush the first half of the CPU 0 buffer, then we read the CPU 1 buffer where we have events that were on the flush slice and then the reordering fails. It's simple to reproduce with: perf lock record perf bench sched messaging To solve this, we use a new solution that doesn't rely on an heuristical time slice period anymore but on a deterministic basis based on how perf record does its job. perf record saves the buffers through passes. A pass is a tour on every buffers from every CPUs. This is made in order: for each CPU we read the buffers of every counters. So the more buffers we visit, the later will be the timstamps of their events. When perf record finishes a pass it records a PERF_RECORD_FINISHED_ROUND pseudo event. We record the max timestamp t found in the pass n. Assuming these timestamps are monotonic across cpus, we know that if a buffer still has events with timestamps below t, they will be all available and then read in the pass n + 1. Hence when we start to read the pass n + 2, we can safely flush every events with timestamps below t. ============ PASS n ================= CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 1 | 2 2 | 3 - | 4 <--- max recorded ============ PASS n + 1 ============== CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 3 | 5 4 | 6 5 | 7 <---- max recorded Flush every events below timestamp 4 ============ PASS n + 2 ============== CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 6 | 8 7 | 9 - | 10 Flush every events below timestamp 7 etc... It also works on perf.data versions that don't have PERF_RECORD_FINISHED_ROUND pseudo events. The difference is that the events will be only flushed in the end of the perf.data processing. It will then consume more memory and scale less with large perf.data files. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Tom Zanussi <tzanussi@gmail.com> Cc: Masami Hiramatsu <mhiramat@redhat.com>
2010-05-03 21:14:33 +08:00
/*
* When perf record finishes a pass on every buffers, it records this pseudo
* event.
* We record the max timestamp t found in the pass n.
* Assuming these timestamps are monotonic across cpus, we know that if
* a buffer still has events with timestamps below t, they will be all
* available and then read in the pass n + 1.
* Hence when we start to read the pass n + 2, we can safely flush every
* events with timestamps below t.
*
* ============ PASS n =================
* CPU 0 | CPU 1
* |
* cnt1 timestamps | cnt2 timestamps
* 1 | 2
* 2 | 3
* - | 4 <--- max recorded
*
* ============ PASS n + 1 ==============
* CPU 0 | CPU 1
* |
* cnt1 timestamps | cnt2 timestamps
* 3 | 5
* 4 | 6
* 5 | 7 <---- max recorded
*
* Flush every events below timestamp 4
*
* ============ PASS n + 2 ==============
* CPU 0 | CPU 1
* |
* cnt1 timestamps | cnt2 timestamps
* 6 | 8
* 7 | 9
* - | 10
*
* Flush every events below timestamp 7
* etc...
*/
static int process_finished_round(event_t *event __used,
struct perf_session *session,
struct perf_event_ops *ops)
{
flush_sample_queue(session, ops);
session->ordered_samples.next_flush = session->ordered_samples.max_timestamp;
return 0;
}
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
/* The queue is ordered by time */
static void __queue_sample_event(struct sample_queue *new,
struct perf_session *s)
{
struct ordered_samples *os = &s->ordered_samples;
struct sample_queue *sample = os->last_sample;
u64 timestamp = new->timestamp;
struct list_head *p;
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
os->last_sample = new;
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
if (!sample) {
list_add(&new->list, &os->samples);
os->max_timestamp = timestamp;
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
return;
}
/*
* last_sample might point to some random place in the list as it's
* the last queued event. We expect that the new event is close to
* this.
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
*/
if (sample->timestamp <= timestamp) {
while (sample->timestamp <= timestamp) {
p = sample->list.next;
if (p == &os->samples) {
list_add_tail(&new->list, &os->samples);
os->max_timestamp = timestamp;
return;
}
sample = list_entry(p, struct sample_queue, list);
}
list_add_tail(&new->list, &sample->list);
} else {
while (sample->timestamp > timestamp) {
p = sample->list.prev;
if (p == &os->samples) {
list_add(&new->list, &os->samples);
return;
}
sample = list_entry(p, struct sample_queue, list);
}
list_add(&new->list, &sample->list);
}
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
}
static int queue_sample_event(event_t *event, struct sample_data *data,
perf: Provide a new deterministic events reordering algorithm The current events reordering algorithm is based on a heuristic that gets broken once we deal with a very fast flow of events. Indeed the time period based flushing is not suitable anymore in the following case, assuming we have a flush period of two seconds. CPU 0 | CPU 1 | cnt1 timestamps | cnt1 timestamps | 0 | 0 1 | 1 2 | 2 3 | 3 [...] | [...] 4 seconds later If we spend too much time to read the buffers (case of a lot of events to record in each buffers or when we have a lot of CPU buffers to read), in the next pass the CPU 0 buffer could contain a slice of several seconds of events. We'll read them all and notice we've reached the period to flush. In the above example we flush the first half of the CPU 0 buffer, then we read the CPU 1 buffer where we have events that were on the flush slice and then the reordering fails. It's simple to reproduce with: perf lock record perf bench sched messaging To solve this, we use a new solution that doesn't rely on an heuristical time slice period anymore but on a deterministic basis based on how perf record does its job. perf record saves the buffers through passes. A pass is a tour on every buffers from every CPUs. This is made in order: for each CPU we read the buffers of every counters. So the more buffers we visit, the later will be the timstamps of their events. When perf record finishes a pass it records a PERF_RECORD_FINISHED_ROUND pseudo event. We record the max timestamp t found in the pass n. Assuming these timestamps are monotonic across cpus, we know that if a buffer still has events with timestamps below t, they will be all available and then read in the pass n + 1. Hence when we start to read the pass n + 2, we can safely flush every events with timestamps below t. ============ PASS n ================= CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 1 | 2 2 | 3 - | 4 <--- max recorded ============ PASS n + 1 ============== CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 3 | 5 4 | 6 5 | 7 <---- max recorded Flush every events below timestamp 4 ============ PASS n + 2 ============== CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 6 | 8 7 | 9 - | 10 Flush every events below timestamp 7 etc... It also works on perf.data versions that don't have PERF_RECORD_FINISHED_ROUND pseudo events. The difference is that the events will be only flushed in the end of the perf.data processing. It will then consume more memory and scale less with large perf.data files. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Tom Zanussi <tzanussi@gmail.com> Cc: Masami Hiramatsu <mhiramat@redhat.com>
2010-05-03 21:14:33 +08:00
struct perf_session *s)
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
{
u64 timestamp = data->time;
struct sample_queue *new;
if (timestamp < s->ordered_samples.last_flush) {
printf("Warning: Timestamp below last timeslice flush\n");
return -EINVAL;
}
new = malloc(sizeof(*new));
if (!new)
return -ENOMEM;
new->timestamp = timestamp;
new->event = malloc(event->header.size);
if (!new->event) {
free(new);
return -ENOMEM;
}
memcpy(new->event, event, event->header.size);
__queue_sample_event(new, s);
return 0;
}
static int perf_session__process_sample(event_t *event, struct perf_session *s,
struct perf_event_ops *ops)
{
struct sample_data data;
if (!ops->ordered_samples)
return ops->sample(event, s);
bzero(&data, sizeof(struct sample_data));
event__parse_sample(event, s->sample_type, &data);
perf: Provide a new deterministic events reordering algorithm The current events reordering algorithm is based on a heuristic that gets broken once we deal with a very fast flow of events. Indeed the time period based flushing is not suitable anymore in the following case, assuming we have a flush period of two seconds. CPU 0 | CPU 1 | cnt1 timestamps | cnt1 timestamps | 0 | 0 1 | 1 2 | 2 3 | 3 [...] | [...] 4 seconds later If we spend too much time to read the buffers (case of a lot of events to record in each buffers or when we have a lot of CPU buffers to read), in the next pass the CPU 0 buffer could contain a slice of several seconds of events. We'll read them all and notice we've reached the period to flush. In the above example we flush the first half of the CPU 0 buffer, then we read the CPU 1 buffer where we have events that were on the flush slice and then the reordering fails. It's simple to reproduce with: perf lock record perf bench sched messaging To solve this, we use a new solution that doesn't rely on an heuristical time slice period anymore but on a deterministic basis based on how perf record does its job. perf record saves the buffers through passes. A pass is a tour on every buffers from every CPUs. This is made in order: for each CPU we read the buffers of every counters. So the more buffers we visit, the later will be the timstamps of their events. When perf record finishes a pass it records a PERF_RECORD_FINISHED_ROUND pseudo event. We record the max timestamp t found in the pass n. Assuming these timestamps are monotonic across cpus, we know that if a buffer still has events with timestamps below t, they will be all available and then read in the pass n + 1. Hence when we start to read the pass n + 2, we can safely flush every events with timestamps below t. ============ PASS n ================= CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 1 | 2 2 | 3 - | 4 <--- max recorded ============ PASS n + 1 ============== CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 3 | 5 4 | 6 5 | 7 <---- max recorded Flush every events below timestamp 4 ============ PASS n + 2 ============== CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 6 | 8 7 | 9 - | 10 Flush every events below timestamp 7 etc... It also works on perf.data versions that don't have PERF_RECORD_FINISHED_ROUND pseudo events. The difference is that the events will be only flushed in the end of the perf.data processing. It will then consume more memory and scale less with large perf.data files. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Tom Zanussi <tzanussi@gmail.com> Cc: Masami Hiramatsu <mhiramat@redhat.com>
2010-05-03 21:14:33 +08:00
queue_sample_event(event, &data, s);
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
return 0;
}
static int perf_session__process_event(struct perf_session *self,
event_t *event,
struct perf_event_ops *ops,
u64 file_offset)
{
trace_event(event);
if (event->header.type < PERF_RECORD_HEADER_MAX) {
perf tools: Cross platform perf.data analysis support There are still some problems related to loading vmlinux files, but those are unrelated to the feature implemented in this patch, so will get fixed in the next patches, but here are some results: 1. collect perf.data file on a Fedora 12 machine, x86_64, 64-bit userland 2. transfer it to a Debian Testing machine, PARISC64, 32-bit userland acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | head -5 74f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms] 55fdd56670453ea66c011158c4b9d30179c1d049 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/ipt_MASQUERADE.ko 41adff63c730890480980d5d8ba513f1c216a858 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/iptable_nat.ko 90a33def1077bb8e97b8a78546dc96c2de62df46 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/nf_nat.ko 984c7bea90ce1376d5c8e7ef43a781801286e62d /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/tun.ko acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | tail -5 22492f3753c6a67de5c7ccbd6b863390c92c0723 /usr/lib64/libXt.so.6.0.0 353802bb7e1b895ba43507cc678f951e778e4c6f /usr/lib64/libMagickCore.so.2.0.0 d10c2897558595efe7be8b0584cf7e6398bc776c /usr/lib64/libfprint.so.0.0.0 a83ecfb519a788774a84d5ddde633c9ba56c03ab /home/acme/bin/perf d3ca765a8ecf257d263801d7ad8c49c189082317 /usr/lib64/libdwarf.so.0.0 acme@parisc:~/git/linux-2.6-tip$ acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm The file [kernel.kallsyms] cannot be used, trying to use /proc/kallsyms... ^^^^ The problem related to vmlinux handling, it shouldn't be trying this ^^^^ rather alien /proc/kallsyms at all... /lib64/libpthread-2.10.2.so with build id 5c68f7afeb33309c78037e374b0deee84dd441f6 not found, continuing without symbols /lib64/libc-2.10.2.so with build id eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 not found, continuing without symbols /home/acme/bin/perf with build id a83ecfb519a788774a84d5ddde633c9ba56c03ab not found, continuing without symbols /usr/sbin/openvpn with build id f2037a091ef36b591187a858d75e203690ea9409 not found, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwlcore.ko, continuing without symbols <SNIP more complaints about not finding the right build-ids, those will have to wait for 'perf archive' or plain copying what was collected by 'perf record' on the x86_64, source machine, see further below for an example of this > # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 acme@parisc:~/git/linux-2.6-tip$ Which matches what we get when running the same command for the same perf.data file on the F12, x86_64, source machine: [root@doppio linux-2.6-tip]# perf report --sort comm # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 [root@doppio linux-2.6-tip]# The other modes work as well, modulo the problem with vmlinux: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object # ........ ............... ................................. # 35.11% find ffffffff81002b5a 18.25% perf ffffffff8102235f 16.17% find libc-2.10.2.so 9.07% find find 5.80% swapper ffffffff8102235f 3.95% perf libc-2.10.2.so 2.33% init ffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k 1.35% find [e1000e] 0.68% sleep libc-2.10.2.so acme@parisc:~/git/linux-2.6-tip$ And the lack of the right buildids: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso,symbol 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object Symbol # ........ ............... ................................. ...... # 35.11% find ffffffff81002b5a [k] 0xffffffff81002b5a 18.25% perf ffffffff8102235f [k] 0xffffffff8102235f 16.17% find libc-2.10.2.so [.] 0x00000000045782 9.07% find find [.] 0x0000000000fb0e 5.80% swapper ffffffff8102235f [k] 0xffffffff8102235f 3.95% perf libc-2.10.2.so [.] 0x0000000007f398 2.33% init ffffffff810091b9 [k] 0xffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k [.] 0x00000000105440 1.35% find [e1000e] [k] 0x00000000010948 0.68% sleep libc-2.10.2.so [.] 0x0000000011ad5b acme@parisc:~/git/linux-2.6-tip$ But if we: acme@parisc:~/git/linux-2.6-tip$ ls ~/.debug ls: cannot access /home/acme/.debug: No such file or directory acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/lib64/libc-2.10.2.so/ acme@parisc:~/git/linux-2.6-tip$ scp doppio:.debug/lib64/libc-2.10.2.so/* ~/.debug/lib64/libc-2.10.2.so/ acme@doppio's password: eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 100% 1783KB 714.7KB/s 00:02 acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/.build-id/eb acme@parisc:~/git/linux-2.6-tip$ ln -s ../../lib64/libc-2.10.2.so/eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 ~/.debug/.build-id/eb/4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 acme@parisc:~/git/linux-2.6-tip$ perf report --dsos libc-2.10.2.so 2> /dev/null # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # acme@parisc:~/git/linux-2.6-tip$ Which matches what we get on the source, F12, x86_64 machine: [root@doppio linux-2.6-tip]# perf report --dsos libc-2.10.2.so # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # [root@doppio linux-2.6-tip]# So I think this is really, really nice in that it demonstrates the portability of perf.data files and the use of build-ids accross such aliens worlds :-) There are some things to fix tho, like the bitmap on the header, but things are looking good. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1263478990-8200-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-14 22:23:10 +08:00
dump_printf("%#Lx [%#x]: PERF_RECORD_%s",
file_offset, event->header.size,
event__name[event->header.type]);
hists__inc_nr_events(&self->hists, event->header.type);
}
perf tools: Cross platform perf.data analysis support There are still some problems related to loading vmlinux files, but those are unrelated to the feature implemented in this patch, so will get fixed in the next patches, but here are some results: 1. collect perf.data file on a Fedora 12 machine, x86_64, 64-bit userland 2. transfer it to a Debian Testing machine, PARISC64, 32-bit userland acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | head -5 74f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms] 55fdd56670453ea66c011158c4b9d30179c1d049 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/ipt_MASQUERADE.ko 41adff63c730890480980d5d8ba513f1c216a858 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/iptable_nat.ko 90a33def1077bb8e97b8a78546dc96c2de62df46 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/nf_nat.ko 984c7bea90ce1376d5c8e7ef43a781801286e62d /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/tun.ko acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | tail -5 22492f3753c6a67de5c7ccbd6b863390c92c0723 /usr/lib64/libXt.so.6.0.0 353802bb7e1b895ba43507cc678f951e778e4c6f /usr/lib64/libMagickCore.so.2.0.0 d10c2897558595efe7be8b0584cf7e6398bc776c /usr/lib64/libfprint.so.0.0.0 a83ecfb519a788774a84d5ddde633c9ba56c03ab /home/acme/bin/perf d3ca765a8ecf257d263801d7ad8c49c189082317 /usr/lib64/libdwarf.so.0.0 acme@parisc:~/git/linux-2.6-tip$ acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm The file [kernel.kallsyms] cannot be used, trying to use /proc/kallsyms... ^^^^ The problem related to vmlinux handling, it shouldn't be trying this ^^^^ rather alien /proc/kallsyms at all... /lib64/libpthread-2.10.2.so with build id 5c68f7afeb33309c78037e374b0deee84dd441f6 not found, continuing without symbols /lib64/libc-2.10.2.so with build id eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 not found, continuing without symbols /home/acme/bin/perf with build id a83ecfb519a788774a84d5ddde633c9ba56c03ab not found, continuing without symbols /usr/sbin/openvpn with build id f2037a091ef36b591187a858d75e203690ea9409 not found, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwlcore.ko, continuing without symbols <SNIP more complaints about not finding the right build-ids, those will have to wait for 'perf archive' or plain copying what was collected by 'perf record' on the x86_64, source machine, see further below for an example of this > # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 acme@parisc:~/git/linux-2.6-tip$ Which matches what we get when running the same command for the same perf.data file on the F12, x86_64, source machine: [root@doppio linux-2.6-tip]# perf report --sort comm # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 [root@doppio linux-2.6-tip]# The other modes work as well, modulo the problem with vmlinux: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object # ........ ............... ................................. # 35.11% find ffffffff81002b5a 18.25% perf ffffffff8102235f 16.17% find libc-2.10.2.so 9.07% find find 5.80% swapper ffffffff8102235f 3.95% perf libc-2.10.2.so 2.33% init ffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k 1.35% find [e1000e] 0.68% sleep libc-2.10.2.so acme@parisc:~/git/linux-2.6-tip$ And the lack of the right buildids: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso,symbol 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object Symbol # ........ ............... ................................. ...... # 35.11% find ffffffff81002b5a [k] 0xffffffff81002b5a 18.25% perf ffffffff8102235f [k] 0xffffffff8102235f 16.17% find libc-2.10.2.so [.] 0x00000000045782 9.07% find find [.] 0x0000000000fb0e 5.80% swapper ffffffff8102235f [k] 0xffffffff8102235f 3.95% perf libc-2.10.2.so [.] 0x0000000007f398 2.33% init ffffffff810091b9 [k] 0xffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k [.] 0x00000000105440 1.35% find [e1000e] [k] 0x00000000010948 0.68% sleep libc-2.10.2.so [.] 0x0000000011ad5b acme@parisc:~/git/linux-2.6-tip$ But if we: acme@parisc:~/git/linux-2.6-tip$ ls ~/.debug ls: cannot access /home/acme/.debug: No such file or directory acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/lib64/libc-2.10.2.so/ acme@parisc:~/git/linux-2.6-tip$ scp doppio:.debug/lib64/libc-2.10.2.so/* ~/.debug/lib64/libc-2.10.2.so/ acme@doppio's password: eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 100% 1783KB 714.7KB/s 00:02 acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/.build-id/eb acme@parisc:~/git/linux-2.6-tip$ ln -s ../../lib64/libc-2.10.2.so/eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 ~/.debug/.build-id/eb/4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 acme@parisc:~/git/linux-2.6-tip$ perf report --dsos libc-2.10.2.so 2> /dev/null # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # acme@parisc:~/git/linux-2.6-tip$ Which matches what we get on the source, F12, x86_64 machine: [root@doppio linux-2.6-tip]# perf report --dsos libc-2.10.2.so # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # [root@doppio linux-2.6-tip]# So I think this is really, really nice in that it demonstrates the portability of perf.data files and the use of build-ids accross such aliens worlds :-) There are some things to fix tho, like the bitmap on the header, but things are looking good. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1263478990-8200-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-14 22:23:10 +08:00
if (self->header.needs_swap && event__swap_ops[event->header.type])
event__swap_ops[event->header.type](event);
switch (event->header.type) {
case PERF_RECORD_SAMPLE:
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
return perf_session__process_sample(event, self, ops);
case PERF_RECORD_MMAP:
return ops->mmap(event, self);
case PERF_RECORD_COMM:
return ops->comm(event, self);
case PERF_RECORD_FORK:
return ops->fork(event, self);
case PERF_RECORD_EXIT:
return ops->exit(event, self);
case PERF_RECORD_LOST:
return ops->lost(event, self);
case PERF_RECORD_READ:
return ops->read(event, self);
case PERF_RECORD_THROTTLE:
return ops->throttle(event, self);
case PERF_RECORD_UNTHROTTLE:
return ops->unthrottle(event, self);
case PERF_RECORD_HEADER_ATTR:
return ops->attr(event, self);
case PERF_RECORD_HEADER_EVENT_TYPE:
return ops->event_type(event, self);
case PERF_RECORD_HEADER_TRACING_DATA:
/* setup for reading amidst mmap */
lseek(self->fd, file_offset, SEEK_SET);
return ops->tracing_data(event, self);
case PERF_RECORD_HEADER_BUILD_ID:
return ops->build_id(event, self);
perf: Provide a new deterministic events reordering algorithm The current events reordering algorithm is based on a heuristic that gets broken once we deal with a very fast flow of events. Indeed the time period based flushing is not suitable anymore in the following case, assuming we have a flush period of two seconds. CPU 0 | CPU 1 | cnt1 timestamps | cnt1 timestamps | 0 | 0 1 | 1 2 | 2 3 | 3 [...] | [...] 4 seconds later If we spend too much time to read the buffers (case of a lot of events to record in each buffers or when we have a lot of CPU buffers to read), in the next pass the CPU 0 buffer could contain a slice of several seconds of events. We'll read them all and notice we've reached the period to flush. In the above example we flush the first half of the CPU 0 buffer, then we read the CPU 1 buffer where we have events that were on the flush slice and then the reordering fails. It's simple to reproduce with: perf lock record perf bench sched messaging To solve this, we use a new solution that doesn't rely on an heuristical time slice period anymore but on a deterministic basis based on how perf record does its job. perf record saves the buffers through passes. A pass is a tour on every buffers from every CPUs. This is made in order: for each CPU we read the buffers of every counters. So the more buffers we visit, the later will be the timstamps of their events. When perf record finishes a pass it records a PERF_RECORD_FINISHED_ROUND pseudo event. We record the max timestamp t found in the pass n. Assuming these timestamps are monotonic across cpus, we know that if a buffer still has events with timestamps below t, they will be all available and then read in the pass n + 1. Hence when we start to read the pass n + 2, we can safely flush every events with timestamps below t. ============ PASS n ================= CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 1 | 2 2 | 3 - | 4 <--- max recorded ============ PASS n + 1 ============== CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 3 | 5 4 | 6 5 | 7 <---- max recorded Flush every events below timestamp 4 ============ PASS n + 2 ============== CPU 0 | CPU 1 | cnt1 timestamps | cnt2 timestamps 6 | 8 7 | 9 - | 10 Flush every events below timestamp 7 etc... It also works on perf.data versions that don't have PERF_RECORD_FINISHED_ROUND pseudo events. The difference is that the events will be only flushed in the end of the perf.data processing. It will then consume more memory and scale less with large perf.data files. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Tom Zanussi <tzanussi@gmail.com> Cc: Masami Hiramatsu <mhiramat@redhat.com>
2010-05-03 21:14:33 +08:00
case PERF_RECORD_FINISHED_ROUND:
return ops->finished_round(event, self, ops);
default:
++self->hists.stats.nr_unknown_events;
return -1;
}
}
perf tools: Cross platform perf.data analysis support There are still some problems related to loading vmlinux files, but those are unrelated to the feature implemented in this patch, so will get fixed in the next patches, but here are some results: 1. collect perf.data file on a Fedora 12 machine, x86_64, 64-bit userland 2. transfer it to a Debian Testing machine, PARISC64, 32-bit userland acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | head -5 74f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms] 55fdd56670453ea66c011158c4b9d30179c1d049 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/ipt_MASQUERADE.ko 41adff63c730890480980d5d8ba513f1c216a858 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/iptable_nat.ko 90a33def1077bb8e97b8a78546dc96c2de62df46 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/nf_nat.ko 984c7bea90ce1376d5c8e7ef43a781801286e62d /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/tun.ko acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | tail -5 22492f3753c6a67de5c7ccbd6b863390c92c0723 /usr/lib64/libXt.so.6.0.0 353802bb7e1b895ba43507cc678f951e778e4c6f /usr/lib64/libMagickCore.so.2.0.0 d10c2897558595efe7be8b0584cf7e6398bc776c /usr/lib64/libfprint.so.0.0.0 a83ecfb519a788774a84d5ddde633c9ba56c03ab /home/acme/bin/perf d3ca765a8ecf257d263801d7ad8c49c189082317 /usr/lib64/libdwarf.so.0.0 acme@parisc:~/git/linux-2.6-tip$ acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm The file [kernel.kallsyms] cannot be used, trying to use /proc/kallsyms... ^^^^ The problem related to vmlinux handling, it shouldn't be trying this ^^^^ rather alien /proc/kallsyms at all... /lib64/libpthread-2.10.2.so with build id 5c68f7afeb33309c78037e374b0deee84dd441f6 not found, continuing without symbols /lib64/libc-2.10.2.so with build id eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 not found, continuing without symbols /home/acme/bin/perf with build id a83ecfb519a788774a84d5ddde633c9ba56c03ab not found, continuing without symbols /usr/sbin/openvpn with build id f2037a091ef36b591187a858d75e203690ea9409 not found, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwlcore.ko, continuing without symbols <SNIP more complaints about not finding the right build-ids, those will have to wait for 'perf archive' or plain copying what was collected by 'perf record' on the x86_64, source machine, see further below for an example of this > # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 acme@parisc:~/git/linux-2.6-tip$ Which matches what we get when running the same command for the same perf.data file on the F12, x86_64, source machine: [root@doppio linux-2.6-tip]# perf report --sort comm # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 [root@doppio linux-2.6-tip]# The other modes work as well, modulo the problem with vmlinux: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object # ........ ............... ................................. # 35.11% find ffffffff81002b5a 18.25% perf ffffffff8102235f 16.17% find libc-2.10.2.so 9.07% find find 5.80% swapper ffffffff8102235f 3.95% perf libc-2.10.2.so 2.33% init ffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k 1.35% find [e1000e] 0.68% sleep libc-2.10.2.so acme@parisc:~/git/linux-2.6-tip$ And the lack of the right buildids: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso,symbol 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object Symbol # ........ ............... ................................. ...... # 35.11% find ffffffff81002b5a [k] 0xffffffff81002b5a 18.25% perf ffffffff8102235f [k] 0xffffffff8102235f 16.17% find libc-2.10.2.so [.] 0x00000000045782 9.07% find find [.] 0x0000000000fb0e 5.80% swapper ffffffff8102235f [k] 0xffffffff8102235f 3.95% perf libc-2.10.2.so [.] 0x0000000007f398 2.33% init ffffffff810091b9 [k] 0xffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k [.] 0x00000000105440 1.35% find [e1000e] [k] 0x00000000010948 0.68% sleep libc-2.10.2.so [.] 0x0000000011ad5b acme@parisc:~/git/linux-2.6-tip$ But if we: acme@parisc:~/git/linux-2.6-tip$ ls ~/.debug ls: cannot access /home/acme/.debug: No such file or directory acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/lib64/libc-2.10.2.so/ acme@parisc:~/git/linux-2.6-tip$ scp doppio:.debug/lib64/libc-2.10.2.so/* ~/.debug/lib64/libc-2.10.2.so/ acme@doppio's password: eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 100% 1783KB 714.7KB/s 00:02 acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/.build-id/eb acme@parisc:~/git/linux-2.6-tip$ ln -s ../../lib64/libc-2.10.2.so/eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 ~/.debug/.build-id/eb/4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 acme@parisc:~/git/linux-2.6-tip$ perf report --dsos libc-2.10.2.so 2> /dev/null # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # acme@parisc:~/git/linux-2.6-tip$ Which matches what we get on the source, F12, x86_64 machine: [root@doppio linux-2.6-tip]# perf report --dsos libc-2.10.2.so # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # [root@doppio linux-2.6-tip]# So I think this is really, really nice in that it demonstrates the portability of perf.data files and the use of build-ids accross such aliens worlds :-) There are some things to fix tho, like the bitmap on the header, but things are looking good. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1263478990-8200-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-14 22:23:10 +08:00
void perf_event_header__bswap(struct perf_event_header *self)
{
self->type = bswap_32(self->type);
self->misc = bswap_16(self->misc);
self->size = bswap_16(self->size);
}
static struct thread *perf_session__register_idle_thread(struct perf_session *self)
{
struct thread *thread = perf_session__findnew(self, 0);
if (thread == NULL || thread__set_comm(thread, "swapper")) {
pr_err("problem inserting idle task.\n");
thread = NULL;
}
return thread;
}
int do_read(int fd, void *buf, size_t size)
{
void *buf_start = buf;
while (size) {
int ret = read(fd, buf, size);
if (ret <= 0)
return ret;
size -= ret;
buf += ret;
}
return buf - buf_start;
}
#define session_done() (*(volatile int *)(&session_done))
volatile int session_done;
static int __perf_session__process_pipe_events(struct perf_session *self,
struct perf_event_ops *ops)
{
event_t event;
uint32_t size;
int skip = 0;
u64 head;
int err;
void *p;
perf_event_ops__fill_defaults(ops);
head = 0;
more:
err = do_read(self->fd, &event, sizeof(struct perf_event_header));
if (err <= 0) {
if (err == 0)
goto done;
pr_err("failed to read event header\n");
goto out_err;
}
if (self->header.needs_swap)
perf_event_header__bswap(&event.header);
size = event.header.size;
if (size == 0)
size = 8;
p = &event;
p += sizeof(struct perf_event_header);
if (size - sizeof(struct perf_event_header)) {
err = do_read(self->fd, p,
size - sizeof(struct perf_event_header));
if (err <= 0) {
if (err == 0) {
pr_err("unexpected end of event stream\n");
goto done;
}
pr_err("failed to read event data\n");
goto out_err;
}
}
if (size == 0 ||
(skip = perf_session__process_event(self, &event, ops, head)) < 0) {
dump_printf("%#Lx [%#x]: skipping unknown header type: %d\n",
head, event.header.size, event.header.type);
/*
* assume we lost track of the stream, check alignment, and
* increment a single u64 in the hope to catch on again 'soon'.
*/
if (unlikely(head & 7))
head &= ~7ULL;
size = 8;
}
head += size;
dump_printf("\n%#Lx [%#x]: event: %d\n",
head, event.header.size, event.header.type);
if (skip > 0)
head += skip;
if (!session_done())
goto more;
done:
err = 0;
out_err:
return err;
}
int __perf_session__process_events(struct perf_session *session,
u64 data_offset, u64 data_size,
u64 file_size, struct perf_event_ops *ops)
{
u64 head, page_offset, file_offset;
perf tools: Cross platform perf.data analysis support There are still some problems related to loading vmlinux files, but those are unrelated to the feature implemented in this patch, so will get fixed in the next patches, but here are some results: 1. collect perf.data file on a Fedora 12 machine, x86_64, 64-bit userland 2. transfer it to a Debian Testing machine, PARISC64, 32-bit userland acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | head -5 74f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms] 55fdd56670453ea66c011158c4b9d30179c1d049 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/ipt_MASQUERADE.ko 41adff63c730890480980d5d8ba513f1c216a858 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/iptable_nat.ko 90a33def1077bb8e97b8a78546dc96c2de62df46 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/nf_nat.ko 984c7bea90ce1376d5c8e7ef43a781801286e62d /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/tun.ko acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | tail -5 22492f3753c6a67de5c7ccbd6b863390c92c0723 /usr/lib64/libXt.so.6.0.0 353802bb7e1b895ba43507cc678f951e778e4c6f /usr/lib64/libMagickCore.so.2.0.0 d10c2897558595efe7be8b0584cf7e6398bc776c /usr/lib64/libfprint.so.0.0.0 a83ecfb519a788774a84d5ddde633c9ba56c03ab /home/acme/bin/perf d3ca765a8ecf257d263801d7ad8c49c189082317 /usr/lib64/libdwarf.so.0.0 acme@parisc:~/git/linux-2.6-tip$ acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm The file [kernel.kallsyms] cannot be used, trying to use /proc/kallsyms... ^^^^ The problem related to vmlinux handling, it shouldn't be trying this ^^^^ rather alien /proc/kallsyms at all... /lib64/libpthread-2.10.2.so with build id 5c68f7afeb33309c78037e374b0deee84dd441f6 not found, continuing without symbols /lib64/libc-2.10.2.so with build id eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 not found, continuing without symbols /home/acme/bin/perf with build id a83ecfb519a788774a84d5ddde633c9ba56c03ab not found, continuing without symbols /usr/sbin/openvpn with build id f2037a091ef36b591187a858d75e203690ea9409 not found, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwlcore.ko, continuing without symbols <SNIP more complaints about not finding the right build-ids, those will have to wait for 'perf archive' or plain copying what was collected by 'perf record' on the x86_64, source machine, see further below for an example of this > # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 acme@parisc:~/git/linux-2.6-tip$ Which matches what we get when running the same command for the same perf.data file on the F12, x86_64, source machine: [root@doppio linux-2.6-tip]# perf report --sort comm # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 [root@doppio linux-2.6-tip]# The other modes work as well, modulo the problem with vmlinux: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object # ........ ............... ................................. # 35.11% find ffffffff81002b5a 18.25% perf ffffffff8102235f 16.17% find libc-2.10.2.so 9.07% find find 5.80% swapper ffffffff8102235f 3.95% perf libc-2.10.2.so 2.33% init ffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k 1.35% find [e1000e] 0.68% sleep libc-2.10.2.so acme@parisc:~/git/linux-2.6-tip$ And the lack of the right buildids: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso,symbol 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object Symbol # ........ ............... ................................. ...... # 35.11% find ffffffff81002b5a [k] 0xffffffff81002b5a 18.25% perf ffffffff8102235f [k] 0xffffffff8102235f 16.17% find libc-2.10.2.so [.] 0x00000000045782 9.07% find find [.] 0x0000000000fb0e 5.80% swapper ffffffff8102235f [k] 0xffffffff8102235f 3.95% perf libc-2.10.2.so [.] 0x0000000007f398 2.33% init ffffffff810091b9 [k] 0xffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k [.] 0x00000000105440 1.35% find [e1000e] [k] 0x00000000010948 0.68% sleep libc-2.10.2.so [.] 0x0000000011ad5b acme@parisc:~/git/linux-2.6-tip$ But if we: acme@parisc:~/git/linux-2.6-tip$ ls ~/.debug ls: cannot access /home/acme/.debug: No such file or directory acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/lib64/libc-2.10.2.so/ acme@parisc:~/git/linux-2.6-tip$ scp doppio:.debug/lib64/libc-2.10.2.so/* ~/.debug/lib64/libc-2.10.2.so/ acme@doppio's password: eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 100% 1783KB 714.7KB/s 00:02 acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/.build-id/eb acme@parisc:~/git/linux-2.6-tip$ ln -s ../../lib64/libc-2.10.2.so/eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 ~/.debug/.build-id/eb/4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 acme@parisc:~/git/linux-2.6-tip$ perf report --dsos libc-2.10.2.so 2> /dev/null # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # acme@parisc:~/git/linux-2.6-tip$ Which matches what we get on the source, F12, x86_64 machine: [root@doppio linux-2.6-tip]# perf report --dsos libc-2.10.2.so # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # [root@doppio linux-2.6-tip]# So I think this is really, really nice in that it demonstrates the portability of perf.data files and the use of build-ids accross such aliens worlds :-) There are some things to fix tho, like the bitmap on the header, but things are looking good. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1263478990-8200-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-14 22:23:10 +08:00
int err, mmap_prot, mmap_flags;
struct ui_progress *progress;
size_t page_size;
event_t *event;
uint32_t size;
char *buf;
progress = ui_progress__new("Processing events...", session->size);
if (progress == NULL)
return -1;
perf_event_ops__fill_defaults(ops);
page_size = sysconf(_SC_PAGESIZE);
page_offset = page_size * (data_offset / page_size);
file_offset = page_offset;
head = data_offset - page_offset;
perf tools: Cross platform perf.data analysis support There are still some problems related to loading vmlinux files, but those are unrelated to the feature implemented in this patch, so will get fixed in the next patches, but here are some results: 1. collect perf.data file on a Fedora 12 machine, x86_64, 64-bit userland 2. transfer it to a Debian Testing machine, PARISC64, 32-bit userland acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | head -5 74f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms] 55fdd56670453ea66c011158c4b9d30179c1d049 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/ipt_MASQUERADE.ko 41adff63c730890480980d5d8ba513f1c216a858 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/iptable_nat.ko 90a33def1077bb8e97b8a78546dc96c2de62df46 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/nf_nat.ko 984c7bea90ce1376d5c8e7ef43a781801286e62d /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/tun.ko acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | tail -5 22492f3753c6a67de5c7ccbd6b863390c92c0723 /usr/lib64/libXt.so.6.0.0 353802bb7e1b895ba43507cc678f951e778e4c6f /usr/lib64/libMagickCore.so.2.0.0 d10c2897558595efe7be8b0584cf7e6398bc776c /usr/lib64/libfprint.so.0.0.0 a83ecfb519a788774a84d5ddde633c9ba56c03ab /home/acme/bin/perf d3ca765a8ecf257d263801d7ad8c49c189082317 /usr/lib64/libdwarf.so.0.0 acme@parisc:~/git/linux-2.6-tip$ acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm The file [kernel.kallsyms] cannot be used, trying to use /proc/kallsyms... ^^^^ The problem related to vmlinux handling, it shouldn't be trying this ^^^^ rather alien /proc/kallsyms at all... /lib64/libpthread-2.10.2.so with build id 5c68f7afeb33309c78037e374b0deee84dd441f6 not found, continuing without symbols /lib64/libc-2.10.2.so with build id eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 not found, continuing without symbols /home/acme/bin/perf with build id a83ecfb519a788774a84d5ddde633c9ba56c03ab not found, continuing without symbols /usr/sbin/openvpn with build id f2037a091ef36b591187a858d75e203690ea9409 not found, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwlcore.ko, continuing without symbols <SNIP more complaints about not finding the right build-ids, those will have to wait for 'perf archive' or plain copying what was collected by 'perf record' on the x86_64, source machine, see further below for an example of this > # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 acme@parisc:~/git/linux-2.6-tip$ Which matches what we get when running the same command for the same perf.data file on the F12, x86_64, source machine: [root@doppio linux-2.6-tip]# perf report --sort comm # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 [root@doppio linux-2.6-tip]# The other modes work as well, modulo the problem with vmlinux: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object # ........ ............... ................................. # 35.11% find ffffffff81002b5a 18.25% perf ffffffff8102235f 16.17% find libc-2.10.2.so 9.07% find find 5.80% swapper ffffffff8102235f 3.95% perf libc-2.10.2.so 2.33% init ffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k 1.35% find [e1000e] 0.68% sleep libc-2.10.2.so acme@parisc:~/git/linux-2.6-tip$ And the lack of the right buildids: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso,symbol 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object Symbol # ........ ............... ................................. ...... # 35.11% find ffffffff81002b5a [k] 0xffffffff81002b5a 18.25% perf ffffffff8102235f [k] 0xffffffff8102235f 16.17% find libc-2.10.2.so [.] 0x00000000045782 9.07% find find [.] 0x0000000000fb0e 5.80% swapper ffffffff8102235f [k] 0xffffffff8102235f 3.95% perf libc-2.10.2.so [.] 0x0000000007f398 2.33% init ffffffff810091b9 [k] 0xffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k [.] 0x00000000105440 1.35% find [e1000e] [k] 0x00000000010948 0.68% sleep libc-2.10.2.so [.] 0x0000000011ad5b acme@parisc:~/git/linux-2.6-tip$ But if we: acme@parisc:~/git/linux-2.6-tip$ ls ~/.debug ls: cannot access /home/acme/.debug: No such file or directory acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/lib64/libc-2.10.2.so/ acme@parisc:~/git/linux-2.6-tip$ scp doppio:.debug/lib64/libc-2.10.2.so/* ~/.debug/lib64/libc-2.10.2.so/ acme@doppio's password: eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 100% 1783KB 714.7KB/s 00:02 acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/.build-id/eb acme@parisc:~/git/linux-2.6-tip$ ln -s ../../lib64/libc-2.10.2.so/eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 ~/.debug/.build-id/eb/4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 acme@parisc:~/git/linux-2.6-tip$ perf report --dsos libc-2.10.2.so 2> /dev/null # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # acme@parisc:~/git/linux-2.6-tip$ Which matches what we get on the source, F12, x86_64 machine: [root@doppio linux-2.6-tip]# perf report --dsos libc-2.10.2.so # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # [root@doppio linux-2.6-tip]# So I think this is really, really nice in that it demonstrates the portability of perf.data files and the use of build-ids accross such aliens worlds :-) There are some things to fix tho, like the bitmap on the header, but things are looking good. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1263478990-8200-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-14 22:23:10 +08:00
mmap_prot = PROT_READ;
mmap_flags = MAP_SHARED;
if (session->header.needs_swap) {
perf tools: Cross platform perf.data analysis support There are still some problems related to loading vmlinux files, but those are unrelated to the feature implemented in this patch, so will get fixed in the next patches, but here are some results: 1. collect perf.data file on a Fedora 12 machine, x86_64, 64-bit userland 2. transfer it to a Debian Testing machine, PARISC64, 32-bit userland acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | head -5 74f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms] 55fdd56670453ea66c011158c4b9d30179c1d049 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/ipt_MASQUERADE.ko 41adff63c730890480980d5d8ba513f1c216a858 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/iptable_nat.ko 90a33def1077bb8e97b8a78546dc96c2de62df46 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/nf_nat.ko 984c7bea90ce1376d5c8e7ef43a781801286e62d /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/tun.ko acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | tail -5 22492f3753c6a67de5c7ccbd6b863390c92c0723 /usr/lib64/libXt.so.6.0.0 353802bb7e1b895ba43507cc678f951e778e4c6f /usr/lib64/libMagickCore.so.2.0.0 d10c2897558595efe7be8b0584cf7e6398bc776c /usr/lib64/libfprint.so.0.0.0 a83ecfb519a788774a84d5ddde633c9ba56c03ab /home/acme/bin/perf d3ca765a8ecf257d263801d7ad8c49c189082317 /usr/lib64/libdwarf.so.0.0 acme@parisc:~/git/linux-2.6-tip$ acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm The file [kernel.kallsyms] cannot be used, trying to use /proc/kallsyms... ^^^^ The problem related to vmlinux handling, it shouldn't be trying this ^^^^ rather alien /proc/kallsyms at all... /lib64/libpthread-2.10.2.so with build id 5c68f7afeb33309c78037e374b0deee84dd441f6 not found, continuing without symbols /lib64/libc-2.10.2.so with build id eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 not found, continuing without symbols /home/acme/bin/perf with build id a83ecfb519a788774a84d5ddde633c9ba56c03ab not found, continuing without symbols /usr/sbin/openvpn with build id f2037a091ef36b591187a858d75e203690ea9409 not found, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwlcore.ko, continuing without symbols <SNIP more complaints about not finding the right build-ids, those will have to wait for 'perf archive' or plain copying what was collected by 'perf record' on the x86_64, source machine, see further below for an example of this > # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 acme@parisc:~/git/linux-2.6-tip$ Which matches what we get when running the same command for the same perf.data file on the F12, x86_64, source machine: [root@doppio linux-2.6-tip]# perf report --sort comm # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 [root@doppio linux-2.6-tip]# The other modes work as well, modulo the problem with vmlinux: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object # ........ ............... ................................. # 35.11% find ffffffff81002b5a 18.25% perf ffffffff8102235f 16.17% find libc-2.10.2.so 9.07% find find 5.80% swapper ffffffff8102235f 3.95% perf libc-2.10.2.so 2.33% init ffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k 1.35% find [e1000e] 0.68% sleep libc-2.10.2.so acme@parisc:~/git/linux-2.6-tip$ And the lack of the right buildids: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso,symbol 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object Symbol # ........ ............... ................................. ...... # 35.11% find ffffffff81002b5a [k] 0xffffffff81002b5a 18.25% perf ffffffff8102235f [k] 0xffffffff8102235f 16.17% find libc-2.10.2.so [.] 0x00000000045782 9.07% find find [.] 0x0000000000fb0e 5.80% swapper ffffffff8102235f [k] 0xffffffff8102235f 3.95% perf libc-2.10.2.so [.] 0x0000000007f398 2.33% init ffffffff810091b9 [k] 0xffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k [.] 0x00000000105440 1.35% find [e1000e] [k] 0x00000000010948 0.68% sleep libc-2.10.2.so [.] 0x0000000011ad5b acme@parisc:~/git/linux-2.6-tip$ But if we: acme@parisc:~/git/linux-2.6-tip$ ls ~/.debug ls: cannot access /home/acme/.debug: No such file or directory acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/lib64/libc-2.10.2.so/ acme@parisc:~/git/linux-2.6-tip$ scp doppio:.debug/lib64/libc-2.10.2.so/* ~/.debug/lib64/libc-2.10.2.so/ acme@doppio's password: eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 100% 1783KB 714.7KB/s 00:02 acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/.build-id/eb acme@parisc:~/git/linux-2.6-tip$ ln -s ../../lib64/libc-2.10.2.so/eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 ~/.debug/.build-id/eb/4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 acme@parisc:~/git/linux-2.6-tip$ perf report --dsos libc-2.10.2.so 2> /dev/null # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # acme@parisc:~/git/linux-2.6-tip$ Which matches what we get on the source, F12, x86_64 machine: [root@doppio linux-2.6-tip]# perf report --dsos libc-2.10.2.so # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # [root@doppio linux-2.6-tip]# So I think this is really, really nice in that it demonstrates the portability of perf.data files and the use of build-ids accross such aliens worlds :-) There are some things to fix tho, like the bitmap on the header, but things are looking good. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1263478990-8200-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-14 22:23:10 +08:00
mmap_prot |= PROT_WRITE;
mmap_flags = MAP_PRIVATE;
}
remap:
buf = mmap(NULL, page_size * session->mmap_window, mmap_prot,
mmap_flags, session->fd, file_offset);
if (buf == MAP_FAILED) {
pr_err("failed to mmap file\n");
err = -errno;
goto out_err;
}
more:
event = (event_t *)(buf + head);
ui_progress__update(progress, file_offset);
if (session->header.needs_swap)
perf tools: Cross platform perf.data analysis support There are still some problems related to loading vmlinux files, but those are unrelated to the feature implemented in this patch, so will get fixed in the next patches, but here are some results: 1. collect perf.data file on a Fedora 12 machine, x86_64, 64-bit userland 2. transfer it to a Debian Testing machine, PARISC64, 32-bit userland acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | head -5 74f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms] 55fdd56670453ea66c011158c4b9d30179c1d049 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/ipt_MASQUERADE.ko 41adff63c730890480980d5d8ba513f1c216a858 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/iptable_nat.ko 90a33def1077bb8e97b8a78546dc96c2de62df46 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/nf_nat.ko 984c7bea90ce1376d5c8e7ef43a781801286e62d /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/tun.ko acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | tail -5 22492f3753c6a67de5c7ccbd6b863390c92c0723 /usr/lib64/libXt.so.6.0.0 353802bb7e1b895ba43507cc678f951e778e4c6f /usr/lib64/libMagickCore.so.2.0.0 d10c2897558595efe7be8b0584cf7e6398bc776c /usr/lib64/libfprint.so.0.0.0 a83ecfb519a788774a84d5ddde633c9ba56c03ab /home/acme/bin/perf d3ca765a8ecf257d263801d7ad8c49c189082317 /usr/lib64/libdwarf.so.0.0 acme@parisc:~/git/linux-2.6-tip$ acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm The file [kernel.kallsyms] cannot be used, trying to use /proc/kallsyms... ^^^^ The problem related to vmlinux handling, it shouldn't be trying this ^^^^ rather alien /proc/kallsyms at all... /lib64/libpthread-2.10.2.so with build id 5c68f7afeb33309c78037e374b0deee84dd441f6 not found, continuing without symbols /lib64/libc-2.10.2.so with build id eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 not found, continuing without symbols /home/acme/bin/perf with build id a83ecfb519a788774a84d5ddde633c9ba56c03ab not found, continuing without symbols /usr/sbin/openvpn with build id f2037a091ef36b591187a858d75e203690ea9409 not found, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwlcore.ko, continuing without symbols <SNIP more complaints about not finding the right build-ids, those will have to wait for 'perf archive' or plain copying what was collected by 'perf record' on the x86_64, source machine, see further below for an example of this > # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 acme@parisc:~/git/linux-2.6-tip$ Which matches what we get when running the same command for the same perf.data file on the F12, x86_64, source machine: [root@doppio linux-2.6-tip]# perf report --sort comm # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 [root@doppio linux-2.6-tip]# The other modes work as well, modulo the problem with vmlinux: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object # ........ ............... ................................. # 35.11% find ffffffff81002b5a 18.25% perf ffffffff8102235f 16.17% find libc-2.10.2.so 9.07% find find 5.80% swapper ffffffff8102235f 3.95% perf libc-2.10.2.so 2.33% init ffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k 1.35% find [e1000e] 0.68% sleep libc-2.10.2.so acme@parisc:~/git/linux-2.6-tip$ And the lack of the right buildids: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso,symbol 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object Symbol # ........ ............... ................................. ...... # 35.11% find ffffffff81002b5a [k] 0xffffffff81002b5a 18.25% perf ffffffff8102235f [k] 0xffffffff8102235f 16.17% find libc-2.10.2.so [.] 0x00000000045782 9.07% find find [.] 0x0000000000fb0e 5.80% swapper ffffffff8102235f [k] 0xffffffff8102235f 3.95% perf libc-2.10.2.so [.] 0x0000000007f398 2.33% init ffffffff810091b9 [k] 0xffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k [.] 0x00000000105440 1.35% find [e1000e] [k] 0x00000000010948 0.68% sleep libc-2.10.2.so [.] 0x0000000011ad5b acme@parisc:~/git/linux-2.6-tip$ But if we: acme@parisc:~/git/linux-2.6-tip$ ls ~/.debug ls: cannot access /home/acme/.debug: No such file or directory acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/lib64/libc-2.10.2.so/ acme@parisc:~/git/linux-2.6-tip$ scp doppio:.debug/lib64/libc-2.10.2.so/* ~/.debug/lib64/libc-2.10.2.so/ acme@doppio's password: eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 100% 1783KB 714.7KB/s 00:02 acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/.build-id/eb acme@parisc:~/git/linux-2.6-tip$ ln -s ../../lib64/libc-2.10.2.so/eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 ~/.debug/.build-id/eb/4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 acme@parisc:~/git/linux-2.6-tip$ perf report --dsos libc-2.10.2.so 2> /dev/null # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # acme@parisc:~/git/linux-2.6-tip$ Which matches what we get on the source, F12, x86_64 machine: [root@doppio linux-2.6-tip]# perf report --dsos libc-2.10.2.so # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # [root@doppio linux-2.6-tip]# So I think this is really, really nice in that it demonstrates the portability of perf.data files and the use of build-ids accross such aliens worlds :-) There are some things to fix tho, like the bitmap on the header, but things are looking good. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1263478990-8200-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-14 22:23:10 +08:00
perf_event_header__bswap(&event->header);
size = event->header.size;
if (size == 0)
size = 8;
if (head + event->header.size >= page_size * session->mmap_window) {
int munmap_ret;
munmap_ret = munmap(buf, page_size * session->mmap_window);
assert(munmap_ret == 0);
page_offset = page_size * (head / page_size);
file_offset += page_offset;
head -= page_offset;
goto remap;
}
size = event->header.size;
perf tools: Cross platform perf.data analysis support There are still some problems related to loading vmlinux files, but those are unrelated to the feature implemented in this patch, so will get fixed in the next patches, but here are some results: 1. collect perf.data file on a Fedora 12 machine, x86_64, 64-bit userland 2. transfer it to a Debian Testing machine, PARISC64, 32-bit userland acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | head -5 74f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms] 55fdd56670453ea66c011158c4b9d30179c1d049 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/ipt_MASQUERADE.ko 41adff63c730890480980d5d8ba513f1c216a858 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/iptable_nat.ko 90a33def1077bb8e97b8a78546dc96c2de62df46 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/nf_nat.ko 984c7bea90ce1376d5c8e7ef43a781801286e62d /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/tun.ko acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | tail -5 22492f3753c6a67de5c7ccbd6b863390c92c0723 /usr/lib64/libXt.so.6.0.0 353802bb7e1b895ba43507cc678f951e778e4c6f /usr/lib64/libMagickCore.so.2.0.0 d10c2897558595efe7be8b0584cf7e6398bc776c /usr/lib64/libfprint.so.0.0.0 a83ecfb519a788774a84d5ddde633c9ba56c03ab /home/acme/bin/perf d3ca765a8ecf257d263801d7ad8c49c189082317 /usr/lib64/libdwarf.so.0.0 acme@parisc:~/git/linux-2.6-tip$ acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm The file [kernel.kallsyms] cannot be used, trying to use /proc/kallsyms... ^^^^ The problem related to vmlinux handling, it shouldn't be trying this ^^^^ rather alien /proc/kallsyms at all... /lib64/libpthread-2.10.2.so with build id 5c68f7afeb33309c78037e374b0deee84dd441f6 not found, continuing without symbols /lib64/libc-2.10.2.so with build id eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 not found, continuing without symbols /home/acme/bin/perf with build id a83ecfb519a788774a84d5ddde633c9ba56c03ab not found, continuing without symbols /usr/sbin/openvpn with build id f2037a091ef36b591187a858d75e203690ea9409 not found, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwlcore.ko, continuing without symbols <SNIP more complaints about not finding the right build-ids, those will have to wait for 'perf archive' or plain copying what was collected by 'perf record' on the x86_64, source machine, see further below for an example of this > # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 acme@parisc:~/git/linux-2.6-tip$ Which matches what we get when running the same command for the same perf.data file on the F12, x86_64, source machine: [root@doppio linux-2.6-tip]# perf report --sort comm # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 [root@doppio linux-2.6-tip]# The other modes work as well, modulo the problem with vmlinux: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object # ........ ............... ................................. # 35.11% find ffffffff81002b5a 18.25% perf ffffffff8102235f 16.17% find libc-2.10.2.so 9.07% find find 5.80% swapper ffffffff8102235f 3.95% perf libc-2.10.2.so 2.33% init ffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k 1.35% find [e1000e] 0.68% sleep libc-2.10.2.so acme@parisc:~/git/linux-2.6-tip$ And the lack of the right buildids: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso,symbol 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object Symbol # ........ ............... ................................. ...... # 35.11% find ffffffff81002b5a [k] 0xffffffff81002b5a 18.25% perf ffffffff8102235f [k] 0xffffffff8102235f 16.17% find libc-2.10.2.so [.] 0x00000000045782 9.07% find find [.] 0x0000000000fb0e 5.80% swapper ffffffff8102235f [k] 0xffffffff8102235f 3.95% perf libc-2.10.2.so [.] 0x0000000007f398 2.33% init ffffffff810091b9 [k] 0xffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k [.] 0x00000000105440 1.35% find [e1000e] [k] 0x00000000010948 0.68% sleep libc-2.10.2.so [.] 0x0000000011ad5b acme@parisc:~/git/linux-2.6-tip$ But if we: acme@parisc:~/git/linux-2.6-tip$ ls ~/.debug ls: cannot access /home/acme/.debug: No such file or directory acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/lib64/libc-2.10.2.so/ acme@parisc:~/git/linux-2.6-tip$ scp doppio:.debug/lib64/libc-2.10.2.so/* ~/.debug/lib64/libc-2.10.2.so/ acme@doppio's password: eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 100% 1783KB 714.7KB/s 00:02 acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/.build-id/eb acme@parisc:~/git/linux-2.6-tip$ ln -s ../../lib64/libc-2.10.2.so/eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 ~/.debug/.build-id/eb/4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 acme@parisc:~/git/linux-2.6-tip$ perf report --dsos libc-2.10.2.so 2> /dev/null # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # acme@parisc:~/git/linux-2.6-tip$ Which matches what we get on the source, F12, x86_64 machine: [root@doppio linux-2.6-tip]# perf report --dsos libc-2.10.2.so # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # [root@doppio linux-2.6-tip]# So I think this is really, really nice in that it demonstrates the portability of perf.data files and the use of build-ids accross such aliens worlds :-) There are some things to fix tho, like the bitmap on the header, but things are looking good. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1263478990-8200-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-14 22:23:10 +08:00
dump_printf("\n%#Lx [%#x]: event: %d\n",
file_offset + head, event->header.size, event->header.type);
if (size == 0 || perf_session__process_event(session, event, ops,
file_offset + head) < 0) {
perf tools: Cross platform perf.data analysis support There are still some problems related to loading vmlinux files, but those are unrelated to the feature implemented in this patch, so will get fixed in the next patches, but here are some results: 1. collect perf.data file on a Fedora 12 machine, x86_64, 64-bit userland 2. transfer it to a Debian Testing machine, PARISC64, 32-bit userland acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | head -5 74f9930ee94475b6b3238caf3725a50d59cb994b [kernel.kallsyms] 55fdd56670453ea66c011158c4b9d30179c1d049 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/ipt_MASQUERADE.ko 41adff63c730890480980d5d8ba513f1c216a858 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/iptable_nat.ko 90a33def1077bb8e97b8a78546dc96c2de62df46 /lib/modules/2.6.33-rc4-tip+/kernel/net/ipv4/netfilter/nf_nat.ko 984c7bea90ce1376d5c8e7ef43a781801286e62d /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/tun.ko acme@parisc:~/git/linux-2.6-tip$ perf buildid-list | tail -5 22492f3753c6a67de5c7ccbd6b863390c92c0723 /usr/lib64/libXt.so.6.0.0 353802bb7e1b895ba43507cc678f951e778e4c6f /usr/lib64/libMagickCore.so.2.0.0 d10c2897558595efe7be8b0584cf7e6398bc776c /usr/lib64/libfprint.so.0.0.0 a83ecfb519a788774a84d5ddde633c9ba56c03ab /home/acme/bin/perf d3ca765a8ecf257d263801d7ad8c49c189082317 /usr/lib64/libdwarf.so.0.0 acme@parisc:~/git/linux-2.6-tip$ acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm The file [kernel.kallsyms] cannot be used, trying to use /proc/kallsyms... ^^^^ The problem related to vmlinux handling, it shouldn't be trying this ^^^^ rather alien /proc/kallsyms at all... /lib64/libpthread-2.10.2.so with build id 5c68f7afeb33309c78037e374b0deee84dd441f6 not found, continuing without symbols /lib64/libc-2.10.2.so with build id eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 not found, continuing without symbols /home/acme/bin/perf with build id a83ecfb519a788774a84d5ddde633c9ba56c03ab not found, continuing without symbols /usr/sbin/openvpn with build id f2037a091ef36b591187a858d75e203690ea9409 not found, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/e1000e/e1000e.ko, continuing without symbols Failed to open /lib/modules/2.6.33-rc4-tip+/kernel/drivers/net/wireless/iwlwifi/iwlcore.ko, continuing without symbols <SNIP more complaints about not finding the right build-ids, those will have to wait for 'perf archive' or plain copying what was collected by 'perf record' on the x86_64, source machine, see further below for an example of this > # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 acme@parisc:~/git/linux-2.6-tip$ Which matches what we get when running the same command for the same perf.data file on the F12, x86_64, source machine: [root@doppio linux-2.6-tip]# perf report --sort comm # Samples: 293085637 # # Overhead Command # ........ ............... # 61.70% find 23.50% perf 5.86% swapper 3.12% sshd 2.39% init 0.87% bash 0.86% sleep 0.59% dbus-daemon 0.25% hald 0.24% NetworkManager 0.19% hald-addon-rfki 0.15% openvpn 0.07% phy0 0.07% events/0 0.05% iwl3945 0.05% events/1 0.03% kondemand/0 [root@doppio linux-2.6-tip]# The other modes work as well, modulo the problem with vmlinux: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object # ........ ............... ................................. # 35.11% find ffffffff81002b5a 18.25% perf ffffffff8102235f 16.17% find libc-2.10.2.so 9.07% find find 5.80% swapper ffffffff8102235f 3.95% perf libc-2.10.2.so 2.33% init ffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k 1.35% find [e1000e] 0.68% sleep libc-2.10.2.so acme@parisc:~/git/linux-2.6-tip$ And the lack of the right buildids: acme@parisc:~/git/linux-2.6-tip$ perf report --sort comm,dso,symbol 2> /dev/null | head -15 # Samples: 293085637 # # Overhead Command Shared Object Symbol # ........ ............... ................................. ...... # 35.11% find ffffffff81002b5a [k] 0xffffffff81002b5a 18.25% perf ffffffff8102235f [k] 0xffffffff8102235f 16.17% find libc-2.10.2.so [.] 0x00000000045782 9.07% find find [.] 0x0000000000fb0e 5.80% swapper ffffffff8102235f [k] 0xffffffff8102235f 3.95% perf libc-2.10.2.so [.] 0x0000000007f398 2.33% init ffffffff810091b9 [k] 0xffffffff810091b9 1.65% sshd libcrypto.so.0.9.8k [.] 0x00000000105440 1.35% find [e1000e] [k] 0x00000000010948 0.68% sleep libc-2.10.2.so [.] 0x0000000011ad5b acme@parisc:~/git/linux-2.6-tip$ But if we: acme@parisc:~/git/linux-2.6-tip$ ls ~/.debug ls: cannot access /home/acme/.debug: No such file or directory acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/lib64/libc-2.10.2.so/ acme@parisc:~/git/linux-2.6-tip$ scp doppio:.debug/lib64/libc-2.10.2.so/* ~/.debug/lib64/libc-2.10.2.so/ acme@doppio's password: eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 100% 1783KB 714.7KB/s 00:02 acme@parisc:~/git/linux-2.6-tip$ mkdir -p ~/.debug/.build-id/eb acme@parisc:~/git/linux-2.6-tip$ ln -s ../../lib64/libc-2.10.2.so/eb4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 ~/.debug/.build-id/eb/4ec8fa8b2a5eb18cad173c92f27ed8887ed1c1 acme@parisc:~/git/linux-2.6-tip$ perf report --dsos libc-2.10.2.so 2> /dev/null # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # acme@parisc:~/git/linux-2.6-tip$ Which matches what we get on the source, F12, x86_64 machine: [root@doppio linux-2.6-tip]# perf report --dsos libc-2.10.2.so # dso: libc-2.10.2.so # Samples: 64281170 # # Overhead Command Symbol # ........ ............... ...... # 14.98% perf [.] __GI_strcmp 12.30% find [.] __GI_memmove 9.25% find [.] _int_malloc 7.60% find [.] _IO_vfprintf_internal 6.10% find [.] _IO_new_file_xsputn 6.02% find [.] __GI_close 3.08% find [.] _IO_file_overflow_internal 3.08% find [.] malloc_consolidate 3.08% find [.] _int_free 3.08% find [.] __strchrnul 3.08% find [.] __getdents64 3.08% find [.] __write_nocancel 3.08% sleep [.] __GI__dl_addr 3.08% sshd [.] __libc_select 3.08% find [.] _IO_new_file_write 3.07% find [.] _IO_new_do_write 3.06% find [.] __GI___errno_location 3.05% find [.] __GI___libc_malloc 3.04% perf [.] __GI_memcpy 1.71% find [.] __fprintf_chk 1.29% bash [.] __gconv_transform_utf8_internal 0.79% dbus-daemon [.] __GI_strlen # # (For a higher level overview, try: perf report --sort comm,dso) # [root@doppio linux-2.6-tip]# So I think this is really, really nice in that it demonstrates the portability of perf.data files and the use of build-ids accross such aliens worlds :-) There are some things to fix tho, like the bitmap on the header, but things are looking good. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> LKML-Reference: <1263478990-8200-2-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-14 22:23:10 +08:00
dump_printf("%#Lx [%#x]: skipping unknown header type: %d\n",
file_offset + head, event->header.size,
event->header.type);
/*
* assume we lost track of the stream, check alignment, and
* increment a single u64 in the hope to catch on again 'soon'.
*/
if (unlikely(head & 7))
head &= ~7ULL;
size = 8;
}
head += size;
if (file_offset + head >= data_offset + data_size)
goto done;
if (file_offset + head < file_size)
goto more;
done:
err = 0;
perf: Generalize perf lock's sample event reordering to the session layer The sample events recorded by perf record are not time ordered because we have one buffer per cpu for each event (even demultiplexed per task/per cpu for task bound events). But when we read trace events we want them to be ordered by time because many state machines are involved. There are currently two ways perf tools deal with that: - use -M to multiplex every buffers (perf sched, perf kmem) But this creates a lot of contention in SMP machines on record time. - use a post-processing time reordering (perf timechart, perf lock) The reordering used by timechart is simple but doesn't scale well with huge flow of events, in terms of performance and memory use (unusable with perf lock for example). Perf lock has its own samples reordering that flushes its memory use in a regular basis and that uses a sorting based on the previous event queued (a new event to be queued is close to the previous one most of the time). This patch proposes to export perf lock's samples reordering facility to the session layer that reads the events. So if a tool wants to get ordered sample events, it needs to set its struct perf_event_ops::ordered_samples to true and that's it. This prepares tracing based perf tools to get rid of the need to use buffers multiplexing (-M) or to implement their own reordering. Also lower the flush period to 2 as it's sufficient already. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp> Cc: Ingo Molnar <mingo@elte.hu> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com>
2010-04-24 06:04:12 +08:00
/* do the final flush for ordered samples */
session->ordered_samples.next_flush = ULLONG_MAX;
flush_sample_queue(session, ops);
out_err:
ui_progress__delete(progress);
if (ops->lost == event__process_lost &&
session->hists.stats.total_lost != 0) {
ui__warning("Processed %Lu events and LOST %Lu!\n\n"
"Check IO/CPU overload!\n\n",
session->hists.stats.total_period,
session->hists.stats.total_lost);
}
if (session->hists.stats.nr_unknown_events != 0) {
ui__warning("Found %u unknown events!\n\n"
"Is this an older tool processing a perf.data "
"file generated by a more recent tool?\n\n"
"If that is not the case, consider "
"reporting to linux-kernel@vger.kernel.org.\n\n",
session->hists.stats.nr_unknown_events);
}
return err;
}
int perf_session__process_events(struct perf_session *self,
struct perf_event_ops *ops)
{
int err;
if (perf_session__register_idle_thread(self) == NULL)
return -ENOMEM;
if (!self->fd_pipe)
err = __perf_session__process_events(self,
self->header.data_offset,
self->header.data_size,
self->size, ops);
else
err = __perf_session__process_pipe_events(self, ops);
return err;
}
bool perf_session__has_traces(struct perf_session *self, const char *msg)
{
if (!(self->sample_type & PERF_SAMPLE_RAW)) {
pr_err("No trace sample to read. Did you call 'perf %s'?\n", msg);
return false;
}
return true;
}
perf tools: Handle relocatable kernels DSOs don't have this problem because the kernel emits a PERF_MMAP for each new executable mapping it performs on monitored threads. To fix the kernel case we simulate the same behaviour, by having 'perf record' to synthesize a PERF_MMAP for the kernel, encoded like this: [root@doppio ~]# perf record -a -f sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.344 MB perf.data (~15038 samples) ] [root@doppio ~]# perf report -D | head -10 0xd0 [0x40]: event: 1 . . ... raw event: size 64 bytes . 0000: 01 00 00 00 00 00 40 00 00 00 00 00 00 00 00 00 ......@........ . 0010: 00 00 00 81 ff ff ff ff 00 00 00 00 00 00 00 00 ............... . 0020: 00 00 00 00 00 00 00 00 5b 6b 65 72 6e 65 6c 2e ........ [kernel . 0030: 6b 61 6c 6c 73 79 6d 73 2e 5f 74 65 78 74 5d 00 kallsyms._text] . 0xd0 [0x40]: PERF_RECORD_MMAP 0/0: [0xffffffff81000000((nil)) @ (nil)]: [kernel.kallsyms._text] I.e. we identify such event as having: .pid = 0 .filename = [kernel.kallsyms.REFNAME] .start = REFNAME addr in /proc/kallsyms at 'perf record' time and use now a hardcoded value of '.text' for REFNAME. Then, later, in 'perf report', if there are any kernel hits and thus we need to resolve kernel symbols, we search for REFNAME and if its address changed, relocation happened and we thus must change the kernel mapping routines to one that uses .pgoff as the relocation to apply. This way we use the same mechanism used for the other DSOs and don't have to do a two pass in all the kernel symbols. Reported-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> LKML-Reference: <1262717431-1246-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-06 02:50:31 +08:00
int perf_session__set_kallsyms_ref_reloc_sym(struct map **maps,
perf tools: Handle relocatable kernels DSOs don't have this problem because the kernel emits a PERF_MMAP for each new executable mapping it performs on monitored threads. To fix the kernel case we simulate the same behaviour, by having 'perf record' to synthesize a PERF_MMAP for the kernel, encoded like this: [root@doppio ~]# perf record -a -f sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.344 MB perf.data (~15038 samples) ] [root@doppio ~]# perf report -D | head -10 0xd0 [0x40]: event: 1 . . ... raw event: size 64 bytes . 0000: 01 00 00 00 00 00 40 00 00 00 00 00 00 00 00 00 ......@........ . 0010: 00 00 00 81 ff ff ff ff 00 00 00 00 00 00 00 00 ............... . 0020: 00 00 00 00 00 00 00 00 5b 6b 65 72 6e 65 6c 2e ........ [kernel . 0030: 6b 61 6c 6c 73 79 6d 73 2e 5f 74 65 78 74 5d 00 kallsyms._text] . 0xd0 [0x40]: PERF_RECORD_MMAP 0/0: [0xffffffff81000000((nil)) @ (nil)]: [kernel.kallsyms._text] I.e. we identify such event as having: .pid = 0 .filename = [kernel.kallsyms.REFNAME] .start = REFNAME addr in /proc/kallsyms at 'perf record' time and use now a hardcoded value of '.text' for REFNAME. Then, later, in 'perf report', if there are any kernel hits and thus we need to resolve kernel symbols, we search for REFNAME and if its address changed, relocation happened and we thus must change the kernel mapping routines to one that uses .pgoff as the relocation to apply. This way we use the same mechanism used for the other DSOs and don't have to do a two pass in all the kernel symbols. Reported-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> LKML-Reference: <1262717431-1246-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-06 02:50:31 +08:00
const char *symbol_name,
u64 addr)
{
char *bracket;
enum map_type i;
struct ref_reloc_sym *ref;
ref = zalloc(sizeof(struct ref_reloc_sym));
if (ref == NULL)
return -ENOMEM;
perf tools: Handle relocatable kernels DSOs don't have this problem because the kernel emits a PERF_MMAP for each new executable mapping it performs on monitored threads. To fix the kernel case we simulate the same behaviour, by having 'perf record' to synthesize a PERF_MMAP for the kernel, encoded like this: [root@doppio ~]# perf record -a -f sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.344 MB perf.data (~15038 samples) ] [root@doppio ~]# perf report -D | head -10 0xd0 [0x40]: event: 1 . . ... raw event: size 64 bytes . 0000: 01 00 00 00 00 00 40 00 00 00 00 00 00 00 00 00 ......@........ . 0010: 00 00 00 81 ff ff ff ff 00 00 00 00 00 00 00 00 ............... . 0020: 00 00 00 00 00 00 00 00 5b 6b 65 72 6e 65 6c 2e ........ [kernel . 0030: 6b 61 6c 6c 73 79 6d 73 2e 5f 74 65 78 74 5d 00 kallsyms._text] . 0xd0 [0x40]: PERF_RECORD_MMAP 0/0: [0xffffffff81000000((nil)) @ (nil)]: [kernel.kallsyms._text] I.e. we identify such event as having: .pid = 0 .filename = [kernel.kallsyms.REFNAME] .start = REFNAME addr in /proc/kallsyms at 'perf record' time and use now a hardcoded value of '.text' for REFNAME. Then, later, in 'perf report', if there are any kernel hits and thus we need to resolve kernel symbols, we search for REFNAME and if its address changed, relocation happened and we thus must change the kernel mapping routines to one that uses .pgoff as the relocation to apply. This way we use the same mechanism used for the other DSOs and don't have to do a two pass in all the kernel symbols. Reported-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> LKML-Reference: <1262717431-1246-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-06 02:50:31 +08:00
ref->name = strdup(symbol_name);
if (ref->name == NULL) {
free(ref);
perf tools: Handle relocatable kernels DSOs don't have this problem because the kernel emits a PERF_MMAP for each new executable mapping it performs on monitored threads. To fix the kernel case we simulate the same behaviour, by having 'perf record' to synthesize a PERF_MMAP for the kernel, encoded like this: [root@doppio ~]# perf record -a -f sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.344 MB perf.data (~15038 samples) ] [root@doppio ~]# perf report -D | head -10 0xd0 [0x40]: event: 1 . . ... raw event: size 64 bytes . 0000: 01 00 00 00 00 00 40 00 00 00 00 00 00 00 00 00 ......@........ . 0010: 00 00 00 81 ff ff ff ff 00 00 00 00 00 00 00 00 ............... . 0020: 00 00 00 00 00 00 00 00 5b 6b 65 72 6e 65 6c 2e ........ [kernel . 0030: 6b 61 6c 6c 73 79 6d 73 2e 5f 74 65 78 74 5d 00 kallsyms._text] . 0xd0 [0x40]: PERF_RECORD_MMAP 0/0: [0xffffffff81000000((nil)) @ (nil)]: [kernel.kallsyms._text] I.e. we identify such event as having: .pid = 0 .filename = [kernel.kallsyms.REFNAME] .start = REFNAME addr in /proc/kallsyms at 'perf record' time and use now a hardcoded value of '.text' for REFNAME. Then, later, in 'perf report', if there are any kernel hits and thus we need to resolve kernel symbols, we search for REFNAME and if its address changed, relocation happened and we thus must change the kernel mapping routines to one that uses .pgoff as the relocation to apply. This way we use the same mechanism used for the other DSOs and don't have to do a two pass in all the kernel symbols. Reported-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> LKML-Reference: <1262717431-1246-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-06 02:50:31 +08:00
return -ENOMEM;
}
perf tools: Handle relocatable kernels DSOs don't have this problem because the kernel emits a PERF_MMAP for each new executable mapping it performs on monitored threads. To fix the kernel case we simulate the same behaviour, by having 'perf record' to synthesize a PERF_MMAP for the kernel, encoded like this: [root@doppio ~]# perf record -a -f sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.344 MB perf.data (~15038 samples) ] [root@doppio ~]# perf report -D | head -10 0xd0 [0x40]: event: 1 . . ... raw event: size 64 bytes . 0000: 01 00 00 00 00 00 40 00 00 00 00 00 00 00 00 00 ......@........ . 0010: 00 00 00 81 ff ff ff ff 00 00 00 00 00 00 00 00 ............... . 0020: 00 00 00 00 00 00 00 00 5b 6b 65 72 6e 65 6c 2e ........ [kernel . 0030: 6b 61 6c 6c 73 79 6d 73 2e 5f 74 65 78 74 5d 00 kallsyms._text] . 0xd0 [0x40]: PERF_RECORD_MMAP 0/0: [0xffffffff81000000((nil)) @ (nil)]: [kernel.kallsyms._text] I.e. we identify such event as having: .pid = 0 .filename = [kernel.kallsyms.REFNAME] .start = REFNAME addr in /proc/kallsyms at 'perf record' time and use now a hardcoded value of '.text' for REFNAME. Then, later, in 'perf report', if there are any kernel hits and thus we need to resolve kernel symbols, we search for REFNAME and if its address changed, relocation happened and we thus must change the kernel mapping routines to one that uses .pgoff as the relocation to apply. This way we use the same mechanism used for the other DSOs and don't have to do a two pass in all the kernel symbols. Reported-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> LKML-Reference: <1262717431-1246-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-06 02:50:31 +08:00
bracket = strchr(ref->name, ']');
perf tools: Handle relocatable kernels DSOs don't have this problem because the kernel emits a PERF_MMAP for each new executable mapping it performs on monitored threads. To fix the kernel case we simulate the same behaviour, by having 'perf record' to synthesize a PERF_MMAP for the kernel, encoded like this: [root@doppio ~]# perf record -a -f sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.344 MB perf.data (~15038 samples) ] [root@doppio ~]# perf report -D | head -10 0xd0 [0x40]: event: 1 . . ... raw event: size 64 bytes . 0000: 01 00 00 00 00 00 40 00 00 00 00 00 00 00 00 00 ......@........ . 0010: 00 00 00 81 ff ff ff ff 00 00 00 00 00 00 00 00 ............... . 0020: 00 00 00 00 00 00 00 00 5b 6b 65 72 6e 65 6c 2e ........ [kernel . 0030: 6b 61 6c 6c 73 79 6d 73 2e 5f 74 65 78 74 5d 00 kallsyms._text] . 0xd0 [0x40]: PERF_RECORD_MMAP 0/0: [0xffffffff81000000((nil)) @ (nil)]: [kernel.kallsyms._text] I.e. we identify such event as having: .pid = 0 .filename = [kernel.kallsyms.REFNAME] .start = REFNAME addr in /proc/kallsyms at 'perf record' time and use now a hardcoded value of '.text' for REFNAME. Then, later, in 'perf report', if there are any kernel hits and thus we need to resolve kernel symbols, we search for REFNAME and if its address changed, relocation happened and we thus must change the kernel mapping routines to one that uses .pgoff as the relocation to apply. This way we use the same mechanism used for the other DSOs and don't have to do a two pass in all the kernel symbols. Reported-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> LKML-Reference: <1262717431-1246-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-06 02:50:31 +08:00
if (bracket)
*bracket = '\0';
ref->addr = addr;
for (i = 0; i < MAP__NR_TYPES; ++i) {
struct kmap *kmap = map__kmap(maps[i]);
kmap->ref_reloc_sym = ref;
}
perf tools: Handle relocatable kernels DSOs don't have this problem because the kernel emits a PERF_MMAP for each new executable mapping it performs on monitored threads. To fix the kernel case we simulate the same behaviour, by having 'perf record' to synthesize a PERF_MMAP for the kernel, encoded like this: [root@doppio ~]# perf record -a -f sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.344 MB perf.data (~15038 samples) ] [root@doppio ~]# perf report -D | head -10 0xd0 [0x40]: event: 1 . . ... raw event: size 64 bytes . 0000: 01 00 00 00 00 00 40 00 00 00 00 00 00 00 00 00 ......@........ . 0010: 00 00 00 81 ff ff ff ff 00 00 00 00 00 00 00 00 ............... . 0020: 00 00 00 00 00 00 00 00 5b 6b 65 72 6e 65 6c 2e ........ [kernel . 0030: 6b 61 6c 6c 73 79 6d 73 2e 5f 74 65 78 74 5d 00 kallsyms._text] . 0xd0 [0x40]: PERF_RECORD_MMAP 0/0: [0xffffffff81000000((nil)) @ (nil)]: [kernel.kallsyms._text] I.e. we identify such event as having: .pid = 0 .filename = [kernel.kallsyms.REFNAME] .start = REFNAME addr in /proc/kallsyms at 'perf record' time and use now a hardcoded value of '.text' for REFNAME. Then, later, in 'perf report', if there are any kernel hits and thus we need to resolve kernel symbols, we search for REFNAME and if its address changed, relocation happened and we thus must change the kernel mapping routines to one that uses .pgoff as the relocation to apply. This way we use the same mechanism used for the other DSOs and don't have to do a two pass in all the kernel symbols. Reported-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Frédéric Weisbecker <fweisbec@gmail.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Mike Galbraith <efault@gmx.de> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Paul Mackerras <paulus@samba.org> Cc: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com> LKML-Reference: <1262717431-1246-1-git-send-email-acme@infradead.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-01-06 02:50:31 +08:00
return 0;
}
size_t perf_session__fprintf_dsos(struct perf_session *self, FILE *fp)
{
return __dsos__fprintf(&self->host_machine.kernel_dsos, fp) +
__dsos__fprintf(&self->host_machine.user_dsos, fp) +
machines__fprintf_dsos(&self->machines, fp);
}
size_t perf_session__fprintf_dsos_buildid(struct perf_session *self, FILE *fp,
bool with_hits)
{
size_t ret = machine__fprintf_dsos_buildid(&self->host_machine, fp, with_hits);
return ret + machines__fprintf_dsos_buildid(&self->machines, fp, with_hits);
}