2010-02-02 20:53:14 -07:00
|
|
|
#define _FILE_OFFSET_BITS 64
|
|
|
|
|
2009-12-11 16:24:02 -07:00
|
|
|
#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 07:23:10 -07:00
|
|
|
#include <byteswap.h>
|
2009-12-11 16:24:02 -07:00
|
|
|
#include <unistd.h>
|
|
|
|
#include <sys/types.h>
|
2010-05-18 15:29:23 -06:00
|
|
|
#include <sys/mman.h>
|
2009-12-11 16:24:02 -07:00
|
|
|
|
|
|
|
#include "session.h"
|
2009-12-14 09:22:59 -07:00
|
|
|
#include "sort.h"
|
2009-12-11 16:24:02 -07:00
|
|
|
#include "util.h"
|
|
|
|
|
|
|
|
static int perf_session__open(struct perf_session *self, bool force)
|
|
|
|
{
|
|
|
|
struct stat input_stat;
|
|
|
|
|
2010-04-01 22:59:15 -06:00
|
|
|
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;
|
|
|
|
}
|
|
|
|
|
2010-02-04 01:46:42 -07:00
|
|
|
self->fd = open(self->filename, O_RDONLY);
|
2009-12-11 16:24:02 -07:00
|
|
|
if (self->fd < 0) {
|
2010-06-11 21:36:15 -06:00
|
|
|
int err = errno;
|
|
|
|
|
|
|
|
pr_err("failed to open %s: %s", self->filename, strerror(err));
|
|
|
|
if (err == ENOENT && !strcmp(self->filename, "perf.data"))
|
2009-12-11 16:24:02 -07:00
|
|
|
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;
|
|
|
|
}
|
|
|
|
|
2010-04-01 22:59:15 -06:00
|
|
|
if (perf_header__read(self, self->fd) < 0) {
|
2009-12-11 16:24:02 -07:00
|
|
|
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;
|
|
|
|
}
|
|
|
|
|
2010-12-02 05:25:28 -07:00
|
|
|
static void perf_session__id_header_size(struct perf_session *session)
|
2010-04-01 22:59:15 -06:00
|
|
|
{
|
2010-12-02 05:25:28 -07:00
|
|
|
struct sample_data *data;
|
|
|
|
u64 sample_type = session->sample_type;
|
|
|
|
u16 size = 0;
|
|
|
|
|
|
|
|
if (!session->sample_id_all)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (sample_type & PERF_SAMPLE_TID)
|
|
|
|
size += sizeof(data->tid) * 2;
|
|
|
|
|
|
|
|
if (sample_type & PERF_SAMPLE_TIME)
|
|
|
|
size += sizeof(data->time);
|
|
|
|
|
|
|
|
if (sample_type & PERF_SAMPLE_ID)
|
|
|
|
size += sizeof(data->id);
|
|
|
|
|
|
|
|
if (sample_type & PERF_SAMPLE_STREAM_ID)
|
|
|
|
size += sizeof(data->stream_id);
|
|
|
|
|
|
|
|
if (sample_type & PERF_SAMPLE_CPU)
|
|
|
|
size += sizeof(data->cpu) * 2;
|
|
|
|
out:
|
|
|
|
session->id_hdr_size = size;
|
|
|
|
}
|
|
|
|
|
|
|
|
void perf_session__set_sample_id_all(struct perf_session *session, bool value)
|
|
|
|
{
|
|
|
|
session->sample_id_all = value;
|
|
|
|
perf_session__id_header_size(session);
|
2010-04-01 22:59:15 -06:00
|
|
|
}
|
|
|
|
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
void perf_session__set_sample_type(struct perf_session *session, u64 type)
|
|
|
|
{
|
|
|
|
session->sample_type = type;
|
|
|
|
}
|
|
|
|
|
2010-12-02 05:25:28 -07:00
|
|
|
void perf_session__update_sample_type(struct perf_session *self)
|
|
|
|
{
|
|
|
|
self->sample_type = perf_header__sample_type(&self->header);
|
|
|
|
self->sample_id_all = perf_header__sample_id_all(&self->header);
|
|
|
|
perf_session__id_header_size(self);
|
|
|
|
}
|
|
|
|
|
2010-04-18 23:32:50 -06:00
|
|
|
int perf_session__create_kernel_maps(struct perf_session *self)
|
|
|
|
{
|
2010-05-10 09:51:05 -06:00
|
|
|
int ret = machine__create_kernel_maps(&self->host_machine);
|
2010-04-18 23:32:50 -06:00
|
|
|
|
|
|
|
if (ret >= 0)
|
2010-05-10 09:51:05 -06:00
|
|
|
ret = machines__create_guest_kernel_maps(&self->machines);
|
2010-04-18 23:32:50 -06:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
perf session: Free the ref_reloc_sym memory at the right place
Which is at perf_session__destroy_kernel_maps, counterpart to the
perf_session__create_kernel_maps where the kmap structure is located, just
after the vmlinux_maps.
Make it also check if the kernel maps were actually created, which may not
be the case if, for instance, perf_session__new can't complete due to
permission problems in, for instance, a 'perf report' case, when a
segfault will take place, that is how this was noticed.
The problem was introduced in d65a458, thus post .35.
This also adds code to release guest machines as them are also created
in perf_session__create_kernel_maps, so should be deleted on this newly
introduced counterpart, perf_session__destroy_kernel_maps.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-08-02 15:18:28 -06:00
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
2010-05-01 00:41:20 -06:00
|
|
|
struct perf_session *perf_session__new(const char *filename, int mode, bool force, bool repipe)
|
2009-12-11 16:24:02 -07:00
|
|
|
{
|
2009-12-13 14:50:28 -07:00
|
|
|
size_t len = filename ? strlen(filename) + 1 : 0;
|
2009-12-11 16:24:02 -07:00
|
|
|
struct perf_session *self = zalloc(sizeof(*self) + len);
|
|
|
|
|
|
|
|
if (self == NULL)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (perf_header__init(&self->header) < 0)
|
perf session: Move kmaps to perf_session
There is still some more work to do to disentangle map creation
from DSO loading, but this happens only for the kernel, and for
the early adopters of perf diff, where this disentanglement
matters most, we'll be testing different kernels, so no problem
here.
Further clarification: right now we create the kernel maps for
the various modules and discontiguous kernel text maps when
loading the DSO, we should do it as a two step process, first
creating the maps, for multiple mappings with the same DSO
store, then doing the dso load just once, for the first hit on
one of the maps sharing this DSO backing store.
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: <1260741029-4430-6-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-13 14:50:29 -07:00
|
|
|
goto out_free;
|
2009-12-11 16:24:02 -07:00
|
|
|
|
|
|
|
memcpy(self->filename, filename, len);
|
2009-12-13 14:50:28 -07:00
|
|
|
self->threads = RB_ROOT;
|
2010-06-17 05:37:44 -06:00
|
|
|
INIT_LIST_HEAD(&self->dead_threads);
|
perf hist: Introduce hists class and move lots of methods to it
In cbbc79a we introduced support for multiple events by introducing a
new "event_stat_id" struct and then made several perf_session methods
receive a point to it instead of a pointer to perf_session, and kept the
event_stats and hists rb_tree in perf_session.
While working on the new newt based browser, I realised that it would be
better to introduce a new class, "hists" (short for "histograms"),
renaming the "event_stat_id" struct and the perf_session methods that
were really "hists" methods, as they manipulate only struct hists
members, not touching anything in the other perf_session members.
Other optimizations, such as calculating the maximum lenght of a symbol
name present in an hists instance will be possible as we add them,
avoiding a re-traversal just for finding that information.
The rationale for the name "hists" to replace "event_stat_id" is that we
may have multiple sets of hists for the same event_stat id, as, for
instance, the 'perf diff' tool has, so event stat id is not what
characterizes what this struct and the functions that manipulate it do.
Cc: Eric B Munson <ebmunson@us.ibm.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-05-10 10:04:11 -06:00
|
|
|
self->hists_tree = RB_ROOT;
|
2009-12-13 14:50:28 -07:00
|
|
|
self->last_match = NULL;
|
2010-11-30 10:49:46 -07:00
|
|
|
/*
|
|
|
|
* On 64bit we can mmap the data file in one go. No need for tiny mmap
|
|
|
|
* slices. On 32bit we use 32MB.
|
|
|
|
*/
|
|
|
|
#if BITS_PER_LONG == 64
|
|
|
|
self->mmap_window = ULLONG_MAX;
|
|
|
|
#else
|
|
|
|
self->mmap_window = 32 * 1024 * 1024ULL;
|
|
|
|
#endif
|
2010-04-27 18:17:50 -06:00
|
|
|
self->machines = RB_ROOT;
|
2010-05-01 00:41:20 -06:00
|
|
|
self->repipe = repipe;
|
2010-11-30 10:49:33 -07:00
|
|
|
INIT_LIST_HEAD(&self->ordered_samples.samples);
|
2010-11-30 10:49:53 -07:00
|
|
|
INIT_LIST_HEAD(&self->ordered_samples.sample_cache);
|
2010-11-30 10:49:55 -07:00
|
|
|
INIT_LIST_HEAD(&self->ordered_samples.to_free);
|
2010-05-09 16:57:08 -06:00
|
|
|
machine__init(&self->host_machine, "", HOST_KERNEL_ID);
|
2009-12-11 16:24:02 -07:00
|
|
|
|
2010-01-27 16:05:52 -07:00
|
|
|
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;
|
|
|
|
}
|
2009-12-27 16:37:02 -07:00
|
|
|
|
2010-04-01 22:59:15 -06:00
|
|
|
perf_session__update_sample_type(self);
|
2009-12-11 16:24:02 -07:00
|
|
|
out:
|
|
|
|
return self;
|
perf session: Move kmaps to perf_session
There is still some more work to do to disentangle map creation
from DSO loading, but this happens only for the kernel, and for
the early adopters of perf diff, where this disentanglement
matters most, we'll be testing different kernels, so no problem
here.
Further clarification: right now we create the kernel maps for
the various modules and discontiguous kernel text maps when
loading the DSO, we should do it as a two step process, first
creating the maps, for multiple mappings with the same DSO
store, then doing the dso load just once, for the first hit on
one of the maps sharing this DSO backing store.
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: <1260741029-4430-6-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-13 14:50:29 -07:00
|
|
|
out_free:
|
2009-12-11 16:24:02 -07:00
|
|
|
free(self);
|
|
|
|
return NULL;
|
perf session: Move kmaps to perf_session
There is still some more work to do to disentangle map creation
from DSO loading, but this happens only for the kernel, and for
the early adopters of perf diff, where this disentanglement
matters most, we'll be testing different kernels, so no problem
here.
Further clarification: right now we create the kernel maps for
the various modules and discontiguous kernel text maps when
loading the DSO, we should do it as a two step process, first
creating the maps, for multiple mappings with the same DSO
store, then doing the dso load just once, for the first hit on
one of the maps sharing this DSO backing store.
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: <1260741029-4430-6-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-12-13 14:50:29 -07:00
|
|
|
out_delete:
|
|
|
|
perf_session__delete(self);
|
|
|
|
return NULL;
|
2009-12-11 16:24:02 -07:00
|
|
|
}
|
|
|
|
|
2010-07-30 15:31:28 -06:00
|
|
|
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);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2009-12-11 16:24:02 -07:00
|
|
|
void perf_session__delete(struct perf_session *self)
|
|
|
|
{
|
|
|
|
perf_header__exit(&self->header);
|
perf session: Free the ref_reloc_sym memory at the right place
Which is at perf_session__destroy_kernel_maps, counterpart to the
perf_session__create_kernel_maps where the kmap structure is located, just
after the vmlinux_maps.
Make it also check if the kernel maps were actually created, which may not
be the case if, for instance, perf_session__new can't complete due to
permission problems in, for instance, a 'perf report' case, when a
segfault will take place, that is how this was noticed.
The problem was introduced in d65a458, thus post .35.
This also adds code to release guest machines as them are also created
in perf_session__create_kernel_maps, so should be deleted on this newly
introduced counterpart, perf_session__destroy_kernel_maps.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-08-02 15:18:28 -06:00
|
|
|
perf_session__destroy_kernel_maps(self);
|
2010-07-30 15:31:28 -06:00
|
|
|
perf_session__delete_dead_threads(self);
|
|
|
|
perf_session__delete_threads(self);
|
|
|
|
machine__exit(&self->host_machine);
|
2009-12-11 16:24:02 -07:00
|
|
|
close(self->fd);
|
|
|
|
free(self);
|
|
|
|
}
|
2009-12-14 09:22:59 -07:00
|
|
|
|
2010-06-17 05:37:44 -06:00
|
|
|
void perf_session__remove_thread(struct perf_session *self, struct thread *th)
|
|
|
|
{
|
2010-08-02 15:59:28 -06:00
|
|
|
self->last_match = NULL;
|
2010-06-17 05:37:44 -06:00
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
2009-12-14 09:22:59 -07:00
|
|
|
static bool symbol__match_parent_regex(struct symbol *sym)
|
|
|
|
{
|
|
|
|
if (sym->name && !regexec(&parent_regex, sym->name, 0, NULL, 0))
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-03-24 13:40:18 -06:00
|
|
|
struct map_symbol *perf_session__resolve_callchain(struct perf_session *self,
|
|
|
|
struct thread *thread,
|
|
|
|
struct ip_callchain *chain,
|
|
|
|
struct symbol **parent)
|
2009-12-14 09:22:59 -07:00
|
|
|
{
|
|
|
|
u8 cpumode = PERF_RECORD_MISC_USER;
|
|
|
|
unsigned int i;
|
2010-04-02 07:04:18 -06:00
|
|
|
struct map_symbol *syms = calloc(chain->nr, sizeof(*syms));
|
2009-12-14 09:22:59 -07:00
|
|
|
|
2010-04-02 07:04:18 -06:00
|
|
|
if (!syms)
|
|
|
|
return NULL;
|
2009-12-14 09:22:59 -07:00
|
|
|
|
|
|
|
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;
|
|
|
|
}
|
|
|
|
|
2010-04-18 23:32:50 -06:00
|
|
|
al.filtered = false;
|
2009-12-14 09:22:59 -07:00
|
|
|
thread__find_addr_location(thread, self, cpumode,
|
2010-04-18 23:32:50 -06:00
|
|
|
MAP__FUNCTION, thread->pid, ip, &al, NULL);
|
2009-12-14 09:22:59 -07:00
|
|
|
if (al.sym != NULL) {
|
|
|
|
if (sort__has_parent && !*parent &&
|
|
|
|
symbol__match_parent_regex(al.sym))
|
|
|
|
*parent = al.sym;
|
2009-12-15 15:04:42 -07:00
|
|
|
if (!symbol_conf.use_callchain)
|
2009-12-14 09:22:59 -07:00
|
|
|
break;
|
2010-03-24 13:40:18 -06:00
|
|
|
syms[i].map = al.map;
|
|
|
|
syms[i].sym = al.sym;
|
2009-12-14 09:22:59 -07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return syms;
|
|
|
|
}
|
2009-12-27 16:36:59 -07:00
|
|
|
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
static int process_event_synth_stub(event_t *event __used,
|
|
|
|
struct perf_session *session __used)
|
|
|
|
{
|
|
|
|
dump_printf(": unhandled!\n");
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-12-27 16:36:59 -07:00
|
|
|
static int process_event_stub(event_t *event __used,
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
struct sample_data *sample __used,
|
2009-12-27 16:36:59 -07:00
|
|
|
struct perf_session *session __used)
|
|
|
|
{
|
|
|
|
dump_printf(": unhandled!\n");
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2010-05-03 07:14:33 -06:00
|
|
|
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);
|
|
|
|
|
2009-12-27 16:36:59 -07:00
|
|
|
static void perf_event_ops__fill_defaults(struct perf_event_ops *handler)
|
|
|
|
{
|
2009-12-27 16:37:05 -07:00
|
|
|
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)
|
2010-11-26 13:31:54 -07:00
|
|
|
handler->lost = event__process_lost;
|
2009-12-27 16:37:05 -07:00
|
|
|
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;
|
2010-04-01 22:59:19 -06:00
|
|
|
if (handler->attr == NULL)
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
handler->attr = process_event_synth_stub;
|
2010-04-01 22:59:20 -06:00
|
|
|
if (handler->event_type == NULL)
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
handler->event_type = process_event_synth_stub;
|
2010-04-01 22:59:21 -06:00
|
|
|
if (handler->tracing_data == NULL)
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
handler->tracing_data = process_event_synth_stub;
|
2010-04-01 22:59:22 -06:00
|
|
|
if (handler->build_id == NULL)
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
handler->build_id = process_event_synth_stub;
|
2010-05-03 07:14:33 -06:00
|
|
|
if (handler->finished_round == NULL) {
|
|
|
|
if (handler->ordered_samples)
|
|
|
|
handler->finished_round = process_finished_round;
|
|
|
|
else
|
|
|
|
handler->finished_round = process_finished_round_stub;
|
|
|
|
}
|
2009-12-27 16:36:59 -07:00
|
|
|
}
|
|
|
|
|
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 07:23:10 -07:00
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
2010-04-01 22:59:19 -06:00
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
2010-04-01 22:59:20 -06:00
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
2010-04-01 22:59:21 -06:00
|
|
|
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 07:23:10 -07:00
|
|
|
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,
|
2010-04-01 22:59:19 -06:00
|
|
|
[PERF_RECORD_HEADER_ATTR] = event__attr_swap,
|
2010-04-01 22:59:20 -06:00
|
|
|
[PERF_RECORD_HEADER_EVENT_TYPE] = event__event_type_swap,
|
2010-04-01 22:59:21 -06:00
|
|
|
[PERF_RECORD_HEADER_TRACING_DATA] = event__tracing_data_swap,
|
2010-04-01 22:59:22 -06:00
|
|
|
[PERF_RECORD_HEADER_BUILD_ID] = NULL,
|
2010-04-01 22:59:15 -06:00
|
|
|
[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 07:23:10 -07:00
|
|
|
};
|
|
|
|
|
2010-04-23 16:04:12 -06:00
|
|
|
struct sample_queue {
|
|
|
|
u64 timestamp;
|
2010-12-07 05:48:50 -07:00
|
|
|
u64 file_offset;
|
2010-11-30 10:49:35 -07:00
|
|
|
event_t *event;
|
2010-04-23 16:04:12 -06:00
|
|
|
struct list_head list;
|
|
|
|
};
|
|
|
|
|
2010-11-30 10:49:53 -07:00
|
|
|
static void perf_session_free_sample_buffers(struct perf_session *session)
|
|
|
|
{
|
|
|
|
struct ordered_samples *os = &session->ordered_samples;
|
|
|
|
|
2010-11-30 10:49:55 -07:00
|
|
|
while (!list_empty(&os->to_free)) {
|
2010-11-30 10:49:53 -07:00
|
|
|
struct sample_queue *sq;
|
|
|
|
|
2010-11-30 10:49:55 -07:00
|
|
|
sq = list_entry(os->to_free.next, struct sample_queue, list);
|
2010-11-30 10:49:53 -07:00
|
|
|
list_del(&sq->list);
|
|
|
|
free(sq);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-12-05 06:32:55 -07:00
|
|
|
static int perf_session_deliver_event(struct perf_session *session,
|
|
|
|
event_t *event,
|
|
|
|
struct sample_data *sample,
|
2010-12-07 05:48:53 -07:00
|
|
|
struct perf_event_ops *ops,
|
|
|
|
u64 file_offset);
|
2010-12-05 06:32:55 -07:00
|
|
|
|
2010-04-23 16:04:12 -06:00
|
|
|
static void flush_sample_queue(struct perf_session *s,
|
|
|
|
struct perf_event_ops *ops)
|
|
|
|
{
|
2010-11-30 10:49:33 -07:00
|
|
|
struct ordered_samples *os = &s->ordered_samples;
|
|
|
|
struct list_head *head = &os->samples;
|
2010-04-23 16:04:12 -06:00
|
|
|
struct sample_queue *tmp, *iter;
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
struct sample_data sample;
|
2010-11-30 10:49:33 -07:00
|
|
|
u64 limit = os->next_flush;
|
|
|
|
u64 last_ts = os->last_sample ? os->last_sample->timestamp : 0ULL;
|
2010-04-23 16:04:12 -06:00
|
|
|
|
2010-05-03 07:14:33 -06:00
|
|
|
if (!ops->ordered_samples || !limit)
|
2010-04-23 16:04:12 -06:00
|
|
|
return;
|
|
|
|
|
|
|
|
list_for_each_entry_safe(iter, tmp, head, list) {
|
|
|
|
if (iter->timestamp > limit)
|
2010-11-30 10:49:33 -07:00
|
|
|
break;
|
2010-04-23 16:04:12 -06:00
|
|
|
|
2010-12-02 05:25:28 -07:00
|
|
|
event__parse_sample(iter->event, s, &sample);
|
2010-12-07 05:48:53 -07:00
|
|
|
perf_session_deliver_event(s, iter->event, &sample, ops,
|
|
|
|
iter->file_offset);
|
2010-04-23 16:04:12 -06:00
|
|
|
|
2010-11-30 10:49:33 -07:00
|
|
|
os->last_flush = iter->timestamp;
|
2010-04-23 16:04:12 -06:00
|
|
|
list_del(&iter->list);
|
2010-11-30 10:49:53 -07:00
|
|
|
list_add(&iter->list, &os->sample_cache);
|
2010-04-23 16:04:12 -06:00
|
|
|
}
|
2010-11-30 10:49:33 -07:00
|
|
|
|
|
|
|
if (list_empty(head)) {
|
|
|
|
os->last_sample = NULL;
|
|
|
|
} else if (last_ts <= limit) {
|
|
|
|
os->last_sample =
|
|
|
|
list_entry(head->prev, struct sample_queue, list);
|
|
|
|
}
|
2010-04-23 16:04:12 -06:00
|
|
|
}
|
|
|
|
|
2010-05-03 07:14:33 -06:00
|
|
|
/*
|
|
|
|
* 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;
|
|
|
|
}
|
|
|
|
|
2010-04-23 16:04:12 -06:00
|
|
|
/* The queue is ordered by time */
|
2010-12-05 06:32:55 -07:00
|
|
|
static void __queue_event(struct sample_queue *new, struct perf_session *s)
|
2010-04-23 16:04:12 -06:00
|
|
|
{
|
2010-11-30 10:49:33 -07:00
|
|
|
struct ordered_samples *os = &s->ordered_samples;
|
|
|
|
struct sample_queue *sample = os->last_sample;
|
|
|
|
u64 timestamp = new->timestamp;
|
|
|
|
struct list_head *p;
|
2010-04-23 16:04:12 -06:00
|
|
|
|
2010-11-30 10:49:33 -07:00
|
|
|
os->last_sample = new;
|
2010-04-23 16:04:12 -06:00
|
|
|
|
2010-11-30 10:49:33 -07:00
|
|
|
if (!sample) {
|
|
|
|
list_add(&new->list, &os->samples);
|
|
|
|
os->max_timestamp = timestamp;
|
2010-04-23 16:04:12 -06:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2010-11-30 10:49:33 -07:00
|
|
|
* 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.
|
2010-04-23 16:04:12 -06:00
|
|
|
*/
|
2010-11-30 10:49:33 -07:00
|
|
|
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);
|
|
|
|
}
|
2010-04-23 16:04:12 -06:00
|
|
|
}
|
|
|
|
|
2010-11-30 10:49:55 -07:00
|
|
|
#define MAX_SAMPLE_BUFFER (64 * 1024 / sizeof(struct sample_queue))
|
|
|
|
|
2010-12-05 06:32:55 -07:00
|
|
|
static int perf_session_queue_event(struct perf_session *s, event_t *event,
|
2010-12-07 05:48:50 -07:00
|
|
|
struct sample_data *data, u64 file_offset)
|
2010-04-23 16:04:12 -06:00
|
|
|
{
|
2010-11-30 10:49:55 -07:00
|
|
|
struct ordered_samples *os = &s->ordered_samples;
|
|
|
|
struct list_head *sc = &os->sample_cache;
|
2010-04-23 16:04:12 -06:00
|
|
|
u64 timestamp = data->time;
|
|
|
|
struct sample_queue *new;
|
|
|
|
|
2010-12-07 05:48:44 -07:00
|
|
|
if (!timestamp || timestamp == ~0ULL)
|
2010-12-05 06:32:55 -07:00
|
|
|
return -ETIME;
|
|
|
|
|
2010-04-23 16:04:12 -06:00
|
|
|
if (timestamp < s->ordered_samples.last_flush) {
|
|
|
|
printf("Warning: Timestamp below last timeslice flush\n");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
2010-11-30 10:49:53 -07:00
|
|
|
if (!list_empty(sc)) {
|
|
|
|
new = list_entry(sc->next, struct sample_queue, list);
|
|
|
|
list_del(&new->list);
|
2010-11-30 10:49:55 -07:00
|
|
|
} else if (os->sample_buffer) {
|
|
|
|
new = os->sample_buffer + os->sample_buffer_idx;
|
|
|
|
if (++os->sample_buffer_idx == MAX_SAMPLE_BUFFER)
|
|
|
|
os->sample_buffer = NULL;
|
2010-11-30 10:49:53 -07:00
|
|
|
} else {
|
2010-11-30 10:49:55 -07:00
|
|
|
os->sample_buffer = malloc(MAX_SAMPLE_BUFFER * sizeof(*new));
|
|
|
|
if (!os->sample_buffer)
|
2010-11-30 10:49:53 -07:00
|
|
|
return -ENOMEM;
|
2010-11-30 10:49:55 -07:00
|
|
|
list_add(&os->sample_buffer->list, &os->to_free);
|
|
|
|
os->sample_buffer_idx = 2;
|
|
|
|
new = os->sample_buffer + 1;
|
2010-11-30 10:49:53 -07:00
|
|
|
}
|
2010-04-23 16:04:12 -06:00
|
|
|
|
|
|
|
new->timestamp = timestamp;
|
2010-12-07 05:48:50 -07:00
|
|
|
new->file_offset = file_offset;
|
2010-11-30 10:49:49 -07:00
|
|
|
new->event = event;
|
2010-04-23 16:04:12 -06:00
|
|
|
|
2010-12-05 06:32:55 -07:00
|
|
|
__queue_event(new, s);
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
2010-04-23 16:04:12 -06:00
|
|
|
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
static void callchain__dump(struct sample_data *sample)
|
|
|
|
{
|
|
|
|
unsigned int i;
|
2010-04-23 16:04:12 -06:00
|
|
|
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
if (!dump_trace)
|
|
|
|
return;
|
2010-04-23 16:04:12 -06:00
|
|
|
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
printf("... chain: nr:%Lu\n", sample->callchain->nr);
|
|
|
|
|
|
|
|
for (i = 0; i < sample->callchain->nr; i++)
|
|
|
|
printf("..... %2d: %016Lx\n", i, sample->callchain->ips[i]);
|
2010-04-23 16:04:12 -06:00
|
|
|
}
|
|
|
|
|
2010-12-02 05:25:28 -07:00
|
|
|
static void perf_session__print_tstamp(struct perf_session *session,
|
|
|
|
event_t *event,
|
|
|
|
struct sample_data *sample)
|
|
|
|
{
|
|
|
|
if (event->header.type != PERF_RECORD_SAMPLE &&
|
|
|
|
!session->sample_id_all) {
|
|
|
|
fputs("-1 -1 ", stdout);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((session->sample_type & PERF_SAMPLE_CPU))
|
|
|
|
printf("%u ", sample->cpu);
|
|
|
|
|
|
|
|
if (session->sample_type & PERF_SAMPLE_TIME)
|
|
|
|
printf("%Lu ", sample->time);
|
|
|
|
}
|
|
|
|
|
2010-12-07 05:48:47 -07:00
|
|
|
static void dump_event(struct perf_session *session, event_t *event,
|
|
|
|
u64 file_offset, struct sample_data *sample)
|
|
|
|
{
|
|
|
|
if (!dump_trace)
|
|
|
|
return;
|
|
|
|
|
|
|
|
dump_printf("\n%#Lx [%#x]: event: %d\n", file_offset,
|
|
|
|
event->header.size, event->header.type);
|
|
|
|
|
|
|
|
trace_event(event);
|
|
|
|
|
|
|
|
if (sample)
|
|
|
|
perf_session__print_tstamp(session, event, sample);
|
|
|
|
|
|
|
|
dump_printf("%#Lx [%#x]: PERF_RECORD_%s",
|
|
|
|
file_offset, event->header.size,
|
|
|
|
event__get_event_name(event->header.type));
|
|
|
|
}
|
|
|
|
|
|
|
|
static void dump_sample(struct perf_session *session, event_t *event,
|
|
|
|
struct sample_data *sample)
|
|
|
|
{
|
|
|
|
dump_printf("(IP, %d): %d/%d: %#Lx period: %Ld\n", event->header.misc,
|
|
|
|
sample->pid, sample->tid, sample->ip, sample->period);
|
|
|
|
|
|
|
|
if (session->sample_type & PERF_SAMPLE_CALLCHAIN)
|
|
|
|
callchain__dump(sample);
|
|
|
|
}
|
|
|
|
|
2010-12-05 06:32:55 -07:00
|
|
|
static int perf_session_deliver_event(struct perf_session *session,
|
|
|
|
event_t *event,
|
|
|
|
struct sample_data *sample,
|
2010-12-07 05:48:53 -07:00
|
|
|
struct perf_event_ops *ops,
|
|
|
|
u64 file_offset __used)
|
2010-12-05 06:32:55 -07:00
|
|
|
{
|
|
|
|
switch (event->header.type) {
|
|
|
|
case PERF_RECORD_SAMPLE:
|
|
|
|
return ops->sample(event, sample, session);
|
|
|
|
case PERF_RECORD_MMAP:
|
|
|
|
return ops->mmap(event, sample, session);
|
|
|
|
case PERF_RECORD_COMM:
|
|
|
|
return ops->comm(event, sample, session);
|
|
|
|
case PERF_RECORD_FORK:
|
|
|
|
return ops->fork(event, sample, session);
|
|
|
|
case PERF_RECORD_EXIT:
|
|
|
|
return ops->exit(event, sample, session);
|
|
|
|
case PERF_RECORD_LOST:
|
|
|
|
return ops->lost(event, sample, session);
|
|
|
|
case PERF_RECORD_READ:
|
|
|
|
return ops->read(event, sample, session);
|
|
|
|
case PERF_RECORD_THROTTLE:
|
|
|
|
return ops->throttle(event, sample, session);
|
|
|
|
case PERF_RECORD_UNTHROTTLE:
|
|
|
|
return ops->unthrottle(event, sample, session);
|
|
|
|
default:
|
|
|
|
++session->hists.stats.nr_unknown_events;
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int perf_session__process_event(struct perf_session *session,
|
2009-12-27 16:36:59 -07:00
|
|
|
event_t *event,
|
|
|
|
struct perf_event_ops *ops,
|
2010-11-30 10:49:38 -07:00
|
|
|
u64 file_offset)
|
2009-12-27 16:36:59 -07:00
|
|
|
{
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
struct sample_data sample;
|
2010-12-05 06:32:55 -07:00
|
|
|
int ret;
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
|
2010-12-05 06:32:55 -07:00
|
|
|
if (session->header.needs_swap && event__swap_ops[event->header.type])
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
event__swap_ops[event->header.type](event);
|
|
|
|
|
2010-12-07 05:48:47 -07:00
|
|
|
if (event->header.type >= PERF_RECORD_HEADER_MAX)
|
|
|
|
return -EINVAL;
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
|
2010-12-07 05:48:47 -07:00
|
|
|
hists__inc_nr_events(&session->hists, event->header.type);
|
|
|
|
|
|
|
|
if (event->header.type >= PERF_RECORD_USER_TYPE_START)
|
|
|
|
dump_event(session, event, file_offset, NULL);
|
|
|
|
else {
|
|
|
|
event__parse_sample(event, session, &sample);
|
|
|
|
dump_event(session, event, file_offset, &sample);
|
2009-12-27 16:36:59 -07:00
|
|
|
}
|
|
|
|
|
2010-12-05 06:32:55 -07:00
|
|
|
/* These events are processed right away */
|
2009-12-27 16:36:59 -07:00
|
|
|
switch (event->header.type) {
|
|
|
|
case PERF_RECORD_SAMPLE:
|
2010-12-05 06:32:55 -07:00
|
|
|
if (session->sample_type & PERF_SAMPLE_CALLCHAIN) {
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
if (!ip_callchain__valid(sample.callchain, event)) {
|
|
|
|
pr_debug("call-chain problem with event, "
|
|
|
|
"skipping it.\n");
|
2010-12-05 06:32:55 -07:00
|
|
|
++session->hists.stats.nr_invalid_chains;
|
|
|
|
session->hists.stats.total_invalid_chains +=
|
|
|
|
sample.period;
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
}
|
2010-12-07 05:48:47 -07:00
|
|
|
dump_sample(session, event, &sample);
|
2010-12-05 06:32:55 -07:00
|
|
|
break;
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
|
2010-04-01 22:59:19 -06:00
|
|
|
case PERF_RECORD_HEADER_ATTR:
|
2010-12-05 06:32:55 -07:00
|
|
|
return ops->attr(event, session);
|
2010-04-01 22:59:20 -06:00
|
|
|
case PERF_RECORD_HEADER_EVENT_TYPE:
|
2010-12-05 06:32:55 -07:00
|
|
|
return ops->event_type(event, session);
|
2010-04-01 22:59:21 -06:00
|
|
|
case PERF_RECORD_HEADER_TRACING_DATA:
|
|
|
|
/* setup for reading amidst mmap */
|
2010-12-05 06:32:55 -07:00
|
|
|
lseek(session->fd, file_offset, SEEK_SET);
|
|
|
|
return ops->tracing_data(event, session);
|
2010-04-01 22:59:22 -06:00
|
|
|
case PERF_RECORD_HEADER_BUILD_ID:
|
2010-12-05 06:32:55 -07:00
|
|
|
return ops->build_id(event, session);
|
2010-05-03 07:14:33 -06:00
|
|
|
case PERF_RECORD_FINISHED_ROUND:
|
2010-12-05 06:32:55 -07:00
|
|
|
return ops->finished_round(event, session, ops);
|
2009-12-27 16:36:59 -07:00
|
|
|
default:
|
2010-12-05 06:32:55 -07:00
|
|
|
break;
|
2009-12-27 16:36:59 -07:00
|
|
|
}
|
2010-12-05 06:32:55 -07:00
|
|
|
|
|
|
|
if (ops->ordered_samples) {
|
2010-12-07 05:48:50 -07:00
|
|
|
ret = perf_session_queue_event(session, event, &sample,
|
|
|
|
file_offset);
|
2010-12-05 06:32:55 -07:00
|
|
|
if (ret != -ETIME)
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2010-12-07 05:48:53 -07:00
|
|
|
return perf_session_deliver_event(session, event, &sample, ops,
|
|
|
|
file_offset);
|
2009-12-27 16:36:59 -07:00
|
|
|
}
|
|
|
|
|
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 07:23:10 -07:00
|
|
|
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);
|
|
|
|
}
|
|
|
|
|
2009-12-27 16:36:59 -07:00
|
|
|
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;
|
|
|
|
}
|
|
|
|
|
2010-04-01 22:59:15 -06:00
|
|
|
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);
|
|
|
|
|
2010-05-04 23:27:40 -06:00
|
|
|
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;
|
|
|
|
}
|
2010-04-01 22:59:15 -06:00
|
|
|
|
2010-05-04 23:27:40 -06:00
|
|
|
pr_err("failed to read event data\n");
|
|
|
|
goto out_err;
|
|
|
|
}
|
2010-04-01 22:59:15 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
if (size == 0 ||
|
2010-11-30 10:49:38 -07:00
|
|
|
(skip = perf_session__process_event(self, &event, ops, head)) < 0) {
|
2010-04-01 22:59:15 -06:00
|
|
|
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;
|
|
|
|
|
|
|
|
if (skip > 0)
|
|
|
|
head += skip;
|
|
|
|
|
|
|
|
if (!session_done())
|
|
|
|
goto more;
|
|
|
|
done:
|
|
|
|
err = 0;
|
|
|
|
out_err:
|
2010-11-30 10:49:53 -07:00
|
|
|
perf_session_free_sample_buffers(self);
|
2010-04-01 22:59:15 -06:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2010-11-30 10:49:38 -07:00
|
|
|
int __perf_session__process_events(struct perf_session *session,
|
2010-02-03 11:52:05 -07:00
|
|
|
u64 data_offset, u64 data_size,
|
|
|
|
u64 file_size, struct perf_event_ops *ops)
|
2009-12-27 16:36:59 -07:00
|
|
|
{
|
2010-11-30 10:49:46 -07:00
|
|
|
u64 head, page_offset, file_offset, file_pos, progress_next;
|
2010-11-30 10:49:49 -07:00
|
|
|
int err, mmap_prot, mmap_flags, map_idx = 0;
|
2010-11-30 10:49:38 -07:00
|
|
|
struct ui_progress *progress;
|
2010-11-30 10:49:46 -07:00
|
|
|
size_t page_size, mmap_size;
|
2010-11-30 10:49:49 -07:00
|
|
|
char *buf, *mmaps[8];
|
2009-12-27 16:36:59 -07:00
|
|
|
event_t *event;
|
|
|
|
uint32_t size;
|
2010-11-30 10:49:38 -07:00
|
|
|
|
2009-12-27 16:36:59 -07:00
|
|
|
perf_event_ops__fill_defaults(ops);
|
|
|
|
|
2010-01-14 13:30:04 -07:00
|
|
|
page_size = sysconf(_SC_PAGESIZE);
|
2009-12-27 16:36:59 -07:00
|
|
|
|
2010-11-30 10:49:38 -07:00
|
|
|
page_offset = page_size * (data_offset / page_size);
|
|
|
|
file_offset = page_offset;
|
|
|
|
head = data_offset - page_offset;
|
2009-12-27 16:36:59 -07:00
|
|
|
|
2010-11-30 10:49:44 -07:00
|
|
|
if (data_offset + data_size < file_size)
|
|
|
|
file_size = data_offset + data_size;
|
|
|
|
|
2010-11-30 10:49:46 -07:00
|
|
|
progress_next = file_size / 16;
|
|
|
|
progress = ui_progress__new("Processing events...", file_size);
|
|
|
|
if (progress == NULL)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
mmap_size = session->mmap_window;
|
|
|
|
if (mmap_size > file_size)
|
|
|
|
mmap_size = file_size;
|
|
|
|
|
2010-11-30 10:49:49 -07:00
|
|
|
memset(mmaps, 0, sizeof(mmaps));
|
|
|
|
|
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 07:23:10 -07:00
|
|
|
mmap_prot = PROT_READ;
|
|
|
|
mmap_flags = MAP_SHARED;
|
|
|
|
|
2010-11-30 10:49:38 -07:00
|
|
|
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 07:23:10 -07:00
|
|
|
mmap_prot |= PROT_WRITE;
|
|
|
|
mmap_flags = MAP_PRIVATE;
|
|
|
|
}
|
2009-12-27 16:36:59 -07:00
|
|
|
remap:
|
2010-11-30 10:49:46 -07:00
|
|
|
buf = mmap(NULL, mmap_size, mmap_prot, mmap_flags, session->fd,
|
|
|
|
file_offset);
|
2009-12-27 16:36:59 -07:00
|
|
|
if (buf == MAP_FAILED) {
|
|
|
|
pr_err("failed to mmap file\n");
|
|
|
|
err = -errno;
|
|
|
|
goto out_err;
|
|
|
|
}
|
2010-11-30 10:49:49 -07:00
|
|
|
mmaps[map_idx] = buf;
|
|
|
|
map_idx = (map_idx + 1) & (ARRAY_SIZE(mmaps) - 1);
|
2010-11-30 10:49:44 -07:00
|
|
|
file_pos = file_offset + head;
|
2009-12-27 16:36:59 -07:00
|
|
|
|
|
|
|
more:
|
|
|
|
event = (event_t *)(buf + head);
|
|
|
|
|
2010-11-30 10:49:38 -07:00
|
|
|
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 07:23:10 -07:00
|
|
|
perf_event_header__bswap(&event->header);
|
2009-12-27 16:36:59 -07:00
|
|
|
size = event->header.size;
|
|
|
|
if (size == 0)
|
|
|
|
size = 8;
|
|
|
|
|
2010-11-30 10:49:46 -07:00
|
|
|
if (head + event->header.size >= mmap_size) {
|
2010-11-30 10:49:49 -07:00
|
|
|
if (mmaps[map_idx]) {
|
|
|
|
munmap(mmaps[map_idx], mmap_size);
|
|
|
|
mmaps[map_idx] = NULL;
|
|
|
|
}
|
2009-12-27 16:36:59 -07:00
|
|
|
|
2010-11-30 10:49:38 -07:00
|
|
|
page_offset = page_size * (head / page_size);
|
|
|
|
file_offset += page_offset;
|
|
|
|
head -= page_offset;
|
2009-12-27 16:36:59 -07:00
|
|
|
goto remap;
|
|
|
|
}
|
|
|
|
|
|
|
|
size = event->header.size;
|
|
|
|
|
2010-11-30 10:49:44 -07:00
|
|
|
if (size == 0 ||
|
|
|
|
perf_session__process_event(session, event, ops, file_pos) < 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 07:23:10 -07:00
|
|
|
dump_printf("%#Lx [%#x]: skipping unknown header type: %d\n",
|
2010-11-30 10:49:38 -07:00
|
|
|
file_offset + head, event->header.size,
|
2009-12-27 16:36:59 -07:00
|
|
|
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;
|
2010-11-30 10:49:44 -07:00
|
|
|
file_pos += size;
|
2009-12-27 16:36:59 -07:00
|
|
|
|
2010-11-30 10:49:46 -07:00
|
|
|
if (file_pos >= progress_next) {
|
|
|
|
progress_next += file_size / 16;
|
|
|
|
ui_progress__update(progress, file_pos);
|
|
|
|
}
|
|
|
|
|
2010-11-30 10:49:44 -07:00
|
|
|
if (file_pos < file_size)
|
2009-12-27 16:36:59 -07:00
|
|
|
goto more;
|
2010-11-30 10:49:44 -07:00
|
|
|
|
2009-12-27 16:36:59 -07:00
|
|
|
err = 0;
|
2010-04-23 16:04:12 -06:00
|
|
|
/* do the final flush for ordered samples */
|
2010-11-30 10:49:38 -07:00
|
|
|
session->ordered_samples.next_flush = ULLONG_MAX;
|
|
|
|
flush_sample_queue(session, ops);
|
2009-12-27 16:36:59 -07:00
|
|
|
out_err:
|
2010-03-26 18:16:22 -06:00
|
|
|
ui_progress__delete(progress);
|
2010-11-26 21:41:01 -07:00
|
|
|
|
|
|
|
if (ops->lost == event__process_lost &&
|
2010-11-30 10:49:38 -07:00
|
|
|
session->hists.stats.total_lost != 0) {
|
2010-11-26 21:41:01 -07:00
|
|
|
ui__warning("Processed %Lu events and LOST %Lu!\n\n"
|
|
|
|
"Check IO/CPU overload!\n\n",
|
2010-11-30 10:49:38 -07:00
|
|
|
session->hists.stats.total_period,
|
|
|
|
session->hists.stats.total_lost);
|
2010-11-26 21:41:01 -07:00
|
|
|
}
|
2010-11-30 10:49:38 -07:00
|
|
|
|
|
|
|
if (session->hists.stats.nr_unknown_events != 0) {
|
2010-11-26 21:41:01 -07:00
|
|
|
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",
|
2010-11-30 10:49:38 -07:00
|
|
|
session->hists.stats.nr_unknown_events);
|
2010-11-26 21:41:01 -07:00
|
|
|
}
|
2010-11-30 10:49:38 -07:00
|
|
|
|
perf session: Parse sample earlier
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <1291318772-30880-4-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2010-12-02 09:10:21 -07:00
|
|
|
if (session->hists.stats.nr_invalid_chains != 0) {
|
|
|
|
ui__warning("Found invalid callchains!\n\n"
|
|
|
|
"%u out of %u events were discarded for this reason.\n\n"
|
|
|
|
"Consider reporting to linux-kernel@vger.kernel.org.\n\n",
|
|
|
|
session->hists.stats.nr_invalid_chains,
|
|
|
|
session->hists.stats.nr_events[PERF_RECORD_SAMPLE]);
|
|
|
|
}
|
|
|
|
|
2010-11-30 10:49:53 -07:00
|
|
|
perf_session_free_sample_buffers(session);
|
2009-12-27 16:36:59 -07:00
|
|
|
return err;
|
|
|
|
}
|
2009-12-27 16:37:01 -07:00
|
|
|
|
2010-02-03 11:52:05 -07:00
|
|
|
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;
|
|
|
|
|
2010-04-01 22:59:15 -06:00
|
|
|
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);
|
2010-07-27 08:46:12 -06:00
|
|
|
|
2010-02-03 11:52:05 -07:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2009-12-27 16:37:02 -07:00
|
|
|
bool perf_session__has_traces(struct perf_session *self, const char *msg)
|
2009-12-27 16:37:01 -07:00
|
|
|
{
|
|
|
|
if (!(self->sample_type & PERF_SAMPLE_RAW)) {
|
2009-12-27 16:37:02 -07:00
|
|
|
pr_err("No trace sample to read. Did you call 'perf %s'?\n", msg);
|
|
|
|
return false;
|
2009-12-27 16:37:01 -07:00
|
|
|
}
|
|
|
|
|
2009-12-27 16:37:02 -07:00
|
|
|
return true;
|
2009-12-27 16:37:01 -07:00
|
|
|
}
|
2010-01-05 11:50:31 -07:00
|
|
|
|
2010-04-18 23:32:50 -06:00
|
|
|
int perf_session__set_kallsyms_ref_reloc_sym(struct map **maps,
|
2010-01-05 11:50:31 -07:00
|
|
|
const char *symbol_name,
|
|
|
|
u64 addr)
|
|
|
|
{
|
|
|
|
char *bracket;
|
2010-02-03 11:52:00 -07:00
|
|
|
enum map_type i;
|
2010-04-18 23:32:50 -06:00
|
|
|
struct ref_reloc_sym *ref;
|
|
|
|
|
|
|
|
ref = zalloc(sizeof(struct ref_reloc_sym));
|
|
|
|
if (ref == NULL)
|
|
|
|
return -ENOMEM;
|
2010-01-05 11:50:31 -07:00
|
|
|
|
2010-04-18 23:32:50 -06:00
|
|
|
ref->name = strdup(symbol_name);
|
|
|
|
if (ref->name == NULL) {
|
|
|
|
free(ref);
|
2010-01-05 11:50:31 -07:00
|
|
|
return -ENOMEM;
|
2010-04-18 23:32:50 -06:00
|
|
|
}
|
2010-01-05 11:50:31 -07:00
|
|
|
|
2010-04-18 23:32:50 -06:00
|
|
|
bracket = strchr(ref->name, ']');
|
2010-01-05 11:50:31 -07:00
|
|
|
if (bracket)
|
|
|
|
*bracket = '\0';
|
|
|
|
|
2010-04-18 23:32:50 -06:00
|
|
|
ref->addr = addr;
|
2010-02-03 11:52:00 -07:00
|
|
|
|
|
|
|
for (i = 0; i < MAP__NR_TYPES; ++i) {
|
2010-04-18 23:32:50 -06:00
|
|
|
struct kmap *kmap = map__kmap(maps[i]);
|
|
|
|
kmap->ref_reloc_sym = ref;
|
2010-02-03 11:52:00 -07:00
|
|
|
}
|
|
|
|
|
2010-01-05 11:50:31 -07:00
|
|
|
return 0;
|
|
|
|
}
|
2010-05-09 16:57:08 -06:00
|
|
|
|
|
|
|
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);
|
|
|
|
}
|
2010-05-19 10:41:23 -06:00
|
|
|
|
|
|
|
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);
|
|
|
|
}
|