2009-04-20 20:00:56 +07:00
|
|
|
|
/*
|
2009-06-03 04:37:05 +07:00
|
|
|
|
* builtin-top.c
|
|
|
|
|
*
|
|
|
|
|
* Builtin top command: Display a continuously updated profile of
|
|
|
|
|
* any workload, CPU or specific PID.
|
|
|
|
|
*
|
|
|
|
|
* Copyright (C) 2008, Red Hat Inc, Ingo Molnar <mingo@redhat.com>
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
* 2011, Red Hat Inc, Arnaldo Carvalho de Melo <acme@redhat.com>
|
2009-06-03 04:37:05 +07:00
|
|
|
|
*
|
|
|
|
|
* Improvements and fixes by:
|
|
|
|
|
*
|
|
|
|
|
* Arjan van de Ven <arjan@linux.intel.com>
|
|
|
|
|
* Yanmin Zhang <yanmin.zhang@intel.com>
|
|
|
|
|
* Wu Fengguang <fengguang.wu@intel.com>
|
|
|
|
|
* Mike Galbraith <efault@gmx.de>
|
|
|
|
|
* Paul Mackerras <paulus@samba.org>
|
|
|
|
|
*
|
|
|
|
|
* Released under the GPL v2. (and only v2, not any later version)
|
2009-04-20 20:00:56 +07:00
|
|
|
|
*/
|
2009-06-03 04:37:05 +07:00
|
|
|
|
#include "builtin.h"
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2009-05-23 23:28:58 +07:00
|
|
|
|
#include "perf.h"
|
2009-06-03 04:37:05 +07:00
|
|
|
|
|
2011-02-06 23:54:44 +07:00
|
|
|
|
#include "util/annotate.h"
|
2011-02-01 03:19:33 +07:00
|
|
|
|
#include "util/cache.h"
|
2009-06-04 20:19:47 +07:00
|
|
|
|
#include "util/color.h"
|
2011-01-12 05:56:53 +07:00
|
|
|
|
#include "util/evlist.h"
|
2011-01-04 01:39:04 +07:00
|
|
|
|
#include "util/evsel.h"
|
2009-12-14 04:50:28 +07:00
|
|
|
|
#include "util/session.h"
|
|
|
|
|
#include "util/symbol.h"
|
2009-10-02 13:29:58 +07:00
|
|
|
|
#include "util/thread.h"
|
2011-01-19 00:15:24 +07:00
|
|
|
|
#include "util/thread_map.h"
|
2011-01-31 23:50:39 +07:00
|
|
|
|
#include "util/top.h"
|
2009-04-27 13:02:14 +07:00
|
|
|
|
#include "util/util.h"
|
2009-07-01 22:28:37 +07:00
|
|
|
|
#include <linux/rbtree.h>
|
2009-05-26 14:17:18 +07:00
|
|
|
|
#include "util/parse-options.h"
|
|
|
|
|
#include "util/parse-events.h"
|
perf tools: Fix sparse CPU numbering related bugs
At present, the perf subcommands that do system-wide monitoring
(perf stat, perf record and perf top) don't work properly unless
the online cpus are numbered 0, 1, ..., N-1. These tools ask
for the number of online cpus with sysconf(_SC_NPROCESSORS_ONLN)
and then try to create events for cpus 0, 1, ..., N-1.
This creates problems for systems where the online cpus are
numbered sparsely. For example, a POWER6 system in
single-threaded mode (i.e. only running 1 hardware thread per
core) will have only even-numbered cpus online.
This fixes the problem by reading the /sys/devices/system/cpu/online
file to find out which cpus are online. The code that does that is in
tools/perf/util/cpumap.[ch], and consists of a read_cpu_map()
function that sets up a cpumap[] array and returns the number of
online cpus. If /sys/devices/system/cpu/online can't be read or
can't be parsed successfully, it falls back to using sysconf to
ask how many cpus are online and sets up an identity map in cpumap[].
The perf record, perf stat and perf top code then calls
read_cpu_map() in the system-wide monitoring case (instead of
sysconf) and uses cpumap[] to get the cpu numbers to pass to
perf_event_open.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Cc: Anton Blanchard <anton@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
LKML-Reference: <20100310093609.GA3959@brick.ozlabs.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2010-03-10 16:36:09 +07:00
|
|
|
|
#include "util/cpumap.h"
|
2011-01-04 01:39:04 +07:00
|
|
|
|
#include "util/xyarray.h"
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
#include "util/sort.h"
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2009-08-17 03:05:48 +07:00
|
|
|
|
#include "util/debug.h"
|
|
|
|
|
|
2009-04-20 20:00:56 +07:00
|
|
|
|
#include <assert.h>
|
|
|
|
|
#include <fcntl.h>
|
2009-05-26 14:17:18 +07:00
|
|
|
|
|
2009-04-20 20:00:56 +07:00
|
|
|
|
#include <stdio.h>
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
#include <termios.h>
|
|
|
|
|
#include <unistd.h>
|
2011-01-23 05:37:02 +07:00
|
|
|
|
#include <inttypes.h>
|
2009-05-26 14:17:18 +07:00
|
|
|
|
|
2009-04-20 20:00:56 +07:00
|
|
|
|
#include <errno.h>
|
|
|
|
|
#include <time.h>
|
|
|
|
|
#include <sched.h>
|
|
|
|
|
|
|
|
|
|
#include <sys/syscall.h>
|
|
|
|
|
#include <sys/ioctl.h>
|
|
|
|
|
#include <sys/poll.h>
|
|
|
|
|
#include <sys/prctl.h>
|
|
|
|
|
#include <sys/wait.h>
|
|
|
|
|
#include <sys/uio.h>
|
|
|
|
|
#include <sys/mman.h>
|
|
|
|
|
|
|
|
|
|
#include <linux/unistd.h>
|
|
|
|
|
#include <linux/types.h>
|
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
static struct perf_top top = {
|
|
|
|
|
.count_filter = 5,
|
|
|
|
|
.delay_secs = 2,
|
|
|
|
|
.target_pid = -1,
|
|
|
|
|
.target_tid = -1,
|
|
|
|
|
.freq = 1000, /* 1 KHz */
|
|
|
|
|
};
|
2011-01-12 05:56:53 +07:00
|
|
|
|
|
2010-04-13 15:37:33 +07:00
|
|
|
|
static bool system_wide = false;
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2011-02-01 03:19:33 +07:00
|
|
|
|
static bool use_tui, use_stdio;
|
|
|
|
|
|
2011-10-06 05:30:22 +07:00
|
|
|
|
static bool sort_has_symbols;
|
|
|
|
|
|
|
|
|
|
static bool dont_use_callchains;
|
|
|
|
|
static char callchain_default_opt[] = "fractal,0.5,callee";
|
|
|
|
|
|
|
|
|
|
|
2009-10-12 12:56:03 +07:00
|
|
|
|
static int default_interval = 0;
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2011-05-27 21:53:28 +07:00
|
|
|
|
static bool kptr_restrict_warned;
|
2011-05-27 23:42:16 +07:00
|
|
|
|
static bool vmlinux_warned;
|
2010-04-13 15:37:33 +07:00
|
|
|
|
static bool inherit = false;
|
2010-05-18 01:39:16 +07:00
|
|
|
|
static int realtime_prio = 0;
|
2010-04-13 15:37:33 +07:00
|
|
|
|
static bool group = false;
|
2011-10-29 21:15:04 +07:00
|
|
|
|
static bool sample_id_all_avail = true;
|
2011-01-13 07:39:13 +07:00
|
|
|
|
static unsigned int mmap_pages = 128;
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2010-04-13 15:37:33 +07:00
|
|
|
|
static bool dump_symtab = false;
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2009-11-18 00:40:54 +07:00
|
|
|
|
static struct winsize winsize;
|
2009-11-17 06:45:24 +07:00
|
|
|
|
|
2010-05-18 02:22:41 +07:00
|
|
|
|
static const char *sym_filter = NULL;
|
2009-10-06 20:14:21 +07:00
|
|
|
|
static int sym_pcnt_filter = 5;
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
/*
|
|
|
|
|
* Source functions
|
|
|
|
|
*/
|
|
|
|
|
|
2010-03-12 06:12:41 +07:00
|
|
|
|
void get_term_dimensions(struct winsize *ws)
|
2009-11-17 04:30:27 +07:00
|
|
|
|
{
|
2009-11-18 00:40:54 +07:00
|
|
|
|
char *s = getenv("LINES");
|
|
|
|
|
|
|
|
|
|
if (s != NULL) {
|
|
|
|
|
ws->ws_row = atoi(s);
|
|
|
|
|
s = getenv("COLUMNS");
|
|
|
|
|
if (s != NULL) {
|
|
|
|
|
ws->ws_col = atoi(s);
|
|
|
|
|
if (ws->ws_row && ws->ws_col)
|
|
|
|
|
return;
|
|
|
|
|
}
|
2009-11-17 04:30:27 +07:00
|
|
|
|
}
|
2009-11-18 00:40:54 +07:00
|
|
|
|
#ifdef TIOCGWINSZ
|
|
|
|
|
if (ioctl(1, TIOCGWINSZ, ws) == 0 &&
|
|
|
|
|
ws->ws_row && ws->ws_col)
|
|
|
|
|
return;
|
2009-11-17 04:30:27 +07:00
|
|
|
|
#endif
|
2009-11-18 00:40:54 +07:00
|
|
|
|
ws->ws_row = 25;
|
|
|
|
|
ws->ws_col = 80;
|
2009-11-17 04:30:27 +07:00
|
|
|
|
}
|
|
|
|
|
|
2009-11-18 00:40:54 +07:00
|
|
|
|
static void update_print_entries(struct winsize *ws)
|
2009-11-17 04:30:27 +07:00
|
|
|
|
{
|
2011-01-31 23:50:39 +07:00
|
|
|
|
top.print_entries = ws->ws_row;
|
2009-11-18 00:40:54 +07:00
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
if (top.print_entries > 9)
|
|
|
|
|
top.print_entries -= 9;
|
2009-11-17 04:30:27 +07:00
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void sig_winch_handler(int sig __used)
|
|
|
|
|
{
|
2009-11-18 00:40:54 +07:00
|
|
|
|
get_term_dimensions(&winsize);
|
|
|
|
|
update_print_entries(&winsize);
|
2009-11-17 04:30:27 +07:00
|
|
|
|
}
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
static int parse_source(struct hist_entry *he)
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
{
|
|
|
|
|
struct symbol *sym;
|
2011-02-08 22:27:39 +07:00
|
|
|
|
struct annotation *notes;
|
2009-10-02 13:29:58 +07:00
|
|
|
|
struct map *map;
|
2011-02-06 23:54:44 +07:00
|
|
|
|
int err = -1;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
if (!he || !he->ms.sym)
|
2010-03-15 21:46:58 +07:00
|
|
|
|
return -1;
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
sym = he->ms.sym;
|
|
|
|
|
map = he->ms.map;
|
2010-03-15 21:46:58 +07:00
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* We can't annotate with just /proc/kallsyms
|
|
|
|
|
*/
|
2011-03-11 23:13:36 +07:00
|
|
|
|
if (map->dso->symtab_type == SYMTAB__KALLSYMS) {
|
2011-02-08 22:27:39 +07:00
|
|
|
|
pr_err("Can't annotate %s: No vmlinux file was found in the "
|
|
|
|
|
"path\n", sym->name);
|
|
|
|
|
sleep(1);
|
2010-03-15 21:46:58 +07:00
|
|
|
|
return -1;
|
2009-11-18 03:38:02 +07:00
|
|
|
|
}
|
|
|
|
|
|
2011-02-08 22:27:39 +07:00
|
|
|
|
notes = symbol__annotation(sym);
|
|
|
|
|
if (notes->src != NULL) {
|
|
|
|
|
pthread_mutex_lock(¬es->lock);
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
goto out_assign;
|
|
|
|
|
}
|
|
|
|
|
|
2011-02-08 22:27:39 +07:00
|
|
|
|
pthread_mutex_lock(¬es->lock);
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
|
2011-02-06 23:54:44 +07:00
|
|
|
|
if (symbol__alloc_hist(sym, top.evlist->nr_entries) < 0) {
|
2011-02-22 22:02:07 +07:00
|
|
|
|
pthread_mutex_unlock(¬es->lock);
|
2011-02-06 23:54:44 +07:00
|
|
|
|
pr_err("Not enough memory for annotating '%s' symbol!\n",
|
|
|
|
|
sym->name);
|
2011-02-08 22:27:39 +07:00
|
|
|
|
sleep(1);
|
2011-02-22 22:02:07 +07:00
|
|
|
|
return err;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
}
|
2011-02-06 23:54:44 +07:00
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
err = symbol__annotate(sym, map, 0);
|
2011-02-06 23:54:44 +07:00
|
|
|
|
if (err == 0) {
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
out_assign:
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
top.sym_filter_entry = he;
|
2011-02-06 23:54:44 +07:00
|
|
|
|
}
|
2011-02-22 22:02:07 +07:00
|
|
|
|
|
2011-02-08 22:27:39 +07:00
|
|
|
|
pthread_mutex_unlock(¬es->lock);
|
2011-02-06 23:54:44 +07:00
|
|
|
|
return err;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
}
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
static void __zero_source_counters(struct hist_entry *he)
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
{
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
struct symbol *sym = he->ms.sym;
|
2011-02-06 23:54:44 +07:00
|
|
|
|
symbol__annotate_zero_histograms(sym);
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
}
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
static void record_precise_ip(struct hist_entry *he, int counter, u64 ip)
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
{
|
2011-02-08 22:27:39 +07:00
|
|
|
|
struct annotation *notes;
|
|
|
|
|
struct symbol *sym;
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
if (he == NULL || he->ms.sym == NULL ||
|
2011-11-03 20:31:26 +07:00
|
|
|
|
((top.sym_filter_entry == NULL ||
|
|
|
|
|
top.sym_filter_entry->ms.sym != he->ms.sym) && use_browser != 1))
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
return;
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
sym = he->ms.sym;
|
2011-02-08 22:27:39 +07:00
|
|
|
|
notes = symbol__annotation(sym);
|
|
|
|
|
|
|
|
|
|
if (pthread_mutex_trylock(¬es->lock))
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
return;
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
if (notes->src == NULL &&
|
|
|
|
|
symbol__alloc_hist(sym, top.evlist->nr_entries) < 0) {
|
|
|
|
|
pthread_mutex_unlock(¬es->lock);
|
|
|
|
|
pr_err("Not enough memory for annotating '%s' symbol!\n",
|
|
|
|
|
sym->name);
|
|
|
|
|
sleep(1);
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
ip = he->ms.map->map_ip(he->ms.map, ip);
|
|
|
|
|
symbol__inc_addr_samples(sym, he->ms.map, counter, ip);
|
2010-02-23 02:14:22 +07:00
|
|
|
|
|
2011-02-08 22:27:39 +07:00
|
|
|
|
pthread_mutex_unlock(¬es->lock);
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
}
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
static void show_details(struct hist_entry *he)
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
{
|
2011-02-08 22:27:39 +07:00
|
|
|
|
struct annotation *notes;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
struct symbol *symbol;
|
2011-02-06 23:54:44 +07:00
|
|
|
|
int more;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
if (!he)
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
return;
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
symbol = he->ms.sym;
|
2011-02-08 22:27:39 +07:00
|
|
|
|
notes = symbol__annotation(symbol);
|
|
|
|
|
|
|
|
|
|
pthread_mutex_lock(¬es->lock);
|
|
|
|
|
|
|
|
|
|
if (notes->src == NULL)
|
|
|
|
|
goto out_unlock;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
printf("Showing %s for %s\n", event_name(top.sym_evsel), symbol->name);
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
printf(" Events Pcnt (>=%d%%)\n", sym_pcnt_filter);
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
more = symbol__annotate_printf(symbol, he->ms.map, top.sym_evsel->idx,
|
2011-02-09 00:29:25 +07:00
|
|
|
|
0, sym_pcnt_filter, top.print_entries, 4);
|
2011-02-06 23:54:44 +07:00
|
|
|
|
if (top.zero)
|
|
|
|
|
symbol__annotate_zero_histogram(symbol, top.sym_evsel->idx);
|
|
|
|
|
else
|
2011-02-08 22:27:39 +07:00
|
|
|
|
symbol__annotate_decay_histogram(symbol, top.sym_evsel->idx);
|
2011-02-06 23:54:44 +07:00
|
|
|
|
if (more != 0)
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
printf("%d lines not displayed, maybe increase display entries [e]\n", more);
|
2011-02-08 22:27:39 +07:00
|
|
|
|
out_unlock:
|
|
|
|
|
pthread_mutex_unlock(¬es->lock);
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
}
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
|
|
|
|
static const char CONSOLE_CLEAR[] = "[H[2J";
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
static struct hist_entry *
|
|
|
|
|
perf_session__add_hist_entry(struct perf_session *session,
|
|
|
|
|
struct addr_location *al,
|
|
|
|
|
struct perf_sample *sample,
|
|
|
|
|
struct perf_evsel *evsel)
|
2009-05-29 00:55:41 +07:00
|
|
|
|
{
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
struct hist_entry *he;
|
|
|
|
|
|
|
|
|
|
he = __hists__add_entry(&evsel->hists, al, NULL, sample->period);
|
|
|
|
|
if (he == NULL)
|
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
|
|
session->hists.stats.total_period += sample->period;
|
|
|
|
|
hists__inc_nr_events(&evsel->hists, PERF_RECORD_SAMPLE);
|
|
|
|
|
return he;
|
2009-05-29 00:55:41 +07:00
|
|
|
|
}
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2011-09-02 00:27:58 +07:00
|
|
|
|
static void print_sym_table(void)
|
2009-04-20 20:00:56 +07:00
|
|
|
|
{
|
2011-01-31 23:50:39 +07:00
|
|
|
|
char bf[160];
|
|
|
|
|
int printed = 0;
|
2009-11-18 00:40:54 +07:00
|
|
|
|
const int win_width = winsize.ws_col - 1;
|
perf top: Reduce display overhead
Iterate over the symbol table once per display interval, and
copy/sort/tally/decay only those symbols which are active.
Before:
top - 10:14:53 up 4:08, 17 users, load average: 1.17, 1.53, 1.49
Tasks: 273 total, 5 running, 268 sleeping, 0 stopped, 0 zombie
Cpu(s): 6.9%us, 38.2%sy, 0.0%ni, 19.9%id, 0.0%wa, 0.0%hi, 35.0%si, 0.0%st
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ P COMMAND
28504 root 20 0 1044 260 164 S 58 0.0 0:04.19 2 netserver
28499 root 20 0 1040 412 316 R 51 0.0 0:04.15 0 netperf
28500 root 20 0 1040 408 316 R 50 0.0 0:04.14 1 netperf
28503 root 20 0 1044 260 164 S 50 0.0 0:04.01 1 netserver
28501 root 20 0 1044 260 164 S 49 0.0 0:03.99 0 netserver
28502 root 20 0 1040 412 316 S 43 0.0 0:03.96 2 netperf
28468 root 20 0 1892m 325m 972 S 16 10.8 0:10.50 3 perf
28467 root 20 0 1892m 325m 972 R 2 10.8 0:00.72 3 perf
After:
top - 10:16:30 up 4:10, 17 users, load average: 2.27, 1.88, 1.62
Tasks: 273 total, 6 running, 267 sleeping, 0 stopped, 0 zombie
Cpu(s): 2.5%us, 39.7%sy, 0.0%ni, 24.6%id, 0.0%wa, 0.0%hi, 33.3%si, 0.0%st
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ P COMMAND
28590 root 20 0 1040 412 316 S 54 0.0 0:07.85 2 netperf
28589 root 20 0 1044 260 164 R 54 0.0 0:07.84 0 netserver
28588 root 20 0 1040 412 316 R 50 0.0 0:07.89 1 netperf
28591 root 20 0 1044 256 164 S 50 0.0 0:07.82 1 netserver
28587 root 20 0 1040 408 316 R 47 0.0 0:07.61 0 netperf
28592 root 20 0 1044 260 164 R 47 0.0 0:07.85 2 netserver
28378 root 20 0 8732 1300 860 R 2 0.0 0:01.81 3 top
28577 root 20 0 1892m 165m 972 R 2 5.5 0:00.48 3 perf
28578 root 20 0 1892m 165m 972 S 2 5.5 0:00.04 3 perf
[ Impact: optimization ]
Signed-off-by: Mike Galbraith <efault@gmx.de>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-05-25 14:57:56 +07:00
|
|
|
|
|
2009-06-05 01:48:04 +07:00
|
|
|
|
puts(CONSOLE_CLEAR);
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
perf_top__header_snprintf(&top, bf, sizeof(bf));
|
|
|
|
|
printf("%s\n", bf);
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
perf_top__reset_sample_counters(&top);
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2009-11-18 00:40:55 +07:00
|
|
|
|
printf("%-*.*s\n", win_width, win_width, graph_dotted_line);
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2011-10-29 21:15:04 +07:00
|
|
|
|
if (top.sym_evsel->hists.stats.nr_lost_warned !=
|
|
|
|
|
top.sym_evsel->hists.stats.nr_events[PERF_RECORD_LOST]) {
|
|
|
|
|
top.sym_evsel->hists.stats.nr_lost_warned =
|
|
|
|
|
top.sym_evsel->hists.stats.nr_events[PERF_RECORD_LOST];
|
|
|
|
|
color_fprintf(stdout, PERF_COLOR_RED,
|
|
|
|
|
"WARNING: LOST %d chunks, Check IO/CPU overload",
|
|
|
|
|
top.sym_evsel->hists.stats.nr_lost_warned);
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
++printed;
|
2011-01-29 21:08:00 +07:00
|
|
|
|
}
|
|
|
|
|
|
2011-02-22 22:02:07 +07:00
|
|
|
|
if (top.sym_filter_entry) {
|
|
|
|
|
show_details(top.sym_filter_entry);
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
hists__collapse_resort_threaded(&top.sym_evsel->hists);
|
|
|
|
|
hists__output_resort_threaded(&top.sym_evsel->hists);
|
2011-10-17 18:05:04 +07:00
|
|
|
|
hists__decay_entries_threaded(&top.sym_evsel->hists,
|
|
|
|
|
top.hide_user_symbols,
|
|
|
|
|
top.hide_kernel_symbols);
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
hists__output_recalc_col_len(&top.sym_evsel->hists, winsize.ws_row - 3);
|
2009-11-24 21:05:14 +07:00
|
|
|
|
putchar('\n');
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
hists__fprintf(&top.sym_evsel->hists, NULL, false, false,
|
|
|
|
|
winsize.ws_row - 4 - printed, win_width, stdout);
|
2009-04-20 20:00:56 +07:00
|
|
|
|
}
|
|
|
|
|
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
static void prompt_integer(int *target, const char *msg)
|
|
|
|
|
{
|
|
|
|
|
char *buf = malloc(0), *p;
|
|
|
|
|
size_t dummy = 0;
|
|
|
|
|
int tmp;
|
|
|
|
|
|
|
|
|
|
fprintf(stdout, "\n%s: ", msg);
|
|
|
|
|
if (getline(&buf, &dummy, stdin) < 0)
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
p = strchr(buf, '\n');
|
|
|
|
|
if (p)
|
|
|
|
|
*p = 0;
|
|
|
|
|
|
|
|
|
|
p = buf;
|
|
|
|
|
while(*p) {
|
|
|
|
|
if (!isdigit(*p))
|
|
|
|
|
goto out_free;
|
|
|
|
|
p++;
|
|
|
|
|
}
|
|
|
|
|
tmp = strtoul(buf, NULL, 10);
|
|
|
|
|
*target = tmp;
|
|
|
|
|
out_free:
|
|
|
|
|
free(buf);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static void prompt_percent(int *target, const char *msg)
|
|
|
|
|
{
|
|
|
|
|
int tmp = 0;
|
|
|
|
|
|
|
|
|
|
prompt_integer(&tmp, msg);
|
|
|
|
|
if (tmp >= 0 && tmp <= 100)
|
|
|
|
|
*target = tmp;
|
|
|
|
|
}
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
static void prompt_symbol(struct hist_entry **target, const char *msg)
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
{
|
|
|
|
|
char *buf = malloc(0), *p;
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
struct hist_entry *syme = *target, *n, *found = NULL;
|
|
|
|
|
struct rb_node *next;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
size_t dummy = 0;
|
|
|
|
|
|
|
|
|
|
/* zero counters of active symbol */
|
|
|
|
|
if (syme) {
|
|
|
|
|
__zero_source_counters(syme);
|
|
|
|
|
*target = NULL;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
fprintf(stdout, "\n%s: ", msg);
|
|
|
|
|
if (getline(&buf, &dummy, stdin) < 0)
|
|
|
|
|
goto out_free;
|
|
|
|
|
|
|
|
|
|
p = strchr(buf, '\n');
|
|
|
|
|
if (p)
|
|
|
|
|
*p = 0;
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
next = rb_first(&top.sym_evsel->hists.entries);
|
|
|
|
|
while (next) {
|
|
|
|
|
n = rb_entry(next, struct hist_entry, rb_node);
|
|
|
|
|
if (n->ms.sym && !strcmp(buf, n->ms.sym->name)) {
|
|
|
|
|
found = n;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
break;
|
|
|
|
|
}
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
next = rb_next(&n->rb_node);
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (!found) {
|
2010-01-13 22:22:19 +07:00
|
|
|
|
fprintf(stderr, "Sorry, %s is not active.\n", buf);
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
sleep(1);
|
|
|
|
|
return;
|
|
|
|
|
} else
|
|
|
|
|
parse_source(found);
|
|
|
|
|
|
|
|
|
|
out_free:
|
|
|
|
|
free(buf);
|
|
|
|
|
}
|
|
|
|
|
|
2009-08-04 15:21:23 +07:00
|
|
|
|
static void print_mapped_keys(void)
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
{
|
2009-08-04 15:21:23 +07:00
|
|
|
|
char *name = NULL;
|
|
|
|
|
|
2011-02-22 22:02:07 +07:00
|
|
|
|
if (top.sym_filter_entry) {
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
struct symbol *sym = top.sym_filter_entry->ms.sym;
|
2009-08-04 15:21:23 +07:00
|
|
|
|
name = sym->name;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
fprintf(stdout, "\nMapped keys:\n");
|
2011-01-31 23:50:39 +07:00
|
|
|
|
fprintf(stdout, "\t[d] display refresh delay. \t(%d)\n", top.delay_secs);
|
|
|
|
|
fprintf(stdout, "\t[e] display entries (lines). \t(%d)\n", top.print_entries);
|
2009-08-04 15:21:23 +07:00
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
if (top.evlist->nr_entries > 1)
|
|
|
|
|
fprintf(stdout, "\t[E] active event counter. \t(%s)\n", event_name(top.sym_evsel));
|
2009-08-04 15:21:23 +07:00
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
fprintf(stdout, "\t[f] profile display filter (count). \t(%d)\n", top.count_filter);
|
2009-08-04 15:21:23 +07:00
|
|
|
|
|
2010-02-04 01:52:08 +07:00
|
|
|
|
fprintf(stdout, "\t[F] annotate display filter (percent). \t(%d%%)\n", sym_pcnt_filter);
|
|
|
|
|
fprintf(stdout, "\t[s] annotate symbol. \t(%s)\n", name?: "NULL");
|
|
|
|
|
fprintf(stdout, "\t[S] stop annotation.\n");
|
2009-08-04 15:21:23 +07:00
|
|
|
|
|
2009-11-17 06:45:24 +07:00
|
|
|
|
fprintf(stdout,
|
2010-02-12 23:20:59 +07:00
|
|
|
|
"\t[K] hide kernel_symbols symbols. \t(%s)\n",
|
2011-01-31 23:50:39 +07:00
|
|
|
|
top.hide_kernel_symbols ? "yes" : "no");
|
2009-11-17 06:45:24 +07:00
|
|
|
|
fprintf(stdout,
|
|
|
|
|
"\t[U] hide user symbols. \t(%s)\n",
|
2011-01-31 23:50:39 +07:00
|
|
|
|
top.hide_user_symbols ? "yes" : "no");
|
|
|
|
|
fprintf(stdout, "\t[z] toggle sample zeroing. \t(%d)\n", top.zero ? 1 : 0);
|
2009-08-04 15:21:23 +07:00
|
|
|
|
fprintf(stdout, "\t[qQ] quit.\n");
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int key_mapped(int c)
|
|
|
|
|
{
|
|
|
|
|
switch (c) {
|
|
|
|
|
case 'd':
|
|
|
|
|
case 'e':
|
|
|
|
|
case 'f':
|
|
|
|
|
case 'z':
|
|
|
|
|
case 'q':
|
|
|
|
|
case 'Q':
|
2009-11-17 06:45:24 +07:00
|
|
|
|
case 'K':
|
|
|
|
|
case 'U':
|
2010-02-04 01:52:08 +07:00
|
|
|
|
case 'F':
|
|
|
|
|
case 's':
|
|
|
|
|
case 'S':
|
2009-08-04 15:21:23 +07:00
|
|
|
|
return 1;
|
|
|
|
|
case 'E':
|
2011-10-17 18:10:24 +07:00
|
|
|
|
return top.evlist->nr_entries > 1 ? 1 : 0;
|
2009-08-15 17:26:57 +07:00
|
|
|
|
default:
|
|
|
|
|
break;
|
2009-08-04 15:21:23 +07:00
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return 0;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
}
|
|
|
|
|
|
2011-09-02 00:27:58 +07:00
|
|
|
|
static void handle_keypress(int c)
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
{
|
2009-08-04 15:21:23 +07:00
|
|
|
|
if (!key_mapped(c)) {
|
|
|
|
|
struct pollfd stdin_poll = { .fd = 0, .events = POLLIN };
|
|
|
|
|
struct termios tc, save;
|
|
|
|
|
|
|
|
|
|
print_mapped_keys();
|
|
|
|
|
fprintf(stdout, "\nEnter selection, or unmapped key to continue: ");
|
|
|
|
|
fflush(stdout);
|
|
|
|
|
|
|
|
|
|
tcgetattr(0, &save);
|
|
|
|
|
tc = save;
|
|
|
|
|
tc.c_lflag &= ~(ICANON | ECHO);
|
|
|
|
|
tc.c_cc[VMIN] = 0;
|
|
|
|
|
tc.c_cc[VTIME] = 0;
|
|
|
|
|
tcsetattr(0, TCSANOW, &tc);
|
|
|
|
|
|
|
|
|
|
poll(&stdin_poll, 1, -1);
|
|
|
|
|
c = getc(stdin);
|
|
|
|
|
|
|
|
|
|
tcsetattr(0, TCSAFLUSH, &save);
|
|
|
|
|
if (!key_mapped(c))
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
switch (c) {
|
|
|
|
|
case 'd':
|
2011-01-31 23:50:39 +07:00
|
|
|
|
prompt_integer(&top.delay_secs, "Enter display delay");
|
|
|
|
|
if (top.delay_secs < 1)
|
|
|
|
|
top.delay_secs = 1;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
break;
|
|
|
|
|
case 'e':
|
2011-01-31 23:50:39 +07:00
|
|
|
|
prompt_integer(&top.print_entries, "Enter display entries (lines)");
|
|
|
|
|
if (top.print_entries == 0) {
|
2009-11-18 00:40:54 +07:00
|
|
|
|
sig_winch_handler(SIGWINCH);
|
2009-11-17 04:30:27 +07:00
|
|
|
|
signal(SIGWINCH, sig_winch_handler);
|
|
|
|
|
} else
|
|
|
|
|
signal(SIGWINCH, SIG_DFL);
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
break;
|
|
|
|
|
case 'E':
|
2011-01-31 23:50:39 +07:00
|
|
|
|
if (top.evlist->nr_entries > 1) {
|
2011-03-23 14:29:39 +07:00
|
|
|
|
/* Select 0 as the default event: */
|
|
|
|
|
int counter = 0;
|
|
|
|
|
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
fprintf(stderr, "\nAvailable events:");
|
2011-01-04 01:39:04 +07:00
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
list_for_each_entry(top.sym_evsel, &top.evlist->entries, node)
|
|
|
|
|
fprintf(stderr, "\n\t%d %s", top.sym_evsel->idx, event_name(top.sym_evsel));
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
|
2011-03-11 20:11:48 +07:00
|
|
|
|
prompt_integer(&counter, "Enter details event counter");
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
|
2011-03-11 20:11:48 +07:00
|
|
|
|
if (counter >= top.evlist->nr_entries) {
|
2011-01-31 23:50:39 +07:00
|
|
|
|
top.sym_evsel = list_entry(top.evlist->entries.next, struct perf_evsel, node);
|
|
|
|
|
fprintf(stderr, "Sorry, no such event, using %s.\n", event_name(top.sym_evsel));
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
sleep(1);
|
2011-01-04 01:39:04 +07:00
|
|
|
|
break;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
}
|
2011-01-31 23:50:39 +07:00
|
|
|
|
list_for_each_entry(top.sym_evsel, &top.evlist->entries, node)
|
2011-03-11 20:11:48 +07:00
|
|
|
|
if (top.sym_evsel->idx == counter)
|
2011-01-04 01:39:04 +07:00
|
|
|
|
break;
|
2011-03-11 20:11:48 +07:00
|
|
|
|
} else
|
|
|
|
|
top.sym_evsel = list_entry(top.evlist->entries.next, struct perf_evsel, node);
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
break;
|
|
|
|
|
case 'f':
|
2011-01-31 23:50:39 +07:00
|
|
|
|
prompt_integer(&top.count_filter, "Enter display event count filter");
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
break;
|
|
|
|
|
case 'F':
|
|
|
|
|
prompt_percent(&sym_pcnt_filter, "Enter details display event filter (percent)");
|
|
|
|
|
break;
|
2009-11-17 06:45:24 +07:00
|
|
|
|
case 'K':
|
2011-01-31 23:50:39 +07:00
|
|
|
|
top.hide_kernel_symbols = !top.hide_kernel_symbols;
|
2009-11-17 06:45:24 +07:00
|
|
|
|
break;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
case 'q':
|
|
|
|
|
case 'Q':
|
|
|
|
|
printf("exiting.\n");
|
2009-11-21 05:51:27 +07:00
|
|
|
|
if (dump_symtab)
|
2011-09-02 00:27:58 +07:00
|
|
|
|
perf_session__fprintf_dsos(top.session, stderr);
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
exit(0);
|
|
|
|
|
case 's':
|
2011-02-22 22:02:07 +07:00
|
|
|
|
prompt_symbol(&top.sym_filter_entry, "Enter details symbol");
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
break;
|
|
|
|
|
case 'S':
|
2011-02-22 22:02:07 +07:00
|
|
|
|
if (!top.sym_filter_entry)
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
break;
|
|
|
|
|
else {
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
struct hist_entry *syme = top.sym_filter_entry;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
|
2011-02-22 22:02:07 +07:00
|
|
|
|
top.sym_filter_entry = NULL;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
__zero_source_counters(syme);
|
|
|
|
|
}
|
|
|
|
|
break;
|
2009-11-17 06:45:24 +07:00
|
|
|
|
case 'U':
|
2011-01-31 23:50:39 +07:00
|
|
|
|
top.hide_user_symbols = !top.hide_user_symbols;
|
2009-11-17 06:45:24 +07:00
|
|
|
|
break;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
case 'z':
|
2011-01-31 23:50:39 +07:00
|
|
|
|
top.zero = !top.zero;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
break;
|
2009-08-15 17:26:57 +07:00
|
|
|
|
default:
|
|
|
|
|
break;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
static void perf_top__sort_new_samples(void *arg)
|
|
|
|
|
{
|
|
|
|
|
struct perf_top *t = arg;
|
|
|
|
|
perf_top__reset_sample_counters(t);
|
|
|
|
|
|
|
|
|
|
if (t->evlist->selected != NULL)
|
|
|
|
|
t->sym_evsel = t->evlist->selected;
|
|
|
|
|
|
|
|
|
|
hists__collapse_resort_threaded(&t->sym_evsel->hists);
|
|
|
|
|
hists__output_resort_threaded(&t->sym_evsel->hists);
|
2011-10-17 18:05:04 +07:00
|
|
|
|
hists__decay_entries_threaded(&t->sym_evsel->hists,
|
|
|
|
|
top.hide_user_symbols,
|
|
|
|
|
top.hide_kernel_symbols);
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
}
|
|
|
|
|
|
2011-02-01 03:19:33 +07:00
|
|
|
|
static void *display_thread_tui(void *arg __used)
|
|
|
|
|
{
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
const char *help = "For a higher level overview, try: perf top --sort comm,dso";
|
|
|
|
|
|
|
|
|
|
perf_top__sort_new_samples(&top);
|
|
|
|
|
perf_evlist__tui_browse_hists(top.evlist, help,
|
|
|
|
|
perf_top__sort_new_samples,
|
|
|
|
|
&top, top.delay_secs);
|
|
|
|
|
|
2011-02-01 03:19:33 +07:00
|
|
|
|
exit_browser(0);
|
|
|
|
|
exit(0);
|
|
|
|
|
return NULL;
|
|
|
|
|
}
|
|
|
|
|
|
2009-07-01 17:37:06 +07:00
|
|
|
|
static void *display_thread(void *arg __used)
|
2009-04-20 20:00:56 +07:00
|
|
|
|
{
|
2009-06-05 01:48:04 +07:00
|
|
|
|
struct pollfd stdin_poll = { .fd = 0, .events = POLLIN };
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
struct termios tc, save;
|
|
|
|
|
int delay_msecs, c;
|
|
|
|
|
|
|
|
|
|
tcgetattr(0, &save);
|
|
|
|
|
tc = save;
|
|
|
|
|
tc.c_lflag &= ~(ICANON | ECHO);
|
|
|
|
|
tc.c_cc[VMIN] = 0;
|
|
|
|
|
tc.c_cc[VTIME] = 0;
|
2009-08-04 15:21:23 +07:00
|
|
|
|
|
2011-10-13 18:52:46 +07:00
|
|
|
|
pthread__unblock_sigwinch();
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
repeat:
|
2011-01-31 23:50:39 +07:00
|
|
|
|
delay_msecs = top.delay_secs * 1000;
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
tcsetattr(0, TCSANOW, &tc);
|
|
|
|
|
/* trash return*/
|
|
|
|
|
getc(stdin);
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2011-10-13 18:52:46 +07:00
|
|
|
|
while (1) {
|
2011-09-02 00:27:58 +07:00
|
|
|
|
print_sym_table();
|
2011-10-13 18:52:46 +07:00
|
|
|
|
/*
|
|
|
|
|
* Either timeout expired or we got an EINTR due to SIGWINCH,
|
|
|
|
|
* refresh screen in both cases.
|
|
|
|
|
*/
|
|
|
|
|
switch (poll(&stdin_poll, 1, delay_msecs)) {
|
|
|
|
|
case 0:
|
|
|
|
|
continue;
|
|
|
|
|
case -1:
|
|
|
|
|
if (errno == EINTR)
|
|
|
|
|
continue;
|
|
|
|
|
/* Fall trhu */
|
|
|
|
|
default:
|
|
|
|
|
goto process_hotkey;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
process_hotkey:
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
c = getc(stdin);
|
|
|
|
|
tcsetattr(0, TCSAFLUSH, &save);
|
|
|
|
|
|
2011-09-02 00:27:58 +07:00
|
|
|
|
handle_keypress(c);
|
perf_counter tools: Fix/resurrect perf top annotation in a simple interactive form
perf top used to have annotation support, but it has bitrotted and
removed.
This patch restores that: it allows the user to select any symbol
in kernel space for source level annotation on the fly, switch
between event counters and alter display variables. When symbol
details are being displayed, stopping annotation reverts to normal.
known keys:
[d] select display delay.
[e] select display entries (lines).
[E] select annotation event counter.
[f] select normal display count filter.
[F] select annotation display count filter (percentage).
[qQ] quit.
[s] select annotation symbol and start annotation.
[S] stop annotation, revert to normal display.
[z] toggle event count zeroing.
Sample:
------------------------------------------------------------------------------
PerfTop: 16719 irqs/sec kernel:78.7% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing cache-misses for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% /* adjust length to remove Ethernet CRC */
0 0.0% if (!(adapter->flags2 & FLAG2_CRC_STRIPPING))
0 0.0% length -= 4;
436 5.0% f039: 41 f6 84 24 5c 29 00 testb $0x1,0x295c(%r12)
0 0.0% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
0 0.0% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
811 9.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
0 0.0% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
7226 82.6% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Available events:
0 cache-misses
1 cache-references
2 instructions
3 cycles
Enter details event counter: 2
------------------------------------------------------------------------------
PerfTop: 15035 irqs/sec kernel:79.0% [cache-misses/cache-references/instructions/cycles], (all, 4 CPUs)
------------------------------------------------------------------------------
Showing instructions for e1000_clean_rx_irq
Events Pcnt (>=3%)
0 0.0% int *work_done, int work_to_do)
0 0.0% {
175 0.9% eebf: 55 push %rbp
1898 9.8% eec0: 48 89 e5 mov %rsp,%rbp
0 0.0%
0 0.0% i = rx_ring->next_to_clean;
140 0.7% ef0a: 0f b7 41 1a movzwl 0x1a(%rcx),%eax
670 3.4% ef0e: 89 45 ac mov %eax,-0x54(%rbp)
0 0.0% {
0 0.0% memcpy(skb->data + offset, from, len);
91 0.5% f07b: 49 8b b6 e8 00 00 00 mov 0xe8(%r14),%rsi
1153 5.9% f082: 48 8b b8 e8 00 00 00 mov 0xe8(%rax),%rdi
42 0.2% f089: 8b 4d 84 mov -0x7c(%rbp),%ecx
14 0.1% f08c: 48 83 ef 02 sub $0x2,%rdi
0 0.0% f090: 48 83 ee 02 sub $0x2,%rsi
1618 8.3% f094: f3 a4 rep movsb %ds:(%rsi),%es:(%rdi)
0 0.0%
0 0.0% /* return some buffers to hardware, one at a time is too slow */
0 0.0% if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
867 4.5% f0e7: 83 7d b0 0f cmpl $0xf,-0x50(%rbp)
0 0.0%
0 0.0% while (rx_desc->status & E1000_RXD_STAT_DD) {
37 0.2% f114: 41 f6 47 0c 01 testb $0x1,0xc(%r15)
4047 20.8% f119: 0f 85 24 fe ff ff jne ef43 <e1000_clean_rx_irq+0x84>
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-07-23 01:36:03 +07:00
|
|
|
|
goto repeat;
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
|
}
|
|
|
|
|
|
2009-07-01 06:00:46 +07:00
|
|
|
|
/* Tag samples to be skipped. */
|
2009-07-01 17:37:06 +07:00
|
|
|
|
static const char *skip_symbols[] = {
|
2009-07-01 06:00:46 +07:00
|
|
|
|
"default_idle",
|
2011-01-17 02:39:15 +07:00
|
|
|
|
"native_safe_halt",
|
2009-07-01 06:00:46 +07:00
|
|
|
|
"cpu_idle",
|
|
|
|
|
"enter_idle",
|
|
|
|
|
"exit_idle",
|
|
|
|
|
"mwait_idle",
|
2009-07-27 05:06:19 +07:00
|
|
|
|
"mwait_idle_with_hints",
|
2009-09-26 05:02:39 +07:00
|
|
|
|
"poll_idle",
|
2009-07-01 06:00:47 +07:00
|
|
|
|
"ppc64_runlatch_off",
|
|
|
|
|
"pseries_dedicated_idle_sleep",
|
2009-07-01 06:00:46 +07:00
|
|
|
|
NULL
|
|
|
|
|
};
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
static int symbol_filter(struct map *map __used, struct symbol *sym)
|
2009-04-20 20:00:56 +07:00
|
|
|
|
{
|
2009-05-29 00:55:41 +07:00
|
|
|
|
const char *name = sym->name;
|
2009-07-01 06:00:46 +07:00
|
|
|
|
int i;
|
2009-05-29 00:55:41 +07:00
|
|
|
|
|
2009-07-01 06:00:47 +07:00
|
|
|
|
/*
|
|
|
|
|
* ppc64 uses function descriptors and appends a '.' to the
|
|
|
|
|
* start of every instruction address. Remove it.
|
|
|
|
|
*/
|
|
|
|
|
if (name[0] == '.')
|
|
|
|
|
name++;
|
|
|
|
|
|
2009-05-29 00:55:41 +07:00
|
|
|
|
if (!strcmp(name, "_text") ||
|
|
|
|
|
!strcmp(name, "_etext") ||
|
|
|
|
|
!strcmp(name, "_sinittext") ||
|
|
|
|
|
!strncmp("init_module", name, 11) ||
|
|
|
|
|
!strncmp("cleanup_module", name, 14) ||
|
|
|
|
|
strstr(name, "_text_start") ||
|
|
|
|
|
strstr(name, "_text_end"))
|
2009-04-20 20:00:56 +07:00
|
|
|
|
return 1;
|
|
|
|
|
|
2009-07-01 06:00:46 +07:00
|
|
|
|
for (i = 0; skip_symbols[i]; i++) {
|
|
|
|
|
if (!strcmp(skip_symbols[i], name)) {
|
2011-03-11 23:36:01 +07:00
|
|
|
|
sym->ignore = true;
|
2009-07-01 06:00:46 +07:00
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
}
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2011-01-29 23:01:45 +07:00
|
|
|
|
static void perf_event__process_sample(const union perf_event *event,
|
2011-10-29 21:15:04 +07:00
|
|
|
|
struct perf_evsel *evsel,
|
2011-01-29 23:01:45 +07:00
|
|
|
|
struct perf_sample *sample,
|
|
|
|
|
struct perf_session *session)
|
2009-04-20 20:00:56 +07:00
|
|
|
|
{
|
2011-10-06 05:30:22 +07:00
|
|
|
|
struct symbol *parent = NULL;
|
2011-01-29 23:01:45 +07:00
|
|
|
|
u64 ip = event->ip.ip;
|
perf tools: Consolidate symbol resolving across all tools
Now we have a very high level routine for simple tools to
process IP sample events:
int event__preprocess_sample(const event_t *self,
struct addr_location *al,
symbol_filter_t filter)
It receives the event itself and will insert new threads in the
global threads list and resolve the map and symbol, filling all
this info into the new addr_location struct, so that tools like
annotate and report can further process the event by creating
hist_entries in their specific way (with or without callgraphs,
etc).
It in turn uses the new next layer function:
void thread__find_addr_location(struct thread *self, u8 cpumode,
enum map_type type, u64 addr,
struct addr_location *al,
symbol_filter_t filter)
This one will, given a thread (userspace or the kernel kthread
one), will find the given type (MAP__FUNCTION now, MAP__VARIABLE
too in the near future) at the given cpumode, taking vdsos into
account (userspace hit, but kernel symbol) and will fill all
these details in the addr_location given.
Tools that need a more compact API for plain function
resolution, like 'kmem', can use this other one:
struct symbol *thread__find_function(struct thread *self, u64 addr,
symbol_filter_t filter)
So, to resolve a kernel symbol, that is all the 'kmem' tool
needs, its just a matter of calling:
sym = thread__find_function(kthread, addr, NULL);
The 'filter' parameter is needed because we do lazy
parsing/loading of ELF symtabs or /proc/kallsyms.
With this we remove more code duplication all around, which is
always good, huh? :-)
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: John Kacur <jkacur@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1259346563-12568-12-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-28 01:29:23 +07:00
|
|
|
|
struct addr_location al;
|
2010-04-28 07:17:50 +07:00
|
|
|
|
struct machine *machine;
|
2011-10-06 05:30:22 +07:00
|
|
|
|
int err;
|
2011-01-29 23:01:45 +07:00
|
|
|
|
u8 origin = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK;
|
2009-10-27 04:23:19 +07:00
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
++top.samples;
|
2010-01-21 22:04:43 +07:00
|
|
|
|
|
2009-11-17 06:45:24 +07:00
|
|
|
|
switch (origin) {
|
perf tools: Consolidate symbol resolving across all tools
Now we have a very high level routine for simple tools to
process IP sample events:
int event__preprocess_sample(const event_t *self,
struct addr_location *al,
symbol_filter_t filter)
It receives the event itself and will insert new threads in the
global threads list and resolve the map and symbol, filling all
this info into the new addr_location struct, so that tools like
annotate and report can further process the event by creating
hist_entries in their specific way (with or without callgraphs,
etc).
It in turn uses the new next layer function:
void thread__find_addr_location(struct thread *self, u8 cpumode,
enum map_type type, u64 addr,
struct addr_location *al,
symbol_filter_t filter)
This one will, given a thread (userspace or the kernel kthread
one), will find the given type (MAP__FUNCTION now, MAP__VARIABLE
too in the near future) at the given cpumode, taking vdsos into
account (userspace hit, but kernel symbol) and will fill all
these details in the addr_location given.
Tools that need a more compact API for plain function
resolution, like 'kmem', can use this other one:
struct symbol *thread__find_function(struct thread *self, u64 addr,
symbol_filter_t filter)
So, to resolve a kernel symbol, that is all the 'kmem' tool
needs, its just a matter of calling:
sym = thread__find_function(kthread, addr, NULL);
The 'filter' parameter is needed because we do lazy
parsing/loading of ELF symtabs or /proc/kallsyms.
With this we remove more code duplication all around, which is
always good, huh? :-)
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: John Kacur <jkacur@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1259346563-12568-12-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-28 01:29:23 +07:00
|
|
|
|
case PERF_RECORD_MISC_USER:
|
2011-01-31 23:50:39 +07:00
|
|
|
|
++top.us_samples;
|
|
|
|
|
if (top.hide_user_symbols)
|
2009-11-17 06:45:24 +07:00
|
|
|
|
return;
|
2010-04-28 07:17:50 +07:00
|
|
|
|
machine = perf_session__find_host_machine(session);
|
perf tools: Consolidate symbol resolving across all tools
Now we have a very high level routine for simple tools to
process IP sample events:
int event__preprocess_sample(const event_t *self,
struct addr_location *al,
symbol_filter_t filter)
It receives the event itself and will insert new threads in the
global threads list and resolve the map and symbol, filling all
this info into the new addr_location struct, so that tools like
annotate and report can further process the event by creating
hist_entries in their specific way (with or without callgraphs,
etc).
It in turn uses the new next layer function:
void thread__find_addr_location(struct thread *self, u8 cpumode,
enum map_type type, u64 addr,
struct addr_location *al,
symbol_filter_t filter)
This one will, given a thread (userspace or the kernel kthread
one), will find the given type (MAP__FUNCTION now, MAP__VARIABLE
too in the near future) at the given cpumode, taking vdsos into
account (userspace hit, but kernel symbol) and will fill all
these details in the addr_location given.
Tools that need a more compact API for plain function
resolution, like 'kmem', can use this other one:
struct symbol *thread__find_function(struct thread *self, u64 addr,
symbol_filter_t filter)
So, to resolve a kernel symbol, that is all the 'kmem' tool
needs, its just a matter of calling:
sym = thread__find_function(kthread, addr, NULL);
The 'filter' parameter is needed because we do lazy
parsing/loading of ELF symtabs or /proc/kallsyms.
With this we remove more code duplication all around, which is
always good, huh? :-)
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: John Kacur <jkacur@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1259346563-12568-12-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-28 01:29:23 +07:00
|
|
|
|
break;
|
2009-10-27 04:23:19 +07:00
|
|
|
|
case PERF_RECORD_MISC_KERNEL:
|
2011-01-31 23:50:39 +07:00
|
|
|
|
++top.kernel_samples;
|
|
|
|
|
if (top.hide_kernel_symbols)
|
2009-11-17 06:45:24 +07:00
|
|
|
|
return;
|
2010-04-28 07:17:50 +07:00
|
|
|
|
machine = perf_session__find_host_machine(session);
|
2010-04-19 12:32:50 +07:00
|
|
|
|
break;
|
|
|
|
|
case PERF_RECORD_MISC_GUEST_KERNEL:
|
2011-01-31 23:50:39 +07:00
|
|
|
|
++top.guest_kernel_samples;
|
2011-01-29 23:01:45 +07:00
|
|
|
|
machine = perf_session__find_machine(session, event->ip.pid);
|
2009-10-27 04:23:19 +07:00
|
|
|
|
break;
|
2010-04-19 12:32:50 +07:00
|
|
|
|
case PERF_RECORD_MISC_GUEST_USER:
|
2011-01-31 23:50:39 +07:00
|
|
|
|
++top.guest_us_samples;
|
2010-04-19 12:32:50 +07:00
|
|
|
|
/*
|
|
|
|
|
* TODO: we don't process guest user from host side
|
|
|
|
|
* except simple counting.
|
|
|
|
|
*/
|
|
|
|
|
return;
|
2009-10-27 04:23:19 +07:00
|
|
|
|
default:
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
2010-04-28 07:17:50 +07:00
|
|
|
|
if (!machine && perf_guest) {
|
2010-04-19 12:32:50 +07:00
|
|
|
|
pr_err("Can't find guest [%d]'s kernel information\n",
|
2011-01-29 23:01:45 +07:00
|
|
|
|
event->ip.pid);
|
2010-04-19 12:32:50 +07:00
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
2011-01-29 23:01:45 +07:00
|
|
|
|
if (event->header.misc & PERF_RECORD_MISC_EXACT_IP)
|
2011-01-31 23:50:39 +07:00
|
|
|
|
top.exact_samples++;
|
2010-03-04 20:19:36 +07:00
|
|
|
|
|
2011-01-29 23:01:45 +07:00
|
|
|
|
if (perf_event__preprocess_sample(event, session, &al, sample,
|
|
|
|
|
symbol_filter) < 0 ||
|
2010-01-28 06:05:49 +07:00
|
|
|
|
al.filtered)
|
perf tools: Consolidate symbol resolving across all tools
Now we have a very high level routine for simple tools to
process IP sample events:
int event__preprocess_sample(const event_t *self,
struct addr_location *al,
symbol_filter_t filter)
It receives the event itself and will insert new threads in the
global threads list and resolve the map and symbol, filling all
this info into the new addr_location struct, so that tools like
annotate and report can further process the event by creating
hist_entries in their specific way (with or without callgraphs,
etc).
It in turn uses the new next layer function:
void thread__find_addr_location(struct thread *self, u8 cpumode,
enum map_type type, u64 addr,
struct addr_location *al,
symbol_filter_t filter)
This one will, given a thread (userspace or the kernel kthread
one), will find the given type (MAP__FUNCTION now, MAP__VARIABLE
too in the near future) at the given cpumode, taking vdsos into
account (userspace hit, but kernel symbol) and will fill all
these details in the addr_location given.
Tools that need a more compact API for plain function
resolution, like 'kmem', can use this other one:
struct symbol *thread__find_function(struct thread *self, u64 addr,
symbol_filter_t filter)
So, to resolve a kernel symbol, that is all the 'kmem' tool
needs, its just a matter of calling:
sym = thread__find_function(kthread, addr, NULL);
The 'filter' parameter is needed because we do lazy
parsing/loading of ELF symtabs or /proc/kallsyms.
With this we remove more code duplication all around, which is
always good, huh? :-)
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: John Kacur <jkacur@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1259346563-12568-12-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-11-28 01:29:23 +07:00
|
|
|
|
return;
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2011-05-27 21:53:28 +07:00
|
|
|
|
if (!kptr_restrict_warned &&
|
|
|
|
|
symbol_conf.kptr_restrict &&
|
|
|
|
|
al.cpumode == PERF_RECORD_MISC_KERNEL) {
|
|
|
|
|
ui__warning(
|
|
|
|
|
"Kernel address maps (/proc/{kallsyms,modules}) are restricted.\n\n"
|
|
|
|
|
"Check /proc/sys/kernel/kptr_restrict.\n\n"
|
|
|
|
|
"Kernel%s samples will not be resolved.\n",
|
|
|
|
|
!RB_EMPTY_ROOT(&al.map->dso->symbols[MAP__FUNCTION]) ?
|
|
|
|
|
" modules" : "");
|
|
|
|
|
if (use_browser <= 0)
|
|
|
|
|
sleep(5);
|
|
|
|
|
kptr_restrict_warned = true;
|
|
|
|
|
}
|
|
|
|
|
|
2010-01-28 06:05:49 +07:00
|
|
|
|
if (al.sym == NULL) {
|
2011-05-27 23:42:16 +07:00
|
|
|
|
const char *msg = "Kernel samples will not be resolved.\n";
|
2010-01-28 06:05:49 +07:00
|
|
|
|
/*
|
|
|
|
|
* As we do lazy loading of symtabs we only will know if the
|
|
|
|
|
* specified vmlinux file is invalid when we actually have a
|
|
|
|
|
* hit in kernel space and then try to load it. So if we get
|
|
|
|
|
* here and there are _no_ symbols in the DSO backing the
|
|
|
|
|
* kernel map, bail out.
|
|
|
|
|
*
|
|
|
|
|
* We may never get here, for instance, if we use -K/
|
|
|
|
|
* --hide-kernel-symbols, even if the user specifies an
|
|
|
|
|
* invalid --vmlinux ;-)
|
|
|
|
|
*/
|
2011-05-27 23:42:16 +07:00
|
|
|
|
if (!kptr_restrict_warned && !vmlinux_warned &&
|
|
|
|
|
al.map == machine->vmlinux_maps[MAP__FUNCTION] &&
|
2010-01-28 06:05:49 +07:00
|
|
|
|
RB_EMPTY_ROOT(&al.map->dso->symbols[MAP__FUNCTION])) {
|
2011-05-27 23:42:16 +07:00
|
|
|
|
if (symbol_conf.vmlinux_name) {
|
|
|
|
|
ui__warning("The %s file can't be used.\n%s",
|
|
|
|
|
symbol_conf.vmlinux_name, msg);
|
|
|
|
|
} else {
|
|
|
|
|
ui__warning("A vmlinux file was not found.\n%s",
|
|
|
|
|
msg);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (use_browser <= 0)
|
|
|
|
|
sleep(5);
|
|
|
|
|
vmlinux_warned = true;
|
2010-01-28 06:05:49 +07:00
|
|
|
|
}
|
2010-02-04 01:52:08 +07:00
|
|
|
|
}
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
if (al.sym == NULL || !al.sym->ignore) {
|
|
|
|
|
struct hist_entry *he;
|
2011-01-13 07:39:13 +07:00
|
|
|
|
|
2011-10-06 05:30:22 +07:00
|
|
|
|
if ((sort__has_parent || symbol_conf.use_callchain) &&
|
|
|
|
|
sample->callchain) {
|
|
|
|
|
err = perf_session__resolve_callchain(session, al.thread,
|
|
|
|
|
sample->callchain, &parent);
|
|
|
|
|
if (err)
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
he = perf_session__add_hist_entry(session, &al, sample, evsel);
|
|
|
|
|
if (he == NULL) {
|
|
|
|
|
pr_err("Problem incrementing symbol period, skipping event\n");
|
|
|
|
|
return;
|
2011-03-01 20:43:03 +07:00
|
|
|
|
}
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
|
2011-10-06 05:30:22 +07:00
|
|
|
|
if (symbol_conf.use_callchain) {
|
|
|
|
|
err = callchain_append(he->callchain, &session->callchain_cursor,
|
|
|
|
|
sample->period);
|
|
|
|
|
if (err)
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (sort_has_symbols)
|
|
|
|
|
record_precise_ip(he, evsel->idx, ip);
|
2009-10-27 04:23:19 +07:00
|
|
|
|
}
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
|
|
|
|
|
return;
|
2009-04-20 20:00:56 +07:00
|
|
|
|
}
|
|
|
|
|
|
perf evlist: Fix per thread mmap setup
The PERF_EVENT_IOC_SET_OUTPUT ioctl was returning -EINVAL when using
--pid when monitoring multithreaded apps, as we can only share a ring
buffer for events on the same thread if not doing per cpu.
Fix it by using per thread ring buffers.
Tested with:
[root@felicio ~]# tuna -t 26131 -CP | nl
1 thread ctxt_switches
2 pid SCHED_ rtpri affinity voluntary nonvoluntary cmd
3 26131 OTHER 0 0,1 10814276 2397830 chromium-browse
4 642 OTHER 0 0,1 14688 0 chromium-browse
5 26148 OTHER 0 0,1 713602 115479 chromium-browse
6 26149 OTHER 0 0,1 801958 2262 chromium-browse
7 26150 OTHER 0 0,1 1271128 248 chromium-browse
8 26151 OTHER 0 0,1 3 0 chromium-browse
9 27049 OTHER 0 0,1 36796 9 chromium-browse
10 618 OTHER 0 0,1 14711 0 chromium-browse
11 661 OTHER 0 0,1 14593 0 chromium-browse
12 29048 OTHER 0 0,1 28125 0 chromium-browse
13 26143 OTHER 0 0,1 2202789 781 chromium-browse
[root@felicio ~]#
So 11 threads under pid 26131, then:
[root@felicio ~]# perf record -F 50000 --pid 26131
[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
1 7fa4a2538000-7fa4a25b9000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
2 7fa4a25b9000-7fa4a263a000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
3 7fa4a263a000-7fa4a26bb000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
4 7fa4a26bb000-7fa4a273c000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
5 7fa4a273c000-7fa4a27bd000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
6 7fa4a27bd000-7fa4a283e000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
7 7fa4a283e000-7fa4a28bf000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
8 7fa4a28bf000-7fa4a2940000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
9 7fa4a2940000-7fa4a29c1000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
10 7fa4a29c1000-7fa4a2a42000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
11 7fa4a2a42000-7fa4a2ac3000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@felicio ~]#
11 mmaps, one per thread since we didn't specify any CPU list, so we need one
mmap per thread and:
[root@felicio ~]# perf record -F 50000 --pid 26131
^M
^C[ perf record: Woken up 79 times to write data ]
[ perf record: Captured and wrote 20.614 MB perf.data (~900639 samples) ]
[root@felicio ~]# perf report -D | grep PERF_RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort -n | uniq -c | sort -nr | nl
1 371310 26131
2 96516 26148
3 95694 26149
4 95203 26150
5 7291 26143
6 87 27049
7 76 661
8 60 29048
9 47 618
10 43 642
[root@felicio ~]#
Ok, one of the threads, 26151 was quiescent, so no samples there, but all the
others are there.
Then, if I specify one CPU:
[root@felicio ~]# perf record -F 50000 --pid 26131 --cpu 1
^C[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.680 MB perf.data (~29730 samples) ]
[root@felicio ~]# perf report -D | grep PERF_RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort -n | uniq -c | sort -nr | nl
1 8444 26131
2 2584 26149
3 2518 26148
4 2324 26150
5 123 26143
6 9 661
7 9 29048
[root@felicio ~]#
This machine has two cores, so fewer threads appeared on the radar, and:
[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
1 7f484b922000-7f484b9a3000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@felicio ~]#
Just one mmap, as now we can use just one per-cpu buffer instead of the
per-thread needed in the previous case.
For global profiling:
[root@felicio ~]# perf record -F 50000 -a
^C[ perf record: Woken up 26 times to write data ]
[ perf record: Captured and wrote 7.128 MB perf.data (~311412 samples) ]
[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
1 7fb49b435000-7fb49b4b6000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
2 7fb49b4b6000-7fb49b537000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@felicio ~]#
It uses per-cpu buffers.
For just one thread:
[root@felicio ~]# perf record -F 50000 --tid 26148
^C[ perf record: Woken up 2 times to write data ]
[ perf record: Captured and wrote 0.330 MB perf.data (~14426 samples) ]
[root@felicio ~]# perf report -D | grep PERF_RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort -n | uniq -c | sort -nr | nl
1 9969 26148
[root@felicio ~]#
[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
1 7f286a51b000-7f286a59c000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@felicio ~]#
Tested-by: David Ahern <dsahern@gmail.com>
Tested-by: Lin Ming <ming.m.lin@intel.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Link: http://lkml.kernel.org/r/20110426204401.GB1746@ghostprotocols.net
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-05-15 19:39:00 +07:00
|
|
|
|
static void perf_session__mmap_read_idx(struct perf_session *self, int idx)
|
2009-04-20 20:00:56 +07:00
|
|
|
|
{
|
2011-01-29 22:02:00 +07:00
|
|
|
|
struct perf_sample sample;
|
2011-10-29 21:15:04 +07:00
|
|
|
|
struct perf_evsel *evsel;
|
2011-01-29 23:01:45 +07:00
|
|
|
|
union perf_event *event;
|
2011-05-22 07:17:22 +07:00
|
|
|
|
int ret;
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
perf evlist: Fix per thread mmap setup
The PERF_EVENT_IOC_SET_OUTPUT ioctl was returning -EINVAL when using
--pid when monitoring multithreaded apps, as we can only share a ring
buffer for events on the same thread if not doing per cpu.
Fix it by using per thread ring buffers.
Tested with:
[root@felicio ~]# tuna -t 26131 -CP | nl
1 thread ctxt_switches
2 pid SCHED_ rtpri affinity voluntary nonvoluntary cmd
3 26131 OTHER 0 0,1 10814276 2397830 chromium-browse
4 642 OTHER 0 0,1 14688 0 chromium-browse
5 26148 OTHER 0 0,1 713602 115479 chromium-browse
6 26149 OTHER 0 0,1 801958 2262 chromium-browse
7 26150 OTHER 0 0,1 1271128 248 chromium-browse
8 26151 OTHER 0 0,1 3 0 chromium-browse
9 27049 OTHER 0 0,1 36796 9 chromium-browse
10 618 OTHER 0 0,1 14711 0 chromium-browse
11 661 OTHER 0 0,1 14593 0 chromium-browse
12 29048 OTHER 0 0,1 28125 0 chromium-browse
13 26143 OTHER 0 0,1 2202789 781 chromium-browse
[root@felicio ~]#
So 11 threads under pid 26131, then:
[root@felicio ~]# perf record -F 50000 --pid 26131
[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
1 7fa4a2538000-7fa4a25b9000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
2 7fa4a25b9000-7fa4a263a000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
3 7fa4a263a000-7fa4a26bb000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
4 7fa4a26bb000-7fa4a273c000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
5 7fa4a273c000-7fa4a27bd000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
6 7fa4a27bd000-7fa4a283e000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
7 7fa4a283e000-7fa4a28bf000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
8 7fa4a28bf000-7fa4a2940000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
9 7fa4a2940000-7fa4a29c1000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
10 7fa4a29c1000-7fa4a2a42000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
11 7fa4a2a42000-7fa4a2ac3000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@felicio ~]#
11 mmaps, one per thread since we didn't specify any CPU list, so we need one
mmap per thread and:
[root@felicio ~]# perf record -F 50000 --pid 26131
^M
^C[ perf record: Woken up 79 times to write data ]
[ perf record: Captured and wrote 20.614 MB perf.data (~900639 samples) ]
[root@felicio ~]# perf report -D | grep PERF_RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort -n | uniq -c | sort -nr | nl
1 371310 26131
2 96516 26148
3 95694 26149
4 95203 26150
5 7291 26143
6 87 27049
7 76 661
8 60 29048
9 47 618
10 43 642
[root@felicio ~]#
Ok, one of the threads, 26151 was quiescent, so no samples there, but all the
others are there.
Then, if I specify one CPU:
[root@felicio ~]# perf record -F 50000 --pid 26131 --cpu 1
^C[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.680 MB perf.data (~29730 samples) ]
[root@felicio ~]# perf report -D | grep PERF_RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort -n | uniq -c | sort -nr | nl
1 8444 26131
2 2584 26149
3 2518 26148
4 2324 26150
5 123 26143
6 9 661
7 9 29048
[root@felicio ~]#
This machine has two cores, so fewer threads appeared on the radar, and:
[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
1 7f484b922000-7f484b9a3000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@felicio ~]#
Just one mmap, as now we can use just one per-cpu buffer instead of the
per-thread needed in the previous case.
For global profiling:
[root@felicio ~]# perf record -F 50000 -a
^C[ perf record: Woken up 26 times to write data ]
[ perf record: Captured and wrote 7.128 MB perf.data (~311412 samples) ]
[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
1 7fb49b435000-7fb49b4b6000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
2 7fb49b4b6000-7fb49b537000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@felicio ~]#
It uses per-cpu buffers.
For just one thread:
[root@felicio ~]# perf record -F 50000 --tid 26148
^C[ perf record: Woken up 2 times to write data ]
[ perf record: Captured and wrote 0.330 MB perf.data (~14426 samples) ]
[root@felicio ~]# perf report -D | grep PERF_RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort -n | uniq -c | sort -nr | nl
1 9969 26148
[root@felicio ~]#
[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
1 7f286a51b000-7f286a59c000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@felicio ~]#
Tested-by: David Ahern <dsahern@gmail.com>
Tested-by: Lin Ming <ming.m.lin@intel.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Link: http://lkml.kernel.org/r/20110426204401.GB1746@ghostprotocols.net
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-05-15 19:39:00 +07:00
|
|
|
|
while ((event = perf_evlist__mmap_read(top.evlist, idx)) != NULL) {
|
2011-05-22 07:17:22 +07:00
|
|
|
|
ret = perf_session__parse_sample(self, event, &sample);
|
|
|
|
|
if (ret) {
|
|
|
|
|
pr_err("Can't parse sample, err = %d\n", ret);
|
|
|
|
|
continue;
|
|
|
|
|
}
|
2011-01-15 19:40:59 +07:00
|
|
|
|
|
2011-10-29 21:15:04 +07:00
|
|
|
|
evsel = perf_evlist__id2evsel(self->evlist, sample.id);
|
|
|
|
|
assert(evsel != NULL);
|
|
|
|
|
|
2009-10-27 04:23:19 +07:00
|
|
|
|
if (event->header.type == PERF_RECORD_SAMPLE)
|
2011-10-29 21:15:04 +07:00
|
|
|
|
perf_event__process_sample(event, evsel, &sample, self);
|
|
|
|
|
else if (event->header.type < PERF_RECORD_MAX) {
|
|
|
|
|
hists__inc_nr_events(&evsel->hists, event->header.type);
|
2011-01-29 23:01:45 +07:00
|
|
|
|
perf_event__process(event, &sample, self);
|
2011-10-29 21:15:04 +07:00
|
|
|
|
} else
|
|
|
|
|
++self->hists.stats.nr_unknown_events;
|
2009-04-20 20:00:56 +07:00
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2009-12-14 04:50:24 +07:00
|
|
|
|
static void perf_session__mmap_read(struct perf_session *self)
|
2009-06-07 04:10:43 +07:00
|
|
|
|
{
|
2011-01-13 07:39:13 +07:00
|
|
|
|
int i;
|
|
|
|
|
|
perf evlist: Fix per thread mmap setup
The PERF_EVENT_IOC_SET_OUTPUT ioctl was returning -EINVAL when using
--pid when monitoring multithreaded apps, as we can only share a ring
buffer for events on the same thread if not doing per cpu.
Fix it by using per thread ring buffers.
Tested with:
[root@felicio ~]# tuna -t 26131 -CP | nl
1 thread ctxt_switches
2 pid SCHED_ rtpri affinity voluntary nonvoluntary cmd
3 26131 OTHER 0 0,1 10814276 2397830 chromium-browse
4 642 OTHER 0 0,1 14688 0 chromium-browse
5 26148 OTHER 0 0,1 713602 115479 chromium-browse
6 26149 OTHER 0 0,1 801958 2262 chromium-browse
7 26150 OTHER 0 0,1 1271128 248 chromium-browse
8 26151 OTHER 0 0,1 3 0 chromium-browse
9 27049 OTHER 0 0,1 36796 9 chromium-browse
10 618 OTHER 0 0,1 14711 0 chromium-browse
11 661 OTHER 0 0,1 14593 0 chromium-browse
12 29048 OTHER 0 0,1 28125 0 chromium-browse
13 26143 OTHER 0 0,1 2202789 781 chromium-browse
[root@felicio ~]#
So 11 threads under pid 26131, then:
[root@felicio ~]# perf record -F 50000 --pid 26131
[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
1 7fa4a2538000-7fa4a25b9000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
2 7fa4a25b9000-7fa4a263a000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
3 7fa4a263a000-7fa4a26bb000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
4 7fa4a26bb000-7fa4a273c000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
5 7fa4a273c000-7fa4a27bd000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
6 7fa4a27bd000-7fa4a283e000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
7 7fa4a283e000-7fa4a28bf000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
8 7fa4a28bf000-7fa4a2940000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
9 7fa4a2940000-7fa4a29c1000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
10 7fa4a29c1000-7fa4a2a42000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
11 7fa4a2a42000-7fa4a2ac3000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@felicio ~]#
11 mmaps, one per thread since we didn't specify any CPU list, so we need one
mmap per thread and:
[root@felicio ~]# perf record -F 50000 --pid 26131
^M
^C[ perf record: Woken up 79 times to write data ]
[ perf record: Captured and wrote 20.614 MB perf.data (~900639 samples) ]
[root@felicio ~]# perf report -D | grep PERF_RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort -n | uniq -c | sort -nr | nl
1 371310 26131
2 96516 26148
3 95694 26149
4 95203 26150
5 7291 26143
6 87 27049
7 76 661
8 60 29048
9 47 618
10 43 642
[root@felicio ~]#
Ok, one of the threads, 26151 was quiescent, so no samples there, but all the
others are there.
Then, if I specify one CPU:
[root@felicio ~]# perf record -F 50000 --pid 26131 --cpu 1
^C[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.680 MB perf.data (~29730 samples) ]
[root@felicio ~]# perf report -D | grep PERF_RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort -n | uniq -c | sort -nr | nl
1 8444 26131
2 2584 26149
3 2518 26148
4 2324 26150
5 123 26143
6 9 661
7 9 29048
[root@felicio ~]#
This machine has two cores, so fewer threads appeared on the radar, and:
[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
1 7f484b922000-7f484b9a3000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@felicio ~]#
Just one mmap, as now we can use just one per-cpu buffer instead of the
per-thread needed in the previous case.
For global profiling:
[root@felicio ~]# perf record -F 50000 -a
^C[ perf record: Woken up 26 times to write data ]
[ perf record: Captured and wrote 7.128 MB perf.data (~311412 samples) ]
[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
1 7fb49b435000-7fb49b4b6000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
2 7fb49b4b6000-7fb49b537000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@felicio ~]#
It uses per-cpu buffers.
For just one thread:
[root@felicio ~]# perf record -F 50000 --tid 26148
^C[ perf record: Woken up 2 times to write data ]
[ perf record: Captured and wrote 0.330 MB perf.data (~14426 samples) ]
[root@felicio ~]# perf report -D | grep PERF_RECORD_SAMPLE | cut -d/ -f2 | cut -d: -f1 | sort -n | uniq -c | sort -nr | nl
1 9969 26148
[root@felicio ~]#
[root@felicio ~]# grep perf_event /proc/`pidof perf`/maps | nl
1 7f286a51b000-7f286a59c000 rwxs 00000000 00:09 4064 anon_inode:[perf_event]
[root@felicio ~]#
Tested-by: David Ahern <dsahern@gmail.com>
Tested-by: Lin Ming <ming.m.lin@intel.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
Link: http://lkml.kernel.org/r/20110426204401.GB1746@ghostprotocols.net
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-05-15 19:39:00 +07:00
|
|
|
|
for (i = 0; i < top.evlist->nr_mmaps; i++)
|
|
|
|
|
perf_session__mmap_read_idx(self, i);
|
2009-06-07 04:10:43 +07:00
|
|
|
|
}
|
|
|
|
|
|
2011-01-12 19:52:47 +07:00
|
|
|
|
static void start_counters(struct perf_evlist *evlist)
|
|
|
|
|
{
|
2011-10-25 19:42:19 +07:00
|
|
|
|
struct perf_evsel *counter, *first;
|
|
|
|
|
|
|
|
|
|
first = list_entry(evlist->entries.next, struct perf_evsel, node);
|
2009-10-12 12:56:03 +07:00
|
|
|
|
|
2011-01-12 19:52:47 +07:00
|
|
|
|
list_for_each_entry(counter, &evlist->entries, node) {
|
|
|
|
|
struct perf_event_attr *attr = &counter->attr;
|
2011-10-25 19:42:19 +07:00
|
|
|
|
struct xyarray *group_fd = NULL;
|
|
|
|
|
|
|
|
|
|
if (group && counter != first)
|
|
|
|
|
group_fd = first->fd;
|
2009-06-07 22:31:52 +07:00
|
|
|
|
|
2011-01-12 19:52:47 +07:00
|
|
|
|
attr->sample_type = PERF_SAMPLE_IP | PERF_SAMPLE_TID;
|
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
if (top.freq) {
|
2011-01-12 19:52:47 +07:00
|
|
|
|
attr->sample_type |= PERF_SAMPLE_PERIOD;
|
|
|
|
|
attr->freq = 1;
|
2011-01-31 23:50:39 +07:00
|
|
|
|
attr->sample_freq = top.freq;
|
2011-01-12 19:52:47 +07:00
|
|
|
|
}
|
2010-03-18 21:36:05 +07:00
|
|
|
|
|
2011-01-13 07:39:13 +07:00
|
|
|
|
if (evlist->nr_entries > 1) {
|
|
|
|
|
attr->sample_type |= PERF_SAMPLE_ID;
|
|
|
|
|
attr->read_format |= PERF_FORMAT_ID;
|
|
|
|
|
}
|
|
|
|
|
|
2011-10-06 05:30:22 +07:00
|
|
|
|
if (symbol_conf.use_callchain)
|
|
|
|
|
attr->sample_type |= PERF_SAMPLE_CALLCHAIN;
|
|
|
|
|
|
2011-01-12 19:52:47 +07:00
|
|
|
|
attr->mmap = 1;
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
attr->comm = 1;
|
2011-04-14 21:20:14 +07:00
|
|
|
|
attr->inherit = inherit;
|
2011-10-29 21:15:04 +07:00
|
|
|
|
retry_sample_id:
|
|
|
|
|
attr->sample_id_all = sample_id_all_avail ? 1 : 0;
|
2011-01-12 19:52:47 +07:00
|
|
|
|
try_again:
|
2011-01-31 23:50:39 +07:00
|
|
|
|
if (perf_evsel__open(counter, top.evlist->cpus,
|
2011-10-25 19:42:19 +07:00
|
|
|
|
top.evlist->threads, group,
|
|
|
|
|
group_fd) < 0) {
|
2010-03-18 21:36:05 +07:00
|
|
|
|
int err = errno;
|
|
|
|
|
|
2011-03-28 19:50:11 +07:00
|
|
|
|
if (err == EPERM || err == EACCES) {
|
2011-10-26 17:02:55 +07:00
|
|
|
|
ui__error_paranoid();
|
2011-03-28 19:50:11 +07:00
|
|
|
|
goto out_err;
|
2011-10-29 21:15:04 +07:00
|
|
|
|
} else if (err == EINVAL && sample_id_all_avail) {
|
|
|
|
|
/*
|
|
|
|
|
* Old kernel, no attr->sample_id_type_all field
|
|
|
|
|
*/
|
|
|
|
|
sample_id_all_avail = false;
|
|
|
|
|
goto retry_sample_id;
|
2011-03-28 19:50:11 +07:00
|
|
|
|
}
|
2010-03-18 21:36:05 +07:00
|
|
|
|
/*
|
|
|
|
|
* If it's cycles then fall back to hrtimer
|
|
|
|
|
* based cpu-clock-tick sw counter, which
|
|
|
|
|
* is always available even if no PMU support:
|
|
|
|
|
*/
|
2011-01-12 19:52:47 +07:00
|
|
|
|
if (attr->type == PERF_TYPE_HARDWARE &&
|
|
|
|
|
attr->config == PERF_COUNT_HW_CPU_CYCLES) {
|
2010-03-18 21:36:05 +07:00
|
|
|
|
if (verbose)
|
2011-03-28 19:50:11 +07:00
|
|
|
|
ui__warning("Cycles event not supported,\n"
|
|
|
|
|
"trying to fall back to cpu-clock-ticks\n");
|
2010-03-18 21:36:05 +07:00
|
|
|
|
|
|
|
|
|
attr->type = PERF_TYPE_SOFTWARE;
|
|
|
|
|
attr->config = PERF_COUNT_SW_CPU_CLOCK;
|
|
|
|
|
goto try_again;
|
|
|
|
|
}
|
2011-03-28 19:50:11 +07:00
|
|
|
|
|
2011-03-26 02:11:11 +07:00
|
|
|
|
if (err == ENOENT) {
|
|
|
|
|
ui__warning("The %s event is not supported.\n",
|
|
|
|
|
event_name(counter));
|
|
|
|
|
goto out_err;
|
|
|
|
|
}
|
|
|
|
|
|
2011-03-28 19:50:11 +07:00
|
|
|
|
ui__warning("The sys_perf_event_open() syscall "
|
|
|
|
|
"returned with %d (%s). /bin/dmesg "
|
|
|
|
|
"may provide additional information.\n"
|
|
|
|
|
"No CONFIG_PERF_EVENTS=y kernel support "
|
|
|
|
|
"configured?\n", err, strerror(err));
|
|
|
|
|
goto out_err;
|
2010-03-18 21:36:05 +07:00
|
|
|
|
}
|
2009-06-07 22:31:52 +07:00
|
|
|
|
}
|
2011-01-13 07:39:13 +07:00
|
|
|
|
|
2011-03-28 19:50:11 +07:00
|
|
|
|
if (perf_evlist__mmap(evlist, mmap_pages, false) < 0) {
|
|
|
|
|
ui__warning("Failed to mmap with %d (%s)\n",
|
|
|
|
|
errno, strerror(errno));
|
|
|
|
|
goto out_err;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return;
|
|
|
|
|
|
|
|
|
|
out_err:
|
|
|
|
|
exit_browser(0);
|
|
|
|
|
exit(0);
|
2009-06-07 22:31:52 +07:00
|
|
|
|
}
|
|
|
|
|
|
2011-10-06 05:30:22 +07:00
|
|
|
|
static int setup_sample_type(void)
|
|
|
|
|
{
|
|
|
|
|
if (!sort_has_symbols) {
|
|
|
|
|
if (symbol_conf.use_callchain) {
|
|
|
|
|
ui__warning("Selected -g but \"sym\" not present in --sort/-s.");
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
} else if (!dont_use_callchains && callchain_param.mode != CHAIN_NONE) {
|
|
|
|
|
if (callchain_register_param(&callchain_param) < 0) {
|
|
|
|
|
ui__warning("Can't register callchain params.\n");
|
|
|
|
|
return -EINVAL;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
2009-06-07 22:31:52 +07:00
|
|
|
|
static int __cmd_top(void)
|
|
|
|
|
{
|
|
|
|
|
pthread_t thread;
|
2011-10-06 05:30:22 +07:00
|
|
|
|
int ret;
|
2009-12-14 04:50:24 +07:00
|
|
|
|
/*
|
2009-12-14 04:50:28 +07:00
|
|
|
|
* FIXME: perf_session__new should allow passing a O_MMAP, so that all this
|
|
|
|
|
* mmap reading, etc is encapsulated in it. Use O_WRONLY for now.
|
2009-12-14 04:50:24 +07:00
|
|
|
|
*/
|
2011-09-02 00:27:58 +07:00
|
|
|
|
top.session = perf_session__new(NULL, O_WRONLY, false, false, NULL);
|
|
|
|
|
if (top.session == NULL)
|
2009-12-14 04:50:28 +07:00
|
|
|
|
return -ENOMEM;
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2011-10-06 05:30:22 +07:00
|
|
|
|
ret = setup_sample_type();
|
|
|
|
|
if (ret)
|
|
|
|
|
goto out_delete;
|
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
if (top.target_tid != -1)
|
2011-02-11 20:45:54 +07:00
|
|
|
|
perf_event__synthesize_thread_map(top.evlist->threads,
|
2011-09-02 00:27:58 +07:00
|
|
|
|
perf_event__process, top.session);
|
2009-10-27 04:23:19 +07:00
|
|
|
|
else
|
2011-09-02 00:27:58 +07:00
|
|
|
|
perf_event__synthesize_threads(perf_event__process, top.session);
|
2009-10-27 04:23:19 +07:00
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
start_counters(top.evlist);
|
2011-09-02 00:27:58 +07:00
|
|
|
|
top.session->evlist = top.evlist;
|
|
|
|
|
perf_session__update_sample_type(top.session);
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2009-06-07 04:10:43 +07:00
|
|
|
|
/* Wait for a minimal set of events before starting the snapshot */
|
2011-01-31 23:50:39 +07:00
|
|
|
|
poll(top.evlist->pollfd, top.evlist->nr_fds, 100);
|
2009-06-07 04:10:43 +07:00
|
|
|
|
|
2011-09-02 00:27:58 +07:00
|
|
|
|
perf_session__mmap_read(top.session);
|
2009-06-07 04:10:43 +07:00
|
|
|
|
|
2011-02-01 03:19:33 +07:00
|
|
|
|
if (pthread_create(&thread, NULL, (use_browser > 0 ? display_thread_tui :
|
2011-09-02 00:27:58 +07:00
|
|
|
|
display_thread), NULL)) {
|
2009-04-20 20:00:56 +07:00
|
|
|
|
printf("Could not create display thread.\n");
|
|
|
|
|
exit(-1);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (realtime_prio) {
|
|
|
|
|
struct sched_param param;
|
|
|
|
|
|
|
|
|
|
param.sched_priority = realtime_prio;
|
|
|
|
|
if (sched_setscheduler(0, SCHED_FIFO, ¶m)) {
|
|
|
|
|
printf("Could not set realtime priority.\n");
|
|
|
|
|
exit(-1);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
while (1) {
|
2011-01-31 23:50:39 +07:00
|
|
|
|
u64 hits = top.samples;
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2011-09-02 00:27:58 +07:00
|
|
|
|
perf_session__mmap_read(top.session);
|
2009-04-20 20:00:56 +07:00
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
if (hits == top.samples)
|
|
|
|
|
ret = poll(top.evlist->pollfd, top.evlist->nr_fds, 100);
|
2009-04-20 20:00:56 +07:00
|
|
|
|
}
|
|
|
|
|
|
2011-10-06 05:30:22 +07:00
|
|
|
|
out_delete:
|
|
|
|
|
perf_session__delete(top.session);
|
|
|
|
|
top.session = NULL;
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
static int
|
|
|
|
|
parse_callchain_opt(const struct option *opt __used, const char *arg,
|
|
|
|
|
int unset)
|
|
|
|
|
{
|
|
|
|
|
char *tok, *tok2;
|
|
|
|
|
char *endptr;
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
|
* --no-call-graph
|
|
|
|
|
*/
|
|
|
|
|
if (unset) {
|
|
|
|
|
dont_use_callchains = true;
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
symbol_conf.use_callchain = true;
|
|
|
|
|
|
|
|
|
|
if (!arg)
|
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
|
|
tok = strtok((char *)arg, ",");
|
|
|
|
|
if (!tok)
|
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
|
|
/* get the output mode */
|
|
|
|
|
if (!strncmp(tok, "graph", strlen(arg)))
|
|
|
|
|
callchain_param.mode = CHAIN_GRAPH_ABS;
|
|
|
|
|
|
|
|
|
|
else if (!strncmp(tok, "flat", strlen(arg)))
|
|
|
|
|
callchain_param.mode = CHAIN_FLAT;
|
|
|
|
|
|
|
|
|
|
else if (!strncmp(tok, "fractal", strlen(arg)))
|
|
|
|
|
callchain_param.mode = CHAIN_GRAPH_REL;
|
|
|
|
|
|
|
|
|
|
else if (!strncmp(tok, "none", strlen(arg))) {
|
|
|
|
|
callchain_param.mode = CHAIN_NONE;
|
|
|
|
|
symbol_conf.use_callchain = false;
|
|
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
else
|
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
|
|
/* get the min percentage */
|
|
|
|
|
tok = strtok(NULL, ",");
|
|
|
|
|
if (!tok)
|
|
|
|
|
goto setup;
|
|
|
|
|
|
|
|
|
|
callchain_param.min_percent = strtod(tok, &endptr);
|
|
|
|
|
if (tok == endptr)
|
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
|
|
/* get the print limit */
|
|
|
|
|
tok2 = strtok(NULL, ",");
|
|
|
|
|
if (!tok2)
|
|
|
|
|
goto setup;
|
|
|
|
|
|
|
|
|
|
if (tok2[0] != 'c') {
|
|
|
|
|
callchain_param.print_limit = strtod(tok2, &endptr);
|
|
|
|
|
tok2 = strtok(NULL, ",");
|
|
|
|
|
if (!tok2)
|
|
|
|
|
goto setup;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* get the call chain order */
|
|
|
|
|
if (!strcmp(tok2, "caller"))
|
|
|
|
|
callchain_param.order = ORDER_CALLER;
|
|
|
|
|
else if (!strcmp(tok2, "callee"))
|
|
|
|
|
callchain_param.order = ORDER_CALLEE;
|
|
|
|
|
else
|
|
|
|
|
return -1;
|
|
|
|
|
setup:
|
|
|
|
|
if (callchain_register_param(&callchain_param) < 0) {
|
|
|
|
|
fprintf(stderr, "Can't register callchain params\n");
|
|
|
|
|
return -1;
|
|
|
|
|
}
|
2009-04-20 20:00:56 +07:00
|
|
|
|
return 0;
|
|
|
|
|
}
|
2009-05-26 14:17:18 +07:00
|
|
|
|
|
|
|
|
|
static const char * const top_usage[] = {
|
|
|
|
|
"perf top [<options>]",
|
|
|
|
|
NULL
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
static const struct option options[] = {
|
2011-01-31 23:50:39 +07:00
|
|
|
|
OPT_CALLBACK('e', "event", &top.evlist, "event",
|
2009-06-06 17:24:17 +07:00
|
|
|
|
"event selector. use 'perf list' to list available events",
|
2011-07-14 16:25:32 +07:00
|
|
|
|
parse_events_option),
|
2009-05-26 14:17:18 +07:00
|
|
|
|
OPT_INTEGER('c', "count", &default_interval,
|
|
|
|
|
"event period to sample"),
|
2011-01-31 23:50:39 +07:00
|
|
|
|
OPT_INTEGER('p', "pid", &top.target_pid,
|
2010-03-18 21:36:05 +07:00
|
|
|
|
"profile events on existing process id"),
|
2011-01-31 23:50:39 +07:00
|
|
|
|
OPT_INTEGER('t', "tid", &top.target_tid,
|
2010-03-18 21:36:05 +07:00
|
|
|
|
"profile events on existing thread id"),
|
2009-05-26 14:17:18 +07:00
|
|
|
|
OPT_BOOLEAN('a', "all-cpus", &system_wide,
|
|
|
|
|
"system-wide collection from all CPUs"),
|
2011-01-31 23:50:39 +07:00
|
|
|
|
OPT_STRING('C', "cpu", &top.cpu_list, "cpu",
|
2010-05-28 17:00:01 +07:00
|
|
|
|
"list of cpus to monitor"),
|
2009-11-24 21:05:15 +07:00
|
|
|
|
OPT_STRING('k', "vmlinux", &symbol_conf.vmlinux_name,
|
|
|
|
|
"file", "vmlinux pathname"),
|
2011-01-31 23:50:39 +07:00
|
|
|
|
OPT_BOOLEAN('K', "hide_kernel_symbols", &top.hide_kernel_symbols,
|
2009-11-17 06:45:24 +07:00
|
|
|
|
"hide kernel symbols"),
|
2010-05-18 01:39:16 +07:00
|
|
|
|
OPT_UINTEGER('m', "mmap-pages", &mmap_pages, "number of mmap data pages"),
|
2009-05-26 14:17:18 +07:00
|
|
|
|
OPT_INTEGER('r', "realtime", &realtime_prio,
|
|
|
|
|
"collect data with this RT SCHED_FIFO priority"),
|
2011-01-31 23:50:39 +07:00
|
|
|
|
OPT_INTEGER('d', "delay", &top.delay_secs,
|
2009-05-26 14:17:18 +07:00
|
|
|
|
"number of seconds to delay between refreshes"),
|
|
|
|
|
OPT_BOOLEAN('D', "dump-symtab", &dump_symtab,
|
|
|
|
|
"dump the symbol table used for profiling"),
|
2011-01-31 23:50:39 +07:00
|
|
|
|
OPT_INTEGER('f', "count-filter", &top.count_filter,
|
2009-05-26 14:17:18 +07:00
|
|
|
|
"only display functions with more events than this"),
|
|
|
|
|
OPT_BOOLEAN('g', "group", &group,
|
|
|
|
|
"put the counters into a counter group"),
|
2009-07-21 15:30:36 +07:00
|
|
|
|
OPT_BOOLEAN('i', "inherit", &inherit,
|
|
|
|
|
"child tasks inherit counters"),
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
OPT_STRING(0, "sym-annotate", &sym_filter, "symbol name",
|
2010-02-04 01:52:08 +07:00
|
|
|
|
"symbol to annotate"),
|
2011-01-31 23:50:39 +07:00
|
|
|
|
OPT_BOOLEAN('z', "zero", &top.zero,
|
2009-05-26 14:17:18 +07:00
|
|
|
|
"zero history across updates"),
|
2011-01-31 23:50:39 +07:00
|
|
|
|
OPT_INTEGER('F', "freq", &top.freq,
|
2009-05-26 14:17:18 +07:00
|
|
|
|
"profile at this frequency"),
|
2011-01-31 23:50:39 +07:00
|
|
|
|
OPT_INTEGER('E', "entries", &top.print_entries,
|
2009-06-04 13:53:05 +07:00
|
|
|
|
"display this many functions"),
|
2011-01-31 23:50:39 +07:00
|
|
|
|
OPT_BOOLEAN('U', "hide_user_symbols", &top.hide_user_symbols,
|
2009-11-17 06:45:24 +07:00
|
|
|
|
"hide user symbols"),
|
2011-02-01 03:19:33 +07:00
|
|
|
|
OPT_BOOLEAN(0, "tui", &use_tui, "Use the TUI interface"),
|
|
|
|
|
OPT_BOOLEAN(0, "stdio", &use_stdio, "Use the stdio interface"),
|
2010-04-13 15:37:33 +07:00
|
|
|
|
OPT_INCR('v', "verbose", &verbose,
|
2009-06-07 22:39:02 +07:00
|
|
|
|
"be more verbose (show counter open errors, etc)"),
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
OPT_STRING('s', "sort", &sort_order, "key[,key2...]",
|
|
|
|
|
"sort by key(s): pid, comm, dso, symbol, parent"),
|
|
|
|
|
OPT_BOOLEAN('n', "show-nr-samples", &symbol_conf.show_nr_samples,
|
|
|
|
|
"Show a column with the number of samples"),
|
2011-10-06 05:30:22 +07:00
|
|
|
|
OPT_CALLBACK_DEFAULT('G', "call-graph", NULL, "output_type,min_percent, call_order",
|
|
|
|
|
"Display callchains using output_type (graph, flat, fractal, or none), min percent threshold and callchain order. "
|
|
|
|
|
"Default: fractal,0.5,callee", &parse_callchain_opt,
|
|
|
|
|
callchain_default_opt),
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
OPT_BOOLEAN(0, "show-total-period", &symbol_conf.show_total_period,
|
|
|
|
|
"Show a column with the sum of periods"),
|
|
|
|
|
OPT_STRING(0, "dsos", &symbol_conf.dso_list_str, "dso[,dso...]",
|
|
|
|
|
"only consider symbols in these dsos"),
|
|
|
|
|
OPT_STRING(0, "comms", &symbol_conf.comm_list_str, "comm[,comm...]",
|
|
|
|
|
"only consider symbols in these comms"),
|
|
|
|
|
OPT_STRING(0, "symbols", &symbol_conf.sym_list_str, "symbol[,symbol...]",
|
|
|
|
|
"only consider these symbols"),
|
2011-10-06 22:48:31 +07:00
|
|
|
|
OPT_BOOLEAN(0, "source", &symbol_conf.annotate_src,
|
|
|
|
|
"Interleave source code with assembly code (default)"),
|
|
|
|
|
OPT_BOOLEAN(0, "asm-raw", &symbol_conf.annotate_asm_raw,
|
|
|
|
|
"Display raw encoding of assembly instructions (default)"),
|
|
|
|
|
OPT_STRING('M', "disassembler-style", &disassembler_style, "disassembler style",
|
|
|
|
|
"Specify disassembler style (e.g. -M intel for intel syntax)"),
|
2009-05-26 14:17:18 +07:00
|
|
|
|
OPT_END()
|
|
|
|
|
};
|
|
|
|
|
|
2009-07-01 17:37:06 +07:00
|
|
|
|
int cmd_top(int argc, const char **argv, const char *prefix __used)
|
2009-05-26 14:17:18 +07:00
|
|
|
|
{
|
2011-01-04 01:39:04 +07:00
|
|
|
|
struct perf_evsel *pos;
|
|
|
|
|
int status = -ENOMEM;
|
2009-05-26 14:17:18 +07:00
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
top.evlist = perf_evlist__new(NULL, NULL);
|
|
|
|
|
if (top.evlist == NULL)
|
2011-01-12 05:56:53 +07:00
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
symbol_conf.exclude_other = false;
|
2009-05-26 14:17:18 +07:00
|
|
|
|
|
|
|
|
|
argc = parse_options(argc, argv, options, top_usage, 0);
|
|
|
|
|
if (argc)
|
|
|
|
|
usage_with_options(top_usage, options);
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
if (sort_order == default_sort_order)
|
|
|
|
|
sort_order = "dso,symbol";
|
|
|
|
|
|
|
|
|
|
setup_sorting(top_usage, options);
|
|
|
|
|
|
2011-02-01 03:19:33 +07:00
|
|
|
|
if (use_stdio)
|
|
|
|
|
use_browser = 0;
|
|
|
|
|
else if (use_tui)
|
|
|
|
|
use_browser = 1;
|
|
|
|
|
|
|
|
|
|
setup_browser(false);
|
|
|
|
|
|
2009-05-26 14:17:18 +07:00
|
|
|
|
/* CPU and PID are mutually exclusive */
|
2011-01-31 23:50:39 +07:00
|
|
|
|
if (top.target_tid > 0 && top.cpu_list) {
|
2009-05-26 14:17:18 +07:00
|
|
|
|
printf("WARNING: PID switch overriding CPU\n");
|
|
|
|
|
sleep(1);
|
2011-01-31 23:50:39 +07:00
|
|
|
|
top.cpu_list = NULL;
|
2009-05-26 14:17:18 +07:00
|
|
|
|
}
|
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
if (top.target_pid != -1)
|
|
|
|
|
top.target_tid = top.target_pid;
|
2011-01-30 20:59:43 +07:00
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
if (perf_evlist__create_maps(top.evlist, top.target_pid,
|
|
|
|
|
top.target_tid, top.cpu_list) < 0)
|
2011-01-30 20:59:43 +07:00
|
|
|
|
usage_with_options(top_usage, options);
|
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
if (!top.evlist->nr_entries &&
|
|
|
|
|
perf_evlist__add_default(top.evlist) < 0) {
|
2011-01-04 01:39:04 +07:00
|
|
|
|
pr_err("Not enough memory for event selector list\n");
|
|
|
|
|
return -ENOMEM;
|
|
|
|
|
}
|
2009-11-18 03:38:01 +07:00
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
if (top.delay_secs < 1)
|
|
|
|
|
top.delay_secs = 1;
|
2009-06-06 00:31:01 +07:00
|
|
|
|
|
2009-10-12 12:56:03 +07:00
|
|
|
|
/*
|
|
|
|
|
* User specified count overrides default frequency.
|
|
|
|
|
*/
|
|
|
|
|
if (default_interval)
|
2011-01-31 23:50:39 +07:00
|
|
|
|
top.freq = 0;
|
|
|
|
|
else if (top.freq) {
|
|
|
|
|
default_interval = top.freq;
|
2009-10-12 12:56:03 +07:00
|
|
|
|
} else {
|
|
|
|
|
fprintf(stderr, "frequency and count are zero, aborting\n");
|
|
|
|
|
exit(EXIT_FAILURE);
|
|
|
|
|
}
|
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
list_for_each_entry(pos, &top.evlist->entries, node) {
|
|
|
|
|
if (perf_evsel__alloc_fd(pos, top.evlist->cpus->nr,
|
|
|
|
|
top.evlist->threads->nr) < 0)
|
2011-01-04 01:39:04 +07:00
|
|
|
|
goto out_free_fd;
|
|
|
|
|
/*
|
|
|
|
|
* Fill in the ones not specifically initialized via -c:
|
|
|
|
|
*/
|
|
|
|
|
if (pos->attr.sample_period)
|
|
|
|
|
continue;
|
|
|
|
|
|
|
|
|
|
pos->attr.sample_period = default_interval;
|
|
|
|
|
}
|
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
if (perf_evlist__alloc_pollfd(top.evlist) < 0 ||
|
|
|
|
|
perf_evlist__alloc_mmap(top.evlist) < 0)
|
2011-01-12 07:30:02 +07:00
|
|
|
|
goto out_free_fd;
|
|
|
|
|
|
2011-01-31 23:50:39 +07:00
|
|
|
|
top.sym_evsel = list_entry(top.evlist->entries.next, struct perf_evsel, node);
|
2011-01-12 00:16:52 +07:00
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
symbol_conf.priv_size = sizeof(struct annotation);
|
2011-01-04 01:39:04 +07:00
|
|
|
|
|
|
|
|
|
symbol_conf.try_vmlinux_path = (symbol_conf.vmlinux_name == NULL);
|
|
|
|
|
if (symbol__init() < 0)
|
|
|
|
|
return -1;
|
|
|
|
|
|
perf top: Reuse the 'report' hist_entry/hists classes
This actually fixes several problems we had in the old 'perf top':
1. Unresolved symbols not show, limitation that came from the old
"KernelTop" codebase, to solve it we would need to do changes
that would make sym_entry have most of the hist_entry fields.
2. It was using the number of samples, not the sum of sample->period.
And brings the --sort code that allows us to have all the views in
'perf report', for instance:
[root@emilia ~]# perf top --sort dso
PerfTop: 5903 irqs/sec kernel:77.5% exact: 0.0% [1000Hz cycles], (all, 8 CPUs)
------------------------------------------------------------------------------
31.59% libcrypto.so.1.0.0
21.55% [kernel]
18.57% libpython2.6.so.1.0
7.04% libc-2.12.so
6.99% _backend_agg.so
4.72% sshd
1.48% multiarray.so
1.39% libfreetype.so.6.3.22
1.37% perf
0.71% libgobject-2.0.so.0.2200.5
0.53% [tg3]
0.48% libglib-2.0.so.0.2200.5
0.44% libstdc++.so.6.0.13
0.40% libcairo.so.2.10800.8
0.38% libm-2.12.so
0.34% umath.so
0.30% libgdk-x11-2.0.so.0.1800.9
0.22% libpthread-2.12.so
0.20% libgtk-x11-2.0.so.0.1800.9
0.20% librt-2.12.so
0.15% _path.so
0.13% libpango-1.0.so.0.2800.1
0.11% libatlas.so.3.0
0.09% ft2font.so
0.09% libpangoft2-1.0.so.0.2800.1
0.08% libX11.so.6.3.0
0.07% [vdso]
0.06% cyclictest
^C
All the filter lists can be used as well: --dsos, --comms, --symbols,
etc.
The 'perf report' TUI is also reused, being possible to apply all the
zoom operations, do annotation, etc.
This change will allow multiple simplifications in the symbol system as
well, that will be detailed in upcoming changesets.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-xzaaldxq7zhqrrxdxjifk1mh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2011-10-06 05:16:15 +07:00
|
|
|
|
sort_entry__setup_elide(&sort_dso, symbol_conf.dso_list, "dso", stdout);
|
|
|
|
|
sort_entry__setup_elide(&sort_comm, symbol_conf.comm_list, "comm", stdout);
|
|
|
|
|
sort_entry__setup_elide(&sort_sym, symbol_conf.sym_list, "symbol", stdout);
|
|
|
|
|
|
2011-10-06 05:30:22 +07:00
|
|
|
|
/*
|
|
|
|
|
* Avoid annotation data structures overhead when symbols aren't on the
|
|
|
|
|
* sort list.
|
|
|
|
|
*/
|
|
|
|
|
sort_has_symbols = sort_sym.list.next != NULL;
|
|
|
|
|
|
2009-11-18 00:40:54 +07:00
|
|
|
|
get_term_dimensions(&winsize);
|
2011-01-31 23:50:39 +07:00
|
|
|
|
if (top.print_entries == 0) {
|
2009-11-18 00:40:54 +07:00
|
|
|
|
update_print_entries(&winsize);
|
2009-11-17 04:30:27 +07:00
|
|
|
|
signal(SIGWINCH, sig_winch_handler);
|
|
|
|
|
}
|
|
|
|
|
|
2011-01-04 01:39:04 +07:00
|
|
|
|
status = __cmd_top();
|
|
|
|
|
out_free_fd:
|
2011-01-31 23:50:39 +07:00
|
|
|
|
perf_evlist__delete(top.evlist);
|
2011-01-04 01:39:04 +07:00
|
|
|
|
|
|
|
|
|
return status;
|
2009-05-26 14:17:18 +07:00
|
|
|
|
}
|