From: Arnaldo Carvalho de Melo <acme@kernel.org>
To: Ingo Molnar <mingo@kernel.org>, Thomas Gleixner <tglx@linutronix.de>
Cc: Jiri Olsa <jolsa@kernel.org>, Namhyung Kim <namhyung@kernel.org>,
Clark Williams <williams@redhat.com>,
linux-kernel@vger.kernel.org, linux-perf-users@vger.kernel.org,
Ian Rogers <irogers@google.com>,
Arnaldo Carvalho de Melo <acme@redhat.com>,
Alexander Shishkin <alexander.shishkin@linux.intel.com>,
Jiri Olsa <jolsa@redhat.com>, Mark Rutland <mark.rutland@arm.com>,
Peter Zijlstra <peterz@infradead.org>,
Stephane Eranian <eranian@google.com>
Subject: [PATCH 88/91] libsymbols kallsyms: Parse using io api
Date: Wed, 6 May 2020 12:22:31 -0300 [thread overview]
Message-ID: <20200506152234.21977-89-acme@kernel.org> (raw)
In-Reply-To: <20200506152234.21977-1-acme@kernel.org>
From: Ian Rogers <irogers@google.com>
'perf record' will call kallsyms__parse 4 times during startup and
process megabytes of data. This changes kallsyms__parse to use the io
library rather than fgets to improve performance of the user code by
over 8%.
Before:
Running 'internals/kallsyms-parse' benchmark:
Average kallsyms__parse took: 103.988 ms (+- 0.203 ms)
After:
Running 'internals/kallsyms-parse' benchmark:
Average kallsyms__parse took: 95.571 ms (+- 0.006 ms)
For a workload like:
$ perf record /bin/true
Run under 'perf record -e cycles:u -g' the time goes from:
Before
30.10% 1.67% perf perf [.] kallsyms__parse
After
25.55% 20.04% perf perf [.] kallsyms__parse
So a little under 5% of the start-up time is removed. A lot of what
remains is on the kernel side, but caching kallsyms within perf would at
least impact memory footprint.
Committer notes:
The internal/kallsyms-parse bench is run using:
[root@five ~]# perf bench internals kallsyms-parse
# Running 'internals/kallsyms-parse' benchmark:
Average kallsyms__parse took: 80.381 ms (+- 0.115 ms)
[root@five ~]#
And this pre-existing test uses these routines to parse kallsyms and
then compare with the info obtained from the matching ELF symtab:
[root@five ~]# perf test vmlinux
1: vmlinux symtab matches kallsyms : Ok
[root@five ~]#
Also we can't remove hex2u64() in this patch as this breaks the build:
/usr/bin/ld: /tmp/build/perf/perf-in.o: in function `modules__parse':
/home/acme/git/perf/tools/perf/util/symbol.c:607: undefined reference to `hex2u64'
/usr/bin/ld: /home/acme/git/perf/tools/perf/util/symbol.c:607: undefined reference to `hex2u64'
/usr/bin/ld: /tmp/build/perf/perf-in.o: in function `dso__load_perf_map':
/home/acme/git/perf/tools/perf/util/symbol.c:1477: undefined reference to `hex2u64'
/usr/bin/ld: /home/acme/git/perf/tools/perf/util/symbol.c:1483: undefined reference to `hex2u64'
collect2: error: ld returned 1 exit status
Leave it there, move it in the next patch.
Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lore.kernel.org/lkml/20200501221315.54715-3-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
tools/lib/api/io.h | 3 ++
tools/lib/symbol/kallsyms.c | 93 +++++++++++++++++++------------------
2 files changed, 51 insertions(+), 45 deletions(-)
diff --git a/tools/lib/api/io.h b/tools/lib/api/io.h
index b7e55b5f8a4a..777c20f6b604 100644
--- a/tools/lib/api/io.h
+++ b/tools/lib/api/io.h
@@ -7,6 +7,9 @@
#ifndef __API_IO__
#define __API_IO__
+#include <stdlib.h>
+#include <unistd.h>
+
struct io {
/* File descriptor being read/ */
int fd;
diff --git a/tools/lib/symbol/kallsyms.c b/tools/lib/symbol/kallsyms.c
index 1a7a9f877095..a5edc75c346c 100644
--- a/tools/lib/symbol/kallsyms.c
+++ b/tools/lib/symbol/kallsyms.c
@@ -1,7 +1,9 @@
// SPDX-License-Identifier: GPL-2.0
#include "symbol/kallsyms.h"
+#include "api/io.h"
#include <stdio.h>
-#include <stdlib.h>
+#include <sys/stat.h>
+#include <fcntl.h>
u8 kallsyms2elf_type(char type)
{
@@ -9,12 +11,6 @@ u8 kallsyms2elf_type(char type)
return (type == 't' || type == 'w') ? STT_FUNC : STT_OBJECT;
}
-bool kallsyms__is_function(char symbol_type)
-{
- symbol_type = toupper(symbol_type);
- return symbol_type == 'T' || symbol_type == 'W';
-}
-
/*
* While we find nice hex chars, build a long_val.
* Return number of chars processed.
@@ -28,61 +24,68 @@ int hex2u64(const char *ptr, u64 *long_val)
return p - ptr;
}
+bool kallsyms__is_function(char symbol_type)
+{
+ symbol_type = toupper(symbol_type);
+ return symbol_type == 'T' || symbol_type == 'W';
+}
+
+static void read_to_eol(struct io *io)
+{
+ int ch;
+
+ for (;;) {
+ ch = io__get_char(io);
+ if (ch < 0 || ch == '\n')
+ return;
+ }
+}
+
int kallsyms__parse(const char *filename, void *arg,
int (*process_symbol)(void *arg, const char *name,
char type, u64 start))
{
- char *line = NULL;
- size_t n;
- int err = -1;
- FILE *file = fopen(filename, "r");
-
- if (file == NULL)
- goto out_failure;
-
- err = 0;
+ struct io io;
+ char bf[BUFSIZ];
+ int err;
- while (!feof(file)) {
- u64 start;
- int line_len, len;
- char symbol_type;
- char *symbol_name;
+ io.fd = open(filename, O_RDONLY, 0);
- line_len = getline(&line, &n, file);
- if (line_len < 0 || !line)
- break;
+ if (io.fd < 0)
+ return -1;
- line[--line_len] = '\0'; /* \n */
+ io__init(&io, io.fd, bf, sizeof(bf));
- len = hex2u64(line, &start);
+ err = 0;
+ while (!io.eof) {
+ __u64 start;
+ int ch;
+ size_t i;
+ char symbol_type;
+ char symbol_name[KSYM_NAME_LEN + 1];
- /* Skip the line if we failed to parse the address. */
- if (!len)
+ if (io__get_hex(&io, &start) != ' ') {
+ read_to_eol(&io);
continue;
-
- len++;
- if (len + 2 >= line_len)
+ }
+ symbol_type = io__get_char(&io);
+ if (io__get_char(&io) != ' ') {
+ read_to_eol(&io);
continue;
-
- symbol_type = line[len];
- len += 2;
- symbol_name = line + len;
- len = line_len - len;
-
- if (len >= KSYM_NAME_LEN) {
- err = -1;
- break;
}
+ for (i = 0; i < sizeof(symbol_name); i++) {
+ ch = io__get_char(&io);
+ if (ch < 0 || ch == '\n')
+ break;
+ symbol_name[i] = ch;
+ }
+ symbol_name[i] = '\0';
err = process_symbol(arg, symbol_name, symbol_type, start);
if (err)
break;
}
- free(line);
- fclose(file);
+ close(io.fd);
return err;
-
-out_failure:
- return -1;
}
--
2.21.1
next prev parent reply other threads:[~2020-05-06 15:27 UTC|newest]
Thread overview: 92+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-05-06 15:21 [GIT PULL] perf/core improvements and fixes Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 01/91] perf cgroup: Avoid needless closing of unopened fd Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 02/91] perf bench: Fix div-by-zero if runtime is zero Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 03/91] perf evlist: Remove duplicate headers Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 04/91] perf script: Avoid NULL dereference on symbol Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 05/91] perf stat: Zero all the 'ena' and 'run' array slot stats for interval mode Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 06/91] perf stat: Improve runtime stat " Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 07/91] perf test session topology: Fix data path Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 08/91] perf record: Add num-synthesize-threads option Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 09/91] perf bench: Add a multi-threaded synthesize benchmark Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 10/91] tools api: Add a lightweight buffered reading api Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 11/91] perf synthetic events: Remove use of sscanf from /proc reading Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 12/91] perf script: Remove extraneous newline in perf_sample__fprintf_regs() Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 13/91] libtraceevent: Remove unneeded semicolon Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 14/91] perf c2c: " Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 15/91] perf tools: Remove unneeded semicolons Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 16/91] perf report: Fix warning assignment of 0/1 to bool variable Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 17/91] perf pmu: Fix function name in comment, its get_cpuid_str(), not get_cpustr() Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 18/91] perf metricgroups: Enhance JSON/metric infrastructure to handle "?" Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 19/91] perf tests expr: Added test for runtime param in metric expression Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 20/91] perf tools: Enable Hz/hz prinitg for --metric-only option Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 21/91] perf vendor events power9: Add hv_24x7 socket/chip level metric events Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 22/91] perf tools: Move routines that probe for perf API features to separate file Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 23/91] perf record: Move sb_evlist to 'struct record' Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 24/91] perf top: Move sb_evlist to 'struct perf_top' Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 25/91] perf bpf: Decouple creating the evlist from adding the SB event Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 26/91] perf parse-events: Add parse_events_option() variant that creates evlist Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 27/91] perf evlist: Move the sideband thread routines to separate object Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 28/91] perf evlist: Allow reusing the side band thread for more purposes Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 29/91] libsubcmd: Introduce OPT_CALLBACK_SET() Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 30/91] perf record: Introduce --switch-output-event Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 31/91] perf record: Move side band evlist setup to separate routine Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 32/91] libperf: Add NULL pointer check for cpu_map iteration and NULL assignment for all_cpus Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 33/91] perf parse-events: Fix memory leaks found on parse_events Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 34/91] " Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 35/91] perf parse-events: Fix another memory leaks found on parse_events() Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 36/91] libperf evlist: Fix a refcount leak Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 37/91] perf tools: Fix reading new topology attribute "core_cpus" Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 38/91] perf tools: Simplify checking if SMT is active Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 39/91] perf thread-stack: Add branch stack support Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 40/91] perf intel-pt: Consolidate thread-stack use condition Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 41/91] perf intel-pt: Change branch stack support to use thread-stacks Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 42/91] perf auxtrace: Add option to synthesize branch stack for regular events Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 43/91] perf evsel: Add support for synthesized branch stack sample type Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 44/91] perf thread-stack: Add thread_stack__br_sample_late() Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 45/91] perf intel-pt: Add support for synthesizing branch stacks for regular events Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 46/91] perf intel-pt: Update documentation about itrace G and L options Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 47/91] perf intel-pt: Update documentation about using /proc/kcore Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 48/91] perf evsel: Rename 'struct perf_evsel__sb_cb_t' to 'struct evsel__sb_cb_t' Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 49/91] perf evsel: Rename perf_evsel__nr_cpus() to evsel__nr_cpus() Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 50/91] perf evsel: Rename perf_evsel__compute_deltas() to evsel__compute_deltas() Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 51/91] perf evsel: Rename perf_evsel__find_pmu() to evsel__find_pmu() Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 52/91] perf evsel: Rename perf_evsel__is_aux_event() to evsel__is_aux_event() Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 53/91] perf evsel: Rename perf_evsel__exit() to evsel__exit() Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 54/91] perf evsel: Rename perf_evsel__config*() to evsel__config*() Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 55/91] perf evsel: Rename perf_evsel__calc_id_pos() to evsel__calc_id_pos() Arnaldo Carvalho de Melo
2020-05-06 15:21 ` [PATCH 56/91] perf evsel: Rename __perf_evsel__sample_size() to __evsel__sample_size() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 57/91] perf evsel: Rename *perf_evsel__*name() to *evsel__*name() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 58/91] perf evsel: Rename perf_evsel__group_desc() to evsel__group_desc() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 59/91] perf evsel: Rename *perf_evsel__*set_sample_*() to *evsel__*set_sample_*() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 60/91] perf evsel: Rename perf_evsel__*filter*() to evsel__*filter*() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 61/91] perf evsel: Rename perf_evsel__open_per_*() to evsel__open_per_*() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 62/91] perf evsel: Rename perf_evsel__{str,int}val() and other tracepoint field metehods to to evsel__*() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 63/91] perf mem2node: Avoid double free related to realloc Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 64/91] perf doc: Pass ASCIIDOC_EXTRA as an argument Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 65/91] tools feature: Add support for detecting libpfm4 Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 66/91] perf pmu: Add perf_pmu__find_by_type helper Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 67/91] perf evsel: Rename perf_evsel__is_*() to evsel__is*() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 68/91] perf evsel: Ditch perf_evsel__cmp(), not used for quite a while Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 69/91] perf evsel: Rename *perf_evsel__read*() to *evsel__read() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 70/91] perf evsel: Rename perf_evsel__parse_sample*() to evsel__parse_sample*() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 71/91] perf evsel: Rename perf_evsel__{prev,next}() to evsel__{prev,next}() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 72/91] perf evsel: Rename perf_evsel__has*() to evsel__has*() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 73/91] perf evsel: Rename perf_evsel__fallback() to evsel__fallback() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 74/91] perf evsel: Rename perf_evsel__group_idx() to evsel__group_idx() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 75/91] perf evsel: Rename perf_evsel__env() to evsel__env() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 76/91] perf evsel: Rename perf_evsel__store_ids() to evsel__store_id() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 77/91] perf stat: Rename perf_evsel__*() operating on 'struct evsel *' to evsel__*() Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 78/91] perf kmem: " Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 79/91] perf lock: " Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 80/91] perf sched: " Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 81/91] perf script: " Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 82/91] perf trace: " Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 83/91] perf annotate: " Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 84/91] perf inject: " Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 85/91] perf symbol: Fix kernel symbol address display Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 86/91] perf: cs-etm: Update to build with latest opencsd version Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 87/91] perf bench: Add kallsyms parsing Arnaldo Carvalho de Melo
2020-05-06 15:22 ` Arnaldo Carvalho de Melo [this message]
2020-05-06 15:22 ` [PATCH 89/91] libsymbols kallsyms: Move hex2u64 out of header Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 90/91] perf cs-etm: Move definition of 'traceid_list' global variable from header file Arnaldo Carvalho de Melo
2020-05-06 15:22 ` [PATCH 91/91] perf flamegraph: Use /bin/bash for report and record scripts Arnaldo Carvalho de Melo
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20200506152234.21977-89-acme@kernel.org \
--to=acme@kernel.org \
--cc=acme@redhat.com \
--cc=alexander.shishkin@linux.intel.com \
--cc=eranian@google.com \
--cc=irogers@google.com \
--cc=jolsa@kernel.org \
--cc=jolsa@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-perf-users@vger.kernel.org \
--cc=mark.rutland@arm.com \
--cc=mingo@kernel.org \
--cc=namhyung@kernel.org \
--cc=peterz@infradead.org \
--cc=tglx@linutronix.de \
--cc=williams@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox