很多的公司，包括我們公司，還在使用CentOS5。但內核早已換成el6系列的2.6.32內核。perf工具是內核自帶的一個性能評估工具，功能很強大，爲了在CentOS5下進行性能優化，我們要在CentOS5下對2.6.32內核的perf源碼進行編譯。我這裏使用了linux-2.6.32-358.el6的內核源碼。
進入源碼目錄，打上以下補丁：
1、004-150-perf-tools-Fix-build-with-bison-2.3-and-older..patch
--- a/tools/perf/Makefile
+++ b/tools/perf/Makefile
@@ -276,13 +276,13 @@ $(OUTPUT)util/parse-events-flex.c: util/
$(QUIET_FLEX)$(FLEX) --header-file=$(OUTPUT)util/parse-events-flex.h $(PARSER_DEBUG_FLEX) -t util/parse-events.l > $(OUTPUT)util/parse-events-flex.c
$(OUTPUT)util/parse-events-bison.c: util/parse-events.y
- $(QUIET_BISON)$(BISON) -v util/parse-events.y -d $(PARSER_DEBUG_BISON) -o $(OUTPUT)util/parse-events-bison.c
+ $(QUIET_BISON)$(BISON) -v util/parse-events.y -d $(PARSER_DEBUG_BISON) -o $(OUTPUT)util/parse-events-bison.c -p parse_events_
$(OUTPUT)util/pmu-flex.c: util/pmu.l $(OUTPUT)util/pmu-bison.c
$(QUIET_FLEX)$(FLEX) --header-file=$(OUTPUT)util/pmu-flex.h -t util/pmu.l > $(OUTPUT)util/pmu-flex.c
$(OUTPUT)util/pmu-bison.c: util/pmu.y
- $(QUIET_BISON)$(BISON) -v util/pmu.y -d -o $(OUTPUT)util/pmu-bison.c
+ $(QUIET_BISON)$(BISON) -v util/pmu.y -d -o $(OUTPUT)util/pmu-bison.c -p perf_pmu_
$(OUTPUT)util/parse-events.o: $(OUTPUT)util/parse-events-flex.c $(OUTPUT)util/parse-events-bison.c
$(OUTPUT)util/pmu.o: $(OUTPUT)util/pmu-flex.c $(OUTPUT)util/pmu-bison.c
--- a/tools/perf/util/parse-events.y
+++ b/tools/perf/util/parse-events.y
@@ -1,5 +1,4 @@
%pure-parser
-%name-prefix "parse_events_"
%parse-param {void *_data}
%parse-param {void *scanner}
%lex-param {void* scanner}
--- a/tools/perf/util/pmu.y
+++ b/tools/perf/util/pmu.y
@@ -1,5 +1,4 @@
-%name-prefix "perf_pmu_"
%parse-param {struct list_head *format}
%parse-param {char *name}
diff --git a/tools/perf/util/dso-test-data.c b/tools/perf/util/dso-test-data.c
index 541cdc7..c6caede 100644
--- a/tools/perf/util/dso-test-data.c
+++ b/tools/perf/util/dso-test-data.c
@@ -23,7 +23,7 @@ static char *test_file(int size)
int fd, i;
unsigned char *buf;
- fd = mkostemp(templ, O_CREAT|O_WRONLY|O_TRUNC);
+ fd = mkstemp(templ);
buf = malloc(size);
if (!buf) {
3、tip-perf-core-perf-test-fix-a-build-error-on-builtin-test.patch
diff --git a/tools/perf/builtin-test.c b/tools/perf/builtin-test.c
index b5a544d..5d4354e 100644
--- a/tools/perf/builtin-test.c
+++ b/tools/perf/builtin-test.c
@@ -605,19 +605,13 @@ out_free_threads:
#undef nsyscalls
}
-static int sched__get_first_possible_cpu(pid_t pid, cpu_set_t **maskp,
- size_t *sizep)
+static int sched__get_first_possible_cpu(pid_t pid, cpu_set_t *maskp)
{
- cpu_set_t *mask;
- size_t size;
int i, cpu = -1, nrcpus = 1024;
realloc:
- mask = CPU_ALLOC(nrcpus);
- size = CPU_ALLOC_SIZE(nrcpus);
- CPU_ZERO_S(size, mask);
+ CPU_ZERO(maskp);
- if (sched_getaffinity(pid, size, mask) == -1) {
- CPU_FREE(mask);
+ if (sched_getaffinity(pid, sizeof(*maskp), maskp) == -1) {
if (errno == EINVAL && nrcpus < (1024 << 8)) {
nrcpus = nrcpus << 2;
goto realloc;
@@ -627,19 +621,14 @@ realloc:
}
for (i = 0; i < nrcpus; i++) {
- if (CPU_ISSET_S(i, size, mask)) {
- if (cpu == -1) {
+ if (CPU_ISSET(i, maskp)) {
+ if (cpu == -1)
cpu = i;
- *maskp = mask;
- *sizep = size;
- } else
- CPU_CLR_S(i, size, mask);
+ else
+ CPU_CLR(i, maskp);
}
}
- if (cpu == -1)
- CPU_FREE(mask);
-
return cpu;
}
@@ -654,8 +643,8 @@ static int test__PERF_RECORD(void)
.freq = 10,
.mmap_pages = 256,
};
- cpu_set_t *cpu_mask = NULL;
- size_t cpu_mask_size = 0;
+ cpu_set_t cpu_mask;
+ size_t cpu_mask_size = sizeof(cpu_mask);
struct perf_evlist *evlist = perf_evlist__new(NULL, NULL);
struct perf_evsel *evsel;
struct perf_sample sample;
@@ -719,8 +708,7 @@ static int test__PERF_RECORD(void)
evsel->attr.sample_type |= PERF_SAMPLE_TIME;
perf_evlist__config_attrs(evlist, &opts);
- err = sched__get_first_possible_cpu(evlist->workload.pid, &cpu_mask,
- &cpu_mask_size);
+ err = sched__get_first_possible_cpu(evlist->workload.pid, &cpu_mask);
if (err < 0) {
pr_debug("sched__get_first_possible_cpu: %s\n", strerror(errno));
goto out_delete_evlist;
@@ -731,9 +719,9 @@ static int test__PERF_RECORD(void)
/*
* So that we can check perf_sample.cpu on all the samples.
*/
- if (sched_setaffinity(evlist->workload.pid, cpu_mask_size, cpu_mask) < 0) {
+ if (sched_setaffinity(evlist->workload.pid, cpu_mask_size, &cpu_mask) < 0) {
pr_debug("sched_setaffinity: %s\n", strerror(errno));
- goto out_free_cpu_mask;
+ goto out_delete_evlist;
}
/*
@@ -917,8 +905,6 @@ found_exit:
}
out_err:
perf_evlist__munmap(evlist);
-out_free_cpu_mask:
- CPU_FREE(cpu_mask);
out_delete_evlist:
perf_evlist__delete(evlist);
out:
然後再安裝
binutils220-2.20.51.0.2-5.29.el5.x86_64.rpm
flex-2.5.35-0.8.el5.rfb.x86_64.rpm
libgomp-4.4.7-1.el5.x86_64.rpm
這幾個包。
進入tools/perf，make一下就可以了。
這幾個包和patch，以及我編譯出來的用於CentOS5下的32內核的perf靜態文件在這可以下載：
http://download.csdn.net/detail/abigwc/9799082