Skip to content

L
linux
Commit ddcacfa0 authored by Ingo Molnar's avatar Ingo Molnar
Browse files
Options

perf_counter tools: separate kerneltop into 'perf top' and 'perf stat' 

Lets use the Git framework of built-in commands.
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
parent 6f06ccbc
Hide whitespace changes
Inline Side-by-side
Showing with 601 additions and 201 deletions
Documentation/perf_counter/Makefile
View file @ ddcacfa0
......@@ -309,6 +309,7 @@ LIB_OBJS += usage.o
LIB_OBJS += wrapper.o
BUILTIN_OBJS += builtin-help.o
BUILTIN_OBJS += builtin-stat.o
BUILTIN_OBJS += builtin-top.o
PERFLIBS = $(LIB_FILE)
......
Documentation/perf_counter/kerneltop.c → Documentation/perf_counter/builtin-stat.c
View file @ ddcacfa0
......@@ -61,15 +61,8 @@
* Released under the GPL v2. (and only v2, not any later version)
*/
#define _GNU_SOURCE
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/time.h>
#include <unistd.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
#include <limits.h>
#include "util.h"
#include <getopt.h>
#include <assert.h>
#include <fcntl.h>
......@@ -101,8 +94,6 @@
#define PR_TASK_PERF_COUNTERS_DISABLE 31
#define PR_TASK_PERF_COUNTERS_ENABLE 32
#define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
#define rdclock() \
({ \
struct timespec ts; \
......@@ -142,23 +133,18 @@
(void) (&_min1 == &_min2); \
_min1 < _min2 ? _min1 : _min2; })
asmlinkage int sys_perf_counter_open(
extern asmlinkage int sys_perf_counter_open(
struct perf_counter_hw_event *hw_event_uptr __user,
pid_t pid,
int cpu,
int group_fd,
unsigned long flags)
{
return syscall(
__NR_perf_counter_open, hw_event_uptr, pid, cpu, group_fd, flags);
}
unsigned long flags);
#define MAX_COUNTERS 64
#define MAX_NR_CPUS 256
#define EID(type, id) (((__u64)(type) << PERF_COUNTER_TYPE_SHIFT) | (id))
static int run_perfstat = 0;
static int system_wide = 0;
static int nr_counters = 0;
......@@ -177,42 +163,18 @@ static int default_interval = 100000;
static int event_count[MAX_COUNTERS];
static int fd[MAX_NR_CPUS][MAX_COUNTERS];
static __u64 count_filter = 100;
static int tid = -1;
static int profile_cpu = -1;
static int nr_cpus = 0;
static int nmi = 1;
static unsigned int realtime_prio = 0;
static int group = 0;
static unsigned int page_size;
static unsigned int mmap_pages = 16;
static int use_mmap = 0;
static int use_munmap = 0;
static char *vmlinux;
static char *sym_filter;
static unsigned long filter_start;
static unsigned long filter_end;
static int delay_secs = 2;
static int zero;
static int dump_symtab;
static int scale;
struct source_line {
uint64_t EIP;
unsigned long count;
char *line;
struct source_line *next;
};
static struct source_line *lines;
static struct source_line **lines_tail;
const unsigned int default_count[] = {
static const unsigned int default_count[] = {
1000000,
1000000,
10000,
......@@ -300,7 +262,7 @@ static void display_events_help(void)
" rNNN: raw PMU events (eventsel+umask)\n\n");
}
static void display_perfstat_help(void)
static void display_help(void)
{
printf(
"Usage: perfstat [<events...>] <cmd...>\n\n"
......@@ -315,41 +277,6 @@ static void display_perfstat_help(void)
exit(0);
}
static void display_help(void)
{
if (run_perfstat)
return display_perfstat_help();
printf(
"Usage: kerneltop [<options>]\n"
" Or: kerneltop -S [<options>] COMMAND [ARGS]\n\n"
"KernelTop Options (up to %d event types can be specified at once):\n\n",
MAX_COUNTERS);
display_events_help();
printf(
" -S --stat # perfstat COMMAND\n"
" -a # system-wide collection (for perfstat)\n\n"
" -c CNT --count=CNT # event period to sample\n\n"
" -C CPU --cpu=CPU # CPU (-1 for all) [default: -1]\n"
" -p PID --pid=PID # PID of sampled task (-1 for all) [default: -1]\n\n"
" -l # show scale factor for RR events\n"
" -d delay --delay=<seconds> # sampling/display delay [default: 2]\n"
" -f CNT --filter=CNT # min-event-count filter [default: 100]\n\n"
" -r prio --realtime=<prio> # event acquisition runs with SCHED_FIFO policy\n"
" -s symbol --symbol=<symbol> # function to be showed annotated one-shot\n"
" -x path --vmlinux=<path> # the vmlinux binary, required for -s use\n"
" -z --zero # zero counts after display\n"
" -D --dump_symtab # dump symbol table to stderr on startup\n"
" -m pages --mmap_pages=<pages> # number of mmap data pages\n"
" -M --mmap_info # print mmap info stream\n"
" -U --munmap_info # print munmap info stream\n"
);
exit(0);
}
static char *event_name(int ctr)
{
__u64 config = event_id[ctr];
......@@ -574,514 +501,10 @@ int do_perfstat(int argc, char *argv[])
return 0;
}
/*
* Symbols
*/
static uint64_t min_ip;
static uint64_t max_ip = -1ll;
struct sym_entry {
unsigned long long addr;
char *sym;
unsigned long count[MAX_COUNTERS];
int skip;
struct source_line *source;
};
#define MAX_SYMS 100000
static int sym_table_count;
struct sym_entry *sym_filter_entry;
static struct sym_entry sym_table[MAX_SYMS];
static void show_details(struct sym_entry *sym);
/*
* Ordering weight: count-1 * count-2 * ... / count-n
*/
static double sym_weight(const struct sym_entry *sym)
{
double weight;
int counter;
weight = sym->count[0];
for (counter = 1; counter < nr_counters-1; counter++)
weight *= sym->count[counter];
weight /= (sym->count[counter] + 1);
return weight;
}
static int compare(const void *__sym1, const void *__sym2)
{
const struct sym_entry *sym1 = __sym1, *sym2 = __sym2;
return sym_weight(sym1) < sym_weight(sym2);
}
static long events;
static long userspace_events;
static const char CONSOLE_CLEAR[] = "";
static struct sym_entry tmp[MAX_SYMS];
static void print_sym_table(void)
{
int i, printed;
int counter;
float events_per_sec = events/delay_secs;
float kevents_per_sec = (events-userspace_events)/delay_secs;
float sum_kevents = 0.0;
events = userspace_events = 0;
memcpy(tmp, sym_table, sizeof(sym_table[0])*sym_table_count);
qsort(tmp, sym_table_count, sizeof(tmp[0]), compare);
for (i = 0; i < sym_table_count && tmp[i].count[0]; i++)
sum_kevents += tmp[i].count[0];
write(1, CONSOLE_CLEAR, strlen(CONSOLE_CLEAR));
printf(
"------------------------------------------------------------------------------\n");
printf( " KernelTop:%8.0f irqs/sec kernel:%4.1f%% [%s, ",
events_per_sec,
100.0 - (100.0*((events_per_sec-kevents_per_sec)/events_per_sec)),
nmi ? "NMI" : "IRQ");
if (nr_counters == 1)
printf("%d ", event_count[0]);
for (counter = 0; counter < nr_counters; counter++) {
if (counter)
printf("/");
printf("%s", event_name(counter));
}
printf( "], ");
if (tid != -1)
printf(" (tid: %d", tid);
else
printf(" (all");
if (profile_cpu != -1)
printf(", cpu: %d)\n", profile_cpu);
else {
if (tid != -1)
printf(")\n");
else
printf(", %d CPUs)\n", nr_cpus);
}
printf("------------------------------------------------------------------------------\n\n");
if (nr_counters == 1)
printf(" events pcnt");
else
printf(" weight events pcnt");
printf(" RIP kernel function\n"
" ______ ______ _____ ________________ _______________\n\n"
);
for (i = 0, printed = 0; i < sym_table_count; i++) {
float pcnt;
int count;
if (printed <= 18 && tmp[i].count[0] >= count_filter) {
pcnt = 100.0 - (100.0*((sum_kevents-tmp[i].count[0])/sum_kevents));
if (nr_counters == 1)
printf("%19.2f - %4.1f%% - %016llx : %s\n",
sym_weight(tmp + i),
pcnt, tmp[i].addr, tmp[i].sym);
else
printf("%8.1f %10ld - %4.1f%% - %016llx : %s\n",
sym_weight(tmp + i),
tmp[i].count[0],
pcnt, tmp[i].addr, tmp[i].sym);
printed++;
}
/*
* Add decay to the counts:
*/
for (count = 0; count < nr_counters; count++)
sym_table[i].count[count] = zero ? 0 : sym_table[i].count[count] * 7 / 8;
}
if (sym_filter_entry)
show_details(sym_filter_entry);
{
struct pollfd stdin_poll = { .fd = 0, .events = POLLIN };
if (poll(&stdin_poll, 1, 0) == 1) {
printf("key pressed - exiting.\n");
exit(0);
}
}
}
static void *display_thread(void *arg)
{
printf("KernelTop refresh period: %d seconds\n", delay_secs);
while (!sleep(delay_secs))
print_sym_table();
return NULL;
}
static int read_symbol(FILE *in, struct sym_entry *s)
{
static int filter_match = 0;
char *sym, stype;
char str[500];
int rc, pos;
rc = fscanf(in, "%llx %c %499s", &s->addr, &stype, str);
if (rc == EOF)
return -1;
assert(rc == 3);
/* skip until end of line: */
pos = strlen(str);
do {
rc = fgetc(in);
if (rc == '\n' || rc == EOF || pos >= 499)
break;
str[pos] = rc;
pos++;
} while (1);
str[pos] = 0;
sym = str;
/* Filter out known duplicates and non-text symbols. */
if (!strcmp(sym, "_text"))
return 1;
if (!min_ip && !strcmp(sym, "_stext"))
return 1;
if (!strcmp(sym, "_etext") || !strcmp(sym, "_sinittext"))
return 1;
if (stype != 'T' && stype != 't')
return 1;
if (!strncmp("init_module", sym, 11) || !strncmp("cleanup_module", sym, 14))
return 1;
if (strstr(sym, "_text_start") || strstr(sym, "_text_end"))
return 1;
s->sym = malloc(strlen(str));
assert(s->sym);
strcpy((char *)s->sym, str);
s->skip = 0;
/* Tag events to be skipped. */
if (!strcmp("default_idle", s->sym) || !strcmp("cpu_idle", s->sym))
s->skip = 1;
else if (!strcmp("enter_idle", s->sym) || !strcmp("exit_idle", s->sym))
s->skip = 1;
else if (!strcmp("mwait_idle", s->sym))
s->skip = 1;
if (filter_match == 1) {
filter_end = s->addr;
filter_match = -1;
if (filter_end - filter_start > 10000) {
printf("hm, too large filter symbol <%s> - skipping.\n",
sym_filter);
printf("symbol filter start: %016lx\n", filter_start);
printf(" end: %016lx\n", filter_end);
filter_end = filter_start = 0;
sym_filter = NULL;
sleep(1);
}
}
if (filter_match == 0 && sym_filter && !strcmp(s->sym, sym_filter)) {
filter_match = 1;
filter_start = s->addr;
}
return 0;
}
int compare_addr(const void *__sym1, const void *__sym2)
{
const struct sym_entry *sym1 = __sym1, *sym2 = __sym2;
return sym1->addr > sym2->addr;
}
static void sort_symbol_table(void)
{
int i, dups;
do {
qsort(sym_table, sym_table_count, sizeof(sym_table[0]), compare_addr);
for (i = 0, dups = 0; i < sym_table_count; i++) {
if (sym_table[i].addr == sym_table[i+1].addr) {
sym_table[i+1].addr = -1ll;
dups++;
}
}
sym_table_count -= dups;
} while(dups);
}
static void parse_symbols(void)
{
struct sym_entry *last;
FILE *kallsyms = fopen("/proc/kallsyms", "r");
if (!kallsyms) {
printf("Could not open /proc/kallsyms - no CONFIG_KALLSYMS_ALL=y?\n");
exit(-1);
}
while (!feof(kallsyms)) {
if (read_symbol(kallsyms, &sym_table[sym_table_count]) == 0) {
sym_table_count++;
assert(sym_table_count <= MAX_SYMS);
}
}
sort_symbol_table();
min_ip = sym_table[0].addr;
max_ip = sym_table[sym_table_count-1].addr;
last = sym_table + sym_table_count++;
last->addr = -1ll;
last->sym = "<end>";
if (filter_end) {
int count;
for (count=0; count < sym_table_count; count ++) {
if (!strcmp(sym_table[count].sym, sym_filter)) {
sym_filter_entry = &sym_table[count];
break;
}
}
}
if (dump_symtab) {
int i;
for (i = 0; i < sym_table_count; i++)
fprintf(stderr, "%llx %s\n",
sym_table[i].addr, sym_table[i].sym);
}
}
/*
* Source lines
*/
static void parse_vmlinux(char *filename)
{
FILE *file;
char command[PATH_MAX*2];
if (!filename)
return;
sprintf(command, "objdump --start-address=0x%016lx --stop-address=0x%016lx -dS %s", filter_start, filter_end, filename);
file = popen(command, "r");
if (!file)
return;
lines_tail = &lines;
while (!feof(file)) {
struct source_line *src;
size_t dummy = 0;
char *c;
src = malloc(sizeof(struct source_line));
assert(src != NULL);
memset(src, 0, sizeof(struct source_line));
if (getline(&src->line, &dummy, file) < 0)
break;
if (!src->line)
break;
c = strchr(src->line, '\n');
if (c)
*c = 0;
src->next = NULL;
*lines_tail = src;
lines_tail = &src->next;
if (strlen(src->line)>8 && src->line[8] == ':')
src->EIP = strtoull(src->line, NULL, 16);
if (strlen(src->line)>8 && src->line[16] == ':')
src->EIP = strtoull(src->line, NULL, 16);
}
pclose(file);
}
static void record_precise_ip(uint64_t ip)
{
struct source_line *line;
for (line = lines; line; line = line->next) {
if (line->EIP == ip)
line->count++;
if (line->EIP > ip)
break;
}
}
static void lookup_sym_in_vmlinux(struct sym_entry *sym)
{
struct source_line *line;
char pattern[PATH_MAX];
sprintf(pattern, "<%s>:", sym->sym);
for (line = lines; line; line = line->next) {
if (strstr(line->line, pattern)) {
sym->source = line;
break;
}
}
}
static void show_lines(struct source_line *line_queue, int line_queue_count)
{
int i;
struct source_line *line;
line = line_queue;
for (i = 0; i < line_queue_count; i++) {
printf("%8li\t%s\n", line->count, line->line);
line = line->next;
}
}
#define TRACE_COUNT 3
static void show_details(struct sym_entry *sym)
{
struct source_line *line;
struct source_line *line_queue = NULL;
int displayed = 0;
int line_queue_count = 0;
if (!sym->source)
lookup_sym_in_vmlinux(sym);
if (!sym->source)
return;
printf("Showing details for %s\n", sym->sym);
line = sym->source;
while (line) {
if (displayed && strstr(line->line, ">:"))
break;
if (!line_queue_count)
line_queue = line;
line_queue_count ++;
if (line->count >= count_filter) {
show_lines(line_queue, line_queue_count);
line_queue_count = 0;
line_queue = NULL;
} else if (line_queue_count > TRACE_COUNT) {
line_queue = line_queue->next;
line_queue_count --;
}
line->count = 0;
displayed++;
if (displayed > 300)
break;
line = line->next;
}
}
/*
* Binary search in the histogram table and record the hit:
*/
static void record_ip(uint64_t ip, int counter)
{
int left_idx, middle_idx, right_idx, idx;
unsigned long left, middle, right;
record_precise_ip(ip);
left_idx = 0;
right_idx = sym_table_count-1;
assert(ip <= max_ip && ip >= min_ip);
while (left_idx + 1 < right_idx) {
middle_idx = (left_idx + right_idx) / 2;
left = sym_table[ left_idx].addr;
middle = sym_table[middle_idx].addr;
right = sym_table[ right_idx].addr;
if (!(left <= middle && middle <= right)) {
printf("%016lx...\n%016lx...\n%016lx\n", left, middle, right);
printf("%d %d %d\n", left_idx, middle_idx, right_idx);
}
assert(left <= middle && middle <= right);
if (!(left <= ip && ip <= right)) {
printf(" left: %016lx\n", left);
printf(" ip: %016lx\n", (unsigned long)ip);
printf("right: %016lx\n", right);
}
assert(left <= ip && ip <= right);
/*
* [ left .... target .... middle .... right ]
* => right := middle
*/
if (ip < middle) {
right_idx = middle_idx;
continue;
}
/*
* [ left .... middle ... target ... right ]
* => left := middle
*/
left_idx = middle_idx;
}
idx = left_idx;
if (!sym_table[idx].skip)
sym_table[idx].count[counter]++;
else events--;
}
static void process_event(uint64_t ip, int counter)
{
events++;
if (ip < min_ip || ip > max_ip) {
userspace_events++;
return;
}
record_ip(ip, counter);
}
static void process_options(int argc, char *argv[])
static void process_options(int argc, char **argv)
{
int error = 0, counter;
if (strstr(argv[0], "perfstat"))
run_perfstat = 1;
for (;;) {
int option_index = 0;
/** Options for getopt */
......@@ -1095,8 +518,6 @@ static void process_options(int argc, char *argv[])
{"group", required_argument, NULL, 'g'},
{"help", no_argument, NULL, 'h'},
{"nmi", required_argument, NULL, 'n'},
{"mmap_info", no_argument, NULL, 'M'},
{"mmap_pages", required_argument, NULL, 'm'},
{"munmap_info", no_argument, NULL, 'U'},
{"pid", required_argument, NULL, 'p'},
{"realtime", required_argument, NULL, 'r'},
......@@ -1123,12 +544,9 @@ static void process_options(int argc, char *argv[])
tid = -1;
}
profile_cpu = atoi(optarg); break;
case 'd': delay_secs = atoi(optarg); break;
case 'D': dump_symtab = 1; break;
case 'e': error = parse_events(optarg); break;
case 'f': count_filter = atoi(optarg); break;
case 'g': group = atoi(optarg); break;
case 'h': display_help(); break;
case 'l': scale = 1; break;
......@@ -1141,14 +559,7 @@ static void process_options(int argc, char *argv[])
profile_cpu = -1;
}
tid = atoi(optarg); break;
case 'r': realtime_prio = atoi(optarg); break;
case 's': sym_filter = strdup(optarg); break;
case 'S': run_perfstat = 1; break;
case 'x': vmlinux = strdup(optarg); break;
case 'z': zero = 1; break;
case 'm': mmap_pages = atoi(optarg); break;
case 'M': use_mmap = 1; break;
case 'U': use_munmap = 1; break;
default: error = 1; break;
}
}
......@@ -1156,12 +567,7 @@ static void process_options(int argc, char *argv[])
display_help();
if (!nr_counters) {
if (run_perfstat)
nr_counters = 8;
else {
nr_counters = 1;
event_id[0] = 0;
}
nr_counters = 8;
}
for (counter = 0; counter < nr_counters; counter++) {
......@@ -1172,143 +578,8 @@ static void process_options(int argc, char *argv[])
}
}
struct mmap_data {
int counter;
void *base;
unsigned int mask;
unsigned int prev;
};
static unsigned int mmap_read_head(struct mmap_data *md)
{
struct perf_counter_mmap_page *pc = md->base;
int head;
head = pc->data_head;
rmb();
return head;
}
struct timeval last_read, this_read;
static void mmap_read(struct mmap_data *md)
{
unsigned int head = mmap_read_head(md);
unsigned int old = md->prev;
unsigned char *data = md->base + page_size;
int diff;
gettimeofday(&this_read, NULL);
/*
* If we're further behind than half the buffer, there's a chance
* the writer will bite our tail and screw up the events under us.
*
* If we somehow ended up ahead of the head, we got messed up.
*
* In either case, truncate and restart at head.
*/
diff = head - old;
if (diff > md->mask / 2 || diff < 0) {
struct timeval iv;
unsigned long msecs;
timersub(&this_read, &last_read, &iv);
msecs = iv.tv_sec*1000 + iv.tv_usec/1000;
fprintf(stderr, "WARNING: failed to keep up with mmap data."
" Last read %lu msecs ago.\n", msecs);
/*
* head points to a known good entry, start there.
*/
old = head;
}
last_read = this_read;
for (; old != head;) {
struct ip_event {
struct perf_event_header header;
__u64 ip;
__u32 pid, tid;
};
struct mmap_event {
struct perf_event_header header;
__u32 pid, tid;
__u64 start;
__u64 len;
__u64 pgoff;
char filename[PATH_MAX];
};
typedef union event_union {
struct perf_event_header header;
struct ip_event ip;
struct mmap_event mmap;
} event_t;
event_t *event = (event_t *)&data[old & md->mask];
event_t event_copy;
unsigned int size = event->header.size;
/*
* Event straddles the mmap boundary -- header should always
* be inside due to u64 alignment of output.
*/
if ((old & md->mask) + size != ((old + size) & md->mask)) {
unsigned int offset = old;
unsigned int len = min(sizeof(*event), size), cpy;
void *dst = &event_copy;
do {
cpy = min(md->mask + 1 - (offset & md->mask), len);
memcpy(dst, &data[offset & md->mask], cpy);
offset += cpy;
dst += cpy;
len -= cpy;
} while (len);
event = &event_copy;
}
old += size;
if (event->header.misc & PERF_EVENT_MISC_OVERFLOW) {
if (event->header.type & PERF_RECORD_IP)
process_event(event->ip.ip, md->counter);
} else {
switch (event->header.type) {
case PERF_EVENT_MMAP:
case PERF_EVENT_MUNMAP:
printf("%s: %Lu %Lu %Lu %s\n",
event->header.type == PERF_EVENT_MMAP
? "mmap" : "munmap",
event->mmap.start,
event->mmap.len,
event->mmap.pgoff,
event->mmap.filename);
break;
}
}
}
md->prev = old;
}
int main(int argc, char *argv[])
int cmd_stat(int argc, char **argv, const char *prefix)
{
struct pollfd event_array[MAX_NR_CPUS * MAX_COUNTERS];
struct mmap_data mmap_array[MAX_NR_CPUS][MAX_COUNTERS];
struct perf_counter_hw_event hw_event;
pthread_t thread;
int i, counter, group_fd, nr_poll = 0;
unsigned int cpu;
int ret;
page_size = sysconf(_SC_PAGE_SIZE);
process_options(argc, argv);
......@@ -1317,93 +588,5 @@ int main(int argc, char *argv[])
assert(nr_cpus <= MAX_NR_CPUS);
assert(nr_cpus >= 0);
if (run_perfstat)
return do_perfstat(argc, argv);
if (tid != -1 || profile_cpu != -1)
nr_cpus = 1;
parse_symbols();
if (vmlinux && sym_filter_entry)
parse_vmlinux(vmlinux);
for (i = 0; i < nr_cpus; i++) {
group_fd = -1;
for (counter = 0; counter < nr_counters; counter++) {
cpu = profile_cpu;
if (tid == -1 && profile_cpu == -1)
cpu = i;
memset(&hw_event, 0, sizeof(hw_event));
hw_event.config = event_id[counter];
hw_event.irq_period = event_count[counter];
hw_event.record_type = PERF_RECORD_IP | PERF_RECORD_TID;
hw_event.nmi = nmi;
hw_event.mmap = use_mmap;
hw_event.munmap = use_munmap;
fd[i][counter] = sys_perf_counter_open(&hw_event, tid, cpu, group_fd, 0);
if (fd[i][counter] < 0) {
int err = errno;
printf("kerneltop error: syscall returned with %d (%s)\n",
fd[i][counter], strerror(err));
if (err == EPERM)
printf("Are you root?\n");
exit(-1);
}
assert(fd[i][counter] >= 0);
fcntl(fd[i][counter], F_SETFL, O_NONBLOCK);
/*
* First counter acts as the group leader:
*/
if (group && group_fd == -1)
group_fd = fd[i][counter];
event_array[nr_poll].fd = fd[i][counter];
event_array[nr_poll].events = POLLIN;
nr_poll++;
mmap_array[i][counter].counter = counter;
mmap_array[i][counter].prev = 0;
mmap_array[i][counter].mask = mmap_pages*page_size - 1;
mmap_array[i][counter].base = mmap(NULL, (mmap_pages+1)*page_size,
PROT_READ, MAP_SHARED, fd[i][counter], 0);
if (mmap_array[i][counter].base == MAP_FAILED) {
printf("kerneltop error: failed to mmap with %d (%s)\n",
errno, strerror(errno));
exit(-1);
}
}
}
if (pthread_create(&thread, NULL, display_thread, NULL)) {
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, &param)) {
printf("Could not set realtime priority.\n");
exit(-1);
}
}
while (1) {
int hits = events;
for (i = 0; i < nr_cpus; i++) {
for (counter = 0; counter < nr_counters; counter++)
mmap_read(&mmap_array[i][counter]);
}
if (hits == events)
ret = poll(event_array, nr_poll, 100);
}
return 0;
return do_perfstat(argc, argv);
}
Documentation/perf_counter/builtin-top.c
View file @ ddcacfa0
......@@ -28,25 +28,6 @@
8.54 - ffffffff805001a3 : ip_queue_xmit
*/
/*
* perfstat: /usr/bin/time -alike performance counter statistics utility
It summarizes the counter events of all tasks (and child tasks),
covering all CPUs that the command (or workload) executes on.
It only counts the per-task events of the workload started,
independent of how many other tasks run on those CPUs.
Sample output:
$ ./perfstat -e 1 -e 3 -e 5 ls -lR /usr/include/ >/dev/null
Performance counter stats for 'ls':
163516953 instructions
2295 cache-misses
2855182 branch-misses
*/
/*
* Copyright (C) 2008, Red Hat Inc, Ingo Molnar <mingo@redhat.com>
*
......@@ -149,7 +130,6 @@ asmlinkage int sys_perf_counter_open(
#define EID(type, id) (((__u64)(type) << PERF_COUNTER_TYPE_SHIFT) | (id))
static int run_perfstat = 0;
static int system_wide = 0;
static int nr_counters = 0;
......@@ -203,7 +183,7 @@ struct source_line {
static struct source_line *lines;
static struct source_line **lines_tail;
const unsigned int default_count[] = {
static const unsigned int default_count[] = {
1000000,
1000000,
10000,
......@@ -291,26 +271,8 @@ static void display_events_help(void)
" rNNN: raw PMU events (eventsel+umask)\n\n");
}
static void display_perfstat_help(void)
{
printf(
"Usage: perfstat [<events...>] <cmd...>\n\n"
"PerfStat Options (up to %d event types can be specified):\n\n",
MAX_COUNTERS);
display_events_help();
printf(
" -l # scale counter values\n"
" -a # system-wide collection\n");
exit(0);
}
static void display_help(void)
{
if (run_perfstat)
return display_perfstat_help();
printf(
"Usage: kerneltop [<options>]\n"
" Or: kerneltop -S [<options>] COMMAND [ARGS]\n\n"
......@@ -320,8 +282,6 @@ static void display_help(void)
display_events_help();
printf(
" -S --stat # perfstat COMMAND\n"
" -a # system-wide collection (for perfstat)\n\n"
" -c CNT --count=CNT # event period to sample\n\n"
" -C CPU --cpu=CPU # CPU (-1 for all) [default: -1]\n"
" -p PID --pid=PID # PID of sampled task (-1 for all) [default: -1]\n\n"
......@@ -420,151 +380,6 @@ static int parse_events(char *str)
return 0;
}
/*
* perfstat
*/
char fault_here[1000000];
static void create_perfstat_counter(int counter)
{
struct perf_counter_hw_event hw_event;
memset(&hw_event, 0, sizeof(hw_event));
hw_event.config = event_id[counter];
hw_event.record_type = 0;
hw_event.nmi = 0;
if (scale)
hw_event.read_format = PERF_FORMAT_TOTAL_TIME_ENABLED |
PERF_FORMAT_TOTAL_TIME_RUNNING;
if (system_wide) {
int cpu;
for (cpu = 0; cpu < nr_cpus; cpu ++) {
fd[cpu][counter] = sys_perf_counter_open(&hw_event, -1, cpu, -1, 0);
if (fd[cpu][counter] < 0) {
printf("perfstat error: syscall returned with %d (%s)\n",
fd[cpu][counter], strerror(errno));
exit(-1);
}
}
} else {
hw_event.inherit = 1;
hw_event.disabled = 1;
fd[0][counter] = sys_perf_counter_open(&hw_event, 0, -1, -1, 0);
if (fd[0][counter] < 0) {
printf("perfstat error: syscall returned with %d (%s)\n",
fd[0][counter], strerror(errno));
exit(-1);
}
}
}
int do_perfstat(int argc, char *argv[])
{
unsigned long long t0, t1;
int counter;
ssize_t res;
int status;
int pid;
if (!system_wide)
nr_cpus = 1;
for (counter = 0; counter < nr_counters; counter++)
create_perfstat_counter(counter);
argc -= optind;
argv += optind;
if (!argc)
display_help();
/*
* Enable counters and exec the command:
*/
t0 = rdclock();
prctl(PR_TASK_PERF_COUNTERS_ENABLE);
if ((pid = fork()) < 0)
perror("failed to fork");
if (!pid) {
if (execvp(argv[0], argv)) {
perror(argv[0]);
exit(-1);
}
}
while (wait(&status) >= 0)
;
prctl(PR_TASK_PERF_COUNTERS_DISABLE);
t1 = rdclock();
fflush(stdout);
fprintf(stderr, "\n");
fprintf(stderr, " Performance counter stats for \'%s\':\n",
argv[0]);
fprintf(stderr, "\n");
for (counter = 0; counter < nr_counters; counter++) {
int cpu, nv;
__u64 count[3], single_count[3];
int scaled;
count[0] = count[1] = count[2] = 0;
nv = scale ? 3 : 1;
for (cpu = 0; cpu < nr_cpus; cpu ++) {
res = read(fd[cpu][counter],
single_count, nv * sizeof(__u64));
assert(res == nv * sizeof(__u64));
count[0] += single_count[0];
if (scale) {
count[1] += single_count[1];
count[2] += single_count[2];
}
}
scaled = 0;
if (scale) {
if (count[2] == 0) {
fprintf(stderr, " %14s %-20s\n",
"<not counted>", event_name(counter));
continue;
}
if (count[2] < count[1]) {
scaled = 1;
count[0] = (unsigned long long)
((double)count[0] * count[1] / count[2] + 0.5);
}
}
if (event_id[counter] == EID(PERF_TYPE_SOFTWARE, PERF_COUNT_CPU_CLOCK) ||
event_id[counter] == EID(PERF_TYPE_SOFTWARE, PERF_COUNT_TASK_CLOCK)) {
double msecs = (double)count[0] / 1000000;
fprintf(stderr, " %14.6f %-20s (msecs)",
msecs, event_name(counter));
} else {
fprintf(stderr, " %14Ld %-20s (events)",
count[0], event_name(counter));
}
if (scaled)
fprintf(stderr, " (scaled from %.2f%%)",
(double) count[2] / count[1] * 100);
fprintf(stderr, "\n");
}
fprintf(stderr, "\n");
fprintf(stderr, " Wall-clock time elapsed: %12.6f msecs\n",
(double)(t1-t0)/1e6);
fprintf(stderr, "\n");
return 0;
}
/*
* Symbols
*/
......@@ -805,7 +620,7 @@ static int read_symbol(FILE *in, struct sym_entry *s)
return 0;
}
int compare_addr(const void *__sym1, const void *__sym2)
static int compare_addr(const void *__sym1, const void *__sym2)
{
const struct sym_entry *sym1 = __sym1, *sym2 = __sym2;
......@@ -1070,9 +885,6 @@ static void process_options(int argc, char **argv)
{
int error = 0, counter;
if (strstr(argv[0], "perfstat"))
run_perfstat = 1;
for (;;) {
int option_index = 0;
/** Options for getopt */
......@@ -1134,7 +946,6 @@ static void process_options(int argc, char **argv)
tid = atoi(optarg); break;
case 'r': realtime_prio = atoi(optarg); break;
case 's': sym_filter = strdup(optarg); break;
case 'S': run_perfstat = 1; break;
case 'x': vmlinux = strdup(optarg); break;
case 'z': zero = 1; break;
case 'm': mmap_pages = atoi(optarg); break;
......@@ -1147,12 +958,8 @@ static void process_options(int argc, char **argv)
display_help();
if (!nr_counters) {
if (run_perfstat)
nr_counters = 8;
else {
nr_counters = 1;
event_id[0] = 0;
}
nr_counters = 1;
event_id[0] = 0;
}
for (counter = 0; counter < nr_counters; counter++) {
......@@ -1308,9 +1115,6 @@ int cmd_top(int argc, char **argv, const char *prefix)
assert(nr_cpus <= MAX_NR_CPUS);
assert(nr_cpus >= 0);
if (run_perfstat)
return do_perfstat(argc, argv);
if (tid != -1 || profile_cpu != -1)
nr_cpus = 1;
......
Documentation/perf_counter/builtin.h
View file @ ddcacfa0
......@@ -14,5 +14,6 @@ extern void prune_packed_objects(int);
extern int read_line_with_nul(char *buf, int size, FILE *file);
extern int check_pager_config(const char *cmd);
extern int cmd_top(int argc, const char **argv, const char *prefix);
extern int cmd_top(int argc, char **argv, const char *prefix);
extern int cmd_stat(int argc, char **argv, const char *prefix);
#endif
Documentation/perf_counter/command-list.txt
View file @ ddcacfa0
# List of known perf commands.
# command name category [deprecated] [common]
perf-top mainporcelain common
perf-stat mainporcelain common
Documentation/perf_counter/perf.c
View file @ ddcacfa0
......@@ -249,6 +249,7 @@ static void handle_internal_command(int argc, const char **argv)
const char *cmd = argv[0];
static struct cmd_struct commands[] = {
{ "top", cmd_top, 0 },
{ "stat", cmd_stat, 0 },
};
int i;
static const char ext[] = STRIP_EXTENSION;
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7