perf_event_open - 如何监控多个事件

Question

有谁知道如何设置perf_event_attr struct 可以触发PMU 通过perf_event_open() 监控多个（类型）事件？

和perf record -e cycles,faults ls一样，它有两种不同的事件类型（PERF_TYPE_HARDWARE和PERF_TYPE_SOFTWARE），但在perf_event_open's manpage的例子中，perf_event_attr.type只能分配一个值。

任何建议将不胜感激，谢谢！

20170208 更新 感谢@gudok 为我指明方向，但结果似乎有些异常。 演示程序如下（用于测量整个系统的 CPU 周期和缓存未命中）：

#define _GNU_SOURCE
#include <stdlib.h>
#include <stdio.h>
#include <unistd.h>
#include <sys/syscall.h>
#include <string.h>
#include <sys/ioctl.h>
#include <linux/perf_event.h>
#include <linux/hw_breakpoint.h>
#include <asm/unistd.h>
#include <errno.h>
#include <stdint.h>
#include <inttypes.h>
#include <time.h>

struct read_format {
  uint64_t nr;
  struct {
    uint64_t value;
    uint64_t id;
  } values[];
};

int main(int argc, char* argv[]) {
  struct perf_event_attr pea;
  int fd1, fd2;
  uint64_t id1, id2;
  uint64_t val1, val2;
  char buf[4096];
  struct read_format* rf = (struct read_format*) buf;
  int i,j;
  struct timespec time, time2;

  time.tv_sec = 1;
  time.tv_nsec = 0;

  memset(&pea, 0, sizeof(struct perf_event_attr));
  pea.type = PERF_TYPE_HARDWARE;
  pea.size = sizeof(struct perf_event_attr);
  pea.config = PERF_COUNT_HW_CPU_CYCLES;
  pea.disabled = 1;
  pea.exclude_kernel = 1;
  pea.exclude_hv = 1;
  pea.read_format = PERF_FORMAT_GROUP | PERF_FORMAT_ID;
  fd1 = syscall(__NR_perf_event_open, &pea, 0, -1, -1, 0);
  ioctl(fd1, PERF_EVENT_IOC_ID, &id1);

  memset(&pea, 0, sizeof(struct perf_event_attr));
  pea.type = PERF_TYPE_HARDWARE;
  pea.size = sizeof(struct perf_event_attr);
  pea.config = PERF_COUNT_HW_CACHE_MISSES;
  pea.disabled = 1;
  pea.exclude_kernel = 1;
  pea.exclude_hv = 1;
  pea.precise_ip = 2;  // want to using PEBS 
  pea.read_format = PERF_FORMAT_GROUP | PERF_FORMAT_ID;
  fd2 = syscall(__NR_perf_event_open, &pea, 0, -1, fd1 /*!!!*/, 0);
  ioctl(fd2, PERF_EVENT_IOC_ID, &id2);

  ioctl(fd1, PERF_EVENT_IOC_RESET, PERF_IOC_FLAG_GROUP);
  ioctl(fd1, PERF_EVENT_IOC_ENABLE, PERF_IOC_FLAG_GROUP);
  while (1) {
    nanosleep(&time, &time2);

    //ioctl(fd1, PERF_EVENT_IOC_DISABLE, PERF_IOC_FLAG_GROUP);

    read(fd1, buf, sizeof(buf));
    for (i = 0; i < rf->nr; i++) {
      if (rf->values[i].id == id1) {
        val1 = rf->values[i].value;
      } else if (rf->values[i].id == id2) {
        val2 = rf->values[i].value;
      }
    }

    printf("cpu cycles: %"PRIu64"\n", val1);
    printf("cache misses: %"PRIu64"\n", val2);

  }

  return 0;
}

输出是：

cpu cycles: 120   // Just have about 120 CPU cycles in a second
cache misses: 0   // and doesn't have any cache miss?
cpu cycles: 233
cache misses: 0
cpu cycles: 352
cache misses: 0
cpu cycles: 455
cache misses: 0
cpu cycles: 562
cache misses: 0
cpu cycles: 673
cache misses: 0
cpu cycles: 794
cache misses: 0
cpu cycles: 907
cache misses: 0
cpu cycles: 1011
cache misses: 0
cpu cycles: 1129
cache misses: 3
cpu cycles: 1269
cache misses: 4
cpu cycles: 1423

Answer 1

这有点棘手。

我们像往常一样创建第一个计数器。此外，我们通过PERF_FORMAT_GROUP 和PERF_FORMAT_ID 能够同时使用多个计数器。这个柜台将是我们的组长。

struct perf_event_attr pea;
int fd1, fd2;
uint64_t id1, id2; 

memset(&pea, 0, sizeof(struct perf_event_attr));
pea.type = PERF_TYPE_HARDWARE;
pea.size = sizeof(struct perf_event_attr);
pea.config = PERF_COUNT_HW_CPU_CYCLES;
pea.disabled = 1;
pea.exclude_kernel = 1;
pea.exclude_hv = 1;
pea.read_format = PERF_FORMAT_GROUP | PERF_FORMAT_ID;
fd1 = syscall(__NR_perf_event_open, &pea, 0, -1, -1, 0);

接下来，我们检索第一个计数器的标识符：

ioctl(fd1, PERF_EVENT_IOC_ID, &id1);

第二个（以及所有其他计数器）以相同的方式创建，只有一个例外：我们将 fd1 值作为组长参数传递：

memset(&pea, 0, sizeof(struct perf_event_attr));
pea.type = PERF_TYPE_SOFTWARE;
pea.size = sizeof(struct perf_event_attr);
pea.config = PERF_COUNT_SW_PAGE_FAULTS;
pea.disabled = 1;
pea.exclude_kernel = 1;
pea.exclude_hv = 1;
pea.read_format = PERF_FORMAT_GROUP | PERF_FORMAT_ID;
fd2 = syscall(__NR_perf_event_open, &pea, 0, -1, fd1, 0); // <-- here
ioctl(fd2, PERF_EVENT_IOC_ID, &id2);

接下来我们需要声明一个数据结构来一次读取多个计数器。您必须根据传递给perf_event_open 的标志来声明不同的字段集。手册页提到了所有可能的字段。在我们的例子中，我们传递了PERF_FORMAT_ID 标志，它添加了id 字段。这将使我们能够区分不同的计数器。

struct read_format {
    uint64_t nr;
    struct {
        uint64_t value;
        uint64_t id;
    } values[/*2*/];
};

现在我们调用标准分析 ioctls：

ioctl(fd1, PERF_EVENT_IOC_RESET, PERF_IOC_FLAG_GROUP);
ioctl(fd1, PERF_EVENT_IOC_ENABLE, PERF_IOC_FLAG_GROUP);
do_something();
ioctl(fd1, PERF_EVENT_IOC_DISABLE, PERF_IOC_FLAG_GROUP);

最后，我们从组长文件描述符中读取计数器。两个计数器都以我们声明的单个 read_format 结构返回：

char buf[4096];
struct read_format* rf = (struct read_format*) buf;
uint64_t val1, val2;

read(fd1, buf, sizeof(buf));
for (i = 0; i < rf->nr; i++) {
  if (rf->values[i].id == id1) {
    val1 = rf->values[i].value;
  } else if (rf->values[i].id == id2) {
    val2 = rf->values[i].value;
  }
}
printf("cpu cycles: %"PRIu64"\n", val1);
printf("page faults: %"PRIu64"\n", val2);

---

以下是完整的节目清单：

#define _GNU_SOURCE
#include <stdlib.h>
#include <stdio.h>
#include <unistd.h>
#include <sys/syscall.h>
#include <string.h>
#include <sys/ioctl.h>
#include <linux/perf_event.h>
#include <linux/hw_breakpoint.h>
#include <asm/unistd.h>
#include <errno.h>
#include <stdint.h>
#include <inttypes.h>

struct read_format {
  uint64_t nr;
  struct {
    uint64_t value;
    uint64_t id;
  } values[];
};

void do_something() {
  int i;
  char* ptr;

  ptr = malloc(100*1024*1024);
  for (i = 0; i < 100*1024*1024; i++) {
    ptr[i] = (char) (i & 0xff); // pagefault
  }
  free(ptr);
}

int main(int argc, char* argv[]) {
  struct perf_event_attr pea;
  int fd1, fd2;
  uint64_t id1, id2;
  uint64_t val1, val2;
  char buf[4096];
  struct read_format* rf = (struct read_format*) buf;
  int i;

  memset(&pea, 0, sizeof(struct perf_event_attr));
  pea.type = PERF_TYPE_HARDWARE;
  pea.size = sizeof(struct perf_event_attr);
  pea.config = PERF_COUNT_HW_CPU_CYCLES;
  pea.disabled = 1;
  pea.exclude_kernel = 1;
  pea.exclude_hv = 1;
  pea.read_format = PERF_FORMAT_GROUP | PERF_FORMAT_ID;
  fd1 = syscall(__NR_perf_event_open, &pea, 0, -1, -1, 0);
  ioctl(fd1, PERF_EVENT_IOC_ID, &id1);

  memset(&pea, 0, sizeof(struct perf_event_attr));
  pea.type = PERF_TYPE_SOFTWARE;
  pea.size = sizeof(struct perf_event_attr);
  pea.config = PERF_COUNT_SW_PAGE_FAULTS;
  pea.disabled = 1;
  pea.exclude_kernel = 1;
  pea.exclude_hv = 1;
  pea.read_format = PERF_FORMAT_GROUP | PERF_FORMAT_ID;
  fd2 = syscall(__NR_perf_event_open, &pea, 0, -1, fd1 /*!!!*/, 0);
  ioctl(fd2, PERF_EVENT_IOC_ID, &id2);


  ioctl(fd1, PERF_EVENT_IOC_RESET, PERF_IOC_FLAG_GROUP);
  ioctl(fd1, PERF_EVENT_IOC_ENABLE, PERF_IOC_FLAG_GROUP);
  do_something();
  ioctl(fd1, PERF_EVENT_IOC_DISABLE, PERF_IOC_FLAG_GROUP);


  read(fd1, buf, sizeof(buf));
  for (i = 0; i < rf->nr; i++) {
    if (rf->values[i].id == id1) {
      val1 = rf->values[i].value;
    } else if (rf->values[i].id == id2) {
      val2 = rf->values[i].value;
    }
  }

  printf("cpu cycles: %"PRIu64"\n", val1);
  printf("page faults: %"PRIu64"\n", val2);

  return 0;
}