]>
git.proxmox.com Git - ovs.git/blob - vswitchd/system-stats.c
1 /* Copyright (c) 2010, 2012, 2013, 2014 Nicira, Inc.
3 * Licensed under the Apache License, Version 2.0 (the "License");
4 * you may not use this file except in compliance with the License.
5 * You may obtain a copy of the License at:
7 * http://www.apache.org/licenses/LICENSE-2.0
9 * Unless required by applicable law or agreed to in writing, software
10 * distributed under the License is distributed on an "AS IS" BASIS,
11 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 * See the License for the specific language governing permissions and
13 * limitations under the License.
18 #include "system-stats.h"
29 #if HAVE_SYS_STATVFS_H
30 #include <sys/statvfs.h>
36 #include "dynamic-string.h"
40 #include "ovs-thread.h"
41 #include "poll-loop.h"
47 VLOG_DEFINE_THIS_MODULE(system_stats
);
49 /* #ifdefs make it a pain to maintain code: you have to try to build both ways.
50 * Thus, this file tries to compile as much of the code as possible regardless
51 * of the target, by writing "if (LINUX)" instead of "#ifdef __linux__" where
52 * this is possible. */
55 #include <asm/param.h>
61 get_cpu_cores(struct smap
*stats
)
63 long int n_cores
= count_cpu_cores();
65 smap_add_format(stats
, "cpu", "%ld", n_cores
);
70 get_load_average(struct smap
*stats OVS_UNUSED
)
75 if (getloadavg(loadavg
, 3) == 3) {
76 smap_add_format(stats
, "load_average", "%.2f,%.2f,%.2f",
77 loadavg
[0], loadavg
[1], loadavg
[2]);
85 static unsigned int cached
;
88 long int value
= sysconf(_SC_PAGESIZE
);
98 get_memory_stats(struct smap
*stats
)
101 unsigned int pagesize
= get_page_size();
102 #ifdef _SC_PHYS_PAGES
103 long int phys_pages
= sysconf(_SC_PHYS_PAGES
);
105 long int phys_pages
= 0;
107 #ifdef _SC_AVPHYS_PAGES
108 long int avphys_pages
= sysconf(_SC_AVPHYS_PAGES
);
110 long int avphys_pages
= 0;
112 int mem_total
, mem_used
;
114 if (pagesize
<= 0 || phys_pages
<= 0 || avphys_pages
<= 0) {
118 mem_total
= phys_pages
* (pagesize
/ 1024);
119 mem_used
= (phys_pages
- avphys_pages
) * (pagesize
/ 1024);
120 smap_add_format(stats
, "memory", "%d,%d", mem_total
, mem_used
);
122 static const char file_name
[] = "/proc/meminfo";
123 int mem_used
, mem_cache
, swap_used
;
134 stream
= fopen(file_name
, "r");
136 VLOG_WARN_ONCE("%s: open failed (%s)",
137 file_name
, ovs_strerror(errno
));
142 shash_add(&dict
, "MemTotal", &mem_total
);
143 shash_add(&dict
, "MemFree", &mem_free
);
144 shash_add(&dict
, "Buffers", &buffers
);
145 shash_add(&dict
, "Cached", &cached
);
146 shash_add(&dict
, "SwapTotal", &swap_total
);
147 shash_add(&dict
, "SwapFree", &swap_free
);
148 while (fgets(line
, sizeof line
, stream
)) {
152 if (ovs_scan(line
, "%15[^:]: %u", key
, &value
)) {
153 int *valuep
= shash_find_data(&dict
, key
);
160 shash_destroy(&dict
);
162 mem_used
= mem_total
- mem_free
;
163 mem_cache
= buffers
+ cached
;
164 swap_used
= swap_total
- swap_free
;
165 smap_add_format(stats
, "memory", "%d,%d,%d,%d,%d",
166 mem_total
, mem_used
, mem_cache
, swap_total
, swap_used
);
170 /* Returns the time at which the system booted, as the number of milliseconds
171 * since the epoch, or 0 if the time of boot cannot be determined. */
175 static long long int cache_expiration
= LLONG_MIN
;
176 static long long int boot_time
;
180 if (time_msec() >= cache_expiration
) {
181 static const char stat_file
[] = "/proc/stat";
185 cache_expiration
= time_msec() + 5 * 1000;
187 stream
= fopen(stat_file
, "r");
189 VLOG_ERR_ONCE("%s: open failed (%s)",
190 stat_file
, ovs_strerror(errno
));
194 while (fgets(line
, sizeof line
, stream
)) {
196 if (ovs_scan(line
, "btime %lld", &btime
)) {
197 boot_time
= btime
* 1000;
201 VLOG_ERR_ONCE("%s: btime not found", stat_file
);
208 static unsigned long long int
209 ticks_to_ms(unsigned long long int ticks
)
217 #if USER_HZ == 100 /* Common case. */
218 return ticks
* (1000 / USER_HZ
);
219 #else /* Alpha and some other architectures. */
220 double factor
= 1000.0 / USER_HZ
;
221 return ticks
* factor
+ 0.5;
225 struct raw_process_info
{
226 unsigned long int vsz
; /* Virtual size, in kB. */
227 unsigned long int rss
; /* Resident set size, in kB. */
228 long long int uptime
; /* ms since started. */
229 long long int cputime
; /* ms of CPU used during 'uptime'. */
230 pid_t ppid
; /* Parent. */
231 char name
[18]; /* Name (surrounded by parentheses). */
235 get_raw_process_info(pid_t pid
, struct raw_process_info
*raw
)
237 unsigned long long int vsize
, rss
, start_time
, utime
, stime
;
238 long long int start_msec
;
246 sprintf(file_name
, "/proc/%lu/stat", (unsigned long int) pid
);
247 stream
= fopen(file_name
, "r");
249 VLOG_ERR_ONCE("%s: open failed (%s)",
250 file_name
, ovs_strerror(errno
));
255 "%*d " /* (1. pid) */
256 "%17s " /* 2. process name */
257 "%*c " /* (3. state) */
259 "%*d " /* (5. pgid) */
260 "%*d " /* (6. sid) */
261 "%*d " /* (7. tty_nr) */
262 "%*d " /* (8. tty_pgrp) */
263 "%*u " /* (9. flags) */
264 "%*u " /* (10. min_flt) */
265 "%*u " /* (11. cmin_flt) */
266 "%*u " /* (12. maj_flt) */
267 "%*u " /* (13. cmaj_flt) */
268 "%llu " /* 14. utime */
269 "%llu " /* 15. stime */
270 "%*d " /* (16. cutime) */
271 "%*d " /* (17. cstime) */
272 "%*d " /* (18. priority) */
273 "%*d " /* (19. nice) */
274 "%*d " /* (20. num_threads) */
275 "%*d " /* (21. always 0) */
276 "%llu " /* 22. start_time */
277 "%llu " /* 23. vsize */
278 "%llu " /* 24. rss */
280 /* These are here for documentation but #if'd out to save
281 * actually parsing them from the stream for no benefit. */
282 "%*lu " /* (25. rsslim) */
283 "%*lu " /* (26. start_code) */
284 "%*lu " /* (27. end_code) */
285 "%*lu " /* (28. start_stack) */
286 "%*lu " /* (29. esp) */
287 "%*lu " /* (30. eip) */
288 "%*lu " /* (31. pending signals) */
289 "%*lu " /* (32. blocked signals) */
290 "%*lu " /* (33. ignored signals) */
291 "%*lu " /* (34. caught signals) */
292 "%*lu " /* (35. whcan) */
293 "%*lu " /* (36. always 0) */
294 "%*lu " /* (37. always 0) */
295 "%*d " /* (38. exit_signal) */
296 "%*d " /* (39. task_cpu) */
297 "%*u " /* (40. rt_priority) */
298 "%*u " /* (41. policy) */
299 "%*llu " /* (42. blkio_ticks) */
300 "%*lu " /* (43. gtime) */
301 "%*ld" /* (44. cgtime) */
303 , raw
->name
, &ppid
, &utime
, &stime
, &start_time
, &vsize
, &rss
);
306 VLOG_ERR_ONCE("%s: fscanf failed", file_name
);
310 start_msec
= get_boot_time() + ticks_to_ms(start_time
);
312 raw
->vsz
= vsize
/ 1024;
313 raw
->rss
= rss
* (getpagesize() / 1024);
314 raw
->uptime
= time_wall_msec() - start_msec
;
315 raw
->cputime
= ticks_to_ms(utime
+ stime
);
322 count_crashes(pid_t pid
)
332 sprintf(file_name
, "/proc/%lu/cmdline", (unsigned long int) pid
);
333 stream
= fopen(file_name
, "r");
335 VLOG_WARN_ONCE("%s: open failed (%s)", file_name
, ovs_strerror(errno
));
339 if (!fgets(line
, sizeof line
, stream
)) {
340 VLOG_WARN_ONCE("%s: read failed (%s)", file_name
,
341 feof(stream
) ? "end of file" : ovs_strerror(errno
));
345 paren
= strchr(line
, '(');
348 if (ovs_scan(paren
+ 1, "%d", &x
)) {
359 struct process_info
{
360 unsigned long int vsz
; /* Virtual size, in kB. */
361 unsigned long int rss
; /* Resident set size, in kB. */
362 long long int booted
; /* ms since monitor started. */
363 int crashes
; /* # of crashes (usually 0). */
364 long long int uptime
; /* ms since last (re)started by monitor. */
365 long long int cputime
; /* ms of CPU used during 'uptime'. */
369 get_process_info(pid_t pid
, struct process_info
*pinfo
)
371 struct raw_process_info child
;
374 if (!get_raw_process_info(pid
, &child
)) {
378 pinfo
->vsz
= child
.vsz
;
379 pinfo
->rss
= child
.rss
;
380 pinfo
->booted
= child
.uptime
;
382 pinfo
->uptime
= child
.uptime
;
383 pinfo
->cputime
= child
.cputime
;
386 struct raw_process_info parent
;
388 get_raw_process_info(child
.ppid
, &parent
);
389 if (!strcmp(child
.name
, parent
.name
)) {
390 pinfo
->booted
= parent
.uptime
;
391 pinfo
->crashes
= count_crashes(child
.ppid
);
399 get_process_stats(struct smap
*stats
)
404 dir
= opendir(ovs_rundir());
406 VLOG_ERR_ONCE("%s: open failed (%s)",
407 ovs_rundir(), ovs_strerror(errno
));
411 while ((de
= readdir(dir
)) != NULL
) {
412 struct process_info pinfo
;
418 #ifdef _DIRENT_HAVE_D_TYPE
419 if (de
->d_type
!= DT_UNKNOWN
&& de
->d_type
!= DT_REG
) {
424 extension
= strrchr(de
->d_name
, '.');
425 if (!extension
|| strcmp(extension
, ".pid")) {
429 file_name
= xasprintf("%s/%s", ovs_rundir(), de
->d_name
);
430 pid
= read_pidfile(file_name
);
436 key
= xasprintf("process_%.*s",
437 (int) (extension
- de
->d_name
), de
->d_name
);
438 if (!smap_get(stats
, key
)) {
439 if (LINUX
&& get_process_info(pid
, &pinfo
)) {
440 smap_add_format(stats
, key
, "%lu,%lu,%lld,%d,%lld,%lld",
441 pinfo
.vsz
, pinfo
.rss
, pinfo
.cputime
,
442 pinfo
.crashes
, pinfo
.booted
, pinfo
.uptime
);
444 smap_add(stats
, key
, "");
454 get_filesys_stats(struct smap
*stats OVS_UNUSED
)
456 #if HAVE_GETMNTENT_R && HAVE_STATVFS
457 static const char file_name
[] = "/etc/mtab";
458 struct mntent mntent
;
464 stream
= setmntent(file_name
, "r");
466 VLOG_ERR_ONCE("%s: open failed (%s)", file_name
, ovs_strerror(errno
));
471 while ((me
= getmntent_r(stream
, &mntent
, buf
, sizeof buf
)) != NULL
) {
472 unsigned long long int total
, free
;
476 /* Skip non-local and read-only filesystems. */
477 if (strncmp(me
->mnt_fsname
, "/dev", 4)
478 || !strstr(me
->mnt_opts
, "rw")) {
482 /* Given the mount point we can stat the file system. */
483 if (statvfs(me
->mnt_dir
, &vfs
) && vfs
.f_flag
& ST_RDONLY
) {
488 /* Now format the data. */
490 ds_put_char(&s
, ' ');
492 for (p
= me
->mnt_dir
; *p
!= '\0'; p
++) {
493 ds_put_char(&s
, *p
== ' ' || *p
== ',' ? '_' : *p
);
495 total
= (unsigned long long int) vfs
.f_frsize
* vfs
.f_blocks
/ 1024;
496 free
= (unsigned long long int) vfs
.f_frsize
* vfs
.f_bfree
/ 1024;
497 ds_put_format(&s
, ",%llu,%llu", total
, total
- free
);
502 smap_add(stats
, "file_systems", ds_cstr(&s
));
505 #endif /* HAVE_GETMNTENT_R && HAVE_STATVFS */
508 #define SYSTEM_STATS_INTERVAL (5 * 1000) /* In milliseconds. */
510 static struct ovs_mutex mutex
= OVS_MUTEX_INITIALIZER
;
511 static pthread_cond_t cond
= PTHREAD_COND_INITIALIZER
;
512 static struct latch latch
OVS_GUARDED_BY(mutex
);
514 static bool started
OVS_GUARDED_BY(mutex
);
515 static struct smap
*system_stats
OVS_GUARDED_BY(mutex
);
517 static void *system_stats_thread_func(void *);
518 static void discard_stats(void);
520 /* Enables or disables system stats collection, according to 'enable'. */
522 system_stats_enable(bool enable
)
524 if (enabled
!= enable
) {
525 ovs_mutex_lock(&mutex
);
528 xpthread_create(NULL
, NULL
, system_stats_thread_func
, NULL
);
533 xpthread_cond_signal(&cond
);
536 ovs_mutex_unlock(&mutex
);
540 /* Tries to obtain a new snapshot of system stats every SYSTEM_STATS_INTERVAL
543 * When a new snapshot is available (which only occurs if system stats are
544 * enabled), returns it as an smap owned by the caller. The caller must use
545 * both smap_destroy() and free() to completely free the returned data.
547 * When no new snapshot is available, returns NULL. */
549 system_stats_run(void)
551 struct smap
*stats
= NULL
;
553 ovs_mutex_lock(&mutex
);
558 stats
= system_stats
;
564 ovs_mutex_unlock(&mutex
);
569 /* Causes poll_block() to wake up when system_stats_run() needs to be
572 system_stats_wait(void)
580 discard_stats(void) OVS_REQUIRES(mutex
)
583 smap_destroy(system_stats
);
589 static void * NO_RETURN
590 system_stats_thread_func(void *arg OVS_UNUSED
)
592 pthread_detach(pthread_self());
595 long long int next_refresh
;
598 ovs_mutex_lock(&mutex
);
600 ovs_mutex_cond_wait(&cond
, &mutex
);
602 ovs_mutex_unlock(&mutex
);
604 stats
= xmalloc(sizeof *stats
);
606 get_cpu_cores(stats
);
607 get_load_average(stats
);
608 get_memory_stats(stats
);
609 get_process_stats(stats
);
610 get_filesys_stats(stats
);
612 ovs_mutex_lock(&mutex
);
614 system_stats
= stats
;
616 ovs_mutex_unlock(&mutex
);
618 next_refresh
= time_msec() + SYSTEM_STATS_INTERVAL
;
620 poll_timer_wait_until(next_refresh
);
622 } while (time_msec() < next_refresh
);