]> arthur.barton.de Git - netdata.git/blob - src/sys_fs_cgroup.c
disable by default cgroups starting with lxc/ and ending with /ns; fixes #1397
[netdata.git] / src / sys_fs_cgroup.c
1 #include "common.h"
2
3 // ----------------------------------------------------------------------------
4 // cgroup globals
5
6 static int cgroup_enable_cpuacct_stat = CONFIG_ONDEMAND_ONDEMAND;
7 static int cgroup_enable_cpuacct_usage = CONFIG_ONDEMAND_ONDEMAND;
8 static int cgroup_enable_memory = CONFIG_ONDEMAND_ONDEMAND;
9 static int cgroup_enable_devices = CONFIG_ONDEMAND_ONDEMAND;
10 static int cgroup_enable_blkio = CONFIG_ONDEMAND_ONDEMAND;
11 static int cgroup_enable_new_cgroups_detected_at_runtime = 1;
12 static int cgroup_check_for_new_every = 10;
13 static char *cgroup_cpuacct_base = NULL;
14 static char *cgroup_blkio_base = NULL;
15 static char *cgroup_memory_base = NULL;
16 static char *cgroup_devices_base = NULL;
17
18 static int cgroup_root_count = 0;
19 static int cgroup_root_max = 500;
20 static int cgroup_max_depth = 0;
21
22 void read_cgroup_plugin_configuration() {
23     cgroup_check_for_new_every = config_get_number("plugin:cgroups", "check for new cgroups every", cgroup_check_for_new_every);
24
25     cgroup_enable_cpuacct_stat = config_get_boolean_ondemand("plugin:cgroups", "enable cpuacct stat", cgroup_enable_cpuacct_stat);
26     cgroup_enable_cpuacct_usage = config_get_boolean_ondemand("plugin:cgroups", "enable cpuacct usage", cgroup_enable_cpuacct_usage);
27     cgroup_enable_memory = config_get_boolean_ondemand("plugin:cgroups", "enable memory", cgroup_enable_memory);
28     cgroup_enable_blkio = config_get_boolean_ondemand("plugin:cgroups", "enable blkio", cgroup_enable_blkio);
29
30     char filename[FILENAME_MAX + 1], *s;
31     struct mountinfo *mi, *root = mountinfo_read();
32
33     mi = mountinfo_find_by_filesystem_super_option(root, "cgroup", "cpuacct");
34     if(!mi) mi = mountinfo_find_by_filesystem_mount_source(root, "cgroup", "cpuacct");
35     if(!mi) {
36         error("Cannot find cgroup cpuacct mountinfo. Assuming default: /sys/fs/cgroup/cpuacct");
37         s = "/sys/fs/cgroup/cpuacct";
38     }
39     else s = mi->mount_point;
40     snprintfz(filename, FILENAME_MAX, "%s%s", global_host_prefix, s);
41     cgroup_cpuacct_base = config_get("plugin:cgroups", "path to /sys/fs/cgroup/cpuacct", filename);
42
43     mi = mountinfo_find_by_filesystem_super_option(root, "cgroup", "blkio");
44     if(!mi) mi = mountinfo_find_by_filesystem_mount_source(root, "cgroup", "blkio");
45     if(!mi) {
46         error("Cannot find cgroup blkio mountinfo. Assuming default: /sys/fs/cgroup/blkio");
47         s = "/sys/fs/cgroup/blkio";
48     }
49     else s = mi->mount_point;
50     snprintfz(filename, FILENAME_MAX, "%s%s", global_host_prefix, s);
51     cgroup_blkio_base = config_get("plugin:cgroups", "path to /sys/fs/cgroup/blkio", filename);
52
53     mi = mountinfo_find_by_filesystem_super_option(root, "cgroup", "memory");
54     if(!mi) mi = mountinfo_find_by_filesystem_mount_source(root, "cgroup", "memory");
55     if(!mi) {
56         error("Cannot find cgroup memory mountinfo. Assuming default: /sys/fs/cgroup/memory");
57         s = "/sys/fs/cgroup/memory";
58     }
59     else s = mi->mount_point;
60     snprintfz(filename, FILENAME_MAX, "%s%s", global_host_prefix, s);
61     cgroup_memory_base = config_get("plugin:cgroups", "path to /sys/fs/cgroup/memory", filename);
62
63     mi = mountinfo_find_by_filesystem_super_option(root, "cgroup", "devices");
64     if(!mi) mi = mountinfo_find_by_filesystem_mount_source(root, "cgroup", "devices");
65     if(!mi) {
66         error("Cannot find cgroup devices mountinfo. Assuming default: /sys/fs/cgroup/devices");
67         s = "/sys/fs/cgroup/devices";
68     }
69     else s = mi->mount_point;
70     snprintfz(filename, FILENAME_MAX, "%s%s", global_host_prefix, s);
71     cgroup_devices_base = config_get("plugin:cgroups", "path to /sys/fs/cgroup/devices", filename);
72
73     cgroup_root_max = config_get_number("plugin:cgroups", "max cgroups to allow", cgroup_root_max);
74     cgroup_max_depth = config_get_number("plugin:cgroups", "max cgroups depth to monitor", cgroup_max_depth);
75
76     cgroup_enable_new_cgroups_detected_at_runtime = config_get_boolean("plugin:cgroups", "enable new cgroups detected at run time", cgroup_enable_new_cgroups_detected_at_runtime);
77
78     mountinfo_free(root);
79 }
80
81 // ----------------------------------------------------------------------------
82 // cgroup objects
83
84 struct blkio {
85     int updated;
86
87     char *filename;
88
89     unsigned long long Read;
90     unsigned long long Write;
91 /*
92     unsigned long long Sync;
93     unsigned long long Async;
94     unsigned long long Total;
95 */
96 };
97
98 // https://www.kernel.org/doc/Documentation/cgroup-v1/memory.txt
99 struct memory {
100     int updated;
101
102     char *filename;
103
104     int has_dirty_swap;
105
106     unsigned long long cache;
107     unsigned long long rss;
108     unsigned long long rss_huge;
109     unsigned long long mapped_file;
110     unsigned long long writeback;
111     unsigned long long dirty;
112     unsigned long long swap;
113     unsigned long long pgpgin;
114     unsigned long long pgpgout;
115     unsigned long long pgfault;
116     unsigned long long pgmajfault;
117 /*
118     unsigned long long inactive_anon;
119     unsigned long long active_anon;
120     unsigned long long inactive_file;
121     unsigned long long active_file;
122     unsigned long long unevictable;
123     unsigned long long hierarchical_memory_limit;
124     unsigned long long total_cache;
125     unsigned long long total_rss;
126     unsigned long long total_rss_huge;
127     unsigned long long total_mapped_file;
128     unsigned long long total_writeback;
129     unsigned long long total_dirty;
130     unsigned long long total_swap;
131     unsigned long long total_pgpgin;
132     unsigned long long total_pgpgout;
133     unsigned long long total_pgfault;
134     unsigned long long total_pgmajfault;
135     unsigned long long total_inactive_anon;
136     unsigned long long total_active_anon;
137     unsigned long long total_inactive_file;
138     unsigned long long total_active_file;
139     unsigned long long total_unevictable;
140 */
141
142     int usage_in_bytes_updated;
143     char *filename_usage_in_bytes;
144     unsigned long long usage_in_bytes;
145
146     int msw_usage_in_bytes_updated;
147     char *filename_msw_usage_in_bytes;
148     unsigned long long msw_usage_in_bytes;
149
150     int failcnt_updated;
151     char *filename_failcnt;
152     unsigned long long failcnt;
153 };
154
155 // https://www.kernel.org/doc/Documentation/cgroup-v1/cpuacct.txt
156 struct cpuacct_stat {
157     int updated;
158
159     char *filename;
160
161     unsigned long long user;
162     unsigned long long system;
163 };
164
165 // https://www.kernel.org/doc/Documentation/cgroup-v1/cpuacct.txt
166 struct cpuacct_usage {
167     int updated;
168
169     char *filename;
170
171     unsigned int cpus;
172     unsigned long long *cpu_percpu;
173 };
174
175 #define CGROUP_OPTIONS_DISABLED_DUPLICATE 0x00000001
176
177 struct cgroup {
178     uint32_t options;
179
180     char available;      // found in the filesystem
181     char enabled;        // enabled in the config
182
183     char *id;
184     uint32_t hash;
185
186     char *chart_id;
187     uint32_t hash_chart;
188
189     char *chart_title;
190
191     struct cpuacct_stat cpuacct_stat;
192     struct cpuacct_usage cpuacct_usage;
193
194     struct memory memory;
195
196     struct blkio io_service_bytes;              // bytes
197     struct blkio io_serviced;                   // operations
198
199     struct blkio throttle_io_service_bytes;     // bytes
200     struct blkio throttle_io_serviced;          // operations
201
202     struct blkio io_merged;                     // operations
203     struct blkio io_queued;                     // operations
204
205     struct cgroup *next;
206
207 } *cgroup_root = NULL;
208
209 // ----------------------------------------------------------------------------
210 // read values from /sys
211
212 void cgroup_read_cpuacct_stat(struct cpuacct_stat *cp) {
213     static procfile *ff = NULL;
214
215     static uint32_t user_hash = 0;
216     static uint32_t system_hash = 0;
217
218     if(unlikely(user_hash == 0)) {
219         user_hash = simple_hash("user");
220         system_hash = simple_hash("system");
221     }
222
223     cp->updated = 0;
224     if(cp->filename) {
225         ff = procfile_reopen(ff, cp->filename, NULL, PROCFILE_FLAG_DEFAULT);
226         if(!ff) return;
227
228         ff = procfile_readall(ff);
229         if(!ff) return;
230
231         unsigned long i, lines = procfile_lines(ff);
232
233         if(lines < 1) {
234             error("File '%s' should have 1+ lines.", cp->filename);
235             return;
236         }
237
238         for(i = 0; i < lines ; i++) {
239             char *s = procfile_lineword(ff, i, 0);
240             uint32_t hash = simple_hash(s);
241
242             if(hash == user_hash && !strcmp(s, "user"))
243                 cp->user = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
244
245             else if(hash == system_hash && !strcmp(s, "system"))
246                 cp->system = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
247         }
248
249         cp->updated = 1;
250
251         // fprintf(stderr, "READ '%s': user: %llu, system: %llu\n", cp->filename, cp->user, cp->system);
252     }
253 }
254
255 void cgroup_read_cpuacct_usage(struct cpuacct_usage *ca) {
256     static procfile *ff = NULL;
257
258     ca->updated = 0;
259     if(ca->filename) {
260         ff = procfile_reopen(ff, ca->filename, NULL, PROCFILE_FLAG_DEFAULT);
261         if(!ff) return;
262
263         ff = procfile_readall(ff);
264         if(!ff) return;
265
266         if(procfile_lines(ff) < 1) {
267             error("File '%s' should have 1+ lines but has %u.", ca->filename, procfile_lines(ff));
268             return;
269         }
270
271         unsigned long i = procfile_linewords(ff, 0);
272         if(i == 0) return;
273
274         // we may have 1 more CPU reported
275         while(i > 0) {
276             char *s = procfile_lineword(ff, 0, i - 1);
277             if(!*s) i--;
278             else break;
279         }
280
281         if(i != ca->cpus) {
282             freez(ca->cpu_percpu);
283             ca->cpu_percpu = mallocz(sizeof(unsigned long long) * i);
284             ca->cpus = (unsigned int)i;
285         }
286
287         for(i = 0; i < ca->cpus ;i++) {
288             ca->cpu_percpu[i] = strtoull(procfile_lineword(ff, 0, i), NULL, 10);
289             // fprintf(stderr, "READ '%s': cpu%d/%d: %llu ('%s')\n", ca->filename, i, ca->cpus, ca->cpu_percpu[i], procfile_lineword(ff, 0, i));
290         }
291
292         ca->updated = 1;
293     }
294 }
295
296 void cgroup_read_blkio(struct blkio *io) {
297     static procfile *ff = NULL;
298
299     static uint32_t Read_hash = 0;
300     static uint32_t Write_hash = 0;
301 /*
302     static uint32_t Sync_hash = 0;
303     static uint32_t Async_hash = 0;
304     static uint32_t Total_hash = 0;
305 */
306
307     if(unlikely(Read_hash == 0)) {
308         Read_hash = simple_hash("Read");
309         Write_hash = simple_hash("Write");
310 /*
311         Sync_hash = simple_hash("Sync");
312         Async_hash = simple_hash("Async");
313         Total_hash = simple_hash("Total");
314 */
315     }
316
317     io->updated = 0;
318     if(io->filename) {
319         ff = procfile_reopen(ff, io->filename, NULL, PROCFILE_FLAG_DEFAULT);
320         if(!ff) return;
321
322         ff = procfile_readall(ff);
323         if(!ff) return;
324
325         unsigned long i, lines = procfile_lines(ff);
326
327         if(lines < 1) {
328             error("File '%s' should have 1+ lines.", io->filename);
329             return;
330         }
331
332         io->Read = 0;
333         io->Write = 0;
334 /*
335         io->Sync = 0;
336         io->Async = 0;
337         io->Total = 0;
338 */
339
340         for(i = 0; i < lines ; i++) {
341             char *s = procfile_lineword(ff, i, 1);
342             uint32_t hash = simple_hash(s);
343
344             if(hash == Read_hash && !strcmp(s, "Read"))
345                 io->Read += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
346
347             else if(hash == Write_hash && !strcmp(s, "Write"))
348                 io->Write += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
349
350 /*
351             else if(hash == Sync_hash && !strcmp(s, "Sync"))
352                 io->Sync += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
353
354             else if(hash == Async_hash && !strcmp(s, "Async"))
355                 io->Async += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
356
357             else if(hash == Total_hash && !strcmp(s, "Total"))
358                 io->Total += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
359 */
360         }
361
362         io->updated = 1;
363         // fprintf(stderr, "READ '%s': Read: %llu, Write: %llu, Sync: %llu, Async: %llu, Total: %llu\n", io->filename, io->Read, io->Write, io->Sync, io->Async, io->Total);
364     }
365 }
366
367 void cgroup_read_memory(struct memory *mem) {
368     static procfile *ff = NULL;
369
370     static uint32_t cache_hash = 0;
371     static uint32_t rss_hash = 0;
372     static uint32_t rss_huge_hash = 0;
373     static uint32_t mapped_file_hash = 0;
374     static uint32_t writeback_hash = 0;
375     static uint32_t dirty_hash = 0;
376     static uint32_t swap_hash = 0;
377     static uint32_t pgpgin_hash = 0;
378     static uint32_t pgpgout_hash = 0;
379     static uint32_t pgfault_hash = 0;
380     static uint32_t pgmajfault_hash = 0;
381 /*
382     static uint32_t inactive_anon_hash = 0;
383     static uint32_t active_anon_hash = 0;
384     static uint32_t inactive_file_hash = 0;
385     static uint32_t active_file_hash = 0;
386     static uint32_t unevictable_hash = 0;
387     static uint32_t hierarchical_memory_limit_hash = 0;
388     static uint32_t total_cache_hash = 0;
389     static uint32_t total_rss_hash = 0;
390     static uint32_t total_rss_huge_hash = 0;
391     static uint32_t total_mapped_file_hash = 0;
392     static uint32_t total_writeback_hash = 0;
393     static uint32_t total_dirty_hash = 0;
394     static uint32_t total_swap_hash = 0;
395     static uint32_t total_pgpgin_hash = 0;
396     static uint32_t total_pgpgout_hash = 0;
397     static uint32_t total_pgfault_hash = 0;
398     static uint32_t total_pgmajfault_hash = 0;
399     static uint32_t total_inactive_anon_hash = 0;
400     static uint32_t total_active_anon_hash = 0;
401     static uint32_t total_inactive_file_hash = 0;
402     static uint32_t total_active_file_hash = 0;
403     static uint32_t total_unevictable_hash = 0;
404 */
405     if(unlikely(cache_hash == 0)) {
406         cache_hash = simple_hash("cache");
407         rss_hash = simple_hash("rss");
408         rss_huge_hash = simple_hash("rss_huge");
409         mapped_file_hash = simple_hash("mapped_file");
410         writeback_hash = simple_hash("writeback");
411         dirty_hash = simple_hash("dirty");
412         swap_hash = simple_hash("swap");
413         pgpgin_hash = simple_hash("pgpgin");
414         pgpgout_hash = simple_hash("pgpgout");
415         pgfault_hash = simple_hash("pgfault");
416         pgmajfault_hash = simple_hash("pgmajfault");
417 /*
418         inactive_anon_hash = simple_hash("inactive_anon");
419         active_anon_hash = simple_hash("active_anon");
420         inactive_file_hash = simple_hash("inactive_file");
421         active_file_hash = simple_hash("active_file");
422         unevictable_hash = simple_hash("unevictable");
423         hierarchical_memory_limit_hash = simple_hash("hierarchical_memory_limit");
424         total_cache_hash = simple_hash("total_cache");
425         total_rss_hash = simple_hash("total_rss");
426         total_rss_huge_hash = simple_hash("total_rss_huge");
427         total_mapped_file_hash = simple_hash("total_mapped_file");
428         total_writeback_hash = simple_hash("total_writeback");
429         total_dirty_hash = simple_hash("total_dirty");
430         total_swap_hash = simple_hash("total_swap");
431         total_pgpgin_hash = simple_hash("total_pgpgin");
432         total_pgpgout_hash = simple_hash("total_pgpgout");
433         total_pgfault_hash = simple_hash("total_pgfault");
434         total_pgmajfault_hash = simple_hash("total_pgmajfault");
435         total_inactive_anon_hash = simple_hash("total_inactive_anon");
436         total_active_anon_hash = simple_hash("total_active_anon");
437         total_inactive_file_hash = simple_hash("total_inactive_file");
438         total_active_file_hash = simple_hash("total_active_file");
439         total_unevictable_hash = simple_hash("total_unevictable");
440 */
441     }
442
443     mem->updated = 0;
444     if(mem->filename) {
445         ff = procfile_reopen(ff, mem->filename, NULL, PROCFILE_FLAG_DEFAULT);
446         if(!ff) return;
447
448         ff = procfile_readall(ff);
449         if(!ff) return;
450
451         unsigned long i, lines = procfile_lines(ff);
452
453         if(lines < 1) {
454             error("File '%s' should have 1+ lines.", mem->filename);
455             return;
456         }
457
458         for(i = 0; i < lines ; i++) {
459             char *s = procfile_lineword(ff, i, 0);
460             uint32_t hash = simple_hash(s);
461
462             if(hash == cache_hash && !strcmp(s, "cache"))
463                 mem->cache = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
464
465             else if(hash == rss_hash && !strcmp(s, "rss"))
466                 mem->rss = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
467
468             else if(hash == rss_huge_hash && !strcmp(s, "rss_huge"))
469                 mem->rss_huge = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
470
471             else if(hash == mapped_file_hash && !strcmp(s, "mapped_file"))
472                 mem->mapped_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
473
474             else if(hash == writeback_hash && !strcmp(s, "writeback"))
475                 mem->writeback = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
476
477             else if(hash == dirty_hash && !strcmp(s, "dirty")) {
478                 mem->dirty = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
479                 mem->has_dirty_swap = 1;
480             }
481
482             else if(hash == swap_hash && !strcmp(s, "swap")) {
483                 mem->swap = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
484                 mem->has_dirty_swap = 1;
485             }
486
487             else if(hash == pgpgin_hash && !strcmp(s, "pgpgin"))
488                 mem->pgpgin = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
489
490             else if(hash == pgpgout_hash && !strcmp(s, "pgpgout"))
491                 mem->pgpgout = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
492
493             else if(hash == pgfault_hash && !strcmp(s, "pgfault"))
494                 mem->pgfault = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
495
496             else if(hash == pgmajfault_hash && !strcmp(s, "pgmajfault"))
497                 mem->pgmajfault = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
498
499 /*
500             else if(hash == inactive_anon_hash && !strcmp(s, "inactive_anon"))
501                 mem->inactive_anon = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
502
503             else if(hash == active_anon_hash && !strcmp(s, "active_anon"))
504                 mem->active_anon = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
505
506             else if(hash == inactive_file_hash && !strcmp(s, "inactive_file"))
507                 mem->inactive_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
508
509             else if(hash == active_file_hash && !strcmp(s, "active_file"))
510                 mem->active_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
511
512             else if(hash == unevictable_hash && !strcmp(s, "unevictable"))
513                 mem->unevictable = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
514
515             else if(hash == hierarchical_memory_limit_hash && !strcmp(s, "hierarchical_memory_limit"))
516                 mem->hierarchical_memory_limit = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
517
518             else if(hash == total_cache_hash && !strcmp(s, "total_cache"))
519                 mem->total_cache = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
520
521             else if(hash == total_rss_hash && !strcmp(s, "total_rss"))
522                 mem->total_rss = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
523
524             else if(hash == total_rss_huge_hash && !strcmp(s, "total_rss_huge"))
525                 mem->total_rss_huge = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
526
527             else if(hash == total_mapped_file_hash && !strcmp(s, "total_mapped_file"))
528                 mem->total_mapped_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
529
530             else if(hash == total_writeback_hash && !strcmp(s, "total_writeback"))
531                 mem->total_writeback = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
532
533             else if(hash == total_dirty_hash && !strcmp(s, "total_dirty"))
534                 mem->total_dirty = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
535
536             else if(hash == total_swap_hash && !strcmp(s, "total_swap"))
537                 mem->total_swap = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
538
539             else if(hash == total_pgpgin_hash && !strcmp(s, "total_pgpgin"))
540                 mem->total_pgpgin = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
541
542             else if(hash == total_pgpgout_hash && !strcmp(s, "total_pgpgout"))
543                 mem->total_pgpgout = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
544
545             else if(hash == total_pgfault_hash && !strcmp(s, "total_pgfault"))
546                 mem->total_pgfault = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
547
548             else if(hash == total_pgmajfault_hash && !strcmp(s, "total_pgmajfault"))
549                 mem->total_pgmajfault = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
550
551             else if(hash == total_inactive_anon_hash && !strcmp(s, "total_inactive_anon"))
552                 mem->total_inactive_anon = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
553
554             else if(hash == total_active_anon_hash && !strcmp(s, "total_active_anon"))
555                 mem->total_active_anon = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
556
557             else if(hash == total_inactive_file_hash && !strcmp(s, "total_inactive_file"))
558                 mem->total_inactive_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
559
560             else if(hash == total_active_file_hash && !strcmp(s, "total_active_file"))
561                 mem->total_active_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
562
563             else if(hash == total_unevictable_hash && !strcmp(s, "total_unevictable"))
564                 mem->total_unevictable = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
565 */
566         }
567
568         // fprintf(stderr, "READ: '%s', cache: %llu, rss: %llu, rss_huge: %llu, mapped_file: %llu, writeback: %llu, dirty: %llu, swap: %llu, pgpgin: %llu, pgpgout: %llu, pgfault: %llu, pgmajfault: %llu, inactive_anon: %llu, active_anon: %llu, inactive_file: %llu, active_file: %llu, unevictable: %llu, hierarchical_memory_limit: %llu, total_cache: %llu, total_rss: %llu, total_rss_huge: %llu, total_mapped_file: %llu, total_writeback: %llu, total_dirty: %llu, total_swap: %llu, total_pgpgin: %llu, total_pgpgout: %llu, total_pgfault: %llu, total_pgmajfault: %llu, total_inactive_anon: %llu, total_active_anon: %llu, total_inactive_file: %llu, total_active_file: %llu, total_unevictable: %llu\n", mem->filename, mem->cache, mem->rss, mem->rss_huge, mem->mapped_file, mem->writeback, mem->dirty, mem->swap, mem->pgpgin, mem->pgpgout, mem->pgfault, mem->pgmajfault, mem->inactive_anon, mem->active_anon, mem->inactive_file, mem->active_file, mem->unevictable, mem->hierarchical_memory_limit, mem->total_cache, mem->total_rss, mem->total_rss_huge, mem->total_mapped_file, mem->total_writeback, mem->total_dirty, mem->total_swap, mem->total_pgpgin, mem->total_pgpgout, mem->total_pgfault, mem->total_pgmajfault, mem->total_inactive_anon, mem->total_active_anon, mem->total_inactive_file, mem->total_active_file, mem->total_unevictable);
569
570         mem->updated = 1;
571     }
572
573     mem->usage_in_bytes_updated = 0;
574     if(mem->filename_usage_in_bytes) {
575         if(likely(!read_single_number_file(mem->filename_usage_in_bytes, &mem->usage_in_bytes)))
576             mem->usage_in_bytes_updated = 1;
577     }
578
579     mem->msw_usage_in_bytes_updated = 0;
580     if(mem->filename_msw_usage_in_bytes) {
581         if(likely(!read_single_number_file(mem->filename_msw_usage_in_bytes, &mem->msw_usage_in_bytes)))
582             mem->msw_usage_in_bytes_updated = 1;
583     }
584
585     mem->failcnt_updated = 0;
586     if(mem->filename_failcnt) {
587         if(likely(!read_single_number_file(mem->filename_failcnt, &mem->failcnt)))
588             mem->failcnt_updated = 1;
589     }
590 }
591
592 void cgroup_read(struct cgroup *cg) {
593     debug(D_CGROUP, "reading metrics for cgroups '%s'", cg->id);
594
595     cgroup_read_cpuacct_stat(&cg->cpuacct_stat);
596     cgroup_read_cpuacct_usage(&cg->cpuacct_usage);
597     cgroup_read_memory(&cg->memory);
598     cgroup_read_blkio(&cg->io_service_bytes);
599     cgroup_read_blkio(&cg->io_serviced);
600     cgroup_read_blkio(&cg->throttle_io_service_bytes);
601     cgroup_read_blkio(&cg->throttle_io_serviced);
602     cgroup_read_blkio(&cg->io_merged);
603     cgroup_read_blkio(&cg->io_queued);
604 }
605
606 void read_all_cgroups(struct cgroup *root) {
607     debug(D_CGROUP, "reading metrics for all cgroups");
608
609     struct cgroup *cg;
610
611     for(cg = root; cg ; cg = cg->next)
612         if(cg->enabled && cg->available)
613             cgroup_read(cg);
614 }
615
616 // ----------------------------------------------------------------------------
617 // add/remove/find cgroup objects
618
619 #define CGROUP_CHARTID_LINE_MAX 1024
620
621 void cgroup_get_chart_id(struct cgroup *cg) {
622     debug(D_CGROUP, "getting the name of cgroup '%s'", cg->id);
623
624     pid_t cgroup_pid;
625     char buffer[CGROUP_CHARTID_LINE_MAX + 1];
626
627     snprintfz(buffer, CGROUP_CHARTID_LINE_MAX, "exec %s '%s'",
628              config_get("plugin:cgroups", "script to get cgroup names", PLUGINS_DIR "/cgroup-name.sh"), cg->chart_id);
629
630     debug(D_CGROUP, "executing command '%s' for cgroup '%s'", buffer, cg->id);
631     FILE *fp = mypopen(buffer, &cgroup_pid);
632     if(!fp) {
633         error("CGROUP: Cannot popen(\"%s\", \"r\").", buffer);
634         return;
635     }
636     debug(D_CGROUP, "reading from command '%s' for cgroup '%s'", buffer, cg->id);
637     char *s = fgets(buffer, CGROUP_CHARTID_LINE_MAX, fp);
638     debug(D_CGROUP, "closing command for cgroup '%s'", cg->id);
639     mypclose(fp, cgroup_pid);
640     debug(D_CGROUP, "closed command for cgroup '%s'", cg->id);
641
642     if(s && *s && *s != '\n') {
643         debug(D_CGROUP, "cgroup '%s' should be renamed to '%s'", cg->id, s);
644
645         trim(s);
646
647         freez(cg->chart_title);
648         cg->chart_title = strdupz(s);
649         netdata_fix_chart_name(cg->chart_title);
650
651         freez(cg->chart_id);
652         cg->chart_id = strdupz(s);
653         netdata_fix_chart_id(cg->chart_id);
654         cg->hash_chart = simple_hash(cg->chart_id);
655
656         debug(D_CGROUP, "cgroup '%s' renamed to '%s' (title: '%s')", cg->id, cg->chart_id, cg->chart_title);
657     }
658     else debug(D_CGROUP, "cgroup '%s' is not to be renamed (will be shown as '%s')", cg->id, cg->chart_id);
659 }
660
661 struct cgroup *cgroup_add(const char *id) {
662     debug(D_CGROUP, "adding cgroup '%s'", id);
663
664     if(cgroup_root_count >= cgroup_root_max) {
665         info("Maximum number of cgroups reached (%d). Not adding cgroup '%s'", cgroup_root_count, id);
666         return NULL;
667     }
668
669     int def = cgroup_enable_new_cgroups_detected_at_runtime;
670     const char *chart_id = id;
671     if(!*chart_id) {
672         chart_id = "/";
673
674         // disable by default the root cgroup
675         def = 0;
676         debug(D_CGROUP, "cgroup '%s' is the root container (by default %s)", id, (def)?"enabled":"disabled");
677     }
678     else {
679         if(*chart_id == '/') chart_id++;
680
681         size_t len = strlen(chart_id);
682
683         // disable by default the parent cgroup
684         // for known cgroup managers
685         if(!strcmp(chart_id, "lxc") ||
686                 !strcmp(chart_id, "docker") ||
687                 !strcmp(chart_id, "libvirt") ||
688                 !strcmp(chart_id, "qemu") ||
689                 !strcmp(chart_id, "systemd") ||
690                 !strcmp(chart_id, "system.slice") ||
691                 !strcmp(chart_id, "machine.slice") ||
692                 !strcmp(chart_id, "init.scope") ||
693                 !strcmp(chart_id, "user") ||
694                 !strcmp(chart_id, "system") ||
695                 !strcmp(chart_id, "machine") ||
696                 // starts with them
697                 (len >  6 && !strncmp(chart_id, "user/", 6)) ||
698                 (len > 11 && !strncmp(chart_id, "user.slice/", 11)) ||
699                 // ends with them
700                 (len >  5 && !strncmp(&chart_id[len -  5], ".user", 5)) ||
701                 (len >  5 && !strncmp(&chart_id[len -  5], ".swap", 5)) ||
702                 (len >  6 && !strncmp(&chart_id[len -  6], ".slice", 6)) ||
703                 (len >  6 && !strncmp(&chart_id[len -  6], ".mount", 6)) ||
704                 (len >  8 && !strncmp(&chart_id[len -  8], ".session", 8)) ||
705                 (len >  8 && !strncmp(&chart_id[len -  8], ".service", 8)) ||
706                 (len > 10 && !strncmp(&chart_id[len - 10], ".partition", 10)) ||
707                 // starts and ends with them
708                 (len > 7 && !strncmp(chart_id, "lxc/", 4) && !strncmp(&chart_id[len - 3], "/ns", 3)) // #1397
709                 ) {
710             def = 0;
711             debug(D_CGROUP, "cgroup '%s' is %s (by default)", id, (def)?"enabled":"disabled");
712         }
713     }
714
715     struct cgroup *cg = callocz(1, sizeof(struct cgroup));
716
717     cg->id = strdupz(id);
718     cg->hash = simple_hash(cg->id);
719
720     cg->chart_id = strdupz(chart_id);
721     netdata_fix_chart_id(cg->chart_id);
722     cg->hash_chart = simple_hash(cg->chart_id);
723
724     cg->chart_title = strdupz(chart_id);
725
726     if(!cgroup_root)
727         cgroup_root = cg;
728     else {
729         // append it
730         struct cgroup *e;
731         for(e = cgroup_root; e->next ;e = e->next) ;
732         e->next = cg;
733     }
734
735     cgroup_root_count++;
736
737     // fix the name by calling the external script
738     cgroup_get_chart_id(cg);
739
740     debug(D_CGROUP, "adding cgroup '%s' with chart id '%s'", id, chart_id);
741
742     char option[FILENAME_MAX + 1];
743     snprintfz(option, FILENAME_MAX, "enable cgroup %s", cg->chart_title);
744     cg->enabled = config_get_boolean("plugin:cgroups", option, def);
745
746     if(cg->enabled) {
747         struct cgroup *t;
748         for (t = cgroup_root; t; t = t->next) {
749             if (t != cg && t->enabled && t->hash_chart == cg->hash_chart && !strcmp(t->chart_id, cg->chart_id)) {
750                 if (!strncmp(t->chart_id, "/system.slice/", 14) && !strncmp(cg->chart_id, "/init.scope/system.slice/", 25)) {
751                     error("Control group with chart id '%s' already exists with id '%s' and is enabled. Swapping them by enabling cgroup with id '%s' and disabling cgroup with id '%s'.",
752                           cg->chart_id, t->id, cg->id, t->id);
753                     debug(D_CGROUP, "Control group with chart id '%s' already exists with id '%s' and is enabled. Swapping them by enabling cgroup with id '%s' and disabling cgroup with id '%s'.",
754                           cg->chart_id, t->id, cg->id, t->id);
755                     t->enabled = 0;
756                     t->options |= CGROUP_OPTIONS_DISABLED_DUPLICATE;
757                 }
758                 else {
759                     error("Control group with chart id '%s' already exists with id '%s' and is enabled and available. Disabling cgroup with id '%s'.",
760                           cg->chart_id, t->id, cg->id);
761                     debug(D_CGROUP, "Control group with chart id '%s' already exists with id '%s' and is enabled and available. Disabling cgroup with id '%s'.",
762                           cg->chart_id, t->id, cg->id);
763                     cg->enabled = 0;
764                     cg->options |= CGROUP_OPTIONS_DISABLED_DUPLICATE;
765                 }
766
767                 break;
768             }
769         }
770     }
771
772     debug(D_CGROUP, "Added cgroup '%s' with chart id '%s' and title '%s' as %s (default was %s)", cg->id, cg->chart_id, cg->chart_title, (cg->enabled)?"enabled":"disabled", (def)?"enabled":"disabled");
773
774     return cg;
775 }
776
777 void cgroup_free(struct cgroup *cg) {
778     debug(D_CGROUP, "Removing cgroup '%s' with chart id '%s' (was %s and %s)", cg->id, cg->chart_id, (cg->enabled)?"enabled":"disabled", (cg->available)?"available":"not available");
779
780     freez(cg->cpuacct_usage.cpu_percpu);
781
782     freez(cg->cpuacct_stat.filename);
783     freez(cg->cpuacct_usage.filename);
784     freez(cg->memory.filename);
785     freez(cg->io_service_bytes.filename);
786     freez(cg->io_serviced.filename);
787     freez(cg->throttle_io_service_bytes.filename);
788     freez(cg->throttle_io_serviced.filename);
789     freez(cg->io_merged.filename);
790     freez(cg->io_queued.filename);
791
792     freez(cg->id);
793     freez(cg->chart_id);
794     freez(cg->chart_title);
795     freez(cg);
796
797     cgroup_root_count--;
798 }
799
800 // find if a given cgroup exists
801 struct cgroup *cgroup_find(const char *id) {
802     debug(D_CGROUP, "searching for cgroup '%s'", id);
803
804     uint32_t hash = simple_hash(id);
805
806     struct cgroup *cg;
807     for(cg = cgroup_root; cg ; cg = cg->next) {
808         if(hash == cg->hash && strcmp(id, cg->id) == 0)
809             break;
810     }
811
812     debug(D_CGROUP, "cgroup_find('%s') %s", id, (cg)?"found":"not found");
813     return cg;
814 }
815
816 // ----------------------------------------------------------------------------
817 // detect running cgroups
818
819 // callback for find_file_in_subdirs()
820 void found_subdir_in_dir(const char *dir) {
821     debug(D_CGROUP, "examining cgroup dir '%s'", dir);
822
823     struct cgroup *cg = cgroup_find(dir);
824     if(!cg) {
825         if(*dir && cgroup_max_depth > 0) {
826             int depth = 0;
827             const char *s;
828
829             for(s = dir; *s ;s++)
830                 if(unlikely(*s == '/'))
831                     depth++;
832
833             if(depth > cgroup_max_depth) {
834                 info("cgroup '%s' is too deep (%d, while max is %d)", dir, depth, cgroup_max_depth);
835                 return;
836             }
837         }
838         debug(D_CGROUP, "will add dir '%s' as cgroup", dir);
839         cg = cgroup_add(dir);
840     }
841
842     if(cg) cg->available = 1;
843 }
844
845 int find_dir_in_subdirs(const char *base, const char *this, void (*callback)(const char *)) {
846     debug(D_CGROUP, "searching for directories in '%s'", base);
847
848     int ret = -1;
849     int enabled = -1;
850     if(!this) this = base;
851     size_t dirlen = strlen(this), baselen = strlen(base);
852     const char *relative_path = &this[baselen];
853
854     DIR *dir = opendir(this);
855     if(!dir) {
856         error("Cannot read cgroups directory '%s'", base);
857         return ret;
858     }
859     ret = 1;
860
861     callback(relative_path);
862
863     struct dirent *de = NULL;
864     while((de = readdir(dir))) {
865         if(de->d_type == DT_DIR
866             && (
867                 (de->d_name[0] == '.' && de->d_name[1] == '\0')
868                 || (de->d_name[0] == '.' && de->d_name[1] == '.' && de->d_name[2] == '\0')
869                 ))
870             continue;
871
872         debug(D_CGROUP, "examining '%s/%s'", this, de->d_name);
873
874         if(de->d_type == DT_DIR) {
875             if(enabled == -1) {
876                 const char *r = relative_path;
877                 if(*r == '\0') r = "/";
878                 else if (*r == '/') r++;
879
880                 // do not decent in directories we are not interested
881                 // https://github.com/firehol/netdata/issues/345
882                 int def = 1;
883                 size_t len = strlen(r);
884                 if(len >  5 && !strncmp(&r[len -  5], "-qemu", 5))
885                     def = 0;
886
887                 // we check for this option here
888                 // so that the config will not have settings
889                 // for leaf directories
890                 char option[FILENAME_MAX + 1];
891                 snprintfz(option, FILENAME_MAX, "search for cgroups under %s", r);
892                 option[FILENAME_MAX] = '\0';
893                 enabled = config_get_boolean("plugin:cgroups", option, def);
894             }
895
896             if(enabled) {
897                 char *s = mallocz(dirlen + strlen(de->d_name) + 2);
898                 strcpy(s, this);
899                 strcat(s, "/");
900                 strcat(s, de->d_name);
901                 int ret2 = find_dir_in_subdirs(base, s, callback);
902                 if(ret2 > 0) ret += ret2;
903                 freez(s);
904             }
905         }
906     }
907
908     closedir(dir);
909     return ret;
910 }
911
912 void mark_all_cgroups_as_not_available() {
913     debug(D_CGROUP, "marking all cgroups as not available");
914
915     struct cgroup *cg;
916
917     // mark all as not available
918     for(cg = cgroup_root; cg ; cg = cg->next) {
919         cg->available = 0;
920     }
921 }
922
923 void cleanup_all_cgroups() {
924     struct cgroup *cg = cgroup_root, *last = NULL;
925
926     for(; cg ;) {
927         if(!cg->available) {
928             // enable the first duplicate cgroup
929             {
930                 struct cgroup *t;
931                 for(t = cgroup_root; t ; t = t->next) {
932                     if(t != cg && t->available && !t->enabled && t->options & CGROUP_OPTIONS_DISABLED_DUPLICATE && t->hash_chart == cg->hash_chart && !strcmp(t->chart_id, cg->chart_id)) {
933                         debug(D_CGROUP, "Enabling duplicate of cgroup '%s' with id '%s', because the original with id '%s' stopped.", t->chart_id, t->id, cg->id);
934                         t->enabled = 1;
935                         t->options &= ~CGROUP_OPTIONS_DISABLED_DUPLICATE;
936                         break;
937                     }
938                 }
939             }
940
941             if(!last)
942                 cgroup_root = cg->next;
943             else
944                 last->next = cg->next;
945
946             cgroup_free(cg);
947
948             if(!last)
949                 cg = cgroup_root;
950             else
951                 cg = last->next;
952         }
953         else {
954             last = cg;
955             cg = cg->next;
956         }
957     }
958 }
959
960 void find_all_cgroups() {
961     debug(D_CGROUP, "searching for cgroups");
962
963     mark_all_cgroups_as_not_available();
964
965     if(cgroup_enable_cpuacct_stat || cgroup_enable_cpuacct_usage) {
966         if (find_dir_in_subdirs(cgroup_cpuacct_base, NULL, found_subdir_in_dir) == -1) {
967             cgroup_enable_cpuacct_stat = cgroup_enable_cpuacct_usage = 0;
968             error("disabled cgroup cpu statistics.");
969         }
970     }
971
972     if(cgroup_enable_blkio) {
973         if (find_dir_in_subdirs(cgroup_blkio_base, NULL, found_subdir_in_dir) == -1) {
974             cgroup_enable_blkio = 0;
975             error("disabled cgroup blkio statistics.");
976         }
977     }
978
979     if(cgroup_enable_memory) {
980         if(find_dir_in_subdirs(cgroup_memory_base, NULL, found_subdir_in_dir) == -1) {
981             cgroup_enable_memory = 0;
982             error("disabled cgroup memory statistics.");
983         }
984     }
985
986     if(cgroup_enable_devices) {
987         if(find_dir_in_subdirs(cgroup_devices_base, NULL, found_subdir_in_dir) == -1) {
988             cgroup_enable_devices = 0;
989             error("disabled cgroup devices statistics.");
990         }
991     }
992
993     // remove any non-existing cgroups
994     cleanup_all_cgroups();
995
996     struct cgroup *cg;
997     struct stat buf;
998     for(cg = cgroup_root; cg ; cg = cg->next) {
999         // fprintf(stderr, " >>> CGROUP '%s' (%u - %s) with name '%s'\n", cg->id, cg->hash, cg->available?"available":"stopped", cg->name);
1000
1001         if(unlikely(!cg->available))
1002             continue;
1003
1004         debug(D_CGROUP, "checking paths for cgroup '%s'", cg->id);
1005
1006         // check for newly added cgroups
1007         // and update the filenames they read
1008         char filename[FILENAME_MAX + 1];
1009         if(cgroup_enable_cpuacct_stat && !cg->cpuacct_stat.filename) {
1010             snprintfz(filename, FILENAME_MAX, "%s%s/cpuacct.stat", cgroup_cpuacct_base, cg->id);
1011             if(stat(filename, &buf) != -1) {
1012                 cg->cpuacct_stat.filename = strdupz(filename);
1013                 debug(D_CGROUP, "cpuacct.stat filename for cgroup '%s': '%s'", cg->id, cg->cpuacct_stat.filename);
1014             }
1015             else debug(D_CGROUP, "cpuacct.stat file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1016         }
1017         if(cgroup_enable_cpuacct_usage && !cg->cpuacct_usage.filename) {
1018             snprintfz(filename, FILENAME_MAX, "%s%s/cpuacct.usage_percpu", cgroup_cpuacct_base, cg->id);
1019             if(stat(filename, &buf) != -1) {
1020                 cg->cpuacct_usage.filename = strdupz(filename);
1021                 debug(D_CGROUP, "cpuacct.usage_percpu filename for cgroup '%s': '%s'", cg->id, cg->cpuacct_usage.filename);
1022             }
1023             else debug(D_CGROUP, "cpuacct.usage_percpu file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1024         }
1025         if(cgroup_enable_memory && !cg->memory.filename) {
1026             snprintfz(filename, FILENAME_MAX, "%s%s/memory.stat", cgroup_memory_base, cg->id);
1027             if(stat(filename, &buf) != -1) {
1028                 cg->memory.filename = strdupz(filename);
1029                 debug(D_CGROUP, "memory.stat filename for cgroup '%s': '%s'", cg->id, cg->memory.filename);
1030             }
1031             else debug(D_CGROUP, "memory.stat file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1032
1033             snprintfz(filename, FILENAME_MAX, "%s%s/memory.usage_in_bytes", cgroup_memory_base, cg->id);
1034             if(stat(filename, &buf) != -1) {
1035                 cg->memory.filename_usage_in_bytes = strdupz(filename);
1036                 debug(D_CGROUP, "memory.usage_in_bytes filename for cgroup '%s': '%s'", cg->id, cg->memory.filename_usage_in_bytes);
1037             }
1038             else debug(D_CGROUP, "memory.usage_in_bytes file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1039
1040             snprintfz(filename, FILENAME_MAX, "%s%s/memory.msw_usage_in_bytes", cgroup_memory_base, cg->id);
1041             if(stat(filename, &buf) != -1) {
1042                 cg->memory.filename_msw_usage_in_bytes = strdupz(filename);
1043                 debug(D_CGROUP, "memory.msw_usage_in_bytes filename for cgroup '%s': '%s'", cg->id, cg->memory.filename_msw_usage_in_bytes);
1044             }
1045             else debug(D_CGROUP, "memory.msw_usage_in_bytes file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1046
1047             snprintfz(filename, FILENAME_MAX, "%s%s/memory.failcnt", cgroup_memory_base, cg->id);
1048             if(stat(filename, &buf) != -1) {
1049                 cg->memory.filename_failcnt = strdupz(filename);
1050                 debug(D_CGROUP, "memory.failcnt filename for cgroup '%s': '%s'", cg->id, cg->memory.filename_failcnt);
1051             }
1052             else debug(D_CGROUP, "memory.failcnt file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1053         }
1054         if(cgroup_enable_blkio) {
1055             if(!cg->io_service_bytes.filename) {
1056                 snprintfz(filename, FILENAME_MAX, "%s%s/blkio.io_service_bytes", cgroup_blkio_base, cg->id);
1057                 if(stat(filename, &buf) != -1) {
1058                     cg->io_service_bytes.filename = strdupz(filename);
1059                     debug(D_CGROUP, "io_service_bytes filename for cgroup '%s': '%s'", cg->id, cg->io_service_bytes.filename);
1060                 }
1061                 else debug(D_CGROUP, "io_service_bytes file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1062             }
1063             if(!cg->io_serviced.filename) {
1064                 snprintfz(filename, FILENAME_MAX, "%s%s/blkio.io_serviced", cgroup_blkio_base, cg->id);
1065                 if(stat(filename, &buf) != -1) {
1066                     cg->io_serviced.filename = strdupz(filename);
1067                     debug(D_CGROUP, "io_serviced filename for cgroup '%s': '%s'", cg->id, cg->io_serviced.filename);
1068                 }
1069                 else debug(D_CGROUP, "io_serviced file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1070             }
1071             if(!cg->throttle_io_service_bytes.filename) {
1072                 snprintfz(filename, FILENAME_MAX, "%s%s/blkio.throttle.io_service_bytes", cgroup_blkio_base, cg->id);
1073                 if(stat(filename, &buf) != -1) {
1074                     cg->throttle_io_service_bytes.filename = strdupz(filename);
1075                     debug(D_CGROUP, "throttle_io_service_bytes filename for cgroup '%s': '%s'", cg->id, cg->throttle_io_service_bytes.filename);
1076                 }
1077                 else debug(D_CGROUP, "throttle_io_service_bytes file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1078             }
1079             if(!cg->throttle_io_serviced.filename) {
1080                 snprintfz(filename, FILENAME_MAX, "%s%s/blkio.throttle.io_serviced", cgroup_blkio_base, cg->id);
1081                 if(stat(filename, &buf) != -1) {
1082                     cg->throttle_io_serviced.filename = strdupz(filename);
1083                     debug(D_CGROUP, "throttle_io_serviced filename for cgroup '%s': '%s'", cg->id, cg->throttle_io_serviced.filename);
1084                 }
1085                 else debug(D_CGROUP, "throttle_io_serviced file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1086             }
1087             if(!cg->io_merged.filename) {
1088                 snprintfz(filename, FILENAME_MAX, "%s%s/blkio.io_merged", cgroup_blkio_base, cg->id);
1089                 if(stat(filename, &buf) != -1) {
1090                     cg->io_merged.filename = strdupz(filename);
1091                     debug(D_CGROUP, "io_merged filename for cgroup '%s': '%s'", cg->id, cg->io_merged.filename);
1092                 }
1093                 else debug(D_CGROUP, "io_merged file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1094             }
1095             if(!cg->io_queued.filename) {
1096                 snprintfz(filename, FILENAME_MAX, "%s%s/blkio.io_queued", cgroup_blkio_base, cg->id);
1097                 if(stat(filename, &buf) != -1) {
1098                     cg->io_queued.filename = strdupz(filename);
1099                     debug(D_CGROUP, "io_queued filename for cgroup '%s': '%s'", cg->id, cg->io_queued.filename);
1100                 }
1101                 else debug(D_CGROUP, "io_queued file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1102             }
1103         }
1104     }
1105
1106     debug(D_CGROUP, "done searching for cgroups");
1107     return;
1108 }
1109
1110 // ----------------------------------------------------------------------------
1111 // generate charts
1112
1113 #define CHART_TITLE_MAX 300
1114
1115 void update_cgroup_charts(int update_every) {
1116     debug(D_CGROUP, "updating cgroups charts");
1117
1118     char type[RRD_ID_LENGTH_MAX + 1];
1119     char title[CHART_TITLE_MAX + 1];
1120
1121     struct cgroup *cg;
1122     RRDSET *st;
1123
1124     for(cg = cgroup_root; cg ; cg = cg->next) {
1125         if(!cg->available || !cg->enabled)
1126             continue;
1127
1128         if(cg->id[0] == '\0')
1129             strcpy(type, "cgroup_root");
1130         else if(cg->id[0] == '/')
1131             snprintfz(type, RRD_ID_LENGTH_MAX, "cgroup_%s", cg->chart_id);
1132         else
1133             snprintfz(type, RRD_ID_LENGTH_MAX, "cgroup_%s", cg->chart_id);
1134
1135         netdata_fix_chart_id(type);
1136
1137         if(cg->cpuacct_stat.updated) {
1138             st = rrdset_find_bytype(type, "cpu");
1139             if(!st) {
1140                 snprintfz(title, CHART_TITLE_MAX, "CPU Usage (%d%% = %d core%s) for cgroup %s", (processors * 100), processors, (processors>1)?"s":"", cg->chart_title);
1141                 st = rrdset_create(type, "cpu", NULL, "cpu", "cgroup.cpu", title, "%", 40000, update_every, RRDSET_TYPE_STACKED);
1142
1143                 rrddim_add(st, "user", NULL, 100, hz, RRDDIM_INCREMENTAL);
1144                 rrddim_add(st, "system", NULL, 100, hz, RRDDIM_INCREMENTAL);
1145             }
1146             else rrdset_next(st);
1147
1148             rrddim_set(st, "user", cg->cpuacct_stat.user);
1149             rrddim_set(st, "system", cg->cpuacct_stat.system);
1150             rrdset_done(st);
1151         }
1152
1153         if(cg->cpuacct_usage.updated) {
1154             char id[RRD_ID_LENGTH_MAX + 1];
1155             unsigned int i;
1156
1157             st = rrdset_find_bytype(type, "cpu_per_core");
1158             if(!st) {
1159                 snprintfz(title, CHART_TITLE_MAX, "CPU Usage (%d%% = %d core%s) Per Core for cgroup %s", (processors * 100), processors, (processors>1)?"s":"", cg->chart_title);
1160                 st = rrdset_create(type, "cpu_per_core", NULL, "cpu", "cgroup.cpu_per_core", title, "%", 40100, update_every, RRDSET_TYPE_STACKED);
1161
1162                 for(i = 0; i < cg->cpuacct_usage.cpus ;i++) {
1163                     snprintfz(id, CHART_TITLE_MAX, "cpu%u", i);
1164                     rrddim_add(st, id, NULL, 100, 1000000000, RRDDIM_INCREMENTAL);
1165                 }
1166             }
1167             else rrdset_next(st);
1168
1169             for(i = 0; i < cg->cpuacct_usage.cpus ;i++) {
1170                 snprintfz(id, CHART_TITLE_MAX, "cpu%u", i);
1171                 rrddim_set(st, id, cg->cpuacct_usage.cpu_percpu[i]);
1172             }
1173             rrdset_done(st);
1174         }
1175
1176         if(cg->memory.updated) {
1177             if(cg->memory.cache + cg->memory.rss + cg->memory.rss_huge + cg->memory.mapped_file > 0) {
1178                 st = rrdset_find_bytype(type, "mem");
1179                 if(!st) {
1180                     snprintfz(title, CHART_TITLE_MAX, "Memory Usage for cgroup %s", cg->chart_title);
1181                     st = rrdset_create(type, "mem", NULL, "mem", "cgroup.mem", title, "MB", 40210, update_every,
1182                                        RRDSET_TYPE_STACKED);
1183
1184                     rrddim_add(st, "cache", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1185                     rrddim_add(st, "rss", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1186                     if(cg->memory.has_dirty_swap)
1187                         rrddim_add(st, "swap", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1188                     rrddim_add(st, "rss_huge", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1189                     rrddim_add(st, "mapped_file", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1190                 }
1191                 else rrdset_next(st);
1192
1193                 rrddim_set(st, "cache", cg->memory.cache);
1194                 rrddim_set(st, "rss", cg->memory.rss);
1195                 if(cg->memory.has_dirty_swap)
1196                     rrddim_set(st, "swap", cg->memory.swap);
1197                 rrddim_set(st, "rss_huge", cg->memory.rss_huge);
1198                 rrddim_set(st, "mapped_file", cg->memory.mapped_file);
1199                 rrdset_done(st);
1200             }
1201
1202             st = rrdset_find_bytype(type, "writeback");
1203             if(!st) {
1204                 snprintfz(title, CHART_TITLE_MAX, "Writeback Memory for cgroup %s", cg->chart_title);
1205                 st = rrdset_create(type, "writeback", NULL, "mem", "cgroup.writeback", title, "MB", 40300,
1206                                    update_every, RRDSET_TYPE_AREA);
1207
1208                 if(cg->memory.has_dirty_swap)
1209                     rrddim_add(st, "dirty", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1210                 rrddim_add(st, "writeback", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1211             }
1212             else rrdset_next(st);
1213
1214             if(cg->memory.has_dirty_swap)
1215                 rrddim_set(st, "dirty", cg->memory.dirty);
1216             rrddim_set(st, "writeback", cg->memory.writeback);
1217             rrdset_done(st);
1218
1219             if(cg->memory.pgpgin + cg->memory.pgpgout > 0) {
1220                 st = rrdset_find_bytype(type, "mem_activity");
1221                 if(!st) {
1222                     snprintfz(title, CHART_TITLE_MAX, "Memory Activity for cgroup %s", cg->chart_title);
1223                     st = rrdset_create(type, "mem_activity", NULL, "mem", "cgroup.mem_activity", title, "MB/s",
1224                                        40400, update_every, RRDSET_TYPE_LINE);
1225
1226                     rrddim_add(st, "pgpgin", "in", sysconf(_SC_PAGESIZE), 1024 * 1024, RRDDIM_INCREMENTAL);
1227                     rrddim_add(st, "pgpgout", "out", -sysconf(_SC_PAGESIZE), 1024 * 1024, RRDDIM_INCREMENTAL);
1228                 }
1229                 else rrdset_next(st);
1230
1231                 rrddim_set(st, "pgpgin", cg->memory.pgpgin);
1232                 rrddim_set(st, "pgpgout", cg->memory.pgpgout);
1233                 rrdset_done(st);
1234             }
1235
1236             if(cg->memory.pgfault + cg->memory.pgmajfault > 0) {
1237                 st = rrdset_find_bytype(type, "pgfaults");
1238                 if(!st) {
1239                     snprintfz(title, CHART_TITLE_MAX, "Memory Page Faults for cgroup %s", cg->chart_title);
1240                     st = rrdset_create(type, "pgfaults", NULL, "mem", "cgroup.pgfaults", title, "MB/s", 40500,
1241                                        update_every, RRDSET_TYPE_LINE);
1242
1243                     rrddim_add(st, "pgfault", NULL, sysconf(_SC_PAGESIZE), 1024 * 1024, RRDDIM_INCREMENTAL);
1244                     rrddim_add(st, "pgmajfault", "swap", -sysconf(_SC_PAGESIZE), 1024 * 1024, RRDDIM_INCREMENTAL);
1245                 }
1246                 else rrdset_next(st);
1247
1248                 rrddim_set(st, "pgfault", cg->memory.pgfault);
1249                 rrddim_set(st, "pgmajfault", cg->memory.pgmajfault);
1250                 rrdset_done(st);
1251             }
1252         }
1253
1254         if(cg->memory.usage_in_bytes_updated) {
1255             st = rrdset_find_bytype(type, "mem_usage");
1256             if(!st) {
1257                 snprintfz(title, CHART_TITLE_MAX, "Total Memory for cgroup %s", cg->chart_title);
1258                 st = rrdset_create(type, "mem_usage", NULL, "mem", "cgroup.mem_usage", title, "MB", 40200,
1259                                    update_every, RRDSET_TYPE_STACKED);
1260
1261                 rrddim_add(st, "ram", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1262                 rrddim_add(st, "swap", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1263             }
1264             else rrdset_next(st);
1265
1266             rrddim_set(st, "ram", cg->memory.usage_in_bytes);
1267             rrddim_set(st, "swap", (cg->memory.msw_usage_in_bytes > cg->memory.usage_in_bytes)?cg->memory.msw_usage_in_bytes - cg->memory.usage_in_bytes:0);
1268             rrdset_done(st);
1269         }
1270
1271         if(cg->memory.failcnt_updated && cg->memory.failcnt > 0) {
1272             st = rrdset_find_bytype(type, "mem_failcnt");
1273             if(!st) {
1274                 snprintfz(title, CHART_TITLE_MAX, "Memory Limit Failures for cgroup %s", cg->chart_title);
1275                 st = rrdset_create(type, "mem_failcnt", NULL, "mem", "cgroup.mem_failcnt", title, "MB", 40250,
1276                                    update_every, RRDSET_TYPE_LINE);
1277
1278                 rrddim_add(st, "failures", NULL, 1, 1, RRDDIM_INCREMENTAL);
1279             }
1280             else rrdset_next(st);
1281
1282             rrddim_set(st, "failures", cg->memory.failcnt);
1283             rrdset_done(st);
1284         }
1285
1286         if(cg->io_service_bytes.updated && cg->io_service_bytes.Read + cg->io_service_bytes.Write > 0) {
1287             st = rrdset_find_bytype(type, "io");
1288             if(!st) {
1289                 snprintfz(title, CHART_TITLE_MAX, "I/O Bandwidth (all disks) for cgroup %s", cg->chart_title);
1290                 st = rrdset_create(type, "io", NULL, "disk", "cgroup.io", title, "KB/s", 41200,
1291                                    update_every, RRDSET_TYPE_AREA);
1292
1293                 rrddim_add(st, "read", NULL, 1, 1024, RRDDIM_INCREMENTAL);
1294                 rrddim_add(st, "write", NULL, -1, 1024, RRDDIM_INCREMENTAL);
1295             }
1296             else rrdset_next(st);
1297
1298             rrddim_set(st, "read", cg->io_service_bytes.Read);
1299             rrddim_set(st, "write", cg->io_service_bytes.Write);
1300             rrdset_done(st);
1301         }
1302
1303         if(cg->io_serviced.updated && cg->io_serviced.Read + cg->io_serviced.Write > 0) {
1304             st = rrdset_find_bytype(type, "serviced_ops");
1305             if(!st) {
1306                 snprintfz(title, CHART_TITLE_MAX, "Serviced I/O Operations (all disks) for cgroup %s", cg->chart_title);
1307                 st = rrdset_create(type, "serviced_ops", NULL, "disk", "cgroup.serviced_ops", title, "operations/s", 41200,
1308                                    update_every, RRDSET_TYPE_LINE);
1309
1310                 rrddim_add(st, "read", NULL, 1, 1, RRDDIM_INCREMENTAL);
1311                 rrddim_add(st, "write", NULL, -1, 1, RRDDIM_INCREMENTAL);
1312             }
1313             else rrdset_next(st);
1314
1315             rrddim_set(st, "read", cg->io_serviced.Read);
1316             rrddim_set(st, "write", cg->io_serviced.Write);
1317             rrdset_done(st);
1318         }
1319
1320         if(cg->throttle_io_service_bytes.updated && cg->throttle_io_service_bytes.Read + cg->throttle_io_service_bytes.Write > 0) {
1321             st = rrdset_find_bytype(type, "throttle_io");
1322             if(!st) {
1323                 snprintfz(title, CHART_TITLE_MAX, "Throttle I/O Bandwidth (all disks) for cgroup %s", cg->chart_title);
1324                 st = rrdset_create(type, "throttle_io", NULL, "disk", "cgroup.throttle_io", title, "KB/s", 41200,
1325                                    update_every, RRDSET_TYPE_AREA);
1326
1327                 rrddim_add(st, "read", NULL, 1, 1024, RRDDIM_INCREMENTAL);
1328                 rrddim_add(st, "write", NULL, -1, 1024, RRDDIM_INCREMENTAL);
1329             }
1330             else rrdset_next(st);
1331
1332             rrddim_set(st, "read", cg->throttle_io_service_bytes.Read);
1333             rrddim_set(st, "write", cg->throttle_io_service_bytes.Write);
1334             rrdset_done(st);
1335         }
1336
1337
1338         if(cg->throttle_io_serviced.updated && cg->throttle_io_serviced.Read + cg->throttle_io_serviced.Write > 0) {
1339             st = rrdset_find_bytype(type, "throttle_serviced_ops");
1340             if(!st) {
1341                 snprintfz(title, CHART_TITLE_MAX, "Throttle Serviced I/O Operations (all disks) for cgroup %s", cg->chart_title);
1342                 st = rrdset_create(type, "throttle_serviced_ops", NULL, "disk", "cgroup.throttle_serviced_ops", title, "operations/s", 41200,
1343                                    update_every, RRDSET_TYPE_LINE);
1344
1345                 rrddim_add(st, "read", NULL, 1, 1, RRDDIM_INCREMENTAL);
1346                 rrddim_add(st, "write", NULL, -1, 1, RRDDIM_INCREMENTAL);
1347             }
1348             else rrdset_next(st);
1349
1350             rrddim_set(st, "read", cg->throttle_io_serviced.Read);
1351             rrddim_set(st, "write", cg->throttle_io_serviced.Write);
1352             rrdset_done(st);
1353         }
1354
1355         if(cg->io_queued.updated) {
1356             st = rrdset_find_bytype(type, "queued_ops");
1357             if(!st) {
1358                 snprintfz(title, CHART_TITLE_MAX, "Queued I/O Operations (all disks) for cgroup %s", cg->chart_title);
1359                 st = rrdset_create(type, "queued_ops", NULL, "disk", "cgroup.queued_ops", title, "operations", 42000,
1360                                    update_every, RRDSET_TYPE_LINE);
1361
1362                 rrddim_add(st, "read", NULL, 1, 1, RRDDIM_ABSOLUTE);
1363                 rrddim_add(st, "write", NULL, -1, 1, RRDDIM_ABSOLUTE);
1364             }
1365             else rrdset_next(st);
1366
1367             rrddim_set(st, "read", cg->io_queued.Read);
1368             rrddim_set(st, "write", cg->io_queued.Write);
1369             rrdset_done(st);
1370         }
1371
1372         if(cg->io_merged.updated && cg->io_merged.Read + cg->io_merged.Write > 0) {
1373             st = rrdset_find_bytype(type, "merged_ops");
1374             if(!st) {
1375                 snprintfz(title, CHART_TITLE_MAX, "Merged I/O Operations (all disks) for cgroup %s", cg->chart_title);
1376                 st = rrdset_create(type, "merged_ops", NULL, "disk", "cgroup.merged_ops", title, "operations/s", 42100,
1377                                    update_every, RRDSET_TYPE_LINE);
1378
1379                 rrddim_add(st, "read", NULL, 1, 1024, RRDDIM_INCREMENTAL);
1380                 rrddim_add(st, "write", NULL, -1, 1024, RRDDIM_INCREMENTAL);
1381             }
1382             else rrdset_next(st);
1383
1384             rrddim_set(st, "read", cg->io_merged.Read);
1385             rrddim_set(st, "write", cg->io_merged.Write);
1386             rrdset_done(st);
1387         }
1388     }
1389
1390     debug(D_CGROUP, "done updating cgroups charts");
1391 }
1392
1393 // ----------------------------------------------------------------------------
1394 // cgroups main
1395
1396 int do_sys_fs_cgroup(int update_every, usec_t dt) {
1397     (void)dt;
1398
1399     static int cgroup_global_config_read = 0;
1400     static time_t last_run = 0;
1401     time_t now = now_realtime_sec();
1402
1403     if(unlikely(!cgroup_global_config_read)) {
1404         read_cgroup_plugin_configuration();
1405         cgroup_global_config_read = 1;
1406     }
1407
1408     if(unlikely(cgroup_enable_new_cgroups_detected_at_runtime && now - last_run > cgroup_check_for_new_every)) {
1409         find_all_cgroups();
1410         last_run = now;
1411     }
1412
1413     read_all_cgroups(cgroup_root);
1414     update_cgroup_charts(update_every);
1415
1416     return 0;
1417 }
1418
1419 void *cgroups_main(void *ptr)
1420 {
1421     (void)ptr;
1422
1423     info("CGROUP Plugin thread created with task id %d", gettid());
1424
1425     if(pthread_setcanceltype(PTHREAD_CANCEL_DEFERRED, NULL) != 0)
1426         error("Cannot set pthread cancel type to DEFERRED.");
1427
1428     if(pthread_setcancelstate(PTHREAD_CANCEL_ENABLE, NULL) != 0)
1429         error("Cannot set pthread cancel state to ENABLE.");
1430
1431     struct rusage thread;
1432
1433     // when ZERO, attempt to do it
1434     int vdo_sys_fs_cgroup           = 0;
1435     int vdo_cpu_netdata             = !config_get_boolean("plugin:cgroups", "cgroups plugin resources", 1);
1436
1437     // keep track of the time each module was called
1438     usec_t sutime_sys_fs_cgroup = 0ULL;
1439
1440     // the next time we will run - aligned properly
1441     usec_t sunext = (now_realtime_sec() - (now_realtime_sec() % rrd_update_every) + rrd_update_every) * USEC_PER_SEC;
1442
1443     RRDSET *stcpu_thread = NULL;
1444
1445     for(;;) {
1446         usec_t sunow;
1447         if(unlikely(netdata_exit)) break;
1448
1449         // delay until it is our time to run
1450         while((sunow = now_realtime_usec()) < sunext)
1451             sleep_usec(sunext - sunow);
1452
1453         // find the next time we need to run
1454         while(now_realtime_usec() > sunext)
1455             sunext += rrd_update_every * USEC_PER_SEC;
1456
1457         if(unlikely(netdata_exit)) break;
1458
1459         // BEGIN -- the job to be done
1460
1461         if(!vdo_sys_fs_cgroup) {
1462             debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_sys_fs_cgroup().");
1463             sunow = now_realtime_usec();
1464             vdo_sys_fs_cgroup = do_sys_fs_cgroup(rrd_update_every, (sutime_sys_fs_cgroup > 0)?sunow - sutime_sys_fs_cgroup:0ULL);
1465             sutime_sys_fs_cgroup = sunow;
1466         }
1467         if(unlikely(netdata_exit)) break;
1468
1469         // END -- the job is done
1470
1471         // --------------------------------------------------------------------
1472
1473         if(!vdo_cpu_netdata) {
1474             getrusage(RUSAGE_THREAD, &thread);
1475
1476             if(!stcpu_thread) stcpu_thread = rrdset_find("netdata.plugin_cgroups_cpu");
1477             if(!stcpu_thread) {
1478                 stcpu_thread = rrdset_create("netdata", "plugin_cgroups_cpu", NULL, "proc.internal", NULL, "NetData CGroups Plugin CPU usage", "milliseconds/s", 132000, rrd_update_every, RRDSET_TYPE_STACKED);
1479
1480                 rrddim_add(stcpu_thread, "user",  NULL,  1, 1000, RRDDIM_INCREMENTAL);
1481                 rrddim_add(stcpu_thread, "system", NULL, 1, 1000, RRDDIM_INCREMENTAL);
1482             }
1483             else rrdset_next(stcpu_thread);
1484
1485             rrddim_set(stcpu_thread, "user"  , thread.ru_utime.tv_sec * 1000000ULL + thread.ru_utime.tv_usec);
1486             rrddim_set(stcpu_thread, "system", thread.ru_stime.tv_sec * 1000000ULL + thread.ru_stime.tv_usec);
1487             rrdset_done(stcpu_thread);
1488         }
1489     }
1490
1491     info("CGROUP thread exiting");
1492
1493     pthread_exit(NULL);
1494     return NULL;
1495 }