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