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