src/sys_fs_cgroup.c

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