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 = 50;
  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) s = "/sys/fs/cgroup/cpuacct";
  51         else s = mi->mount_point;
  52         snprintf(filename, FILENAME_MAX, "%s%s", global_host_prefix, s);
  53         cgroup_cpuacct_base = config_get("plugin:cgroups", "path to /sys/fs/cgroup/cpuacct", filename);
  54
  55         mi = mountinfo_find_by_filesystem_mount_source(root, "cgroup", "blkio");
  56         if(!mi) s = "/sys/fs/cgroup/blkio";
  57         else s = mi->mount_point;
  58         snprintf(filename, FILENAME_MAX, "%s%s", global_host_prefix, s);
  59         cgroup_blkio_base = config_get("plugin:cgroups", "path to /sys/fs/cgroup/blkio", filename);
  60
  61         mi = mountinfo_find_by_filesystem_mount_source(root, "cgroup", "memory");
  62         if(!mi) s = "/sys/fs/cgroup/memory";
  63         else s = mi->mount_point;
  64         snprintf(filename, FILENAME_MAX, "%s%s", global_host_prefix, s);
  65         cgroup_memory_base = config_get("plugin:cgroups", "path to /sys/fs/cgroup/memory", filename);
  66
  67         cgroup_root_max = config_get_number("plugin:cgroups", "max cgroups to allow", cgroup_root_max);
  68         cgroup_max_depth = config_get_number("plugin:cgroups", "max cgroups depth to monitor", cgroup_max_depth);
  69
  70         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);
  71
  72         mountinfo_free(root);
  73 }
  74
  75 // ----------------------------------------------------------------------------
  76 // cgroup objects
  77
  78 struct blkio {
  79         int updated;
  80
  81         char *filename;
  82
  83         unsigned long long Read;
  84         unsigned long long Write;
  85 /*
  86         unsigned long long Sync;
  87         unsigned long long Async;
  88         unsigned long long Total;
  89 */
  90 };
  91
  92 // https://www.kernel.org/doc/Documentation/cgroup-v1/memory.txt
  93 struct memory {
  94         int updated;
  95
  96         char *filename;
  97
  98         int has_dirty_swap;
  99
 100         unsigned long long cache;
 101         unsigned long long rss;
 102         unsigned long long rss_huge;
 103         unsigned long long mapped_file;
 104         unsigned long long writeback;
 105         unsigned long long dirty;
 106         unsigned long long swap;
 107         unsigned long long pgpgin;
 108         unsigned long long pgpgout;
 109         unsigned long long pgfault;
 110         unsigned long long pgmajfault;
 111 /*
 112         unsigned long long inactive_anon;
 113         unsigned long long active_anon;
 114         unsigned long long inactive_file;
 115         unsigned long long active_file;
 116         unsigned long long unevictable;
 117         unsigned long long hierarchical_memory_limit;
 118         unsigned long long total_cache;
 119         unsigned long long total_rss;
 120         unsigned long long total_rss_huge;
 121         unsigned long long total_mapped_file;
 122         unsigned long long total_writeback;
 123         unsigned long long total_dirty;
 124         unsigned long long total_swap;
 125         unsigned long long total_pgpgin;
 126         unsigned long long total_pgpgout;
 127         unsigned long long total_pgfault;
 128         unsigned long long total_pgmajfault;
 129         unsigned long long total_inactive_anon;
 130         unsigned long long total_active_anon;
 131         unsigned long long total_inactive_file;
 132         unsigned long long total_active_file;
 133         unsigned long long total_unevictable;
 134 */
 135 };
 136
 137 // https://www.kernel.org/doc/Documentation/cgroup-v1/cpuacct.txt
 138 struct cpuacct_stat {
 139         int updated;
 140
 141         char *filename;
 142
 143         unsigned long long user;
 144         unsigned long long system;
 145 };
 146
 147 // https://www.kernel.org/doc/Documentation/cgroup-v1/cpuacct.txt
 148 struct cpuacct_usage {
 149         int updated;
 150
 151         char *filename;
 152
 153         unsigned int cpus;
 154         unsigned long long *cpu_percpu;
 155 };
 156
 157 struct cgroup {
 158         int available;          // found in the filesystem
 159         int enabled;            // enabled in the config
 160
 161         char *id;
 162         uint32_t hash;
 163
 164         char *chart_id;
 165
 166         struct cpuacct_stat cpuacct_stat;
 167         struct cpuacct_usage cpuacct_usage;
 168
 169         struct memory memory;
 170
 171         struct blkio io_service_bytes;                          // bytes
 172         struct blkio io_serviced;                                       // operations
 173
 174         struct blkio throttle_io_service_bytes;         // bytes
 175         struct blkio throttle_io_serviced;                      // operations
 176
 177         struct blkio io_merged;                                         // operations
 178         struct blkio io_queued;                                         // operations
 179
 180         struct cgroup *next;
 181
 182 } *cgroup_root = NULL;
 183
 184 // ----------------------------------------------------------------------------
 185 // read values from /sys
 186
 187 void cgroup_read_cpuacct_stat(struct cpuacct_stat *cp) {
 188         static procfile *ff = NULL;
 189
 190         static uint32_t user_hash = 0;
 191         static uint32_t system_hash = 0;
 192
 193         if(unlikely(user_hash == 0)) {
 194                 user_hash = simple_hash("user");
 195                 system_hash = simple_hash("system");
 196         }
 197
 198         cp->updated = 0;
 199         if(cp->filename) {
 200                 ff = procfile_reopen(ff, cp->filename, NULL, PROCFILE_FLAG_DEFAULT);
 201                 if(!ff) return;
 202
 203                 ff = procfile_readall(ff);
 204                 if(!ff) return;
 205
 206                 unsigned long i, lines = procfile_lines(ff);
 207
 208                 if(lines < 1) {
 209                         error("File '%s' should have 1+ lines.", cp->filename);
 210                         return;
 211                 }
 212
 213                 for(i = 0; i < lines ; i++) {
 214                         char *s = procfile_lineword(ff, i, 0);
 215                         uint32_t hash = simple_hash(s);
 216
 217                         if(hash == user_hash && !strcmp(s, "user"))
 218                                 cp->user = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 219
 220                         else if(hash == system_hash && !strcmp(s, "system"))
 221                                 cp->system = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 222                 }
 223
 224                 cp->updated = 1;
 225
 226                 // fprintf(stderr, "READ '%s': user: %llu, system: %llu\n", cp->filename, cp->user, cp->system);
 227         }
 228 }
 229
 230 void cgroup_read_cpuacct_usage(struct cpuacct_usage *ca) {
 231         static procfile *ff = NULL;
 232
 233         ca->updated = 0;
 234         if(ca->filename) {
 235                 ff = procfile_reopen(ff, ca->filename, NULL, PROCFILE_FLAG_DEFAULT);
 236                 if(!ff) return;
 237
 238                 ff = procfile_readall(ff);
 239                 if(!ff) return;
 240
 241                 if(procfile_lines(ff) < 1) {
 242                         error("File '%s' should have 1+ lines but has %d.", ca->filename, procfile_lines(ff));
 243                         return;
 244                 }
 245
 246                 unsigned long i = procfile_linewords(ff, 0);
 247                 if(i <= 0) return;
 248
 249                 // we may have 1 more CPU reported
 250                 while(i > 0) {
 251                         char *s = procfile_lineword(ff, 0, i - 1);
 252                         if(!*s) i--;
 253                         else break;
 254                 }
 255
 256                 if(i != ca->cpus) {
 257                         free(ca->cpu_percpu);
 258
 259                         ca->cpu_percpu = malloc(sizeof(unsigned long long) * i);
 260                         if(!ca->cpu_percpu)
 261                                 fatal("Cannot allocate memory (%z bytes)", sizeof(unsigned long long) * i);
 262
 263                         ca->cpus = i;
 264                 }
 265
 266                 for(i = 0; i < ca->cpus ;i++) {
 267                         ca->cpu_percpu[i] = strtoull(procfile_lineword(ff, 0, i), NULL, 10);
 268                         // fprintf(stderr, "READ '%s': cpu%d/%d: %llu ('%s')\n", ca->filename, i, ca->cpus, ca->cpu_percpu[i], procfile_lineword(ff, 0, i));
 269                 }
 270
 271                 ca->updated = 1;
 272         }
 273 }
 274
 275 void cgroup_read_blkio(struct blkio *io) {
 276         static procfile *ff = NULL;
 277
 278         static uint32_t Read_hash = 0;
 279         static uint32_t Write_hash = 0;
 280 /*
 281         static uint32_t Sync_hash = 0;
 282         static uint32_t Async_hash = 0;
 283         static uint32_t Total_hash = 0;
 284 */
 285
 286         if(unlikely(Read_hash == 0)) {
 287                 Read_hash = simple_hash("Read");
 288                 Write_hash = simple_hash("Write");
 289 /*
 290                 Sync_hash = simple_hash("Sync");
 291                 Async_hash = simple_hash("Async");
 292                 Total_hash = simple_hash("Total");
 293 */
 294         }
 295
 296         io->updated = 0;
 297         if(io->filename) {
 298                 ff = procfile_reopen(ff, io->filename, NULL, PROCFILE_FLAG_DEFAULT);
 299                 if(!ff) return;
 300
 301                 ff = procfile_readall(ff);
 302                 if(!ff) return;
 303
 304                 unsigned long i, lines = procfile_lines(ff);
 305
 306                 if(lines < 1) {
 307                         error("File '%s' should have 1+ lines.", io->filename);
 308                         return;
 309                 }
 310
 311                 io->Read = 0;
 312                 io->Write = 0;
 313 /*
 314                 io->Sync = 0;
 315                 io->Async = 0;
 316                 io->Total = 0;
 317 */
 318
 319                 for(i = 0; i < lines ; i++) {
 320                         char *s = procfile_lineword(ff, i, 1);
 321                         uint32_t hash = simple_hash(s);
 322
 323                         if(hash == Read_hash && !strcmp(s, "Read"))
 324                                 io->Read += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
 325
 326                         else if(hash == Write_hash && !strcmp(s, "Write"))
 327                                 io->Write += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
 328
 329 /*
 330                         else if(hash == Sync_hash && !strcmp(s, "Sync"))
 331                                 io->Sync += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
 332
 333                         else if(hash == Async_hash && !strcmp(s, "Async"))
 334                                 io->Async += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
 335
 336                         else if(hash == Total_hash && !strcmp(s, "Total"))
 337                                 io->Total += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
 338 */
 339                 }
 340
 341                 io->updated = 1;
 342                 // 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);
 343         }
 344 }
 345
 346 void cgroup_read_memory(struct memory *mem) {
 347         static procfile *ff = NULL;
 348
 349         static uint32_t cache_hash = 0;
 350         static uint32_t rss_hash = 0;
 351         static uint32_t rss_huge_hash = 0;
 352         static uint32_t mapped_file_hash = 0;
 353         static uint32_t writeback_hash = 0;
 354         static uint32_t dirty_hash = 0;
 355         static uint32_t swap_hash = 0;
 356         static uint32_t pgpgin_hash = 0;
 357         static uint32_t pgpgout_hash = 0;
 358         static uint32_t pgfault_hash = 0;
 359         static uint32_t pgmajfault_hash = 0;
 360 /*
 361         static uint32_t inactive_anon_hash = 0;
 362         static uint32_t active_anon_hash = 0;
 363         static uint32_t inactive_file_hash = 0;
 364         static uint32_t active_file_hash = 0;
 365         static uint32_t unevictable_hash = 0;
 366         static uint32_t hierarchical_memory_limit_hash = 0;
 367         static uint32_t total_cache_hash = 0;
 368         static uint32_t total_rss_hash = 0;
 369         static uint32_t total_rss_huge_hash = 0;
 370         static uint32_t total_mapped_file_hash = 0;
 371         static uint32_t total_writeback_hash = 0;
 372         static uint32_t total_dirty_hash = 0;
 373         static uint32_t total_swap_hash = 0;
 374         static uint32_t total_pgpgin_hash = 0;
 375         static uint32_t total_pgpgout_hash = 0;
 376         static uint32_t total_pgfault_hash = 0;
 377         static uint32_t total_pgmajfault_hash = 0;
 378         static uint32_t total_inactive_anon_hash = 0;
 379         static uint32_t total_active_anon_hash = 0;
 380         static uint32_t total_inactive_file_hash = 0;
 381         static uint32_t total_active_file_hash = 0;
 382         static uint32_t total_unevictable_hash = 0;
 383 */
 384         if(unlikely(cache_hash == 0)) {
 385                 cache_hash = simple_hash("cache");
 386                 rss_hash = simple_hash("rss");
 387                 rss_huge_hash = simple_hash("rss_huge");
 388                 mapped_file_hash = simple_hash("mapped_file");
 389                 writeback_hash = simple_hash("writeback");
 390                 dirty_hash = simple_hash("dirty");
 391                 swap_hash = simple_hash("swap");
 392                 pgpgin_hash = simple_hash("pgpgin");
 393                 pgpgout_hash = simple_hash("pgpgout");
 394                 pgfault_hash = simple_hash("pgfault");
 395                 pgmajfault_hash = simple_hash("pgmajfault");
 396 /*
 397                 inactive_anon_hash = simple_hash("inactive_anon");
 398                 active_anon_hash = simple_hash("active_anon");
 399                 inactive_file_hash = simple_hash("inactive_file");
 400                 active_file_hash = simple_hash("active_file");
 401                 unevictable_hash = simple_hash("unevictable");
 402                 hierarchical_memory_limit_hash = simple_hash("hierarchical_memory_limit");
 403                 total_cache_hash = simple_hash("total_cache");
 404                 total_rss_hash = simple_hash("total_rss");
 405                 total_rss_huge_hash = simple_hash("total_rss_huge");
 406                 total_mapped_file_hash = simple_hash("total_mapped_file");
 407                 total_writeback_hash = simple_hash("total_writeback");
 408                 total_dirty_hash = simple_hash("total_dirty");
 409                 total_swap_hash = simple_hash("total_swap");
 410                 total_pgpgin_hash = simple_hash("total_pgpgin");
 411                 total_pgpgout_hash = simple_hash("total_pgpgout");
 412                 total_pgfault_hash = simple_hash("total_pgfault");
 413                 total_pgmajfault_hash = simple_hash("total_pgmajfault");
 414                 total_inactive_anon_hash = simple_hash("total_inactive_anon");
 415                 total_active_anon_hash = simple_hash("total_active_anon");
 416                 total_inactive_file_hash = simple_hash("total_inactive_file");
 417                 total_active_file_hash = simple_hash("total_active_file");
 418                 total_unevictable_hash = simple_hash("total_unevictable");
 419 */
 420         }
 421
 422         mem->updated = 0;
 423         if(mem->filename) {
 424                 ff = procfile_reopen(ff, mem->filename, NULL, PROCFILE_FLAG_DEFAULT);
 425                 if(!ff) return;
 426
 427                 ff = procfile_readall(ff);
 428                 if(!ff) return;
 429
 430                 unsigned long i, lines = procfile_lines(ff);
 431
 432                 if(lines < 1) {
 433                         error("File '%s' should have 1+ lines.", mem->filename);
 434                         return;
 435                 }
 436
 437                 for(i = 0; i < lines ; i++) {
 438                         char *s = procfile_lineword(ff, i, 0);
 439                         uint32_t hash = simple_hash(s);
 440
 441                         if(hash == cache_hash && !strcmp(s, "cache"))
 442                                 mem->cache = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 443
 444                         else if(hash == rss_hash && !strcmp(s, "rss"))
 445                                 mem->rss = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 446
 447                         else if(hash == rss_huge_hash && !strcmp(s, "rss_huge"))
 448                                 mem->rss_huge = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 449
 450                         else if(hash == mapped_file_hash && !strcmp(s, "mapped_file"))
 451                                 mem->mapped_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 452
 453                         else if(hash == writeback_hash && !strcmp(s, "writeback"))
 454                                 mem->writeback = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 455
 456                         else if(hash == dirty_hash && !strcmp(s, "dirty")) {
 457                                 mem->dirty = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 458                                 mem->has_dirty_swap = 1;
 459                         }
 460
 461                         else if(hash == swap_hash && !strcmp(s, "swap")) {
 462                                 mem->swap = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 463                                 mem->has_dirty_swap = 1;
 464                         }
 465
 466                         else if(hash == pgpgin_hash && !strcmp(s, "pgpgin"))
 467                                 mem->pgpgin = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 468
 469                         else if(hash == pgpgout_hash && !strcmp(s, "pgpgout"))
 470                                 mem->pgpgout = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 471
 472                         else if(hash == pgfault_hash && !strcmp(s, "pgfault"))
 473                                 mem->pgfault = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 474
 475                         else if(hash == pgmajfault_hash && !strcmp(s, "pgmajfault"))
 476                                 mem->pgmajfault = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 477
 478 /*
 479                         else if(hash == inactive_anon_hash && !strcmp(s, "inactive_anon"))
 480                                 mem->inactive_anon = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 481
 482                         else if(hash == active_anon_hash && !strcmp(s, "active_anon"))
 483                                 mem->active_anon = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 484
 485                         else if(hash == inactive_file_hash && !strcmp(s, "inactive_file"))
 486                                 mem->inactive_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 487
 488                         else if(hash == active_file_hash && !strcmp(s, "active_file"))
 489                                 mem->active_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 490
 491                         else if(hash == unevictable_hash && !strcmp(s, "unevictable"))
 492                                 mem->unevictable = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 493
 494                         else if(hash == hierarchical_memory_limit_hash && !strcmp(s, "hierarchical_memory_limit"))
 495                                 mem->hierarchical_memory_limit = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 496
 497                 else if(hash == total_cache_hash && !strcmp(s, "total_cache"))
 498                                 mem->total_cache = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 499
 500                         else if(hash == total_rss_hash && !strcmp(s, "total_rss"))
 501                                 mem->total_rss = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 502
 503                         else if(hash == total_rss_huge_hash && !strcmp(s, "total_rss_huge"))
 504                                 mem->total_rss_huge = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 505
 506                         else if(hash == total_mapped_file_hash && !strcmp(s, "total_mapped_file"))
 507                                 mem->total_mapped_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 508
 509                         else if(hash == total_writeback_hash && !strcmp(s, "total_writeback"))
 510                                 mem->total_writeback = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 511
 512                         else if(hash == total_dirty_hash && !strcmp(s, "total_dirty"))
 513                                 mem->total_dirty = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 514
 515                         else if(hash == total_swap_hash && !strcmp(s, "total_swap"))
 516                                 mem->total_swap = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 517
 518                         else if(hash == total_pgpgin_hash && !strcmp(s, "total_pgpgin"))
 519                                 mem->total_pgpgin = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 520
 521                         else if(hash == total_pgpgout_hash && !strcmp(s, "total_pgpgout"))
 522                                 mem->total_pgpgout = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 523
 524                         else if(hash == total_pgfault_hash && !strcmp(s, "total_pgfault"))
 525                                 mem->total_pgfault = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 526
 527                         else if(hash == total_pgmajfault_hash && !strcmp(s, "total_pgmajfault"))
 528                                 mem->total_pgmajfault = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 529
 530                         else if(hash == total_inactive_anon_hash && !strcmp(s, "total_inactive_anon"))
 531                                 mem->total_inactive_anon = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 532
 533                         else if(hash == total_active_anon_hash && !strcmp(s, "total_active_anon"))
 534                                 mem->total_active_anon = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 535
 536                         else if(hash == total_inactive_file_hash && !strcmp(s, "total_inactive_file"))
 537                                 mem->total_inactive_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 538
 539                         else if(hash == total_active_file_hash && !strcmp(s, "total_active_file"))
 540                                 mem->total_active_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 541
 542                         else if(hash == total_unevictable_hash && !strcmp(s, "total_unevictable"))
 543                                 mem->total_unevictable = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 544 */
 545                 }
 546
 547                 // 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);
 548
 549                 mem->updated = 1;
 550         }
 551 }
 552
 553 void cgroup_read(struct cgroup *cg) {
 554         debug(D_CGROUP, "reading metrics for cgroups '%s'", cg->id);
 555
 556         cgroup_read_cpuacct_stat(&cg->cpuacct_stat);
 557         cgroup_read_cpuacct_usage(&cg->cpuacct_usage);
 558         cgroup_read_memory(&cg->memory);
 559         cgroup_read_blkio(&cg->io_service_bytes);
 560         cgroup_read_blkio(&cg->io_serviced);
 561         cgroup_read_blkio(&cg->throttle_io_service_bytes);
 562         cgroup_read_blkio(&cg->throttle_io_serviced);
 563         cgroup_read_blkio(&cg->io_merged);
 564         cgroup_read_blkio(&cg->io_queued);
 565 }
 566
 567 void read_all_cgroups(struct cgroup *root) {
 568         debug(D_CGROUP, "reading metrics for all cgroups");
 569
 570         struct cgroup *cg;
 571
 572         for(cg = root; cg ; cg = cg->next)
 573                 if(cg->enabled && cg->available)
 574                         cgroup_read(cg);
 575 }
 576
 577 // ----------------------------------------------------------------------------
 578 // add/remove/find cgroup objects
 579
 580 #define CGROUP_CHARTID_LINE_MAX 1024
 581
 582 void cgroup_get_chart_id(struct cgroup *cg) {
 583         debug(D_CGROUP, "getting the name of cgroup '%s'", cg->id);
 584
 585         pid_t cgroup_pid;
 586         char buffer[CGROUP_CHARTID_LINE_MAX + 1];
 587
 588         snprintf(buffer, CGROUP_CHARTID_LINE_MAX, "exec %s '%s'",
 589                  config_get("plugin:cgroups", "script to get cgroup names", PLUGINS_DIR "/cgroup-name.sh"), cg->chart_id);
 590
 591         debug(D_CGROUP, "executing command '%s' for cgroup '%s'", buffer, cg->id);
 592         FILE *fp = mypopen(buffer, &cgroup_pid);
 593         if(!fp) {
 594                 error("CGROUP: Cannot popen(\"%s\", \"r\").", buffer);
 595                 return;
 596         }
 597         debug(D_CGROUP, "reading from command '%s' for cgroup '%s'", buffer, cg->id);
 598         char *s = fgets(buffer, CGROUP_CHARTID_LINE_MAX, fp);
 599         debug(D_CGROUP, "closing command for cgroup '%s'", cg->id);
 600         mypclose(fp, cgroup_pid);
 601         debug(D_CGROUP, "closed command for cgroup '%s'", cg->id);
 602
 603         if(s && *s && *s != '\n') {
 604                 debug(D_CGROUP, "cgroup '%s' should be renamed to '%s'", cg->id, s);
 605
 606                 trim(s);
 607                 netdata_fix_chart_id(s);
 608                 free(cg->chart_id);
 609                 cg->chart_id = strdup(s);
 610                 if(!cg->chart_id)
 611                         fatal("CGROUP: Cannot allocate memory for chart id of cgroup '%s' chart id: '%s'", cg->id, s);
 612
 613                 debug(D_CGROUP, "cgroup '%s' renamed to '%s'", cg->id, cg->chart_id);
 614         }
 615         else debug(D_CGROUP, "cgroup '%s' is not to be renamed (will be shown as '%s')", cg->id, cg->chart_id);
 616 }
 617
 618 struct cgroup *cgroup_add(const char *id) {
 619         debug(D_CGROUP, "adding cgroup '%s'", id);
 620
 621         if(cgroup_root_count >= cgroup_root_max) {
 622                 info("Maximum number of cgroups reached (%d). Not adding cgroup '%s'", cgroup_root_count, id);
 623                 return NULL;
 624         }
 625
 626         int def = cgroup_enable_new_cgroups_detected_at_runtime;
 627         const char *chart_id = id;
 628         if(!*chart_id) {
 629                 chart_id = "/";
 630
 631                 // disable by default the host cgroup
 632                 def = 0;
 633                 debug(D_CGROUP, "cgroup '%s' is the host container (by default %s)", id, (def)?"enabled":"disabled");
 634         }
 635         else {
 636                 if(*chart_id == '/') chart_id++;
 637
 638                 // disable by default the parent cgroup
 639                 // for known cgroup managers
 640                 if(!strcmp(chart_id, "lxc") ||
 641                                 !strcmp(chart_id, "docker") ||
 642                                 !strcmp(chart_id, "systemd") ||
 643                                 !strncmp(chart_id, "user.slice", 10) ||
 644                                 !strncmp(chart_id, "system.slice", 12)
 645                                 ) {
 646                         def = 0;
 647                         debug(D_CGROUP, "cgroup '%s' is container manager (by default %s)", id, (def)?"enabled":"disabled");
 648                 }
 649         }
 650
 651         struct cgroup *cg = calloc(1, sizeof(struct cgroup));
 652         if(!cg) fatal("Cannot allocate memory for cgroup '%s'", id);
 653
 654         debug(D_CGROUP, "adding cgroup '%s'", id);
 655
 656         cg->id = strdup(id);
 657         if(!cg->id) fatal("Cannot allocate memory for cgroup '%s'", id);
 658
 659         cg->hash = simple_hash(cg->id);
 660
 661         cg->chart_id = strdup(chart_id);
 662         if(!cg->chart_id) fatal("Cannot allocate memory for cgroup '%s'", id);
 663
 664         if(!cgroup_root)
 665                 cgroup_root = cg;
 666         else {
 667                 // append it
 668                 struct cgroup *e;
 669                 for(e = cgroup_root; e->next ;e = e->next) ;
 670                 e->next = cg;
 671         }
 672
 673         cgroup_root_count++;
 674
 675         // fix the name by calling the external script
 676         cgroup_get_chart_id(cg);
 677
 678         char option[FILENAME_MAX + 1];
 679         snprintf(option, FILENAME_MAX, "enable cgroup %s", chart_id);
 680         cg->enabled = config_get_boolean("plugin:cgroups", option, def);
 681
 682         // FIXME
 683         // switch this to debug(D_CGROUP, ...
 684         info("Added cgroup '%s' with chart id '%s' as %s (default was %s)", cg->id, cg->chart_id, (cg->enabled)?"enabled":"disabled", (def)?"enabled":"disabled");
 685
 686         return cg;
 687 }
 688
 689 void cgroup_free(struct cgroup *cg) {
 690         debug(D_CGROUP, "removing cgroup '%s'", cg->id);
 691
 692         // FIXME
 693         // switch this to debug(D_CGROUP, ...
 694         info("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");
 695
 696         free(cg->cpuacct_usage.cpu_percpu);
 697
 698         free(cg->cpuacct_stat.filename);
 699         free(cg->cpuacct_usage.filename);
 700         free(cg->memory.filename);
 701         free(cg->io_service_bytes.filename);
 702         free(cg->io_serviced.filename);
 703         free(cg->throttle_io_service_bytes.filename);
 704         free(cg->throttle_io_serviced.filename);
 705         free(cg->io_merged.filename);
 706         free(cg->io_queued.filename);
 707
 708         free(cg->id);
 709         free(cg->chart_id);
 710         free(cg);
 711
 712         cgroup_root_count--;
 713 }
 714
 715 // find if a given cgroup exists
 716 struct cgroup *cgroup_find(const char *id) {
 717         debug(D_CGROUP, "searching for cgroup '%s'", id);
 718
 719         uint32_t hash = simple_hash(id);
 720
 721         struct cgroup *cg;
 722         for(cg = cgroup_root; cg ; cg = cg->next) {
 723                 if(hash == cg->hash && strcmp(id, cg->id) == 0)
 724                         break;
 725         }
 726
 727         debug(D_CGROUP, "cgroup_find('%s') %s", id, (cg)?"found":"not found");
 728         return cg;
 729 }
 730
 731 // ----------------------------------------------------------------------------
 732 // detect running cgroups
 733
 734 // callback for find_file_in_subdirs()
 735 void found_subdir_in_dir(const char *dir) {
 736         debug(D_CGROUP, "examining cgroup dir '%s'", dir);
 737
 738         struct cgroup *cg = cgroup_find(dir);
 739         if(!cg) {
 740                 if(*dir && cgroup_max_depth > 0) {
 741                         int depth = 0;
 742                         const char *s;
 743
 744                         for(s = dir; *s ;s++)
 745                                 if(unlikely(*s == '/'))
 746                                         depth++;
 747
 748                         if(depth > cgroup_max_depth) {
 749                                 info("cgroup '%s' is too deep (%d, while max is %d)", dir, depth, cgroup_max_depth);
 750                                 return;
 751                         }
 752                 }
 753                 debug(D_CGROUP, "will add dir '%s' as cgroup", dir);
 754                 cg = cgroup_add(dir);
 755         }
 756
 757         if(cg) cg->available = 1;
 758 }
 759
 760 void find_dir_in_subdirs(const char *base, const char *this, void (*callback)(const char *)) {
 761         int enabled = -1;
 762         if(!this) this = base;
 763         size_t dirlen = strlen(this), baselen = strlen(base);
 764         const char *relative_path = &this[baselen];
 765
 766         DIR *dir = opendir(this);
 767         if(!dir) return;
 768
 769         callback(relative_path);
 770
 771         struct dirent *de = NULL;
 772         while((de = readdir(dir))) {
 773                 if(de->d_type == DT_DIR
 774                         && (
 775                                 (de->d_name[0] == '.' && de->d_name[1] == '\0')
 776                                 || (de->d_name[0] == '.' && de->d_name[1] == '.' && de->d_name[2] == '\0')
 777                                 ))
 778                         continue;
 779
 780                 debug(D_CGROUP, "examining '%s/%s'", this, de->d_name);
 781
 782                 if(de->d_type == DT_DIR) {
 783                         if(enabled == -1) {
 784                                 // we check for this option here
 785                                 // so that the config will not have settings
 786                                 // for leaf directories
 787                                 char option[FILENAME_MAX + 1];
 788                                 snprintf(option, FILENAME_MAX, "search for cgroups under %s", (*relative_path == '\0')?"/":relative_path);
 789
 790                                 int def = 1;
 791                                 if(!strcmp(this, "system.slice") ||
 792                                         !strcmp(this, "user.slice") ||
 793                                         !strcmp(this, "systemd"))
 794                                         def = 0;
 795
 796                                 enabled = config_get_boolean("plugin:cgroups", option, def);
 797                         }
 798
 799                         if(enabled) {
 800                                 char *s = malloc(dirlen + strlen(de->d_name) + 2);
 801                                 if(s) {
 802                                         strcpy(s, this);
 803                                         strcat(s, "/");
 804                                         strcat(s, de->d_name);
 805                                         find_dir_in_subdirs(base, s, callback);
 806                                         free(s);
 807                                 }
 808                         }
 809                 }
 810         }
 811
 812         closedir(dir);
 813 }
 814
 815 void mark_all_cgroups_as_not_available() {
 816         debug(D_CGROUP, "marking all cgroups as not available");
 817
 818         struct cgroup *cg;
 819
 820         // mark all as not available
 821         for(cg = cgroup_root; cg ; cg = cg->next)
 822                 cg->available = 0;
 823 }
 824
 825 void cleanup_all_cgroups() {
 826         struct cgroup *cg = cgroup_root, *last = NULL;
 827
 828         for(; cg ;) {
 829                 if(!cg->available) {
 830
 831                         if(!last)
 832                                 cgroup_root = cg->next;
 833                         else
 834                                 last->next = cg->next;
 835
 836                         cgroup_free(cg);
 837
 838                         if(!last)
 839                                 cg = cgroup_root;
 840                         else
 841                                 cg = last->next;
 842                 }
 843                 else {
 844                         last = cg;
 845                         cg = cg->next;
 846                 }
 847         }
 848 }
 849
 850 void find_all_cgroups() {
 851         debug(D_CGROUP, "searching for cgroups");
 852
 853         mark_all_cgroups_as_not_available();
 854
 855         if(cgroup_enable_cpuacct_stat || cgroup_enable_cpuacct_usage)
 856                 find_dir_in_subdirs(cgroup_cpuacct_base, NULL, found_subdir_in_dir);
 857
 858         if(cgroup_enable_blkio)
 859                 find_dir_in_subdirs(cgroup_blkio_base, NULL, found_subdir_in_dir);
 860
 861         if(cgroup_enable_memory)
 862                 find_dir_in_subdirs(cgroup_memory_base, NULL, found_subdir_in_dir);
 863
 864         // remove any non-existing cgroups
 865         cleanup_all_cgroups();
 866
 867         struct cgroup *cg;
 868         for(cg = cgroup_root; cg ; cg = cg->next) {
 869                 // fprintf(stderr, " >>> CGROUP '%s' (%u - %s) with name '%s'\n", cg->id, cg->hash, cg->available?"available":"stopped", cg->name);
 870
 871                 if(unlikely(!cg->available))
 872                         continue;
 873
 874                 debug(D_CGROUP, "checking paths for cgroup '%s'", cg->id);
 875
 876                 // check for newly added cgroups
 877                 // and update the filenames they read
 878                 char filename[FILENAME_MAX + 1];
 879                 if(cgroup_enable_cpuacct_stat && !cg->cpuacct_stat.filename) {
 880                         snprintf(filename, FILENAME_MAX, "%s%s/cpuacct.stat", cgroup_cpuacct_base, cg->id);
 881                         cg->cpuacct_stat.filename = strdup(filename);
 882                         debug(D_CGROUP, "cpuacct.stat filename for cgroup '%s': '%s'", cg->id, cg->cpuacct_stat.filename);
 883                 }
 884                 if(cgroup_enable_cpuacct_usage && !cg->cpuacct_usage.filename) {
 885                         snprintf(filename, FILENAME_MAX, "%s%s/cpuacct.usage_percpu", cgroup_cpuacct_base, cg->id);
 886                         cg->cpuacct_usage.filename = strdup(filename);
 887                         debug(D_CGROUP, "cpuacct.usage_percpu filename for cgroup '%s': '%s'", cg->id, cg->cpuacct_usage.filename);
 888                 }
 889                 if(cgroup_enable_memory && !cg->memory.filename) {
 890                         snprintf(filename, FILENAME_MAX, "%s%s/memory.stat", cgroup_memory_base, cg->id);
 891                         cg->memory.filename = strdup(filename);
 892                         debug(D_CGROUP, "memory.stat filename for cgroup '%s': '%s'", cg->id, cg->memory.filename);
 893                 }
 894                 if(cgroup_enable_blkio) {
 895                         if(!cg->io_service_bytes.filename) {
 896                                 snprintf(filename, FILENAME_MAX, "%s%s/blkio.io_service_bytes", cgroup_blkio_base, cg->id);
 897                                 cg->io_service_bytes.filename = strdup(filename);
 898                                 debug(D_CGROUP, "io_service_bytes filename for cgroup '%s': '%s'", cg->id, cg->io_service_bytes.filename);
 899                         }
 900                         if(!cg->io_serviced.filename) {
 901                                 snprintf(filename, FILENAME_MAX, "%s%s/blkio.io_serviced", cgroup_blkio_base, cg->id);
 902                                 cg->io_serviced.filename = strdup(filename);
 903                                 debug(D_CGROUP, "io_serviced filename for cgroup '%s': '%s'", cg->id, cg->io_serviced.filename);
 904                         }
 905                         if(!cg->throttle_io_service_bytes.filename) {
 906                                 snprintf(filename, FILENAME_MAX, "%s%s/blkio.throttle.io_service_bytes", cgroup_blkio_base, cg->id);
 907                                 cg->throttle_io_service_bytes.filename = strdup(filename);
 908                                 debug(D_CGROUP, "throttle_io_service_bytes filename for cgroup '%s': '%s'", cg->id, cg->throttle_io_service_bytes.filename);
 909                         }
 910                         if(!cg->throttle_io_serviced.filename) {
 911                                 snprintf(filename, FILENAME_MAX, "%s%s/blkio.throttle.io_serviced", cgroup_blkio_base, cg->id);
 912                                 cg->throttle_io_serviced.filename = strdup(filename);
 913                                 debug(D_CGROUP, "throttle_io_serviced filename for cgroup '%s': '%s'", cg->id, cg->throttle_io_serviced.filename);
 914                         }
 915                         if(!cg->io_merged.filename) {
 916                                 snprintf(filename, FILENAME_MAX, "%s%s/blkio.io_merged", cgroup_blkio_base, cg->id);
 917                                 cg->io_merged.filename = strdup(filename);
 918                                 debug(D_CGROUP, "io_merged filename for cgroup '%s': '%s'", cg->id, cg->io_merged.filename);
 919                         }
 920                         if(!cg->io_queued.filename) {
 921                                 snprintf(filename, FILENAME_MAX, "%s%s/blkio.io_queued", cgroup_blkio_base, cg->id);
 922                                 cg->io_queued.filename = strdup(filename);
 923                                 debug(D_CGROUP, "io_queued filename for cgroup '%s': '%s'", cg->id, cg->io_queued.filename);
 924                         }
 925                 }
 926         }
 927
 928         debug(D_CGROUP, "done searching for cgroups");
 929         return;
 930 }
 931
 932 // ----------------------------------------------------------------------------
 933 // generate charts
 934
 935 #define CHART_TITLE_MAX 300
 936
 937 void update_cgroup_charts(int update_every) {
 938         debug(D_CGROUP, "updating cgroups charts");
 939
 940         char type[RRD_ID_LENGTH_MAX + 1];
 941         char title[CHART_TITLE_MAX + 1];
 942
 943         struct cgroup *cg;
 944         RRDSET *st;
 945
 946         for(cg = cgroup_root; cg ; cg = cg->next) {
 947                 if(!cg->available || !cg->enabled)
 948                         continue;
 949
 950                 if(cg->id[0] == '\0')
 951                         strcpy(type, "cgroup_host");
 952                 else if(cg->id[0] == '/')
 953                         snprintf(type, RRD_ID_LENGTH_MAX, "cgroup_%s", cg->chart_id);
 954                 else
 955                         snprintf(type, RRD_ID_LENGTH_MAX, "cgroup_%s", cg->chart_id);
 956
 957                 netdata_fix_chart_id(type);
 958
 959                 if(cg->cpuacct_stat.updated) {
 960                         st = rrdset_find_bytype(type, "cpu");
 961                         if(!st) {
 962                                 snprintf(title, CHART_TITLE_MAX, "CPU Usage for cgroup %s", cg->chart_id);
 963                                 st = rrdset_create(type, "cpu", NULL, "cpu", "cgroup.cpu", title, "%", 40000, update_every, RRDSET_TYPE_STACKED);
 964
 965                                 rrddim_add(st, "user", NULL, 100, hz, RRDDIM_INCREMENTAL);
 966                                 rrddim_add(st, "system", NULL, 100, hz, RRDDIM_INCREMENTAL);
 967                         }
 968                         else rrdset_next(st);
 969
 970                         rrddim_set(st, "user", cg->cpuacct_stat.user);
 971                         rrddim_set(st, "system", cg->cpuacct_stat.system);
 972                         rrdset_done(st);
 973                 }
 974
 975                 if(cg->cpuacct_usage.updated) {
 976                         char id[RRD_ID_LENGTH_MAX + 1];
 977                         unsigned int i;
 978
 979                         st = rrdset_find_bytype(type, "cpu_per_core");
 980                         if(!st) {
 981                                 snprintf(title, CHART_TITLE_MAX, "CPU Usage Per Core for cgroup %s", cg->chart_id);
 982                                 st = rrdset_create(type, "cpu_per_core", NULL, "cpu", "cgroup.cpu_per_core", title, "%", 40100, update_every, RRDSET_TYPE_STACKED);
 983
 984                                 for(i = 0; i < cg->cpuacct_usage.cpus ;i++) {
 985                                         snprintf(id, CHART_TITLE_MAX, "cpu%d", i);
 986                                         rrddim_add(st, id, NULL, 100, 1000000, RRDDIM_INCREMENTAL);
 987                                 }
 988                         }
 989                         else rrdset_next(st);
 990
 991                         for(i = 0; i < cg->cpuacct_usage.cpus ;i++) {
 992                                 snprintf(id, CHART_TITLE_MAX, "cpu%d", i);
 993                                 rrddim_set(st, id, cg->cpuacct_usage.cpu_percpu[i]);
 994                         }
 995                         rrdset_done(st);
 996                 }
 997
 998                 if(cg->memory.updated) {
 999                         if(cg->memory.cache + cg->memory.rss + cg->memory.rss_huge + cg->memory.mapped_file > 0) {
1000                                 st = rrdset_find_bytype(type, "mem");
1001                                 if(!st) {
1002                                         snprintf(title, CHART_TITLE_MAX, "Memory Usage for cgroup %s", cg->chart_id);
1003                                         st = rrdset_create(type, "mem", NULL, "mem", "cgroup.mem", title, "MB", 40200, update_every,
1004                                                            RRDSET_TYPE_STACKED);
1005
1006                                         rrddim_add(st, "cache", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1007                                         rrddim_add(st, "rss", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1008                                         if(cg->memory.has_dirty_swap)
1009                                                 rrddim_add(st, "swap", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1010                                         rrddim_add(st, "rss_huge", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1011                                         rrddim_add(st, "mapped_file", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1012                                 }
1013                                 else rrdset_next(st);
1014
1015                                 rrddim_set(st, "cache", cg->memory.cache);
1016                                 rrddim_set(st, "rss", cg->memory.rss);
1017                                 if(cg->memory.has_dirty_swap)
1018                                         rrddim_set(st, "swap", cg->memory.swap);
1019                                 rrddim_set(st, "rss_huge", cg->memory.rss_huge);
1020                                 rrddim_set(st, "mapped_file", cg->memory.mapped_file);
1021                                 rrdset_done(st);
1022                         }
1023
1024                         st = rrdset_find_bytype(type, "writeback");
1025                         if(!st) {
1026                                 snprintf(title, CHART_TITLE_MAX, "Writeback Memory for cgroup %s", cg->chart_id);
1027                                 st = rrdset_create(type, "writeback", NULL, "mem", "cgroup.writeback", title, "MB", 40300,
1028                                                    update_every, RRDSET_TYPE_AREA);
1029
1030                                 if(cg->memory.has_dirty_swap)
1031                                         rrddim_add(st, "dirty", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1032                                 rrddim_add(st, "writeback", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1033                         }
1034                         else rrdset_next(st);
1035
1036                         if(cg->memory.has_dirty_swap)
1037                                 rrddim_set(st, "dirty", cg->memory.dirty);
1038                         rrddim_set(st, "writeback", cg->memory.writeback);
1039                         rrdset_done(st);
1040
1041                         if(cg->memory.pgpgin + cg->memory.pgpgout > 0) {
1042                                 st = rrdset_find_bytype(type, "mem_activity");
1043                                 if(!st) {
1044                                         snprintf(title, CHART_TITLE_MAX, "Memory Activity for cgroup %s", cg->chart_id);
1045                                         st = rrdset_create(type, "mem_activity", NULL, "mem", "cgroup.mem_activity", title, "MB/s",
1046                                                            40400, update_every, RRDSET_TYPE_LINE);
1047
1048                                         rrddim_add(st, "pgpgin", "in", sysconf(_SC_PAGESIZE), 1024 * 1024, RRDDIM_INCREMENTAL);
1049                                         rrddim_add(st, "pgpgout", "out", -sysconf(_SC_PAGESIZE), 1024 * 1024, RRDDIM_INCREMENTAL);
1050                                 }
1051                                 else rrdset_next(st);
1052
1053                                 rrddim_set(st, "pgpgin", cg->memory.pgpgin);
1054                                 rrddim_set(st, "pgpgout", cg->memory.pgpgout);
1055                                 rrdset_done(st);
1056                         }
1057
1058                         if(cg->memory.pgfault + cg->memory.pgmajfault > 0) {
1059                                 st = rrdset_find_bytype(type, "pgfaults");
1060                                 if(!st) {
1061                                         snprintf(title, CHART_TITLE_MAX, "Memory Page Faults for cgroup %s", cg->chart_id);
1062                                         st = rrdset_create(type, "pgfaults", NULL, "mem", "cgroup.pgfaults", title, "MB/s", 40500,
1063                                                            update_every, RRDSET_TYPE_LINE);
1064
1065                                         rrddim_add(st, "pgfault", NULL, sysconf(_SC_PAGESIZE), 1024 * 1024, RRDDIM_INCREMENTAL);
1066                                         rrddim_add(st, "pgmajfault", "swap", -sysconf(_SC_PAGESIZE), 1024 * 1024, RRDDIM_INCREMENTAL);
1067                                 }
1068                                 else rrdset_next(st);
1069
1070                                 rrddim_set(st, "pgfault", cg->memory.pgfault);
1071                                 rrddim_set(st, "pgmajfault", cg->memory.pgmajfault);
1072                                 rrdset_done(st);
1073                         }
1074                 }
1075
1076                 if(cg->io_service_bytes.updated && cg->io_service_bytes.Read + cg->io_service_bytes.Write > 0) {
1077                         st = rrdset_find_bytype(type, "io");
1078                         if(!st) {
1079                                 snprintf(title, CHART_TITLE_MAX, "I/O Bandwidth (all disks) for cgroup %s", cg->chart_id);
1080                                 st = rrdset_create(type, "io", NULL, "disk", "cgroup.io", title, "KB/s", 41200,
1081                                                    update_every, RRDSET_TYPE_LINE);
1082
1083                                 rrddim_add(st, "read", NULL, 1, 1024, RRDDIM_INCREMENTAL);
1084                                 rrddim_add(st, "write", NULL, -1, 1024, RRDDIM_INCREMENTAL);
1085                         }
1086                         else rrdset_next(st);
1087
1088                         rrddim_set(st, "read", cg->io_service_bytes.Read);
1089                         rrddim_set(st, "write", cg->io_service_bytes.Write);
1090                         rrdset_done(st);
1091                 }
1092
1093                 if(cg->io_serviced.updated && cg->io_serviced.Read + cg->io_serviced.Write > 0) {
1094                         st = rrdset_find_bytype(type, "serviced_ops");
1095                         if(!st) {
1096                                 snprintf(title, CHART_TITLE_MAX, "Serviced I/O Operations (all disks) for cgroup %s", cg->chart_id);
1097                                 st = rrdset_create(type, "serviced_ops", NULL, "disk", "cgroup.serviced_ops", title, "operations/s", 41200,
1098                                                    update_every, RRDSET_TYPE_LINE);
1099
1100                                 rrddim_add(st, "read", NULL, 1, 1, RRDDIM_INCREMENTAL);
1101                                 rrddim_add(st, "write", NULL, -1, 1, RRDDIM_INCREMENTAL);
1102                         }
1103                         else rrdset_next(st);
1104
1105                         rrddim_set(st, "read", cg->io_serviced.Read);
1106                         rrddim_set(st, "write", cg->io_serviced.Write);
1107                         rrdset_done(st);
1108                 }
1109
1110                 if(cg->throttle_io_service_bytes.updated && cg->throttle_io_service_bytes.Read + cg->throttle_io_service_bytes.Write > 0) {
1111                         st = rrdset_find_bytype(type, "io");
1112                         if(!st) {
1113                                 snprintf(title, CHART_TITLE_MAX, "Throttle I/O Bandwidth (all disks) for cgroup %s", cg->chart_id);
1114                                 st = rrdset_create(type, "io", NULL, "disk", "cgroup.io", title, "KB/s", 41200,
1115                                                    update_every, RRDSET_TYPE_LINE);
1116
1117                                 rrddim_add(st, "read", NULL, 1, 1024, RRDDIM_INCREMENTAL);
1118                                 rrddim_add(st, "write", NULL, -1, 1024, RRDDIM_INCREMENTAL);
1119                         }
1120                         else rrdset_next(st);
1121
1122                         rrddim_set(st, "read", cg->throttle_io_service_bytes.Read);
1123                         rrddim_set(st, "write", cg->throttle_io_service_bytes.Write);
1124                         rrdset_done(st);
1125                 }
1126
1127
1128                 if(cg->throttle_io_serviced.updated && cg->throttle_io_serviced.Read + cg->throttle_io_serviced.Write > 0) {
1129                         st = rrdset_find_bytype(type, "throttle_serviced_ops");
1130                         if(!st) {
1131                                 snprintf(title, CHART_TITLE_MAX, "Throttle Serviced I/O Operations (all disks) for cgroup %s", cg->chart_id);
1132                                 st = rrdset_create(type, "throttle_serviced_ops", NULL, "disk", "cgroup.throttle_serviced_ops", title, "operations/s", 41200,
1133                                                    update_every, RRDSET_TYPE_LINE);
1134
1135                                 rrddim_add(st, "read", NULL, 1, 1, RRDDIM_INCREMENTAL);
1136                                 rrddim_add(st, "write", NULL, -1, 1, RRDDIM_INCREMENTAL);
1137                         }
1138                         else rrdset_next(st);
1139
1140                         rrddim_set(st, "read", cg->throttle_io_serviced.Read);
1141                         rrddim_set(st, "write", cg->throttle_io_serviced.Write);
1142                         rrdset_done(st);
1143                 }
1144
1145                 if(cg->io_queued.updated) {
1146                         st = rrdset_find_bytype(type, "queued_ops");
1147                         if(!st) {
1148                                 snprintf(title, CHART_TITLE_MAX, "Queued I/O Operations (all disks) for cgroup %s", cg->chart_id);
1149                                 st = rrdset_create(type, "queued_ops", NULL, "disk", "cgroup.queued_ops", title, "operations", 42000,
1150                                                    update_every, RRDSET_TYPE_LINE);
1151
1152                                 rrddim_add(st, "read", NULL, 1, 1, RRDDIM_ABSOLUTE);
1153                                 rrddim_add(st, "write", NULL, -1, 1, RRDDIM_ABSOLUTE);
1154                         }
1155                         else rrdset_next(st);
1156
1157                         rrddim_set(st, "read", cg->io_queued.Read);
1158                         rrddim_set(st, "write", cg->io_queued.Write);
1159                         rrdset_done(st);
1160                 }
1161
1162                 if(cg->io_merged.updated && cg->io_merged.Read + cg->io_merged.Write > 0) {
1163                         st = rrdset_find_bytype(type, "merged_ops");
1164                         if(!st) {
1165                                 snprintf(title, CHART_TITLE_MAX, "Merged I/O Operations (all disks) for cgroup %s", cg->chart_id);
1166                                 st = rrdset_create(type, "merged_ops", NULL, "disk", "cgroup.merged_ops", title, "operations/s", 42100,
1167                                                    update_every, RRDSET_TYPE_LINE);
1168
1169                                 rrddim_add(st, "read", NULL, 1, 1024, RRDDIM_INCREMENTAL);
1170                                 rrddim_add(st, "write", NULL, -1, 1024, RRDDIM_INCREMENTAL);
1171                         }
1172                         else rrdset_next(st);
1173
1174                         rrddim_set(st, "read", cg->io_merged.Read);
1175                         rrddim_set(st, "write", cg->io_merged.Write);
1176                         rrdset_done(st);
1177                 }
1178         }
1179
1180         debug(D_CGROUP, "done updating cgroups charts");
1181 }
1182
1183 // ----------------------------------------------------------------------------
1184 // cgroups main
1185
1186 int do_sys_fs_cgroup(int update_every, unsigned long long dt) {
1187         static int cgroup_global_config_read = 0;
1188         static time_t last_run = 0;
1189         time_t now = time(NULL);
1190
1191         if(dt) {};
1192
1193         if(unlikely(!cgroup_global_config_read)) {
1194                 read_cgroup_plugin_configuration();
1195                 cgroup_global_config_read = 1;
1196         }
1197
1198         if(unlikely(cgroup_enable_new_cgroups_detected_at_runtime && now - last_run > cgroup_check_for_new_every)) {
1199                 find_all_cgroups();
1200                 last_run = now;
1201         }
1202
1203         read_all_cgroups(cgroup_root);
1204         update_cgroup_charts(update_every);
1205
1206         return 0;
1207 }
1208
1209 void *cgroups_main(void *ptr)
1210 {
1211         if(ptr) { ; }
1212
1213         info("CGROUP Plugin thread created with task id %d", gettid());
1214
1215         if(pthread_setcanceltype(PTHREAD_CANCEL_DEFERRED, NULL) != 0)
1216                 error("Cannot set pthread cancel type to DEFERRED.");
1217
1218         if(pthread_setcancelstate(PTHREAD_CANCEL_ENABLE, NULL) != 0)
1219                 error("Cannot set pthread cancel state to ENABLE.");
1220
1221         struct rusage thread;
1222
1223         // when ZERO, attempt to do it
1224         int vdo_sys_fs_cgroup                   = 0;
1225         int vdo_cpu_netdata                     = !config_get_boolean("plugin:cgroups", "cgroups plugin resources", 1);
1226
1227         // keep track of the time each module was called
1228         unsigned long long sutime_sys_fs_cgroup = 0ULL;
1229
1230         // the next time we will run - aligned properly
1231         unsigned long long sunext = (time(NULL) - (time(NULL) % rrd_update_every) + rrd_update_every) * 1000000ULL;
1232         unsigned long long sunow;
1233
1234         RRDSET *stcpu_thread = NULL;
1235
1236         for(;1;) {
1237                 if(unlikely(netdata_exit)) break;
1238
1239                 // delay until it is our time to run
1240                 while((sunow = timems()) < sunext)
1241                         usleep((useconds_t)(sunext - sunow));
1242
1243                 // find the next time we need to run
1244                 while(timems() > sunext)
1245                         sunext += rrd_update_every * 1000000ULL;
1246
1247                 if(unlikely(netdata_exit)) break;
1248
1249                 // BEGIN -- the job to be done
1250
1251                 if(!vdo_sys_fs_cgroup) {
1252                         debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_sys_fs_cgroup().");
1253                         sunow = timems();
1254                         vdo_sys_fs_cgroup = do_sys_fs_cgroup(rrd_update_every, (sutime_sys_fs_cgroup > 0)?sunow - sutime_sys_fs_cgroup:0ULL);
1255                         sutime_sys_fs_cgroup = sunow;
1256                 }
1257                 if(unlikely(netdata_exit)) break;
1258
1259                 // END -- the job is done
1260
1261                 // --------------------------------------------------------------------
1262
1263                 if(!vdo_cpu_netdata) {
1264                         getrusage(RUSAGE_THREAD, &thread);
1265
1266                         if(!stcpu_thread) stcpu_thread = rrdset_find("netdata.plugin_cgroups_cpu");
1267                         if(!stcpu_thread) {
1268                                 stcpu_thread = rrdset_create("netdata", "plugin_cgroups_cpu", NULL, "proc.internal", NULL, "NetData CGroups Plugin CPU usage", "milliseconds/s", 131000, rrd_update_every, RRDSET_TYPE_STACKED);
1269
1270                                 rrddim_add(stcpu_thread, "user",  NULL,  1, 1000, RRDDIM_INCREMENTAL);
1271                                 rrddim_add(stcpu_thread, "system", NULL, 1, 1000, RRDDIM_INCREMENTAL);
1272                         }
1273                         else rrdset_next(stcpu_thread);
1274
1275                         rrddim_set(stcpu_thread, "user"  , thread.ru_utime.tv_sec * 1000000ULL + thread.ru_utime.tv_usec);
1276                         rrddim_set(stcpu_thread, "system", thread.ru_stime.tv_sec * 1000000ULL + thread.ru_stime.tv_usec);
1277                         rrdset_done(stcpu_thread);
1278                 }
1279         }
1280
1281         pthread_exit(NULL);
1282         return NULL;
1283 }