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