src/processes.c

   1 /**
   2  * collectd - src/processes.c
   3  * Copyright (C) 2005  Lyonel Vincent
   4  * Copyright (C) 2006  Florian Forster (Mach code)
   5  *
   6  * This program is free software; you can redistribute it and/or modify it
   7  * under the terms of the GNU General Public License as published by the
   8  * Free Software Foundation; either version 2 of the License, or (at your
   9  * option) any later version.
  10  *
  11  * This program is distributed in the hope that it will be useful, but
  12  * WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU General Public License along
  17  * with this program; if not, write to the Free Software Foundation, Inc.,
  18  * 51 Franklin St, Fifth Floor, Boston, MA  02110-1301 USA
  19  *
  20  * Authors:
  21  *   Lyonel Vincent <lyonel at ezix.org>
  22  *   Florian octo Forster <octo at verplant.org>
  23  **/
  24
  25 #include "collectd.h"
  26 #include "common.h"
  27 #include "plugin.h"
  28 #include "utils_debug.h"
  29 #include "configfile.h"
  30
  31 /* Include header files for the mach system, if they exist.. */
  32 #if HAVE_THREAD_INFO
  33 #  if HAVE_MACH_MACH_INIT_H
  34 #    include <mach/mach_init.h>
  35 #  endif
  36 #  if HAVE_MACH_HOST_PRIV_H
  37 #    include <mach/host_priv.h>
  38 #  endif
  39 #  if HAVE_MACH_MACH_ERROR_H
  40 #    include <mach/mach_error.h>
  41 #  endif
  42 #  if HAVE_MACH_MACH_HOST_H
  43 #    include <mach/mach_host.h>
  44 #  endif
  45 #  if HAVE_MACH_MACH_PORT_H
  46 #    include <mach/mach_port.h>
  47 #  endif
  48 #  if HAVE_MACH_MACH_TYPES_H
  49 #    include <mach/mach_types.h>
  50 #  endif
  51 #  if HAVE_MACH_MESSAGE_H
  52 #    include <mach/message.h>
  53 #  endif
  54 #  if HAVE_MACH_PROCESSOR_SET_H
  55 #    include <mach/processor_set.h>
  56 #  endif
  57 #  if HAVE_MACH_TASK_H
  58 #    include <mach/task.h>
  59 #  endif
  60 #  if HAVE_MACH_THREAD_ACT_H
  61 #    include <mach/thread_act.h>
  62 #  endif
  63 #  if HAVE_MACH_VM_REGION_H
  64 #    include <mach/vm_region.h>
  65 #  endif
  66 #  if HAVE_MACH_VM_MAP_H
  67 #    include <mach/vm_map.h>
  68 #  endif
  69 #  if HAVE_MACH_VM_PROT_H
  70 #    include <mach/vm_prot.h>
  71 #  endif
  72 #  if HAVE_SYS_SYSCTL_H
  73 #    include <sys/sysctl.h>
  74 #  endif
  75 /* #endif HAVE_THREAD_INFO */
  76
  77 #elif KERNEL_LINUX
  78 #  if HAVE_LINUX_CONFIG_H
  79 #    include <linux/config.h>
  80 #  endif
  81 #  ifndef CONFIG_HZ
  82 #    define CONFIG_HZ 100
  83 #  endif
  84 #endif /* KERNEL_LINUX */
  85
  86 #define MODULE_NAME "processes"
  87
  88 #if HAVE_THREAD_INFO || KERNEL_LINUX
  89 # define PROCESSES_HAVE_READ 1
  90 #else
  91 # define PROCESSES_HAVE_READ 0
  92 #endif
  93
  94 #define BUFSIZE 256
  95
  96 static char *processes_file = "processes.rrd";
  97 static char *processes_ds_def[] =
  98 {
  99         "DS:running:GAUGE:"COLLECTD_HEARTBEAT":0:65535",
 100         "DS:sleeping:GAUGE:"COLLECTD_HEARTBEAT":0:65535",
 101         "DS:zombies:GAUGE:"COLLECTD_HEARTBEAT":0:65535",
 102         "DS:stopped:GAUGE:"COLLECTD_HEARTBEAT":0:65535",
 103         "DS:paging:GAUGE:"COLLECTD_HEARTBEAT":0:65535",
 104         "DS:blocked:GAUGE:"COLLECTD_HEARTBEAT":0:65535",
 105         NULL
 106 };
 107 static int processes_ds_num = 6;
 108
 109 static char *ps_rss_file = "processes/ps_rss-%s.rrd";
 110 static char *ps_rss_ds_def[] =
 111 {
 112         /* max = 2^63 - 1 */
 113         "DS:byte:GAUGE:"COLLECTD_HEARTBEAT":0:9223372036854775807",
 114         NULL
 115 };
 116 static int ps_rss_ds_num = 1;
 117
 118 static char *ps_cputime_file = "processes/ps_cputime-%s.rrd";
 119 static char *ps_cputime_ds_def[] =
 120 {
 121         /* 1 second in user-mode per second ought to be enough.. */
 122         "DS:user:COUNTER:"COLLECTD_HEARTBEAT":0:1000000",
 123         "DS:syst:COUNTER:"COLLECTD_HEARTBEAT":0:1000000",
 124         NULL
 125 };
 126 static int ps_cputime_ds_num = 2;
 127
 128 static char *ps_count_file = "processes/ps_count-%s.rrd";
 129 static char *ps_count_ds_def[] =
 130 {
 131         "DS:processes:GAUGE:"COLLECTD_HEARTBEAT":0:65535",
 132         "DS:threads:GAUGE:"COLLECTD_HEARTBEAT":0:65535",
 133         NULL
 134 };
 135 static int ps_count_ds_num = 2;
 136
 137 static char *ps_pagefaults_file = "processes/ps_pagefaults-%s.rrd";
 138 static char *ps_pagefaults_ds_def[] =
 139 {
 140         /* max = 2^63 - 1 */
 141         "DS:minflt:COUNTER:"COLLECTD_HEARTBEAT":0:9223372036854775807",
 142         "DS:majflt:COUNTER:"COLLECTD_HEARTBEAT":0:9223372036854775807",
 143         NULL
 144 };
 145 static int ps_pagefaults_ds_num = 2;
 146
 147 static char *config_keys[] =
 148 {
 149         "Process",
 150         NULL
 151 };
 152 static int config_keys_num = 1;
 153
 154 typedef struct procstat_entry_s
 155 {
 156         unsigned long id;
 157         unsigned long age;
 158
 159         unsigned long num_proc;
 160         unsigned long num_lwp;
 161         unsigned long vmem_rss;
 162
 163         unsigned long vmem_minflt;
 164         unsigned long vmem_majflt;
 165         unsigned long vmem_minflt_counter;
 166         unsigned long vmem_majflt_counter;
 167
 168         unsigned long cpu_user;
 169         unsigned long cpu_system;
 170         unsigned long cpu_user_counter;
 171         unsigned long cpu_system_counter;
 172
 173         struct procstat_entry_s *next;
 174 } procstat_entry_t;
 175
 176 #define PROCSTAT_NAME_LEN 256
 177 typedef struct procstat
 178 {
 179         char          name[PROCSTAT_NAME_LEN];
 180
 181         unsigned long num_proc;
 182         unsigned long num_lwp;
 183         unsigned long vmem_rss;
 184
 185         unsigned long vmem_minflt_counter;
 186         unsigned long vmem_majflt_counter;
 187
 188         unsigned long cpu_user_counter;
 189         unsigned long cpu_system_counter;
 190
 191         struct procstat   *next;
 192         struct procstat_entry_s *instances;
 193 } procstat_t;
 194
 195 static procstat_t *list_head_g = NULL;
 196
 197 #if HAVE_THREAD_INFO
 198 static mach_port_t port_host_self;
 199 static mach_port_t port_task_self;
 200
 201 static processor_set_name_array_t pset_list;
 202 static mach_msg_type_number_t     pset_list_len;
 203 /* #endif HAVE_THREAD_INFO */
 204
 205 #elif KERNEL_LINUX
 206 static long pagesize_g;
 207 #endif /* KERNEL_LINUX */
 208
 209 #if HAVE_THREAD_INFO | KERNEL_LINUX
 210 static void ps_list_register (const char *name)
 211 {
 212         procstat_t *new;
 213         procstat_t *ptr;
 214
 215         if ((new = (procstat_t *) malloc (sizeof (procstat_t))) == NULL)
 216                 return;
 217         memset (new, 0, sizeof (procstat_t));
 218         strncpy (new->name, name, PROCSTAT_NAME_LEN);
 219
 220         for (ptr = list_head_g; ptr != NULL; ptr = ptr->next)
 221         {
 222                 if (strcmp (ptr->name, name) == 0)
 223                         return;
 224                 if (ptr->next == NULL)
 225                         break;
 226         }
 227
 228         if (ptr == NULL)
 229                 list_head_g = new;
 230         else
 231                 ptr->next = new;
 232 }
 233
 234 static procstat_t *ps_list_search (const char *name)
 235 {
 236         procstat_t *ptr;
 237
 238         for (ptr = list_head_g; ptr != NULL; ptr = ptr->next)
 239                 if (strcmp (ptr->name, name) == 0)
 240                         break;
 241
 242         return (ptr);
 243 }
 244
 245 static void ps_list_add (const char *name, procstat_entry_t *entry)
 246 {
 247         procstat_t *ps;
 248         procstat_entry_t *pse;
 249
 250         if (entry->id == 0)
 251                 return;
 252
 253         if ((ps = ps_list_search (name)) == NULL)
 254                 return;
 255
 256         for (pse = ps->instances; pse != NULL; pse = pse->next)
 257                 if ((pse->id == entry->id) || (pse->next == NULL))
 258                         break;
 259
 260         if ((pse == NULL) || (pse->id != entry->id))
 261         {
 262                 procstat_entry_t *new;
 263
 264                 new = (procstat_entry_t *) malloc (sizeof (procstat_entry_t));
 265                 if (new == NULL)
 266                         return;
 267                 memset (new, 0, sizeof (procstat_entry_t));
 268                 new->id = entry->id;
 269
 270                 if (pse == NULL)
 271                         ps->instances = new;
 272                 else
 273                         pse->next = new;
 274
 275                 pse = new;
 276         }
 277
 278         pse->age = 0;
 279         pse->num_proc = entry->num_proc;
 280         pse->num_lwp  = entry->num_lwp;
 281         pse->vmem_rss = entry->vmem_rss;
 282
 283         ps->num_proc += pse->num_proc;
 284         ps->num_lwp  += pse->num_lwp;
 285         ps->vmem_rss += pse->vmem_rss;
 286
 287         if ((entry->vmem_minflt_counter == 0)
 288                         && (entry->vmem_majflt_counter == 0))
 289         {
 290                 pse->vmem_minflt_counter += entry->vmem_minflt;
 291                 pse->vmem_minflt = entry->vmem_minflt;
 292
 293                 pse->vmem_majflt_counter += entry->vmem_majflt;
 294                 pse->vmem_majflt = entry->vmem_majflt;
 295         }
 296         else
 297         {
 298                 if (entry->vmem_minflt_counter < pse->vmem_minflt_counter)
 299                 {
 300                         pse->vmem_minflt = entry->vmem_minflt_counter
 301                                 + (ULONG_MAX - pse->vmem_minflt_counter);
 302                 }
 303                 else
 304                 {
 305                         pse->vmem_minflt = entry->vmem_minflt_counter - pse->vmem_minflt_counter;
 306                 }
 307                 pse->vmem_minflt_counter = entry->vmem_minflt_counter;
 308
 309                 if (entry->vmem_majflt_counter < pse->vmem_majflt_counter)
 310                 {
 311                         pse->vmem_majflt = entry->vmem_majflt_counter
 312                                 + (ULONG_MAX - pse->vmem_majflt_counter);
 313                 }
 314                 else
 315                 {
 316                         pse->vmem_majflt = entry->vmem_majflt_counter - pse->vmem_majflt_counter;
 317                 }
 318                 pse->vmem_majflt_counter = entry->vmem_majflt_counter;
 319         }
 320
 321         ps->vmem_minflt_counter += pse->vmem_minflt;
 322         ps->vmem_majflt_counter += pse->vmem_majflt;
 323
 324         if ((entry->cpu_user_counter == 0)
 325                         && (entry->cpu_system_counter == 0))
 326         {
 327                 pse->cpu_user_counter += entry->cpu_user;
 328                 pse->cpu_user = entry->cpu_user;
 329
 330                 pse->cpu_system_counter += entry->cpu_system;
 331                 pse->cpu_system = entry->cpu_system;
 332         }
 333         else
 334         {
 335                 if (entry->cpu_user_counter < pse->cpu_user_counter)
 336                 {
 337                         pse->cpu_user = entry->cpu_user_counter
 338                                 + (ULONG_MAX - pse->cpu_user_counter);
 339                 }
 340                 else
 341                 {
 342                         pse->cpu_user = entry->cpu_user_counter - pse->cpu_user_counter;
 343                 }
 344                 pse->cpu_user_counter = entry->cpu_user_counter;
 345
 346                 if (entry->cpu_system_counter < pse->cpu_system_counter)
 347                 {
 348                         pse->cpu_system = entry->cpu_system_counter
 349                                 + (ULONG_MAX - pse->cpu_system_counter);
 350                 }
 351                 else
 352                 {
 353                         pse->cpu_system = entry->cpu_system_counter - pse->cpu_system_counter;
 354                 }
 355                 pse->cpu_system_counter = entry->cpu_system_counter;
 356         }
 357
 358         ps->cpu_user_counter   += pse->cpu_user;
 359         ps->cpu_system_counter += pse->cpu_system;
 360 }
 361
 362 static void ps_list_reset (void)
 363 {
 364         procstat_t *ps;
 365         procstat_entry_t *pse;
 366         procstat_entry_t *pse_prev;
 367
 368         for (ps = list_head_g; ps != NULL; ps = ps->next)
 369         {
 370                 ps->num_proc    = 0;
 371                 ps->num_lwp     = 0;
 372                 ps->vmem_rss    = 0;
 373
 374                 pse_prev = NULL;
 375                 pse = ps->instances;
 376                 while (pse != NULL)
 377                 {
 378                         if (pse->age > 10)
 379                         {
 380                                 DBG ("Removing this procstat entry cause it's too old: "
 381                                                 "id = %lu; name = %s;",
 382                                                 pse->id, ps->name);
 383
 384                                 if (pse_prev == NULL)
 385                                 {
 386                                         ps->instances = pse->next;
 387                                         free (pse);
 388                                         pse = ps->instances;
 389                                 }
 390                                 else
 391                                 {
 392                                         pse_prev->next = pse->next;
 393                                         free (pse);
 394                                         pse = pse_prev->next;
 395                                 }
 396                         }
 397                         else
 398                         {
 399                                 pse->age++;
 400                                 pse_prev = pse;
 401                                 pse = pse->next;
 402                         }
 403                 } /* while (pse != NULL) */
 404         } /* for (ps = list_head_g; ps != NULL; ps = ps->next) */
 405 }
 406 #endif /* HAVE_THREAD_INFO | KERNEL_LINUX */
 407
 408 static int ps_config (char *key, char *value)
 409 {
 410         if (strcasecmp (key, "Process") == 0)
 411         {
 412                 ps_list_register (value);
 413         }
 414         else
 415         {
 416                 return (-1);
 417         }
 418
 419         return (0);
 420 }
 421
 422 static void ps_init (void)
 423 {
 424 #if HAVE_THREAD_INFO
 425         kern_return_t status;
 426
 427         port_host_self = mach_host_self ();
 428         port_task_self = mach_task_self ();
 429
 430         if (pset_list != NULL)
 431         {
 432                 vm_deallocate (port_task_self,
 433                                 (vm_address_t) pset_list,
 434                                 pset_list_len * sizeof (processor_set_t));
 435                 pset_list = NULL;
 436                 pset_list_len = 0;
 437         }
 438
 439         if ((status = host_processor_sets (port_host_self,
 440                                         &pset_list,
 441                                         &pset_list_len)) != KERN_SUCCESS)
 442         {
 443                 syslog (LOG_ERR, "host_processor_sets failed: %s\n",
 444                                 mach_error_string (status));
 445                 pset_list = NULL;
 446                 pset_list_len = 0;
 447                 return;
 448         }
 449 /* #endif HAVE_THREAD_INFO */
 450
 451 #elif KERNEL_LINUX
 452         pagesize_g = sysconf(_SC_PAGESIZE);
 453         DBG ("pagesize_g = %li; CONFIG_HZ = %i;",
 454                         pagesize_g, CONFIG_HZ);
 455 #endif /* KERNEL_LINUX */
 456
 457         return;
 458 }
 459
 460 static void ps_write (char *host, char *inst, char *val)
 461 {
 462         rrd_update_file (host, processes_file, val,
 463                         processes_ds_def, processes_ds_num);
 464 }
 465
 466 static void ps_rss_write (char *host, char *inst, char *val)
 467 {
 468         char filename[256];
 469         int status;
 470
 471         status = snprintf (filename, 256, ps_rss_file, inst);
 472         if ((status < 1) || (status >= 256))
 473                 return;
 474
 475         rrd_update_file (host, filename, val, ps_rss_ds_def, ps_rss_ds_num);
 476 }
 477
 478 static void ps_cputime_write (char *host, char *inst, char *val)
 479 {
 480         char filename[256];
 481         int status;
 482
 483         status = snprintf (filename, 256, ps_cputime_file, inst);
 484         if ((status < 1) || (status >= 256))
 485                 return;
 486
 487         DBG ("host = %s; filename = %s; val = %s;",
 488                         host, filename, val);
 489         rrd_update_file (host, filename, val,
 490                         ps_cputime_ds_def, ps_cputime_ds_num);
 491 }
 492
 493 static void ps_count_write (char *host, char *inst, char *val)
 494 {
 495         char filename[256];
 496         int status;
 497
 498         status = snprintf (filename, 256, ps_count_file, inst);
 499         if ((status < 1) || (status >= 256))
 500                 return;
 501
 502         DBG ("host = %s; filename = %s; val = %s;",
 503                         host, filename, val);
 504         rrd_update_file (host, filename, val,
 505                         ps_count_ds_def, ps_count_ds_num);
 506 }
 507
 508 static void ps_pagefaults_write (char *host, char *inst, char *val)
 509 {
 510         char filename[256];
 511         int status;
 512
 513         status = snprintf (filename, 256, ps_pagefaults_file, inst);
 514         if ((status < 1) || (status >= 256))
 515                 return;
 516
 517         DBG ("host = %s; filename = %s; val = %s;",
 518                         host, filename, val);
 519         rrd_update_file (host, filename, val,
 520                         ps_pagefaults_ds_def, ps_pagefaults_ds_num);
 521 }
 522
 523 #if PROCESSES_HAVE_READ
 524 static void ps_submit (int running,
 525                 int sleeping,
 526                 int zombies,
 527                 int stopped,
 528                 int paging,
 529                 int blocked)
 530 {
 531         char buf[BUFSIZE];
 532
 533         if (snprintf (buf, BUFSIZE, "%u:%i:%i:%i:%i:%i:%i",
 534                                 (unsigned int) curtime,
 535                                 running, sleeping, zombies, stopped, paging,
 536                                 blocked) >= BUFSIZE)
 537                 return;
 538
 539         DBG ("running = %i; sleeping = %i; zombies = %i; stopped = %i; paging = %i; blocked = %i;",
 540                         running, sleeping, zombies, stopped, paging, blocked);
 541
 542         plugin_submit (MODULE_NAME, "-", buf);
 543 }
 544
 545 static void ps_submit_proc_list (procstat_t *ps)
 546 {
 547         char buffer[64];
 548
 549         if (ps == NULL)
 550                 return;
 551
 552         snprintf (buffer, 64, "%u:%lu",
 553                         (unsigned int) curtime,
 554                         ps->vmem_rss);
 555         buffer[63] = '\0';
 556         plugin_submit ("ps_rss", ps->name, buffer);
 557
 558         snprintf (buffer, 64, "%u:%u:%u",
 559                         (unsigned int) curtime,
 560                         /* Make the counter overflow */
 561                         (unsigned int) (ps->cpu_user_counter   & 0xFFFFFFFF),
 562                         (unsigned int) (ps->cpu_system_counter & 0xFFFFFFFF));
 563         buffer[63] = '\0';
 564         plugin_submit ("ps_cputime", ps->name, buffer);
 565
 566         snprintf (buffer, 64, "%u:%lu:%lu",
 567                         (unsigned int) curtime,
 568                         ps->num_proc, ps->num_lwp);
 569         buffer[63] = '\0';
 570         plugin_submit ("ps_count", ps->name, buffer);
 571
 572         snprintf (buffer, 64, "%u:%lu:%lu",
 573                         (unsigned int) curtime,
 574                         ps->vmem_minflt_counter, ps->vmem_majflt_counter);
 575         buffer[63] = '\0';
 576         plugin_submit ("ps_pagefaults", ps->name, buffer);
 577
 578         DBG ("name = %s; num_proc = %lu; num_lwp = %lu; vmem_rss = %lu; "
 579                         "vmem_minflt_counter = %lu; vmem_majflt_counter = %lu; "
 580                         "cpu_user_counter = %lu; cpu_system_counter = %lu;",
 581                         ps->name, ps->num_proc, ps->num_lwp, ps->vmem_rss,
 582                         ps->vmem_minflt_counter, ps->vmem_majflt_counter, ps->cpu_user_counter,
 583                         ps->cpu_system_counter);
 584
 585 }
 586
 587 #if KERNEL_LINUX
 588 static int *ps_read_tasks (int pid)
 589 {
 590         int *list = NULL;
 591         int  list_size = 1; /* size of allocated space, in elements */
 592         int  list_len = 0;  /* number of currently used elements */
 593
 594         char           dirname[64];
 595         DIR           *dh;
 596         struct dirent *ent;
 597
 598         snprintf (dirname, 64, "/proc/%i/task", pid);
 599         dirname[63] = '\0';
 600
 601         if ((dh = opendir (dirname)) == NULL)
 602         {
 603                 DBG ("Failed to open directory `%s'", dirname);
 604                 return (NULL);
 605         }
 606
 607         while ((ent = readdir (dh)) != NULL)
 608         {
 609                 if (!isdigit (ent->d_name[0]))
 610                         continue;
 611
 612                 if ((list_len + 1) >= list_size)
 613                 {
 614                         int *new_ptr;
 615                         int  new_size = 2 * list_size;
 616                         /* Comes in sizes: 2, 4, 8, 16, ... */
 617
 618                         new_ptr = (int *) realloc (list, (size_t) (sizeof (int) * new_size));
 619                         if (new_ptr == NULL)
 620                         {
 621                                 if (list != NULL)
 622                                         free (list);
 623                                 syslog (LOG_ERR, "processes plugin: "
 624                                                 "Failed to allocate more memory.");
 625                                 return (NULL);
 626                         }
 627
 628                         list = new_ptr;
 629                         list_size = new_size;
 630
 631                         memset (list + list_len, 0, sizeof (int) * (list_size - list_len));
 632                 }
 633
 634                 list[list_len] = atoi (ent->d_name);
 635                 if (list[list_len] != 0)
 636                         list_len++;
 637         }
 638
 639         closedir (dh);
 640
 641         assert (list_len < list_size);
 642         assert (list[list_len] == 0);
 643
 644         return (list);
 645 }
 646
 647 int ps_read_process (int pid, procstat_t *ps, char *state)
 648 {
 649         char  filename[64];
 650         char  buffer[1024];
 651         FILE *fh;
 652
 653         char *fields[64];
 654         char  fields_len;
 655
 656         int  *tasks;
 657         int   i;
 658
 659         int   ppid;
 660         int   name_len;
 661
 662         long long unsigned cpu_user_counter;
 663         long long unsigned cpu_system_counter;
 664         long long unsigned vmem_rss;
 665
 666         memset (ps, 0, sizeof (procstat_t));
 667
 668         snprintf (filename, 64, "/proc/%i/stat", pid);
 669         filename[63] = '\0';
 670
 671         if ((fh = fopen (filename, "r")) == NULL)
 672                 return (-1);
 673
 674         if (fgets (buffer, 1024, fh) == NULL)
 675         {
 676                 fclose (fh);
 677                 return (-1);
 678         }
 679
 680         fclose (fh);
 681
 682         fields_len = strsplit (buffer, fields, 64);
 683         if (fields_len < 24)
 684         {
 685                 DBG ("`%s' has only %i fields..",
 686                                 filename, fields_len);
 687                 return (-1);
 688         }
 689         else if (fields_len != 41)
 690         {
 691                 DBG ("WARNING: (fields_len = %i) != 41", fields_len);
 692         }
 693
 694         /* copy the name, strip brackets in the process */
 695         name_len = strlen (fields[1]) - 2;
 696         if ((fields[1][0] != '(') || (fields[1][name_len + 1] != ')'))
 697         {
 698                 DBG ("No brackets found in process name: `%s'", fields[1]);
 699                 return (-1);
 700         }
 701         fields[1] = fields[1] + 1;
 702         fields[1][name_len] = '\0';
 703         strncpy (ps->name, fields[1], PROCSTAT_NAME_LEN);
 704
 705         ppid = atoi (fields[3]);
 706
 707         if ((tasks = ps_read_tasks (pid)) == NULL)
 708         {
 709                 /* This happends for zombied, e.g. */
 710                 DBG ("ps_read_tasks (%i) failed.", pid);
 711                 *state = 'Z';
 712                 ps->num_lwp  = 0;
 713                 ps->num_proc = 0;
 714         }
 715         else
 716         {
 717                 *state = '\0';
 718                 ps->num_lwp  = 0;
 719                 ps->num_proc = 1;
 720                 for (i = 0; tasks[i] != 0; i++)
 721                         ps->num_lwp++;
 722
 723                 free (tasks);
 724                 tasks = NULL;
 725         }
 726
 727         /* Leave the rest at zero if this is only an LWP */
 728         if (ps->num_proc == 0)
 729         {
 730                 DBG ("This is only an LWP: pid = %i; name = %s;",
 731                                 pid, ps->name);
 732                 return (0);
 733         }
 734
 735         cpu_user_counter   = atoll (fields[13]);
 736         cpu_system_counter = atoll (fields[14]);
 737         vmem_rss = atoll (fields[23]);
 738         ps->vmem_minflt_counter = atol (fields[9]);
 739         ps->vmem_majflt_counter = atol (fields[11]);
 740
 741         /* Convert jiffies to useconds */
 742         cpu_user_counter   = cpu_user_counter   * 1000000 / CONFIG_HZ;
 743         cpu_system_counter = cpu_system_counter * 1000000 / CONFIG_HZ;
 744         vmem_rss = vmem_rss * pagesize_g;
 745
 746         ps->cpu_user_counter = (unsigned long) cpu_user_counter;
 747         ps->cpu_system_counter = (unsigned long) cpu_system_counter;
 748         ps->vmem_rss = (unsigned long) vmem_rss;
 749
 750         *state = fields[2][0];
 751
 752         /* success */
 753         return (0);
 754 } /* int ps_read_process (...) */
 755 #endif /* KERNEL_LINUX */
 756
 757 #if HAVE_THREAD_INFO
 758 static int mach_get_task_name (task_t t, int *pid, char *name, size_t name_max_len)
 759 {
 760         int mib[4];
 761
 762         struct kinfo_proc kp;
 763         size_t            kp_size;
 764
 765         mib[0] = CTL_KERN;
 766         mib[1] = KERN_PROC;
 767         mib[2] = KERN_PROC_PID;
 768
 769         if (pid_for_task (t, pid) != KERN_SUCCESS)
 770                 return (-1);
 771         mib[3] = *pid;
 772
 773         kp_size = sizeof (kp);
 774         if (sysctl (mib, 4, &kp, &kp_size, NULL, 0) != 0)
 775                 return (-1);
 776
 777         if (name_max_len > (MAXCOMLEN + 1))
 778                 name_max_len = MAXCOMLEN + 1;
 779
 780         strncpy (name, kp.kp_proc.p_comm, name_max_len - 1);
 781         name[name_max_len - 1] = '\0';
 782
 783         DBG ("pid = %i; name = %s;", *pid, name);
 784
 785         /* We don't do the special handling for `p_comm == "LaunchCFMApp"' as
 786          * `top' does it, because it is a lot of work and only used when
 787          * debugging. -octo */
 788
 789         return (0);
 790 }
 791 #endif /* HAVE_THREAD_INFO */
 792
 793 static void ps_read (void)
 794 {
 795 #if HAVE_THREAD_INFO
 796         kern_return_t            status;
 797
 798         int                      pset;
 799         processor_set_t          port_pset_priv;
 800
 801         int                      task;
 802         task_array_t             task_list;
 803         mach_msg_type_number_t   task_list_len;
 804
 805         int                      task_pid;
 806         char                     task_name[MAXCOMLEN + 1];
 807
 808         int                      thread;
 809         thread_act_array_t       thread_list;
 810         mach_msg_type_number_t   thread_list_len;
 811         thread_basic_info_data_t thread_data;
 812         mach_msg_type_number_t   thread_data_len;
 813
 814         int running  = 0;
 815         int sleeping = 0;
 816         int zombies  = 0;
 817         int stopped  = 0;
 818         int blocked  = 0;
 819
 820         procstat_t *ps;
 821         procstat_entry_t pse;
 822
 823         ps_list_reset ();
 824
 825         /*
 826          * The Mach-concept is a little different from the traditional UNIX
 827          * concept: All the work is done in threads. Threads are contained in
 828          * `tasks'. Therefore, `task status' doesn't make much sense, since
 829          * it's actually a `thread status'.
 830          * Tasks are assigned to sets of processors, so that's where you go to
 831          * get a list.
 832          */
 833         for (pset = 0; pset < pset_list_len; pset++)
 834         {
 835                 if ((status = host_processor_set_priv (port_host_self,
 836                                                 pset_list[pset],
 837                                                 &port_pset_priv)) != KERN_SUCCESS)
 838                 {
 839                         syslog (LOG_ERR, "host_processor_set_priv failed: %s\n",
 840                                         mach_error_string (status));
 841                         continue;
 842                 }
 843
 844                 if ((status = processor_set_tasks (port_pset_priv,
 845                                                 &task_list,
 846                                                 &task_list_len)) != KERN_SUCCESS)
 847                 {
 848                         syslog (LOG_ERR, "processor_set_tasks failed: %s\n",
 849                                         mach_error_string (status));
 850                         mach_port_deallocate (port_task_self, port_pset_priv);
 851                         continue;
 852                 }
 853
 854                 for (task = 0; task < task_list_len; task++)
 855                 {
 856                         ps = NULL;
 857                         if (mach_get_task_name (task_list[task],
 858                                                 &task_pid,
 859                                                 task_name, PROCSTAT_NAME_LEN) == 0)
 860                                 ps = ps_list_search (task_name);
 861
 862                         /* Collect more detailed statistics for this process */
 863                         if (ps != NULL)
 864                         {
 865                                 task_basic_info_data_t        task_basic_info;
 866                                 mach_msg_type_number_t        task_basic_info_len;
 867                                 task_events_info_data_t       task_events_info;
 868                                 mach_msg_type_number_t        task_events_info_len;
 869                                 task_absolutetime_info_data_t task_absolutetime_info;
 870                                 mach_msg_type_number_t        task_absolutetime_info_len;
 871
 872                                 memset (&pse, '\0', sizeof (pse));
 873                                 pse.id = task_pid;
 874
 875                                 task_basic_info_len = TASK_BASIC_INFO_COUNT;
 876                                 status = task_info (task_list[task],
 877                                                 TASK_BASIC_INFO,
 878                                                 (task_info_t) &task_basic_info,
 879                                                 &task_basic_info_len);
 880                                 if (status != KERN_SUCCESS)
 881                                 {
 882                                         syslog (LOG_ERR, "task_info failed: %s",
 883                                                         mach_error_string (status));
 884                                         continue; /* with next thread_list */
 885                                 }
 886
 887                                 task_events_info_len = TASK_EVENTS_INFO_COUNT;
 888                                 status = task_info (task_list[task],
 889                                                 TASK_EVENTS_INFO,
 890                                                 (task_info_t) &task_events_info,
 891                                                 &task_events_info_len);
 892                                 if (status != KERN_SUCCESS)
 893                                 {
 894                                         syslog (LOG_ERR, "task_info failed: %s",
 895                                                         mach_error_string (status));
 896                                         continue; /* with next thread_list */
 897                                 }
 898
 899                                 task_absolutetime_info_len = TASK_ABSOLUTETIME_INFO_COUNT;
 900                                 status = task_info (task_list[task],
 901                                                 TASK_ABSOLUTETIME_INFO,
 902                                                 (task_info_t) &task_absolutetime_info,
 903                                                 &task_absolutetime_info_len);
 904                                 if (status != KERN_SUCCESS)
 905                                 {
 906                                         syslog (LOG_ERR, "task_info failed: %s",
 907                                                         mach_error_string (status));
 908                                         continue; /* with next thread_list */
 909                                 }
 910
 911                                 pse.num_proc++;
 912                                 pse.vmem_rss = task_basic_info.resident_size;
 913
 914                                 pse.vmem_minflt_counter = task_events_info.cow_faults;
 915                                 pse.vmem_majflt_counter = task_events_info.faults;
 916
 917                                 pse.cpu_user_counter = task_absolutetime_info.total_user;
 918                                 pse.cpu_system_counter = task_absolutetime_info.total_system;
 919                         }
 920
 921                         status = task_threads (task_list[task], &thread_list,
 922                                         &thread_list_len);
 923                         if (status != KERN_SUCCESS)
 924                         {
 925                                 /* Apple's `top' treats this case a zombie. It
 926                                  * makes sense to some extend: A `zombie'
 927                                  * thread is nonsense, since the task/process
 928                                  * is dead. */
 929                                 zombies++;
 930                                 DBG ("task_threads failed: %s",
 931                                                 mach_error_string (status));
 932                                 if (task_list[task] != port_task_self)
 933                                         mach_port_deallocate (port_task_self,
 934                                                         task_list[task]);
 935                                 continue; /* with next task_list */
 936                         }
 937
 938                         for (thread = 0; thread < thread_list_len; thread++)
 939                         {
 940                                 thread_data_len = THREAD_BASIC_INFO_COUNT;
 941                                 status = thread_info (thread_list[thread],
 942                                                 THREAD_BASIC_INFO,
 943                                                 (thread_info_t) &thread_data,
 944                                                 &thread_data_len);
 945                                 if (status != KERN_SUCCESS)
 946                                 {
 947                                         syslog (LOG_ERR, "thread_info failed: %s",
 948                                                         mach_error_string (status));
 949                                         if (task_list[task] != port_task_self)
 950                                                 mach_port_deallocate (port_task_self,
 951                                                                 thread_list[thread]);
 952                                         continue; /* with next thread_list */
 953                                 }
 954
 955                                 if (ps != NULL)
 956                                         pse.num_lwp++;
 957
 958                                 switch (thread_data.run_state)
 959                                 {
 960                                         case TH_STATE_RUNNING:
 961                                                 running++;
 962                                                 break;
 963                                         case TH_STATE_STOPPED:
 964                                         /* What exactly is `halted'? */
 965                                         case TH_STATE_HALTED:
 966                                                 stopped++;
 967                                                 break;
 968                                         case TH_STATE_WAITING:
 969                                                 sleeping++;
 970                                                 break;
 971                                         case TH_STATE_UNINTERRUPTIBLE:
 972                                                 blocked++;
 973                                                 break;
 974                                         /* There is no `zombie' case here,
 975                                          * since there are no zombie-threads.
 976                                          * There's only zombie tasks, which are
 977                                          * handled above. */
 978                                         default:
 979                                                 syslog (LOG_WARNING,
 980                                                                 "Unknown thread status: %s",
 981                                                                 thread_data.run_state);
 982                                                 break;
 983                                 } /* switch (thread_data.run_state) */
 984
 985                                 if (task_list[task] != port_task_self)
 986                                 {
 987                                         status = mach_port_deallocate (port_task_self,
 988                                                         thread_list[thread]);
 989                                         if (status != KERN_SUCCESS)
 990                                                 syslog (LOG_ERR, "mach_port_deallocate failed: %s",
 991                                                                 mach_error_string (status));
 992                                 }
 993                         } /* for (thread_list) */
 994
 995                         if ((status = vm_deallocate (port_task_self,
 996                                                         (vm_address_t) thread_list,
 997                                                         thread_list_len * sizeof (thread_act_t)))
 998                                         != KERN_SUCCESS)
 999                         {
1000                                 syslog (LOG_ERR, "vm_deallocate failed: %s",
1001                                                 mach_error_string (status));
1002                         }
1003                         thread_list = NULL;
1004                         thread_list_len = 0;
1005
1006                         /* Only deallocate the task port, if it isn't our own.
1007                          * Don't know what would happen in that case, but this
1008                          * is what Apple's top does.. ;) */
1009                         if (task_list[task] != port_task_self)
1010                         {
1011                                 status = mach_port_deallocate (port_task_self,
1012                                                 task_list[task]);
1013                                 if (status != KERN_SUCCESS)
1014                                         syslog (LOG_ERR, "mach_port_deallocate failed: %s",
1015                                                         mach_error_string (status));
1016                         }
1017
1018                         if (ps != NULL)
1019                                 ps_list_add (task_name, &pse);
1020                 } /* for (task_list) */
1021
1022                 if ((status = vm_deallocate (port_task_self,
1023                                 (vm_address_t) task_list,
1024                                 task_list_len * sizeof (task_t))) != KERN_SUCCESS)
1025                 {
1026                         syslog (LOG_ERR, "vm_deallocate failed: %s",
1027                                         mach_error_string (status));
1028                 }
1029                 task_list = NULL;
1030                 task_list_len = 0;
1031
1032                 if ((status = mach_port_deallocate (port_task_self, port_pset_priv))
1033                                 != KERN_SUCCESS)
1034                 {
1035                         syslog (LOG_ERR, "mach_port_deallocate failed: %s",
1036                                         mach_error_string (status));
1037                 }
1038         } /* for (pset_list) */
1039
1040         ps_submit (running, sleeping, zombies, stopped, -1, blocked);
1041
1042         for (ps = list_head_g; ps != NULL; ps = ps->next)
1043                 ps_submit_proc_list (ps);
1044 /* #endif HAVE_THREAD_INFO */
1045
1046 #elif KERNEL_LINUX
1047         int running  = 0;
1048         int sleeping = 0;
1049         int zombies  = 0;
1050         int stopped  = 0;
1051         int paging   = 0;
1052         int blocked  = 0;
1053
1054         struct dirent *ent;
1055         DIR           *proc;
1056         int            pid;
1057
1058         int        status;
1059         procstat_t ps;
1060         procstat_entry_t pse;
1061         char       state;
1062
1063         procstat_t *ps_ptr;
1064
1065         running = sleeping = zombies = stopped = paging = blocked = 0;
1066         ps_list_reset ();
1067
1068         if ((proc = opendir ("/proc")) == NULL)
1069         {
1070                 syslog (LOG_ERR, "Cannot open `/proc': %s", strerror (errno));
1071                 return;
1072         }
1073
1074         while ((ent = readdir (proc)) != NULL)
1075         {
1076                 if (!isdigit (ent->d_name[0]))
1077                         continue;
1078
1079                 if ((pid = atoi (ent->d_name)) < 1)
1080                         continue;
1081
1082                 status = ps_read_process (pid, &ps, &state);
1083                 if (status != 0)
1084                 {
1085                         DBG ("ps_read_process failed: %i", status);
1086                         continue;
1087                 }
1088
1089                 pse.id       = pid;
1090                 pse.age      = 0;
1091
1092                 pse.num_proc = ps.num_proc;
1093                 pse.num_lwp  = ps.num_lwp;
1094                 pse.vmem_rss = ps.vmem_rss;
1095
1096                 pse.vmem_minflt = 0;
1097                 pse.vmem_minflt_counter = ps.vmem_minflt_counter;
1098                 pse.vmem_majflt = 0;
1099                 pse.vmem_majflt_counter = ps.vmem_majflt_counter;
1100
1101                 pse.cpu_user = 0;
1102                 pse.cpu_user_counter = ps.cpu_user_counter;
1103                 pse.cpu_system = 0;
1104                 pse.cpu_system_counter = ps.cpu_system_counter;
1105
1106                 switch (state)
1107                 {
1108                         case 'R': running++;  break;
1109                         case 'S': sleeping++; break;
1110                         case 'D': blocked++;  break;
1111                         case 'Z': zombies++;  break;
1112                         case 'T': stopped++;  break;
1113                         case 'W': paging++;   break;
1114                 }
1115
1116                 ps_list_add (ps.name, &pse);
1117         }
1118
1119         closedir (proc);
1120
1121         ps_submit (running, sleeping, zombies, stopped, paging, blocked);
1122
1123         for (ps_ptr = list_head_g; ps_ptr != NULL; ps_ptr = ps_ptr->next)
1124                 ps_submit_proc_list (ps_ptr);
1125 #endif /* KERNEL_LINUX */
1126 }
1127 #else
1128 # define ps_read NULL
1129 #endif /* PROCESSES_HAVE_READ */
1130
1131 void module_register (void)
1132 {
1133         plugin_register (MODULE_NAME, ps_init, ps_read, ps_write);
1134         plugin_register ("ps_rss", NULL, NULL, ps_rss_write);
1135         plugin_register ("ps_cputime", NULL, NULL, ps_cputime_write);
1136         plugin_register ("ps_count", NULL, NULL, ps_count_write);
1137         plugin_register ("ps_pagefaults", NULL, NULL, ps_pagefaults_write);
1138         cf_register (MODULE_NAME, ps_config, config_keys, config_keys_num);
1139 }
1140
1141 #undef BUFSIZE
1142 #undef MODULE_NAME