src/processes.c

   1 /**
   2  * collectd - src/processes.c
   3  * Copyright (C) 2005  Lyonel Vincent
   4  * Copyright (C) 2006-2008  Florian Forster (Mach code)
   5  * Copyright (C) 2008  Oleg King
   6  *
   7  * This program is free software; you can redistribute it and/or modify it
   8  * under the terms of the GNU General Public License as published by the
   9  * Free Software Foundation; either version 2 of the License, or (at your
  10  * option) any later version.
  11  *
  12  * This program is distributed in the hope that it will be useful, but
  13  * WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU General Public License along
  18  * with this program; if not, write to the Free Software Foundation, Inc.,
  19  * 51 Franklin St, Fifth Floor, Boston, MA  02110-1301 USA
  20  *
  21  * Authors:
  22  *   Lyonel Vincent <lyonel at ezix.org>
  23  *   Florian octo Forster <octo at verplant.org>
  24  *   Oleg King <king2 at kaluga.ru>
  25  **/
  26
  27 #include "collectd.h"
  28 #include "common.h"
  29 #include "plugin.h"
  30 #include "configfile.h"
  31
  32 /* Include header files for the mach system, if they exist.. */
  33 #if HAVE_THREAD_INFO
  34 #  if HAVE_MACH_MACH_INIT_H
  35 #    include <mach/mach_init.h>
  36 #  endif
  37 #  if HAVE_MACH_HOST_PRIV_H
  38 #    include <mach/host_priv.h>
  39 #  endif
  40 #  if HAVE_MACH_MACH_ERROR_H
  41 #    include <mach/mach_error.h>
  42 #  endif
  43 #  if HAVE_MACH_MACH_HOST_H
  44 #    include <mach/mach_host.h>
  45 #  endif
  46 #  if HAVE_MACH_MACH_PORT_H
  47 #    include <mach/mach_port.h>
  48 #  endif
  49 #  if HAVE_MACH_MACH_TYPES_H
  50 #    include <mach/mach_types.h>
  51 #  endif
  52 #  if HAVE_MACH_MESSAGE_H
  53 #    include <mach/message.h>
  54 #  endif
  55 #  if HAVE_MACH_PROCESSOR_SET_H
  56 #    include <mach/processor_set.h>
  57 #  endif
  58 #  if HAVE_MACH_TASK_H
  59 #    include <mach/task.h>
  60 #  endif
  61 #  if HAVE_MACH_THREAD_ACT_H
  62 #    include <mach/thread_act.h>
  63 #  endif
  64 #  if HAVE_MACH_VM_REGION_H
  65 #    include <mach/vm_region.h>
  66 #  endif
  67 #  if HAVE_MACH_VM_MAP_H
  68 #    include <mach/vm_map.h>
  69 #  endif
  70 #  if HAVE_MACH_VM_PROT_H
  71 #    include <mach/vm_prot.h>
  72 #  endif
  73 #  if HAVE_SYS_SYSCTL_H
  74 #    include <sys/sysctl.h>
  75 #  endif
  76 /* #endif HAVE_THREAD_INFO */
  77
  78 #elif KERNEL_LINUX
  79 #  if HAVE_LINUX_CONFIG_H
  80 #    include <linux/config.h>
  81 #  endif
  82 #  ifndef CONFIG_HZ
  83 #    define CONFIG_HZ 100
  84 #  endif
  85 /* #endif KERNEL_LINUX */
  86
  87 #elif HAVE_LIBKVM_GETPROCS
  88 #  include <kvm.h>
  89 #  include <sys/user.h>
  90 #  include <sys/proc.h>
  91 #  if HAVE_SYS_SYSCTL_H
  92 #    include <sys/sysctl.h>
  93 #  endif
  94 /* #endif HAVE_LIBKVM_GETPROCS */
  95
  96 #else
  97 # error "No applicable input method."
  98 #endif
  99
 100 #if HAVE_REGEX_H
 101 # include <regex.h>
 102 #endif
 103
 104 #define BUFSIZE 256
 105
 106 static const char *config_keys[] =
 107 {
 108         "Process",
 109         "ProcessMatch",
 110         NULL
 111 };
 112 static int config_keys_num = STATIC_ARRAY_SIZE (config_keys);
 113
 114 typedef struct procstat_entry_s
 115 {
 116         unsigned long id;
 117         unsigned long age;
 118
 119         unsigned long num_proc;
 120         unsigned long num_lwp;
 121         unsigned long vmem_rss;
 122
 123         unsigned long vmem_minflt;
 124         unsigned long vmem_majflt;
 125         unsigned long vmem_minflt_counter;
 126         unsigned long vmem_majflt_counter;
 127
 128         unsigned long cpu_user;
 129         unsigned long cpu_system;
 130         unsigned long cpu_user_counter;
 131         unsigned long cpu_system_counter;
 132
 133         struct procstat_entry_s *next;
 134 } procstat_entry_t;
 135
 136 #define PROCSTAT_NAME_LEN 256
 137 typedef struct procstat
 138 {
 139         char          name[PROCSTAT_NAME_LEN];
 140 #if HAVE_REGEX_H
 141         regex_t *re;
 142 #endif
 143
 144         unsigned long num_proc;
 145         unsigned long num_lwp;
 146         unsigned long vmem_rss;
 147
 148         unsigned long vmem_minflt_counter;
 149         unsigned long vmem_majflt_counter;
 150
 151         unsigned long cpu_user_counter;
 152         unsigned long cpu_system_counter;
 153
 154         struct procstat   *next;
 155         struct procstat_entry_s *instances;
 156 } procstat_t;
 157
 158 static procstat_t *list_head_g = NULL;
 159
 160 #if HAVE_THREAD_INFO
 161 static mach_port_t port_host_self;
 162 static mach_port_t port_task_self;
 163
 164 static processor_set_name_array_t pset_list;
 165 static mach_msg_type_number_t     pset_list_len;
 166 /* #endif HAVE_THREAD_INFO */
 167
 168 #elif KERNEL_LINUX
 169 static long pagesize_g;
 170 /* #endif KERNEL_LINUX */
 171
 172 #elif HAVE_LIBKVM_GETPROCS
 173 /* no global variables */
 174 #endif /* HAVE_LIBKVM_GETPROCS */
 175
 176 /* put name of process from config to list_head_g tree
 177    list_head_g is a list of 'procstat_t' structs with
 178    processes names we want to watch */
 179 static void ps_list_register (const char *name, const char *regexp)
 180 {
 181         procstat_t *new;
 182         procstat_t *ptr;
 183         int status;
 184
 185         new = (procstat_t *) malloc (sizeof (procstat_t));
 186         if (new == NULL)
 187         {
 188                 ERROR ("processes plugin: ps_list_register: malloc failed.");
 189                 return;
 190         }
 191         memset (new, 0, sizeof (procstat_t));
 192         sstrncpy (new->name, name, sizeof (new->name));
 193
 194 #if HAVE_REGEX_H
 195         if (regexp != NULL)
 196         {
 197                 DEBUG ("ProcessMatch: adding \"%s\" as criteria to process %s.", regexp, name);
 198                 new->re = (regex_t *) malloc (sizeof (regex_t));
 199                 if (new->re == NULL)
 200                 {
 201                         ERROR ("processes plugin: ps_list_register: malloc failed.");
 202                         sfree (new);
 203                         return;
 204                 }
 205
 206                 status = regcomp (new->re, regexp, REG_EXTENDED | REG_NOSUB);
 207                 if (status != 0)
 208                 {
 209                         DEBUG ("ProcessMatch: compiling the regular expression \"%s\" failed.", regexp);
 210                         sfree(new->re);
 211                         return;
 212                 }
 213         }
 214 #else
 215         if (regexp != NULL)
 216         {
 217                 ERROR ("processes plugin: ps_list_register: "
 218                                 "Regular expression \"%s\" found in config "
 219                                 "file, but support for regular expressions "
 220                                 "has been dispabled at compile time.",
 221                                 regexp);
 222                 sfree (new);
 223                 return;
 224         }
 225 #endif
 226
 227         for (ptr = list_head_g; ptr != NULL; ptr = ptr->next)
 228         {
 229                 if (strcmp (ptr->name, name) == 0)
 230                 {
 231                         WARNING ("processes plugin: You have configured more "
 232                                         "than one `Process' or "
 233                                         "`ProcessMatch' with the same name. "
 234                                         "All but the first setting will be "
 235                                         "ignored.");
 236                         sfree (new->re);
 237                         sfree (new);
 238                         return;
 239                 }
 240
 241                 if (ptr->next == NULL)
 242                         break;
 243         }
 244
 245         if (ptr == NULL)
 246                 list_head_g = new;
 247         else
 248                 ptr->next = new;
 249 } /* void ps_list_register */
 250
 251 /* try to match name against entry, returns 1 if success */
 252 static int ps_list_match (const char *name, const char *cmdline, procstat_t *ps)
 253 {
 254 #if HAVE_REGEX_H
 255         if (ps->re != NULL)
 256         {
 257                 int status;
 258                 const char *str;
 259
 260                 str = cmdline;
 261                 if ((str == NULL) || (str[0] == 0))
 262                         str = name;
 263
 264                 assert (str != NULL);
 265
 266                 status = regexec (ps->re, str,
 267                                 /* nmatch = */ 0,
 268                                 /* pmatch = */ NULL,
 269                                 /* eflags = */ 0);
 270                 if (status == 0)
 271                         return (1);
 272         }
 273         else
 274 #endif
 275         if (strcmp (ps->name, name) == 0)
 276                 return (1);
 277
 278         return (0);
 279 } /* int ps_list_match */
 280
 281 /* add process entry to 'instances' of process 'name' (or refresh it) */
 282 static void ps_list_add (const char *name, const char *cmdline, procstat_entry_t *entry)
 283 {
 284         procstat_t *ps;
 285         procstat_entry_t *pse;
 286
 287         if (entry->id == 0)
 288                 return;
 289
 290         for (ps = list_head_g; ps != NULL; ps = ps->next)
 291         {
 292                 if ((ps_list_match (name, cmdline, ps)) == 0)
 293                         continue;
 294
 295                 for (pse = ps->instances; pse != NULL; pse = pse->next)
 296                         if ((pse->id == entry->id) || (pse->next == NULL))
 297                                 break;
 298
 299                 if ((pse == NULL) || (pse->id != entry->id))
 300                 {
 301                         procstat_entry_t *new;
 302
 303                         new = (procstat_entry_t *) malloc (sizeof (procstat_entry_t));
 304                         if (new == NULL)
 305                                 return;
 306                         memset (new, 0, sizeof (procstat_entry_t));
 307                         new->id = entry->id;
 308
 309                         if (pse == NULL)
 310                                 ps->instances = new;
 311                         else
 312                                 pse->next = new;
 313
 314                         pse = new;
 315                 }
 316
 317                 pse->age = 0;
 318                 pse->num_proc = entry->num_proc;
 319                 pse->num_lwp  = entry->num_lwp;
 320                 pse->vmem_rss = entry->vmem_rss;
 321
 322                 ps->num_proc += pse->num_proc;
 323                 ps->num_lwp  += pse->num_lwp;
 324                 ps->vmem_rss += pse->vmem_rss;
 325
 326                 if ((entry->vmem_minflt_counter == 0)
 327                                 && (entry->vmem_majflt_counter == 0))
 328                 {
 329                         pse->vmem_minflt_counter += entry->vmem_minflt;
 330                         pse->vmem_minflt = entry->vmem_minflt;
 331
 332                         pse->vmem_majflt_counter += entry->vmem_majflt;
 333                         pse->vmem_majflt = entry->vmem_majflt;
 334                 }
 335                 else
 336                 {
 337                         if (entry->vmem_minflt_counter < pse->vmem_minflt_counter)
 338                         {
 339                                 pse->vmem_minflt = entry->vmem_minflt_counter
 340                                         + (ULONG_MAX - pse->vmem_minflt_counter);
 341                         }
 342                         else
 343                         {
 344                                 pse->vmem_minflt = entry->vmem_minflt_counter - pse->vmem_minflt_counter;
 345                         }
 346                         pse->vmem_minflt_counter = entry->vmem_minflt_counter;
 347
 348                         if (entry->vmem_majflt_counter < pse->vmem_majflt_counter)
 349                         {
 350                                 pse->vmem_majflt = entry->vmem_majflt_counter
 351                                         + (ULONG_MAX - pse->vmem_majflt_counter);
 352                         }
 353                         else
 354                         {
 355                                 pse->vmem_majflt = entry->vmem_majflt_counter - pse->vmem_majflt_counter;
 356                         }
 357                         pse->vmem_majflt_counter = entry->vmem_majflt_counter;
 358                 }
 359
 360                 ps->vmem_minflt_counter += pse->vmem_minflt;
 361                 ps->vmem_majflt_counter += pse->vmem_majflt;
 362
 363                 if ((entry->cpu_user_counter == 0)
 364                                 && (entry->cpu_system_counter == 0))
 365                 {
 366                         pse->cpu_user_counter += entry->cpu_user;
 367                         pse->cpu_user = entry->cpu_user;
 368
 369                         pse->cpu_system_counter += entry->cpu_system;
 370                         pse->cpu_system = entry->cpu_system;
 371                 }
 372                 else
 373                 {
 374                         if (entry->cpu_user_counter < pse->cpu_user_counter)
 375                         {
 376                                 pse->cpu_user = entry->cpu_user_counter
 377                                         + (ULONG_MAX - pse->cpu_user_counter);
 378                         }
 379                         else
 380                         {
 381                                 pse->cpu_user = entry->cpu_user_counter - pse->cpu_user_counter;
 382                         }
 383                         pse->cpu_user_counter = entry->cpu_user_counter;
 384
 385                         if (entry->cpu_system_counter < pse->cpu_system_counter)
 386                         {
 387                                 pse->cpu_system = entry->cpu_system_counter
 388                                         + (ULONG_MAX - pse->cpu_system_counter);
 389                         }
 390                         else
 391                         {
 392                                 pse->cpu_system = entry->cpu_system_counter - pse->cpu_system_counter;
 393                         }
 394                         pse->cpu_system_counter = entry->cpu_system_counter;
 395                 }
 396
 397                 ps->cpu_user_counter   += pse->cpu_user;
 398                 ps->cpu_system_counter += pse->cpu_system;
 399         }
 400 }
 401
 402 /* remove old entries from instances of processes in list_head_g */
 403 static void ps_list_reset (void)
 404 {
 405         procstat_t *ps;
 406         procstat_entry_t *pse;
 407         procstat_entry_t *pse_prev;
 408
 409         for (ps = list_head_g; ps != NULL; ps = ps->next)
 410         {
 411                 ps->num_proc    = 0;
 412                 ps->num_lwp     = 0;
 413                 ps->vmem_rss    = 0;
 414
 415                 pse_prev = NULL;
 416                 pse = ps->instances;
 417                 while (pse != NULL)
 418                 {
 419                         if (pse->age > 10)
 420                         {
 421                                 DEBUG ("Removing this procstat entry cause it's too old: "
 422                                                 "id = %lu; name = %s;",
 423                                                 pse->id, ps->name);
 424
 425                                 if (pse_prev == NULL)
 426                                 {
 427                                         ps->instances = pse->next;
 428                                         free (pse);
 429                                         pse = ps->instances;
 430                                 }
 431                                 else
 432                                 {
 433                                         pse_prev->next = pse->next;
 434                                         free (pse);
 435                                         pse = pse_prev->next;
 436                                 }
 437                         }
 438                         else
 439                         {
 440                                 pse->age++;
 441                                 pse_prev = pse;
 442                                 pse = pse->next;
 443                         }
 444                 } /* while (pse != NULL) */
 445         } /* for (ps = list_head_g; ps != NULL; ps = ps->next) */
 446 }
 447
 448 /* put all pre-defined 'Process' names from config to list_head_g tree */
 449 static int ps_config (const char *key, const char *value)
 450 {
 451         if (strcasecmp (key, "Process") == 0)
 452         {
 453                 ps_list_register (value, NULL);
 454         }
 455         else if (strcasecmp (key, "ProcessMatch") == 0)
 456         {
 457                 char *new_val;
 458                 char *fields[3];
 459                 int fields_num;
 460
 461                 new_val = strdup (value);
 462                 if (new_val == NULL)
 463                         return (1);
 464                 fields_num = strsplit (new_val, fields,
 465                                 STATIC_ARRAY_SIZE (fields));
 466                 if (fields_num != 2)
 467                 {
 468                         sfree (new_val);
 469                         return (1);
 470                 }
 471                 ps_list_register (fields[0], fields[1]);
 472                 sfree (new_val);
 473         }
 474         else
 475         {
 476                 return (-1);
 477         }
 478
 479         return (0);
 480 }
 481
 482 static int ps_init (void)
 483 {
 484 #if HAVE_THREAD_INFO
 485         kern_return_t status;
 486
 487         port_host_self = mach_host_self ();
 488         port_task_self = mach_task_self ();
 489
 490         if (pset_list != NULL)
 491         {
 492                 vm_deallocate (port_task_self,
 493                                 (vm_address_t) pset_list,
 494                                 pset_list_len * sizeof (processor_set_t));
 495                 pset_list = NULL;
 496                 pset_list_len = 0;
 497         }
 498
 499         if ((status = host_processor_sets (port_host_self,
 500                                         &pset_list,
 501                                         &pset_list_len)) != KERN_SUCCESS)
 502         {
 503                 ERROR ("host_processor_sets failed: %s\n",
 504                                 mach_error_string (status));
 505                 pset_list = NULL;
 506                 pset_list_len = 0;
 507                 return (-1);
 508         }
 509 /* #endif HAVE_THREAD_INFO */
 510
 511 #elif KERNEL_LINUX
 512         pagesize_g = sysconf(_SC_PAGESIZE);
 513         DEBUG ("pagesize_g = %li; CONFIG_HZ = %i;",
 514                         pagesize_g, CONFIG_HZ);
 515 /* #endif KERNEL_LINUX */
 516
 517 #elif HAVE_LIBKVM_GETPROCS
 518 /* no initialization */
 519 #endif /* HAVE_LIBKVM_GETPROCS */
 520
 521         return (0);
 522 } /* int ps_init */
 523
 524 /* submit global state (e.g.: qty of zombies, running, etc..) */
 525 static void ps_submit_state (const char *state, double value)
 526 {
 527         value_t values[1];
 528         value_list_t vl = VALUE_LIST_INIT;
 529
 530         values[0].gauge = value;
 531
 532         vl.values = values;
 533         vl.values_len = 1;
 534         vl.time = time (NULL);
 535         sstrncpy (vl.host, hostname_g, sizeof (vl.host));
 536         sstrncpy (vl.plugin, "processes", sizeof (vl.plugin));
 537         sstrncpy (vl.plugin_instance, "", sizeof (vl.plugin_instance));
 538         sstrncpy (vl.type, "ps_state", sizeof (vl.type));
 539         sstrncpy (vl.type_instance, state, sizeof (vl.type_instance));
 540
 541         plugin_dispatch_values (&vl);
 542 }
 543
 544 /* submit info about specific process (e.g.: memory taken, cpu usage, etc..) */
 545 static void ps_submit_proc_list (procstat_t *ps)
 546 {
 547         value_t values[2];
 548         value_list_t vl = VALUE_LIST_INIT;
 549
 550         vl.values = values;
 551         vl.values_len = 2;
 552         vl.time = time (NULL);
 553         sstrncpy (vl.host, hostname_g, sizeof (vl.host));
 554         sstrncpy (vl.plugin, "processes", sizeof (vl.plugin));
 555         sstrncpy (vl.plugin_instance, ps->name, sizeof (vl.plugin_instance));
 556
 557         sstrncpy (vl.type, "ps_rss", sizeof (vl.type));
 558         vl.values[0].gauge = ps->vmem_rss;
 559         vl.values_len = 1;
 560         plugin_dispatch_values (&vl);
 561
 562         sstrncpy (vl.type, "ps_cputime", sizeof (vl.type));
 563         vl.values[0].counter = ps->cpu_user_counter;
 564         vl.values[1].counter = ps->cpu_system_counter;
 565         vl.values_len = 2;
 566         plugin_dispatch_values (&vl);
 567
 568         sstrncpy (vl.type, "ps_count", sizeof (vl.type));
 569         vl.values[0].gauge = ps->num_proc;
 570         vl.values[1].gauge = ps->num_lwp;
 571         vl.values_len = 2;
 572         plugin_dispatch_values (&vl);
 573
 574         sstrncpy (vl.type, "ps_pagefaults", sizeof (vl.type));
 575         vl.values[0].counter = ps->vmem_minflt_counter;
 576         vl.values[1].counter = ps->vmem_majflt_counter;
 577         vl.values_len = 2;
 578         plugin_dispatch_values (&vl);
 579
 580         DEBUG ("name = %s; num_proc = %lu; num_lwp = %lu; vmem_rss = %lu; "
 581                         "vmem_minflt_counter = %lu; vmem_majflt_counter = %lu; "
 582                         "cpu_user_counter = %lu; cpu_system_counter = %lu;",
 583                         ps->name, ps->num_proc, ps->num_lwp, ps->vmem_rss,
 584                         ps->vmem_minflt_counter, ps->vmem_majflt_counter,
 585                         ps->cpu_user_counter, ps->cpu_system_counter);
 586 } /* void ps_submit_proc_list */
 587
 588 /* ------- additional functions for KERNEL_LINUX/HAVE_THREAD_INFO ------- */
 589 #if KERNEL_LINUX
 590 static int *ps_read_tasks (int pid)
 591 {
 592         int *list = NULL;
 593         int  list_size = 1; /* size of allocated space, in elements */
 594         int  list_len = 0;  /* number of currently used elements */
 595
 596         char           dirname[64];
 597         DIR           *dh;
 598         struct dirent *ent;
 599
 600         ssnprintf (dirname, sizeof (dirname), "/proc/%i/task", pid);
 601
 602         if ((dh = opendir (dirname)) == NULL)
 603         {
 604                 DEBUG ("Failed to open directory `%s'", dirname);
 605                 return (NULL);
 606         }
 607
 608         while ((ent = readdir (dh)) != NULL)
 609         {
 610                 if (!isdigit (ent->d_name[0]))
 611                         continue;
 612
 613                 if ((list_len + 1) >= list_size)
 614                 {
 615                         int *new_ptr;
 616                         int  new_size = 2 * list_size;
 617                         /* Comes in sizes: 2, 4, 8, 16, ... */
 618
 619                         new_ptr = (int *) realloc (list, (size_t) (sizeof (int) * new_size));
 620                         if (new_ptr == NULL)
 621                         {
 622                                 if (list != NULL)
 623                                         free (list);
 624                                 ERROR ("processes plugin: "
 625                                                 "Failed to allocate more memory.");
 626                                 return (NULL);
 627                         }
 628
 629                         list = new_ptr;
 630                         list_size = new_size;
 631
 632                         memset (list + list_len, 0, sizeof (int) * (list_size - list_len));
 633                 }
 634
 635                 list[list_len] = atoi (ent->d_name);
 636                 if (list[list_len] != 0)
 637                         list_len++;
 638         }
 639
 640         closedir (dh);
 641
 642         if (list_len == 0)
 643                 return (NULL);
 644
 645         assert (list_len < list_size);
 646         assert (list[list_len] == 0);
 647
 648         return (list);
 649 } /* int *ps_read_tasks */
 650
 651 int ps_read_process (int pid, procstat_t *ps, char *state)
 652 {
 653         char  filename[64];
 654         char  buffer[1024];
 655
 656         char *fields[64];
 657         char  fields_len;
 658
 659         int  *tasks;
 660         int   i;
 661
 662         int   ppid;
 663         int   name_len;
 664
 665         long long unsigned cpu_user_counter;
 666         long long unsigned cpu_system_counter;
 667         long long unsigned vmem_rss;
 668
 669         memset (ps, 0, sizeof (procstat_t));
 670
 671         ssnprintf (filename, sizeof (filename), "/proc/%i/stat", pid);
 672
 673         i = read_file_contents (filename, buffer, sizeof(buffer) - 1);
 674         if (i <= 0)
 675                 return (-1);
 676         buffer[i] = 0;
 677
 678         fields_len = strsplit (buffer, fields, 64);
 679         if (fields_len < 24)
 680         {
 681                 DEBUG ("processes plugin: ps_read_process (pid = %i):"
 682                                 " `%s' has only %i fields..",
 683                                 (int) pid, filename, fields_len);
 684                 return (-1);
 685         }
 686
 687         /* copy the name, strip brackets in the process */
 688         name_len = strlen (fields[1]) - 2;
 689         if ((fields[1][0] != '(') || (fields[1][name_len + 1] != ')'))
 690         {
 691                 DEBUG ("No brackets found in process name: `%s'", fields[1]);
 692                 return (-1);
 693         }
 694         fields[1] = fields[1] + 1;
 695         fields[1][name_len] = '\0';
 696         strncpy (ps->name, fields[1], PROCSTAT_NAME_LEN);
 697
 698         ppid = atoi (fields[3]);
 699
 700         *state = fields[2][0];
 701
 702         if (*state == 'Z')
 703         {
 704                 ps->num_lwp  = 0;
 705                 ps->num_proc = 0;
 706         }
 707         else if ((tasks = ps_read_tasks (pid)) == NULL)
 708         {
 709                 /* Kernel 2.4 or so */
 710                 ps->num_lwp  = 1;
 711                 ps->num_proc = 1;
 712         }
 713         else
 714         {
 715                 ps->num_lwp  = 0;
 716                 ps->num_proc = 1;
 717                 for (i = 0; tasks[i] != 0; i++)
 718                         ps->num_lwp++;
 719
 720                 free (tasks);
 721                 tasks = NULL;
 722         }
 723
 724         /* Leave the rest at zero if this is only a zombi */
 725         if (ps->num_proc == 0)
 726         {
 727                 DEBUG ("processes plugin: This is only a zombi: pid = %i; "
 728                                 "name = %s;", pid, ps->name);
 729                 return (0);
 730         }
 731
 732         cpu_user_counter   = atoll (fields[13]);
 733         cpu_system_counter = atoll (fields[14]);
 734         vmem_rss = atoll (fields[23]);
 735         ps->vmem_minflt_counter = atol (fields[9]);
 736         ps->vmem_majflt_counter = atol (fields[11]);
 737
 738         /* Convert jiffies to useconds */
 739         cpu_user_counter   = cpu_user_counter   * 1000000 / CONFIG_HZ;
 740         cpu_system_counter = cpu_system_counter * 1000000 / CONFIG_HZ;
 741         vmem_rss = vmem_rss * pagesize_g;
 742
 743         ps->cpu_user_counter = (unsigned long) cpu_user_counter;
 744         ps->cpu_system_counter = (unsigned long) cpu_system_counter;
 745         ps->vmem_rss = (unsigned long) vmem_rss;
 746
 747         /* success */
 748         return (0);
 749 } /* int ps_read_process (...) */
 750 #endif /* KERNEL_LINUX */
 751
 752 #if HAVE_THREAD_INFO
 753 static int mach_get_task_name (task_t t, int *pid, char *name, size_t name_max_len)
 754 {
 755         int mib[4];
 756
 757         struct kinfo_proc kp;
 758         size_t            kp_size;
 759
 760         mib[0] = CTL_KERN;
 761         mib[1] = KERN_PROC;
 762         mib[2] = KERN_PROC_PID;
 763
 764         if (pid_for_task (t, pid) != KERN_SUCCESS)
 765                 return (-1);
 766         mib[3] = *pid;
 767
 768         kp_size = sizeof (kp);
 769         if (sysctl (mib, 4, &kp, &kp_size, NULL, 0) != 0)
 770                 return (-1);
 771
 772         if (name_max_len > (MAXCOMLEN + 1))
 773                 name_max_len = MAXCOMLEN + 1;
 774
 775         strncpy (name, kp.kp_proc.p_comm, name_max_len - 1);
 776         name[name_max_len - 1] = '\0';
 777
 778         DEBUG ("pid = %i; name = %s;", *pid, name);
 779
 780         /* We don't do the special handling for `p_comm == "LaunchCFMApp"' as
 781          * `top' does it, because it is a lot of work and only used when
 782          * debugging. -octo */
 783
 784         return (0);
 785 }
 786 #endif /* HAVE_THREAD_INFO */
 787 /* ------- end of additional functions for KERNEL_LINUX/HAVE_THREAD_INFO ------- */
 788
 789 /* do actual readings from kernel */
 790 static int ps_read (void)
 791 {
 792 #if HAVE_THREAD_INFO
 793         kern_return_t            status;
 794
 795         int                      pset;
 796         processor_set_t          port_pset_priv;
 797
 798         int                      task;
 799         task_array_t             task_list;
 800         mach_msg_type_number_t   task_list_len;
 801
 802         int                      task_pid;
 803         char                     task_name[MAXCOMLEN + 1];
 804
 805         int                      thread;
 806         thread_act_array_t       thread_list;
 807         mach_msg_type_number_t   thread_list_len;
 808         thread_basic_info_data_t thread_data;
 809         mach_msg_type_number_t   thread_data_len;
 810
 811         int running  = 0;
 812         int sleeping = 0;
 813         int zombies  = 0;
 814         int stopped  = 0;
 815         int blocked  = 0;
 816
 817         procstat_t *ps;
 818         procstat_entry_t pse;
 819
 820         ps_list_reset ();
 821
 822         /*
 823          * The Mach-concept is a little different from the traditional UNIX
 824          * concept: All the work is done in threads. Threads are contained in
 825          * `tasks'. Therefore, `task status' doesn't make much sense, since
 826          * it's actually a `thread status'.
 827          * Tasks are assigned to sets of processors, so that's where you go to
 828          * get a list.
 829          */
 830         for (pset = 0; pset < pset_list_len; pset++)
 831         {
 832                 if ((status = host_processor_set_priv (port_host_self,
 833                                                 pset_list[pset],
 834                                                 &port_pset_priv)) != KERN_SUCCESS)
 835                 {
 836                         ERROR ("host_processor_set_priv failed: %s\n",
 837                                         mach_error_string (status));
 838                         continue;
 839                 }
 840
 841                 if ((status = processor_set_tasks (port_pset_priv,
 842                                                 &task_list,
 843                                                 &task_list_len)) != KERN_SUCCESS)
 844                 {
 845                         ERROR ("processor_set_tasks failed: %s\n",
 846                                         mach_error_string (status));
 847                         mach_port_deallocate (port_task_self, port_pset_priv);
 848                         continue;
 849                 }
 850
 851                 for (task = 0; task < task_list_len; task++)
 852                 {
 853                         ps = NULL;
 854                         if (mach_get_task_name (task_list[task],
 855                                                 &task_pid,
 856                                                 task_name, PROCSTAT_NAME_LEN) == 0)
 857                         {
 858                                 /* search for at least one match */
 859                                 for (ps = list_head_g; ps != NULL; ps = ps->next)
 860                                         /* FIXME: cmdline should be here instead of NULL */
 861                                         if (ps_list_match (task_name, NULL, ps) == 1)
 862                                                 break;
 863                         }
 864
 865                         /* Collect more detailed statistics for this process */
 866                         if (ps != NULL)
 867                         {
 868                                 task_basic_info_data_t        task_basic_info;
 869                                 mach_msg_type_number_t        task_basic_info_len;
 870                                 task_events_info_data_t       task_events_info;
 871                                 mach_msg_type_number_t        task_events_info_len;
 872                                 task_absolutetime_info_data_t task_absolutetime_info;
 873                                 mach_msg_type_number_t        task_absolutetime_info_len;
 874
 875                                 memset (&pse, '\0', sizeof (pse));
 876                                 pse.id = task_pid;
 877
 878                                 task_basic_info_len = TASK_BASIC_INFO_COUNT;
 879                                 status = task_info (task_list[task],
 880                                                 TASK_BASIC_INFO,
 881                                                 (task_info_t) &task_basic_info,
 882                                                 &task_basic_info_len);
 883                                 if (status != KERN_SUCCESS)
 884                                 {
 885                                         ERROR ("task_info failed: %s",
 886                                                         mach_error_string (status));
 887                                         continue; /* with next thread_list */
 888                                 }
 889
 890                                 task_events_info_len = TASK_EVENTS_INFO_COUNT;
 891                                 status = task_info (task_list[task],
 892                                                 TASK_EVENTS_INFO,
 893                                                 (task_info_t) &task_events_info,
 894                                                 &task_events_info_len);
 895                                 if (status != KERN_SUCCESS)
 896                                 {
 897                                         ERROR ("task_info failed: %s",
 898                                                         mach_error_string (status));
 899                                         continue; /* with next thread_list */
 900                                 }
 901
 902                                 task_absolutetime_info_len = TASK_ABSOLUTETIME_INFO_COUNT;
 903                                 status = task_info (task_list[task],
 904                                                 TASK_ABSOLUTETIME_INFO,
 905                                                 (task_info_t) &task_absolutetime_info,
 906                                                 &task_absolutetime_info_len);
 907                                 if (status != KERN_SUCCESS)
 908                                 {
 909                                         ERROR ("task_info failed: %s",
 910                                                         mach_error_string (status));
 911                                         continue; /* with next thread_list */
 912                                 }
 913
 914                                 pse.num_proc++;
 915                                 pse.vmem_rss = task_basic_info.resident_size;
 916
 917                                 pse.vmem_minflt_counter = task_events_info.cow_faults;
 918                                 pse.vmem_majflt_counter = task_events_info.faults;
 919
 920                                 pse.cpu_user_counter = task_absolutetime_info.total_user;
 921                                 pse.cpu_system_counter = task_absolutetime_info.total_system;
 922                         }
 923
 924                         status = task_threads (task_list[task], &thread_list,
 925                                         &thread_list_len);
 926                         if (status != KERN_SUCCESS)
 927                         {
 928                                 /* Apple's `top' treats this case a zombie. It
 929                                  * makes sense to some extend: A `zombie'
 930                                  * thread is nonsense, since the task/process
 931                                  * is dead. */
 932                                 zombies++;
 933                                 DEBUG ("task_threads failed: %s",
 934                                                 mach_error_string (status));
 935                                 if (task_list[task] != port_task_self)
 936                                         mach_port_deallocate (port_task_self,
 937                                                         task_list[task]);
 938                                 continue; /* with next task_list */
 939                         }
 940
 941                         for (thread = 0; thread < thread_list_len; thread++)
 942                         {
 943                                 thread_data_len = THREAD_BASIC_INFO_COUNT;
 944                                 status = thread_info (thread_list[thread],
 945                                                 THREAD_BASIC_INFO,
 946                                                 (thread_info_t) &thread_data,
 947                                                 &thread_data_len);
 948                                 if (status != KERN_SUCCESS)
 949                                 {
 950                                         ERROR ("thread_info failed: %s",
 951                                                         mach_error_string (status));
 952                                         if (task_list[task] != port_task_self)
 953                                                 mach_port_deallocate (port_task_self,
 954                                                                 thread_list[thread]);
 955                                         continue; /* with next thread_list */
 956                                 }
 957
 958                                 if (ps != NULL)
 959                                         pse.num_lwp++;
 960
 961                                 switch (thread_data.run_state)
 962                                 {
 963                                         case TH_STATE_RUNNING:
 964                                                 running++;
 965                                                 break;
 966                                         case TH_STATE_STOPPED:
 967                                         /* What exactly is `halted'? */
 968                                         case TH_STATE_HALTED:
 969                                                 stopped++;
 970                                                 break;
 971                                         case TH_STATE_WAITING:
 972                                                 sleeping++;
 973                                                 break;
 974                                         case TH_STATE_UNINTERRUPTIBLE:
 975                                                 blocked++;
 976                                                 break;
 977                                         /* There is no `zombie' case here,
 978                                          * since there are no zombie-threads.
 979                                          * There's only zombie tasks, which are
 980                                          * handled above. */
 981                                         default:
 982                                                 WARNING ("Unknown thread status: %s",
 983                                                                 thread_data.run_state);
 984                                                 break;
 985                                 } /* switch (thread_data.run_state) */
 986
 987                                 if (task_list[task] != port_task_self)
 988                                 {
 989                                         status = mach_port_deallocate (port_task_self,
 990                                                         thread_list[thread]);
 991                                         if (status != KERN_SUCCESS)
 992                                                 ERROR ("mach_port_deallocate failed: %s",
 993                                                                 mach_error_string (status));
 994                                 }
 995                         } /* for (thread_list) */
 996
 997                         if ((status = vm_deallocate (port_task_self,
 998                                                         (vm_address_t) thread_list,
 999                                                         thread_list_len * sizeof (thread_act_t)))
1000                                         != KERN_SUCCESS)
1001                         {
1002                                 ERROR ("vm_deallocate failed: %s",
1003                                                 mach_error_string (status));
1004                         }
1005                         thread_list = NULL;
1006                         thread_list_len = 0;
1007
1008                         /* Only deallocate the task port, if it isn't our own.
1009                          * Don't know what would happen in that case, but this
1010                          * is what Apple's top does.. ;) */
1011                         if (task_list[task] != port_task_self)
1012                         {
1013                                 status = mach_port_deallocate (port_task_self,
1014                                                 task_list[task]);
1015                                 if (status != KERN_SUCCESS)
1016                                         ERROR ("mach_port_deallocate failed: %s",
1017                                                         mach_error_string (status));
1018                         }
1019
1020                         if (ps != NULL)
1021                                 /* FIXME: cmdline should be here instead of NULL */
1022                                 ps_list_add (task_name, NULL, &pse);
1023                 } /* for (task_list) */
1024
1025                 if ((status = vm_deallocate (port_task_self,
1026                                 (vm_address_t) task_list,
1027                                 task_list_len * sizeof (task_t))) != KERN_SUCCESS)
1028                 {
1029                         ERROR ("vm_deallocate failed: %s",
1030                                         mach_error_string (status));
1031                 }
1032                 task_list = NULL;
1033                 task_list_len = 0;
1034
1035                 if ((status = mach_port_deallocate (port_task_self, port_pset_priv))
1036                                 != KERN_SUCCESS)
1037                 {
1038                         ERROR ("mach_port_deallocate failed: %s",
1039                                         mach_error_string (status));
1040                 }
1041         } /* for (pset_list) */
1042
1043         ps_submit_state ("running", running);
1044         ps_submit_state ("sleeping", sleeping);
1045         ps_submit_state ("zombies", zombies);
1046         ps_submit_state ("stopped", stopped);
1047         ps_submit_state ("blocked", blocked);
1048
1049         for (ps = list_head_g; ps != NULL; ps = ps->next)
1050                 ps_submit_proc_list (ps);
1051 /* #endif HAVE_THREAD_INFO */
1052
1053 #elif KERNEL_LINUX
1054         int running  = 0;
1055         int sleeping = 0;
1056         int zombies  = 0;
1057         int stopped  = 0;
1058         int paging   = 0;
1059         int blocked  = 0;
1060
1061         struct dirent *ent;
1062         DIR           *proc;
1063         int            pid;
1064
1065         int        status;
1066         procstat_t ps;
1067         procstat_entry_t pse;
1068         char       state;
1069
1070         procstat_t *ps_ptr;
1071
1072         running = sleeping = zombies = stopped = paging = blocked = 0;
1073         ps_list_reset ();
1074
1075         if ((proc = opendir ("/proc")) == NULL)
1076         {
1077                 char errbuf[1024];
1078                 ERROR ("Cannot open `/proc': %s",
1079                                 sstrerror (errno, errbuf, sizeof (errbuf)));
1080                 return (-1);
1081         }
1082
1083         while ((ent = readdir (proc)) != NULL)
1084         {
1085                 if (!isdigit (ent->d_name[0]))
1086                         continue;
1087
1088                 if ((pid = atoi (ent->d_name)) < 1)
1089                         continue;
1090
1091                 status = ps_read_process (pid, &ps, &state);
1092                 if (status != 0)
1093                 {
1094                         DEBUG ("ps_read_process failed: %i", status);
1095                         continue;
1096                 }
1097
1098                 pse.id       = pid;
1099                 pse.age      = 0;
1100
1101                 pse.num_proc = ps.num_proc;
1102                 pse.num_lwp  = ps.num_lwp;
1103                 pse.vmem_rss = ps.vmem_rss;
1104
1105                 pse.vmem_minflt = 0;
1106                 pse.vmem_minflt_counter = ps.vmem_minflt_counter;
1107                 pse.vmem_majflt = 0;
1108                 pse.vmem_majflt_counter = ps.vmem_majflt_counter;
1109
1110                 pse.cpu_user = 0;
1111                 pse.cpu_user_counter = ps.cpu_user_counter;
1112                 pse.cpu_system = 0;
1113                 pse.cpu_system_counter = ps.cpu_system_counter;
1114
1115                 switch (state)
1116                 {
1117                         case 'R': running++;  break;
1118                         case 'S': sleeping++; break;
1119                         case 'D': blocked++;  break;
1120                         case 'Z': zombies++;  break;
1121                         case 'T': stopped++;  break;
1122                         case 'W': paging++;   break;
1123                 }
1124
1125                 /* FIXME: cmdline should be here instead of NULL */
1126                 ps_list_add (ps.name, NULL, &pse);
1127         }
1128
1129         closedir (proc);
1130
1131         ps_submit_state ("running",  running);
1132         ps_submit_state ("sleeping", sleeping);
1133         ps_submit_state ("zombies",  zombies);
1134         ps_submit_state ("stopped",  stopped);
1135         ps_submit_state ("paging",   paging);
1136         ps_submit_state ("blocked",  blocked);
1137
1138         for (ps_ptr = list_head_g; ps_ptr != NULL; ps_ptr = ps_ptr->next)
1139                 ps_submit_proc_list (ps_ptr);
1140 /* #endif KERNEL_LINUX */
1141
1142 #elif HAVE_LIBKVM_GETPROCS
1143         int running  = 0;
1144         int sleeping = 0;
1145         int zombies  = 0;
1146         int stopped  = 0;
1147         int blocked  = 0;
1148         int idle     = 0;
1149         int wait     = 0;
1150
1151         kvm_t *kd;
1152         char errbuf[1024];
1153         char cmdline[ARG_MAX];
1154         char *cmdline_ptr;
1155         struct kinfo_proc *procs;          /* array of processes */
1156         char **argv;
1157         int count;                         /* returns number of processes */
1158         int i;
1159
1160         procstat_t *ps_ptr;
1161         procstat_entry_t pse;
1162
1163         ps_list_reset ();
1164
1165         /* Open the kvm interface, get a descriptor */
1166         kd = kvm_open (NULL, NULL, NULL, 0, errbuf);
1167         if (kd == NULL)
1168         {
1169                 ERROR ("processes plugin: Cannot open kvm interface: %s",
1170                                 errbuf);
1171                 return (0);
1172         }
1173
1174         /* Get the list of processes. */
1175         procs = kvm_getprocs(kd, KERN_PROC_ALL, 0, &count);
1176         if (procs == NULL)
1177         {
1178                 kvm_close (kd);
1179                 ERROR ("processes plugin: Cannot get kvm processes list: %s",
1180                                 kvm_geterr(kd));
1181                 return (0);
1182         }
1183
1184         /* Iterate through the processes in kinfo_proc */
1185         for (i = 0; i < count; i++)
1186         {
1187                 /* retrieve the arguments */
1188                 cmdline[0] = 0;
1189                 cmdline_ptr = NULL;
1190
1191                 argv = kvm_getargv (kd, (const struct kinfo_proc *) &(procs[i]), 0);
1192                 if (argv != NULL)
1193                 {
1194                         int status;
1195                         int argc;
1196
1197                         argc = 0;
1198                         while (argv[argc] != NULL)
1199                                 argc++;
1200
1201                         status = strjoin (cmdline, sizeof (cmdline),
1202                                         argv, argc, " ");
1203
1204                         if (status < 0)
1205                         {
1206                                 WARNING ("processes plugin: Command line did "
1207                                                 "not fit into buffer.");
1208                         }
1209                         else
1210                         {
1211                                 cmdline_ptr = &cmdline[0];
1212                         }
1213                 }
1214
1215                 pse.id       = procs[i].ki_pid;
1216                 pse.age      = 0;
1217
1218                 pse.num_proc = 1;
1219                 pse.num_lwp  = procs[i].ki_numthreads;
1220
1221                 pse.vmem_rss = procs[i].ki_rssize * getpagesize();
1222                 pse.vmem_minflt = 0;
1223                 pse.vmem_minflt_counter = procs[i].ki_rusage.ru_minflt;
1224                 pse.vmem_majflt = 0;
1225                 pse.vmem_majflt_counter = procs[i].ki_rusage.ru_majflt;
1226
1227                 pse.cpu_user = 0;
1228                 pse.cpu_user_counter = procs[i].ki_rusage.ru_utime.tv_sec
1229                         * 1000
1230                         + procs[i].ki_rusage.ru_utime.tv_usec;
1231                 pse.cpu_system = 0;
1232                 pse.cpu_system_counter = procs[i].ki_rusage.ru_stime.tv_sec
1233                         * 1000
1234                         + procs[i].ki_rusage.ru_stime.tv_usec;
1235
1236                 switch (procs[i].ki_stat)
1237                 {
1238                         case SSTOP:     stopped++;      break;
1239                         case SSLEEP:    sleeping++;     break;
1240                         case SRUN:      running++;      break;
1241                         case SIDL:      idle++;         break;
1242                         case SWAIT:     wait++;         break;
1243                         case SLOCK:     blocked++;      break;
1244                         case SZOMB:     zombies++;      break;
1245                 }
1246
1247                 ps_list_add (procs[i].ki_comm, cmdline_ptr, &pse);
1248         }
1249
1250         kvm_close(kd);
1251
1252         ps_submit_state ("running",  running);
1253         ps_submit_state ("sleeping", sleeping);
1254         ps_submit_state ("zombies",  zombies);
1255         ps_submit_state ("stopped",  stopped);
1256         ps_submit_state ("blocked",  blocked);
1257         ps_submit_state ("idle",     idle);
1258         ps_submit_state ("wait",     wait);
1259
1260         for (ps_ptr = list_head_g; ps_ptr != NULL; ps_ptr = ps_ptr->next)
1261                 ps_submit_proc_list (ps_ptr);
1262 #endif /* HAVE_LIBKVM_GETPROCS */
1263
1264         return (0);
1265 } /* int ps_read */
1266
1267 void module_register (void)
1268 {
1269         plugin_register_config ("processes", ps_config,
1270                         config_keys, config_keys_num);
1271         plugin_register_init ("processes", ps_init);
1272         plugin_register_read ("processes", ps_read);
1273 } /* void module_register */