src/processes.c

   1 /**
   2  * collectd - src/processes.c
   3  * Copyright (C) 2005  Lyonel Vincent
   4  * Copyright (C) 2006-2007  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 "configfile.h"
  29
  30 /* Include header files for the mach system, if they exist.. */
  31 #if HAVE_THREAD_INFO
  32 #  if HAVE_MACH_MACH_INIT_H
  33 #    include <mach/mach_init.h>
  34 #  endif
  35 #  if HAVE_MACH_HOST_PRIV_H
  36 #    include <mach/host_priv.h>
  37 #  endif
  38 #  if HAVE_MACH_MACH_ERROR_H
  39 #    include <mach/mach_error.h>
  40 #  endif
  41 #  if HAVE_MACH_MACH_HOST_H
  42 #    include <mach/mach_host.h>
  43 #  endif
  44 #  if HAVE_MACH_MACH_PORT_H
  45 #    include <mach/mach_port.h>
  46 #  endif
  47 #  if HAVE_MACH_MACH_TYPES_H
  48 #    include <mach/mach_types.h>
  49 #  endif
  50 #  if HAVE_MACH_MESSAGE_H
  51 #    include <mach/message.h>
  52 #  endif
  53 #  if HAVE_MACH_PROCESSOR_SET_H
  54 #    include <mach/processor_set.h>
  55 #  endif
  56 #  if HAVE_MACH_TASK_H
  57 #    include <mach/task.h>
  58 #  endif
  59 #  if HAVE_MACH_THREAD_ACT_H
  60 #    include <mach/thread_act.h>
  61 #  endif
  62 #  if HAVE_MACH_VM_REGION_H
  63 #    include <mach/vm_region.h>
  64 #  endif
  65 #  if HAVE_MACH_VM_MAP_H
  66 #    include <mach/vm_map.h>
  67 #  endif
  68 #  if HAVE_MACH_VM_PROT_H
  69 #    include <mach/vm_prot.h>
  70 #  endif
  71 #  if HAVE_SYS_SYSCTL_H
  72 #    include <sys/sysctl.h>
  73 #  endif
  74 /* #endif HAVE_THREAD_INFO */
  75
  76 #elif KERNEL_LINUX
  77 #  if HAVE_LINUX_CONFIG_H
  78 #    include <linux/config.h>
  79 #  endif
  80 #  ifndef CONFIG_HZ
  81 #    define CONFIG_HZ 100
  82 #  endif
  83 /* #endif KERNEL_LINUX */
  84
  85 #elif HAVE_KVM_H
  86 #  include <kvm.h>
  87 #  include <sys/user.h>
  88 #  include <sys/proc.h>
  89 #  if HAVE_SYS_SYSCTL_H
  90 #    include <sys/sysctl.h>
  91 #  endif
  92 /* #endif HAVE_KVM_H */
  93
  94 #else
  95 # error "No applicable input method."
  96 #endif
  97
  98 #if HAVE_REGEX_H
  99 # include <regex.h>
 100 #endif
 101
 102 #define BUFSIZE 256
 103
 104 static const char *config_keys[] =
 105 {
 106         "Process",
 107         "ProcessMatch",
 108         NULL
 109 };
 110 static int config_keys_num = STATIC_ARRAY_SIZE (config_keys);
 111
 112 typedef struct procstat_entry_s
 113 {
 114         unsigned long id;
 115         unsigned long age;
 116
 117         unsigned long num_proc;
 118         unsigned long num_lwp;
 119         unsigned long vmem_rss;
 120
 121         unsigned long vmem_minflt;
 122         unsigned long vmem_majflt;
 123         unsigned long vmem_minflt_counter;
 124         unsigned long vmem_majflt_counter;
 125
 126         unsigned long cpu_user;
 127         unsigned long cpu_system;
 128         unsigned long cpu_user_counter;
 129         unsigned long cpu_system_counter;
 130
 131         struct procstat_entry_s *next;
 132 } procstat_entry_t;
 133
 134 #define PROCSTAT_NAME_LEN 256
 135 typedef struct procstat
 136 {
 137         char          name[PROCSTAT_NAME_LEN];
 138 #if HAVE_REGEX_H
 139         regex_t *re;
 140 #endif
 141
 142         unsigned long num_proc;
 143         unsigned long num_lwp;
 144         unsigned long vmem_rss;
 145
 146         unsigned long vmem_minflt_counter;
 147         unsigned long vmem_majflt_counter;
 148
 149         unsigned long cpu_user_counter;
 150         unsigned long cpu_system_counter;
 151
 152         struct procstat   *next;
 153         struct procstat_entry_s *instances;
 154 } procstat_t;
 155
 156 static procstat_t *list_head_g = NULL;
 157
 158 #if HAVE_THREAD_INFO
 159 static mach_port_t port_host_self;
 160 static mach_port_t port_task_self;
 161
 162 static processor_set_name_array_t pset_list;
 163 static mach_msg_type_number_t     pset_list_len;
 164 /* #endif HAVE_THREAD_INFO */
 165
 166 #elif KERNEL_LINUX
 167 static long pagesize_g;
 168 #endif /* KERNEL_LINUX */
 169
 170 /* put name of process from config to list_head_g tree
 171    list_head_g is a list of 'procstat_t' structs with
 172    processes names we want to watch */
 173 static void ps_list_register (const char *name, const char *regexp)
 174 {
 175         procstat_t *new;
 176         procstat_t *ptr;
 177         int status;
 178
 179         new = (procstat_t *) malloc (sizeof (procstat_t));
 180         if (new == NULL)
 181         {
 182                 ERROR ("processes plugin: ps_list_register: malloc failed.");
 183                 return;
 184         }
 185         memset (new, 0, sizeof (procstat_t));
 186         sstrncpy (new->name, name, sizeof (new->name));
 187
 188 #if HAVE_REGEX_H
 189         if (regexp != NULL)
 190         {
 191                 DEBUG ("ProcessMatch: adding \"%s\" as criteria to process %s.", regexp, name);
 192                 new->re = (regex_t *) malloc (sizeof (regex_t));
 193                 if (new->re == NULL)
 194                 {
 195                         ERROR ("processes plugin: ps_list_register: malloc failed.");
 196                         sfree (new);
 197                         return;
 198                 }
 199
 200                 status = regcomp (new->re, regexp, REG_EXTENDED | REG_NOSUB);
 201                 if (status != 0)
 202                 {
 203                         DEBUG ("ProcessMatch: compiling the regular expression \"%s\" failed.", regexp);
 204                         sfree(new->re);
 205                         return;
 206                 }
 207         }
 208 #else
 209         if (regexp != NULL)
 210         {
 211                 ERROR ("processes plugin: ps_list_register: "
 212                                 "Regular expression \"%s\" found in config "
 213                                 "file, but support for regular expressions "
 214                                 "has been dispabled at compile time.",
 215                                 regexp);
 216                 sfree (new);
 217                 return;
 218         }
 219 #endif
 220
 221         for (ptr = list_head_g; ptr != NULL; ptr = ptr->next)
 222         {
 223                 if (strcmp (ptr->name, name) == 0)
 224                 {
 225                         WARNING ("processes plugin: You have configured more "
 226                                         "than one `Process' or "
 227                                         "`ProcessMatch' with the same name. "
 228                                         "All but the first setting will be "
 229                                         "ignored.");
 230                         sfree (new->re);
 231                         sfree (new);
 232                         return;
 233                 }
 234
 235                 if (ptr->next == NULL)
 236                         break;
 237         }
 238
 239         if (ptr == NULL)
 240                 list_head_g = new;
 241         else
 242                 ptr->next = new;
 243 } /* void ps_list_register */
 244
 245 /* try to match name against entry, returns 1 if success */
 246 static int ps_list_match (const char *name, const char *cmdline, procstat_t *ps)
 247 {
 248         if ((ps->re != NULL) && (regexec(ps->re, (strlen(cmdline)!=0)?cmdline:name, 0, NULL, 0) == 0))
 249                 return (1);
 250         if (strcmp (ps->name, name) == 0) {
 251                 return (1);
 252         }
 253         return (0);
 254 }
 255
 256 /* add process entry to 'instances' of process 'name' (or refresh it) */
 257 static void ps_list_add (const char *name, const char *cmdline, procstat_entry_t *entry)
 258 {
 259         procstat_t *ps;
 260         procstat_entry_t *pse;
 261
 262         if (entry->id == 0)
 263                 return;
 264
 265         for (ps = list_head_g; ps != NULL; ps = ps->next)
 266         {
 267
 268                 if ((ps_list_match (name, cmdline, ps)) == 0)
 269                         continue;
 270
 271                 for (pse = ps->instances; pse != NULL; pse = pse->next)
 272                         if ((pse->id == entry->id) || (pse->next == NULL))
 273                                 break;
 274
 275                 if ((pse == NULL) || (pse->id != entry->id))
 276                 {
 277                         procstat_entry_t *new;
 278
 279                         new = (procstat_entry_t *) malloc (sizeof (procstat_entry_t));
 280                         if (new == NULL)
 281                                 return;
 282                         memset (new, 0, sizeof (procstat_entry_t));
 283                         new->id = entry->id;
 284
 285                         if (pse == NULL)
 286                                 ps->instances = new;
 287                         else
 288                                 pse->next = new;
 289
 290                         pse = new;
 291                 }
 292
 293                 pse->age = 0;
 294                 pse->num_proc = entry->num_proc;
 295                 pse->num_lwp  = entry->num_lwp;
 296                 pse->vmem_rss = entry->vmem_rss;
 297
 298                 ps->num_proc += pse->num_proc;
 299                 ps->num_lwp  += pse->num_lwp;
 300                 ps->vmem_rss += pse->vmem_rss;
 301
 302                 if ((entry->vmem_minflt_counter == 0)
 303                                 && (entry->vmem_majflt_counter == 0))
 304                 {
 305                         pse->vmem_minflt_counter += entry->vmem_minflt;
 306                         pse->vmem_minflt = entry->vmem_minflt;
 307
 308                         pse->vmem_majflt_counter += entry->vmem_majflt;
 309                         pse->vmem_majflt = entry->vmem_majflt;
 310                 }
 311                 else
 312                 {
 313                         if (entry->vmem_minflt_counter < pse->vmem_minflt_counter)
 314                         {
 315                                 pse->vmem_minflt = entry->vmem_minflt_counter
 316                                         + (ULONG_MAX - pse->vmem_minflt_counter);
 317                         }
 318                         else
 319                         {
 320                                 pse->vmem_minflt = entry->vmem_minflt_counter - pse->vmem_minflt_counter;
 321                         }
 322                         pse->vmem_minflt_counter = entry->vmem_minflt_counter;
 323
 324                         if (entry->vmem_majflt_counter < pse->vmem_majflt_counter)
 325                         {
 326                                 pse->vmem_majflt = entry->vmem_majflt_counter
 327                                         + (ULONG_MAX - pse->vmem_majflt_counter);
 328                         }
 329                         else
 330                         {
 331                                 pse->vmem_majflt = entry->vmem_majflt_counter - pse->vmem_majflt_counter;
 332                         }
 333                         pse->vmem_majflt_counter = entry->vmem_majflt_counter;
 334                 }
 335
 336                 ps->vmem_minflt_counter += pse->vmem_minflt;
 337                 ps->vmem_majflt_counter += pse->vmem_majflt;
 338
 339                 if ((entry->cpu_user_counter == 0)
 340                                 && (entry->cpu_system_counter == 0))
 341                 {
 342                         pse->cpu_user_counter += entry->cpu_user;
 343                         pse->cpu_user = entry->cpu_user;
 344
 345                         pse->cpu_system_counter += entry->cpu_system;
 346                         pse->cpu_system = entry->cpu_system;
 347                 }
 348                 else
 349                 {
 350                         if (entry->cpu_user_counter < pse->cpu_user_counter)
 351                         {
 352                                 pse->cpu_user = entry->cpu_user_counter
 353                                         + (ULONG_MAX - pse->cpu_user_counter);
 354                         }
 355                         else
 356                         {
 357                                 pse->cpu_user = entry->cpu_user_counter - pse->cpu_user_counter;
 358                         }
 359                         pse->cpu_user_counter = entry->cpu_user_counter;
 360
 361                         if (entry->cpu_system_counter < pse->cpu_system_counter)
 362                         {
 363                                 pse->cpu_system = entry->cpu_system_counter
 364                                         + (ULONG_MAX - pse->cpu_system_counter);
 365                         }
 366                         else
 367                         {
 368                                 pse->cpu_system = entry->cpu_system_counter - pse->cpu_system_counter;
 369                         }
 370                         pse->cpu_system_counter = entry->cpu_system_counter;
 371                 }
 372
 373                 ps->cpu_user_counter   += pse->cpu_user;
 374                 ps->cpu_system_counter += pse->cpu_system;
 375         }
 376 }
 377
 378 /* remove old entries from instances of processes in list_head_g */
 379 static void ps_list_reset (void)
 380 {
 381         procstat_t *ps;
 382         procstat_entry_t *pse;
 383         procstat_entry_t *pse_prev;
 384
 385         for (ps = list_head_g; ps != NULL; ps = ps->next)
 386         {
 387                 ps->num_proc    = 0;
 388                 ps->num_lwp     = 0;
 389                 ps->vmem_rss    = 0;
 390
 391                 pse_prev = NULL;
 392                 pse = ps->instances;
 393                 while (pse != NULL)
 394                 {
 395                         if (pse->age > 10)
 396                         {
 397                                 DEBUG ("Removing this procstat entry cause it's too old: "
 398                                                 "id = %lu; name = %s;",
 399                                                 pse->id, ps->name);
 400
 401                                 if (pse_prev == NULL)
 402                                 {
 403                                         ps->instances = pse->next;
 404                                         free (pse);
 405                                         pse = ps->instances;
 406                                 }
 407                                 else
 408                                 {
 409                                         pse_prev->next = pse->next;
 410                                         free (pse);
 411                                         pse = pse_prev->next;
 412                                 }
 413                         }
 414                         else
 415                         {
 416                                 pse->age++;
 417                                 pse_prev = pse;
 418                                 pse = pse->next;
 419                         }
 420                 } /* while (pse != NULL) */
 421         } /* for (ps = list_head_g; ps != NULL; ps = ps->next) */
 422 }
 423
 424 /* put all pre-defined 'Process' names from config to list_head_g tree */
 425 static int ps_config (const char *key, const char *value)
 426 {
 427         char *new_val;
 428         char *fields[2];
 429         int fields_num;
 430
 431         if (strcasecmp (key, "Process") == 0)
 432         {
 433                 ps_list_register (value, NULL);
 434                 return (0);
 435         }
 436
 437         if (strcasecmp (key, "ProcessMatch") == 0)
 438         {
 439                 new_val = strdup (value);
 440                 if (new_val == NULL)
 441                         return (-1);
 442                 fields_num = strsplit (new_val, fields, 2);
 443                 if (fields_num != 2)
 444                 {
 445                         sfree (new_val);
 446                         return (-1);
 447                 }
 448                 ps_list_register (fields[0], fields[1]);
 449                 sfree (new_val);
 450                 return (0);
 451         }
 452
 453         return (-1);
 454 }
 455
 456 static int ps_init (void)
 457 {
 458 #if HAVE_THREAD_INFO
 459         kern_return_t status;
 460
 461         port_host_self = mach_host_self ();
 462         port_task_self = mach_task_self ();
 463
 464         if (pset_list != NULL)
 465         {
 466                 vm_deallocate (port_task_self,
 467                                 (vm_address_t) pset_list,
 468                                 pset_list_len * sizeof (processor_set_t));
 469                 pset_list = NULL;
 470                 pset_list_len = 0;
 471         }
 472
 473         if ((status = host_processor_sets (port_host_self,
 474                                         &pset_list,
 475                                         &pset_list_len)) != KERN_SUCCESS)
 476         {
 477                 ERROR ("host_processor_sets failed: %s\n",
 478                                 mach_error_string (status));
 479                 pset_list = NULL;
 480                 pset_list_len = 0;
 481                 return (-1);
 482         }
 483 /* #endif HAVE_THREAD_INFO */
 484
 485 #elif KERNEL_LINUX
 486         pagesize_g = sysconf(_SC_PAGESIZE);
 487         DEBUG ("pagesize_g = %li; CONFIG_HZ = %i;",
 488                         pagesize_g, CONFIG_HZ);
 489 #endif /* KERNEL_LINUX */
 490
 491         return (0);
 492 } /* int ps_init */
 493
 494 /* submit global state (e.g.: qty of zombies, running, etc..) */
 495 static void ps_submit_state (const char *state, double value)
 496 {
 497         value_t values[1];
 498         value_list_t vl = VALUE_LIST_INIT;
 499
 500         values[0].gauge = value;
 501
 502         vl.values = values;
 503         vl.values_len = 1;
 504         vl.time = time (NULL);
 505         strcpy (vl.host, hostname_g);
 506         strcpy (vl.plugin, "processes");
 507         strcpy (vl.plugin_instance, "");
 508         strcpy (vl.type, "ps_state");
 509         sstrncpy (vl.type_instance, state, sizeof (vl.type_instance));
 510
 511         plugin_dispatch_values (&vl);
 512 }
 513
 514 /* submit info about specific process (e.g.: memory taken, cpu usage, etc..) */
 515 static void ps_submit_proc_list (procstat_t *ps)
 516 {
 517         value_t values[2];
 518         value_list_t vl = VALUE_LIST_INIT;
 519
 520         vl.values = values;
 521         vl.values_len = 2;
 522         vl.time = time (NULL);
 523         strcpy (vl.host, hostname_g);
 524         strcpy (vl.plugin, "processes");
 525         sstrncpy (vl.plugin_instance, ps->name, sizeof (vl.plugin_instance));
 526
 527         strcpy (vl.type, "ps_rss");
 528         vl.values[0].gauge = ps->vmem_rss;
 529         vl.values_len = 1;
 530         plugin_dispatch_values (&vl);
 531
 532         strcpy (vl.type, "ps_cputime");
 533         vl.values[0].counter = ps->cpu_user_counter;
 534         vl.values[1].counter = ps->cpu_system_counter;
 535         vl.values_len = 2;
 536         plugin_dispatch_values (&vl);
 537
 538         strcpy (vl.type, "ps_count");
 539         vl.values[0].gauge = ps->num_proc;
 540         vl.values[1].gauge = ps->num_lwp;
 541         vl.values_len = 2;
 542         plugin_dispatch_values (&vl);
 543
 544         strcpy (vl.type, "ps_pagefaults");
 545         vl.values[0].counter = ps->vmem_minflt_counter;
 546         vl.values[1].counter = ps->vmem_majflt_counter;
 547         vl.values_len = 2;
 548         plugin_dispatch_values (&vl);
 549
 550         DEBUG ("name = %s; num_proc = %lu; num_lwp = %lu; vmem_rss = %lu; "
 551                         "vmem_minflt_counter = %lu; vmem_majflt_counter = %lu; "
 552                         "cpu_user_counter = %lu; cpu_system_counter = %lu;",
 553                         ps->name, ps->num_proc, ps->num_lwp, ps->vmem_rss,
 554                         ps->vmem_minflt_counter, ps->vmem_majflt_counter,
 555                         ps->cpu_user_counter, ps->cpu_system_counter);
 556 } /* void ps_submit_proc_list */
 557
 558 /* ------- additional functions for KERNEL_LINUX/HAVE_THREAD_INFO ------- */
 559 #if KERNEL_LINUX
 560 static int *ps_read_tasks (int pid)
 561 {
 562         int *list = NULL;
 563         int  list_size = 1; /* size of allocated space, in elements */
 564         int  list_len = 0;  /* number of currently used elements */
 565
 566         char           dirname[64];
 567         DIR           *dh;
 568         struct dirent *ent;
 569
 570         ssnprintf (dirname, sizeof (dirname), "/proc/%i/task", pid);
 571
 572         if ((dh = opendir (dirname)) == NULL)
 573         {
 574                 DEBUG ("Failed to open directory `%s'", dirname);
 575                 return (NULL);
 576         }
 577
 578         while ((ent = readdir (dh)) != NULL)
 579         {
 580                 if (!isdigit (ent->d_name[0]))
 581                         continue;
 582
 583                 if ((list_len + 1) >= list_size)
 584                 {
 585                         int *new_ptr;
 586                         int  new_size = 2 * list_size;
 587                         /* Comes in sizes: 2, 4, 8, 16, ... */
 588
 589                         new_ptr = (int *) realloc (list, (size_t) (sizeof (int) * new_size));
 590                         if (new_ptr == NULL)
 591                         {
 592                                 if (list != NULL)
 593                                         free (list);
 594                                 ERROR ("processes plugin: "
 595                                                 "Failed to allocate more memory.");
 596                                 return (NULL);
 597                         }
 598
 599                         list = new_ptr;
 600                         list_size = new_size;
 601
 602                         memset (list + list_len, 0, sizeof (int) * (list_size - list_len));
 603                 }
 604
 605                 list[list_len] = atoi (ent->d_name);
 606                 if (list[list_len] != 0)
 607                         list_len++;
 608         }
 609
 610         closedir (dh);
 611
 612         if (list_len == 0)
 613                 return (NULL);
 614
 615         assert (list_len < list_size);
 616         assert (list[list_len] == 0);
 617
 618         return (list);
 619 } /* int *ps_read_tasks */
 620
 621 int ps_read_process (int pid, procstat_t *ps, char *state)
 622 {
 623         char  filename[64];
 624         char  buffer[1024];
 625         FILE *fh;
 626
 627         char *fields[64];
 628         char  fields_len;
 629
 630         int  *tasks;
 631         int   i;
 632
 633         int   ppid;
 634         int   name_len;
 635
 636         long long unsigned cpu_user_counter;
 637         long long unsigned cpu_system_counter;
 638         long long unsigned vmem_rss;
 639
 640         memset (ps, 0, sizeof (procstat_t));
 641
 642         ssnprintf (filename, sizeof (filename), "/proc/%i/stat", pid);
 643
 644         if ((fh = fopen (filename, "r")) == NULL)
 645                 return (-1);
 646
 647         if (fgets (buffer, 1024, fh) == NULL)
 648         {
 649                 fclose (fh);
 650                 return (-1);
 651         }
 652
 653         fclose (fh);
 654
 655         fields_len = strsplit (buffer, fields, 64);
 656         if (fields_len < 24)
 657         {
 658                 DEBUG ("processes plugin: ps_read_process (pid = %i):"
 659                                 " `%s' has only %i fields..",
 660                                 (int) pid, filename, fields_len);
 661                 return (-1);
 662         }
 663
 664         /* copy the name, strip brackets in the process */
 665         name_len = strlen (fields[1]) - 2;
 666         if ((fields[1][0] != '(') || (fields[1][name_len + 1] != ')'))
 667         {
 668                 DEBUG ("No brackets found in process name: `%s'", fields[1]);
 669                 return (-1);
 670         }
 671         fields[1] = fields[1] + 1;
 672         fields[1][name_len] = '\0';
 673         strncpy (ps->name, fields[1], PROCSTAT_NAME_LEN);
 674
 675         ppid = atoi (fields[3]);
 676
 677         *state = fields[2][0];
 678
 679         if (*state == 'Z')
 680         {
 681                 ps->num_lwp  = 0;
 682                 ps->num_proc = 0;
 683         }
 684         else if ((tasks = ps_read_tasks (pid)) == NULL)
 685         {
 686                 /* Kernel 2.4 or so */
 687                 ps->num_lwp  = 1;
 688                 ps->num_proc = 1;
 689         }
 690         else
 691         {
 692                 ps->num_lwp  = 0;
 693                 ps->num_proc = 1;
 694                 for (i = 0; tasks[i] != 0; i++)
 695                         ps->num_lwp++;
 696
 697                 free (tasks);
 698                 tasks = NULL;
 699         }
 700
 701         /* Leave the rest at zero if this is only a zombi */
 702         if (ps->num_proc == 0)
 703         {
 704                 DEBUG ("processes plugin: This is only a zombi: pid = %i; "
 705                                 "name = %s;", pid, ps->name);
 706                 return (0);
 707         }
 708
 709         cpu_user_counter   = atoll (fields[13]);
 710         cpu_system_counter = atoll (fields[14]);
 711         vmem_rss = atoll (fields[23]);
 712         ps->vmem_minflt_counter = atol (fields[9]);
 713         ps->vmem_majflt_counter = atol (fields[11]);
 714
 715         /* Convert jiffies to useconds */
 716         cpu_user_counter   = cpu_user_counter   * 1000000 / CONFIG_HZ;
 717         cpu_system_counter = cpu_system_counter * 1000000 / CONFIG_HZ;
 718         vmem_rss = vmem_rss * pagesize_g;
 719
 720         ps->cpu_user_counter = (unsigned long) cpu_user_counter;
 721         ps->cpu_system_counter = (unsigned long) cpu_system_counter;
 722         ps->vmem_rss = (unsigned long) vmem_rss;
 723
 724         /* success */
 725         return (0);
 726 } /* int ps_read_process (...) */
 727 #endif /* KERNEL_LINUX */
 728
 729 #if HAVE_THREAD_INFO
 730 static int mach_get_task_name (task_t t, int *pid, char *name, size_t name_max_len)
 731 {
 732         int mib[4];
 733
 734         struct kinfo_proc kp;
 735         size_t            kp_size;
 736
 737         mib[0] = CTL_KERN;
 738         mib[1] = KERN_PROC;
 739         mib[2] = KERN_PROC_PID;
 740
 741         if (pid_for_task (t, pid) != KERN_SUCCESS)
 742                 return (-1);
 743         mib[3] = *pid;
 744
 745         kp_size = sizeof (kp);
 746         if (sysctl (mib, 4, &kp, &kp_size, NULL, 0) != 0)
 747                 return (-1);
 748
 749         if (name_max_len > (MAXCOMLEN + 1))
 750                 name_max_len = MAXCOMLEN + 1;
 751
 752         strncpy (name, kp.kp_proc.p_comm, name_max_len - 1);
 753         name[name_max_len - 1] = '\0';
 754
 755         DEBUG ("pid = %i; name = %s;", *pid, name);
 756
 757         /* We don't do the special handling for `p_comm == "LaunchCFMApp"' as
 758          * `top' does it, because it is a lot of work and only used when
 759          * debugging. -octo */
 760
 761         return (0);
 762 }
 763 #endif /* HAVE_THREAD_INFO */
 764 /* ------- end of additional functions for KERNEL_LINUX/HAVE_THREAD_INFO ------- */
 765
 766 /* do actual readings from kernel */
 767 static int ps_read (void)
 768 {
 769 #if HAVE_THREAD_INFO
 770         kern_return_t            status;
 771
 772         int                      pset;
 773         processor_set_t          port_pset_priv;
 774
 775         int                      task;
 776         task_array_t             task_list;
 777         mach_msg_type_number_t   task_list_len;
 778
 779         int                      task_pid;
 780         char                     task_name[MAXCOMLEN + 1];
 781
 782         int                      thread;
 783         thread_act_array_t       thread_list;
 784         mach_msg_type_number_t   thread_list_len;
 785         thread_basic_info_data_t thread_data;
 786         mach_msg_type_number_t   thread_data_len;
 787
 788         int running  = 0;
 789         int sleeping = 0;
 790         int zombies  = 0;
 791         int stopped  = 0;
 792         int blocked  = 0;
 793
 794         procstat_t *ps;
 795         procstat_entry_t pse;
 796
 797         ps_list_reset ();
 798
 799         /*
 800          * The Mach-concept is a little different from the traditional UNIX
 801          * concept: All the work is done in threads. Threads are contained in
 802          * `tasks'. Therefore, `task status' doesn't make much sense, since
 803          * it's actually a `thread status'.
 804          * Tasks are assigned to sets of processors, so that's where you go to
 805          * get a list.
 806          */
 807         for (pset = 0; pset < pset_list_len; pset++)
 808         {
 809                 if ((status = host_processor_set_priv (port_host_self,
 810                                                 pset_list[pset],
 811                                                 &port_pset_priv)) != KERN_SUCCESS)
 812                 {
 813                         ERROR ("host_processor_set_priv failed: %s\n",
 814                                         mach_error_string (status));
 815                         continue;
 816                 }
 817
 818                 if ((status = processor_set_tasks (port_pset_priv,
 819                                                 &task_list,
 820                                                 &task_list_len)) != KERN_SUCCESS)
 821                 {
 822                         ERROR ("processor_set_tasks failed: %s\n",
 823                                         mach_error_string (status));
 824                         mach_port_deallocate (port_task_self, port_pset_priv);
 825                         continue;
 826                 }
 827
 828                 for (task = 0; task < task_list_len; task++)
 829                 {
 830                         ps = NULL;
 831                         if (mach_get_task_name (task_list[task],
 832                                                 &task_pid,
 833                                                 task_name, PROCSTAT_NAME_LEN) == 0)
 834                         {
 835                                 /* search for at least one match */
 836                                 for (ps = list_head_g; ps != NULL; ps = ps->next)
 837                                         if (ps_list_match(task_name, NULL, ps) == 1) //!!! cmdline should be here instead of NULL
 838                                                 break;
 839                         }
 840
 841                         /* Collect more detailed statistics for this process */
 842                         if (ps != NULL)
 843                         {
 844                                 task_basic_info_data_t        task_basic_info;
 845                                 mach_msg_type_number_t        task_basic_info_len;
 846                                 task_events_info_data_t       task_events_info;
 847                                 mach_msg_type_number_t        task_events_info_len;
 848                                 task_absolutetime_info_data_t task_absolutetime_info;
 849                                 mach_msg_type_number_t        task_absolutetime_info_len;
 850
 851                                 memset (&pse, '\0', sizeof (pse));
 852                                 pse.id = task_pid;
 853
 854                                 task_basic_info_len = TASK_BASIC_INFO_COUNT;
 855                                 status = task_info (task_list[task],
 856                                                 TASK_BASIC_INFO,
 857                                                 (task_info_t) &task_basic_info,
 858                                                 &task_basic_info_len);
 859                                 if (status != KERN_SUCCESS)
 860                                 {
 861                                         ERROR ("task_info failed: %s",
 862                                                         mach_error_string (status));
 863                                         continue; /* with next thread_list */
 864                                 }
 865
 866                                 task_events_info_len = TASK_EVENTS_INFO_COUNT;
 867                                 status = task_info (task_list[task],
 868                                                 TASK_EVENTS_INFO,
 869                                                 (task_info_t) &task_events_info,
 870                                                 &task_events_info_len);
 871                                 if (status != KERN_SUCCESS)
 872                                 {
 873                                         ERROR ("task_info failed: %s",
 874                                                         mach_error_string (status));
 875                                         continue; /* with next thread_list */
 876                                 }
 877
 878                                 task_absolutetime_info_len = TASK_ABSOLUTETIME_INFO_COUNT;
 879                                 status = task_info (task_list[task],
 880                                                 TASK_ABSOLUTETIME_INFO,
 881                                                 (task_info_t) &task_absolutetime_info,
 882                                                 &task_absolutetime_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                                 pse.num_proc++;
 891                                 pse.vmem_rss = task_basic_info.resident_size;
 892
 893                                 pse.vmem_minflt_counter = task_events_info.cow_faults;
 894                                 pse.vmem_majflt_counter = task_events_info.faults;
 895
 896                                 pse.cpu_user_counter = task_absolutetime_info.total_user;
 897                                 pse.cpu_system_counter = task_absolutetime_info.total_system;
 898                         }
 899
 900                         status = task_threads (task_list[task], &thread_list,
 901                                         &thread_list_len);
 902                         if (status != KERN_SUCCESS)
 903                         {
 904                                 /* Apple's `top' treats this case a zombie. It
 905                                  * makes sense to some extend: A `zombie'
 906                                  * thread is nonsense, since the task/process
 907                                  * is dead. */
 908                                 zombies++;
 909                                 DEBUG ("task_threads failed: %s",
 910                                                 mach_error_string (status));
 911                                 if (task_list[task] != port_task_self)
 912                                         mach_port_deallocate (port_task_self,
 913                                                         task_list[task]);
 914                                 continue; /* with next task_list */
 915                         }
 916
 917                         for (thread = 0; thread < thread_list_len; thread++)
 918                         {
 919                                 thread_data_len = THREAD_BASIC_INFO_COUNT;
 920                                 status = thread_info (thread_list[thread],
 921                                                 THREAD_BASIC_INFO,
 922                                                 (thread_info_t) &thread_data,
 923                                                 &thread_data_len);
 924                                 if (status != KERN_SUCCESS)
 925                                 {
 926                                         ERROR ("thread_info failed: %s",
 927                                                         mach_error_string (status));
 928                                         if (task_list[task] != port_task_self)
 929                                                 mach_port_deallocate (port_task_self,
 930                                                                 thread_list[thread]);
 931                                         continue; /* with next thread_list */
 932                                 }
 933
 934                                 if (ps != NULL)
 935                                         pse.num_lwp++;
 936
 937                                 switch (thread_data.run_state)
 938                                 {
 939                                         case TH_STATE_RUNNING:
 940                                                 running++;
 941                                                 break;
 942                                         case TH_STATE_STOPPED:
 943                                         /* What exactly is `halted'? */
 944                                         case TH_STATE_HALTED:
 945                                                 stopped++;
 946                                                 break;
 947                                         case TH_STATE_WAITING:
 948                                                 sleeping++;
 949                                                 break;
 950                                         case TH_STATE_UNINTERRUPTIBLE:
 951                                                 blocked++;
 952                                                 break;
 953                                         /* There is no `zombie' case here,
 954                                          * since there are no zombie-threads.
 955                                          * There's only zombie tasks, which are
 956                                          * handled above. */
 957                                         default:
 958                                                 WARNING ("Unknown thread status: %s",
 959                                                                 thread_data.run_state);
 960                                                 break;
 961                                 } /* switch (thread_data.run_state) */
 962
 963                                 if (task_list[task] != port_task_self)
 964                                 {
 965                                         status = mach_port_deallocate (port_task_self,
 966                                                         thread_list[thread]);
 967                                         if (status != KERN_SUCCESS)
 968                                                 ERROR ("mach_port_deallocate failed: %s",
 969                                                                 mach_error_string (status));
 970                                 }
 971                         } /* for (thread_list) */
 972
 973                         if ((status = vm_deallocate (port_task_self,
 974                                                         (vm_address_t) thread_list,
 975                                                         thread_list_len * sizeof (thread_act_t)))
 976                                         != KERN_SUCCESS)
 977                         {
 978                                 ERROR ("vm_deallocate failed: %s",
 979                                                 mach_error_string (status));
 980                         }
 981                         thread_list = NULL;
 982                         thread_list_len = 0;
 983
 984                         /* Only deallocate the task port, if it isn't our own.
 985                          * Don't know what would happen in that case, but this
 986                          * is what Apple's top does.. ;) */
 987                         if (task_list[task] != port_task_self)
 988                         {
 989                                 status = mach_port_deallocate (port_task_self,
 990                                                 task_list[task]);
 991                                 if (status != KERN_SUCCESS)
 992                                         ERROR ("mach_port_deallocate failed: %s",
 993                                                         mach_error_string (status));
 994                         }
 995
 996                         if (ps != NULL)
 997                                 ps_list_add (task_name, NULL, &pse); //!!! cmdline should be here instead of NULL
 998                 } /* for (task_list) */
 999
1000                 if ((status = vm_deallocate (port_task_self,
1001                                 (vm_address_t) task_list,
1002                                 task_list_len * sizeof (task_t))) != KERN_SUCCESS)
1003                 {
1004                         ERROR ("vm_deallocate failed: %s",
1005                                         mach_error_string (status));
1006                 }
1007                 task_list = NULL;
1008                 task_list_len = 0;
1009
1010                 if ((status = mach_port_deallocate (port_task_self, port_pset_priv))
1011                                 != KERN_SUCCESS)
1012                 {
1013                         ERROR ("mach_port_deallocate failed: %s",
1014                                         mach_error_string (status));
1015                 }
1016         } /* for (pset_list) */
1017
1018         ps_submit_state ("running", running);
1019         ps_submit_state ("sleeping", sleeping);
1020         ps_submit_state ("zombies", zombies);
1021         ps_submit_state ("stopped", stopped);
1022         ps_submit_state ("blocked", blocked);
1023
1024         for (ps = list_head_g; ps != NULL; ps = ps->next)
1025                 ps_submit_proc_list (ps);
1026 /* #endif HAVE_THREAD_INFO */
1027
1028 #elif KERNEL_LINUX
1029         int running  = 0;
1030         int sleeping = 0;
1031         int zombies  = 0;
1032         int stopped  = 0;
1033         int paging   = 0;
1034         int blocked  = 0;
1035
1036         struct dirent *ent;
1037         DIR           *proc;
1038         int            pid;
1039
1040         int        status;
1041         procstat_t ps;
1042         procstat_entry_t pse;
1043         char       state;
1044
1045         procstat_t *ps_ptr;
1046
1047         running = sleeping = zombies = stopped = paging = blocked = 0;
1048         ps_list_reset ();
1049
1050         if ((proc = opendir ("/proc")) == NULL)
1051         {
1052                 char errbuf[1024];
1053                 ERROR ("Cannot open `/proc': %s",
1054                                 sstrerror (errno, errbuf, sizeof (errbuf)));
1055                 return (-1);
1056         }
1057
1058         while ((ent = readdir (proc)) != NULL)
1059         {
1060                 if (!isdigit (ent->d_name[0]))
1061                         continue;
1062
1063                 if ((pid = atoi (ent->d_name)) < 1)
1064                         continue;
1065
1066                 status = ps_read_process (pid, &ps, &state);
1067                 if (status != 0)
1068                 {
1069                         DEBUG ("ps_read_process failed: %i", status);
1070                         continue;
1071                 }
1072
1073                 pse.id       = pid;
1074                 pse.age      = 0;
1075
1076                 pse.num_proc = ps.num_proc;
1077                 pse.num_lwp  = ps.num_lwp;
1078                 pse.vmem_rss = ps.vmem_rss;
1079
1080                 pse.vmem_minflt = 0;
1081                 pse.vmem_minflt_counter = ps.vmem_minflt_counter;
1082                 pse.vmem_majflt = 0;
1083                 pse.vmem_majflt_counter = ps.vmem_majflt_counter;
1084
1085                 pse.cpu_user = 0;
1086                 pse.cpu_user_counter = ps.cpu_user_counter;
1087                 pse.cpu_system = 0;
1088                 pse.cpu_system_counter = ps.cpu_system_counter;
1089
1090                 switch (state)
1091                 {
1092                         case 'R': running++;  break;
1093                         case 'S': sleeping++; break;
1094                         case 'D': blocked++;  break;
1095                         case 'Z': zombies++;  break;
1096                         case 'T': stopped++;  break;
1097                         case 'W': paging++;   break;
1098                 }
1099
1100                 ps_list_add (ps.name, NULL, &pse); //!!! cmdline should be here instead of NULL
1101         }
1102
1103         closedir (proc);
1104
1105         ps_submit_state ("running",  running);
1106         ps_submit_state ("sleeping", sleeping);
1107         ps_submit_state ("zombies",  zombies);
1108         ps_submit_state ("stopped",  stopped);
1109         ps_submit_state ("paging",   paging);
1110         ps_submit_state ("blocked",  blocked);
1111
1112         for (ps_ptr = list_head_g; ps_ptr != NULL; ps_ptr = ps_ptr->next)
1113                 ps_submit_proc_list (ps_ptr);
1114 /* #endif KERNEL_LINUX */
1115
1116 #elif HAVE_LIBKVM
1117         int running  = 0;
1118         int sleeping = 0;
1119         int zombies  = 0;
1120         int stopped  = 0;
1121         int blocked  = 0;
1122         int idle     = 0;
1123         int wait     = 0;
1124
1125         kvm_t *kd;
1126         char errbuf[1024];
1127         char cmdline[ARG_MAX];
1128         struct kinfo_proc *procs;          /* array of processes */
1129         char ** argv;
1130         int count;                         /* returns number of processes */
1131         int i, j;
1132
1133         procstat_t *ps_ptr;
1134         procstat_entry_t pse;
1135
1136         ps_list_reset ();
1137
1138         /* Open the kvm interface, get a descriptor */
1139         if ((kd = kvm_open(NULL, NULL, NULL, 0, errbuf)) == NULL) {
1140                 ERROR ("Cannot open kvm interface: %s", errbuf);
1141                 return (0);
1142         }
1143
1144         /* Get the list of processes. */
1145         if ((procs = kvm_getprocs(kd, KERN_PROC_ALL, 0, &count)) == NULL) {
1146                 kvm_close(kd);
1147                 ERROR ("Cannot get kvm processes list: %s", kvm_geterr(kd));
1148                 return (0);
1149         }
1150
1151         /* Iterate through the processes in kinfo_proc */
1152         for (i=0; i < count; i++) {
1153                 // retrieve the arguments
1154                 *cmdline = '\0';
1155                 argv = kvm_getargv(kd, (const struct kinfo_proc *) &(procs[i]), 0);
1156                 if (argv) {
1157                         j = 0;
1158                         while (argv[j] && strlen(cmdline) <= ARG_MAX) {
1159                                 if (j)
1160                                         strncat(cmdline, " ", 1);
1161                                 strncat(cmdline, argv[j], strlen(argv[j]));
1162                                 j++;
1163                         }
1164                 }
1165
1166                 pse.id       = procs[i].ki_pid;
1167                 pse.age      = 0;
1168
1169                 pse.num_proc = 1;
1170                 pse.num_lwp  = procs[i].ki_numthreads;
1171
1172                 pse.vmem_rss = procs[i].ki_rssize * getpagesize();
1173                 pse.vmem_minflt = 0;
1174                 pse.vmem_minflt_counter = procs[i].ki_rusage.ru_minflt;
1175                 pse.vmem_majflt = 0;
1176                 pse.vmem_majflt_counter = procs[i].ki_rusage.ru_majflt;
1177
1178                 pse.cpu_user = 0;
1179                 pse.cpu_user_counter = procs[i].ki_rusage.ru_utime.tv_sec*1000 + procs[i].ki_rusage.ru_utime.tv_usec;
1180                 pse.cpu_system = 0;
1181                 pse.cpu_system_counter = procs[i].ki_rusage.ru_stime.tv_sec*1000 + procs[i].ki_rusage.ru_stime.tv_usec;
1182
1183                 switch (procs[i].ki_stat) {
1184                         case SSTOP:     stopped++;      break;
1185                         case SSLEEP:    sleeping++;     break;
1186                         case SRUN:      running++;      break;
1187                         case SIDL:      idle++;         break;
1188                         case SWAIT:     wait++;         break;
1189                         case SLOCK:     blocked++;      break;
1190                         case SZOMB:     zombies++;      break;
1191                 }
1192
1193                 ps_list_add (procs[i].ki_comm, cmdline, &pse);
1194         }
1195
1196         if (kd) kvm_close(kd);
1197
1198         ps_submit_state ("running",  running);
1199         ps_submit_state ("sleeping", sleeping);
1200         ps_submit_state ("zombies",  zombies);
1201         ps_submit_state ("stopped",  stopped);
1202         ps_submit_state ("blocked",  blocked);
1203         ps_submit_state ("idle",     idle);
1204         ps_submit_state ("wait",     wait);
1205
1206         for (ps_ptr = list_head_g; ps_ptr != NULL; ps_ptr = ps_ptr->next)
1207                 ps_submit_proc_list (ps_ptr);
1208
1209 #endif /* HAVE_LIBKVM */
1210
1211         return (0);
1212 } /* int ps_read */
1213
1214 void module_register (void)
1215 {
1216         plugin_register_config ("processes", ps_config,
1217                         config_keys, config_keys_num);
1218         plugin_register_init ("processes", ps_init);
1219         plugin_register_read ("processes", ps_read);
1220 } /* void module_register */