fs/proc/array.c

   1 /*
   2  *  linux/fs/proc/array.c
   3  *
   4  *  Copyright (C) 1992  by Linus Torvalds
   5  *  based on ideas by Darren Senn
   6  *
   7  * Fixes:
   8  * Michael. K. Johnson: stat,statm extensions.
   9  *                      <johnsonm@stolaf.edu>
  10  *
  11  * Pauline Middelink :  Made cmdline,envline only break at '\0's, to
  12  *                      make sure SET_PROCTITLE works. Also removed
  13  *                      bad '!' which forced address recalculation for
  14  *                      EVERY character on the current page.
  15  *                      <middelin@polyware.iaf.nl>
  16  *
  17  * Danny ter Haar    :  added cpuinfo
  18  *                      <dth@cistron.nl>
  19  *
  20  * Alessandro Rubini :  profile extension.
  21  *                      <rubini@ipvvis.unipv.it>
  22  *
  23  * Jeff Tranter      :  added BogoMips field to cpuinfo
  24  *                      <Jeff_Tranter@Mitel.COM>
  25  *
  26  * Bruno Haible      :  remove 4K limit for the maps file
  27  *                      <haible@ma2s2.mathematik.uni-karlsruhe.de>
  28  *
  29  * Yves Arrouye      :  remove removal of trailing spaces in get_array.
  30  *                      <Yves.Arrouye@marin.fdn.fr>
  31  *
  32  * Jerome Forissier  :  added per-CPU time information to /proc/stat
  33  *                      and /proc/<pid>/cpu extension
  34  *                      <forissier@isia.cma.fr>
  35  *                      - Incorporation and non-SMP safe operation
  36  *                      of forissier patch in 2.1.78 by
  37  *                      Hans Marcus <crowbar@concepts.nl>
  38  *
  39  * aeb@cwi.nl        :  /proc/partitions
  40  *
  41  *
  42  * Alan Cox          :  security fixes.
  43  *                      <Alan.Cox@linux.org>
  44  *
  45  * Al Viro           :  safe handling of mm_struct
  46  *
  47  * Gerhard Wichert   :  added BIGMEM support
  48  * Siemens AG           <Gerhard.Wichert@pdb.siemens.de>
  49  *
  50  * Al Viro & Jeff Garzik :  moved most of the thing into base.c and
  51  *                       :  proc_misc.c. The rest may eventually go into
  52  *                       :  base.c too.
  53  */
  54
  55 #include <linux/config.h>
  56 #include <linux/types.h>
  57 #include <linux/errno.h>
  58 #include <linux/time.h>
  59 #include <linux/kernel.h>
  60 #include <linux/kernel_stat.h>
  61 #include <linux/tty.h>
  62 #include <linux/string.h>
  63 #include <linux/mman.h>
  64 #include <linux/proc_fs.h>
  65 #include <linux/ioport.h>
  66 #include <linux/mm.h>
  67 #include <linux/hugetlb.h>
  68 #include <linux/pagemap.h>
  69 #include <linux/swap.h>
  70 #include <linux/slab.h>
  71 #include <linux/smp.h>
  72 #include <linux/signal.h>
  73 #include <linux/highmem.h>
  74 #include <linux/file.h>
  75 #include <linux/times.h>
  76 #include <linux/vs_base.h>
  77 #include <linux/vs_context.h>
  78 #include <linux/vs_network.h>
  79 #include <linux/vs_cvirt.h>
  80
  81 #include <asm/uaccess.h>
  82 #include <asm/pgtable.h>
  83 #include <asm/io.h>
  84 #include <asm/processor.h>
  85
  86 /* Gcc optimizes away "strlen(x)" for constant x */
  87 #define ADDBUF(buffer, string) \
  88 do { memcpy(buffer, string, strlen(string)); \
  89      buffer += strlen(string); } while (0)
  90
  91 static inline char * task_name(struct task_struct *p, char * buf)
  92 {
  93         int i;
  94         char * name;
  95         char tcomm[sizeof(p->comm)];
  96
  97         get_task_comm(tcomm, p);
  98
  99         ADDBUF(buf, "Name:\t");
 100         name = tcomm;
 101         i = sizeof(tcomm);
 102         do {
 103                 unsigned char c = *name;
 104                 name++;
 105                 i--;
 106                 *buf = c;
 107                 if (!c)
 108                         break;
 109                 if (c == '\\') {
 110                         buf[1] = c;
 111                         buf += 2;
 112                         continue;
 113                 }
 114                 if (c == '\n') {
 115                         buf[0] = '\\';
 116                         buf[1] = 'n';
 117                         buf += 2;
 118                         continue;
 119                 }
 120                 buf++;
 121         } while (i);
 122         *buf = '\n';
 123         return buf+1;
 124 }
 125
 126 /*
 127  * The task state array is a strange "bitmap" of
 128  * reasons to sleep. Thus "running" is zero, and
 129  * you can test for combinations of others with
 130  * simple bit tests.
 131  */
 132 static const char *task_state_array[] = {
 133         "R (running)",          /*  0 */
 134         "S (sleeping)",         /*  1 */
 135         "D (disk sleep)",       /*  2 */
 136         "T (stopped)",          /*  4 */
 137         "T (tracing stop)",     /*  8 */
 138         "Z (zombie)",           /* 16 */
 139         "X (dead)",             /* 32 */
 140         "H (on hold)"           /* 64 */
 141 };
 142
 143 static inline const char * get_task_state(struct task_struct *tsk)
 144 {
 145         unsigned int state = (tsk->state & (TASK_RUNNING |
 146                                             TASK_INTERRUPTIBLE |
 147                                             TASK_UNINTERRUPTIBLE |
 148                                             TASK_STOPPED |
 149                                             TASK_TRACED |
 150                                             TASK_ONHOLD)) |
 151                         (tsk->exit_state & (EXIT_ZOMBIE |
 152                                             EXIT_DEAD));
 153         const char **p = &task_state_array[0];
 154
 155         while (state) {
 156                 p++;
 157                 state >>= 1;
 158         }
 159         return *p;
 160 }
 161
 162 static inline char * task_state(struct task_struct *p, char *buffer)
 163 {
 164         struct group_info *group_info;
 165         int g;
 166         pid_t pid, ppid, tppid, tgid;
 167
 168         read_lock(&tasklist_lock);
 169         tgid = vx_map_tgid(p->tgid);
 170         pid = vx_map_pid(p->pid);
 171         ppid = vx_map_pid(p->real_parent->pid);
 172         tppid = vx_map_pid(p->parent->pid);
 173         buffer += sprintf(buffer,
 174                 "State:\t%s\n"
 175                 "SleepAVG:\t%lu%%\n"
 176                 "Tgid:\t%d\n"
 177                 "Pid:\t%d\n"
 178                 "PPid:\t%d\n"
 179                 "TracerPid:\t%d\n"
 180                 "Uid:\t%d\t%d\t%d\t%d\n"
 181                 "Gid:\t%d\t%d\t%d\t%d\n",
 182                 get_task_state(p),
 183                 (p->sleep_avg/1024)*100/(1020000000/1024),
 184                 tgid, pid, (pid > 1) ? ppid : 0,
 185                 p->pid && p->ptrace ? tppid : 0,
 186                 p->uid, p->euid, p->suid, p->fsuid,
 187                 p->gid, p->egid, p->sgid, p->fsgid);
 188         read_unlock(&tasklist_lock);
 189         task_lock(p);
 190         buffer += sprintf(buffer,
 191                 "FDSize:\t%d\n"
 192                 "Groups:\t",
 193                 p->files ? p->files->max_fds : 0);
 194
 195         group_info = p->group_info;
 196         get_group_info(group_info);
 197         task_unlock(p);
 198
 199         for (g = 0; g < min(group_info->ngroups,NGROUPS_SMALL); g++)
 200                 buffer += sprintf(buffer, "%d ", GROUP_AT(group_info,g));
 201         put_group_info(group_info);
 202
 203         buffer += sprintf(buffer, "\n");
 204         return buffer;
 205 }
 206
 207 static char * render_sigset_t(const char *header, sigset_t *set, char *buffer)
 208 {
 209         int i, len;
 210
 211         len = strlen(header);
 212         memcpy(buffer, header, len);
 213         buffer += len;
 214
 215         i = _NSIG;
 216         do {
 217                 int x = 0;
 218
 219                 i -= 4;
 220                 if (sigismember(set, i+1)) x |= 1;
 221                 if (sigismember(set, i+2)) x |= 2;
 222                 if (sigismember(set, i+3)) x |= 4;
 223                 if (sigismember(set, i+4)) x |= 8;
 224                 *buffer++ = (x < 10 ? '0' : 'a' - 10) + x;
 225         } while (i >= 4);
 226
 227         *buffer++ = '\n';
 228         *buffer = 0;
 229         return buffer;
 230 }
 231
 232 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
 233                                     sigset_t *catch)
 234 {
 235         struct k_sigaction *k;
 236         int i;
 237
 238         k = p->sighand->action;
 239         for (i = 1; i <= _NSIG; ++i, ++k) {
 240                 if (k->sa.sa_handler == SIG_IGN)
 241                         sigaddset(ign, i);
 242                 else if (k->sa.sa_handler != SIG_DFL)
 243                         sigaddset(catch, i);
 244         }
 245 }
 246
 247 static inline char * task_sig(struct task_struct *p, char *buffer)
 248 {
 249         sigset_t pending, shpending, blocked, ignored, caught;
 250         int num_threads = 0;
 251
 252         sigemptyset(&pending);
 253         sigemptyset(&shpending);
 254         sigemptyset(&blocked);
 255         sigemptyset(&ignored);
 256         sigemptyset(&caught);
 257
 258         /* Gather all the data with the appropriate locks held */
 259         read_lock(&tasklist_lock);
 260         if (p->sighand) {
 261                 spin_lock_irq(&p->sighand->siglock);
 262                 pending = p->pending.signal;
 263                 shpending = p->signal->shared_pending.signal;
 264                 blocked = p->blocked;
 265                 collect_sigign_sigcatch(p, &ignored, &caught);
 266                 num_threads = atomic_read(&p->signal->count);
 267                 spin_unlock_irq(&p->sighand->siglock);
 268         }
 269         read_unlock(&tasklist_lock);
 270
 271         buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
 272
 273         /* render them all */
 274         buffer = render_sigset_t("SigPnd:\t", &pending, buffer);
 275         buffer = render_sigset_t("ShdPnd:\t", &shpending, buffer);
 276         buffer = render_sigset_t("SigBlk:\t", &blocked, buffer);
 277         buffer = render_sigset_t("SigIgn:\t", &ignored, buffer);
 278         buffer = render_sigset_t("SigCgt:\t", &caught, buffer);
 279
 280         return buffer;
 281 }
 282
 283 static inline char *task_cap(struct task_struct *p, char *buffer)
 284 {
 285     return buffer + sprintf(buffer, "CapInh:\t%016x\n"
 286                             "CapPrm:\t%016x\n"
 287                             "CapEff:\t%016x\n",
 288                             cap_t(p->cap_inheritable),
 289                             cap_t(p->cap_permitted),
 290                             cap_t(p->cap_effective));
 291 }
 292
 293 int proc_pid_status(struct task_struct *task, char * buffer)
 294 {
 295         char * orig = buffer;
 296 #ifdef  CONFIG_VSERVER_LEGACY
 297         struct vx_info *vxi;
 298         struct nx_info *nxi;
 299 #endif
 300         struct mm_struct *mm = get_task_mm(task);
 301
 302         buffer = task_name(task, buffer);
 303         buffer = task_state(task, buffer);
 304
 305         if (mm) {
 306                 buffer = task_mem(mm, buffer);
 307                 mmput(mm);
 308         }
 309         buffer = task_sig(task, buffer);
 310         buffer = task_cap(task, buffer);
 311
 312 #ifdef  CONFIG_VSERVER_LEGACY
 313         buffer += sprintf (buffer,"s_context: %d\n", vx_task_xid(task));
 314         vxi = task_get_vx_info(task);
 315         if (vxi) {
 316                 buffer += sprintf (buffer,"ctxflags: %08llx\n"
 317                         ,(unsigned long long)vxi->vx_flags);
 318                 buffer += sprintf (buffer,"initpid: %d\n"
 319                         ,vxi->vx_initpid);
 320         } else {
 321                 buffer += sprintf (buffer,"ctxflags: none\n");
 322                 buffer += sprintf (buffer,"initpid: none\n");
 323         }
 324         put_vx_info(vxi);
 325         nxi = task_get_nx_info(task);
 326         if (nxi) {
 327                 int i;
 328
 329                 buffer += sprintf (buffer,"ipv4root:");
 330                 for (i=0; i<nxi->nbipv4; i++){
 331                         buffer += sprintf (buffer," %08x/%08x"
 332                                 ,nxi->ipv4[i]
 333                                 ,nxi->mask[i]);
 334                 }
 335                 *buffer++ = '\n';
 336                 buffer += sprintf (buffer,"ipv4root_bcast: %08x\n"
 337                         ,nxi->v4_bcast);
 338         } else {
 339                 buffer += sprintf (buffer,"ipv4root: 0\n");
 340                 buffer += sprintf (buffer,"ipv4root_bcast: 0\n");
 341         }
 342         put_nx_info(nxi);
 343 #endif
 344 #if defined(CONFIG_ARCH_S390)
 345         buffer = task_show_regs(task, buffer);
 346 #endif
 347         return buffer - orig;
 348 }
 349
 350 static int do_task_stat(struct task_struct *task, char * buffer, int whole)
 351 {
 352         unsigned long vsize, eip, esp, wchan = ~0UL;
 353         long priority, nice;
 354         unsigned long long bias_uptime = 0;
 355         int tty_pgrp = -1, tty_nr = 0;
 356         sigset_t sigign, sigcatch;
 357         char state;
 358         int res;
 359         pid_t pid, ppid, pgid = -1, sid = -1;
 360         int num_threads = 0;
 361         struct mm_struct *mm;
 362         unsigned long long start_time;
 363         unsigned long cmin_flt = 0, cmaj_flt = 0, cutime = 0, cstime = 0;
 364         unsigned long  min_flt = 0,  maj_flt = 0,  utime = 0,  stime = 0;
 365         unsigned long rsslim = 0;
 366         struct task_struct *t;
 367         char tcomm[sizeof(task->comm)];
 368
 369         state = *get_task_state(task);
 370         vsize = eip = esp = 0;
 371         mm = get_task_mm(task);
 372         if (mm) {
 373                 vsize = task_vsize(mm);
 374                 eip = KSTK_EIP(task);
 375                 esp = KSTK_ESP(task);
 376         }
 377
 378         get_task_comm(tcomm, task);
 379
 380         sigemptyset(&sigign);
 381         sigemptyset(&sigcatch);
 382         read_lock(&tasklist_lock);
 383         if (task->sighand) {
 384                 spin_lock_irq(&task->sighand->siglock);
 385                 num_threads = atomic_read(&task->signal->count);
 386                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
 387
 388                 /* add up live thread stats at the group level */
 389                 if (whole) {
 390                         t = task;
 391                         do {
 392                                 min_flt += t->min_flt;
 393                                 maj_flt += t->maj_flt;
 394                                 utime += t->utime;
 395                                 stime += t->stime;
 396                                 t = next_thread(t);
 397                         } while (t != task);
 398                 }
 399
 400                 spin_unlock_irq(&task->sighand->siglock);
 401         }
 402         if (task->signal) {
 403                 if (task->signal->tty) {
 404                         tty_pgrp = task->signal->tty->pgrp;
 405                         tty_nr = new_encode_dev(tty_devnum(task->signal->tty));
 406                 }
 407                 pgid = process_group(task);
 408                 sid = task->signal->session;
 409                 cmin_flt = task->signal->cmin_flt;
 410                 cmaj_flt = task->signal->cmaj_flt;
 411                 cutime = task->signal->cutime;
 412                 cstime = task->signal->cstime;
 413                 rsslim = task->signal->rlim[RLIMIT_RSS].rlim_cur;
 414                 if (whole) {
 415                         min_flt += task->signal->min_flt;
 416                         maj_flt += task->signal->maj_flt;
 417                         utime += task->signal->utime;
 418                         stime += task->signal->stime;
 419                 }
 420         }
 421         if (task_vx_flags(task, VXF_VIRT_UPTIME, 0)) {
 422                 bias_uptime = task->vx_info->cvirt.bias_uptime.tv_sec * NSEC_PER_SEC
 423                         + task->vx_info->cvirt.bias_uptime.tv_nsec;
 424         }
 425         read_unlock(&tasklist_lock);
 426
 427         if (!whole || num_threads<2) {
 428                 wchan = 0;
 429                 if (current->uid == task->uid || current->euid == task->uid ||
 430                                                 capable(CAP_SYS_NICE))
 431                         wchan = get_wchan(task);
 432         }
 433         if (!whole) {
 434                 min_flt = task->min_flt;
 435                 maj_flt = task->maj_flt;
 436                 utime = task->utime;
 437                 stime = task->stime;
 438         }
 439
 440         /* scale priority and nice values from timeslices to -20..20 */
 441         /* to make it look like a "normal" Unix priority/nice value  */
 442         priority = task_prio(task);
 443         nice = task_nice(task);
 444
 445         read_lock(&tasklist_lock);
 446         pid = vx_info_map_pid(task->vx_info, task->pid);
 447         ppid = (!(pid > 1)) ? 0 :
 448                 vx_info_map_pid(task->vx_info, task->real_parent->pid);
 449         pgid = vx_info_map_pid(task->vx_info, pgid);
 450         read_unlock(&tasklist_lock);
 451
 452         /* Temporary variable needed for gcc-2.96 */
 453         /* convert timespec -> nsec*/
 454         start_time = (unsigned long long)task->start_time.tv_sec * NSEC_PER_SEC
 455                                 + task->start_time.tv_nsec;
 456         /* convert nsec -> ticks */
 457         start_time = nsec_to_clock_t(start_time - bias_uptime);
 458
 459         res = sprintf(buffer,"%d (%s) %c %d %d %d %d %d %lu %lu \
 460 %lu %lu %lu %lu %lu %ld %ld %ld %ld %d %ld %llu %lu %ld %lu %lu %lu %lu %lu \
 461 %lu %lu %lu %lu %lu %lu %lu %lu %d %d %lu %lu\n",
 462                 pid,
 463                 tcomm,
 464                 state,
 465                 ppid,
 466                 pgid,
 467                 sid,
 468                 tty_nr,
 469                 tty_pgrp,
 470                 task->flags,
 471                 min_flt,
 472                 cmin_flt,
 473                 maj_flt,
 474                 cmaj_flt,
 475                 jiffies_to_clock_t(utime),
 476                 jiffies_to_clock_t(stime),
 477                 jiffies_to_clock_t(cutime),
 478                 jiffies_to_clock_t(cstime),
 479                 priority,
 480                 nice,
 481                 num_threads,
 482                 jiffies_to_clock_t(task->it_real_value),
 483                 start_time,
 484                 vsize,
 485                 mm ? mm->rss : 0, /* you might want to shift this left 3 */
 486                 rsslim,
 487                 mm ? mm->start_code : 0,
 488                 mm ? mm->end_code : 0,
 489                 mm ? mm->start_stack : 0,
 490                 esp,
 491                 eip,
 492                 /* The signal information here is obsolete.
 493                  * It must be decimal for Linux 2.0 compatibility.
 494                  * Use /proc/#/status for real-time signals.
 495                  */
 496                 task->pending.signal.sig[0] & 0x7fffffffUL,
 497                 task->blocked.sig[0] & 0x7fffffffUL,
 498                 sigign      .sig[0] & 0x7fffffffUL,
 499                 sigcatch    .sig[0] & 0x7fffffffUL,
 500                 wchan,
 501                 0UL,
 502                 0UL,
 503                 task->exit_signal,
 504                 task_cpu(task),
 505                 task->rt_priority,
 506                 task->policy);
 507         if(mm)
 508                 mmput(mm);
 509         return res;
 510 }
 511
 512 int proc_tid_stat(struct task_struct *task, char * buffer)
 513 {
 514         return do_task_stat(task, buffer, 0);
 515 }
 516
 517 int proc_tgid_stat(struct task_struct *task, char * buffer)
 518 {
 519         return do_task_stat(task, buffer, 1);
 520 }
 521
 522 int proc_pid_statm(struct task_struct *task, char *buffer)
 523 {
 524         int size = 0, resident = 0, shared = 0, text = 0, lib = 0, data = 0;
 525         struct mm_struct *mm = get_task_mm(task);
 526
 527         if (mm) {
 528                 size = task_statm(mm, &shared, &text, &data, &resident);
 529                 mmput(mm);
 530         }
 531
 532         return sprintf(buffer,"%d %d %d %d %d %d %d\n",
 533                        size, resident, shared, text, lib, data, 0);
 534 }
 535
 536
 537 int proc_pid_delay(struct task_struct *task, char * buffer)
 538 {
 539         int res;
 540
 541         res  = sprintf(buffer,"%u %llu %llu %u %llu %u %llu\n",
 542                        get_delay(task,runs),
 543                        (unsigned long long)get_delay(task,runcpu_total),
 544                        (unsigned long long)get_delay(task,waitcpu_total),
 545                        get_delay(task,num_iowaits),
 546                        (unsigned long long)get_delay(task,iowait_total),
 547                        get_delay(task,num_memwaits),
 548                        (unsigned long long)get_delay(task,mem_iowait_total)
 549                 );
 550         return res;
 551 }
 552