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@lxorguk.ukuu.org.uk>
  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/types.h>
  56 #include <linux/errno.h>
  57 #include <linux/time.h>
  58 #include <linux/kernel.h>
  59 #include <linux/kernel_stat.h>
  60 #include <linux/tty.h>
  61 #include <linux/string.h>
  62 #include <linux/mman.h>
  63 #include <linux/proc_fs.h>
  64 #include <linux/ioport.h>
  65 #include <linux/uaccess.h>
  66 #include <linux/io.h>
  67 #include <linux/mm.h>
  68 #include <linux/hugetlb.h>
  69 #include <linux/pagemap.h>
  70 #include <linux/swap.h>
  71 #include <linux/smp.h>
  72 #include <linux/signal.h>
  73 #include <linux/highmem.h>
  74 #include <linux/file.h>
  75 #include <linux/fdtable.h>
  76 #include <linux/times.h>
  77 #include <linux/cpuset.h>
  78 #include <linux/rcupdate.h>
  79 #include <linux/delayacct.h>
  80 #include <linux/seq_file.h>
  81 #include <linux/pid_namespace.h>
  82 #include <linux/ptrace.h>
  83 #include <linux/tracehook.h>
  84 #include <linux/user_namespace.h>
  85
  86 #include <asm/pgtable.h>
  87 #include <asm/processor.h>
  88 #include "internal.h"
  89
  90 static inline void task_name(struct seq_file *m, struct task_struct *p)
  91 {
  92         int i;
  93         char *buf, *end;
  94         char *name;
  95         char tcomm[sizeof(p->comm)];
  96
  97         get_task_comm(tcomm, p);
  98
  99         seq_puts(m, "Name:\t");
 100         end = m->buf + m->size;
 101         buf = m->buf + m->count;
 102         name = tcomm;
 103         i = sizeof(tcomm);
 104         while (i && (buf < end)) {
 105                 unsigned char c = *name;
 106                 name++;
 107                 i--;
 108                 *buf = c;
 109                 if (!c)
 110                         break;
 111                 if (c == '\\') {
 112                         buf++;
 113                         if (buf < end)
 114                                 *buf++ = c;
 115                         continue;
 116                 }
 117                 if (c == '\n') {
 118                         *buf++ = '\\';
 119                         if (buf < end)
 120                                 *buf++ = 'n';
 121                         continue;
 122                 }
 123                 buf++;
 124         }
 125         m->count = buf - m->buf;
 126         seq_putc(m, '\n');
 127 }
 128
 129 /*
 130  * The task state array is a strange "bitmap" of
 131  * reasons to sleep. Thus "running" is zero, and
 132  * you can test for combinations of others with
 133  * simple bit tests.
 134  */
 135 static const char * const task_state_array[] = {
 136         "R (running)",          /*   0 */
 137         "S (sleeping)",         /*   1 */
 138         "D (disk sleep)",       /*   2 */
 139         "T (stopped)",          /*   4 */
 140         "t (tracing stop)",     /*   8 */
 141         "Z (zombie)",           /*  16 */
 142         "X (dead)",             /*  32 */
 143         "x (dead)",             /*  64 */
 144         "K (wakekill)",         /* 128 */
 145         "W (waking)",           /* 256 */
 146         "P (parked)",           /* 512 */
 147 };
 148
 149 static inline const char *get_task_state(struct task_struct *tsk)
 150 {
 151         unsigned int state = (tsk->state & TASK_REPORT) | tsk->exit_state;
 152         const char * const *p = &task_state_array[0];
 153
 154         BUILD_BUG_ON(1 + ilog2(TASK_STATE_MAX) != ARRAY_SIZE(task_state_array));
 155
 156         while (state) {
 157                 p++;
 158                 state >>= 1;
 159         }
 160         return *p;
 161 }
 162
 163 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
 164                                 struct pid *pid, struct task_struct *p)
 165 {
 166         struct user_namespace *user_ns = seq_user_ns(m);
 167         struct group_info *group_info;
 168         int g;
 169         struct fdtable *fdt = NULL;
 170         const struct cred *cred;
 171         pid_t ppid, tpid;
 172
 173         rcu_read_lock();
 174         ppid = pid_alive(p) ?
 175                 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
 176         tpid = 0;
 177         if (pid_alive(p)) {
 178                 struct task_struct *tracer = ptrace_parent(p);
 179                 if (tracer)
 180                         tpid = task_pid_nr_ns(tracer, ns);
 181         }
 182         cred = get_task_cred(p);
 183         seq_printf(m,
 184                 "State:\t%s\n"
 185                 "Tgid:\t%d\n"
 186                 "Pid:\t%d\n"
 187                 "PPid:\t%d\n"
 188                 "TracerPid:\t%d\n"
 189                 "Uid:\t%d\t%d\t%d\t%d\n"
 190                 "Gid:\t%d\t%d\t%d\t%d\n",
 191                 get_task_state(p),
 192                 task_tgid_nr_ns(p, ns),
 193                 pid_nr_ns(pid, ns),
 194                 ppid, tpid,
 195                 from_kuid_munged(user_ns, cred->uid),
 196                 from_kuid_munged(user_ns, cred->euid),
 197                 from_kuid_munged(user_ns, cred->suid),
 198                 from_kuid_munged(user_ns, cred->fsuid),
 199                 from_kgid_munged(user_ns, cred->gid),
 200                 from_kgid_munged(user_ns, cred->egid),
 201                 from_kgid_munged(user_ns, cred->sgid),
 202                 from_kgid_munged(user_ns, cred->fsgid));
 203
 204         task_lock(p);
 205         if (p->files)
 206                 fdt = files_fdtable(p->files);
 207         seq_printf(m,
 208                 "FDSize:\t%d\n"
 209                 "Groups:\t",
 210                 fdt ? fdt->max_fds : 0);
 211         rcu_read_unlock();
 212
 213         group_info = cred->group_info;
 214         task_unlock(p);
 215
 216         for (g = 0; g < group_info->ngroups; g++)
 217                 seq_printf(m, "%d ",
 218                            from_kgid_munged(user_ns, GROUP_AT(group_info, g)));
 219         put_cred(cred);
 220
 221         seq_putc(m, '\n');
 222 }
 223
 224 void render_sigset_t(struct seq_file *m, const char *header,
 225                                 sigset_t *set)
 226 {
 227         int i;
 228
 229         seq_puts(m, header);
 230
 231         i = _NSIG;
 232         do {
 233                 int x = 0;
 234
 235                 i -= 4;
 236                 if (sigismember(set, i+1)) x |= 1;
 237                 if (sigismember(set, i+2)) x |= 2;
 238                 if (sigismember(set, i+3)) x |= 4;
 239                 if (sigismember(set, i+4)) x |= 8;
 240                 seq_printf(m, "%x", x);
 241         } while (i >= 4);
 242
 243         seq_putc(m, '\n');
 244 }
 245
 246 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
 247                                     sigset_t *catch)
 248 {
 249         struct k_sigaction *k;
 250         int i;
 251
 252         k = p->sighand->action;
 253         for (i = 1; i <= _NSIG; ++i, ++k) {
 254                 if (k->sa.sa_handler == SIG_IGN)
 255                         sigaddset(ign, i);
 256                 else if (k->sa.sa_handler != SIG_DFL)
 257                         sigaddset(catch, i);
 258         }
 259 }
 260
 261 static inline void task_sig(struct seq_file *m, struct task_struct *p)
 262 {
 263         unsigned long flags;
 264         sigset_t pending, shpending, blocked, ignored, caught;
 265         int num_threads = 0;
 266         unsigned long qsize = 0;
 267         unsigned long qlim = 0;
 268
 269         sigemptyset(&pending);
 270         sigemptyset(&shpending);
 271         sigemptyset(&blocked);
 272         sigemptyset(&ignored);
 273         sigemptyset(&caught);
 274
 275         if (lock_task_sighand(p, &flags)) {
 276                 pending = p->pending.signal;
 277                 shpending = p->signal->shared_pending.signal;
 278                 blocked = p->blocked;
 279                 collect_sigign_sigcatch(p, &ignored, &caught);
 280                 num_threads = get_nr_threads(p);
 281                 rcu_read_lock();  /* FIXME: is this correct? */
 282                 qsize = atomic_read(&__task_cred(p)->user->sigpending);
 283                 rcu_read_unlock();
 284                 qlim = task_rlimit(p, RLIMIT_SIGPENDING);
 285                 unlock_task_sighand(p, &flags);
 286         }
 287
 288         seq_printf(m, "Threads:\t%d\n", num_threads);
 289         seq_printf(m, "SigQ:\t%lu/%lu\n", qsize, qlim);
 290
 291         /* render them all */
 292         render_sigset_t(m, "SigPnd:\t", &pending);
 293         render_sigset_t(m, "ShdPnd:\t", &shpending);
 294         render_sigset_t(m, "SigBlk:\t", &blocked);
 295         render_sigset_t(m, "SigIgn:\t", &ignored);
 296         render_sigset_t(m, "SigCgt:\t", &caught);
 297 }
 298
 299 static void render_cap_t(struct seq_file *m, const char *header,
 300                         kernel_cap_t *a)
 301 {
 302         unsigned __capi;
 303
 304         seq_puts(m, header);
 305         CAP_FOR_EACH_U32(__capi) {
 306                 seq_printf(m, "%08x",
 307                            a->cap[(_KERNEL_CAPABILITY_U32S-1) - __capi]);
 308         }
 309         seq_putc(m, '\n');
 310 }
 311
 312 /* Remove non-existent capabilities */
 313 #define NORM_CAPS(v) (v.cap[CAP_TO_INDEX(CAP_LAST_CAP)] &= \
 314                                 CAP_TO_MASK(CAP_LAST_CAP + 1) - 1)
 315
 316 static inline void task_cap(struct seq_file *m, struct task_struct *p)
 317 {
 318         const struct cred *cred;
 319         kernel_cap_t cap_inheritable, cap_permitted, cap_effective, cap_bset;
 320
 321         rcu_read_lock();
 322         cred = __task_cred(p);
 323         cap_inheritable = cred->cap_inheritable;
 324         cap_permitted   = cred->cap_permitted;
 325         cap_effective   = cred->cap_effective;
 326         cap_bset        = cred->cap_bset;
 327         rcu_read_unlock();
 328
 329         NORM_CAPS(cap_inheritable);
 330         NORM_CAPS(cap_permitted);
 331         NORM_CAPS(cap_effective);
 332         NORM_CAPS(cap_bset);
 333
 334         render_cap_t(m, "CapInh:\t", &cap_inheritable);
 335         render_cap_t(m, "CapPrm:\t", &cap_permitted);
 336         render_cap_t(m, "CapEff:\t", &cap_effective);
 337         render_cap_t(m, "CapBnd:\t", &cap_bset);
 338 }
 339
 340 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
 341 {
 342 #ifdef CONFIG_SECCOMP
 343         seq_printf(m, "Seccomp:\t%d\n", p->seccomp.mode);
 344 #endif
 345 }
 346
 347 static inline void task_context_switch_counts(struct seq_file *m,
 348                                                 struct task_struct *p)
 349 {
 350         seq_printf(m,   "voluntary_ctxt_switches:\t%lu\n"
 351                         "nonvoluntary_ctxt_switches:\t%lu\n",
 352                         p->nvcsw,
 353                         p->nivcsw);
 354 }
 355
 356 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
 357 {
 358         seq_puts(m, "Cpus_allowed:\t");
 359         seq_cpumask(m, &task->cpus_allowed);
 360         seq_putc(m, '\n');
 361         seq_puts(m, "Cpus_allowed_list:\t");
 362         seq_cpumask_list(m, &task->cpus_allowed);
 363         seq_putc(m, '\n');
 364 }
 365
 366 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
 367                         struct pid *pid, struct task_struct *task)
 368 {
 369         struct mm_struct *mm = get_task_mm(task);
 370
 371         task_name(m, task);
 372         task_state(m, ns, pid, task);
 373
 374         if (mm) {
 375                 task_mem(m, mm);
 376                 mmput(mm);
 377         }
 378         task_sig(m, task);
 379         task_cap(m, task);
 380         task_seccomp(m, task);
 381         task_cpus_allowed(m, task);
 382         cpuset_task_status_allowed(m, task);
 383         task_context_switch_counts(m, task);
 384         return 0;
 385 }
 386
 387 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 388                         struct pid *pid, struct task_struct *task, int whole)
 389 {
 390         unsigned long vsize, eip, esp, wchan = ~0UL;
 391         int priority, nice;
 392         int tty_pgrp = -1, tty_nr = 0;
 393         sigset_t sigign, sigcatch;
 394         char state;
 395         pid_t ppid = 0, pgid = -1, sid = -1;
 396         int num_threads = 0;
 397         int permitted;
 398         struct mm_struct *mm;
 399         unsigned long long start_time;
 400         unsigned long cmin_flt = 0, cmaj_flt = 0;
 401         unsigned long  min_flt = 0,  maj_flt = 0;
 402         cputime_t cutime, cstime, utime, stime;
 403         cputime_t cgtime, gtime;
 404         unsigned long rsslim = 0;
 405         char tcomm[sizeof(task->comm)];
 406         unsigned long flags;
 407
 408         state = *get_task_state(task);
 409         vsize = eip = esp = 0;
 410         permitted = ptrace_may_access(task, PTRACE_MODE_READ | PTRACE_MODE_NOAUDIT);
 411         mm = get_task_mm(task);
 412         if (mm) {
 413                 vsize = task_vsize(mm);
 414                 if (permitted) {
 415                         eip = KSTK_EIP(task);
 416                         esp = KSTK_ESP(task);
 417                 }
 418         }
 419
 420         get_task_comm(tcomm, task);
 421
 422         sigemptyset(&sigign);
 423         sigemptyset(&sigcatch);
 424         cutime = cstime = utime = stime = 0;
 425         cgtime = gtime = 0;
 426
 427         if (lock_task_sighand(task, &flags)) {
 428                 struct signal_struct *sig = task->signal;
 429
 430                 if (sig->tty) {
 431                         struct pid *pgrp = tty_get_pgrp(sig->tty);
 432                         tty_pgrp = pid_nr_ns(pgrp, ns);
 433                         put_pid(pgrp);
 434                         tty_nr = new_encode_dev(tty_devnum(sig->tty));
 435                 }
 436
 437                 num_threads = get_nr_threads(task);
 438                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
 439
 440                 cmin_flt = sig->cmin_flt;
 441                 cmaj_flt = sig->cmaj_flt;
 442                 cutime = sig->cutime;
 443                 cstime = sig->cstime;
 444                 cgtime = sig->cgtime;
 445                 rsslim = ACCESS_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
 446
 447                 /* add up live thread stats at the group level */
 448                 if (whole) {
 449                         struct task_struct *t = task;
 450                         do {
 451                                 min_flt += t->min_flt;
 452                                 maj_flt += t->maj_flt;
 453                                 gtime += task_gtime(t);
 454                                 t = next_thread(t);
 455                         } while (t != task);
 456
 457                         min_flt += sig->min_flt;
 458                         maj_flt += sig->maj_flt;
 459                         thread_group_cputime_adjusted(task, &utime, &stime);
 460                         gtime += sig->gtime;
 461                 }
 462
 463                 sid = task_session_nr_ns(task, ns);
 464                 ppid = task_tgid_nr_ns(task->real_parent, ns);
 465                 pgid = task_pgrp_nr_ns(task, ns);
 466
 467                 unlock_task_sighand(task, &flags);
 468         }
 469
 470         if (permitted && (!whole || num_threads < 2))
 471                 wchan = get_wchan(task);
 472         if (!whole) {
 473                 min_flt = task->min_flt;
 474                 maj_flt = task->maj_flt;
 475                 task_cputime_adjusted(task, &utime, &stime);
 476                 gtime = task_gtime(task);
 477         }
 478
 479         /* scale priority and nice values from timeslices to -20..20 */
 480         /* to make it look like a "normal" Unix priority/nice value  */
 481         priority = task_prio(task);
 482         nice = task_nice(task);
 483
 484         /* Temporary variable needed for gcc-2.96 */
 485         /* convert timespec -> nsec*/
 486         start_time =
 487                 (unsigned long long)task->real_start_time.tv_sec * NSEC_PER_SEC
 488                                 + task->real_start_time.tv_nsec;
 489         /* convert nsec -> ticks */
 490         start_time = nsec_to_clock_t(start_time);
 491
 492         seq_printf(m, "%d (%s) %c", pid_nr_ns(pid, ns), tcomm, state);
 493         seq_put_decimal_ll(m, ' ', ppid);
 494         seq_put_decimal_ll(m, ' ', pgid);
 495         seq_put_decimal_ll(m, ' ', sid);
 496         seq_put_decimal_ll(m, ' ', tty_nr);
 497         seq_put_decimal_ll(m, ' ', tty_pgrp);
 498         seq_put_decimal_ull(m, ' ', task->flags);
 499         seq_put_decimal_ull(m, ' ', min_flt);
 500         seq_put_decimal_ull(m, ' ', cmin_flt);
 501         seq_put_decimal_ull(m, ' ', maj_flt);
 502         seq_put_decimal_ull(m, ' ', cmaj_flt);
 503         seq_put_decimal_ull(m, ' ', cputime_to_clock_t(utime));
 504         seq_put_decimal_ull(m, ' ', cputime_to_clock_t(stime));
 505         seq_put_decimal_ll(m, ' ', cputime_to_clock_t(cutime));
 506         seq_put_decimal_ll(m, ' ', cputime_to_clock_t(cstime));
 507         seq_put_decimal_ll(m, ' ', priority);
 508         seq_put_decimal_ll(m, ' ', nice);
 509         seq_put_decimal_ll(m, ' ', num_threads);
 510         seq_put_decimal_ull(m, ' ', 0);
 511         seq_put_decimal_ull(m, ' ', start_time);
 512         seq_put_decimal_ull(m, ' ', vsize);
 513         seq_put_decimal_ull(m, ' ', mm ? get_mm_rss(mm) : 0);
 514         seq_put_decimal_ull(m, ' ', rsslim);
 515         seq_put_decimal_ull(m, ' ', mm ? (permitted ? mm->start_code : 1) : 0);
 516         seq_put_decimal_ull(m, ' ', mm ? (permitted ? mm->end_code : 1) : 0);
 517         seq_put_decimal_ull(m, ' ', (permitted && mm) ? mm->start_stack : 0);
 518         seq_put_decimal_ull(m, ' ', esp);
 519         seq_put_decimal_ull(m, ' ', eip);
 520         /* The signal information here is obsolete.
 521          * It must be decimal for Linux 2.0 compatibility.
 522          * Use /proc/#/status for real-time signals.
 523          */
 524         seq_put_decimal_ull(m, ' ', task->pending.signal.sig[0] & 0x7fffffffUL);
 525         seq_put_decimal_ull(m, ' ', task->blocked.sig[0] & 0x7fffffffUL);
 526         seq_put_decimal_ull(m, ' ', sigign.sig[0] & 0x7fffffffUL);
 527         seq_put_decimal_ull(m, ' ', sigcatch.sig[0] & 0x7fffffffUL);
 528         seq_put_decimal_ull(m, ' ', wchan);
 529         seq_put_decimal_ull(m, ' ', 0);
 530         seq_put_decimal_ull(m, ' ', 0);
 531         seq_put_decimal_ll(m, ' ', task->exit_signal);
 532         seq_put_decimal_ll(m, ' ', task_cpu(task));
 533         seq_put_decimal_ull(m, ' ', task->rt_priority);
 534         seq_put_decimal_ull(m, ' ', task->policy);
 535         seq_put_decimal_ull(m, ' ', delayacct_blkio_ticks(task));
 536         seq_put_decimal_ull(m, ' ', cputime_to_clock_t(gtime));
 537         seq_put_decimal_ll(m, ' ', cputime_to_clock_t(cgtime));
 538
 539         if (mm && permitted) {
 540                 seq_put_decimal_ull(m, ' ', mm->start_data);
 541                 seq_put_decimal_ull(m, ' ', mm->end_data);
 542                 seq_put_decimal_ull(m, ' ', mm->start_brk);
 543                 seq_put_decimal_ull(m, ' ', mm->arg_start);
 544                 seq_put_decimal_ull(m, ' ', mm->arg_end);
 545                 seq_put_decimal_ull(m, ' ', mm->env_start);
 546                 seq_put_decimal_ull(m, ' ', mm->env_end);
 547         } else
 548                 seq_printf(m, " 0 0 0 0 0 0 0");
 549
 550         if (permitted)
 551                 seq_put_decimal_ll(m, ' ', task->exit_code);
 552         else
 553                 seq_put_decimal_ll(m, ' ', 0);
 554
 555         seq_putc(m, '\n');
 556         if (mm)
 557                 mmput(mm);
 558         return 0;
 559 }
 560
 561 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
 562                         struct pid *pid, struct task_struct *task)
 563 {
 564         return do_task_stat(m, ns, pid, task, 0);
 565 }
 566
 567 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
 568                         struct pid *pid, struct task_struct *task)
 569 {
 570         return do_task_stat(m, ns, pid, task, 1);
 571 }
 572
 573 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
 574                         struct pid *pid, struct task_struct *task)
 575 {
 576         unsigned long size = 0, resident = 0, shared = 0, text = 0, data = 0;
 577         struct mm_struct *mm = get_task_mm(task);
 578
 579         if (mm) {
 580                 size = task_statm(mm, &shared, &text, &data, &resident);
 581                 mmput(mm);
 582         }
 583         /*
 584          * For quick read, open code by putting numbers directly
 585          * expected format is
 586          * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
 587          *               size, resident, shared, text, data);
 588          */
 589         seq_put_decimal_ull(m, 0, size);
 590         seq_put_decimal_ull(m, ' ', resident);
 591         seq_put_decimal_ull(m, ' ', shared);
 592         seq_put_decimal_ull(m, ' ', text);
 593         seq_put_decimal_ull(m, ' ', 0);
 594         seq_put_decimal_ull(m, ' ', data);
 595         seq_put_decimal_ull(m, ' ', 0);
 596         seq_putc(m, '\n');
 597
 598         return 0;
 599 }
 600
 601 #ifdef CONFIG_CHECKPOINT_RESTORE
 602 static struct pid *
 603 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
 604 {
 605         struct task_struct *start, *task;
 606         struct pid *pid = NULL;
 607
 608         read_lock(&tasklist_lock);
 609
 610         start = pid_task(proc_pid(inode), PIDTYPE_PID);
 611         if (!start)
 612                 goto out;
 613
 614         /*
 615          * Lets try to continue searching first, this gives
 616          * us significant speedup on children-rich processes.
 617          */
 618         if (pid_prev) {
 619                 task = pid_task(pid_prev, PIDTYPE_PID);
 620                 if (task && task->real_parent == start &&
 621                     !(list_empty(&task->sibling))) {
 622                         if (list_is_last(&task->sibling, &start->children))
 623                                 goto out;
 624                         task = list_first_entry(&task->sibling,
 625                                                 struct task_struct, sibling);
 626                         pid = get_pid(task_pid(task));
 627                         goto out;
 628                 }
 629         }
 630
 631         /*
 632          * Slow search case.
 633          *
 634          * We might miss some children here if children
 635          * are exited while we were not holding the lock,
 636          * but it was never promised to be accurate that
 637          * much.
 638          *
 639          * "Just suppose that the parent sleeps, but N children
 640          *  exit after we printed their tids. Now the slow paths
 641          *  skips N extra children, we miss N tasks." (c)
 642          *
 643          * So one need to stop or freeze the leader and all
 644          * its children to get a precise result.
 645          */
 646         list_for_each_entry(task, &start->children, sibling) {
 647                 if (pos-- == 0) {
 648                         pid = get_pid(task_pid(task));
 649                         break;
 650                 }
 651         }
 652
 653 out:
 654         read_unlock(&tasklist_lock);
 655         return pid;
 656 }
 657
 658 static int children_seq_show(struct seq_file *seq, void *v)
 659 {
 660         struct inode *inode = seq->private;
 661         pid_t pid;
 662
 663         pid = pid_nr_ns(v, inode->i_sb->s_fs_info);
 664         return seq_printf(seq, "%d ", pid);
 665 }
 666
 667 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
 668 {
 669         return get_children_pid(seq->private, NULL, *pos);
 670 }
 671
 672 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 673 {
 674         struct pid *pid;
 675
 676         pid = get_children_pid(seq->private, v, *pos + 1);
 677         put_pid(v);
 678
 679         ++*pos;
 680         return pid;
 681 }
 682
 683 static void children_seq_stop(struct seq_file *seq, void *v)
 684 {
 685         put_pid(v);
 686 }
 687
 688 static const struct seq_operations children_seq_ops = {
 689         .start  = children_seq_start,
 690         .next   = children_seq_next,
 691         .stop   = children_seq_stop,
 692         .show   = children_seq_show,
 693 };
 694
 695 static int children_seq_open(struct inode *inode, struct file *file)
 696 {
 697         struct seq_file *m;
 698         int ret;
 699
 700         ret = seq_open(file, &children_seq_ops);
 701         if (ret)
 702                 return ret;
 703
 704         m = file->private_data;
 705         m->private = inode;
 706
 707         return ret;
 708 }
 709
 710 int children_seq_release(struct inode *inode, struct file *file)
 711 {
 712         seq_release(inode, file);
 713         return 0;
 714 }
 715
 716 const struct file_operations proc_tid_children_operations = {
 717         .open    = children_seq_open,
 718         .read    = seq_read,
 719         .llseek  = seq_lseek,
 720         .release = children_seq_release,
 721 };
 722 #endif /* CONFIG_CHECKPOINT_RESTORE */