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/sched/mm.h>
  64 #include <linux/sched/numa_balancing.h>
  65 #include <linux/sched/task_stack.h>
  66 #include <linux/sched/task.h>
  67 #include <linux/sched/cputime.h>
  68 #include <linux/proc_fs.h>
  69 #include <linux/ioport.h>
  70 #include <linux/uaccess.h>
  71 #include <linux/io.h>
  72 #include <linux/mm.h>
  73 #include <linux/hugetlb.h>
  74 #include <linux/pagemap.h>
  75 #include <linux/swap.h>
  76 #include <linux/smp.h>
  77 #include <linux/signal.h>
  78 #include <linux/highmem.h>
  79 #include <linux/file.h>
  80 #include <linux/fdtable.h>
  81 #include <linux/times.h>
  82 #include <linux/cpuset.h>
  83 #include <linux/rcupdate.h>
  84 #include <linux/delayacct.h>
  85 #include <linux/seq_file.h>
  86 #include <linux/pid_namespace.h>
  87 #include <linux/nospec.h>
  88 #include <linux/prctl.h>
  89 #include <linux/ptrace.h>
  90 #include <linux/tracehook.h>
  91 #include <linux/string_helpers.h>
  92 #include <linux/user_namespace.h>
  93 #include <linux/fs_struct.h>
  94
  95 #include <asm/pgtable.h>
  96 #include <asm/processor.h>
  97 #include "internal.h"
  98
  99 static inline void task_name(struct seq_file *m, struct task_struct *p)
 100 {
 101         char *buf;
 102         size_t size;
 103         char tcomm[sizeof(p->comm)];
 104         int ret;
 105
 106         get_task_comm(tcomm, p);
 107
 108         seq_puts(m, "Name:\t");
 109
 110         size = seq_get_buf(m, &buf);
 111         ret = string_escape_str(tcomm, buf, size, ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
 112         seq_commit(m, ret < size ? ret : -1);
 113
 114         seq_putc(m, '\n');
 115 }
 116
 117 /*
 118  * The task state array is a strange "bitmap" of
 119  * reasons to sleep. Thus "running" is zero, and
 120  * you can test for combinations of others with
 121  * simple bit tests.
 122  */
 123 static const char * const task_state_array[] = {
 124         "R (running)",          /*   0 */
 125         "S (sleeping)",         /*   1 */
 126         "D (disk sleep)",       /*   2 */
 127         "T (stopped)",          /*   4 */
 128         "t (tracing stop)",     /*   8 */
 129         "X (dead)",             /*  16 */
 130         "Z (zombie)",           /*  32 */
 131 };
 132
 133 static inline const char *get_task_state(struct task_struct *tsk)
 134 {
 135         unsigned int state = (tsk->state | tsk->exit_state) & TASK_REPORT;
 136
 137         /*
 138          * Parked tasks do not run; they sit in __kthread_parkme().
 139          * Without this check, we would report them as running, which is
 140          * clearly wrong, so we report them as sleeping instead.
 141          */
 142         if (tsk->state == TASK_PARKED)
 143                 state = TASK_INTERRUPTIBLE;
 144
 145         BUILD_BUG_ON(1 + ilog2(TASK_REPORT) != ARRAY_SIZE(task_state_array)-1);
 146
 147         return task_state_array[fls(state)];
 148 }
 149
 150 static inline int get_task_umask(struct task_struct *tsk)
 151 {
 152         struct fs_struct *fs;
 153         int umask = -ENOENT;
 154
 155         task_lock(tsk);
 156         fs = tsk->fs;
 157         if (fs)
 158                 umask = fs->umask;
 159         task_unlock(tsk);
 160         return umask;
 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, umask;
 169         struct task_struct *tracer;
 170         const struct cred *cred;
 171         pid_t ppid, tpid = 0, tgid, ngid;
 172         unsigned int max_fds = 0;
 173
 174         rcu_read_lock();
 175         ppid = pid_alive(p) ?
 176                 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
 177
 178         tracer = ptrace_parent(p);
 179         if (tracer)
 180                 tpid = task_pid_nr_ns(tracer, ns);
 181
 182         tgid = task_tgid_nr_ns(p, ns);
 183         ngid = task_numa_group_id(p);
 184         cred = get_task_cred(p);
 185
 186         umask = get_task_umask(p);
 187         if (umask >= 0)
 188                 seq_printf(m, "Umask:\t%#04o\n", umask);
 189
 190         task_lock(p);
 191         if (p->files)
 192                 max_fds = files_fdtable(p->files)->max_fds;
 193         task_unlock(p);
 194         rcu_read_unlock();
 195
 196         seq_printf(m, "State:\t%s", get_task_state(p));
 197
 198         seq_put_decimal_ull(m, "\nTgid:\t", tgid);
 199         seq_put_decimal_ull(m, "\nNgid:\t", ngid);
 200         seq_put_decimal_ull(m, "\nPid:\t", pid_nr_ns(pid, ns));
 201         seq_put_decimal_ull(m, "\nPPid:\t", ppid);
 202         seq_put_decimal_ull(m, "\nTracerPid:\t", tpid);
 203         seq_put_decimal_ull(m, "\nUid:\t", from_kuid_munged(user_ns, cred->uid));
 204         seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->euid));
 205         seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->suid));
 206         seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->fsuid));
 207         seq_put_decimal_ull(m, "\nGid:\t", from_kgid_munged(user_ns, cred->gid));
 208         seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->egid));
 209         seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->sgid));
 210         seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->fsgid));
 211         seq_put_decimal_ull(m, "\nFDSize:\t", max_fds);
 212
 213         seq_puts(m, "\nGroups:\t");
 214         group_info = cred->group_info;
 215         for (g = 0; g < group_info->ngroups; g++)
 216                 seq_put_decimal_ull(m, g ? " " : "",
 217                                 from_kgid_munged(user_ns, group_info->gid[g]));
 218         put_cred(cred);
 219         /* Trailing space shouldn't have been added in the first place. */
 220         seq_putc(m, ' ');
 221
 222 #ifdef CONFIG_PID_NS
 223         seq_puts(m, "\nNStgid:");
 224         for (g = ns->level; g <= pid->level; g++)
 225                 seq_put_decimal_ull(m, "\t", task_tgid_nr_ns(p, pid->numbers[g].ns));
 226         seq_puts(m, "\nNSpid:");
 227         for (g = ns->level; g <= pid->level; g++)
 228                 seq_put_decimal_ull(m, "\t", task_pid_nr_ns(p, pid->numbers[g].ns));
 229         seq_puts(m, "\nNSpgid:");
 230         for (g = ns->level; g <= pid->level; g++)
 231                 seq_put_decimal_ull(m, "\t", task_pgrp_nr_ns(p, pid->numbers[g].ns));
 232         seq_puts(m, "\nNSsid:");
 233         for (g = ns->level; g <= pid->level; g++)
 234                 seq_put_decimal_ull(m, "\t", task_session_nr_ns(p, pid->numbers[g].ns));
 235 #endif
 236         seq_putc(m, '\n');
 237 }
 238
 239 void render_sigset_t(struct seq_file *m, const char *header,
 240                                 sigset_t *set)
 241 {
 242         int i;
 243
 244         seq_puts(m, header);
 245
 246         i = _NSIG;
 247         do {
 248                 int x = 0;
 249
 250                 i -= 4;
 251                 if (sigismember(set, i+1)) x |= 1;
 252                 if (sigismember(set, i+2)) x |= 2;
 253                 if (sigismember(set, i+3)) x |= 4;
 254                 if (sigismember(set, i+4)) x |= 8;
 255                 seq_putc(m, hex_asc[x]);
 256         } while (i >= 4);
 257
 258         seq_putc(m, '\n');
 259 }
 260
 261 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
 262                                     sigset_t *catch)
 263 {
 264         struct k_sigaction *k;
 265         int i;
 266
 267         k = p->sighand->action;
 268         for (i = 1; i <= _NSIG; ++i, ++k) {
 269                 if (k->sa.sa_handler == SIG_IGN)
 270                         sigaddset(ign, i);
 271                 else if (k->sa.sa_handler != SIG_DFL)
 272                         sigaddset(catch, i);
 273         }
 274 }
 275
 276 static inline void task_sig(struct seq_file *m, struct task_struct *p)
 277 {
 278         unsigned long flags;
 279         sigset_t pending, shpending, blocked, ignored, caught;
 280         int num_threads = 0;
 281         unsigned long qsize = 0;
 282         unsigned long qlim = 0;
 283
 284         sigemptyset(&pending);
 285         sigemptyset(&shpending);
 286         sigemptyset(&blocked);
 287         sigemptyset(&ignored);
 288         sigemptyset(&caught);
 289
 290         if (lock_task_sighand(p, &flags)) {
 291                 pending = p->pending.signal;
 292                 shpending = p->signal->shared_pending.signal;
 293                 blocked = p->blocked;
 294                 collect_sigign_sigcatch(p, &ignored, &caught);
 295                 num_threads = get_nr_threads(p);
 296                 rcu_read_lock();  /* FIXME: is this correct? */
 297                 qsize = atomic_read(&__task_cred(p)->user->sigpending);
 298                 rcu_read_unlock();
 299                 qlim = task_rlimit(p, RLIMIT_SIGPENDING);
 300                 unlock_task_sighand(p, &flags);
 301         }
 302
 303         seq_put_decimal_ull(m, "Threads:\t", num_threads);
 304         seq_put_decimal_ull(m, "\nSigQ:\t", qsize);
 305         seq_put_decimal_ull(m, "/", qlim);
 306
 307         /* render them all */
 308         render_sigset_t(m, "\nSigPnd:\t", &pending);
 309         render_sigset_t(m, "ShdPnd:\t", &shpending);
 310         render_sigset_t(m, "SigBlk:\t", &blocked);
 311         render_sigset_t(m, "SigIgn:\t", &ignored);
 312         render_sigset_t(m, "SigCgt:\t", &caught);
 313 }
 314
 315 static void render_cap_t(struct seq_file *m, const char *header,
 316                         kernel_cap_t *a)
 317 {
 318         unsigned __capi;
 319
 320         seq_puts(m, header);
 321         CAP_FOR_EACH_U32(__capi) {
 322                 seq_printf(m, "%08x",
 323                            a->cap[CAP_LAST_U32 - __capi]);
 324         }
 325         seq_putc(m, '\n');
 326 }
 327
 328 static inline void task_cap(struct seq_file *m, struct task_struct *p)
 329 {
 330         const struct cred *cred;
 331         kernel_cap_t cap_inheritable, cap_permitted, cap_effective,
 332                         cap_bset, cap_ambient;
 333
 334         rcu_read_lock();
 335         cred = __task_cred(p);
 336         cap_inheritable = cred->cap_inheritable;
 337         cap_permitted   = cred->cap_permitted;
 338         cap_effective   = cred->cap_effective;
 339         cap_bset        = cred->cap_bset;
 340         cap_ambient     = cred->cap_ambient;
 341         rcu_read_unlock();
 342
 343         render_cap_t(m, "CapInh:\t", &cap_inheritable);
 344         render_cap_t(m, "CapPrm:\t", &cap_permitted);
 345         render_cap_t(m, "CapEff:\t", &cap_effective);
 346         render_cap_t(m, "CapBnd:\t", &cap_bset);
 347         render_cap_t(m, "CapAmb:\t", &cap_ambient);
 348 }
 349
 350 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
 351 {
 352         seq_put_decimal_ull(m, "NoNewPrivs:\t", task_no_new_privs(p));
 353 #ifdef CONFIG_SECCOMP
 354         seq_put_decimal_ull(m, "\nSeccomp:\t", p->seccomp.mode);
 355 #endif
 356         seq_printf(m, "\nSpeculation Store Bypass:\t");
 357         switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_STORE_BYPASS)) {
 358         case -EINVAL:
 359                 seq_printf(m, "unknown");
 360                 break;
 361         case PR_SPEC_NOT_AFFECTED:
 362                 seq_printf(m, "not vulnerable");
 363                 break;
 364         case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
 365                 seq_printf(m, "thread mitigated");
 366                 break;
 367         case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
 368                 seq_printf(m, "thread vulnerable");
 369                 break;
 370         case PR_SPEC_DISABLE:
 371                 seq_printf(m, "globally mitigated");
 372                 break;
 373         default:
 374                 seq_printf(m, "vulnerable");
 375                 break;
 376         }
 377         seq_putc(m, '\n');
 378 }
 379
 380 static inline void task_context_switch_counts(struct seq_file *m,
 381                                                 struct task_struct *p)
 382 {
 383         seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
 384         seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
 385         seq_putc(m, '\n');
 386 }
 387
 388 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
 389 {
 390         seq_printf(m, "Cpus_allowed:\t%*pb\n",
 391                    cpumask_pr_args(&task->cpus_allowed));
 392         seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
 393                    cpumask_pr_args(&task->cpus_allowed));
 394 }
 395
 396 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
 397                         struct pid *pid, struct task_struct *task)
 398 {
 399         struct mm_struct *mm = get_task_mm(task);
 400
 401         task_name(m, task);
 402         task_state(m, ns, pid, task);
 403
 404         if (mm) {
 405                 task_mem(m, mm);
 406                 mmput(mm);
 407         }
 408         task_sig(m, task);
 409         task_cap(m, task);
 410         task_seccomp(m, task);
 411         task_cpus_allowed(m, task);
 412         cpuset_task_status_allowed(m, task);
 413         task_context_switch_counts(m, task);
 414         return 0;
 415 }
 416
 417 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 418                         struct pid *pid, struct task_struct *task, int whole)
 419 {
 420         unsigned long vsize, eip, esp, wchan = 0;
 421         int priority, nice;
 422         int tty_pgrp = -1, tty_nr = 0;
 423         sigset_t sigign, sigcatch;
 424         char state;
 425         pid_t ppid = 0, pgid = -1, sid = -1;
 426         int num_threads = 0;
 427         int permitted;
 428         struct mm_struct *mm;
 429         unsigned long long start_time;
 430         unsigned long cmin_flt = 0, cmaj_flt = 0;
 431         unsigned long  min_flt = 0,  maj_flt = 0;
 432         u64 cutime, cstime, utime, stime;
 433         u64 cgtime, gtime;
 434         unsigned long rsslim = 0;
 435         char tcomm[sizeof(task->comm)];
 436         unsigned long flags;
 437
 438         state = *get_task_state(task);
 439         vsize = eip = esp = 0;
 440         permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
 441         mm = get_task_mm(task);
 442         if (mm) {
 443                 vsize = task_vsize(mm);
 444                 /*
 445                  * esp and eip are intentionally zeroed out.  There is no
 446                  * non-racy way to read them without freezing the task.
 447                  * Programs that need reliable values can use ptrace(2).
 448                  *
 449                  * The only exception is if the task is core dumping because
 450                  * a program is not able to use ptrace(2) in that case. It is
 451                  * safe because the task has stopped executing permanently.
 452                  */
 453                 if (permitted && (task->flags & PF_DUMPCORE)) {
 454                         eip = KSTK_EIP(task);
 455                         esp = KSTK_ESP(task);
 456                 }
 457         }
 458
 459         get_task_comm(tcomm, task);
 460
 461         sigemptyset(&sigign);
 462         sigemptyset(&sigcatch);
 463         cutime = cstime = utime = stime = 0;
 464         cgtime = gtime = 0;
 465
 466         if (lock_task_sighand(task, &flags)) {
 467                 struct signal_struct *sig = task->signal;
 468
 469                 if (sig->tty) {
 470                         struct pid *pgrp = tty_get_pgrp(sig->tty);
 471                         tty_pgrp = pid_nr_ns(pgrp, ns);
 472                         put_pid(pgrp);
 473                         tty_nr = new_encode_dev(tty_devnum(sig->tty));
 474                 }
 475
 476                 num_threads = get_nr_threads(task);
 477                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
 478
 479                 cmin_flt = sig->cmin_flt;
 480                 cmaj_flt = sig->cmaj_flt;
 481                 cutime = sig->cutime;
 482                 cstime = sig->cstime;
 483                 cgtime = sig->cgtime;
 484                 rsslim = ACCESS_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
 485
 486                 /* add up live thread stats at the group level */
 487                 if (whole) {
 488                         struct task_struct *t = task;
 489                         do {
 490                                 min_flt += t->min_flt;
 491                                 maj_flt += t->maj_flt;
 492                                 gtime += task_gtime(t);
 493                         } while_each_thread(task, t);
 494
 495                         min_flt += sig->min_flt;
 496                         maj_flt += sig->maj_flt;
 497                         thread_group_cputime_adjusted(task, &utime, &stime);
 498                         gtime += sig->gtime;
 499                 }
 500
 501                 sid = task_session_nr_ns(task, ns);
 502                 ppid = task_tgid_nr_ns(task->real_parent, ns);
 503                 pgid = task_pgrp_nr_ns(task, ns);
 504
 505                 unlock_task_sighand(task, &flags);
 506         }
 507
 508         if (permitted && (!whole || num_threads < 2))
 509                 wchan = get_wchan(task);
 510         if (!whole) {
 511                 min_flt = task->min_flt;
 512                 maj_flt = task->maj_flt;
 513                 task_cputime_adjusted(task, &utime, &stime);
 514                 gtime = task_gtime(task);
 515         }
 516
 517         /* scale priority and nice values from timeslices to -20..20 */
 518         /* to make it look like a "normal" Unix priority/nice value  */
 519         priority = task_prio(task);
 520         nice = task_nice(task);
 521
 522         /* convert nsec -> ticks */
 523         start_time = nsec_to_clock_t(task->real_start_time);
 524
 525         seq_printf(m, "%d (%s) %c", pid_nr_ns(pid, ns), tcomm, state);
 526         seq_put_decimal_ll(m, " ", ppid);
 527         seq_put_decimal_ll(m, " ", pgid);
 528         seq_put_decimal_ll(m, " ", sid);
 529         seq_put_decimal_ll(m, " ", tty_nr);
 530         seq_put_decimal_ll(m, " ", tty_pgrp);
 531         seq_put_decimal_ull(m, " ", task->flags);
 532         seq_put_decimal_ull(m, " ", min_flt);
 533         seq_put_decimal_ull(m, " ", cmin_flt);
 534         seq_put_decimal_ull(m, " ", maj_flt);
 535         seq_put_decimal_ull(m, " ", cmaj_flt);
 536         seq_put_decimal_ull(m, " ", nsec_to_clock_t(utime));
 537         seq_put_decimal_ull(m, " ", nsec_to_clock_t(stime));
 538         seq_put_decimal_ll(m, " ", nsec_to_clock_t(cutime));
 539         seq_put_decimal_ll(m, " ", nsec_to_clock_t(cstime));
 540         seq_put_decimal_ll(m, " ", priority);
 541         seq_put_decimal_ll(m, " ", nice);
 542         seq_put_decimal_ll(m, " ", num_threads);
 543         seq_put_decimal_ull(m, " ", 0);
 544         seq_put_decimal_ull(m, " ", start_time);
 545         seq_put_decimal_ull(m, " ", vsize);
 546         seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
 547         seq_put_decimal_ull(m, " ", rsslim);
 548         seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
 549         seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
 550         seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
 551         seq_put_decimal_ull(m, " ", esp);
 552         seq_put_decimal_ull(m, " ", eip);
 553         /* The signal information here is obsolete.
 554          * It must be decimal for Linux 2.0 compatibility.
 555          * Use /proc/#/status for real-time signals.
 556          */
 557         seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
 558         seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
 559         seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
 560         seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
 561
 562         /*
 563          * We used to output the absolute kernel address, but that's an
 564          * information leak - so instead we show a 0/1 flag here, to signal
 565          * to user-space whether there's a wchan field in /proc/PID/wchan.
 566          *
 567          * This works with older implementations of procps as well.
 568          */
 569         if (wchan)
 570                 seq_puts(m, " 1");
 571         else
 572                 seq_puts(m, " 0");
 573
 574         seq_put_decimal_ull(m, " ", 0);
 575         seq_put_decimal_ull(m, " ", 0);
 576         seq_put_decimal_ll(m, " ", task->exit_signal);
 577         seq_put_decimal_ll(m, " ", task_cpu(task));
 578         seq_put_decimal_ull(m, " ", task->rt_priority);
 579         seq_put_decimal_ull(m, " ", task->policy);
 580         seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
 581         seq_put_decimal_ull(m, " ", nsec_to_clock_t(gtime));
 582         seq_put_decimal_ll(m, " ", nsec_to_clock_t(cgtime));
 583
 584         if (mm && permitted) {
 585                 seq_put_decimal_ull(m, " ", mm->start_data);
 586                 seq_put_decimal_ull(m, " ", mm->end_data);
 587                 seq_put_decimal_ull(m, " ", mm->start_brk);
 588                 seq_put_decimal_ull(m, " ", mm->arg_start);
 589                 seq_put_decimal_ull(m, " ", mm->arg_end);
 590                 seq_put_decimal_ull(m, " ", mm->env_start);
 591                 seq_put_decimal_ull(m, " ", mm->env_end);
 592         } else
 593                 seq_puts(m, " 0 0 0 0 0 0 0");
 594
 595         if (permitted)
 596                 seq_put_decimal_ll(m, " ", task->exit_code);
 597         else
 598                 seq_puts(m, " 0");
 599
 600         seq_putc(m, '\n');
 601         if (mm)
 602                 mmput(mm);
 603         return 0;
 604 }
 605
 606 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
 607                         struct pid *pid, struct task_struct *task)
 608 {
 609         return do_task_stat(m, ns, pid, task, 0);
 610 }
 611
 612 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
 613                         struct pid *pid, struct task_struct *task)
 614 {
 615         return do_task_stat(m, ns, pid, task, 1);
 616 }
 617
 618 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
 619                         struct pid *pid, struct task_struct *task)
 620 {
 621         unsigned long size = 0, resident = 0, shared = 0, text = 0, data = 0;
 622         struct mm_struct *mm = get_task_mm(task);
 623
 624         if (mm) {
 625                 size = task_statm(mm, &shared, &text, &data, &resident);
 626                 mmput(mm);
 627         }
 628         /*
 629          * For quick read, open code by putting numbers directly
 630          * expected format is
 631          * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
 632          *               size, resident, shared, text, data);
 633          */
 634         seq_put_decimal_ull(m, "", size);
 635         seq_put_decimal_ull(m, " ", resident);
 636         seq_put_decimal_ull(m, " ", shared);
 637         seq_put_decimal_ull(m, " ", text);
 638         seq_put_decimal_ull(m, " ", 0);
 639         seq_put_decimal_ull(m, " ", data);
 640         seq_put_decimal_ull(m, " ", 0);
 641         seq_putc(m, '\n');
 642
 643         return 0;
 644 }
 645
 646 #ifdef CONFIG_PROC_CHILDREN
 647 static struct pid *
 648 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
 649 {
 650         struct task_struct *start, *task;
 651         struct pid *pid = NULL;
 652
 653         read_lock(&tasklist_lock);
 654
 655         start = pid_task(proc_pid(inode), PIDTYPE_PID);
 656         if (!start)
 657                 goto out;
 658
 659         /*
 660          * Lets try to continue searching first, this gives
 661          * us significant speedup on children-rich processes.
 662          */
 663         if (pid_prev) {
 664                 task = pid_task(pid_prev, PIDTYPE_PID);
 665                 if (task && task->real_parent == start &&
 666                     !(list_empty(&task->sibling))) {
 667                         if (list_is_last(&task->sibling, &start->children))
 668                                 goto out;
 669                         task = list_first_entry(&task->sibling,
 670                                                 struct task_struct, sibling);
 671                         pid = get_pid(task_pid(task));
 672                         goto out;
 673                 }
 674         }
 675
 676         /*
 677          * Slow search case.
 678          *
 679          * We might miss some children here if children
 680          * are exited while we were not holding the lock,
 681          * but it was never promised to be accurate that
 682          * much.
 683          *
 684          * "Just suppose that the parent sleeps, but N children
 685          *  exit after we printed their tids. Now the slow paths
 686          *  skips N extra children, we miss N tasks." (c)
 687          *
 688          * So one need to stop or freeze the leader and all
 689          * its children to get a precise result.
 690          */
 691         list_for_each_entry(task, &start->children, sibling) {
 692                 if (pos-- == 0) {
 693                         pid = get_pid(task_pid(task));
 694                         break;
 695                 }
 696         }
 697
 698 out:
 699         read_unlock(&tasklist_lock);
 700         return pid;
 701 }
 702
 703 static int children_seq_show(struct seq_file *seq, void *v)
 704 {
 705         struct inode *inode = seq->private;
 706         pid_t pid;
 707
 708         pid = pid_nr_ns(v, inode->i_sb->s_fs_info);
 709         seq_printf(seq, "%d ", pid);
 710
 711         return 0;
 712 }
 713
 714 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
 715 {
 716         return get_children_pid(seq->private, NULL, *pos);
 717 }
 718
 719 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 720 {
 721         struct pid *pid;
 722
 723         pid = get_children_pid(seq->private, v, *pos + 1);
 724         put_pid(v);
 725
 726         ++*pos;
 727         return pid;
 728 }
 729
 730 static void children_seq_stop(struct seq_file *seq, void *v)
 731 {
 732         put_pid(v);
 733 }
 734
 735 static const struct seq_operations children_seq_ops = {
 736         .start  = children_seq_start,
 737         .next   = children_seq_next,
 738         .stop   = children_seq_stop,
 739         .show   = children_seq_show,
 740 };
 741
 742 static int children_seq_open(struct inode *inode, struct file *file)
 743 {
 744         struct seq_file *m;
 745         int ret;
 746
 747         ret = seq_open(file, &children_seq_ops);
 748         if (ret)
 749                 return ret;
 750
 751         m = file->private_data;
 752         m->private = inode;
 753
 754         return ret;
 755 }
 756
 757 int children_seq_release(struct inode *inode, struct file *file)
 758 {
 759         seq_release(inode, file);
 760         return 0;
 761 }
 762
 763 const struct file_operations proc_tid_children_operations = {
 764         .open    = children_seq_open,
 765         .read    = seq_read,
 766         .llseek  = seq_lseek,
 767         .release = children_seq_release,
 768 };
 769 #endif /* CONFIG_PROC_CHILDREN */