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.h>
  64 #include <linux/proc_fs.h>
  65 #include <linux/ioport.h>
  66 #include <linux/uaccess.h>
  67 #include <linux/io.h>
  68 #include <linux/mm.h>
  69 #include <linux/hugetlb.h>
  70 #include <linux/pagemap.h>
  71 #include <linux/swap.h>
  72 #include <linux/smp.h>
  73 #include <linux/signal.h>
  74 #include <linux/highmem.h>
  75 #include <linux/file.h>
  76 #include <linux/fdtable.h>
  77 #include <linux/times.h>
  78 #include <linux/cpuset.h>
  79 #include <linux/rcupdate.h>
  80 #include <linux/delayacct.h>
  81 #include <linux/seq_file.h>
  82 #include <linux/pid_namespace.h>
  83 #include <linux/prctl.h>
  84 #include <linux/ptrace.h>
  85 #include <linux/tracehook.h>
  86 #include <linux/string_helpers.h>
  87 #include <linux/user_namespace.h>
  88 #include <linux/fs_struct.h>
  89
  90 #include <asm/pgtable.h>
  91 #include <asm/processor.h>
  92 #include "internal.h"
  93
  94 static inline void task_name(struct seq_file *m, struct task_struct *p)
  95 {
  96         char *buf;
  97         size_t size;
  98         char tcomm[sizeof(p->comm)];
  99         int ret;
 100
 101         get_task_comm(tcomm, p);
 102
 103         seq_puts(m, "Name:\t");
 104
 105         size = seq_get_buf(m, &buf);
 106         ret = string_escape_str(tcomm, buf, size, ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
 107         seq_commit(m, ret < size ? ret : -1);
 108
 109         seq_putc(m, '\n');
 110 }
 111
 112 /*
 113  * The task state array is a strange "bitmap" of
 114  * reasons to sleep. Thus "running" is zero, and
 115  * you can test for combinations of others with
 116  * simple bit tests.
 117  */
 118 static const char * const task_state_array[] = {
 119         "R (running)",          /*   0 */
 120         "S (sleeping)",         /*   1 */
 121         "D (disk sleep)",       /*   2 */
 122         "T (stopped)",          /*   4 */
 123         "t (tracing stop)",     /*   8 */
 124         "X (dead)",             /*  16 */
 125         "Z (zombie)",           /*  32 */
 126 };
 127
 128 static inline const char *get_task_state(struct task_struct *tsk)
 129 {
 130         unsigned int state = (tsk->state | tsk->exit_state) & TASK_REPORT;
 131
 132         /*
 133          * Parked tasks do not run; they sit in __kthread_parkme().
 134          * Without this check, we would report them as running, which is
 135          * clearly wrong, so we report them as sleeping instead.
 136          */
 137         if (tsk->state == TASK_PARKED)
 138                 state = TASK_INTERRUPTIBLE;
 139
 140         BUILD_BUG_ON(1 + ilog2(TASK_REPORT) != ARRAY_SIZE(task_state_array)-1);
 141
 142         return task_state_array[fls(state)];
 143 }
 144
 145 static inline int get_task_umask(struct task_struct *tsk)
 146 {
 147         struct fs_struct *fs;
 148         int umask = -ENOENT;
 149
 150         task_lock(tsk);
 151         fs = tsk->fs;
 152         if (fs)
 153                 umask = fs->umask;
 154         task_unlock(tsk);
 155         return umask;
 156 }
 157
 158 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
 159                                 struct pid *pid, struct task_struct *p)
 160 {
 161         struct user_namespace *user_ns = seq_user_ns(m);
 162         struct group_info *group_info;
 163         int g, umask;
 164         struct task_struct *tracer;
 165         const struct cred *cred;
 166         pid_t ppid, tpid = 0, tgid, ngid;
 167         unsigned int max_fds = 0;
 168
 169         rcu_read_lock();
 170         ppid = pid_alive(p) ?
 171                 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
 172
 173         tracer = ptrace_parent(p);
 174         if (tracer)
 175                 tpid = task_pid_nr_ns(tracer, ns);
 176
 177         tgid = task_tgid_nr_ns(p, ns);
 178         ngid = task_numa_group_id(p);
 179         cred = get_task_cred(p);
 180
 181         umask = get_task_umask(p);
 182         if (umask >= 0)
 183                 seq_printf(m, "Umask:\t%#04o\n", umask);
 184
 185         task_lock(p);
 186         if (p->files)
 187                 max_fds = files_fdtable(p->files)->max_fds;
 188         task_unlock(p);
 189         rcu_read_unlock();
 190
 191         seq_printf(m, "State:\t%s", get_task_state(p));
 192
 193         seq_put_decimal_ull(m, "\nTgid:\t", tgid);
 194         seq_put_decimal_ull(m, "\nNgid:\t", ngid);
 195         seq_put_decimal_ull(m, "\nPid:\t", pid_nr_ns(pid, ns));
 196         seq_put_decimal_ull(m, "\nPPid:\t", ppid);
 197         seq_put_decimal_ull(m, "\nTracerPid:\t", tpid);
 198         seq_put_decimal_ull(m, "\nUid:\t", from_kuid_munged(user_ns, cred->uid));
 199         seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->euid));
 200         seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->suid));
 201         seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->fsuid));
 202         seq_put_decimal_ull(m, "\nGid:\t", from_kgid_munged(user_ns, cred->gid));
 203         seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->egid));
 204         seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->sgid));
 205         seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->fsgid));
 206         seq_put_decimal_ull(m, "\nFDSize:\t", max_fds);
 207
 208         seq_puts(m, "\nGroups:\t");
 209         group_info = cred->group_info;
 210         for (g = 0; g < group_info->ngroups; g++)
 211                 seq_put_decimal_ull(m, g ? " " : "",
 212                                 from_kgid_munged(user_ns, group_info->gid[g]));
 213         put_cred(cred);
 214         /* Trailing space shouldn't have been added in the first place. */
 215         seq_putc(m, ' ');
 216
 217 #ifdef CONFIG_PID_NS
 218         seq_puts(m, "\nNStgid:");
 219         for (g = ns->level; g <= pid->level; g++)
 220                 seq_put_decimal_ull(m, "\t", task_tgid_nr_ns(p, pid->numbers[g].ns));
 221         seq_puts(m, "\nNSpid:");
 222         for (g = ns->level; g <= pid->level; g++)
 223                 seq_put_decimal_ull(m, "\t", task_pid_nr_ns(p, pid->numbers[g].ns));
 224         seq_puts(m, "\nNSpgid:");
 225         for (g = ns->level; g <= pid->level; g++)
 226                 seq_put_decimal_ull(m, "\t", task_pgrp_nr_ns(p, pid->numbers[g].ns));
 227         seq_puts(m, "\nNSsid:");
 228         for (g = ns->level; g <= pid->level; g++)
 229                 seq_put_decimal_ull(m, "\t", task_session_nr_ns(p, pid->numbers[g].ns));
 230 #endif
 231         seq_putc(m, '\n');
 232 }
 233
 234 void render_sigset_t(struct seq_file *m, const char *header,
 235                                 sigset_t *set)
 236 {
 237         int i;
 238
 239         seq_puts(m, header);
 240
 241         i = _NSIG;
 242         do {
 243                 int x = 0;
 244
 245                 i -= 4;
 246                 if (sigismember(set, i+1)) x |= 1;
 247                 if (sigismember(set, i+2)) x |= 2;
 248                 if (sigismember(set, i+3)) x |= 4;
 249                 if (sigismember(set, i+4)) x |= 8;
 250                 seq_printf(m, "%x", x);
 251         } while (i >= 4);
 252
 253         seq_putc(m, '\n');
 254 }
 255
 256 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
 257                                     sigset_t *catch)
 258 {
 259         struct k_sigaction *k;
 260         int i;
 261
 262         k = p->sighand->action;
 263         for (i = 1; i <= _NSIG; ++i, ++k) {
 264                 if (k->sa.sa_handler == SIG_IGN)
 265                         sigaddset(ign, i);
 266                 else if (k->sa.sa_handler != SIG_DFL)
 267                         sigaddset(catch, i);
 268         }
 269 }
 270
 271 static inline void task_sig(struct seq_file *m, struct task_struct *p)
 272 {
 273         unsigned long flags;
 274         sigset_t pending, shpending, blocked, ignored, caught;
 275         int num_threads = 0;
 276         unsigned long qsize = 0;
 277         unsigned long qlim = 0;
 278
 279         sigemptyset(&pending);
 280         sigemptyset(&shpending);
 281         sigemptyset(&blocked);
 282         sigemptyset(&ignored);
 283         sigemptyset(&caught);
 284
 285         if (lock_task_sighand(p, &flags)) {
 286                 pending = p->pending.signal;
 287                 shpending = p->signal->shared_pending.signal;
 288                 blocked = p->blocked;
 289                 collect_sigign_sigcatch(p, &ignored, &caught);
 290                 num_threads = get_nr_threads(p);
 291                 rcu_read_lock();  /* FIXME: is this correct? */
 292                 qsize = atomic_read(&__task_cred(p)->user->sigpending);
 293                 rcu_read_unlock();
 294                 qlim = task_rlimit(p, RLIMIT_SIGPENDING);
 295                 unlock_task_sighand(p, &flags);
 296         }
 297
 298         seq_put_decimal_ull(m, "Threads:\t", num_threads);
 299         seq_put_decimal_ull(m, "\nSigQ:\t", qsize);
 300         seq_put_decimal_ull(m, "/", qlim);
 301
 302         /* render them all */
 303         render_sigset_t(m, "\nSigPnd:\t", &pending);
 304         render_sigset_t(m, "ShdPnd:\t", &shpending);
 305         render_sigset_t(m, "SigBlk:\t", &blocked);
 306         render_sigset_t(m, "SigIgn:\t", &ignored);
 307         render_sigset_t(m, "SigCgt:\t", &caught);
 308 }
 309
 310 static void render_cap_t(struct seq_file *m, const char *header,
 311                         kernel_cap_t *a)
 312 {
 313         unsigned __capi;
 314
 315         seq_puts(m, header);
 316         CAP_FOR_EACH_U32(__capi) {
 317                 seq_printf(m, "%08x",
 318                            a->cap[CAP_LAST_U32 - __capi]);
 319         }
 320         seq_putc(m, '\n');
 321 }
 322
 323 static inline void task_cap(struct seq_file *m, struct task_struct *p)
 324 {
 325         const struct cred *cred;
 326         kernel_cap_t cap_inheritable, cap_permitted, cap_effective,
 327                         cap_bset, cap_ambient;
 328
 329         rcu_read_lock();
 330         cred = __task_cred(p);
 331         cap_inheritable = cred->cap_inheritable;
 332         cap_permitted   = cred->cap_permitted;
 333         cap_effective   = cred->cap_effective;
 334         cap_bset        = cred->cap_bset;
 335         cap_ambient     = cred->cap_ambient;
 336         rcu_read_unlock();
 337
 338         render_cap_t(m, "CapInh:\t", &cap_inheritable);
 339         render_cap_t(m, "CapPrm:\t", &cap_permitted);
 340         render_cap_t(m, "CapEff:\t", &cap_effective);
 341         render_cap_t(m, "CapBnd:\t", &cap_bset);
 342         render_cap_t(m, "CapAmb:\t", &cap_ambient);
 343 }
 344
 345 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
 346 {
 347 #ifdef CONFIG_SECCOMP
 348         seq_put_decimal_ull(m, "Seccomp:\t", p->seccomp.mode);
 349         seq_putc(m, '\n');
 350 #endif
 351         seq_printf(m, "Speculation_Store_Bypass:\t");
 352         switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_STORE_BYPASS)) {
 353         case -EINVAL:
 354                 seq_printf(m, "unknown");
 355                 break;
 356         case PR_SPEC_NOT_AFFECTED:
 357                 seq_printf(m, "not vulnerable");
 358                 break;
 359         case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
 360                 seq_printf(m, "thread force mitigated");
 361                 break;
 362         case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
 363                 seq_printf(m, "thread mitigated");
 364                 break;
 365         case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
 366                 seq_printf(m, "thread vulnerable");
 367                 break;
 368         case PR_SPEC_DISABLE:
 369                 seq_printf(m, "globally mitigated");
 370                 break;
 371         default:
 372                 seq_printf(m, "vulnerable");
 373                 break;
 374         }
 375         seq_putc(m, '\n');
 376 }
 377
 378 static inline void task_context_switch_counts(struct seq_file *m,
 379                                                 struct task_struct *p)
 380 {
 381         seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
 382         seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
 383         seq_putc(m, '\n');
 384 }
 385
 386 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
 387 {
 388         seq_printf(m, "Cpus_allowed:\t%*pb\n",
 389                    cpumask_pr_args(&task->cpus_allowed));
 390         seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
 391                    cpumask_pr_args(&task->cpus_allowed));
 392 }
 393
 394 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
 395                         struct pid *pid, struct task_struct *task)
 396 {
 397         struct mm_struct *mm = get_task_mm(task);
 398
 399         task_name(m, task);
 400         task_state(m, ns, pid, task);
 401
 402         if (mm) {
 403                 task_mem(m, mm);
 404                 mmput(mm);
 405         }
 406         task_sig(m, task);
 407         task_cap(m, task);
 408         task_seccomp(m, task);
 409         task_cpus_allowed(m, task);
 410         cpuset_task_status_allowed(m, task);
 411         task_context_switch_counts(m, task);
 412         return 0;
 413 }
 414
 415 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 416                         struct pid *pid, struct task_struct *task, int whole)
 417 {
 418         unsigned long vsize, eip, esp, wchan = 0;
 419         int priority, nice;
 420         int tty_pgrp = -1, tty_nr = 0;
 421         sigset_t sigign, sigcatch;
 422         char state;
 423         pid_t ppid = 0, pgid = -1, sid = -1;
 424         int num_threads = 0;
 425         int permitted;
 426         struct mm_struct *mm;
 427         unsigned long long start_time;
 428         unsigned long cmin_flt = 0, cmaj_flt = 0;
 429         unsigned long  min_flt = 0,  maj_flt = 0;
 430         cputime_t cutime, cstime, utime, stime;
 431         cputime_t cgtime, gtime;
 432         unsigned long rsslim = 0;
 433         char tcomm[sizeof(task->comm)];
 434         unsigned long flags;
 435
 436         state = *get_task_state(task);
 437         vsize = eip = esp = 0;
 438         permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
 439         mm = get_task_mm(task);
 440         if (mm) {
 441                 vsize = task_vsize(mm);
 442                 /*
 443                  * esp and eip are intentionally zeroed out.  There is no
 444                  * non-racy way to read them without freezing the task.
 445                  * Programs that need reliable values can use ptrace(2).
 446                  *
 447                  * The only exception is if the task is core dumping because
 448                  * a program is not able to use ptrace(2) in that case. It is
 449                  * safe because the task has stopped executing permanently.
 450                  */
 451                 if (permitted && (task->flags & (PF_EXITING|PF_DUMPCORE))) {
 452                         if (try_get_task_stack(task)) {
 453                                 eip = KSTK_EIP(task);
 454                                 esp = KSTK_ESP(task);
 455                                 put_task_stack(task);
 456                         }
 457                 }
 458         }
 459
 460         get_task_comm(tcomm, task);
 461
 462         sigemptyset(&sigign);
 463         sigemptyset(&sigcatch);
 464         cutime = cstime = utime = stime = 0;
 465         cgtime = gtime = 0;
 466
 467         if (lock_task_sighand(task, &flags)) {
 468                 struct signal_struct *sig = task->signal;
 469
 470                 if (sig->tty) {
 471                         struct pid *pgrp = tty_get_pgrp(sig->tty);
 472                         tty_pgrp = pid_nr_ns(pgrp, ns);
 473                         put_pid(pgrp);
 474                         tty_nr = new_encode_dev(tty_devnum(sig->tty));
 475                 }
 476
 477                 num_threads = get_nr_threads(task);
 478                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
 479
 480                 cmin_flt = sig->cmin_flt;
 481                 cmaj_flt = sig->cmaj_flt;
 482                 cutime = sig->cutime;
 483                 cstime = sig->cstime;
 484                 cgtime = sig->cgtime;
 485                 rsslim = ACCESS_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
 486
 487                 /* add up live thread stats at the group level */
 488                 if (whole) {
 489                         struct task_struct *t = task;
 490                         do {
 491                                 min_flt += t->min_flt;
 492                                 maj_flt += t->maj_flt;
 493                                 gtime += task_gtime(t);
 494                         } while_each_thread(task, t);
 495
 496                         min_flt += sig->min_flt;
 497                         maj_flt += sig->maj_flt;
 498                         thread_group_cputime_adjusted(task, &utime, &stime);
 499                         gtime += sig->gtime;
 500                 }
 501
 502                 sid = task_session_nr_ns(task, ns);
 503                 ppid = task_tgid_nr_ns(task->real_parent, ns);
 504                 pgid = task_pgrp_nr_ns(task, ns);
 505
 506                 unlock_task_sighand(task, &flags);
 507         }
 508
 509         if (permitted && (!whole || num_threads < 2))
 510                 wchan = get_wchan(task);
 511         if (!whole) {
 512                 min_flt = task->min_flt;
 513                 maj_flt = task->maj_flt;
 514                 task_cputime_adjusted(task, &utime, &stime);
 515                 gtime = task_gtime(task);
 516         }
 517
 518         /* scale priority and nice values from timeslices to -20..20 */
 519         /* to make it look like a "normal" Unix priority/nice value  */
 520         priority = task_prio(task);
 521         nice = task_nice(task);
 522
 523         /* convert nsec -> ticks */
 524         start_time = nsec_to_clock_t(task->real_start_time);
 525
 526         seq_printf(m, "%d (%s) %c", pid_nr_ns(pid, ns), tcomm, state);
 527         seq_put_decimal_ll(m, " ", ppid);
 528         seq_put_decimal_ll(m, " ", pgid);
 529         seq_put_decimal_ll(m, " ", sid);
 530         seq_put_decimal_ll(m, " ", tty_nr);
 531         seq_put_decimal_ll(m, " ", tty_pgrp);
 532         seq_put_decimal_ull(m, " ", task->flags);
 533         seq_put_decimal_ull(m, " ", min_flt);
 534         seq_put_decimal_ull(m, " ", cmin_flt);
 535         seq_put_decimal_ull(m, " ", maj_flt);
 536         seq_put_decimal_ull(m, " ", cmaj_flt);
 537         seq_put_decimal_ull(m, " ", cputime_to_clock_t(utime));
 538         seq_put_decimal_ull(m, " ", cputime_to_clock_t(stime));
 539         seq_put_decimal_ll(m, " ", cputime_to_clock_t(cutime));
 540         seq_put_decimal_ll(m, " ", cputime_to_clock_t(cstime));
 541         seq_put_decimal_ll(m, " ", priority);
 542         seq_put_decimal_ll(m, " ", nice);
 543         seq_put_decimal_ll(m, " ", num_threads);
 544         seq_put_decimal_ull(m, " ", 0);
 545         seq_put_decimal_ull(m, " ", start_time);
 546         seq_put_decimal_ull(m, " ", vsize);
 547         seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
 548         seq_put_decimal_ull(m, " ", rsslim);
 549         seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
 550         seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
 551         seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
 552         seq_put_decimal_ull(m, " ", esp);
 553         seq_put_decimal_ull(m, " ", eip);
 554         /* The signal information here is obsolete.
 555          * It must be decimal for Linux 2.0 compatibility.
 556          * Use /proc/#/status for real-time signals.
 557          */
 558         seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
 559         seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
 560         seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
 561         seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
 562
 563         /*
 564          * We used to output the absolute kernel address, but that's an
 565          * information leak - so instead we show a 0/1 flag here, to signal
 566          * to user-space whether there's a wchan field in /proc/PID/wchan.
 567          *
 568          * This works with older implementations of procps as well.
 569          */
 570         if (wchan)
 571                 seq_puts(m, " 1");
 572         else
 573                 seq_puts(m, " 0");
 574
 575         seq_put_decimal_ull(m, " ", 0);
 576         seq_put_decimal_ull(m, " ", 0);
 577         seq_put_decimal_ll(m, " ", task->exit_signal);
 578         seq_put_decimal_ll(m, " ", task_cpu(task));
 579         seq_put_decimal_ull(m, " ", task->rt_priority);
 580         seq_put_decimal_ull(m, " ", task->policy);
 581         seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
 582         seq_put_decimal_ull(m, " ", cputime_to_clock_t(gtime));
 583         seq_put_decimal_ll(m, " ", cputime_to_clock_t(cgtime));
 584
 585         if (mm && permitted) {
 586                 seq_put_decimal_ull(m, " ", mm->start_data);
 587                 seq_put_decimal_ull(m, " ", mm->end_data);
 588                 seq_put_decimal_ull(m, " ", mm->start_brk);
 589                 seq_put_decimal_ull(m, " ", mm->arg_start);
 590                 seq_put_decimal_ull(m, " ", mm->arg_end);
 591                 seq_put_decimal_ull(m, " ", mm->env_start);
 592                 seq_put_decimal_ull(m, " ", mm->env_end);
 593         } else
 594                 seq_puts(m, " 0 0 0 0 0 0 0");
 595
 596         if (permitted)
 597                 seq_put_decimal_ll(m, " ", task->exit_code);
 598         else
 599                 seq_puts(m, " 0");
 600
 601         seq_putc(m, '\n');
 602         if (mm)
 603                 mmput(mm);
 604         return 0;
 605 }
 606
 607 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
 608                         struct pid *pid, struct task_struct *task)
 609 {
 610         return do_task_stat(m, ns, pid, task, 0);
 611 }
 612
 613 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
 614                         struct pid *pid, struct task_struct *task)
 615 {
 616         return do_task_stat(m, ns, pid, task, 1);
 617 }
 618
 619 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
 620                         struct pid *pid, struct task_struct *task)
 621 {
 622         unsigned long size = 0, resident = 0, shared = 0, text = 0, data = 0;
 623         struct mm_struct *mm = get_task_mm(task);
 624
 625         if (mm) {
 626                 size = task_statm(mm, &shared, &text, &data, &resident);
 627                 mmput(mm);
 628         }
 629         /*
 630          * For quick read, open code by putting numbers directly
 631          * expected format is
 632          * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
 633          *               size, resident, shared, text, data);
 634          */
 635         seq_put_decimal_ull(m, "", size);
 636         seq_put_decimal_ull(m, " ", resident);
 637         seq_put_decimal_ull(m, " ", shared);
 638         seq_put_decimal_ull(m, " ", text);
 639         seq_put_decimal_ull(m, " ", 0);
 640         seq_put_decimal_ull(m, " ", data);
 641         seq_put_decimal_ull(m, " ", 0);
 642         seq_putc(m, '\n');
 643
 644         return 0;
 645 }
 646
 647 #ifdef CONFIG_PROC_CHILDREN
 648 static struct pid *
 649 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
 650 {
 651         struct task_struct *start, *task;
 652         struct pid *pid = NULL;
 653
 654         read_lock(&tasklist_lock);
 655
 656         start = pid_task(proc_pid(inode), PIDTYPE_PID);
 657         if (!start)
 658                 goto out;
 659
 660         /*
 661          * Lets try to continue searching first, this gives
 662          * us significant speedup on children-rich processes.
 663          */
 664         if (pid_prev) {
 665                 task = pid_task(pid_prev, PIDTYPE_PID);
 666                 if (task && task->real_parent == start &&
 667                     !(list_empty(&task->sibling))) {
 668                         if (list_is_last(&task->sibling, &start->children))
 669                                 goto out;
 670                         task = list_first_entry(&task->sibling,
 671                                                 struct task_struct, sibling);
 672                         pid = get_pid(task_pid(task));
 673                         goto out;
 674                 }
 675         }
 676
 677         /*
 678          * Slow search case.
 679          *
 680          * We might miss some children here if children
 681          * are exited while we were not holding the lock,
 682          * but it was never promised to be accurate that
 683          * much.
 684          *
 685          * "Just suppose that the parent sleeps, but N children
 686          *  exit after we printed their tids. Now the slow paths
 687          *  skips N extra children, we miss N tasks." (c)
 688          *
 689          * So one need to stop or freeze the leader and all
 690          * its children to get a precise result.
 691          */
 692         list_for_each_entry(task, &start->children, sibling) {
 693                 if (pos-- == 0) {
 694                         pid = get_pid(task_pid(task));
 695                         break;
 696                 }
 697         }
 698
 699 out:
 700         read_unlock(&tasklist_lock);
 701         return pid;
 702 }
 703
 704 static int children_seq_show(struct seq_file *seq, void *v)
 705 {
 706         struct inode *inode = seq->private;
 707         pid_t pid;
 708
 709         pid = pid_nr_ns(v, inode->i_sb->s_fs_info);
 710         seq_printf(seq, "%d ", pid);
 711
 712         return 0;
 713 }
 714
 715 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
 716 {
 717         return get_children_pid(seq->private, NULL, *pos);
 718 }
 719
 720 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 721 {
 722         struct pid *pid;
 723
 724         pid = get_children_pid(seq->private, v, *pos + 1);
 725         put_pid(v);
 726
 727         ++*pos;
 728         return pid;
 729 }
 730
 731 static void children_seq_stop(struct seq_file *seq, void *v)
 732 {
 733         put_pid(v);
 734 }
 735
 736 static const struct seq_operations children_seq_ops = {
 737         .start  = children_seq_start,
 738         .next   = children_seq_next,
 739         .stop   = children_seq_stop,
 740         .show   = children_seq_show,
 741 };
 742
 743 static int children_seq_open(struct inode *inode, struct file *file)
 744 {
 745         struct seq_file *m;
 746         int ret;
 747
 748         ret = seq_open(file, &children_seq_ops);
 749         if (ret)
 750                 return ret;
 751
 752         m = file->private_data;
 753         m->private = inode;
 754
 755         return ret;
 756 }
 757
 758 int children_seq_release(struct inode *inode, struct file *file)
 759 {
 760         seq_release(inode, file);
 761         return 0;
 762 }
 763
 764 const struct file_operations proc_tid_children_operations = {
 765         .open    = children_seq_open,
 766         .read    = seq_read,
 767         .llseek  = seq_lseek,
 768         .release = children_seq_release,
 769 };
 770 #endif /* CONFIG_PROC_CHILDREN */