1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/proc/array.c
5 * Copyright (C) 1992 by Linus Torvalds
6 * based on ideas by Darren Senn
9 * Michael. K. Johnson: stat,statm extensions.
10 * <johnsonm@stolaf.edu>
12 * Pauline Middelink : Made cmdline,envline only break at '\0's, to
13 * make sure SET_PROCTITLE works. Also removed
14 * bad '!' which forced address recalculation for
15 * EVERY character on the current page.
16 * <middelin@polyware.iaf.nl>
18 * Danny ter Haar : added cpuinfo
21 * Alessandro Rubini : profile extension.
22 * <rubini@ipvvis.unipv.it>
24 * Jeff Tranter : added BogoMips field to cpuinfo
25 * <Jeff_Tranter@Mitel.COM>
27 * Bruno Haible : remove 4K limit for the maps file
28 * <haible@ma2s2.mathematik.uni-karlsruhe.de>
30 * Yves Arrouye : remove removal of trailing spaces in get_array.
31 * <Yves.Arrouye@marin.fdn.fr>
33 * Jerome Forissier : added per-CPU time information to /proc/stat
34 * and /proc/<pid>/cpu extension
35 * <forissier@isia.cma.fr>
36 * - Incorporation and non-SMP safe operation
37 * of forissier patch in 2.1.78 by
38 * Hans Marcus <crowbar@concepts.nl>
40 * aeb@cwi.nl : /proc/partitions
43 * Alan Cox : security fixes.
44 * <alan@lxorguk.ukuu.org.uk>
46 * Al Viro : safe handling of mm_struct
48 * Gerhard Wichert : added BIGMEM support
49 * Siemens AG <Gerhard.Wichert@pdb.siemens.de>
51 * Al Viro & Jeff Garzik : moved most of the thing into base.c and
52 * : proc_misc.c. The rest may eventually go into
56 #include <linux/types.h>
57 #include <linux/errno.h>
58 #include <linux/time.h>
59 #include <linux/time_namespace.h>
60 #include <linux/kernel.h>
61 #include <linux/kernel_stat.h>
62 #include <linux/tty.h>
63 #include <linux/string.h>
64 #include <linux/mman.h>
65 #include <linux/sched/mm.h>
66 #include <linux/sched/numa_balancing.h>
67 #include <linux/sched/task_stack.h>
68 #include <linux/sched/task.h>
69 #include <linux/sched/cputime.h>
70 #include <linux/proc_fs.h>
71 #include <linux/ioport.h>
72 #include <linux/uaccess.h>
75 #include <linux/hugetlb.h>
76 #include <linux/pagemap.h>
77 #include <linux/swap.h>
78 #include <linux/smp.h>
79 #include <linux/signal.h>
80 #include <linux/highmem.h>
81 #include <linux/file.h>
82 #include <linux/fdtable.h>
83 #include <linux/times.h>
84 #include <linux/cpuset.h>
85 #include <linux/rcupdate.h>
86 #include <linux/delayacct.h>
87 #include <linux/seq_file.h>
88 #include <linux/pid_namespace.h>
89 #include <linux/prctl.h>
90 #include <linux/ptrace.h>
91 #include <linux/tracehook.h>
92 #include <linux/string_helpers.h>
93 #include <linux/user_namespace.h>
94 #include <linux/fs_struct.h>
96 #include <asm/processor.h>
99 void proc_task_name(struct seq_file *m, struct task_struct *p, bool escape)
106 if (p->flags & PF_WQ_WORKER)
107 wq_worker_comm(tcomm, sizeof(tcomm), p);
109 __get_task_comm(tcomm, sizeof(tcomm), p);
111 size = seq_get_buf(m, &buf);
113 ret = string_escape_str(tcomm, buf, size,
114 ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
118 ret = strscpy(buf, tcomm, size);
125 * The task state array is a strange "bitmap" of
126 * reasons to sleep. Thus "running" is zero, and
127 * you can test for combinations of others with
130 static const char * const task_state_array[] = {
132 /* states in TASK_REPORT: */
133 "R (running)", /* 0x00 */
134 "S (sleeping)", /* 0x01 */
135 "D (disk sleep)", /* 0x02 */
136 "T (stopped)", /* 0x04 */
137 "t (tracing stop)", /* 0x08 */
138 "X (dead)", /* 0x10 */
139 "Z (zombie)", /* 0x20 */
140 "P (parked)", /* 0x40 */
142 /* states beyond TASK_REPORT: */
143 "I (idle)", /* 0x80 */
146 static inline const char *get_task_state(struct task_struct *tsk)
148 BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != ARRAY_SIZE(task_state_array));
149 return task_state_array[task_state_index(tsk)];
152 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
153 struct pid *pid, struct task_struct *p)
155 struct user_namespace *user_ns = seq_user_ns(m);
156 struct group_info *group_info;
158 struct task_struct *tracer;
159 const struct cred *cred;
160 pid_t ppid, tpid = 0, tgid, ngid;
161 unsigned int max_fds = 0;
164 ppid = pid_alive(p) ?
165 task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
167 tracer = ptrace_parent(p);
169 tpid = task_pid_nr_ns(tracer, ns);
171 tgid = task_tgid_nr_ns(p, ns);
172 ngid = task_numa_group_id(p);
173 cred = get_task_cred(p);
177 umask = p->fs->umask;
179 max_fds = files_fdtable(p->files)->max_fds;
184 seq_printf(m, "Umask:\t%#04o\n", umask);
185 seq_puts(m, "State:\t");
186 seq_puts(m, get_task_state(p));
188 seq_put_decimal_ull(m, "\nTgid:\t", tgid);
189 seq_put_decimal_ull(m, "\nNgid:\t", ngid);
190 seq_put_decimal_ull(m, "\nPid:\t", pid_nr_ns(pid, ns));
191 seq_put_decimal_ull(m, "\nPPid:\t", ppid);
192 seq_put_decimal_ull(m, "\nTracerPid:\t", tpid);
193 seq_put_decimal_ull(m, "\nUid:\t", from_kuid_munged(user_ns, cred->uid));
194 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->euid));
195 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->suid));
196 seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->fsuid));
197 seq_put_decimal_ull(m, "\nGid:\t", from_kgid_munged(user_ns, cred->gid));
198 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->egid));
199 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->sgid));
200 seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->fsgid));
201 seq_put_decimal_ull(m, "\nFDSize:\t", max_fds);
203 seq_puts(m, "\nGroups:\t");
204 group_info = cred->group_info;
205 for (g = 0; g < group_info->ngroups; g++)
206 seq_put_decimal_ull(m, g ? " " : "",
207 from_kgid_munged(user_ns, group_info->gid[g]));
209 /* Trailing space shouldn't have been added in the first place. */
213 seq_puts(m, "\nNStgid:");
214 for (g = ns->level; g <= pid->level; g++)
215 seq_put_decimal_ull(m, "\t", task_tgid_nr_ns(p, pid->numbers[g].ns));
216 seq_puts(m, "\nNSpid:");
217 for (g = ns->level; g <= pid->level; g++)
218 seq_put_decimal_ull(m, "\t", task_pid_nr_ns(p, pid->numbers[g].ns));
219 seq_puts(m, "\nNSpgid:");
220 for (g = ns->level; g <= pid->level; g++)
221 seq_put_decimal_ull(m, "\t", task_pgrp_nr_ns(p, pid->numbers[g].ns));
222 seq_puts(m, "\nNSsid:");
223 for (g = ns->level; g <= pid->level; g++)
224 seq_put_decimal_ull(m, "\t", task_session_nr_ns(p, pid->numbers[g].ns));
229 void render_sigset_t(struct seq_file *m, const char *header,
241 if (sigismember(set, i+1)) x |= 1;
242 if (sigismember(set, i+2)) x |= 2;
243 if (sigismember(set, i+3)) x |= 4;
244 if (sigismember(set, i+4)) x |= 8;
245 seq_putc(m, hex_asc[x]);
251 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *sigign,
254 struct k_sigaction *k;
257 k = p->sighand->action;
258 for (i = 1; i <= _NSIG; ++i, ++k) {
259 if (k->sa.sa_handler == SIG_IGN)
260 sigaddset(sigign, i);
261 else if (k->sa.sa_handler != SIG_DFL)
262 sigaddset(sigcatch, i);
266 static inline void task_sig(struct seq_file *m, struct task_struct *p)
269 sigset_t pending, shpending, blocked, ignored, caught;
271 unsigned int qsize = 0;
272 unsigned long qlim = 0;
274 sigemptyset(&pending);
275 sigemptyset(&shpending);
276 sigemptyset(&blocked);
277 sigemptyset(&ignored);
278 sigemptyset(&caught);
280 if (lock_task_sighand(p, &flags)) {
281 pending = p->pending.signal;
282 shpending = p->signal->shared_pending.signal;
283 blocked = p->blocked;
284 collect_sigign_sigcatch(p, &ignored, &caught);
285 num_threads = get_nr_threads(p);
286 rcu_read_lock(); /* FIXME: is this correct? */
287 qsize = atomic_read(&__task_cred(p)->user->sigpending);
289 qlim = task_rlimit(p, RLIMIT_SIGPENDING);
290 unlock_task_sighand(p, &flags);
293 seq_put_decimal_ull(m, "Threads:\t", num_threads);
294 seq_put_decimal_ull(m, "\nSigQ:\t", qsize);
295 seq_put_decimal_ull(m, "/", qlim);
297 /* render them all */
298 render_sigset_t(m, "\nSigPnd:\t", &pending);
299 render_sigset_t(m, "ShdPnd:\t", &shpending);
300 render_sigset_t(m, "SigBlk:\t", &blocked);
301 render_sigset_t(m, "SigIgn:\t", &ignored);
302 render_sigset_t(m, "SigCgt:\t", &caught);
305 static void render_cap_t(struct seq_file *m, const char *header,
311 CAP_FOR_EACH_U32(__capi) {
312 seq_put_hex_ll(m, NULL,
313 a->cap[CAP_LAST_U32 - __capi], 8);
318 static inline void task_cap(struct seq_file *m, struct task_struct *p)
320 const struct cred *cred;
321 kernel_cap_t cap_inheritable, cap_permitted, cap_effective,
322 cap_bset, cap_ambient;
325 cred = __task_cred(p);
326 cap_inheritable = cred->cap_inheritable;
327 cap_permitted = cred->cap_permitted;
328 cap_effective = cred->cap_effective;
329 cap_bset = cred->cap_bset;
330 cap_ambient = cred->cap_ambient;
333 render_cap_t(m, "CapInh:\t", &cap_inheritable);
334 render_cap_t(m, "CapPrm:\t", &cap_permitted);
335 render_cap_t(m, "CapEff:\t", &cap_effective);
336 render_cap_t(m, "CapBnd:\t", &cap_bset);
337 render_cap_t(m, "CapAmb:\t", &cap_ambient);
340 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
342 seq_put_decimal_ull(m, "NoNewPrivs:\t", task_no_new_privs(p));
343 #ifdef CONFIG_SECCOMP
344 seq_put_decimal_ull(m, "\nSeccomp:\t", p->seccomp.mode);
345 seq_put_decimal_ull(m, "\nSeccomp_filters:\t",
346 atomic_read(&p->seccomp.filter_count));
348 seq_puts(m, "\nSpeculation_Store_Bypass:\t");
349 switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_STORE_BYPASS)) {
351 seq_puts(m, "unknown");
353 case PR_SPEC_NOT_AFFECTED:
354 seq_puts(m, "not vulnerable");
356 case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
357 seq_puts(m, "thread force mitigated");
359 case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
360 seq_puts(m, "thread mitigated");
362 case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
363 seq_puts(m, "thread vulnerable");
365 case PR_SPEC_DISABLE:
366 seq_puts(m, "globally mitigated");
369 seq_puts(m, "vulnerable");
373 seq_puts(m, "\nSpeculationIndirectBranch:\t");
374 switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_INDIRECT_BRANCH)) {
376 seq_puts(m, "unsupported");
378 case PR_SPEC_NOT_AFFECTED:
379 seq_puts(m, "not affected");
381 case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
382 seq_puts(m, "conditional force disabled");
384 case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
385 seq_puts(m, "conditional disabled");
387 case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
388 seq_puts(m, "conditional enabled");
391 seq_puts(m, "always enabled");
393 case PR_SPEC_DISABLE:
394 seq_puts(m, "always disabled");
397 seq_puts(m, "unknown");
403 static inline void task_context_switch_counts(struct seq_file *m,
404 struct task_struct *p)
406 seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
407 seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
411 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
413 seq_printf(m, "Cpus_allowed:\t%*pb\n",
414 cpumask_pr_args(&task->cpus_mask));
415 seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
416 cpumask_pr_args(&task->cpus_mask));
419 static inline void task_core_dumping(struct seq_file *m, struct mm_struct *mm)
421 seq_put_decimal_ull(m, "CoreDumping:\t", !!mm->core_state);
425 static inline void task_thp_status(struct seq_file *m, struct mm_struct *mm)
427 bool thp_enabled = IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE);
430 thp_enabled = !test_bit(MMF_DISABLE_THP, &mm->flags);
431 seq_printf(m, "THP_enabled:\t%d\n", thp_enabled);
434 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
435 struct pid *pid, struct task_struct *task)
437 struct mm_struct *mm = get_task_mm(task);
439 seq_puts(m, "Name:\t");
440 proc_task_name(m, task, true);
443 task_state(m, ns, pid, task);
447 task_core_dumping(m, mm);
448 task_thp_status(m, mm);
453 task_seccomp(m, task);
454 task_cpus_allowed(m, task);
455 cpuset_task_status_allowed(m, task);
456 task_context_switch_counts(m, task);
460 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
461 struct pid *pid, struct task_struct *task, int whole)
463 unsigned long vsize, eip, esp, wchan = 0;
465 int tty_pgrp = -1, tty_nr = 0;
466 sigset_t sigign, sigcatch;
468 pid_t ppid = 0, pgid = -1, sid = -1;
471 struct mm_struct *mm;
472 unsigned long long start_time;
473 unsigned long cmin_flt = 0, cmaj_flt = 0;
474 unsigned long min_flt = 0, maj_flt = 0;
475 u64 cutime, cstime, utime, stime;
477 unsigned long rsslim = 0;
480 state = *get_task_state(task);
481 vsize = eip = esp = 0;
482 permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
483 mm = get_task_mm(task);
485 vsize = task_vsize(mm);
487 * esp and eip are intentionally zeroed out. There is no
488 * non-racy way to read them without freezing the task.
489 * Programs that need reliable values can use ptrace(2).
491 * The only exception is if the task is core dumping because
492 * a program is not able to use ptrace(2) in that case. It is
493 * safe because the task has stopped executing permanently.
495 if (permitted && (task->flags & (PF_EXITING|PF_DUMPCORE))) {
496 if (try_get_task_stack(task)) {
497 eip = KSTK_EIP(task);
498 esp = KSTK_ESP(task);
499 put_task_stack(task);
504 sigemptyset(&sigign);
505 sigemptyset(&sigcatch);
506 cutime = cstime = utime = stime = 0;
509 if (lock_task_sighand(task, &flags)) {
510 struct signal_struct *sig = task->signal;
513 struct pid *pgrp = tty_get_pgrp(sig->tty);
514 tty_pgrp = pid_nr_ns(pgrp, ns);
516 tty_nr = new_encode_dev(tty_devnum(sig->tty));
519 num_threads = get_nr_threads(task);
520 collect_sigign_sigcatch(task, &sigign, &sigcatch);
522 cmin_flt = sig->cmin_flt;
523 cmaj_flt = sig->cmaj_flt;
524 cutime = sig->cutime;
525 cstime = sig->cstime;
526 cgtime = sig->cgtime;
527 rsslim = READ_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
529 /* add up live thread stats at the group level */
531 struct task_struct *t = task;
533 min_flt += t->min_flt;
534 maj_flt += t->maj_flt;
535 gtime += task_gtime(t);
536 } while_each_thread(task, t);
538 min_flt += sig->min_flt;
539 maj_flt += sig->maj_flt;
540 thread_group_cputime_adjusted(task, &utime, &stime);
544 sid = task_session_nr_ns(task, ns);
545 ppid = task_tgid_nr_ns(task->real_parent, ns);
546 pgid = task_pgrp_nr_ns(task, ns);
548 unlock_task_sighand(task, &flags);
551 if (permitted && (!whole || num_threads < 2))
552 wchan = get_wchan(task);
554 min_flt = task->min_flt;
555 maj_flt = task->maj_flt;
556 task_cputime_adjusted(task, &utime, &stime);
557 gtime = task_gtime(task);
560 /* scale priority and nice values from timeslices to -20..20 */
561 /* to make it look like a "normal" Unix priority/nice value */
562 priority = task_prio(task);
563 nice = task_nice(task);
565 /* apply timens offset for boottime and convert nsec -> ticks */
567 nsec_to_clock_t(timens_add_boottime_ns(task->start_boottime));
569 seq_put_decimal_ull(m, "", pid_nr_ns(pid, ns));
571 proc_task_name(m, task, false);
574 seq_put_decimal_ll(m, " ", ppid);
575 seq_put_decimal_ll(m, " ", pgid);
576 seq_put_decimal_ll(m, " ", sid);
577 seq_put_decimal_ll(m, " ", tty_nr);
578 seq_put_decimal_ll(m, " ", tty_pgrp);
579 seq_put_decimal_ull(m, " ", task->flags);
580 seq_put_decimal_ull(m, " ", min_flt);
581 seq_put_decimal_ull(m, " ", cmin_flt);
582 seq_put_decimal_ull(m, " ", maj_flt);
583 seq_put_decimal_ull(m, " ", cmaj_flt);
584 seq_put_decimal_ull(m, " ", nsec_to_clock_t(utime));
585 seq_put_decimal_ull(m, " ", nsec_to_clock_t(stime));
586 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cutime));
587 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cstime));
588 seq_put_decimal_ll(m, " ", priority);
589 seq_put_decimal_ll(m, " ", nice);
590 seq_put_decimal_ll(m, " ", num_threads);
591 seq_put_decimal_ull(m, " ", 0);
592 seq_put_decimal_ull(m, " ", start_time);
593 seq_put_decimal_ull(m, " ", vsize);
594 seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
595 seq_put_decimal_ull(m, " ", rsslim);
596 seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
597 seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
598 seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
599 seq_put_decimal_ull(m, " ", esp);
600 seq_put_decimal_ull(m, " ", eip);
601 /* The signal information here is obsolete.
602 * It must be decimal for Linux 2.0 compatibility.
603 * Use /proc/#/status for real-time signals.
605 seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
606 seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
607 seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
608 seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
611 * We used to output the absolute kernel address, but that's an
612 * information leak - so instead we show a 0/1 flag here, to signal
613 * to user-space whether there's a wchan field in /proc/PID/wchan.
615 * This works with older implementations of procps as well.
622 seq_put_decimal_ull(m, " ", 0);
623 seq_put_decimal_ull(m, " ", 0);
624 seq_put_decimal_ll(m, " ", task->exit_signal);
625 seq_put_decimal_ll(m, " ", task_cpu(task));
626 seq_put_decimal_ull(m, " ", task->rt_priority);
627 seq_put_decimal_ull(m, " ", task->policy);
628 seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
629 seq_put_decimal_ull(m, " ", nsec_to_clock_t(gtime));
630 seq_put_decimal_ll(m, " ", nsec_to_clock_t(cgtime));
632 if (mm && permitted) {
633 seq_put_decimal_ull(m, " ", mm->start_data);
634 seq_put_decimal_ull(m, " ", mm->end_data);
635 seq_put_decimal_ull(m, " ", mm->start_brk);
636 seq_put_decimal_ull(m, " ", mm->arg_start);
637 seq_put_decimal_ull(m, " ", mm->arg_end);
638 seq_put_decimal_ull(m, " ", mm->env_start);
639 seq_put_decimal_ull(m, " ", mm->env_end);
641 seq_puts(m, " 0 0 0 0 0 0 0");
644 seq_put_decimal_ll(m, " ", task->exit_code);
654 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
655 struct pid *pid, struct task_struct *task)
657 return do_task_stat(m, ns, pid, task, 0);
660 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
661 struct pid *pid, struct task_struct *task)
663 return do_task_stat(m, ns, pid, task, 1);
666 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
667 struct pid *pid, struct task_struct *task)
669 struct mm_struct *mm = get_task_mm(task);
673 unsigned long resident = 0;
674 unsigned long shared = 0;
675 unsigned long text = 0;
676 unsigned long data = 0;
678 size = task_statm(mm, &shared, &text, &data, &resident);
682 * For quick read, open code by putting numbers directly
684 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
685 * size, resident, shared, text, data);
687 seq_put_decimal_ull(m, "", size);
688 seq_put_decimal_ull(m, " ", resident);
689 seq_put_decimal_ull(m, " ", shared);
690 seq_put_decimal_ull(m, " ", text);
691 seq_put_decimal_ull(m, " ", 0);
692 seq_put_decimal_ull(m, " ", data);
693 seq_put_decimal_ull(m, " ", 0);
696 seq_write(m, "0 0 0 0 0 0 0\n", 14);
701 #ifdef CONFIG_PROC_CHILDREN
703 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
705 struct task_struct *start, *task;
706 struct pid *pid = NULL;
708 read_lock(&tasklist_lock);
710 start = pid_task(proc_pid(inode), PIDTYPE_PID);
715 * Lets try to continue searching first, this gives
716 * us significant speedup on children-rich processes.
719 task = pid_task(pid_prev, PIDTYPE_PID);
720 if (task && task->real_parent == start &&
721 !(list_empty(&task->sibling))) {
722 if (list_is_last(&task->sibling, &start->children))
724 task = list_first_entry(&task->sibling,
725 struct task_struct, sibling);
726 pid = get_pid(task_pid(task));
734 * We might miss some children here if children
735 * are exited while we were not holding the lock,
736 * but it was never promised to be accurate that
739 * "Just suppose that the parent sleeps, but N children
740 * exit after we printed their tids. Now the slow paths
741 * skips N extra children, we miss N tasks." (c)
743 * So one need to stop or freeze the leader and all
744 * its children to get a precise result.
746 list_for_each_entry(task, &start->children, sibling) {
748 pid = get_pid(task_pid(task));
754 read_unlock(&tasklist_lock);
758 static int children_seq_show(struct seq_file *seq, void *v)
760 struct inode *inode = file_inode(seq->file);
762 seq_printf(seq, "%d ", pid_nr_ns(v, proc_pid_ns(inode->i_sb)));
766 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
768 return get_children_pid(file_inode(seq->file), NULL, *pos);
771 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
775 pid = get_children_pid(file_inode(seq->file), v, *pos + 1);
782 static void children_seq_stop(struct seq_file *seq, void *v)
787 static const struct seq_operations children_seq_ops = {
788 .start = children_seq_start,
789 .next = children_seq_next,
790 .stop = children_seq_stop,
791 .show = children_seq_show,
794 static int children_seq_open(struct inode *inode, struct file *file)
796 return seq_open(file, &children_seq_ops);
799 const struct file_operations proc_tid_children_operations = {
800 .open = children_seq_open,
803 .release = seq_release,
805 #endif /* CONFIG_PROC_CHILDREN */