xref: /openbmc/linux/fs/proc/array.c (revision a10c9ede)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  linux/fs/proc/array.c
4  *
5  *  Copyright (C) 1992  by Linus Torvalds
6  *  based on ideas by Darren Senn
7  *
8  * Fixes:
9  * Michael. K. Johnson: stat,statm extensions.
10  *                      <johnsonm@stolaf.edu>
11  *
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>
17  *
18  * Danny ter Haar    :	added cpuinfo
19  *			<dth@cistron.nl>
20  *
21  * Alessandro Rubini :  profile extension.
22  *                      <rubini@ipvvis.unipv.it>
23  *
24  * Jeff Tranter      :  added BogoMips field to cpuinfo
25  *                      <Jeff_Tranter@Mitel.COM>
26  *
27  * Bruno Haible      :  remove 4K limit for the maps file
28  *			<haible@ma2s2.mathematik.uni-karlsruhe.de>
29  *
30  * Yves Arrouye      :  remove removal of trailing spaces in get_array.
31  *			<Yves.Arrouye@marin.fdn.fr>
32  *
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>
39  *
40  * aeb@cwi.nl        :  /proc/partitions
41  *
42  *
43  * Alan Cox	     :  security fixes.
44  *			<alan@lxorguk.ukuu.org.uk>
45  *
46  * Al Viro           :  safe handling of mm_struct
47  *
48  * Gerhard Wichert   :  added BIGMEM support
49  * Siemens AG           <Gerhard.Wichert@pdb.siemens.de>
50  *
51  * Al Viro & Jeff Garzik :  moved most of the thing into base.c and
52  *			 :  proc_misc.c. The rest may eventually go into
53  *			 :  base.c too.
54  */
55 
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/io.h>
73 #include <linux/mm.h>
74 #include <linux/hugetlb.h>
75 #include <linux/pagemap.h>
76 #include <linux/swap.h>
77 #include <linux/smp.h>
78 #include <linux/signal.h>
79 #include <linux/highmem.h>
80 #include <linux/file.h>
81 #include <linux/fdtable.h>
82 #include <linux/times.h>
83 #include <linux/cpuset.h>
84 #include <linux/rcupdate.h>
85 #include <linux/delayacct.h>
86 #include <linux/seq_file.h>
87 #include <linux/pid_namespace.h>
88 #include <linux/prctl.h>
89 #include <linux/ptrace.h>
90 #include <linux/string_helpers.h>
91 #include <linux/user_namespace.h>
92 #include <linux/fs_struct.h>
93 #include <linux/kthread.h>
94 
95 #include <asm/processor.h>
96 #include "internal.h"
97 
98 void proc_task_name(struct seq_file *m, struct task_struct *p, bool escape)
99 {
100 	char tcomm[64];
101 
102 	if (p->flags & PF_WQ_WORKER)
103 		wq_worker_comm(tcomm, sizeof(tcomm), p);
104 	else if (p->flags & PF_KTHREAD)
105 		get_kthread_comm(tcomm, sizeof(tcomm), p);
106 	else
107 		__get_task_comm(tcomm, sizeof(tcomm), p);
108 
109 	if (escape)
110 		seq_escape_str(m, tcomm, ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
111 	else
112 		seq_printf(m, "%.64s", tcomm);
113 }
114 
115 /*
116  * The task state array is a strange "bitmap" of
117  * reasons to sleep. Thus "running" is zero, and
118  * you can test for combinations of others with
119  * simple bit tests.
120  */
121 static const char * const task_state_array[] = {
122 
123 	/* states in TASK_REPORT: */
124 	"R (running)",		/* 0x00 */
125 	"S (sleeping)",		/* 0x01 */
126 	"D (disk sleep)",	/* 0x02 */
127 	"T (stopped)",		/* 0x04 */
128 	"t (tracing stop)",	/* 0x08 */
129 	"X (dead)",		/* 0x10 */
130 	"Z (zombie)",		/* 0x20 */
131 	"P (parked)",		/* 0x40 */
132 
133 	/* states beyond TASK_REPORT: */
134 	"I (idle)",		/* 0x80 */
135 };
136 
137 static inline const char *get_task_state(struct task_struct *tsk)
138 {
139 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != ARRAY_SIZE(task_state_array));
140 	return task_state_array[task_state_index(tsk)];
141 }
142 
143 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
144 				struct pid *pid, struct task_struct *p)
145 {
146 	struct user_namespace *user_ns = seq_user_ns(m);
147 	struct group_info *group_info;
148 	int g, umask = -1;
149 	struct task_struct *tracer;
150 	const struct cred *cred;
151 	pid_t ppid, tpid = 0, tgid, ngid;
152 	unsigned int max_fds = 0;
153 
154 	rcu_read_lock();
155 	ppid = pid_alive(p) ?
156 		task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
157 
158 	tracer = ptrace_parent(p);
159 	if (tracer)
160 		tpid = task_pid_nr_ns(tracer, ns);
161 
162 	tgid = task_tgid_nr_ns(p, ns);
163 	ngid = task_numa_group_id(p);
164 	cred = get_task_cred(p);
165 
166 	task_lock(p);
167 	if (p->fs)
168 		umask = p->fs->umask;
169 	if (p->files)
170 		max_fds = files_fdtable(p->files)->max_fds;
171 	task_unlock(p);
172 	rcu_read_unlock();
173 
174 	if (umask >= 0)
175 		seq_printf(m, "Umask:\t%#04o\n", umask);
176 	seq_puts(m, "State:\t");
177 	seq_puts(m, get_task_state(p));
178 
179 	seq_put_decimal_ull(m, "\nTgid:\t", tgid);
180 	seq_put_decimal_ull(m, "\nNgid:\t", ngid);
181 	seq_put_decimal_ull(m, "\nPid:\t", pid_nr_ns(pid, ns));
182 	seq_put_decimal_ull(m, "\nPPid:\t", ppid);
183 	seq_put_decimal_ull(m, "\nTracerPid:\t", tpid);
184 	seq_put_decimal_ull(m, "\nUid:\t", from_kuid_munged(user_ns, cred->uid));
185 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->euid));
186 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->suid));
187 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->fsuid));
188 	seq_put_decimal_ull(m, "\nGid:\t", from_kgid_munged(user_ns, cred->gid));
189 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->egid));
190 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->sgid));
191 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->fsgid));
192 	seq_put_decimal_ull(m, "\nFDSize:\t", max_fds);
193 
194 	seq_puts(m, "\nGroups:\t");
195 	group_info = cred->group_info;
196 	for (g = 0; g < group_info->ngroups; g++)
197 		seq_put_decimal_ull(m, g ? " " : "",
198 				from_kgid_munged(user_ns, group_info->gid[g]));
199 	put_cred(cred);
200 	/* Trailing space shouldn't have been added in the first place. */
201 	seq_putc(m, ' ');
202 
203 #ifdef CONFIG_PID_NS
204 	seq_puts(m, "\nNStgid:");
205 	for (g = ns->level; g <= pid->level; g++)
206 		seq_put_decimal_ull(m, "\t", task_tgid_nr_ns(p, pid->numbers[g].ns));
207 	seq_puts(m, "\nNSpid:");
208 	for (g = ns->level; g <= pid->level; g++)
209 		seq_put_decimal_ull(m, "\t", task_pid_nr_ns(p, pid->numbers[g].ns));
210 	seq_puts(m, "\nNSpgid:");
211 	for (g = ns->level; g <= pid->level; g++)
212 		seq_put_decimal_ull(m, "\t", task_pgrp_nr_ns(p, pid->numbers[g].ns));
213 	seq_puts(m, "\nNSsid:");
214 	for (g = ns->level; g <= pid->level; g++)
215 		seq_put_decimal_ull(m, "\t", task_session_nr_ns(p, pid->numbers[g].ns));
216 #endif
217 	seq_putc(m, '\n');
218 }
219 
220 void render_sigset_t(struct seq_file *m, const char *header,
221 				sigset_t *set)
222 {
223 	int i;
224 
225 	seq_puts(m, header);
226 
227 	i = _NSIG;
228 	do {
229 		int x = 0;
230 
231 		i -= 4;
232 		if (sigismember(set, i+1)) x |= 1;
233 		if (sigismember(set, i+2)) x |= 2;
234 		if (sigismember(set, i+3)) x |= 4;
235 		if (sigismember(set, i+4)) x |= 8;
236 		seq_putc(m, hex_asc[x]);
237 	} while (i >= 4);
238 
239 	seq_putc(m, '\n');
240 }
241 
242 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *sigign,
243 				    sigset_t *sigcatch)
244 {
245 	struct k_sigaction *k;
246 	int i;
247 
248 	k = p->sighand->action;
249 	for (i = 1; i <= _NSIG; ++i, ++k) {
250 		if (k->sa.sa_handler == SIG_IGN)
251 			sigaddset(sigign, i);
252 		else if (k->sa.sa_handler != SIG_DFL)
253 			sigaddset(sigcatch, i);
254 	}
255 }
256 
257 static inline void task_sig(struct seq_file *m, struct task_struct *p)
258 {
259 	unsigned long flags;
260 	sigset_t pending, shpending, blocked, ignored, caught;
261 	int num_threads = 0;
262 	unsigned int qsize = 0;
263 	unsigned long qlim = 0;
264 
265 	sigemptyset(&pending);
266 	sigemptyset(&shpending);
267 	sigemptyset(&blocked);
268 	sigemptyset(&ignored);
269 	sigemptyset(&caught);
270 
271 	if (lock_task_sighand(p, &flags)) {
272 		pending = p->pending.signal;
273 		shpending = p->signal->shared_pending.signal;
274 		blocked = p->blocked;
275 		collect_sigign_sigcatch(p, &ignored, &caught);
276 		num_threads = get_nr_threads(p);
277 		rcu_read_lock();  /* FIXME: is this correct? */
278 		qsize = get_ucounts_value(task_ucounts(p), UCOUNT_RLIMIT_SIGPENDING);
279 		rcu_read_unlock();
280 		qlim = task_rlimit(p, RLIMIT_SIGPENDING);
281 		unlock_task_sighand(p, &flags);
282 	}
283 
284 	seq_put_decimal_ull(m, "Threads:\t", num_threads);
285 	seq_put_decimal_ull(m, "\nSigQ:\t", qsize);
286 	seq_put_decimal_ull(m, "/", qlim);
287 
288 	/* render them all */
289 	render_sigset_t(m, "\nSigPnd:\t", &pending);
290 	render_sigset_t(m, "ShdPnd:\t", &shpending);
291 	render_sigset_t(m, "SigBlk:\t", &blocked);
292 	render_sigset_t(m, "SigIgn:\t", &ignored);
293 	render_sigset_t(m, "SigCgt:\t", &caught);
294 }
295 
296 static void render_cap_t(struct seq_file *m, const char *header,
297 			kernel_cap_t *a)
298 {
299 	unsigned __capi;
300 
301 	seq_puts(m, header);
302 	CAP_FOR_EACH_U32(__capi) {
303 		seq_put_hex_ll(m, NULL,
304 			   a->cap[CAP_LAST_U32 - __capi], 8);
305 	}
306 	seq_putc(m, '\n');
307 }
308 
309 static inline void task_cap(struct seq_file *m, struct task_struct *p)
310 {
311 	const struct cred *cred;
312 	kernel_cap_t cap_inheritable, cap_permitted, cap_effective,
313 			cap_bset, cap_ambient;
314 
315 	rcu_read_lock();
316 	cred = __task_cred(p);
317 	cap_inheritable	= cred->cap_inheritable;
318 	cap_permitted	= cred->cap_permitted;
319 	cap_effective	= cred->cap_effective;
320 	cap_bset	= cred->cap_bset;
321 	cap_ambient	= cred->cap_ambient;
322 	rcu_read_unlock();
323 
324 	render_cap_t(m, "CapInh:\t", &cap_inheritable);
325 	render_cap_t(m, "CapPrm:\t", &cap_permitted);
326 	render_cap_t(m, "CapEff:\t", &cap_effective);
327 	render_cap_t(m, "CapBnd:\t", &cap_bset);
328 	render_cap_t(m, "CapAmb:\t", &cap_ambient);
329 }
330 
331 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
332 {
333 	seq_put_decimal_ull(m, "NoNewPrivs:\t", task_no_new_privs(p));
334 #ifdef CONFIG_SECCOMP
335 	seq_put_decimal_ull(m, "\nSeccomp:\t", p->seccomp.mode);
336 #ifdef CONFIG_SECCOMP_FILTER
337 	seq_put_decimal_ull(m, "\nSeccomp_filters:\t",
338 			    atomic_read(&p->seccomp.filter_count));
339 #endif
340 #endif
341 	seq_puts(m, "\nSpeculation_Store_Bypass:\t");
342 	switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_STORE_BYPASS)) {
343 	case -EINVAL:
344 		seq_puts(m, "unknown");
345 		break;
346 	case PR_SPEC_NOT_AFFECTED:
347 		seq_puts(m, "not vulnerable");
348 		break;
349 	case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
350 		seq_puts(m, "thread force mitigated");
351 		break;
352 	case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
353 		seq_puts(m, "thread mitigated");
354 		break;
355 	case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
356 		seq_puts(m, "thread vulnerable");
357 		break;
358 	case PR_SPEC_DISABLE:
359 		seq_puts(m, "globally mitigated");
360 		break;
361 	default:
362 		seq_puts(m, "vulnerable");
363 		break;
364 	}
365 
366 	seq_puts(m, "\nSpeculationIndirectBranch:\t");
367 	switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_INDIRECT_BRANCH)) {
368 	case -EINVAL:
369 		seq_puts(m, "unsupported");
370 		break;
371 	case PR_SPEC_NOT_AFFECTED:
372 		seq_puts(m, "not affected");
373 		break;
374 	case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
375 		seq_puts(m, "conditional force disabled");
376 		break;
377 	case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
378 		seq_puts(m, "conditional disabled");
379 		break;
380 	case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
381 		seq_puts(m, "conditional enabled");
382 		break;
383 	case PR_SPEC_ENABLE:
384 		seq_puts(m, "always enabled");
385 		break;
386 	case PR_SPEC_DISABLE:
387 		seq_puts(m, "always disabled");
388 		break;
389 	default:
390 		seq_puts(m, "unknown");
391 		break;
392 	}
393 	seq_putc(m, '\n');
394 }
395 
396 static inline void task_context_switch_counts(struct seq_file *m,
397 						struct task_struct *p)
398 {
399 	seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
400 	seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
401 	seq_putc(m, '\n');
402 }
403 
404 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
405 {
406 	seq_printf(m, "Cpus_allowed:\t%*pb\n",
407 		   cpumask_pr_args(&task->cpus_mask));
408 	seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
409 		   cpumask_pr_args(&task->cpus_mask));
410 }
411 
412 static inline void task_core_dumping(struct seq_file *m, struct task_struct *task)
413 {
414 	seq_put_decimal_ull(m, "CoreDumping:\t", !!task->signal->core_state);
415 	seq_putc(m, '\n');
416 }
417 
418 static inline void task_thp_status(struct seq_file *m, struct mm_struct *mm)
419 {
420 	bool thp_enabled = IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE);
421 
422 	if (thp_enabled)
423 		thp_enabled = !test_bit(MMF_DISABLE_THP, &mm->flags);
424 	seq_printf(m, "THP_enabled:\t%d\n", thp_enabled);
425 }
426 
427 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
428 			struct pid *pid, struct task_struct *task)
429 {
430 	struct mm_struct *mm = get_task_mm(task);
431 
432 	seq_puts(m, "Name:\t");
433 	proc_task_name(m, task, true);
434 	seq_putc(m, '\n');
435 
436 	task_state(m, ns, pid, task);
437 
438 	if (mm) {
439 		task_mem(m, mm);
440 		task_core_dumping(m, task);
441 		task_thp_status(m, mm);
442 		mmput(mm);
443 	}
444 	task_sig(m, task);
445 	task_cap(m, task);
446 	task_seccomp(m, task);
447 	task_cpus_allowed(m, task);
448 	cpuset_task_status_allowed(m, task);
449 	task_context_switch_counts(m, task);
450 	return 0;
451 }
452 
453 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
454 			struct pid *pid, struct task_struct *task, int whole)
455 {
456 	unsigned long vsize, eip, esp, wchan = 0;
457 	int priority, nice;
458 	int tty_pgrp = -1, tty_nr = 0;
459 	sigset_t sigign, sigcatch;
460 	char state;
461 	pid_t ppid = 0, pgid = -1, sid = -1;
462 	int num_threads = 0;
463 	int permitted;
464 	struct mm_struct *mm;
465 	unsigned long long start_time;
466 	unsigned long cmin_flt = 0, cmaj_flt = 0;
467 	unsigned long  min_flt = 0,  maj_flt = 0;
468 	u64 cutime, cstime, utime, stime;
469 	u64 cgtime, gtime;
470 	unsigned long rsslim = 0;
471 	unsigned long flags;
472 	int exit_code = task->exit_code;
473 
474 	state = *get_task_state(task);
475 	vsize = eip = esp = 0;
476 	permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
477 	mm = get_task_mm(task);
478 	if (mm) {
479 		vsize = task_vsize(mm);
480 		/*
481 		 * esp and eip are intentionally zeroed out.  There is no
482 		 * non-racy way to read them without freezing the task.
483 		 * Programs that need reliable values can use ptrace(2).
484 		 *
485 		 * The only exception is if the task is core dumping because
486 		 * a program is not able to use ptrace(2) in that case. It is
487 		 * safe because the task has stopped executing permanently.
488 		 */
489 		if (permitted && (task->flags & (PF_EXITING|PF_DUMPCORE))) {
490 			if (try_get_task_stack(task)) {
491 				eip = KSTK_EIP(task);
492 				esp = KSTK_ESP(task);
493 				put_task_stack(task);
494 			}
495 		}
496 	}
497 
498 	sigemptyset(&sigign);
499 	sigemptyset(&sigcatch);
500 	cutime = cstime = utime = stime = 0;
501 	cgtime = gtime = 0;
502 
503 	if (lock_task_sighand(task, &flags)) {
504 		struct signal_struct *sig = task->signal;
505 
506 		if (sig->tty) {
507 			struct pid *pgrp = tty_get_pgrp(sig->tty);
508 			tty_pgrp = pid_nr_ns(pgrp, ns);
509 			put_pid(pgrp);
510 			tty_nr = new_encode_dev(tty_devnum(sig->tty));
511 		}
512 
513 		num_threads = get_nr_threads(task);
514 		collect_sigign_sigcatch(task, &sigign, &sigcatch);
515 
516 		cmin_flt = sig->cmin_flt;
517 		cmaj_flt = sig->cmaj_flt;
518 		cutime = sig->cutime;
519 		cstime = sig->cstime;
520 		cgtime = sig->cgtime;
521 		rsslim = READ_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
522 
523 		/* add up live thread stats at the group level */
524 		if (whole) {
525 			struct task_struct *t = task;
526 			do {
527 				min_flt += t->min_flt;
528 				maj_flt += t->maj_flt;
529 				gtime += task_gtime(t);
530 			} while_each_thread(task, t);
531 
532 			min_flt += sig->min_flt;
533 			maj_flt += sig->maj_flt;
534 			thread_group_cputime_adjusted(task, &utime, &stime);
535 			gtime += sig->gtime;
536 
537 			if (sig->flags & (SIGNAL_GROUP_EXIT | SIGNAL_STOP_STOPPED))
538 				exit_code = sig->group_exit_code;
539 		}
540 
541 		sid = task_session_nr_ns(task, ns);
542 		ppid = task_tgid_nr_ns(task->real_parent, ns);
543 		pgid = task_pgrp_nr_ns(task, ns);
544 
545 		unlock_task_sighand(task, &flags);
546 	}
547 
548 	if (permitted && (!whole || num_threads < 2))
549 		wchan = !task_is_running(task);
550 	if (!whole) {
551 		min_flt = task->min_flt;
552 		maj_flt = task->maj_flt;
553 		task_cputime_adjusted(task, &utime, &stime);
554 		gtime = task_gtime(task);
555 	}
556 
557 	/* scale priority and nice values from timeslices to -20..20 */
558 	/* to make it look like a "normal" Unix priority/nice value  */
559 	priority = task_prio(task);
560 	nice = task_nice(task);
561 
562 	/* apply timens offset for boottime and convert nsec -> ticks */
563 	start_time =
564 		nsec_to_clock_t(timens_add_boottime_ns(task->start_boottime));
565 
566 	seq_put_decimal_ull(m, "", pid_nr_ns(pid, ns));
567 	seq_puts(m, " (");
568 	proc_task_name(m, task, false);
569 	seq_puts(m, ") ");
570 	seq_putc(m, state);
571 	seq_put_decimal_ll(m, " ", ppid);
572 	seq_put_decimal_ll(m, " ", pgid);
573 	seq_put_decimal_ll(m, " ", sid);
574 	seq_put_decimal_ll(m, " ", tty_nr);
575 	seq_put_decimal_ll(m, " ", tty_pgrp);
576 	seq_put_decimal_ull(m, " ", task->flags);
577 	seq_put_decimal_ull(m, " ", min_flt);
578 	seq_put_decimal_ull(m, " ", cmin_flt);
579 	seq_put_decimal_ull(m, " ", maj_flt);
580 	seq_put_decimal_ull(m, " ", cmaj_flt);
581 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(utime));
582 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(stime));
583 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cutime));
584 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cstime));
585 	seq_put_decimal_ll(m, " ", priority);
586 	seq_put_decimal_ll(m, " ", nice);
587 	seq_put_decimal_ll(m, " ", num_threads);
588 	seq_put_decimal_ull(m, " ", 0);
589 	seq_put_decimal_ull(m, " ", start_time);
590 	seq_put_decimal_ull(m, " ", vsize);
591 	seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
592 	seq_put_decimal_ull(m, " ", rsslim);
593 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
594 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
595 	seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
596 	seq_put_decimal_ull(m, " ", esp);
597 	seq_put_decimal_ull(m, " ", eip);
598 	/* The signal information here is obsolete.
599 	 * It must be decimal for Linux 2.0 compatibility.
600 	 * Use /proc/#/status for real-time signals.
601 	 */
602 	seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
603 	seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
604 	seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
605 	seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
606 
607 	/*
608 	 * We used to output the absolute kernel address, but that's an
609 	 * information leak - so instead we show a 0/1 flag here, to signal
610 	 * to user-space whether there's a wchan field in /proc/PID/wchan.
611 	 *
612 	 * This works with older implementations of procps as well.
613 	 */
614 	seq_put_decimal_ull(m, " ", wchan);
615 
616 	seq_put_decimal_ull(m, " ", 0);
617 	seq_put_decimal_ull(m, " ", 0);
618 	seq_put_decimal_ll(m, " ", task->exit_signal);
619 	seq_put_decimal_ll(m, " ", task_cpu(task));
620 	seq_put_decimal_ull(m, " ", task->rt_priority);
621 	seq_put_decimal_ull(m, " ", task->policy);
622 	seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
623 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(gtime));
624 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cgtime));
625 
626 	if (mm && permitted) {
627 		seq_put_decimal_ull(m, " ", mm->start_data);
628 		seq_put_decimal_ull(m, " ", mm->end_data);
629 		seq_put_decimal_ull(m, " ", mm->start_brk);
630 		seq_put_decimal_ull(m, " ", mm->arg_start);
631 		seq_put_decimal_ull(m, " ", mm->arg_end);
632 		seq_put_decimal_ull(m, " ", mm->env_start);
633 		seq_put_decimal_ull(m, " ", mm->env_end);
634 	} else
635 		seq_puts(m, " 0 0 0 0 0 0 0");
636 
637 	if (permitted)
638 		seq_put_decimal_ll(m, " ", exit_code);
639 	else
640 		seq_puts(m, " 0");
641 
642 	seq_putc(m, '\n');
643 	if (mm)
644 		mmput(mm);
645 	return 0;
646 }
647 
648 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
649 			struct pid *pid, struct task_struct *task)
650 {
651 	return do_task_stat(m, ns, pid, task, 0);
652 }
653 
654 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
655 			struct pid *pid, struct task_struct *task)
656 {
657 	return do_task_stat(m, ns, pid, task, 1);
658 }
659 
660 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
661 			struct pid *pid, struct task_struct *task)
662 {
663 	struct mm_struct *mm = get_task_mm(task);
664 
665 	if (mm) {
666 		unsigned long size;
667 		unsigned long resident = 0;
668 		unsigned long shared = 0;
669 		unsigned long text = 0;
670 		unsigned long data = 0;
671 
672 		size = task_statm(mm, &shared, &text, &data, &resident);
673 		mmput(mm);
674 
675 		/*
676 		 * For quick read, open code by putting numbers directly
677 		 * expected format is
678 		 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
679 		 *               size, resident, shared, text, data);
680 		 */
681 		seq_put_decimal_ull(m, "", size);
682 		seq_put_decimal_ull(m, " ", resident);
683 		seq_put_decimal_ull(m, " ", shared);
684 		seq_put_decimal_ull(m, " ", text);
685 		seq_put_decimal_ull(m, " ", 0);
686 		seq_put_decimal_ull(m, " ", data);
687 		seq_put_decimal_ull(m, " ", 0);
688 		seq_putc(m, '\n');
689 	} else {
690 		seq_write(m, "0 0 0 0 0 0 0\n", 14);
691 	}
692 	return 0;
693 }
694 
695 #ifdef CONFIG_PROC_CHILDREN
696 static struct pid *
697 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
698 {
699 	struct task_struct *start, *task;
700 	struct pid *pid = NULL;
701 
702 	read_lock(&tasklist_lock);
703 
704 	start = pid_task(proc_pid(inode), PIDTYPE_PID);
705 	if (!start)
706 		goto out;
707 
708 	/*
709 	 * Lets try to continue searching first, this gives
710 	 * us significant speedup on children-rich processes.
711 	 */
712 	if (pid_prev) {
713 		task = pid_task(pid_prev, PIDTYPE_PID);
714 		if (task && task->real_parent == start &&
715 		    !(list_empty(&task->sibling))) {
716 			if (list_is_last(&task->sibling, &start->children))
717 				goto out;
718 			task = list_first_entry(&task->sibling,
719 						struct task_struct, sibling);
720 			pid = get_pid(task_pid(task));
721 			goto out;
722 		}
723 	}
724 
725 	/*
726 	 * Slow search case.
727 	 *
728 	 * We might miss some children here if children
729 	 * are exited while we were not holding the lock,
730 	 * but it was never promised to be accurate that
731 	 * much.
732 	 *
733 	 * "Just suppose that the parent sleeps, but N children
734 	 *  exit after we printed their tids. Now the slow paths
735 	 *  skips N extra children, we miss N tasks." (c)
736 	 *
737 	 * So one need to stop or freeze the leader and all
738 	 * its children to get a precise result.
739 	 */
740 	list_for_each_entry(task, &start->children, sibling) {
741 		if (pos-- == 0) {
742 			pid = get_pid(task_pid(task));
743 			break;
744 		}
745 	}
746 
747 out:
748 	read_unlock(&tasklist_lock);
749 	return pid;
750 }
751 
752 static int children_seq_show(struct seq_file *seq, void *v)
753 {
754 	struct inode *inode = file_inode(seq->file);
755 
756 	seq_printf(seq, "%d ", pid_nr_ns(v, proc_pid_ns(inode->i_sb)));
757 	return 0;
758 }
759 
760 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
761 {
762 	return get_children_pid(file_inode(seq->file), NULL, *pos);
763 }
764 
765 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
766 {
767 	struct pid *pid;
768 
769 	pid = get_children_pid(file_inode(seq->file), v, *pos + 1);
770 	put_pid(v);
771 
772 	++*pos;
773 	return pid;
774 }
775 
776 static void children_seq_stop(struct seq_file *seq, void *v)
777 {
778 	put_pid(v);
779 }
780 
781 static const struct seq_operations children_seq_ops = {
782 	.start	= children_seq_start,
783 	.next	= children_seq_next,
784 	.stop	= children_seq_stop,
785 	.show	= children_seq_show,
786 };
787 
788 static int children_seq_open(struct inode *inode, struct file *file)
789 {
790 	return seq_open(file, &children_seq_ops);
791 }
792 
793 const struct file_operations proc_tid_children_operations = {
794 	.open    = children_seq_open,
795 	.read    = seq_read,
796 	.llseek  = seq_lseek,
797 	.release = seq_release,
798 };
799 #endif /* CONFIG_PROC_CHILDREN */
800