xref: /linux/fs/proc/array.c (revision 74ce1896c6c65b2f8cccbf59162d542988835835)
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.h>
66 #include <linux/sched/cputime.h>
67 #include <linux/proc_fs.h>
68 #include <linux/ioport.h>
69 #include <linux/uaccess.h>
70 #include <linux/io.h>
71 #include <linux/mm.h>
72 #include <linux/hugetlb.h>
73 #include <linux/pagemap.h>
74 #include <linux/swap.h>
75 #include <linux/smp.h>
76 #include <linux/signal.h>
77 #include <linux/highmem.h>
78 #include <linux/file.h>
79 #include <linux/fdtable.h>
80 #include <linux/times.h>
81 #include <linux/cpuset.h>
82 #include <linux/rcupdate.h>
83 #include <linux/delayacct.h>
84 #include <linux/seq_file.h>
85 #include <linux/pid_namespace.h>
86 #include <linux/ptrace.h>
87 #include <linux/tracehook.h>
88 #include <linux/string_helpers.h>
89 #include <linux/user_namespace.h>
90 #include <linux/fs_struct.h>
91 
92 #include <asm/pgtable.h>
93 #include <asm/processor.h>
94 #include "internal.h"
95 
96 static inline void task_name(struct seq_file *m, struct task_struct *p)
97 {
98 	char *buf;
99 	size_t size;
100 	char tcomm[sizeof(p->comm)];
101 	int ret;
102 
103 	get_task_comm(tcomm, p);
104 
105 	seq_puts(m, "Name:\t");
106 
107 	size = seq_get_buf(m, &buf);
108 	ret = string_escape_str(tcomm, buf, size, ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
109 	seq_commit(m, ret < size ? ret : -1);
110 
111 	seq_putc(m, '\n');
112 }
113 
114 /*
115  * The task state array is a strange "bitmap" of
116  * reasons to sleep. Thus "running" is zero, and
117  * you can test for combinations of others with
118  * simple bit tests.
119  */
120 static const char * const task_state_array[] = {
121 	"R (running)",		/*   0 */
122 	"S (sleeping)",		/*   1 */
123 	"D (disk sleep)",	/*   2 */
124 	"T (stopped)",		/*   4 */
125 	"t (tracing stop)",	/*   8 */
126 	"X (dead)",		/*  16 */
127 	"Z (zombie)",		/*  32 */
128 };
129 
130 static inline const char *get_task_state(struct task_struct *tsk)
131 {
132 	unsigned int state = (tsk->state | tsk->exit_state) & TASK_REPORT;
133 
134 	/*
135 	 * Parked tasks do not run; they sit in __kthread_parkme().
136 	 * Without this check, we would report them as running, which is
137 	 * clearly wrong, so we report them as sleeping instead.
138 	 */
139 	if (tsk->state == TASK_PARKED)
140 		state = TASK_INTERRUPTIBLE;
141 
142 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT) != ARRAY_SIZE(task_state_array)-1);
143 
144 	return task_state_array[fls(state)];
145 }
146 
147 static inline int get_task_umask(struct task_struct *tsk)
148 {
149 	struct fs_struct *fs;
150 	int umask = -ENOENT;
151 
152 	task_lock(tsk);
153 	fs = tsk->fs;
154 	if (fs)
155 		umask = fs->umask;
156 	task_unlock(tsk);
157 	return umask;
158 }
159 
160 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
161 				struct pid *pid, struct task_struct *p)
162 {
163 	struct user_namespace *user_ns = seq_user_ns(m);
164 	struct group_info *group_info;
165 	int g, umask;
166 	struct task_struct *tracer;
167 	const struct cred *cred;
168 	pid_t ppid, tpid = 0, tgid, ngid;
169 	unsigned int max_fds = 0;
170 
171 	rcu_read_lock();
172 	ppid = pid_alive(p) ?
173 		task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
174 
175 	tracer = ptrace_parent(p);
176 	if (tracer)
177 		tpid = task_pid_nr_ns(tracer, ns);
178 
179 	tgid = task_tgid_nr_ns(p, ns);
180 	ngid = task_numa_group_id(p);
181 	cred = get_task_cred(p);
182 
183 	umask = get_task_umask(p);
184 	if (umask >= 0)
185 		seq_printf(m, "Umask:\t%#04o\n", umask);
186 
187 	task_lock(p);
188 	if (p->files)
189 		max_fds = files_fdtable(p->files)->max_fds;
190 	task_unlock(p);
191 	rcu_read_unlock();
192 
193 	seq_printf(m, "State:\t%s", get_task_state(p));
194 
195 	seq_put_decimal_ull(m, "\nTgid:\t", tgid);
196 	seq_put_decimal_ull(m, "\nNgid:\t", ngid);
197 	seq_put_decimal_ull(m, "\nPid:\t", pid_nr_ns(pid, ns));
198 	seq_put_decimal_ull(m, "\nPPid:\t", ppid);
199 	seq_put_decimal_ull(m, "\nTracerPid:\t", tpid);
200 	seq_put_decimal_ull(m, "\nUid:\t", from_kuid_munged(user_ns, cred->uid));
201 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->euid));
202 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->suid));
203 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->fsuid));
204 	seq_put_decimal_ull(m, "\nGid:\t", from_kgid_munged(user_ns, cred->gid));
205 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->egid));
206 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->sgid));
207 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->fsgid));
208 	seq_put_decimal_ull(m, "\nFDSize:\t", max_fds);
209 
210 	seq_puts(m, "\nGroups:\t");
211 	group_info = cred->group_info;
212 	for (g = 0; g < group_info->ngroups; g++)
213 		seq_put_decimal_ull(m, g ? " " : "",
214 				from_kgid_munged(user_ns, group_info->gid[g]));
215 	put_cred(cred);
216 	/* Trailing space shouldn't have been added in the first place. */
217 	seq_putc(m, ' ');
218 
219 #ifdef CONFIG_PID_NS
220 	seq_puts(m, "\nNStgid:");
221 	for (g = ns->level; g <= pid->level; g++)
222 		seq_put_decimal_ull(m, "\t", task_tgid_nr_ns(p, pid->numbers[g].ns));
223 	seq_puts(m, "\nNSpid:");
224 	for (g = ns->level; g <= pid->level; g++)
225 		seq_put_decimal_ull(m, "\t", task_pid_nr_ns(p, pid->numbers[g].ns));
226 	seq_puts(m, "\nNSpgid:");
227 	for (g = ns->level; g <= pid->level; g++)
228 		seq_put_decimal_ull(m, "\t", task_pgrp_nr_ns(p, pid->numbers[g].ns));
229 	seq_puts(m, "\nNSsid:");
230 	for (g = ns->level; g <= pid->level; g++)
231 		seq_put_decimal_ull(m, "\t", task_session_nr_ns(p, pid->numbers[g].ns));
232 #endif
233 	seq_putc(m, '\n');
234 }
235 
236 void render_sigset_t(struct seq_file *m, const char *header,
237 				sigset_t *set)
238 {
239 	int i;
240 
241 	seq_puts(m, header);
242 
243 	i = _NSIG;
244 	do {
245 		int x = 0;
246 
247 		i -= 4;
248 		if (sigismember(set, i+1)) x |= 1;
249 		if (sigismember(set, i+2)) x |= 2;
250 		if (sigismember(set, i+3)) x |= 4;
251 		if (sigismember(set, i+4)) x |= 8;
252 		seq_putc(m, hex_asc[x]);
253 	} while (i >= 4);
254 
255 	seq_putc(m, '\n');
256 }
257 
258 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
259 				    sigset_t *catch)
260 {
261 	struct k_sigaction *k;
262 	int i;
263 
264 	k = p->sighand->action;
265 	for (i = 1; i <= _NSIG; ++i, ++k) {
266 		if (k->sa.sa_handler == SIG_IGN)
267 			sigaddset(ign, i);
268 		else if (k->sa.sa_handler != SIG_DFL)
269 			sigaddset(catch, i);
270 	}
271 }
272 
273 static inline void task_sig(struct seq_file *m, struct task_struct *p)
274 {
275 	unsigned long flags;
276 	sigset_t pending, shpending, blocked, ignored, caught;
277 	int num_threads = 0;
278 	unsigned long qsize = 0;
279 	unsigned long qlim = 0;
280 
281 	sigemptyset(&pending);
282 	sigemptyset(&shpending);
283 	sigemptyset(&blocked);
284 	sigemptyset(&ignored);
285 	sigemptyset(&caught);
286 
287 	if (lock_task_sighand(p, &flags)) {
288 		pending = p->pending.signal;
289 		shpending = p->signal->shared_pending.signal;
290 		blocked = p->blocked;
291 		collect_sigign_sigcatch(p, &ignored, &caught);
292 		num_threads = get_nr_threads(p);
293 		rcu_read_lock();  /* FIXME: is this correct? */
294 		qsize = atomic_read(&__task_cred(p)->user->sigpending);
295 		rcu_read_unlock();
296 		qlim = task_rlimit(p, RLIMIT_SIGPENDING);
297 		unlock_task_sighand(p, &flags);
298 	}
299 
300 	seq_put_decimal_ull(m, "Threads:\t", num_threads);
301 	seq_put_decimal_ull(m, "\nSigQ:\t", qsize);
302 	seq_put_decimal_ull(m, "/", qlim);
303 
304 	/* render them all */
305 	render_sigset_t(m, "\nSigPnd:\t", &pending);
306 	render_sigset_t(m, "ShdPnd:\t", &shpending);
307 	render_sigset_t(m, "SigBlk:\t", &blocked);
308 	render_sigset_t(m, "SigIgn:\t", &ignored);
309 	render_sigset_t(m, "SigCgt:\t", &caught);
310 }
311 
312 static void render_cap_t(struct seq_file *m, const char *header,
313 			kernel_cap_t *a)
314 {
315 	unsigned __capi;
316 
317 	seq_puts(m, header);
318 	CAP_FOR_EACH_U32(__capi) {
319 		seq_printf(m, "%08x",
320 			   a->cap[CAP_LAST_U32 - __capi]);
321 	}
322 	seq_putc(m, '\n');
323 }
324 
325 static inline void task_cap(struct seq_file *m, struct task_struct *p)
326 {
327 	const struct cred *cred;
328 	kernel_cap_t cap_inheritable, cap_permitted, cap_effective,
329 			cap_bset, cap_ambient;
330 
331 	rcu_read_lock();
332 	cred = __task_cred(p);
333 	cap_inheritable	= cred->cap_inheritable;
334 	cap_permitted	= cred->cap_permitted;
335 	cap_effective	= cred->cap_effective;
336 	cap_bset	= cred->cap_bset;
337 	cap_ambient	= cred->cap_ambient;
338 	rcu_read_unlock();
339 
340 	render_cap_t(m, "CapInh:\t", &cap_inheritable);
341 	render_cap_t(m, "CapPrm:\t", &cap_permitted);
342 	render_cap_t(m, "CapEff:\t", &cap_effective);
343 	render_cap_t(m, "CapBnd:\t", &cap_bset);
344 	render_cap_t(m, "CapAmb:\t", &cap_ambient);
345 }
346 
347 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
348 {
349 	seq_put_decimal_ull(m, "NoNewPrivs:\t", task_no_new_privs(p));
350 #ifdef CONFIG_SECCOMP
351 	seq_put_decimal_ull(m, "\nSeccomp:\t", p->seccomp.mode);
352 #endif
353 	seq_putc(m, '\n');
354 }
355 
356 static inline void task_context_switch_counts(struct seq_file *m,
357 						struct task_struct *p)
358 {
359 	seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
360 	seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
361 	seq_putc(m, '\n');
362 }
363 
364 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
365 {
366 	seq_printf(m, "Cpus_allowed:\t%*pb\n",
367 		   cpumask_pr_args(&task->cpus_allowed));
368 	seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
369 		   cpumask_pr_args(&task->cpus_allowed));
370 }
371 
372 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
373 			struct pid *pid, struct task_struct *task)
374 {
375 	struct mm_struct *mm = get_task_mm(task);
376 
377 	task_name(m, task);
378 	task_state(m, ns, pid, task);
379 
380 	if (mm) {
381 		task_mem(m, mm);
382 		mmput(mm);
383 	}
384 	task_sig(m, task);
385 	task_cap(m, task);
386 	task_seccomp(m, task);
387 	task_cpus_allowed(m, task);
388 	cpuset_task_status_allowed(m, task);
389 	task_context_switch_counts(m, task);
390 	return 0;
391 }
392 
393 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
394 			struct pid *pid, struct task_struct *task, int whole)
395 {
396 	unsigned long vsize, eip, esp, wchan = 0;
397 	int priority, nice;
398 	int tty_pgrp = -1, tty_nr = 0;
399 	sigset_t sigign, sigcatch;
400 	char state;
401 	pid_t ppid = 0, pgid = -1, sid = -1;
402 	int num_threads = 0;
403 	int permitted;
404 	struct mm_struct *mm;
405 	unsigned long long start_time;
406 	unsigned long cmin_flt = 0, cmaj_flt = 0;
407 	unsigned long  min_flt = 0,  maj_flt = 0;
408 	u64 cutime, cstime, utime, stime;
409 	u64 cgtime, gtime;
410 	unsigned long rsslim = 0;
411 	char tcomm[sizeof(task->comm)];
412 	unsigned long flags;
413 
414 	state = *get_task_state(task);
415 	vsize = eip = esp = 0;
416 	permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
417 	mm = get_task_mm(task);
418 	if (mm) {
419 		vsize = task_vsize(mm);
420 		/*
421 		 * esp and eip are intentionally zeroed out.  There is no
422 		 * non-racy way to read them without freezing the task.
423 		 * Programs that need reliable values can use ptrace(2).
424 		 */
425 	}
426 
427 	get_task_comm(tcomm, task);
428 
429 	sigemptyset(&sigign);
430 	sigemptyset(&sigcatch);
431 	cutime = cstime = utime = stime = 0;
432 	cgtime = gtime = 0;
433 
434 	if (lock_task_sighand(task, &flags)) {
435 		struct signal_struct *sig = task->signal;
436 
437 		if (sig->tty) {
438 			struct pid *pgrp = tty_get_pgrp(sig->tty);
439 			tty_pgrp = pid_nr_ns(pgrp, ns);
440 			put_pid(pgrp);
441 			tty_nr = new_encode_dev(tty_devnum(sig->tty));
442 		}
443 
444 		num_threads = get_nr_threads(task);
445 		collect_sigign_sigcatch(task, &sigign, &sigcatch);
446 
447 		cmin_flt = sig->cmin_flt;
448 		cmaj_flt = sig->cmaj_flt;
449 		cutime = sig->cutime;
450 		cstime = sig->cstime;
451 		cgtime = sig->cgtime;
452 		rsslim = ACCESS_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
453 
454 		/* add up live thread stats at the group level */
455 		if (whole) {
456 			struct task_struct *t = task;
457 			do {
458 				min_flt += t->min_flt;
459 				maj_flt += t->maj_flt;
460 				gtime += task_gtime(t);
461 			} while_each_thread(task, t);
462 
463 			min_flt += sig->min_flt;
464 			maj_flt += sig->maj_flt;
465 			thread_group_cputime_adjusted(task, &utime, &stime);
466 			gtime += sig->gtime;
467 		}
468 
469 		sid = task_session_nr_ns(task, ns);
470 		ppid = task_tgid_nr_ns(task->real_parent, ns);
471 		pgid = task_pgrp_nr_ns(task, ns);
472 
473 		unlock_task_sighand(task, &flags);
474 	}
475 
476 	if (permitted && (!whole || num_threads < 2))
477 		wchan = get_wchan(task);
478 	if (!whole) {
479 		min_flt = task->min_flt;
480 		maj_flt = task->maj_flt;
481 		task_cputime_adjusted(task, &utime, &stime);
482 		gtime = task_gtime(task);
483 	}
484 
485 	/* scale priority and nice values from timeslices to -20..20 */
486 	/* to make it look like a "normal" Unix priority/nice value  */
487 	priority = task_prio(task);
488 	nice = task_nice(task);
489 
490 	/* convert nsec -> ticks */
491 	start_time = nsec_to_clock_t(task->real_start_time);
492 
493 	seq_printf(m, "%d (%s) %c", pid_nr_ns(pid, ns), tcomm, state);
494 	seq_put_decimal_ll(m, " ", ppid);
495 	seq_put_decimal_ll(m, " ", pgid);
496 	seq_put_decimal_ll(m, " ", sid);
497 	seq_put_decimal_ll(m, " ", tty_nr);
498 	seq_put_decimal_ll(m, " ", tty_pgrp);
499 	seq_put_decimal_ull(m, " ", task->flags);
500 	seq_put_decimal_ull(m, " ", min_flt);
501 	seq_put_decimal_ull(m, " ", cmin_flt);
502 	seq_put_decimal_ull(m, " ", maj_flt);
503 	seq_put_decimal_ull(m, " ", cmaj_flt);
504 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(utime));
505 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(stime));
506 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cutime));
507 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cstime));
508 	seq_put_decimal_ll(m, " ", priority);
509 	seq_put_decimal_ll(m, " ", nice);
510 	seq_put_decimal_ll(m, " ", num_threads);
511 	seq_put_decimal_ull(m, " ", 0);
512 	seq_put_decimal_ull(m, " ", start_time);
513 	seq_put_decimal_ull(m, " ", vsize);
514 	seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
515 	seq_put_decimal_ull(m, " ", rsslim);
516 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
517 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
518 	seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
519 	seq_put_decimal_ull(m, " ", esp);
520 	seq_put_decimal_ull(m, " ", eip);
521 	/* The signal information here is obsolete.
522 	 * It must be decimal for Linux 2.0 compatibility.
523 	 * Use /proc/#/status for real-time signals.
524 	 */
525 	seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
526 	seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
527 	seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
528 	seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
529 
530 	/*
531 	 * We used to output the absolute kernel address, but that's an
532 	 * information leak - so instead we show a 0/1 flag here, to signal
533 	 * to user-space whether there's a wchan field in /proc/PID/wchan.
534 	 *
535 	 * This works with older implementations of procps as well.
536 	 */
537 	if (wchan)
538 		seq_puts(m, " 1");
539 	else
540 		seq_puts(m, " 0");
541 
542 	seq_put_decimal_ull(m, " ", 0);
543 	seq_put_decimal_ull(m, " ", 0);
544 	seq_put_decimal_ll(m, " ", task->exit_signal);
545 	seq_put_decimal_ll(m, " ", task_cpu(task));
546 	seq_put_decimal_ull(m, " ", task->rt_priority);
547 	seq_put_decimal_ull(m, " ", task->policy);
548 	seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
549 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(gtime));
550 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cgtime));
551 
552 	if (mm && permitted) {
553 		seq_put_decimal_ull(m, " ", mm->start_data);
554 		seq_put_decimal_ull(m, " ", mm->end_data);
555 		seq_put_decimal_ull(m, " ", mm->start_brk);
556 		seq_put_decimal_ull(m, " ", mm->arg_start);
557 		seq_put_decimal_ull(m, " ", mm->arg_end);
558 		seq_put_decimal_ull(m, " ", mm->env_start);
559 		seq_put_decimal_ull(m, " ", mm->env_end);
560 	} else
561 		seq_puts(m, " 0 0 0 0 0 0 0");
562 
563 	if (permitted)
564 		seq_put_decimal_ll(m, " ", task->exit_code);
565 	else
566 		seq_puts(m, " 0");
567 
568 	seq_putc(m, '\n');
569 	if (mm)
570 		mmput(mm);
571 	return 0;
572 }
573 
574 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
575 			struct pid *pid, struct task_struct *task)
576 {
577 	return do_task_stat(m, ns, pid, task, 0);
578 }
579 
580 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
581 			struct pid *pid, struct task_struct *task)
582 {
583 	return do_task_stat(m, ns, pid, task, 1);
584 }
585 
586 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
587 			struct pid *pid, struct task_struct *task)
588 {
589 	unsigned long size = 0, resident = 0, shared = 0, text = 0, data = 0;
590 	struct mm_struct *mm = get_task_mm(task);
591 
592 	if (mm) {
593 		size = task_statm(mm, &shared, &text, &data, &resident);
594 		mmput(mm);
595 	}
596 	/*
597 	 * For quick read, open code by putting numbers directly
598 	 * expected format is
599 	 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
600 	 *               size, resident, shared, text, data);
601 	 */
602 	seq_put_decimal_ull(m, "", size);
603 	seq_put_decimal_ull(m, " ", resident);
604 	seq_put_decimal_ull(m, " ", shared);
605 	seq_put_decimal_ull(m, " ", text);
606 	seq_put_decimal_ull(m, " ", 0);
607 	seq_put_decimal_ull(m, " ", data);
608 	seq_put_decimal_ull(m, " ", 0);
609 	seq_putc(m, '\n');
610 
611 	return 0;
612 }
613 
614 #ifdef CONFIG_PROC_CHILDREN
615 static struct pid *
616 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
617 {
618 	struct task_struct *start, *task;
619 	struct pid *pid = NULL;
620 
621 	read_lock(&tasklist_lock);
622 
623 	start = pid_task(proc_pid(inode), PIDTYPE_PID);
624 	if (!start)
625 		goto out;
626 
627 	/*
628 	 * Lets try to continue searching first, this gives
629 	 * us significant speedup on children-rich processes.
630 	 */
631 	if (pid_prev) {
632 		task = pid_task(pid_prev, PIDTYPE_PID);
633 		if (task && task->real_parent == start &&
634 		    !(list_empty(&task->sibling))) {
635 			if (list_is_last(&task->sibling, &start->children))
636 				goto out;
637 			task = list_first_entry(&task->sibling,
638 						struct task_struct, sibling);
639 			pid = get_pid(task_pid(task));
640 			goto out;
641 		}
642 	}
643 
644 	/*
645 	 * Slow search case.
646 	 *
647 	 * We might miss some children here if children
648 	 * are exited while we were not holding the lock,
649 	 * but it was never promised to be accurate that
650 	 * much.
651 	 *
652 	 * "Just suppose that the parent sleeps, but N children
653 	 *  exit after we printed their tids. Now the slow paths
654 	 *  skips N extra children, we miss N tasks." (c)
655 	 *
656 	 * So one need to stop or freeze the leader and all
657 	 * its children to get a precise result.
658 	 */
659 	list_for_each_entry(task, &start->children, sibling) {
660 		if (pos-- == 0) {
661 			pid = get_pid(task_pid(task));
662 			break;
663 		}
664 	}
665 
666 out:
667 	read_unlock(&tasklist_lock);
668 	return pid;
669 }
670 
671 static int children_seq_show(struct seq_file *seq, void *v)
672 {
673 	struct inode *inode = seq->private;
674 	pid_t pid;
675 
676 	pid = pid_nr_ns(v, inode->i_sb->s_fs_info);
677 	seq_printf(seq, "%d ", pid);
678 
679 	return 0;
680 }
681 
682 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
683 {
684 	return get_children_pid(seq->private, NULL, *pos);
685 }
686 
687 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
688 {
689 	struct pid *pid;
690 
691 	pid = get_children_pid(seq->private, v, *pos + 1);
692 	put_pid(v);
693 
694 	++*pos;
695 	return pid;
696 }
697 
698 static void children_seq_stop(struct seq_file *seq, void *v)
699 {
700 	put_pid(v);
701 }
702 
703 static const struct seq_operations children_seq_ops = {
704 	.start	= children_seq_start,
705 	.next	= children_seq_next,
706 	.stop	= children_seq_stop,
707 	.show	= children_seq_show,
708 };
709 
710 static int children_seq_open(struct inode *inode, struct file *file)
711 {
712 	struct seq_file *m;
713 	int ret;
714 
715 	ret = seq_open(file, &children_seq_ops);
716 	if (ret)
717 		return ret;
718 
719 	m = file->private_data;
720 	m->private = inode;
721 
722 	return ret;
723 }
724 
725 int children_seq_release(struct inode *inode, struct file *file)
726 {
727 	seq_release(inode, file);
728 	return 0;
729 }
730 
731 const struct file_operations proc_tid_children_operations = {
732 	.open    = children_seq_open,
733 	.read    = seq_read,
734 	.llseek  = seq_lseek,
735 	.release = children_seq_release,
736 };
737 #endif /* CONFIG_PROC_CHILDREN */
738