xref: /openbmc/linux/fs/proc/array.c (revision 174cd4b1)
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/proc_fs.h>
66 #include <linux/ioport.h>
67 #include <linux/uaccess.h>
68 #include <linux/io.h>
69 #include <linux/mm.h>
70 #include <linux/hugetlb.h>
71 #include <linux/pagemap.h>
72 #include <linux/swap.h>
73 #include <linux/smp.h>
74 #include <linux/signal.h>
75 #include <linux/highmem.h>
76 #include <linux/file.h>
77 #include <linux/fdtable.h>
78 #include <linux/times.h>
79 #include <linux/cpuset.h>
80 #include <linux/rcupdate.h>
81 #include <linux/delayacct.h>
82 #include <linux/seq_file.h>
83 #include <linux/pid_namespace.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_putc(m, hex_asc[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 	seq_put_decimal_ull(m, "NoNewPrivs:\t", task_no_new_privs(p));
348 #ifdef CONFIG_SECCOMP
349 	seq_put_decimal_ull(m, "\nSeccomp:\t", p->seccomp.mode);
350 #endif
351 	seq_putc(m, '\n');
352 }
353 
354 static inline void task_context_switch_counts(struct seq_file *m,
355 						struct task_struct *p)
356 {
357 	seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
358 	seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
359 	seq_putc(m, '\n');
360 }
361 
362 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
363 {
364 	seq_printf(m, "Cpus_allowed:\t%*pb\n",
365 		   cpumask_pr_args(&task->cpus_allowed));
366 	seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
367 		   cpumask_pr_args(&task->cpus_allowed));
368 }
369 
370 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
371 			struct pid *pid, struct task_struct *task)
372 {
373 	struct mm_struct *mm = get_task_mm(task);
374 
375 	task_name(m, task);
376 	task_state(m, ns, pid, task);
377 
378 	if (mm) {
379 		task_mem(m, mm);
380 		mmput(mm);
381 	}
382 	task_sig(m, task);
383 	task_cap(m, task);
384 	task_seccomp(m, task);
385 	task_cpus_allowed(m, task);
386 	cpuset_task_status_allowed(m, task);
387 	task_context_switch_counts(m, task);
388 	return 0;
389 }
390 
391 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
392 			struct pid *pid, struct task_struct *task, int whole)
393 {
394 	unsigned long vsize, eip, esp, wchan = 0;
395 	int priority, nice;
396 	int tty_pgrp = -1, tty_nr = 0;
397 	sigset_t sigign, sigcatch;
398 	char state;
399 	pid_t ppid = 0, pgid = -1, sid = -1;
400 	int num_threads = 0;
401 	int permitted;
402 	struct mm_struct *mm;
403 	unsigned long long start_time;
404 	unsigned long cmin_flt = 0, cmaj_flt = 0;
405 	unsigned long  min_flt = 0,  maj_flt = 0;
406 	u64 cutime, cstime, utime, stime;
407 	u64 cgtime, gtime;
408 	unsigned long rsslim = 0;
409 	char tcomm[sizeof(task->comm)];
410 	unsigned long flags;
411 
412 	state = *get_task_state(task);
413 	vsize = eip = esp = 0;
414 	permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
415 	mm = get_task_mm(task);
416 	if (mm) {
417 		vsize = task_vsize(mm);
418 		/*
419 		 * esp and eip are intentionally zeroed out.  There is no
420 		 * non-racy way to read them without freezing the task.
421 		 * Programs that need reliable values can use ptrace(2).
422 		 */
423 	}
424 
425 	get_task_comm(tcomm, task);
426 
427 	sigemptyset(&sigign);
428 	sigemptyset(&sigcatch);
429 	cutime = cstime = utime = stime = 0;
430 	cgtime = gtime = 0;
431 
432 	if (lock_task_sighand(task, &flags)) {
433 		struct signal_struct *sig = task->signal;
434 
435 		if (sig->tty) {
436 			struct pid *pgrp = tty_get_pgrp(sig->tty);
437 			tty_pgrp = pid_nr_ns(pgrp, ns);
438 			put_pid(pgrp);
439 			tty_nr = new_encode_dev(tty_devnum(sig->tty));
440 		}
441 
442 		num_threads = get_nr_threads(task);
443 		collect_sigign_sigcatch(task, &sigign, &sigcatch);
444 
445 		cmin_flt = sig->cmin_flt;
446 		cmaj_flt = sig->cmaj_flt;
447 		cutime = sig->cutime;
448 		cstime = sig->cstime;
449 		cgtime = sig->cgtime;
450 		rsslim = ACCESS_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
451 
452 		/* add up live thread stats at the group level */
453 		if (whole) {
454 			struct task_struct *t = task;
455 			do {
456 				min_flt += t->min_flt;
457 				maj_flt += t->maj_flt;
458 				gtime += task_gtime(t);
459 			} while_each_thread(task, t);
460 
461 			min_flt += sig->min_flt;
462 			maj_flt += sig->maj_flt;
463 			thread_group_cputime_adjusted(task, &utime, &stime);
464 			gtime += sig->gtime;
465 		}
466 
467 		sid = task_session_nr_ns(task, ns);
468 		ppid = task_tgid_nr_ns(task->real_parent, ns);
469 		pgid = task_pgrp_nr_ns(task, ns);
470 
471 		unlock_task_sighand(task, &flags);
472 	}
473 
474 	if (permitted && (!whole || num_threads < 2))
475 		wchan = get_wchan(task);
476 	if (!whole) {
477 		min_flt = task->min_flt;
478 		maj_flt = task->maj_flt;
479 		task_cputime_adjusted(task, &utime, &stime);
480 		gtime = task_gtime(task);
481 	}
482 
483 	/* scale priority and nice values from timeslices to -20..20 */
484 	/* to make it look like a "normal" Unix priority/nice value  */
485 	priority = task_prio(task);
486 	nice = task_nice(task);
487 
488 	/* convert nsec -> ticks */
489 	start_time = nsec_to_clock_t(task->real_start_time);
490 
491 	seq_printf(m, "%d (%s) %c", pid_nr_ns(pid, ns), tcomm, state);
492 	seq_put_decimal_ll(m, " ", ppid);
493 	seq_put_decimal_ll(m, " ", pgid);
494 	seq_put_decimal_ll(m, " ", sid);
495 	seq_put_decimal_ll(m, " ", tty_nr);
496 	seq_put_decimal_ll(m, " ", tty_pgrp);
497 	seq_put_decimal_ull(m, " ", task->flags);
498 	seq_put_decimal_ull(m, " ", min_flt);
499 	seq_put_decimal_ull(m, " ", cmin_flt);
500 	seq_put_decimal_ull(m, " ", maj_flt);
501 	seq_put_decimal_ull(m, " ", cmaj_flt);
502 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(utime));
503 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(stime));
504 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cutime));
505 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cstime));
506 	seq_put_decimal_ll(m, " ", priority);
507 	seq_put_decimal_ll(m, " ", nice);
508 	seq_put_decimal_ll(m, " ", num_threads);
509 	seq_put_decimal_ull(m, " ", 0);
510 	seq_put_decimal_ull(m, " ", start_time);
511 	seq_put_decimal_ull(m, " ", vsize);
512 	seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
513 	seq_put_decimal_ull(m, " ", rsslim);
514 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
515 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
516 	seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
517 	seq_put_decimal_ull(m, " ", esp);
518 	seq_put_decimal_ull(m, " ", eip);
519 	/* The signal information here is obsolete.
520 	 * It must be decimal for Linux 2.0 compatibility.
521 	 * Use /proc/#/status for real-time signals.
522 	 */
523 	seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
524 	seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
525 	seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
526 	seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
527 
528 	/*
529 	 * We used to output the absolute kernel address, but that's an
530 	 * information leak - so instead we show a 0/1 flag here, to signal
531 	 * to user-space whether there's a wchan field in /proc/PID/wchan.
532 	 *
533 	 * This works with older implementations of procps as well.
534 	 */
535 	if (wchan)
536 		seq_puts(m, " 1");
537 	else
538 		seq_puts(m, " 0");
539 
540 	seq_put_decimal_ull(m, " ", 0);
541 	seq_put_decimal_ull(m, " ", 0);
542 	seq_put_decimal_ll(m, " ", task->exit_signal);
543 	seq_put_decimal_ll(m, " ", task_cpu(task));
544 	seq_put_decimal_ull(m, " ", task->rt_priority);
545 	seq_put_decimal_ull(m, " ", task->policy);
546 	seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
547 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(gtime));
548 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cgtime));
549 
550 	if (mm && permitted) {
551 		seq_put_decimal_ull(m, " ", mm->start_data);
552 		seq_put_decimal_ull(m, " ", mm->end_data);
553 		seq_put_decimal_ull(m, " ", mm->start_brk);
554 		seq_put_decimal_ull(m, " ", mm->arg_start);
555 		seq_put_decimal_ull(m, " ", mm->arg_end);
556 		seq_put_decimal_ull(m, " ", mm->env_start);
557 		seq_put_decimal_ull(m, " ", mm->env_end);
558 	} else
559 		seq_puts(m, " 0 0 0 0 0 0 0");
560 
561 	if (permitted)
562 		seq_put_decimal_ll(m, " ", task->exit_code);
563 	else
564 		seq_puts(m, " 0");
565 
566 	seq_putc(m, '\n');
567 	if (mm)
568 		mmput(mm);
569 	return 0;
570 }
571 
572 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
573 			struct pid *pid, struct task_struct *task)
574 {
575 	return do_task_stat(m, ns, pid, task, 0);
576 }
577 
578 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
579 			struct pid *pid, struct task_struct *task)
580 {
581 	return do_task_stat(m, ns, pid, task, 1);
582 }
583 
584 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
585 			struct pid *pid, struct task_struct *task)
586 {
587 	unsigned long size = 0, resident = 0, shared = 0, text = 0, data = 0;
588 	struct mm_struct *mm = get_task_mm(task);
589 
590 	if (mm) {
591 		size = task_statm(mm, &shared, &text, &data, &resident);
592 		mmput(mm);
593 	}
594 	/*
595 	 * For quick read, open code by putting numbers directly
596 	 * expected format is
597 	 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
598 	 *               size, resident, shared, text, data);
599 	 */
600 	seq_put_decimal_ull(m, "", size);
601 	seq_put_decimal_ull(m, " ", resident);
602 	seq_put_decimal_ull(m, " ", shared);
603 	seq_put_decimal_ull(m, " ", text);
604 	seq_put_decimal_ull(m, " ", 0);
605 	seq_put_decimal_ull(m, " ", data);
606 	seq_put_decimal_ull(m, " ", 0);
607 	seq_putc(m, '\n');
608 
609 	return 0;
610 }
611 
612 #ifdef CONFIG_PROC_CHILDREN
613 static struct pid *
614 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
615 {
616 	struct task_struct *start, *task;
617 	struct pid *pid = NULL;
618 
619 	read_lock(&tasklist_lock);
620 
621 	start = pid_task(proc_pid(inode), PIDTYPE_PID);
622 	if (!start)
623 		goto out;
624 
625 	/*
626 	 * Lets try to continue searching first, this gives
627 	 * us significant speedup on children-rich processes.
628 	 */
629 	if (pid_prev) {
630 		task = pid_task(pid_prev, PIDTYPE_PID);
631 		if (task && task->real_parent == start &&
632 		    !(list_empty(&task->sibling))) {
633 			if (list_is_last(&task->sibling, &start->children))
634 				goto out;
635 			task = list_first_entry(&task->sibling,
636 						struct task_struct, sibling);
637 			pid = get_pid(task_pid(task));
638 			goto out;
639 		}
640 	}
641 
642 	/*
643 	 * Slow search case.
644 	 *
645 	 * We might miss some children here if children
646 	 * are exited while we were not holding the lock,
647 	 * but it was never promised to be accurate that
648 	 * much.
649 	 *
650 	 * "Just suppose that the parent sleeps, but N children
651 	 *  exit after we printed their tids. Now the slow paths
652 	 *  skips N extra children, we miss N tasks." (c)
653 	 *
654 	 * So one need to stop or freeze the leader and all
655 	 * its children to get a precise result.
656 	 */
657 	list_for_each_entry(task, &start->children, sibling) {
658 		if (pos-- == 0) {
659 			pid = get_pid(task_pid(task));
660 			break;
661 		}
662 	}
663 
664 out:
665 	read_unlock(&tasklist_lock);
666 	return pid;
667 }
668 
669 static int children_seq_show(struct seq_file *seq, void *v)
670 {
671 	struct inode *inode = seq->private;
672 	pid_t pid;
673 
674 	pid = pid_nr_ns(v, inode->i_sb->s_fs_info);
675 	seq_printf(seq, "%d ", pid);
676 
677 	return 0;
678 }
679 
680 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
681 {
682 	return get_children_pid(seq->private, NULL, *pos);
683 }
684 
685 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
686 {
687 	struct pid *pid;
688 
689 	pid = get_children_pid(seq->private, v, *pos + 1);
690 	put_pid(v);
691 
692 	++*pos;
693 	return pid;
694 }
695 
696 static void children_seq_stop(struct seq_file *seq, void *v)
697 {
698 	put_pid(v);
699 }
700 
701 static const struct seq_operations children_seq_ops = {
702 	.start	= children_seq_start,
703 	.next	= children_seq_next,
704 	.stop	= children_seq_stop,
705 	.show	= children_seq_show,
706 };
707 
708 static int children_seq_open(struct inode *inode, struct file *file)
709 {
710 	struct seq_file *m;
711 	int ret;
712 
713 	ret = seq_open(file, &children_seq_ops);
714 	if (ret)
715 		return ret;
716 
717 	m = file->private_data;
718 	m->private = inode;
719 
720 	return ret;
721 }
722 
723 int children_seq_release(struct inode *inode, struct file *file)
724 {
725 	seq_release(inode, file);
726 	return 0;
727 }
728 
729 const struct file_operations proc_tid_children_operations = {
730 	.open    = children_seq_open,
731 	.read    = seq_read,
732 	.llseek  = seq_lseek,
733 	.release = children_seq_release,
734 };
735 #endif /* CONFIG_PROC_CHILDREN */
736