xref: /openbmc/linux/fs/proc/array.c (revision 13fe7056bebb4015c6231a07a1be4d3aebbfe979)
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/kernel.h>
60 #include <linux/kernel_stat.h>
61 #include <linux/tty.h>
62 #include <linux/string.h>
63 #include <linux/mman.h>
64 #include <linux/sched/mm.h>
65 #include <linux/sched/numa_balancing.h>
66 #include <linux/sched/task_stack.h>
67 #include <linux/sched/task.h>
68 #include <linux/sched/cputime.h>
69 #include <linux/proc_fs.h>
70 #include <linux/ioport.h>
71 #include <linux/uaccess.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/tracehook.h>
91 #include <linux/string_helpers.h>
92 #include <linux/user_namespace.h>
93 #include <linux/fs_struct.h>
94 
95 #include <asm/pgtable.h>
96 #include <asm/processor.h>
97 #include "internal.h"
98 
99 static inline void task_name(struct seq_file *m, struct task_struct *p)
100 {
101 	char *buf;
102 	size_t size;
103 	char tcomm[sizeof(p->comm)];
104 	int ret;
105 
106 	get_task_comm(tcomm, p);
107 
108 	seq_puts(m, "Name:\t");
109 
110 	size = seq_get_buf(m, &buf);
111 	ret = string_escape_str(tcomm, buf, size, ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
112 	seq_commit(m, ret < size ? ret : -1);
113 
114 	seq_putc(m, '\n');
115 }
116 
117 /*
118  * The task state array is a strange "bitmap" of
119  * reasons to sleep. Thus "running" is zero, and
120  * you can test for combinations of others with
121  * simple bit tests.
122  */
123 static const char * const task_state_array[] = {
124 
125 	/* states in TASK_REPORT: */
126 	"R (running)",		/* 0x00 */
127 	"S (sleeping)",		/* 0x01 */
128 	"D (disk sleep)",	/* 0x02 */
129 	"T (stopped)",		/* 0x04 */
130 	"t (tracing stop)",	/* 0x08 */
131 	"X (dead)",		/* 0x10 */
132 	"Z (zombie)",		/* 0x20 */
133 	"P (parked)",		/* 0x40 */
134 
135 	/* states beyond TASK_REPORT: */
136 	"I (idle)",		/* 0x80 */
137 };
138 
139 static inline const char *get_task_state(struct task_struct *tsk)
140 {
141 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != ARRAY_SIZE(task_state_array));
142 	return task_state_array[task_state_index(tsk)];
143 }
144 
145 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
146 				struct pid *pid, struct task_struct *p)
147 {
148 	struct user_namespace *user_ns = seq_user_ns(m);
149 	struct group_info *group_info;
150 	int g, umask = -1;
151 	struct task_struct *tracer;
152 	const struct cred *cred;
153 	pid_t ppid, tpid = 0, tgid, ngid;
154 	unsigned int max_fds = 0;
155 
156 	rcu_read_lock();
157 	ppid = pid_alive(p) ?
158 		task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
159 
160 	tracer = ptrace_parent(p);
161 	if (tracer)
162 		tpid = task_pid_nr_ns(tracer, ns);
163 
164 	tgid = task_tgid_nr_ns(p, ns);
165 	ngid = task_numa_group_id(p);
166 	cred = get_task_cred(p);
167 
168 	task_lock(p);
169 	if (p->fs)
170 		umask = p->fs->umask;
171 	if (p->files)
172 		max_fds = files_fdtable(p->files)->max_fds;
173 	task_unlock(p);
174 	rcu_read_unlock();
175 
176 	if (umask >= 0)
177 		seq_printf(m, "Umask:\t%#04o\n", umask);
178 	seq_puts(m, "State:\t");
179 	seq_puts(m, get_task_state(p));
180 
181 	seq_put_decimal_ull(m, "\nTgid:\t", tgid);
182 	seq_put_decimal_ull(m, "\nNgid:\t", ngid);
183 	seq_put_decimal_ull(m, "\nPid:\t", pid_nr_ns(pid, ns));
184 	seq_put_decimal_ull(m, "\nPPid:\t", ppid);
185 	seq_put_decimal_ull(m, "\nTracerPid:\t", tpid);
186 	seq_put_decimal_ull(m, "\nUid:\t", from_kuid_munged(user_ns, cred->uid));
187 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->euid));
188 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->suid));
189 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->fsuid));
190 	seq_put_decimal_ull(m, "\nGid:\t", from_kgid_munged(user_ns, cred->gid));
191 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->egid));
192 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->sgid));
193 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->fsgid));
194 	seq_put_decimal_ull(m, "\nFDSize:\t", max_fds);
195 
196 	seq_puts(m, "\nGroups:\t");
197 	group_info = cred->group_info;
198 	for (g = 0; g < group_info->ngroups; g++)
199 		seq_put_decimal_ull(m, g ? " " : "",
200 				from_kgid_munged(user_ns, group_info->gid[g]));
201 	put_cred(cred);
202 	/* Trailing space shouldn't have been added in the first place. */
203 	seq_putc(m, ' ');
204 
205 #ifdef CONFIG_PID_NS
206 	seq_puts(m, "\nNStgid:");
207 	for (g = ns->level; g <= pid->level; g++)
208 		seq_put_decimal_ull(m, "\t", task_tgid_nr_ns(p, pid->numbers[g].ns));
209 	seq_puts(m, "\nNSpid:");
210 	for (g = ns->level; g <= pid->level; g++)
211 		seq_put_decimal_ull(m, "\t", task_pid_nr_ns(p, pid->numbers[g].ns));
212 	seq_puts(m, "\nNSpgid:");
213 	for (g = ns->level; g <= pid->level; g++)
214 		seq_put_decimal_ull(m, "\t", task_pgrp_nr_ns(p, pid->numbers[g].ns));
215 	seq_puts(m, "\nNSsid:");
216 	for (g = ns->level; g <= pid->level; g++)
217 		seq_put_decimal_ull(m, "\t", task_session_nr_ns(p, pid->numbers[g].ns));
218 #endif
219 	seq_putc(m, '\n');
220 }
221 
222 void render_sigset_t(struct seq_file *m, const char *header,
223 				sigset_t *set)
224 {
225 	int i;
226 
227 	seq_puts(m, header);
228 
229 	i = _NSIG;
230 	do {
231 		int x = 0;
232 
233 		i -= 4;
234 		if (sigismember(set, i+1)) x |= 1;
235 		if (sigismember(set, i+2)) x |= 2;
236 		if (sigismember(set, i+3)) x |= 4;
237 		if (sigismember(set, i+4)) x |= 8;
238 		seq_putc(m, hex_asc[x]);
239 	} while (i >= 4);
240 
241 	seq_putc(m, '\n');
242 }
243 
244 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *ign,
245 				    sigset_t *catch)
246 {
247 	struct k_sigaction *k;
248 	int i;
249 
250 	k = p->sighand->action;
251 	for (i = 1; i <= _NSIG; ++i, ++k) {
252 		if (k->sa.sa_handler == SIG_IGN)
253 			sigaddset(ign, i);
254 		else if (k->sa.sa_handler != SIG_DFL)
255 			sigaddset(catch, i);
256 	}
257 }
258 
259 static inline void task_sig(struct seq_file *m, struct task_struct *p)
260 {
261 	unsigned long flags;
262 	sigset_t pending, shpending, blocked, ignored, caught;
263 	int num_threads = 0;
264 	unsigned long qsize = 0;
265 	unsigned long qlim = 0;
266 
267 	sigemptyset(&pending);
268 	sigemptyset(&shpending);
269 	sigemptyset(&blocked);
270 	sigemptyset(&ignored);
271 	sigemptyset(&caught);
272 
273 	if (lock_task_sighand(p, &flags)) {
274 		pending = p->pending.signal;
275 		shpending = p->signal->shared_pending.signal;
276 		blocked = p->blocked;
277 		collect_sigign_sigcatch(p, &ignored, &caught);
278 		num_threads = get_nr_threads(p);
279 		rcu_read_lock();  /* FIXME: is this correct? */
280 		qsize = atomic_read(&__task_cred(p)->user->sigpending);
281 		rcu_read_unlock();
282 		qlim = task_rlimit(p, RLIMIT_SIGPENDING);
283 		unlock_task_sighand(p, &flags);
284 	}
285 
286 	seq_put_decimal_ull(m, "Threads:\t", num_threads);
287 	seq_put_decimal_ull(m, "\nSigQ:\t", qsize);
288 	seq_put_decimal_ull(m, "/", qlim);
289 
290 	/* render them all */
291 	render_sigset_t(m, "\nSigPnd:\t", &pending);
292 	render_sigset_t(m, "ShdPnd:\t", &shpending);
293 	render_sigset_t(m, "SigBlk:\t", &blocked);
294 	render_sigset_t(m, "SigIgn:\t", &ignored);
295 	render_sigset_t(m, "SigCgt:\t", &caught);
296 }
297 
298 static void render_cap_t(struct seq_file *m, const char *header,
299 			kernel_cap_t *a)
300 {
301 	unsigned __capi;
302 
303 	seq_puts(m, header);
304 	CAP_FOR_EACH_U32(__capi) {
305 		seq_put_hex_ll(m, NULL,
306 			   a->cap[CAP_LAST_U32 - __capi], 8);
307 	}
308 	seq_putc(m, '\n');
309 }
310 
311 static inline void task_cap(struct seq_file *m, struct task_struct *p)
312 {
313 	const struct cred *cred;
314 	kernel_cap_t cap_inheritable, cap_permitted, cap_effective,
315 			cap_bset, cap_ambient;
316 
317 	rcu_read_lock();
318 	cred = __task_cred(p);
319 	cap_inheritable	= cred->cap_inheritable;
320 	cap_permitted	= cred->cap_permitted;
321 	cap_effective	= cred->cap_effective;
322 	cap_bset	= cred->cap_bset;
323 	cap_ambient	= cred->cap_ambient;
324 	rcu_read_unlock();
325 
326 	render_cap_t(m, "CapInh:\t", &cap_inheritable);
327 	render_cap_t(m, "CapPrm:\t", &cap_permitted);
328 	render_cap_t(m, "CapEff:\t", &cap_effective);
329 	render_cap_t(m, "CapBnd:\t", &cap_bset);
330 	render_cap_t(m, "CapAmb:\t", &cap_ambient);
331 }
332 
333 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
334 {
335 	seq_put_decimal_ull(m, "NoNewPrivs:\t", task_no_new_privs(p));
336 #ifdef CONFIG_SECCOMP
337 	seq_put_decimal_ull(m, "\nSeccomp:\t", p->seccomp.mode);
338 #endif
339 	seq_printf(m, "\nSpeculation_Store_Bypass:\t");
340 	switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_STORE_BYPASS)) {
341 	case -EINVAL:
342 		seq_printf(m, "unknown");
343 		break;
344 	case PR_SPEC_NOT_AFFECTED:
345 		seq_printf(m, "not vulnerable");
346 		break;
347 	case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
348 		seq_printf(m, "thread force mitigated");
349 		break;
350 	case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
351 		seq_printf(m, "thread mitigated");
352 		break;
353 	case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
354 		seq_printf(m, "thread vulnerable");
355 		break;
356 	case PR_SPEC_DISABLE:
357 		seq_printf(m, "globally mitigated");
358 		break;
359 	default:
360 		seq_printf(m, "vulnerable");
361 		break;
362 	}
363 	seq_putc(m, '\n');
364 }
365 
366 static inline void task_context_switch_counts(struct seq_file *m,
367 						struct task_struct *p)
368 {
369 	seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
370 	seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
371 	seq_putc(m, '\n');
372 }
373 
374 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
375 {
376 	seq_printf(m, "Cpus_allowed:\t%*pb\n",
377 		   cpumask_pr_args(&task->cpus_allowed));
378 	seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
379 		   cpumask_pr_args(&task->cpus_allowed));
380 }
381 
382 static inline void task_core_dumping(struct seq_file *m, struct mm_struct *mm)
383 {
384 	seq_put_decimal_ull(m, "CoreDumping:\t", !!mm->core_state);
385 	seq_putc(m, '\n');
386 }
387 
388 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
389 			struct pid *pid, struct task_struct *task)
390 {
391 	struct mm_struct *mm = get_task_mm(task);
392 
393 	task_name(m, task);
394 	task_state(m, ns, pid, task);
395 
396 	if (mm) {
397 		task_mem(m, mm);
398 		task_core_dumping(m, mm);
399 		mmput(mm);
400 	}
401 	task_sig(m, task);
402 	task_cap(m, task);
403 	task_seccomp(m, task);
404 	task_cpus_allowed(m, task);
405 	cpuset_task_status_allowed(m, task);
406 	task_context_switch_counts(m, task);
407 	return 0;
408 }
409 
410 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
411 			struct pid *pid, struct task_struct *task, int whole)
412 {
413 	unsigned long vsize, eip, esp, wchan = 0;
414 	int priority, nice;
415 	int tty_pgrp = -1, tty_nr = 0;
416 	sigset_t sigign, sigcatch;
417 	char state;
418 	pid_t ppid = 0, pgid = -1, sid = -1;
419 	int num_threads = 0;
420 	int permitted;
421 	struct mm_struct *mm;
422 	unsigned long long start_time;
423 	unsigned long cmin_flt = 0, cmaj_flt = 0;
424 	unsigned long  min_flt = 0,  maj_flt = 0;
425 	u64 cutime, cstime, utime, stime;
426 	u64 cgtime, gtime;
427 	unsigned long rsslim = 0;
428 	char tcomm[sizeof(task->comm)];
429 	unsigned long flags;
430 
431 	state = *get_task_state(task);
432 	vsize = eip = esp = 0;
433 	permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
434 	mm = get_task_mm(task);
435 	if (mm) {
436 		vsize = task_vsize(mm);
437 		/*
438 		 * esp and eip are intentionally zeroed out.  There is no
439 		 * non-racy way to read them without freezing the task.
440 		 * Programs that need reliable values can use ptrace(2).
441 		 *
442 		 * The only exception is if the task is core dumping because
443 		 * a program is not able to use ptrace(2) in that case. It is
444 		 * safe because the task has stopped executing permanently.
445 		 */
446 		if (permitted && (task->flags & PF_DUMPCORE)) {
447 			if (try_get_task_stack(task)) {
448 				eip = KSTK_EIP(task);
449 				esp = KSTK_ESP(task);
450 				put_task_stack(task);
451 			}
452 		}
453 	}
454 
455 	get_task_comm(tcomm, task);
456 
457 	sigemptyset(&sigign);
458 	sigemptyset(&sigcatch);
459 	cutime = cstime = utime = stime = 0;
460 	cgtime = gtime = 0;
461 
462 	if (lock_task_sighand(task, &flags)) {
463 		struct signal_struct *sig = task->signal;
464 
465 		if (sig->tty) {
466 			struct pid *pgrp = tty_get_pgrp(sig->tty);
467 			tty_pgrp = pid_nr_ns(pgrp, ns);
468 			put_pid(pgrp);
469 			tty_nr = new_encode_dev(tty_devnum(sig->tty));
470 		}
471 
472 		num_threads = get_nr_threads(task);
473 		collect_sigign_sigcatch(task, &sigign, &sigcatch);
474 
475 		cmin_flt = sig->cmin_flt;
476 		cmaj_flt = sig->cmaj_flt;
477 		cutime = sig->cutime;
478 		cstime = sig->cstime;
479 		cgtime = sig->cgtime;
480 		rsslim = READ_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
481 
482 		/* add up live thread stats at the group level */
483 		if (whole) {
484 			struct task_struct *t = task;
485 			do {
486 				min_flt += t->min_flt;
487 				maj_flt += t->maj_flt;
488 				gtime += task_gtime(t);
489 			} while_each_thread(task, t);
490 
491 			min_flt += sig->min_flt;
492 			maj_flt += sig->maj_flt;
493 			thread_group_cputime_adjusted(task, &utime, &stime);
494 			gtime += sig->gtime;
495 		}
496 
497 		sid = task_session_nr_ns(task, ns);
498 		ppid = task_tgid_nr_ns(task->real_parent, ns);
499 		pgid = task_pgrp_nr_ns(task, ns);
500 
501 		unlock_task_sighand(task, &flags);
502 	}
503 
504 	if (permitted && (!whole || num_threads < 2))
505 		wchan = get_wchan(task);
506 	if (!whole) {
507 		min_flt = task->min_flt;
508 		maj_flt = task->maj_flt;
509 		task_cputime_adjusted(task, &utime, &stime);
510 		gtime = task_gtime(task);
511 	}
512 
513 	/* scale priority and nice values from timeslices to -20..20 */
514 	/* to make it look like a "normal" Unix priority/nice value  */
515 	priority = task_prio(task);
516 	nice = task_nice(task);
517 
518 	/* convert nsec -> ticks */
519 	start_time = nsec_to_clock_t(task->real_start_time);
520 
521 	seq_put_decimal_ull(m, "", pid_nr_ns(pid, ns));
522 	seq_puts(m, " (");
523 	seq_puts(m, tcomm);
524 	seq_puts(m, ") ");
525 	seq_putc(m, state);
526 	seq_put_decimal_ll(m, " ", ppid);
527 	seq_put_decimal_ll(m, " ", pgid);
528 	seq_put_decimal_ll(m, " ", sid);
529 	seq_put_decimal_ll(m, " ", tty_nr);
530 	seq_put_decimal_ll(m, " ", tty_pgrp);
531 	seq_put_decimal_ull(m, " ", task->flags);
532 	seq_put_decimal_ull(m, " ", min_flt);
533 	seq_put_decimal_ull(m, " ", cmin_flt);
534 	seq_put_decimal_ull(m, " ", maj_flt);
535 	seq_put_decimal_ull(m, " ", cmaj_flt);
536 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(utime));
537 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(stime));
538 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cutime));
539 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cstime));
540 	seq_put_decimal_ll(m, " ", priority);
541 	seq_put_decimal_ll(m, " ", nice);
542 	seq_put_decimal_ll(m, " ", num_threads);
543 	seq_put_decimal_ull(m, " ", 0);
544 	seq_put_decimal_ull(m, " ", start_time);
545 	seq_put_decimal_ull(m, " ", vsize);
546 	seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
547 	seq_put_decimal_ull(m, " ", rsslim);
548 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
549 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
550 	seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
551 	seq_put_decimal_ull(m, " ", esp);
552 	seq_put_decimal_ull(m, " ", eip);
553 	/* The signal information here is obsolete.
554 	 * It must be decimal for Linux 2.0 compatibility.
555 	 * Use /proc/#/status for real-time signals.
556 	 */
557 	seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
558 	seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
559 	seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
560 	seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
561 
562 	/*
563 	 * We used to output the absolute kernel address, but that's an
564 	 * information leak - so instead we show a 0/1 flag here, to signal
565 	 * to user-space whether there's a wchan field in /proc/PID/wchan.
566 	 *
567 	 * This works with older implementations of procps as well.
568 	 */
569 	if (wchan)
570 		seq_puts(m, " 1");
571 	else
572 		seq_puts(m, " 0");
573 
574 	seq_put_decimal_ull(m, " ", 0);
575 	seq_put_decimal_ull(m, " ", 0);
576 	seq_put_decimal_ll(m, " ", task->exit_signal);
577 	seq_put_decimal_ll(m, " ", task_cpu(task));
578 	seq_put_decimal_ull(m, " ", task->rt_priority);
579 	seq_put_decimal_ull(m, " ", task->policy);
580 	seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
581 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(gtime));
582 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cgtime));
583 
584 	if (mm && permitted) {
585 		seq_put_decimal_ull(m, " ", mm->start_data);
586 		seq_put_decimal_ull(m, " ", mm->end_data);
587 		seq_put_decimal_ull(m, " ", mm->start_brk);
588 		seq_put_decimal_ull(m, " ", mm->arg_start);
589 		seq_put_decimal_ull(m, " ", mm->arg_end);
590 		seq_put_decimal_ull(m, " ", mm->env_start);
591 		seq_put_decimal_ull(m, " ", mm->env_end);
592 	} else
593 		seq_puts(m, " 0 0 0 0 0 0 0");
594 
595 	if (permitted)
596 		seq_put_decimal_ll(m, " ", task->exit_code);
597 	else
598 		seq_puts(m, " 0");
599 
600 	seq_putc(m, '\n');
601 	if (mm)
602 		mmput(mm);
603 	return 0;
604 }
605 
606 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
607 			struct pid *pid, struct task_struct *task)
608 {
609 	return do_task_stat(m, ns, pid, task, 0);
610 }
611 
612 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
613 			struct pid *pid, struct task_struct *task)
614 {
615 	return do_task_stat(m, ns, pid, task, 1);
616 }
617 
618 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
619 			struct pid *pid, struct task_struct *task)
620 {
621 	unsigned long size = 0, resident = 0, shared = 0, text = 0, data = 0;
622 	struct mm_struct *mm = get_task_mm(task);
623 
624 	if (mm) {
625 		size = task_statm(mm, &shared, &text, &data, &resident);
626 		mmput(mm);
627 	}
628 	/*
629 	 * For quick read, open code by putting numbers directly
630 	 * expected format is
631 	 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
632 	 *               size, resident, shared, text, data);
633 	 */
634 	seq_put_decimal_ull(m, "", size);
635 	seq_put_decimal_ull(m, " ", resident);
636 	seq_put_decimal_ull(m, " ", shared);
637 	seq_put_decimal_ull(m, " ", text);
638 	seq_put_decimal_ull(m, " ", 0);
639 	seq_put_decimal_ull(m, " ", data);
640 	seq_put_decimal_ull(m, " ", 0);
641 	seq_putc(m, '\n');
642 
643 	return 0;
644 }
645 
646 #ifdef CONFIG_PROC_CHILDREN
647 static struct pid *
648 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
649 {
650 	struct task_struct *start, *task;
651 	struct pid *pid = NULL;
652 
653 	read_lock(&tasklist_lock);
654 
655 	start = pid_task(proc_pid(inode), PIDTYPE_PID);
656 	if (!start)
657 		goto out;
658 
659 	/*
660 	 * Lets try to continue searching first, this gives
661 	 * us significant speedup on children-rich processes.
662 	 */
663 	if (pid_prev) {
664 		task = pid_task(pid_prev, PIDTYPE_PID);
665 		if (task && task->real_parent == start &&
666 		    !(list_empty(&task->sibling))) {
667 			if (list_is_last(&task->sibling, &start->children))
668 				goto out;
669 			task = list_first_entry(&task->sibling,
670 						struct task_struct, sibling);
671 			pid = get_pid(task_pid(task));
672 			goto out;
673 		}
674 	}
675 
676 	/*
677 	 * Slow search case.
678 	 *
679 	 * We might miss some children here if children
680 	 * are exited while we were not holding the lock,
681 	 * but it was never promised to be accurate that
682 	 * much.
683 	 *
684 	 * "Just suppose that the parent sleeps, but N children
685 	 *  exit after we printed their tids. Now the slow paths
686 	 *  skips N extra children, we miss N tasks." (c)
687 	 *
688 	 * So one need to stop or freeze the leader and all
689 	 * its children to get a precise result.
690 	 */
691 	list_for_each_entry(task, &start->children, sibling) {
692 		if (pos-- == 0) {
693 			pid = get_pid(task_pid(task));
694 			break;
695 		}
696 	}
697 
698 out:
699 	read_unlock(&tasklist_lock);
700 	return pid;
701 }
702 
703 static int children_seq_show(struct seq_file *seq, void *v)
704 {
705 	struct inode *inode = seq->private;
706 	pid_t pid;
707 
708 	pid = pid_nr_ns(v, inode->i_sb->s_fs_info);
709 	seq_printf(seq, "%d ", pid);
710 
711 	return 0;
712 }
713 
714 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
715 {
716 	return get_children_pid(seq->private, NULL, *pos);
717 }
718 
719 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
720 {
721 	struct pid *pid;
722 
723 	pid = get_children_pid(seq->private, v, *pos + 1);
724 	put_pid(v);
725 
726 	++*pos;
727 	return pid;
728 }
729 
730 static void children_seq_stop(struct seq_file *seq, void *v)
731 {
732 	put_pid(v);
733 }
734 
735 static const struct seq_operations children_seq_ops = {
736 	.start	= children_seq_start,
737 	.next	= children_seq_next,
738 	.stop	= children_seq_stop,
739 	.show	= children_seq_show,
740 };
741 
742 static int children_seq_open(struct inode *inode, struct file *file)
743 {
744 	struct seq_file *m;
745 	int ret;
746 
747 	ret = seq_open(file, &children_seq_ops);
748 	if (ret)
749 		return ret;
750 
751 	m = file->private_data;
752 	m->private = inode;
753 
754 	return ret;
755 }
756 
757 const struct file_operations proc_tid_children_operations = {
758 	.open    = children_seq_open,
759 	.read    = seq_read,
760 	.llseek  = seq_lseek,
761 	.release = seq_release,
762 };
763 #endif /* CONFIG_PROC_CHILDREN */
764