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