1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  linux/fs/proc/array.c
4  *
5  *  Copyright (C) 1992  by Linus Torvalds
6  *  based on ideas by Darren Senn
7  *
8  * Fixes:
9  * Michael. K. Johnson: stat,statm extensions.
10  *                      <johnsonm@stolaf.edu>
11  *
12  * Pauline Middelink :  Made cmdline,envline only break at '\0's, to
13  *                      make sure SET_PROCTITLE works. Also removed
14  *                      bad '!' which forced address recalculation for
15  *                      EVERY character on the current page.
16  *                      <middelin@polyware.iaf.nl>
17  *
18  * Danny ter Haar    :	added cpuinfo
19  *			<dth@cistron.nl>
20  *
21  * Alessandro Rubini :  profile extension.
22  *                      <rubini@ipvvis.unipv.it>
23  *
24  * Jeff Tranter      :  added BogoMips field to cpuinfo
25  *                      <Jeff_Tranter@Mitel.COM>
26  *
27  * Bruno Haible      :  remove 4K limit for the maps file
28  *			<haible@ma2s2.mathematik.uni-karlsruhe.de>
29  *
30  * Yves Arrouye      :  remove removal of trailing spaces in get_array.
31  *			<Yves.Arrouye@marin.fdn.fr>
32  *
33  * Jerome Forissier  :  added per-CPU time information to /proc/stat
34  *                      and /proc/<pid>/cpu extension
35  *                      <forissier@isia.cma.fr>
36  *			- Incorporation and non-SMP safe operation
37  *			of forissier patch in 2.1.78 by
38  *			Hans Marcus <crowbar@concepts.nl>
39  *
40  * aeb@cwi.nl        :  /proc/partitions
41  *
42  *
43  * Alan Cox	     :  security fixes.
44  *			<alan@lxorguk.ukuu.org.uk>
45  *
46  * Al Viro           :  safe handling of mm_struct
47  *
48  * Gerhard Wichert   :  added BIGMEM support
49  * Siemens AG           <Gerhard.Wichert@pdb.siemens.de>
50  *
51  * Al Viro & Jeff Garzik :  moved most of the thing into base.c and
52  *			 :  proc_misc.c. The rest may eventually go into
53  *			 :  base.c too.
54  */
55 
56 #include <linux/types.h>
57 #include <linux/errno.h>
58 #include <linux/time.h>
59 #include <linux/time_namespace.h>
60 #include <linux/kernel.h>
61 #include <linux/kernel_stat.h>
62 #include <linux/tty.h>
63 #include <linux/string.h>
64 #include <linux/mman.h>
65 #include <linux/sched/mm.h>
66 #include <linux/sched/numa_balancing.h>
67 #include <linux/sched/task_stack.h>
68 #include <linux/sched/task.h>
69 #include <linux/sched/cputime.h>
70 #include <linux/proc_fs.h>
71 #include <linux/ioport.h>
72 #include <linux/uaccess.h>
73 #include <linux/io.h>
74 #include <linux/mm.h>
75 #include <linux/hugetlb.h>
76 #include <linux/pagemap.h>
77 #include <linux/swap.h>
78 #include <linux/smp.h>
79 #include <linux/signal.h>
80 #include <linux/highmem.h>
81 #include <linux/file.h>
82 #include <linux/fdtable.h>
83 #include <linux/times.h>
84 #include <linux/cpuset.h>
85 #include <linux/rcupdate.h>
86 #include <linux/delayacct.h>
87 #include <linux/seq_file.h>
88 #include <linux/pid_namespace.h>
89 #include <linux/prctl.h>
90 #include <linux/ptrace.h>
91 #include <linux/string_helpers.h>
92 #include <linux/user_namespace.h>
93 #include <linux/fs_struct.h>
94 #include <linux/kthread.h>
95 
96 #include <asm/processor.h>
97 #include "internal.h"
98 
proc_task_name(struct seq_file * m,struct task_struct * p,bool escape)99 void proc_task_name(struct seq_file *m, struct task_struct *p, bool escape)
100 {
101 	char tcomm[64];
102 
103 	if (p->flags & PF_WQ_WORKER)
104 		wq_worker_comm(tcomm, sizeof(tcomm), p);
105 	else if (p->flags & PF_KTHREAD)
106 		get_kthread_comm(tcomm, sizeof(tcomm), p);
107 	else
108 		__get_task_comm(tcomm, sizeof(tcomm), p);
109 
110 	if (escape)
111 		seq_escape_str(m, tcomm, ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
112 	else
113 		seq_printf(m, "%.64s", tcomm);
114 }
115 
116 /*
117  * The task state array is a strange "bitmap" of
118  * reasons to sleep. Thus "running" is zero, and
119  * you can test for combinations of others with
120  * simple bit tests.
121  */
122 static const char * const task_state_array[] = {
123 
124 	/* states in TASK_REPORT: */
125 	"R (running)",		/* 0x00 */
126 	"S (sleeping)",		/* 0x01 */
127 	"D (disk sleep)",	/* 0x02 */
128 	"T (stopped)",		/* 0x04 */
129 	"t (tracing stop)",	/* 0x08 */
130 	"X (dead)",		/* 0x10 */
131 	"Z (zombie)",		/* 0x20 */
132 	"P (parked)",		/* 0x40 */
133 
134 	/* states beyond TASK_REPORT: */
135 	"I (idle)",		/* 0x80 */
136 };
137 
get_task_state(struct task_struct * tsk)138 static inline const char *get_task_state(struct task_struct *tsk)
139 {
140 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != ARRAY_SIZE(task_state_array));
141 	return task_state_array[task_state_index(tsk)];
142 }
143 
task_state(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * p)144 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
145 				struct pid *pid, struct task_struct *p)
146 {
147 	struct user_namespace *user_ns = seq_user_ns(m);
148 	struct group_info *group_info;
149 	int g, umask = -1;
150 	struct task_struct *tracer;
151 	const struct cred *cred;
152 	pid_t ppid, tpid = 0, tgid, ngid;
153 	unsigned int max_fds = 0;
154 
155 	rcu_read_lock();
156 	ppid = pid_alive(p) ?
157 		task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
158 
159 	tracer = ptrace_parent(p);
160 	if (tracer)
161 		tpid = task_pid_nr_ns(tracer, ns);
162 
163 	tgid = task_tgid_nr_ns(p, ns);
164 	ngid = task_numa_group_id(p);
165 	cred = get_task_cred(p);
166 
167 	task_lock(p);
168 	if (p->fs)
169 		umask = p->fs->umask;
170 	if (p->files)
171 		max_fds = files_fdtable(p->files)->max_fds;
172 	task_unlock(p);
173 	rcu_read_unlock();
174 
175 	if (umask >= 0)
176 		seq_printf(m, "Umask:\t%#04o\n", umask);
177 	seq_puts(m, "State:\t");
178 	seq_puts(m, get_task_state(p));
179 
180 	seq_put_decimal_ull(m, "\nTgid:\t", tgid);
181 	seq_put_decimal_ull(m, "\nNgid:\t", ngid);
182 	seq_put_decimal_ull(m, "\nPid:\t", pid_nr_ns(pid, ns));
183 	seq_put_decimal_ull(m, "\nPPid:\t", ppid);
184 	seq_put_decimal_ull(m, "\nTracerPid:\t", tpid);
185 	seq_put_decimal_ull(m, "\nUid:\t", from_kuid_munged(user_ns, cred->uid));
186 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->euid));
187 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->suid));
188 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->fsuid));
189 	seq_put_decimal_ull(m, "\nGid:\t", from_kgid_munged(user_ns, cred->gid));
190 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->egid));
191 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->sgid));
192 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->fsgid));
193 	seq_put_decimal_ull(m, "\nFDSize:\t", max_fds);
194 
195 	seq_puts(m, "\nGroups:\t");
196 	group_info = cred->group_info;
197 	for (g = 0; g < group_info->ngroups; g++)
198 		seq_put_decimal_ull(m, g ? " " : "",
199 				from_kgid_munged(user_ns, group_info->gid[g]));
200 	put_cred(cred);
201 	/* Trailing space shouldn't have been added in the first place. */
202 	seq_putc(m, ' ');
203 
204 #ifdef CONFIG_PID_NS
205 	seq_puts(m, "\nNStgid:");
206 	for (g = ns->level; g <= pid->level; g++)
207 		seq_put_decimal_ull(m, "\t", task_tgid_nr_ns(p, pid->numbers[g].ns));
208 	seq_puts(m, "\nNSpid:");
209 	for (g = ns->level; g <= pid->level; g++)
210 		seq_put_decimal_ull(m, "\t", task_pid_nr_ns(p, pid->numbers[g].ns));
211 	seq_puts(m, "\nNSpgid:");
212 	for (g = ns->level; g <= pid->level; g++)
213 		seq_put_decimal_ull(m, "\t", task_pgrp_nr_ns(p, pid->numbers[g].ns));
214 	seq_puts(m, "\nNSsid:");
215 	for (g = ns->level; g <= pid->level; g++)
216 		seq_put_decimal_ull(m, "\t", task_session_nr_ns(p, pid->numbers[g].ns));
217 #endif
218 	seq_putc(m, '\n');
219 }
220 
render_sigset_t(struct seq_file * m,const char * header,sigset_t * set)221 void render_sigset_t(struct seq_file *m, const char *header,
222 				sigset_t *set)
223 {
224 	int i;
225 
226 	seq_puts(m, header);
227 
228 	i = _NSIG;
229 	do {
230 		int x = 0;
231 
232 		i -= 4;
233 		if (sigismember(set, i+1)) x |= 1;
234 		if (sigismember(set, i+2)) x |= 2;
235 		if (sigismember(set, i+3)) x |= 4;
236 		if (sigismember(set, i+4)) x |= 8;
237 		seq_putc(m, hex_asc[x]);
238 	} while (i >= 4);
239 
240 	seq_putc(m, '\n');
241 }
242 
collect_sigign_sigcatch(struct task_struct * p,sigset_t * sigign,sigset_t * sigcatch)243 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *sigign,
244 				    sigset_t *sigcatch)
245 {
246 	struct k_sigaction *k;
247 	int i;
248 
249 	k = p->sighand->action;
250 	for (i = 1; i <= _NSIG; ++i, ++k) {
251 		if (k->sa.sa_handler == SIG_IGN)
252 			sigaddset(sigign, i);
253 		else if (k->sa.sa_handler != SIG_DFL)
254 			sigaddset(sigcatch, i);
255 	}
256 }
257 
task_sig(struct seq_file * m,struct task_struct * p)258 static inline void task_sig(struct seq_file *m, struct task_struct *p)
259 {
260 	unsigned long flags;
261 	sigset_t pending, shpending, blocked, ignored, caught;
262 	int num_threads = 0;
263 	unsigned int qsize = 0;
264 	unsigned long qlim = 0;
265 
266 	sigemptyset(&pending);
267 	sigemptyset(&shpending);
268 	sigemptyset(&blocked);
269 	sigemptyset(&ignored);
270 	sigemptyset(&caught);
271 
272 	if (lock_task_sighand(p, &flags)) {
273 		pending = p->pending.signal;
274 		shpending = p->signal->shared_pending.signal;
275 		blocked = p->blocked;
276 		collect_sigign_sigcatch(p, &ignored, &caught);
277 		num_threads = get_nr_threads(p);
278 		rcu_read_lock();  /* FIXME: is this correct? */
279 		qsize = get_ucounts_value(task_ucounts(p), UCOUNT_RLIMIT_SIGPENDING);
280 		rcu_read_unlock();
281 		qlim = task_rlimit(p, RLIMIT_SIGPENDING);
282 		unlock_task_sighand(p, &flags);
283 	}
284 
285 	seq_put_decimal_ull(m, "Threads:\t", num_threads);
286 	seq_put_decimal_ull(m, "\nSigQ:\t", qsize);
287 	seq_put_decimal_ull(m, "/", qlim);
288 
289 	/* render them all */
290 	render_sigset_t(m, "\nSigPnd:\t", &pending);
291 	render_sigset_t(m, "ShdPnd:\t", &shpending);
292 	render_sigset_t(m, "SigBlk:\t", &blocked);
293 	render_sigset_t(m, "SigIgn:\t", &ignored);
294 	render_sigset_t(m, "SigCgt:\t", &caught);
295 }
296 
render_cap_t(struct seq_file * m,const char * header,kernel_cap_t * a)297 static void render_cap_t(struct seq_file *m, const char *header,
298 			kernel_cap_t *a)
299 {
300 	unsigned __capi;
301 
302 	seq_puts(m, header);
303 	CAP_FOR_EACH_U32(__capi) {
304 		seq_put_hex_ll(m, NULL,
305 			   a->cap[CAP_LAST_U32 - __capi], 8);
306 	}
307 	seq_putc(m, '\n');
308 }
309 
task_cap(struct seq_file * m,struct task_struct * p)310 static inline void task_cap(struct seq_file *m, struct task_struct *p)
311 {
312 	const struct cred *cred;
313 	kernel_cap_t cap_inheritable, cap_permitted, cap_effective,
314 			cap_bset, cap_ambient;
315 
316 	rcu_read_lock();
317 	cred = __task_cred(p);
318 	cap_inheritable	= cred->cap_inheritable;
319 	cap_permitted	= cred->cap_permitted;
320 	cap_effective	= cred->cap_effective;
321 	cap_bset	= cred->cap_bset;
322 	cap_ambient	= cred->cap_ambient;
323 	rcu_read_unlock();
324 
325 	render_cap_t(m, "CapInh:\t", &cap_inheritable);
326 	render_cap_t(m, "CapPrm:\t", &cap_permitted);
327 	render_cap_t(m, "CapEff:\t", &cap_effective);
328 	render_cap_t(m, "CapBnd:\t", &cap_bset);
329 	render_cap_t(m, "CapAmb:\t", &cap_ambient);
330 }
331 
task_seccomp(struct seq_file * m,struct task_struct * p)332 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
333 {
334 	seq_put_decimal_ull(m, "NoNewPrivs:\t", task_no_new_privs(p));
335 #ifdef CONFIG_SECCOMP
336 	seq_put_decimal_ull(m, "\nSeccomp:\t", p->seccomp.mode);
337 #ifdef CONFIG_SECCOMP_FILTER
338 	seq_put_decimal_ull(m, "\nSeccomp_filters:\t",
339 			    atomic_read(&p->seccomp.filter_count));
340 #endif
341 #endif
342 	seq_puts(m, "\nSpeculation_Store_Bypass:\t");
343 	switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_STORE_BYPASS)) {
344 	case -EINVAL:
345 		seq_puts(m, "unknown");
346 		break;
347 	case PR_SPEC_NOT_AFFECTED:
348 		seq_puts(m, "not vulnerable");
349 		break;
350 	case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
351 		seq_puts(m, "thread force mitigated");
352 		break;
353 	case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
354 		seq_puts(m, "thread mitigated");
355 		break;
356 	case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
357 		seq_puts(m, "thread vulnerable");
358 		break;
359 	case PR_SPEC_DISABLE:
360 		seq_puts(m, "globally mitigated");
361 		break;
362 	default:
363 		seq_puts(m, "vulnerable");
364 		break;
365 	}
366 
367 	seq_puts(m, "\nSpeculationIndirectBranch:\t");
368 	switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_INDIRECT_BRANCH)) {
369 	case -EINVAL:
370 		seq_puts(m, "unsupported");
371 		break;
372 	case PR_SPEC_NOT_AFFECTED:
373 		seq_puts(m, "not affected");
374 		break;
375 	case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
376 		seq_puts(m, "conditional force disabled");
377 		break;
378 	case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
379 		seq_puts(m, "conditional disabled");
380 		break;
381 	case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
382 		seq_puts(m, "conditional enabled");
383 		break;
384 	case PR_SPEC_ENABLE:
385 		seq_puts(m, "always enabled");
386 		break;
387 	case PR_SPEC_DISABLE:
388 		seq_puts(m, "always disabled");
389 		break;
390 	default:
391 		seq_puts(m, "unknown");
392 		break;
393 	}
394 	seq_putc(m, '\n');
395 }
396 
task_context_switch_counts(struct seq_file * m,struct task_struct * p)397 static inline void task_context_switch_counts(struct seq_file *m,
398 						struct task_struct *p)
399 {
400 	seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
401 	seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
402 	seq_putc(m, '\n');
403 }
404 
task_cpus_allowed(struct seq_file * m,struct task_struct * task)405 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
406 {
407 	seq_printf(m, "Cpus_allowed:\t%*pb\n",
408 		   cpumask_pr_args(&task->cpus_mask));
409 	seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
410 		   cpumask_pr_args(&task->cpus_mask));
411 }
412 
task_core_dumping(struct seq_file * m,struct task_struct * task)413 static inline void task_core_dumping(struct seq_file *m, struct task_struct *task)
414 {
415 	seq_put_decimal_ull(m, "CoreDumping:\t", !!task->signal->core_state);
416 	seq_putc(m, '\n');
417 }
418 
task_thp_status(struct seq_file * m,struct mm_struct * mm)419 static inline void task_thp_status(struct seq_file *m, struct mm_struct *mm)
420 {
421 	bool thp_enabled = IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE);
422 
423 	if (thp_enabled)
424 		thp_enabled = !test_bit(MMF_DISABLE_THP, &mm->flags);
425 	seq_printf(m, "THP_enabled:\t%d\n", thp_enabled);
426 }
427 
proc_pid_status(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)428 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
429 			struct pid *pid, struct task_struct *task)
430 {
431 	struct mm_struct *mm = get_task_mm(task);
432 
433 	seq_puts(m, "Name:\t");
434 	proc_task_name(m, task, true);
435 	seq_putc(m, '\n');
436 
437 	task_state(m, ns, pid, task);
438 
439 	if (mm) {
440 		task_mem(m, mm);
441 		task_core_dumping(m, task);
442 		task_thp_status(m, mm);
443 		mmput(mm);
444 	}
445 	task_sig(m, task);
446 	task_cap(m, task);
447 	task_seccomp(m, task);
448 	task_cpus_allowed(m, task);
449 	cpuset_task_status_allowed(m, task);
450 	task_context_switch_counts(m, task);
451 	return 0;
452 }
453 
do_task_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task,int whole)454 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
455 			struct pid *pid, struct task_struct *task, int whole)
456 {
457 	unsigned long vsize, eip, esp, wchan = 0;
458 	int priority, nice;
459 	int tty_pgrp = -1, tty_nr = 0;
460 	sigset_t sigign, sigcatch;
461 	char state;
462 	pid_t ppid = 0, pgid = -1, sid = -1;
463 	int num_threads = 0;
464 	int permitted;
465 	struct mm_struct *mm;
466 	unsigned long long start_time;
467 	unsigned long cmin_flt = 0, cmaj_flt = 0;
468 	unsigned long  min_flt = 0,  maj_flt = 0;
469 	u64 cutime, cstime, utime, stime;
470 	u64 cgtime, gtime;
471 	unsigned long rsslim = 0;
472 	unsigned long flags;
473 	int exit_code = task->exit_code;
474 
475 	state = *get_task_state(task);
476 	vsize = eip = esp = 0;
477 	permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
478 	mm = get_task_mm(task);
479 	if (mm) {
480 		vsize = task_vsize(mm);
481 		/*
482 		 * esp and eip are intentionally zeroed out.  There is no
483 		 * non-racy way to read them without freezing the task.
484 		 * Programs that need reliable values can use ptrace(2).
485 		 *
486 		 * The only exception is if the task is core dumping because
487 		 * a program is not able to use ptrace(2) in that case. It is
488 		 * safe because the task has stopped executing permanently.
489 		 */
490 		if (permitted && (task->flags & (PF_EXITING|PF_DUMPCORE))) {
491 			if (try_get_task_stack(task)) {
492 				eip = KSTK_EIP(task);
493 				esp = KSTK_ESP(task);
494 				put_task_stack(task);
495 			}
496 		}
497 	}
498 
499 	sigemptyset(&sigign);
500 	sigemptyset(&sigcatch);
501 	cutime = cstime = utime = stime = 0;
502 	cgtime = gtime = 0;
503 
504 	if (lock_task_sighand(task, &flags)) {
505 		struct signal_struct *sig = task->signal;
506 
507 		if (sig->tty) {
508 			struct pid *pgrp = tty_get_pgrp(sig->tty);
509 			tty_pgrp = pid_nr_ns(pgrp, ns);
510 			put_pid(pgrp);
511 			tty_nr = new_encode_dev(tty_devnum(sig->tty));
512 		}
513 
514 		num_threads = get_nr_threads(task);
515 		collect_sigign_sigcatch(task, &sigign, &sigcatch);
516 
517 		cmin_flt = sig->cmin_flt;
518 		cmaj_flt = sig->cmaj_flt;
519 		cutime = sig->cutime;
520 		cstime = sig->cstime;
521 		cgtime = sig->cgtime;
522 		rsslim = READ_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
523 
524 		/* add up live thread stats at the group level */
525 		if (whole) {
526 			struct task_struct *t = task;
527 			do {
528 				min_flt += t->min_flt;
529 				maj_flt += t->maj_flt;
530 				gtime += task_gtime(t);
531 			} while_each_thread(task, t);
532 
533 			min_flt += sig->min_flt;
534 			maj_flt += sig->maj_flt;
535 			thread_group_cputime_adjusted(task, &utime, &stime);
536 			gtime += sig->gtime;
537 
538 			if (sig->flags & (SIGNAL_GROUP_EXIT | SIGNAL_STOP_STOPPED))
539 				exit_code = sig->group_exit_code;
540 		}
541 
542 		sid = task_session_nr_ns(task, ns);
543 		ppid = task_tgid_nr_ns(task->real_parent, ns);
544 		pgid = task_pgrp_nr_ns(task, ns);
545 
546 		unlock_task_sighand(task, &flags);
547 	}
548 
549 	if (permitted && (!whole || num_threads < 2))
550 		wchan = !task_is_running(task);
551 	if (!whole) {
552 		min_flt = task->min_flt;
553 		maj_flt = task->maj_flt;
554 		task_cputime_adjusted(task, &utime, &stime);
555 		gtime = task_gtime(task);
556 	}
557 
558 	/* scale priority and nice values from timeslices to -20..20 */
559 	/* to make it look like a "normal" Unix priority/nice value  */
560 	priority = task_prio(task);
561 	nice = task_nice(task);
562 
563 	/* apply timens offset for boottime and convert nsec -> ticks */
564 	start_time =
565 		nsec_to_clock_t(timens_add_boottime_ns(task->start_boottime));
566 
567 	seq_put_decimal_ull(m, "", pid_nr_ns(pid, ns));
568 	seq_puts(m, " (");
569 	proc_task_name(m, task, false);
570 	seq_puts(m, ") ");
571 	seq_putc(m, state);
572 	seq_put_decimal_ll(m, " ", ppid);
573 	seq_put_decimal_ll(m, " ", pgid);
574 	seq_put_decimal_ll(m, " ", sid);
575 	seq_put_decimal_ll(m, " ", tty_nr);
576 	seq_put_decimal_ll(m, " ", tty_pgrp);
577 	seq_put_decimal_ull(m, " ", task->flags);
578 	seq_put_decimal_ull(m, " ", min_flt);
579 	seq_put_decimal_ull(m, " ", cmin_flt);
580 	seq_put_decimal_ull(m, " ", maj_flt);
581 	seq_put_decimal_ull(m, " ", cmaj_flt);
582 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(utime));
583 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(stime));
584 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cutime));
585 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cstime));
586 	seq_put_decimal_ll(m, " ", priority);
587 	seq_put_decimal_ll(m, " ", nice);
588 	seq_put_decimal_ll(m, " ", num_threads);
589 	seq_put_decimal_ull(m, " ", 0);
590 	seq_put_decimal_ull(m, " ", start_time);
591 	seq_put_decimal_ull(m, " ", vsize);
592 	seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
593 	seq_put_decimal_ull(m, " ", rsslim);
594 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
595 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
596 	seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
597 	seq_put_decimal_ull(m, " ", esp);
598 	seq_put_decimal_ull(m, " ", eip);
599 	/* The signal information here is obsolete.
600 	 * It must be decimal for Linux 2.0 compatibility.
601 	 * Use /proc/#/status for real-time signals.
602 	 */
603 	seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
604 	seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
605 	seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
606 	seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
607 
608 	/*
609 	 * We used to output the absolute kernel address, but that's an
610 	 * information leak - so instead we show a 0/1 flag here, to signal
611 	 * to user-space whether there's a wchan field in /proc/PID/wchan.
612 	 *
613 	 * This works with older implementations of procps as well.
614 	 */
615 	seq_put_decimal_ull(m, " ", wchan);
616 
617 	seq_put_decimal_ull(m, " ", 0);
618 	seq_put_decimal_ull(m, " ", 0);
619 	seq_put_decimal_ll(m, " ", task->exit_signal);
620 	seq_put_decimal_ll(m, " ", task_cpu(task));
621 	seq_put_decimal_ull(m, " ", task->rt_priority);
622 	seq_put_decimal_ull(m, " ", task->policy);
623 	seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
624 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(gtime));
625 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cgtime));
626 
627 	if (mm && permitted) {
628 		seq_put_decimal_ull(m, " ", mm->start_data);
629 		seq_put_decimal_ull(m, " ", mm->end_data);
630 		seq_put_decimal_ull(m, " ", mm->start_brk);
631 		seq_put_decimal_ull(m, " ", mm->arg_start);
632 		seq_put_decimal_ull(m, " ", mm->arg_end);
633 		seq_put_decimal_ull(m, " ", mm->env_start);
634 		seq_put_decimal_ull(m, " ", mm->env_end);
635 	} else
636 		seq_puts(m, " 0 0 0 0 0 0 0");
637 
638 	if (permitted)
639 		seq_put_decimal_ll(m, " ", exit_code);
640 	else
641 		seq_puts(m, " 0");
642 
643 	seq_putc(m, '\n');
644 	if (mm)
645 		mmput(mm);
646 	return 0;
647 }
648 
proc_tid_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)649 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
650 			struct pid *pid, struct task_struct *task)
651 {
652 	return do_task_stat(m, ns, pid, task, 0);
653 }
654 
proc_tgid_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)655 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
656 			struct pid *pid, struct task_struct *task)
657 {
658 	return do_task_stat(m, ns, pid, task, 1);
659 }
660 
proc_pid_statm(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)661 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
662 			struct pid *pid, struct task_struct *task)
663 {
664 	struct mm_struct *mm = get_task_mm(task);
665 
666 	if (mm) {
667 		unsigned long size;
668 		unsigned long resident = 0;
669 		unsigned long shared = 0;
670 		unsigned long text = 0;
671 		unsigned long data = 0;
672 
673 		size = task_statm(mm, &shared, &text, &data, &resident);
674 		mmput(mm);
675 
676 		/*
677 		 * For quick read, open code by putting numbers directly
678 		 * expected format is
679 		 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
680 		 *               size, resident, shared, text, data);
681 		 */
682 		seq_put_decimal_ull(m, "", size);
683 		seq_put_decimal_ull(m, " ", resident);
684 		seq_put_decimal_ull(m, " ", shared);
685 		seq_put_decimal_ull(m, " ", text);
686 		seq_put_decimal_ull(m, " ", 0);
687 		seq_put_decimal_ull(m, " ", data);
688 		seq_put_decimal_ull(m, " ", 0);
689 		seq_putc(m, '\n');
690 	} else {
691 		seq_write(m, "0 0 0 0 0 0 0\n", 14);
692 	}
693 	return 0;
694 }
695 
696 #ifdef CONFIG_PROC_CHILDREN
697 static struct pid *
get_children_pid(struct inode * inode,struct pid * pid_prev,loff_t pos)698 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
699 {
700 	struct task_struct *start, *task;
701 	struct pid *pid = NULL;
702 
703 	read_lock(&tasklist_lock);
704 
705 	start = pid_task(proc_pid(inode), PIDTYPE_PID);
706 	if (!start)
707 		goto out;
708 
709 	/*
710 	 * Lets try to continue searching first, this gives
711 	 * us significant speedup on children-rich processes.
712 	 */
713 	if (pid_prev) {
714 		task = pid_task(pid_prev, PIDTYPE_PID);
715 		if (task && task->real_parent == start &&
716 		    !(list_empty(&task->sibling))) {
717 			if (list_is_last(&task->sibling, &start->children))
718 				goto out;
719 			task = list_first_entry(&task->sibling,
720 						struct task_struct, sibling);
721 			pid = get_pid(task_pid(task));
722 			goto out;
723 		}
724 	}
725 
726 	/*
727 	 * Slow search case.
728 	 *
729 	 * We might miss some children here if children
730 	 * are exited while we were not holding the lock,
731 	 * but it was never promised to be accurate that
732 	 * much.
733 	 *
734 	 * "Just suppose that the parent sleeps, but N children
735 	 *  exit after we printed their tids. Now the slow paths
736 	 *  skips N extra children, we miss N tasks." (c)
737 	 *
738 	 * So one need to stop or freeze the leader and all
739 	 * its children to get a precise result.
740 	 */
741 	list_for_each_entry(task, &start->children, sibling) {
742 		if (pos-- == 0) {
743 			pid = get_pid(task_pid(task));
744 			break;
745 		}
746 	}
747 
748 out:
749 	read_unlock(&tasklist_lock);
750 	return pid;
751 }
752 
children_seq_show(struct seq_file * seq,void * v)753 static int children_seq_show(struct seq_file *seq, void *v)
754 {
755 	struct inode *inode = file_inode(seq->file);
756 
757 	seq_printf(seq, "%d ", pid_nr_ns(v, proc_pid_ns(inode->i_sb)));
758 	return 0;
759 }
760 
children_seq_start(struct seq_file * seq,loff_t * pos)761 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
762 {
763 	return get_children_pid(file_inode(seq->file), NULL, *pos);
764 }
765 
children_seq_next(struct seq_file * seq,void * v,loff_t * pos)766 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
767 {
768 	struct pid *pid;
769 
770 	pid = get_children_pid(file_inode(seq->file), v, *pos + 1);
771 	put_pid(v);
772 
773 	++*pos;
774 	return pid;
775 }
776 
children_seq_stop(struct seq_file * seq,void * v)777 static void children_seq_stop(struct seq_file *seq, void *v)
778 {
779 	put_pid(v);
780 }
781 
782 static const struct seq_operations children_seq_ops = {
783 	.start	= children_seq_start,
784 	.next	= children_seq_next,
785 	.stop	= children_seq_stop,
786 	.show	= children_seq_show,
787 };
788 
children_seq_open(struct inode * inode,struct file * file)789 static int children_seq_open(struct inode *inode, struct file *file)
790 {
791 	return seq_open(file, &children_seq_ops);
792 }
793 
794 const struct file_operations proc_tid_children_operations = {
795 	.open    = children_seq_open,
796 	.read    = seq_read,
797 	.llseek  = seq_lseek,
798 	.release = seq_release,
799 };
800 #endif /* CONFIG_PROC_CHILDREN */
801