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/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/string_helpers.h>
91 #include <linux/user_namespace.h>
92 #include <linux/fs_struct.h>
93 #include <linux/kthread.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 tcomm[64];
101 
102 	/*
103 	 * Test before PF_KTHREAD because all workqueue worker threads are
104 	 * kernel threads.
105 	 */
106 	if (p->flags & PF_WQ_WORKER)
107 		wq_worker_comm(tcomm, sizeof(tcomm), p);
108 	else if (p->flags & PF_KTHREAD)
109 		get_kthread_comm(tcomm, sizeof(tcomm), p);
110 	else
111 		__get_task_comm(tcomm, sizeof(tcomm), p);
112 
113 	if (escape)
114 		seq_escape_str(m, tcomm, ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
115 	else
116 		seq_printf(m, "%.64s", tcomm);
117 }
118 
119 /*
120  * The task state array is a strange "bitmap" of
121  * reasons to sleep. Thus "running" is zero, and
122  * you can test for combinations of others with
123  * simple bit tests.
124  */
125 static const char * const task_state_array[] = {
126 
127 	/* states in TASK_REPORT: */
128 	"R (running)",		/* 0x00 */
129 	"S (sleeping)",		/* 0x01 */
130 	"D (disk sleep)",	/* 0x02 */
131 	"T (stopped)",		/* 0x04 */
132 	"t (tracing stop)",	/* 0x08 */
133 	"X (dead)",		/* 0x10 */
134 	"Z (zombie)",		/* 0x20 */
135 	"P (parked)",		/* 0x40 */
136 
137 	/* states beyond TASK_REPORT: */
138 	"I (idle)",		/* 0x80 */
139 };
140 
get_task_state(struct task_struct * tsk)141 static inline const char *get_task_state(struct task_struct *tsk)
142 {
143 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != ARRAY_SIZE(task_state_array));
144 	return task_state_array[task_state_index(tsk)];
145 }
146 
task_state(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * p)147 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
148 				struct pid *pid, struct task_struct *p)
149 {
150 	struct user_namespace *user_ns = seq_user_ns(m);
151 	struct group_info *group_info;
152 	int g, umask = -1;
153 	struct task_struct *tracer;
154 	const struct cred *cred;
155 	pid_t ppid, tpid = 0, tgid, ngid;
156 	unsigned int max_fds = 0;
157 
158 	rcu_read_lock();
159 	ppid = pid_alive(p) ?
160 		task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
161 
162 	tracer = ptrace_parent(p);
163 	if (tracer)
164 		tpid = task_pid_nr_ns(tracer, ns);
165 
166 	tgid = task_tgid_nr_ns(p, ns);
167 	ngid = task_numa_group_id(p);
168 	cred = get_task_cred(p);
169 
170 	task_lock(p);
171 	if (p->fs)
172 		umask = p->fs->umask;
173 	if (p->files)
174 		max_fds = files_fdtable(p->files)->max_fds;
175 	task_unlock(p);
176 	rcu_read_unlock();
177 
178 	if (umask >= 0)
179 		seq_printf(m, "Umask:\t%#04o\n", umask);
180 	seq_puts(m, "State:\t");
181 	seq_puts(m, get_task_state(p));
182 
183 	seq_put_decimal_ull(m, "\nTgid:\t", tgid);
184 	seq_put_decimal_ull(m, "\nNgid:\t", ngid);
185 	seq_put_decimal_ull(m, "\nPid:\t", pid_nr_ns(pid, ns));
186 	seq_put_decimal_ull(m, "\nPPid:\t", ppid);
187 	seq_put_decimal_ull(m, "\nTracerPid:\t", tpid);
188 	seq_put_decimal_ull(m, "\nUid:\t", from_kuid_munged(user_ns, cred->uid));
189 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->euid));
190 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->suid));
191 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->fsuid));
192 	seq_put_decimal_ull(m, "\nGid:\t", from_kgid_munged(user_ns, cred->gid));
193 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->egid));
194 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->sgid));
195 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->fsgid));
196 	seq_put_decimal_ull(m, "\nFDSize:\t", max_fds);
197 
198 	seq_puts(m, "\nGroups:\t");
199 	group_info = cred->group_info;
200 	for (g = 0; g < group_info->ngroups; g++)
201 		seq_put_decimal_ull(m, g ? " " : "",
202 				from_kgid_munged(user_ns, group_info->gid[g]));
203 	put_cred(cred);
204 	/* Trailing space shouldn't have been added in the first place. */
205 	seq_putc(m, ' ');
206 
207 #ifdef CONFIG_PID_NS
208 	seq_puts(m, "\nNStgid:");
209 	for (g = ns->level; g <= pid->level; g++)
210 		seq_put_decimal_ull(m, "\t", task_tgid_nr_ns(p, pid->numbers[g].ns));
211 	seq_puts(m, "\nNSpid:");
212 	for (g = ns->level; g <= pid->level; g++)
213 		seq_put_decimal_ull(m, "\t", task_pid_nr_ns(p, pid->numbers[g].ns));
214 	seq_puts(m, "\nNSpgid:");
215 	for (g = ns->level; g <= pid->level; g++)
216 		seq_put_decimal_ull(m, "\t", task_pgrp_nr_ns(p, pid->numbers[g].ns));
217 	seq_puts(m, "\nNSsid:");
218 	for (g = ns->level; g <= pid->level; g++)
219 		seq_put_decimal_ull(m, "\t", task_session_nr_ns(p, pid->numbers[g].ns));
220 #endif
221 	seq_putc(m, '\n');
222 }
223 
render_sigset_t(struct seq_file * m,const char * header,sigset_t * set)224 void render_sigset_t(struct seq_file *m, const char *header,
225 				sigset_t *set)
226 {
227 	int i;
228 
229 	seq_puts(m, header);
230 
231 	i = _NSIG;
232 	do {
233 		int x = 0;
234 
235 		i -= 4;
236 		if (sigismember(set, i+1)) x |= 1;
237 		if (sigismember(set, i+2)) x |= 2;
238 		if (sigismember(set, i+3)) x |= 4;
239 		if (sigismember(set, i+4)) x |= 8;
240 		seq_putc(m, hex_asc[x]);
241 	} while (i >= 4);
242 
243 	seq_putc(m, '\n');
244 }
245 
collect_sigign_sigcatch(struct task_struct * p,sigset_t * sigign,sigset_t * sigcatch)246 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *sigign,
247 				    sigset_t *sigcatch)
248 {
249 	struct k_sigaction *k;
250 	int i;
251 
252 	k = p->sighand->action;
253 	for (i = 1; i <= _NSIG; ++i, ++k) {
254 		if (k->sa.sa_handler == SIG_IGN)
255 			sigaddset(sigign, i);
256 		else if (k->sa.sa_handler != SIG_DFL)
257 			sigaddset(sigcatch, i);
258 	}
259 }
260 
task_sig(struct seq_file * m,struct task_struct * p)261 static inline void task_sig(struct seq_file *m, struct task_struct *p)
262 {
263 	unsigned long flags;
264 	sigset_t pending, shpending, blocked, ignored, caught;
265 	int num_threads = 0;
266 	unsigned int qsize = 0;
267 	unsigned long qlim = 0;
268 
269 	sigemptyset(&pending);
270 	sigemptyset(&shpending);
271 	sigemptyset(&blocked);
272 	sigemptyset(&ignored);
273 	sigemptyset(&caught);
274 
275 	if (lock_task_sighand(p, &flags)) {
276 		pending = p->pending.signal;
277 		shpending = p->signal->shared_pending.signal;
278 		blocked = p->blocked;
279 		collect_sigign_sigcatch(p, &ignored, &caught);
280 		num_threads = get_nr_threads(p);
281 		rcu_read_lock();  /* FIXME: is this correct? */
282 		qsize = get_rlimit_value(task_ucounts(p), UCOUNT_RLIMIT_SIGPENDING);
283 		rcu_read_unlock();
284 		qlim = task_rlimit(p, RLIMIT_SIGPENDING);
285 		unlock_task_sighand(p, &flags);
286 	}
287 
288 	seq_put_decimal_ull(m, "Threads:\t", num_threads);
289 	seq_put_decimal_ull(m, "\nSigQ:\t", qsize);
290 	seq_put_decimal_ull(m, "/", qlim);
291 
292 	/* render them all */
293 	render_sigset_t(m, "\nSigPnd:\t", &pending);
294 	render_sigset_t(m, "ShdPnd:\t", &shpending);
295 	render_sigset_t(m, "SigBlk:\t", &blocked);
296 	render_sigset_t(m, "SigIgn:\t", &ignored);
297 	render_sigset_t(m, "SigCgt:\t", &caught);
298 }
299 
render_cap_t(struct seq_file * m,const char * header,kernel_cap_t * a)300 static void render_cap_t(struct seq_file *m, const char *header,
301 			kernel_cap_t *a)
302 {
303 	unsigned __capi;
304 
305 	seq_puts(m, header);
306 	CAP_FOR_EACH_U32(__capi) {
307 		seq_put_hex_ll(m, NULL,
308 			   a->cap[CAP_LAST_U32 - __capi], 8);
309 	}
310 	seq_putc(m, '\n');
311 }
312 
task_cap(struct seq_file * m,struct task_struct * p)313 static inline void task_cap(struct seq_file *m, struct task_struct *p)
314 {
315 	const struct cred *cred;
316 	kernel_cap_t cap_inheritable, cap_permitted, cap_effective,
317 			cap_bset, cap_ambient;
318 
319 	rcu_read_lock();
320 	cred = __task_cred(p);
321 	cap_inheritable	= cred->cap_inheritable;
322 	cap_permitted	= cred->cap_permitted;
323 	cap_effective	= cred->cap_effective;
324 	cap_bset	= cred->cap_bset;
325 	cap_ambient	= cred->cap_ambient;
326 	rcu_read_unlock();
327 
328 	render_cap_t(m, "CapInh:\t", &cap_inheritable);
329 	render_cap_t(m, "CapPrm:\t", &cap_permitted);
330 	render_cap_t(m, "CapEff:\t", &cap_effective);
331 	render_cap_t(m, "CapBnd:\t", &cap_bset);
332 	render_cap_t(m, "CapAmb:\t", &cap_ambient);
333 }
334 
task_seccomp(struct seq_file * m,struct task_struct * p)335 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
336 {
337 	seq_put_decimal_ull(m, "NoNewPrivs:\t", task_no_new_privs(p));
338 #ifdef CONFIG_SECCOMP
339 	seq_put_decimal_ull(m, "\nSeccomp:\t", p->seccomp.mode);
340 #ifdef CONFIG_SECCOMP_FILTER
341 	seq_put_decimal_ull(m, "\nSeccomp_filters:\t",
342 			    atomic_read(&p->seccomp.filter_count));
343 #endif
344 #endif
345 	seq_puts(m, "\nSpeculation_Store_Bypass:\t");
346 	switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_STORE_BYPASS)) {
347 	case -EINVAL:
348 		seq_puts(m, "unknown");
349 		break;
350 	case PR_SPEC_NOT_AFFECTED:
351 		seq_puts(m, "not vulnerable");
352 		break;
353 	case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
354 		seq_puts(m, "thread force mitigated");
355 		break;
356 	case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
357 		seq_puts(m, "thread mitigated");
358 		break;
359 	case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
360 		seq_puts(m, "thread vulnerable");
361 		break;
362 	case PR_SPEC_DISABLE:
363 		seq_puts(m, "globally mitigated");
364 		break;
365 	default:
366 		seq_puts(m, "vulnerable");
367 		break;
368 	}
369 
370 	seq_puts(m, "\nSpeculationIndirectBranch:\t");
371 	switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_INDIRECT_BRANCH)) {
372 	case -EINVAL:
373 		seq_puts(m, "unsupported");
374 		break;
375 	case PR_SPEC_NOT_AFFECTED:
376 		seq_puts(m, "not affected");
377 		break;
378 	case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
379 		seq_puts(m, "conditional force disabled");
380 		break;
381 	case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
382 		seq_puts(m, "conditional disabled");
383 		break;
384 	case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
385 		seq_puts(m, "conditional enabled");
386 		break;
387 	case PR_SPEC_ENABLE:
388 		seq_puts(m, "always enabled");
389 		break;
390 	case PR_SPEC_DISABLE:
391 		seq_puts(m, "always disabled");
392 		break;
393 	default:
394 		seq_puts(m, "unknown");
395 		break;
396 	}
397 	seq_putc(m, '\n');
398 }
399 
task_context_switch_counts(struct seq_file * m,struct task_struct * p)400 static inline void task_context_switch_counts(struct seq_file *m,
401 						struct task_struct *p)
402 {
403 	seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
404 	seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
405 	seq_putc(m, '\n');
406 }
407 
task_cpus_allowed(struct seq_file * m,struct task_struct * task)408 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
409 {
410 	seq_printf(m, "Cpus_allowed:\t%*pb\n",
411 		   cpumask_pr_args(&task->cpus_mask));
412 	seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
413 		   cpumask_pr_args(&task->cpus_mask));
414 }
415 
task_core_dumping(struct seq_file * m,struct task_struct * task)416 static inline void task_core_dumping(struct seq_file *m, struct task_struct *task)
417 {
418 	seq_put_decimal_ull(m, "CoreDumping:\t", !!task->signal->core_state);
419 	seq_putc(m, '\n');
420 }
421 
task_thp_status(struct seq_file * m,struct mm_struct * mm)422 static inline void task_thp_status(struct seq_file *m, struct mm_struct *mm)
423 {
424 	bool thp_enabled = IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE);
425 
426 	if (thp_enabled)
427 		thp_enabled = !test_bit(MMF_DISABLE_THP, &mm->flags);
428 	seq_printf(m, "THP_enabled:\t%d\n", thp_enabled);
429 }
430 
proc_pid_status(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)431 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
432 			struct pid *pid, struct task_struct *task)
433 {
434 	struct mm_struct *mm = get_task_mm(task);
435 
436 	seq_puts(m, "Name:\t");
437 	proc_task_name(m, task, true);
438 	seq_putc(m, '\n');
439 
440 	task_state(m, ns, pid, task);
441 
442 	if (mm) {
443 		task_mem(m, mm);
444 		task_core_dumping(m, task);
445 		task_thp_status(m, mm);
446 		mmput(mm);
447 	}
448 	task_sig(m, task);
449 	task_cap(m, task);
450 	task_seccomp(m, task);
451 	task_cpus_allowed(m, task);
452 	cpuset_task_status_allowed(m, task);
453 	task_context_switch_counts(m, task);
454 	return 0;
455 }
456 
do_task_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task,int whole)457 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
458 			struct pid *pid, struct task_struct *task, int whole)
459 {
460 	unsigned long vsize, eip, esp, wchan = 0;
461 	int priority, nice;
462 	int tty_pgrp = -1, tty_nr = 0;
463 	sigset_t sigign, sigcatch;
464 	char state;
465 	pid_t ppid = 0, pgid = -1, sid = -1;
466 	int num_threads = 0;
467 	int permitted;
468 	struct mm_struct *mm;
469 	unsigned long long start_time;
470 	unsigned long cmin_flt = 0, cmaj_flt = 0;
471 	unsigned long  min_flt = 0,  maj_flt = 0;
472 	u64 cutime, cstime, utime, stime;
473 	u64 cgtime, gtime;
474 	unsigned long rsslim = 0;
475 	unsigned long flags;
476 	int exit_code = task->exit_code;
477 
478 	state = *get_task_state(task);
479 	vsize = eip = esp = 0;
480 	permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
481 	mm = get_task_mm(task);
482 	if (mm) {
483 		vsize = task_vsize(mm);
484 		/*
485 		 * esp and eip are intentionally zeroed out.  There is no
486 		 * non-racy way to read them without freezing the task.
487 		 * Programs that need reliable values can use ptrace(2).
488 		 *
489 		 * The only exception is if the task is core dumping because
490 		 * a program is not able to use ptrace(2) in that case. It is
491 		 * safe because the task has stopped executing permanently.
492 		 */
493 		if (permitted && (task->flags & (PF_EXITING|PF_DUMPCORE))) {
494 			if (try_get_task_stack(task)) {
495 				eip = KSTK_EIP(task);
496 				esp = KSTK_ESP(task);
497 				put_task_stack(task);
498 			}
499 		}
500 	}
501 
502 	sigemptyset(&sigign);
503 	sigemptyset(&sigcatch);
504 	cutime = cstime = utime = stime = 0;
505 	cgtime = gtime = 0;
506 
507 	if (lock_task_sighand(task, &flags)) {
508 		struct signal_struct *sig = task->signal;
509 
510 		if (sig->tty) {
511 			struct pid *pgrp = tty_get_pgrp(sig->tty);
512 			tty_pgrp = pid_nr_ns(pgrp, ns);
513 			put_pid(pgrp);
514 			tty_nr = new_encode_dev(tty_devnum(sig->tty));
515 		}
516 
517 		num_threads = get_nr_threads(task);
518 		collect_sigign_sigcatch(task, &sigign, &sigcatch);
519 
520 		cmin_flt = sig->cmin_flt;
521 		cmaj_flt = sig->cmaj_flt;
522 		cutime = sig->cutime;
523 		cstime = sig->cstime;
524 		cgtime = sig->cgtime;
525 		rsslim = READ_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
526 
527 		/* add up live thread stats at the group level */
528 		if (whole) {
529 			struct task_struct *t = task;
530 			do {
531 				min_flt += t->min_flt;
532 				maj_flt += t->maj_flt;
533 				gtime += task_gtime(t);
534 			} while_each_thread(task, t);
535 
536 			min_flt += sig->min_flt;
537 			maj_flt += sig->maj_flt;
538 			thread_group_cputime_adjusted(task, &utime, &stime);
539 			gtime += sig->gtime;
540 
541 			if (sig->flags & (SIGNAL_GROUP_EXIT | SIGNAL_STOP_STOPPED))
542 				exit_code = sig->group_exit_code;
543 		}
544 
545 		sid = task_session_nr_ns(task, ns);
546 		ppid = task_tgid_nr_ns(task->real_parent, ns);
547 		pgid = task_pgrp_nr_ns(task, ns);
548 
549 		unlock_task_sighand(task, &flags);
550 	}
551 
552 	if (permitted && (!whole || num_threads < 2))
553 		wchan = !task_is_running(task);
554 	if (!whole) {
555 		min_flt = task->min_flt;
556 		maj_flt = task->maj_flt;
557 		task_cputime_adjusted(task, &utime, &stime);
558 		gtime = task_gtime(task);
559 	}
560 
561 	/* scale priority and nice values from timeslices to -20..20 */
562 	/* to make it look like a "normal" Unix priority/nice value  */
563 	priority = task_prio(task);
564 	nice = task_nice(task);
565 
566 	/* apply timens offset for boottime and convert nsec -> ticks */
567 	start_time =
568 		nsec_to_clock_t(timens_add_boottime_ns(task->start_boottime));
569 
570 	seq_put_decimal_ull(m, "", pid_nr_ns(pid, ns));
571 	seq_puts(m, " (");
572 	proc_task_name(m, task, false);
573 	seq_puts(m, ") ");
574 	seq_putc(m, state);
575 	seq_put_decimal_ll(m, " ", ppid);
576 	seq_put_decimal_ll(m, " ", pgid);
577 	seq_put_decimal_ll(m, " ", sid);
578 	seq_put_decimal_ll(m, " ", tty_nr);
579 	seq_put_decimal_ll(m, " ", tty_pgrp);
580 	seq_put_decimal_ull(m, " ", task->flags);
581 	seq_put_decimal_ull(m, " ", min_flt);
582 	seq_put_decimal_ull(m, " ", cmin_flt);
583 	seq_put_decimal_ull(m, " ", maj_flt);
584 	seq_put_decimal_ull(m, " ", cmaj_flt);
585 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(utime));
586 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(stime));
587 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cutime));
588 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cstime));
589 	seq_put_decimal_ll(m, " ", priority);
590 	seq_put_decimal_ll(m, " ", nice);
591 	seq_put_decimal_ll(m, " ", num_threads);
592 	seq_put_decimal_ull(m, " ", 0);
593 	seq_put_decimal_ull(m, " ", start_time);
594 	seq_put_decimal_ull(m, " ", vsize);
595 	seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
596 	seq_put_decimal_ull(m, " ", rsslim);
597 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
598 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
599 	seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
600 	seq_put_decimal_ull(m, " ", esp);
601 	seq_put_decimal_ull(m, " ", eip);
602 	/* The signal information here is obsolete.
603 	 * It must be decimal for Linux 2.0 compatibility.
604 	 * Use /proc/#/status for real-time signals.
605 	 */
606 	seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
607 	seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
608 	seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
609 	seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
610 
611 	/*
612 	 * We used to output the absolute kernel address, but that's an
613 	 * information leak - so instead we show a 0/1 flag here, to signal
614 	 * to user-space whether there's a wchan field in /proc/PID/wchan.
615 	 *
616 	 * This works with older implementations of procps as well.
617 	 */
618 	seq_put_decimal_ull(m, " ", wchan);
619 
620 	seq_put_decimal_ull(m, " ", 0);
621 	seq_put_decimal_ull(m, " ", 0);
622 	seq_put_decimal_ll(m, " ", task->exit_signal);
623 	seq_put_decimal_ll(m, " ", task_cpu(task));
624 	seq_put_decimal_ull(m, " ", task->rt_priority);
625 	seq_put_decimal_ull(m, " ", task->policy);
626 	seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
627 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(gtime));
628 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cgtime));
629 
630 	if (mm && permitted) {
631 		seq_put_decimal_ull(m, " ", mm->start_data);
632 		seq_put_decimal_ull(m, " ", mm->end_data);
633 		seq_put_decimal_ull(m, " ", mm->start_brk);
634 		seq_put_decimal_ull(m, " ", mm->arg_start);
635 		seq_put_decimal_ull(m, " ", mm->arg_end);
636 		seq_put_decimal_ull(m, " ", mm->env_start);
637 		seq_put_decimal_ull(m, " ", mm->env_end);
638 	} else
639 		seq_puts(m, " 0 0 0 0 0 0 0");
640 
641 	if (permitted)
642 		seq_put_decimal_ll(m, " ", exit_code);
643 	else
644 		seq_puts(m, " 0");
645 
646 	seq_putc(m, '\n');
647 	if (mm)
648 		mmput(mm);
649 	return 0;
650 }
651 
proc_tid_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)652 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
653 			struct pid *pid, struct task_struct *task)
654 {
655 	return do_task_stat(m, ns, pid, task, 0);
656 }
657 
proc_tgid_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)658 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
659 			struct pid *pid, struct task_struct *task)
660 {
661 	return do_task_stat(m, ns, pid, task, 1);
662 }
663 
proc_pid_statm(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)664 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
665 			struct pid *pid, struct task_struct *task)
666 {
667 	struct mm_struct *mm = get_task_mm(task);
668 
669 	if (mm) {
670 		unsigned long size;
671 		unsigned long resident = 0;
672 		unsigned long shared = 0;
673 		unsigned long text = 0;
674 		unsigned long data = 0;
675 
676 		size = task_statm(mm, &shared, &text, &data, &resident);
677 		mmput(mm);
678 
679 		/*
680 		 * For quick read, open code by putting numbers directly
681 		 * expected format is
682 		 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
683 		 *               size, resident, shared, text, data);
684 		 */
685 		seq_put_decimal_ull(m, "", size);
686 		seq_put_decimal_ull(m, " ", resident);
687 		seq_put_decimal_ull(m, " ", shared);
688 		seq_put_decimal_ull(m, " ", text);
689 		seq_put_decimal_ull(m, " ", 0);
690 		seq_put_decimal_ull(m, " ", data);
691 		seq_put_decimal_ull(m, " ", 0);
692 		seq_putc(m, '\n');
693 	} else {
694 		seq_write(m, "0 0 0 0 0 0 0\n", 14);
695 	}
696 	return 0;
697 }
698 
699 #ifdef CONFIG_PROC_CHILDREN
700 static struct pid *
get_children_pid(struct inode * inode,struct pid * pid_prev,loff_t pos)701 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
702 {
703 	struct task_struct *start, *task;
704 	struct pid *pid = NULL;
705 
706 	read_lock(&tasklist_lock);
707 
708 	start = pid_task(proc_pid(inode), PIDTYPE_PID);
709 	if (!start)
710 		goto out;
711 
712 	/*
713 	 * Lets try to continue searching first, this gives
714 	 * us significant speedup on children-rich processes.
715 	 */
716 	if (pid_prev) {
717 		task = pid_task(pid_prev, PIDTYPE_PID);
718 		if (task && task->real_parent == start &&
719 		    !(list_empty(&task->sibling))) {
720 			if (list_is_last(&task->sibling, &start->children))
721 				goto out;
722 			task = list_first_entry(&task->sibling,
723 						struct task_struct, sibling);
724 			pid = get_pid(task_pid(task));
725 			goto out;
726 		}
727 	}
728 
729 	/*
730 	 * Slow search case.
731 	 *
732 	 * We might miss some children here if children
733 	 * are exited while we were not holding the lock,
734 	 * but it was never promised to be accurate that
735 	 * much.
736 	 *
737 	 * "Just suppose that the parent sleeps, but N children
738 	 *  exit after we printed their tids. Now the slow paths
739 	 *  skips N extra children, we miss N tasks." (c)
740 	 *
741 	 * So one need to stop or freeze the leader and all
742 	 * its children to get a precise result.
743 	 */
744 	list_for_each_entry(task, &start->children, sibling) {
745 		if (pos-- == 0) {
746 			pid = get_pid(task_pid(task));
747 			break;
748 		}
749 	}
750 
751 out:
752 	read_unlock(&tasklist_lock);
753 	return pid;
754 }
755 
children_seq_show(struct seq_file * seq,void * v)756 static int children_seq_show(struct seq_file *seq, void *v)
757 {
758 	struct inode *inode = file_inode(seq->file);
759 
760 	seq_printf(seq, "%d ", pid_nr_ns(v, proc_pid_ns(inode->i_sb)));
761 	return 0;
762 }
763 
children_seq_start(struct seq_file * seq,loff_t * pos)764 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
765 {
766 	return get_children_pid(file_inode(seq->file), NULL, *pos);
767 }
768 
children_seq_next(struct seq_file * seq,void * v,loff_t * pos)769 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
770 {
771 	struct pid *pid;
772 
773 	pid = get_children_pid(file_inode(seq->file), v, *pos + 1);
774 	put_pid(v);
775 
776 	++*pos;
777 	return pid;
778 }
779 
children_seq_stop(struct seq_file * seq,void * v)780 static void children_seq_stop(struct seq_file *seq, void *v)
781 {
782 	put_pid(v);
783 }
784 
785 static const struct seq_operations children_seq_ops = {
786 	.start	= children_seq_start,
787 	.next	= children_seq_next,
788 	.stop	= children_seq_stop,
789 	.show	= children_seq_show,
790 };
791 
children_seq_open(struct inode * inode,struct file * file)792 static int children_seq_open(struct inode *inode, struct file *file)
793 {
794 	return seq_open(file, &children_seq_ops);
795 }
796 
797 const struct file_operations proc_tid_children_operations = {
798 	.open    = children_seq_open,
799 	.read    = seq_read,
800 	.llseek  = seq_lseek,
801 	.release = seq_release,
802 };
803 #endif /* CONFIG_PROC_CHILDREN */
804