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 #include <linux/mmu_context.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 	/*
104 	 * Test before PF_KTHREAD because all workqueue worker threads are
105 	 * kernel threads.
106 	 */
107 	if (p->flags & PF_WQ_WORKER)
108 		wq_worker_comm(tcomm, sizeof(tcomm), p);
109 	else if (p->flags & PF_KTHREAD)
110 		get_kthread_comm(tcomm, sizeof(tcomm), p);
111 	else
112 		__get_task_comm(tcomm, sizeof(tcomm), p);
113 
114 	if (escape)
115 		seq_escape_str(m, tcomm, ESCAPE_SPACE | ESCAPE_SPECIAL, "\n\\");
116 	else
117 		seq_printf(m, "%.64s", tcomm);
118 }
119 
120 /*
121  * The task state array is a strange "bitmap" of
122  * reasons to sleep. Thus "running" is zero, and
123  * you can test for combinations of others with
124  * simple bit tests.
125  */
126 static const char * const task_state_array[] = {
127 
128 	/* states in TASK_REPORT: */
129 	"R (running)",		/* 0x00 */
130 	"S (sleeping)",		/* 0x01 */
131 	"D (disk sleep)",	/* 0x02 */
132 	"T (stopped)",		/* 0x04 */
133 	"t (tracing stop)",	/* 0x08 */
134 	"X (dead)",		/* 0x10 */
135 	"Z (zombie)",		/* 0x20 */
136 	"P (parked)",		/* 0x40 */
137 
138 	/* states beyond TASK_REPORT: */
139 	"I (idle)",		/* 0x80 */
140 };
141 
get_task_state(struct task_struct * tsk)142 static inline const char *get_task_state(struct task_struct *tsk)
143 {
144 	BUILD_BUG_ON(1 + ilog2(TASK_REPORT_MAX) != ARRAY_SIZE(task_state_array));
145 	return task_state_array[task_state_index(tsk)];
146 }
147 
task_state(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * p)148 static inline void task_state(struct seq_file *m, struct pid_namespace *ns,
149 				struct pid *pid, struct task_struct *p)
150 {
151 	struct user_namespace *user_ns = seq_user_ns(m);
152 	struct group_info *group_info;
153 	int g, umask = -1;
154 	struct task_struct *tracer;
155 	const struct cred *cred;
156 	pid_t ppid, tpid = 0, tgid, ngid;
157 	unsigned int max_fds = 0;
158 
159 	rcu_read_lock();
160 	ppid = pid_alive(p) ?
161 		task_tgid_nr_ns(rcu_dereference(p->real_parent), ns) : 0;
162 
163 	tracer = ptrace_parent(p);
164 	if (tracer)
165 		tpid = task_pid_nr_ns(tracer, ns);
166 
167 	tgid = task_tgid_nr_ns(p, ns);
168 	ngid = task_numa_group_id(p);
169 	cred = get_task_cred(p);
170 
171 	task_lock(p);
172 	if (p->fs)
173 		umask = p->fs->umask;
174 	if (p->files)
175 		max_fds = files_fdtable(p->files)->max_fds;
176 	task_unlock(p);
177 	rcu_read_unlock();
178 
179 	if (umask >= 0)
180 		seq_printf(m, "Umask:\t%#04o\n", umask);
181 	seq_puts(m, "State:\t");
182 	seq_puts(m, get_task_state(p));
183 
184 	seq_put_decimal_ull(m, "\nTgid:\t", tgid);
185 	seq_put_decimal_ull(m, "\nNgid:\t", ngid);
186 	seq_put_decimal_ull(m, "\nPid:\t", pid_nr_ns(pid, ns));
187 	seq_put_decimal_ull(m, "\nPPid:\t", ppid);
188 	seq_put_decimal_ull(m, "\nTracerPid:\t", tpid);
189 	seq_put_decimal_ull(m, "\nUid:\t", from_kuid_munged(user_ns, cred->uid));
190 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->euid));
191 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->suid));
192 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->fsuid));
193 	seq_put_decimal_ull(m, "\nGid:\t", from_kgid_munged(user_ns, cred->gid));
194 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->egid));
195 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->sgid));
196 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->fsgid));
197 	seq_put_decimal_ull(m, "\nFDSize:\t", max_fds);
198 
199 	seq_puts(m, "\nGroups:\t");
200 	group_info = cred->group_info;
201 	for (g = 0; g < group_info->ngroups; g++)
202 		seq_put_decimal_ull(m, g ? " " : "",
203 				from_kgid_munged(user_ns, group_info->gid[g]));
204 	put_cred(cred);
205 	/* Trailing space shouldn't have been added in the first place. */
206 	seq_putc(m, ' ');
207 
208 #ifdef CONFIG_PID_NS
209 	seq_puts(m, "\nNStgid:");
210 	for (g = ns->level; g <= pid->level; g++)
211 		seq_put_decimal_ull(m, "\t", task_tgid_nr_ns(p, pid->numbers[g].ns));
212 	seq_puts(m, "\nNSpid:");
213 	for (g = ns->level; g <= pid->level; g++)
214 		seq_put_decimal_ull(m, "\t", task_pid_nr_ns(p, pid->numbers[g].ns));
215 	seq_puts(m, "\nNSpgid:");
216 	for (g = ns->level; g <= pid->level; g++)
217 		seq_put_decimal_ull(m, "\t", task_pgrp_nr_ns(p, pid->numbers[g].ns));
218 	seq_puts(m, "\nNSsid:");
219 	for (g = ns->level; g <= pid->level; g++)
220 		seq_put_decimal_ull(m, "\t", task_session_nr_ns(p, pid->numbers[g].ns));
221 #endif
222 	seq_putc(m, '\n');
223 
224 	seq_printf(m, "Kthread:\t%c\n", p->flags & PF_KTHREAD ? '1' : '0');
225 }
226 
render_sigset_t(struct seq_file * m,const char * header,sigset_t * set)227 void render_sigset_t(struct seq_file *m, const char *header,
228 				sigset_t *set)
229 {
230 	int i;
231 
232 	seq_puts(m, header);
233 
234 	i = _NSIG;
235 	do {
236 		int x = 0;
237 
238 		i -= 4;
239 		if (sigismember(set, i+1)) x |= 1;
240 		if (sigismember(set, i+2)) x |= 2;
241 		if (sigismember(set, i+3)) x |= 4;
242 		if (sigismember(set, i+4)) x |= 8;
243 		seq_putc(m, hex_asc[x]);
244 	} while (i >= 4);
245 
246 	seq_putc(m, '\n');
247 }
248 
collect_sigign_sigcatch(struct task_struct * p,sigset_t * sigign,sigset_t * sigcatch)249 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *sigign,
250 				    sigset_t *sigcatch)
251 {
252 	struct k_sigaction *k;
253 	int i;
254 
255 	k = p->sighand->action;
256 	for (i = 1; i <= _NSIG; ++i, ++k) {
257 		if (k->sa.sa_handler == SIG_IGN)
258 			sigaddset(sigign, i);
259 		else if (k->sa.sa_handler != SIG_DFL)
260 			sigaddset(sigcatch, i);
261 	}
262 }
263 
task_sig(struct seq_file * m,struct task_struct * p)264 static inline void task_sig(struct seq_file *m, struct task_struct *p)
265 {
266 	unsigned long flags;
267 	sigset_t pending, shpending, blocked, ignored, caught;
268 	int num_threads = 0;
269 	unsigned int qsize = 0;
270 	unsigned long qlim = 0;
271 
272 	sigemptyset(&pending);
273 	sigemptyset(&shpending);
274 	sigemptyset(&blocked);
275 	sigemptyset(&ignored);
276 	sigemptyset(&caught);
277 
278 	if (lock_task_sighand(p, &flags)) {
279 		pending = p->pending.signal;
280 		shpending = p->signal->shared_pending.signal;
281 		blocked = p->blocked;
282 		collect_sigign_sigcatch(p, &ignored, &caught);
283 		num_threads = get_nr_threads(p);
284 		rcu_read_lock();  /* FIXME: is this correct? */
285 		qsize = get_rlimit_value(task_ucounts(p), UCOUNT_RLIMIT_SIGPENDING);
286 		rcu_read_unlock();
287 		qlim = task_rlimit(p, RLIMIT_SIGPENDING);
288 		unlock_task_sighand(p, &flags);
289 	}
290 
291 	seq_put_decimal_ull(m, "Threads:\t", num_threads);
292 	seq_put_decimal_ull(m, "\nSigQ:\t", qsize);
293 	seq_put_decimal_ull(m, "/", qlim);
294 
295 	/* render them all */
296 	render_sigset_t(m, "\nSigPnd:\t", &pending);
297 	render_sigset_t(m, "ShdPnd:\t", &shpending);
298 	render_sigset_t(m, "SigBlk:\t", &blocked);
299 	render_sigset_t(m, "SigIgn:\t", &ignored);
300 	render_sigset_t(m, "SigCgt:\t", &caught);
301 }
302 
render_cap_t(struct seq_file * m,const char * header,kernel_cap_t * a)303 static void render_cap_t(struct seq_file *m, const char *header,
304 			kernel_cap_t *a)
305 {
306 	seq_puts(m, header);
307 	seq_put_hex_ll(m, NULL, a->val, 16);
308 	seq_putc(m, '\n');
309 }
310 
task_cap(struct seq_file * m,struct task_struct * p)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 
task_seccomp(struct seq_file * m,struct task_struct * p)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 #ifdef CONFIG_SECCOMP_FILTER
339 	seq_put_decimal_ull(m, "\nSeccomp_filters:\t",
340 			    atomic_read(&p->seccomp.filter_count));
341 #endif
342 #endif
343 	seq_puts(m, "\nSpeculation_Store_Bypass:\t");
344 	switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_STORE_BYPASS)) {
345 	case -EINVAL:
346 		seq_puts(m, "unknown");
347 		break;
348 	case PR_SPEC_NOT_AFFECTED:
349 		seq_puts(m, "not vulnerable");
350 		break;
351 	case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
352 		seq_puts(m, "thread force mitigated");
353 		break;
354 	case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
355 		seq_puts(m, "thread mitigated");
356 		break;
357 	case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
358 		seq_puts(m, "thread vulnerable");
359 		break;
360 	case PR_SPEC_DISABLE:
361 		seq_puts(m, "globally mitigated");
362 		break;
363 	default:
364 		seq_puts(m, "vulnerable");
365 		break;
366 	}
367 
368 	seq_puts(m, "\nSpeculationIndirectBranch:\t");
369 	switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_INDIRECT_BRANCH)) {
370 	case -EINVAL:
371 		seq_puts(m, "unsupported");
372 		break;
373 	case PR_SPEC_NOT_AFFECTED:
374 		seq_puts(m, "not affected");
375 		break;
376 	case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
377 		seq_puts(m, "conditional force disabled");
378 		break;
379 	case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
380 		seq_puts(m, "conditional disabled");
381 		break;
382 	case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
383 		seq_puts(m, "conditional enabled");
384 		break;
385 	case PR_SPEC_ENABLE:
386 		seq_puts(m, "always enabled");
387 		break;
388 	case PR_SPEC_DISABLE:
389 		seq_puts(m, "always disabled");
390 		break;
391 	default:
392 		seq_puts(m, "unknown");
393 		break;
394 	}
395 	seq_putc(m, '\n');
396 }
397 
task_context_switch_counts(struct seq_file * m,struct task_struct * p)398 static inline void task_context_switch_counts(struct seq_file *m,
399 						struct task_struct *p)
400 {
401 	seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
402 	seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
403 	seq_putc(m, '\n');
404 }
405 
task_cpus_allowed(struct seq_file * m,struct task_struct * task)406 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
407 {
408 	seq_printf(m, "Cpus_allowed:\t%*pb\n",
409 		   cpumask_pr_args(&task->cpus_mask));
410 	seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
411 		   cpumask_pr_args(&task->cpus_mask));
412 }
413 
task_core_dumping(struct seq_file * m,struct task_struct * task)414 static inline void task_core_dumping(struct seq_file *m, struct task_struct *task)
415 {
416 	seq_put_decimal_ull(m, "CoreDumping:\t", !!task->signal->core_state);
417 	seq_putc(m, '\n');
418 }
419 
task_thp_status(struct seq_file * m,struct mm_struct * mm)420 static inline void task_thp_status(struct seq_file *m, struct mm_struct *mm)
421 {
422 	bool thp_enabled = IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE);
423 
424 	if (thp_enabled)
425 		thp_enabled = !test_bit(MMF_DISABLE_THP, &mm->flags);
426 	seq_printf(m, "THP_enabled:\t%d\n", thp_enabled);
427 }
428 
task_untag_mask(struct seq_file * m,struct mm_struct * mm)429 static inline void task_untag_mask(struct seq_file *m, struct mm_struct *mm)
430 {
431 	seq_printf(m, "untag_mask:\t%#lx\n", mm_untag_mask(mm));
432 }
433 
arch_proc_pid_thread_features(struct seq_file * m,struct task_struct * task)434 __weak void arch_proc_pid_thread_features(struct seq_file *m,
435 					  struct task_struct *task)
436 {
437 }
438 
proc_pid_status(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)439 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
440 			struct pid *pid, struct task_struct *task)
441 {
442 	struct mm_struct *mm = get_task_mm(task);
443 
444 	seq_puts(m, "Name:\t");
445 	proc_task_name(m, task, true);
446 	seq_putc(m, '\n');
447 
448 	task_state(m, ns, pid, task);
449 
450 	if (mm) {
451 		task_mem(m, mm);
452 		task_core_dumping(m, task);
453 		task_thp_status(m, mm);
454 		task_untag_mask(m, mm);
455 		mmput(mm);
456 	}
457 	task_sig(m, task);
458 	task_cap(m, task);
459 	task_seccomp(m, task);
460 	task_cpus_allowed(m, task);
461 	cpuset_task_status_allowed(m, task);
462 	task_context_switch_counts(m, task);
463 	arch_proc_pid_thread_features(m, task);
464 	return 0;
465 }
466 
do_task_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task,int whole)467 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
468 			struct pid *pid, struct task_struct *task, int whole)
469 {
470 	unsigned long vsize, eip, esp, wchan = 0;
471 	int priority, nice;
472 	int tty_pgrp = -1, tty_nr = 0;
473 	sigset_t sigign, sigcatch;
474 	char state;
475 	pid_t ppid = 0, pgid = -1, sid = -1;
476 	int num_threads = 0;
477 	int permitted;
478 	struct mm_struct *mm;
479 	unsigned long long start_time;
480 	unsigned long cmin_flt, cmaj_flt, min_flt, maj_flt;
481 	u64 cutime, cstime, cgtime, utime, stime, gtime;
482 	unsigned long rsslim = 0;
483 	unsigned long flags;
484 	int exit_code = task->exit_code;
485 	struct signal_struct *sig = task->signal;
486 	unsigned int seq = 1;
487 
488 	state = *get_task_state(task);
489 	vsize = eip = esp = 0;
490 	permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
491 	mm = get_task_mm(task);
492 	if (mm) {
493 		vsize = task_vsize(mm);
494 		/*
495 		 * esp and eip are intentionally zeroed out.  There is no
496 		 * non-racy way to read them without freezing the task.
497 		 * Programs that need reliable values can use ptrace(2).
498 		 *
499 		 * The only exception is if the task is core dumping because
500 		 * a program is not able to use ptrace(2) in that case. It is
501 		 * safe because the task has stopped executing permanently.
502 		 */
503 		if (permitted && (task->flags & (PF_EXITING|PF_DUMPCORE))) {
504 			if (try_get_task_stack(task)) {
505 				eip = KSTK_EIP(task);
506 				esp = KSTK_ESP(task);
507 				put_task_stack(task);
508 			}
509 		}
510 	}
511 
512 	sigemptyset(&sigign);
513 	sigemptyset(&sigcatch);
514 
515 	if (lock_task_sighand(task, &flags)) {
516 		if (sig->tty) {
517 			struct pid *pgrp = tty_get_pgrp(sig->tty);
518 			tty_pgrp = pid_nr_ns(pgrp, ns);
519 			put_pid(pgrp);
520 			tty_nr = new_encode_dev(tty_devnum(sig->tty));
521 		}
522 
523 		num_threads = get_nr_threads(task);
524 		collect_sigign_sigcatch(task, &sigign, &sigcatch);
525 
526 		rsslim = READ_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
527 
528 		if (whole) {
529 			if (sig->flags & (SIGNAL_GROUP_EXIT | SIGNAL_STOP_STOPPED))
530 				exit_code = sig->group_exit_code;
531 		}
532 
533 		sid = task_session_nr_ns(task, ns);
534 		ppid = task_tgid_nr_ns(task->real_parent, ns);
535 		pgid = task_pgrp_nr_ns(task, ns);
536 
537 		unlock_task_sighand(task, &flags);
538 	}
539 
540 	if (permitted && (!whole || num_threads < 2))
541 		wchan = !task_is_running(task);
542 
543 	do {
544 		seq++; /* 2 on the 1st/lockless path, otherwise odd */
545 		flags = read_seqbegin_or_lock_irqsave(&sig->stats_lock, &seq);
546 
547 		cmin_flt = sig->cmin_flt;
548 		cmaj_flt = sig->cmaj_flt;
549 		cutime = sig->cutime;
550 		cstime = sig->cstime;
551 		cgtime = sig->cgtime;
552 
553 		if (whole) {
554 			struct task_struct *t;
555 
556 			min_flt = sig->min_flt;
557 			maj_flt = sig->maj_flt;
558 			gtime = sig->gtime;
559 
560 			rcu_read_lock();
561 			__for_each_thread(sig, t) {
562 				min_flt += t->min_flt;
563 				maj_flt += t->maj_flt;
564 				gtime += task_gtime(t);
565 			}
566 			rcu_read_unlock();
567 		}
568 	} while (need_seqretry(&sig->stats_lock, seq));
569 	done_seqretry_irqrestore(&sig->stats_lock, seq, flags);
570 
571 	if (whole) {
572 		thread_group_cputime_adjusted(task, &utime, &stime);
573 	} else {
574 		task_cputime_adjusted(task, &utime, &stime);
575 		min_flt = task->min_flt;
576 		maj_flt = task->maj_flt;
577 		gtime = task_gtime(task);
578 	}
579 
580 	/* scale priority and nice values from timeslices to -20..20 */
581 	/* to make it look like a "normal" Unix priority/nice value  */
582 	priority = task_prio(task);
583 	nice = task_nice(task);
584 
585 	/* apply timens offset for boottime and convert nsec -> ticks */
586 	start_time =
587 		nsec_to_clock_t(timens_add_boottime_ns(task->start_boottime));
588 
589 	seq_put_decimal_ull(m, "", pid_nr_ns(pid, ns));
590 	seq_puts(m, " (");
591 	proc_task_name(m, task, false);
592 	seq_puts(m, ") ");
593 	seq_putc(m, state);
594 	seq_put_decimal_ll(m, " ", ppid);
595 	seq_put_decimal_ll(m, " ", pgid);
596 	seq_put_decimal_ll(m, " ", sid);
597 	seq_put_decimal_ll(m, " ", tty_nr);
598 	seq_put_decimal_ll(m, " ", tty_pgrp);
599 	seq_put_decimal_ull(m, " ", task->flags);
600 	seq_put_decimal_ull(m, " ", min_flt);
601 	seq_put_decimal_ull(m, " ", cmin_flt);
602 	seq_put_decimal_ull(m, " ", maj_flt);
603 	seq_put_decimal_ull(m, " ", cmaj_flt);
604 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(utime));
605 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(stime));
606 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cutime));
607 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cstime));
608 	seq_put_decimal_ll(m, " ", priority);
609 	seq_put_decimal_ll(m, " ", nice);
610 	seq_put_decimal_ll(m, " ", num_threads);
611 	seq_put_decimal_ull(m, " ", 0);
612 	seq_put_decimal_ull(m, " ", start_time);
613 	seq_put_decimal_ull(m, " ", vsize);
614 	seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
615 	seq_put_decimal_ull(m, " ", rsslim);
616 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
617 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
618 	seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
619 	seq_put_decimal_ull(m, " ", esp);
620 	seq_put_decimal_ull(m, " ", eip);
621 	/* The signal information here is obsolete.
622 	 * It must be decimal for Linux 2.0 compatibility.
623 	 * Use /proc/#/status for real-time signals.
624 	 */
625 	seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
626 	seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
627 	seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
628 	seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
629 
630 	/*
631 	 * We used to output the absolute kernel address, but that's an
632 	 * information leak - so instead we show a 0/1 flag here, to signal
633 	 * to user-space whether there's a wchan field in /proc/PID/wchan.
634 	 *
635 	 * This works with older implementations of procps as well.
636 	 */
637 	seq_put_decimal_ull(m, " ", wchan);
638 
639 	seq_put_decimal_ull(m, " ", 0);
640 	seq_put_decimal_ull(m, " ", 0);
641 	seq_put_decimal_ll(m, " ", task->exit_signal);
642 	seq_put_decimal_ll(m, " ", task_cpu(task));
643 	seq_put_decimal_ull(m, " ", task->rt_priority);
644 	seq_put_decimal_ull(m, " ", task->policy);
645 	seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
646 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(gtime));
647 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cgtime));
648 
649 	if (mm && permitted) {
650 		seq_put_decimal_ull(m, " ", mm->start_data);
651 		seq_put_decimal_ull(m, " ", mm->end_data);
652 		seq_put_decimal_ull(m, " ", mm->start_brk);
653 		seq_put_decimal_ull(m, " ", mm->arg_start);
654 		seq_put_decimal_ull(m, " ", mm->arg_end);
655 		seq_put_decimal_ull(m, " ", mm->env_start);
656 		seq_put_decimal_ull(m, " ", mm->env_end);
657 	} else
658 		seq_puts(m, " 0 0 0 0 0 0 0");
659 
660 	if (permitted)
661 		seq_put_decimal_ll(m, " ", exit_code);
662 	else
663 		seq_puts(m, " 0");
664 
665 	seq_putc(m, '\n');
666 	if (mm)
667 		mmput(mm);
668 	return 0;
669 }
670 
proc_tid_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)671 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
672 			struct pid *pid, struct task_struct *task)
673 {
674 	return do_task_stat(m, ns, pid, task, 0);
675 }
676 
proc_tgid_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)677 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
678 			struct pid *pid, struct task_struct *task)
679 {
680 	return do_task_stat(m, ns, pid, task, 1);
681 }
682 
proc_pid_statm(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)683 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
684 			struct pid *pid, struct task_struct *task)
685 {
686 	struct mm_struct *mm = get_task_mm(task);
687 
688 	if (mm) {
689 		unsigned long size;
690 		unsigned long resident = 0;
691 		unsigned long shared = 0;
692 		unsigned long text = 0;
693 		unsigned long data = 0;
694 
695 		size = task_statm(mm, &shared, &text, &data, &resident);
696 		mmput(mm);
697 
698 		/*
699 		 * For quick read, open code by putting numbers directly
700 		 * expected format is
701 		 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
702 		 *               size, resident, shared, text, data);
703 		 */
704 		seq_put_decimal_ull(m, "", size);
705 		seq_put_decimal_ull(m, " ", resident);
706 		seq_put_decimal_ull(m, " ", shared);
707 		seq_put_decimal_ull(m, " ", text);
708 		seq_put_decimal_ull(m, " ", 0);
709 		seq_put_decimal_ull(m, " ", data);
710 		seq_put_decimal_ull(m, " ", 0);
711 		seq_putc(m, '\n');
712 	} else {
713 		seq_write(m, "0 0 0 0 0 0 0\n", 14);
714 	}
715 	return 0;
716 }
717 
718 #ifdef CONFIG_PROC_CHILDREN
719 static struct pid *
get_children_pid(struct inode * inode,struct pid * pid_prev,loff_t pos)720 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
721 {
722 	struct task_struct *start, *task;
723 	struct pid *pid = NULL;
724 
725 	read_lock(&tasklist_lock);
726 
727 	start = pid_task(proc_pid(inode), PIDTYPE_PID);
728 	if (!start)
729 		goto out;
730 
731 	/*
732 	 * Lets try to continue searching first, this gives
733 	 * us significant speedup on children-rich processes.
734 	 */
735 	if (pid_prev) {
736 		task = pid_task(pid_prev, PIDTYPE_PID);
737 		if (task && task->real_parent == start &&
738 		    !(list_empty(&task->sibling))) {
739 			if (list_is_last(&task->sibling, &start->children))
740 				goto out;
741 			task = list_first_entry(&task->sibling,
742 						struct task_struct, sibling);
743 			pid = get_pid(task_pid(task));
744 			goto out;
745 		}
746 	}
747 
748 	/*
749 	 * Slow search case.
750 	 *
751 	 * We might miss some children here if children
752 	 * are exited while we were not holding the lock,
753 	 * but it was never promised to be accurate that
754 	 * much.
755 	 *
756 	 * "Just suppose that the parent sleeps, but N children
757 	 *  exit after we printed their tids. Now the slow paths
758 	 *  skips N extra children, we miss N tasks." (c)
759 	 *
760 	 * So one need to stop or freeze the leader and all
761 	 * its children to get a precise result.
762 	 */
763 	list_for_each_entry(task, &start->children, sibling) {
764 		if (pos-- == 0) {
765 			pid = get_pid(task_pid(task));
766 			break;
767 		}
768 	}
769 
770 out:
771 	read_unlock(&tasklist_lock);
772 	return pid;
773 }
774 
children_seq_show(struct seq_file * seq,void * v)775 static int children_seq_show(struct seq_file *seq, void *v)
776 {
777 	struct inode *inode = file_inode(seq->file);
778 
779 	seq_printf(seq, "%d ", pid_nr_ns(v, proc_pid_ns(inode->i_sb)));
780 	return 0;
781 }
782 
children_seq_start(struct seq_file * seq,loff_t * pos)783 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
784 {
785 	return get_children_pid(file_inode(seq->file), NULL, *pos);
786 }
787 
children_seq_next(struct seq_file * seq,void * v,loff_t * pos)788 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
789 {
790 	struct pid *pid;
791 
792 	pid = get_children_pid(file_inode(seq->file), v, *pos + 1);
793 	put_pid(v);
794 
795 	++*pos;
796 	return pid;
797 }
798 
children_seq_stop(struct seq_file * seq,void * v)799 static void children_seq_stop(struct seq_file *seq, void *v)
800 {
801 	put_pid(v);
802 }
803 
804 static const struct seq_operations children_seq_ops = {
805 	.start	= children_seq_start,
806 	.next	= children_seq_next,
807 	.stop	= children_seq_stop,
808 	.show	= children_seq_show,
809 };
810 
children_seq_open(struct inode * inode,struct file * file)811 static int children_seq_open(struct inode *inode, struct file *file)
812 {
813 	return seq_open(file, &children_seq_ops);
814 }
815 
816 const struct file_operations proc_tid_children_operations = {
817 	.open    = children_seq_open,
818 	.read    = seq_read,
819 	.llseek  = seq_lseek,
820 	.release = seq_release,
821 };
822 #endif /* CONFIG_PROC_CHILDREN */
823