xref: /linux/fs/proc/array.c (revision 8804d970fab45726b3c7cd7f240b31122aa94219)
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, 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 	tracer = ptrace_parent(p);
161 	if (tracer)
162 		tpid = task_pid_nr_ns(tracer, ns);
163 
164 	ppid = task_ppid_nr_ns(p, ns);
165 	tgid = task_tgid_nr_ns(p, ns);
166 	ngid = task_numa_group_id(p);
167 	cred = get_task_cred(p);
168 
169 	task_lock(p);
170 	if (p->fs)
171 		umask = p->fs->umask;
172 	if (p->files)
173 		max_fds = files_fdtable(p->files)->max_fds;
174 	task_unlock(p);
175 	rcu_read_unlock();
176 
177 	if (umask >= 0)
178 		seq_printf(m, "Umask:\t%#04o\n", umask);
179 	seq_puts(m, "State:\t");
180 	seq_puts(m, get_task_state(p));
181 
182 	seq_put_decimal_ull(m, "\nTgid:\t", tgid);
183 	seq_put_decimal_ull(m, "\nNgid:\t", ngid);
184 	seq_put_decimal_ull(m, "\nPid:\t", pid_nr_ns(pid, ns));
185 	seq_put_decimal_ull(m, "\nPPid:\t", ppid);
186 	seq_put_decimal_ull(m, "\nTracerPid:\t", tpid);
187 	seq_put_decimal_ull(m, "\nUid:\t", from_kuid_munged(user_ns, cred->uid));
188 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->euid));
189 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->suid));
190 	seq_put_decimal_ull(m, "\t", from_kuid_munged(user_ns, cred->fsuid));
191 	seq_put_decimal_ull(m, "\nGid:\t", from_kgid_munged(user_ns, cred->gid));
192 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->egid));
193 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->sgid));
194 	seq_put_decimal_ull(m, "\t", from_kgid_munged(user_ns, cred->fsgid));
195 	seq_put_decimal_ull(m, "\nFDSize:\t", max_fds);
196 
197 	seq_puts(m, "\nGroups:\t");
198 	group_info = cred->group_info;
199 	for (g = 0; g < group_info->ngroups; g++)
200 		seq_put_decimal_ull(m, g ? " " : "",
201 				from_kgid_munged(user_ns, group_info->gid[g]));
202 	put_cred(cred);
203 	/* Trailing space shouldn't have been added in the first place. */
204 	seq_putc(m, ' ');
205 
206 #ifdef CONFIG_PID_NS
207 	seq_puts(m, "\nNStgid:");
208 	for (g = ns->level; g <= pid->level; g++)
209 		seq_put_decimal_ull(m, "\t", task_tgid_nr_ns(p, pid->numbers[g].ns));
210 	seq_puts(m, "\nNSpid:");
211 	for (g = ns->level; g <= pid->level; g++)
212 		seq_put_decimal_ull(m, "\t", task_pid_nr_ns(p, pid->numbers[g].ns));
213 	seq_puts(m, "\nNSpgid:");
214 	for (g = ns->level; g <= pid->level; g++)
215 		seq_put_decimal_ull(m, "\t", task_pgrp_nr_ns(p, pid->numbers[g].ns));
216 	seq_puts(m, "\nNSsid:");
217 	for (g = ns->level; g <= pid->level; g++)
218 		seq_put_decimal_ull(m, "\t", task_session_nr_ns(p, pid->numbers[g].ns));
219 #endif
220 	seq_putc(m, '\n');
221 
222 	seq_printf(m, "Kthread:\t%c\n", p->flags & PF_KTHREAD ? '1' : '0');
223 }
224 
render_sigset_t(struct seq_file * m,const char * header,sigset_t * set)225 void render_sigset_t(struct seq_file *m, const char *header,
226 				sigset_t *set)
227 {
228 	int i;
229 
230 	seq_puts(m, header);
231 
232 	i = _NSIG;
233 	do {
234 		int x = 0;
235 
236 		i -= 4;
237 		if (sigismember(set, i+1)) x |= 1;
238 		if (sigismember(set, i+2)) x |= 2;
239 		if (sigismember(set, i+3)) x |= 4;
240 		if (sigismember(set, i+4)) x |= 8;
241 		seq_putc(m, hex_asc[x]);
242 	} while (i >= 4);
243 
244 	seq_putc(m, '\n');
245 }
246 
collect_sigign_sigcatch(struct task_struct * p,sigset_t * sigign,sigset_t * sigcatch)247 static void collect_sigign_sigcatch(struct task_struct *p, sigset_t *sigign,
248 				    sigset_t *sigcatch)
249 {
250 	struct k_sigaction *k;
251 	int i;
252 
253 	k = p->sighand->action;
254 	for (i = 1; i <= _NSIG; ++i, ++k) {
255 		if (k->sa.sa_handler == SIG_IGN)
256 			sigaddset(sigign, i);
257 		else if (k->sa.sa_handler != SIG_DFL)
258 			sigaddset(sigcatch, i);
259 	}
260 }
261 
task_sig(struct seq_file * m,struct task_struct * p)262 static inline void task_sig(struct seq_file *m, struct task_struct *p)
263 {
264 	unsigned long flags;
265 	sigset_t pending, shpending, blocked, ignored, caught;
266 	int num_threads = 0;
267 	unsigned int qsize = 0;
268 	unsigned long qlim = 0;
269 
270 	sigemptyset(&pending);
271 	sigemptyset(&shpending);
272 	sigemptyset(&blocked);
273 	sigemptyset(&ignored);
274 	sigemptyset(&caught);
275 
276 	if (lock_task_sighand(p, &flags)) {
277 		pending = p->pending.signal;
278 		shpending = p->signal->shared_pending.signal;
279 		blocked = p->blocked;
280 		collect_sigign_sigcatch(p, &ignored, &caught);
281 		num_threads = get_nr_threads(p);
282 		rcu_read_lock();  /* FIXME: is this correct? */
283 		qsize = get_rlimit_value(task_ucounts(p), UCOUNT_RLIMIT_SIGPENDING);
284 		rcu_read_unlock();
285 		qlim = task_rlimit(p, RLIMIT_SIGPENDING);
286 		unlock_task_sighand(p, &flags);
287 	}
288 
289 	seq_put_decimal_ull(m, "Threads:\t", num_threads);
290 	seq_put_decimal_ull(m, "\nSigQ:\t", qsize);
291 	seq_put_decimal_ull(m, "/", qlim);
292 
293 	/* render them all */
294 	render_sigset_t(m, "\nSigPnd:\t", &pending);
295 	render_sigset_t(m, "ShdPnd:\t", &shpending);
296 	render_sigset_t(m, "SigBlk:\t", &blocked);
297 	render_sigset_t(m, "SigIgn:\t", &ignored);
298 	render_sigset_t(m, "SigCgt:\t", &caught);
299 }
300 
render_cap_t(struct seq_file * m,const char * header,kernel_cap_t * a)301 static void render_cap_t(struct seq_file *m, const char *header,
302 			kernel_cap_t *a)
303 {
304 	seq_puts(m, header);
305 	seq_put_hex_ll(m, NULL, a->val, 16);
306 	seq_putc(m, '\n');
307 }
308 
task_cap(struct seq_file * m,struct task_struct * p)309 static inline void task_cap(struct seq_file *m, struct task_struct *p)
310 {
311 	const struct cred *cred;
312 	kernel_cap_t cap_inheritable, cap_permitted, cap_effective,
313 			cap_bset, cap_ambient;
314 
315 	rcu_read_lock();
316 	cred = __task_cred(p);
317 	cap_inheritable	= cred->cap_inheritable;
318 	cap_permitted	= cred->cap_permitted;
319 	cap_effective	= cred->cap_effective;
320 	cap_bset	= cred->cap_bset;
321 	cap_ambient	= cred->cap_ambient;
322 	rcu_read_unlock();
323 
324 	render_cap_t(m, "CapInh:\t", &cap_inheritable);
325 	render_cap_t(m, "CapPrm:\t", &cap_permitted);
326 	render_cap_t(m, "CapEff:\t", &cap_effective);
327 	render_cap_t(m, "CapBnd:\t", &cap_bset);
328 	render_cap_t(m, "CapAmb:\t", &cap_ambient);
329 }
330 
task_seccomp(struct seq_file * m,struct task_struct * p)331 static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
332 {
333 	seq_put_decimal_ull(m, "NoNewPrivs:\t", task_no_new_privs(p));
334 #ifdef CONFIG_SECCOMP
335 	seq_put_decimal_ull(m, "\nSeccomp:\t", p->seccomp.mode);
336 #ifdef CONFIG_SECCOMP_FILTER
337 	seq_put_decimal_ull(m, "\nSeccomp_filters:\t",
338 			    atomic_read(&p->seccomp.filter_count));
339 #endif
340 #endif
341 	seq_puts(m, "\nSpeculation_Store_Bypass:\t");
342 	switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_STORE_BYPASS)) {
343 	case -EINVAL:
344 		seq_puts(m, "unknown");
345 		break;
346 	case PR_SPEC_NOT_AFFECTED:
347 		seq_puts(m, "not vulnerable");
348 		break;
349 	case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
350 		seq_puts(m, "thread force mitigated");
351 		break;
352 	case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
353 		seq_puts(m, "thread mitigated");
354 		break;
355 	case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
356 		seq_puts(m, "thread vulnerable");
357 		break;
358 	case PR_SPEC_DISABLE:
359 		seq_puts(m, "globally mitigated");
360 		break;
361 	default:
362 		seq_puts(m, "vulnerable");
363 		break;
364 	}
365 
366 	seq_puts(m, "\nSpeculationIndirectBranch:\t");
367 	switch (arch_prctl_spec_ctrl_get(p, PR_SPEC_INDIRECT_BRANCH)) {
368 	case -EINVAL:
369 		seq_puts(m, "unsupported");
370 		break;
371 	case PR_SPEC_NOT_AFFECTED:
372 		seq_puts(m, "not affected");
373 		break;
374 	case PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE:
375 		seq_puts(m, "conditional force disabled");
376 		break;
377 	case PR_SPEC_PRCTL | PR_SPEC_DISABLE:
378 		seq_puts(m, "conditional disabled");
379 		break;
380 	case PR_SPEC_PRCTL | PR_SPEC_ENABLE:
381 		seq_puts(m, "conditional enabled");
382 		break;
383 	case PR_SPEC_ENABLE:
384 		seq_puts(m, "always enabled");
385 		break;
386 	case PR_SPEC_DISABLE:
387 		seq_puts(m, "always disabled");
388 		break;
389 	default:
390 		seq_puts(m, "unknown");
391 		break;
392 	}
393 	seq_putc(m, '\n');
394 }
395 
task_context_switch_counts(struct seq_file * m,struct task_struct * p)396 static inline void task_context_switch_counts(struct seq_file *m,
397 						struct task_struct *p)
398 {
399 	seq_put_decimal_ull(m, "voluntary_ctxt_switches:\t", p->nvcsw);
400 	seq_put_decimal_ull(m, "\nnonvoluntary_ctxt_switches:\t", p->nivcsw);
401 	seq_putc(m, '\n');
402 }
403 
task_cpus_allowed(struct seq_file * m,struct task_struct * task)404 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)
405 {
406 	seq_printf(m, "Cpus_allowed:\t%*pb\n",
407 		   cpumask_pr_args(&task->cpus_mask));
408 	seq_printf(m, "Cpus_allowed_list:\t%*pbl\n",
409 		   cpumask_pr_args(&task->cpus_mask));
410 }
411 
task_core_dumping(struct seq_file * m,struct task_struct * task)412 static inline void task_core_dumping(struct seq_file *m, struct task_struct *task)
413 {
414 	seq_put_decimal_ull(m, "CoreDumping:\t", !!task->signal->core_state);
415 	seq_putc(m, '\n');
416 }
417 
task_thp_status(struct seq_file * m,struct mm_struct * mm)418 static inline void task_thp_status(struct seq_file *m, struct mm_struct *mm)
419 {
420 	bool thp_enabled = IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE);
421 
422 	if (thp_enabled)
423 		thp_enabled = !mm_flags_test(MMF_DISABLE_THP_COMPLETELY, mm);
424 	seq_printf(m, "THP_enabled:\t%d\n", thp_enabled);
425 }
426 
task_untag_mask(struct seq_file * m,struct mm_struct * mm)427 static inline void task_untag_mask(struct seq_file *m, struct mm_struct *mm)
428 {
429 	seq_printf(m, "untag_mask:\t%#lx\n", mm_untag_mask(mm));
430 }
431 
arch_proc_pid_thread_features(struct seq_file * m,struct task_struct * task)432 __weak void arch_proc_pid_thread_features(struct seq_file *m,
433 					  struct task_struct *task)
434 {
435 }
436 
proc_pid_status(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)437 int proc_pid_status(struct seq_file *m, struct pid_namespace *ns,
438 			struct pid *pid, struct task_struct *task)
439 {
440 	struct mm_struct *mm = get_task_mm(task);
441 
442 	seq_puts(m, "Name:\t");
443 	proc_task_name(m, task, true);
444 	seq_putc(m, '\n');
445 
446 	task_state(m, ns, pid, task);
447 
448 	if (mm) {
449 		task_mem(m, mm);
450 		task_core_dumping(m, task);
451 		task_thp_status(m, mm);
452 		task_untag_mask(m, mm);
453 		mmput(mm);
454 	}
455 	task_sig(m, task);
456 	task_cap(m, task);
457 	task_seccomp(m, task);
458 	task_cpus_allowed(m, task);
459 	cpuset_task_status_allowed(m, task);
460 	task_context_switch_counts(m, task);
461 	arch_proc_pid_thread_features(m, task);
462 	return 0;
463 }
464 
do_task_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task,int whole)465 static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
466 			struct pid *pid, struct task_struct *task, int whole)
467 {
468 	unsigned long vsize, eip, esp, wchan = 0;
469 	int priority, nice;
470 	int tty_pgrp = -1, tty_nr = 0;
471 	sigset_t sigign, sigcatch;
472 	char state;
473 	pid_t ppid = 0, pgid = -1, sid = -1;
474 	int num_threads = 0;
475 	int permitted;
476 	struct mm_struct *mm;
477 	unsigned long long start_time;
478 	unsigned long cmin_flt, cmaj_flt, min_flt, maj_flt;
479 	u64 cutime, cstime, cgtime, utime, stime, gtime;
480 	unsigned long rsslim = 0;
481 	unsigned long flags;
482 	int exit_code = task->exit_code;
483 	struct signal_struct *sig = task->signal;
484 	unsigned int seq = 1;
485 
486 	state = *get_task_state(task);
487 	vsize = eip = esp = 0;
488 	permitted = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS | PTRACE_MODE_NOAUDIT);
489 	mm = get_task_mm(task);
490 	if (mm) {
491 		vsize = task_vsize(mm);
492 		/*
493 		 * esp and eip are intentionally zeroed out.  There is no
494 		 * non-racy way to read them without freezing the task.
495 		 * Programs that need reliable values can use ptrace(2).
496 		 *
497 		 * The only exception is if the task is core dumping because
498 		 * a program is not able to use ptrace(2) in that case. It is
499 		 * safe because the task has stopped executing permanently.
500 		 */
501 		if (permitted && (task->flags & (PF_EXITING|PF_DUMPCORE|PF_POSTCOREDUMP))) {
502 			if (try_get_task_stack(task)) {
503 				eip = KSTK_EIP(task);
504 				esp = KSTK_ESP(task);
505 				put_task_stack(task);
506 			}
507 		}
508 	}
509 
510 	sigemptyset(&sigign);
511 	sigemptyset(&sigcatch);
512 
513 	if (lock_task_sighand(task, &flags)) {
514 		if (sig->tty) {
515 			struct pid *pgrp = tty_get_pgrp(sig->tty);
516 			tty_pgrp = pid_nr_ns(pgrp, ns);
517 			put_pid(pgrp);
518 			tty_nr = new_encode_dev(tty_devnum(sig->tty));
519 		}
520 
521 		num_threads = get_nr_threads(task);
522 		collect_sigign_sigcatch(task, &sigign, &sigcatch);
523 
524 		rsslim = READ_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
525 
526 		if (whole) {
527 			if (sig->flags & (SIGNAL_GROUP_EXIT | SIGNAL_STOP_STOPPED))
528 				exit_code = sig->group_exit_code;
529 		}
530 
531 		sid = task_session_nr_ns(task, ns);
532 		ppid = task_tgid_nr_ns(task->real_parent, ns);
533 		pgid = task_pgrp_nr_ns(task, ns);
534 
535 		unlock_task_sighand(task, &flags);
536 	}
537 
538 	if (permitted && (!whole || num_threads < 2))
539 		wchan = !task_is_running(task);
540 
541 	do {
542 		seq++; /* 2 on the 1st/lockless path, otherwise odd */
543 		flags = read_seqbegin_or_lock_irqsave(&sig->stats_lock, &seq);
544 
545 		cmin_flt = sig->cmin_flt;
546 		cmaj_flt = sig->cmaj_flt;
547 		cutime = sig->cutime;
548 		cstime = sig->cstime;
549 		cgtime = sig->cgtime;
550 
551 		if (whole) {
552 			struct task_struct *t;
553 
554 			min_flt = sig->min_flt;
555 			maj_flt = sig->maj_flt;
556 			gtime = sig->gtime;
557 
558 			rcu_read_lock();
559 			__for_each_thread(sig, t) {
560 				min_flt += t->min_flt;
561 				maj_flt += t->maj_flt;
562 				gtime += task_gtime(t);
563 			}
564 			rcu_read_unlock();
565 		}
566 	} while (need_seqretry(&sig->stats_lock, seq));
567 	done_seqretry_irqrestore(&sig->stats_lock, seq, flags);
568 
569 	if (whole) {
570 		thread_group_cputime_adjusted(task, &utime, &stime);
571 	} else {
572 		task_cputime_adjusted(task, &utime, &stime);
573 		min_flt = task->min_flt;
574 		maj_flt = task->maj_flt;
575 		gtime = task_gtime(task);
576 	}
577 
578 	/* scale priority and nice values from timeslices to -20..20 */
579 	/* to make it look like a "normal" Unix priority/nice value  */
580 	priority = task_prio(task);
581 	nice = task_nice(task);
582 
583 	/* apply timens offset for boottime and convert nsec -> ticks */
584 	start_time =
585 		nsec_to_clock_t(timens_add_boottime_ns(task->start_boottime));
586 
587 	seq_put_decimal_ull(m, "", pid_nr_ns(pid, ns));
588 	seq_puts(m, " (");
589 	proc_task_name(m, task, false);
590 	seq_puts(m, ") ");
591 	seq_putc(m, state);
592 	seq_put_decimal_ll(m, " ", ppid);
593 	seq_put_decimal_ll(m, " ", pgid);
594 	seq_put_decimal_ll(m, " ", sid);
595 	seq_put_decimal_ll(m, " ", tty_nr);
596 	seq_put_decimal_ll(m, " ", tty_pgrp);
597 	seq_put_decimal_ull(m, " ", task->flags);
598 	seq_put_decimal_ull(m, " ", min_flt);
599 	seq_put_decimal_ull(m, " ", cmin_flt);
600 	seq_put_decimal_ull(m, " ", maj_flt);
601 	seq_put_decimal_ull(m, " ", cmaj_flt);
602 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(utime));
603 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(stime));
604 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cutime));
605 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cstime));
606 	seq_put_decimal_ll(m, " ", priority);
607 	seq_put_decimal_ll(m, " ", nice);
608 	seq_put_decimal_ll(m, " ", num_threads);
609 	seq_put_decimal_ull(m, " ", 0);
610 	seq_put_decimal_ull(m, " ", start_time);
611 	seq_put_decimal_ull(m, " ", vsize);
612 	seq_put_decimal_ull(m, " ", mm ? get_mm_rss(mm) : 0);
613 	seq_put_decimal_ull(m, " ", rsslim);
614 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->start_code : 1) : 0);
615 	seq_put_decimal_ull(m, " ", mm ? (permitted ? mm->end_code : 1) : 0);
616 	seq_put_decimal_ull(m, " ", (permitted && mm) ? mm->start_stack : 0);
617 	seq_put_decimal_ull(m, " ", esp);
618 	seq_put_decimal_ull(m, " ", eip);
619 	/* The signal information here is obsolete.
620 	 * It must be decimal for Linux 2.0 compatibility.
621 	 * Use /proc/#/status for real-time signals.
622 	 */
623 	seq_put_decimal_ull(m, " ", task->pending.signal.sig[0] & 0x7fffffffUL);
624 	seq_put_decimal_ull(m, " ", task->blocked.sig[0] & 0x7fffffffUL);
625 	seq_put_decimal_ull(m, " ", sigign.sig[0] & 0x7fffffffUL);
626 	seq_put_decimal_ull(m, " ", sigcatch.sig[0] & 0x7fffffffUL);
627 
628 	/*
629 	 * We used to output the absolute kernel address, but that's an
630 	 * information leak - so instead we show a 0/1 flag here, to signal
631 	 * to user-space whether there's a wchan field in /proc/PID/wchan.
632 	 *
633 	 * This works with older implementations of procps as well.
634 	 */
635 	seq_put_decimal_ull(m, " ", wchan);
636 
637 	seq_put_decimal_ull(m, " ", 0);
638 	seq_put_decimal_ull(m, " ", 0);
639 	seq_put_decimal_ll(m, " ", task->exit_signal);
640 	seq_put_decimal_ll(m, " ", task_cpu(task));
641 	seq_put_decimal_ull(m, " ", task->rt_priority);
642 	seq_put_decimal_ull(m, " ", task->policy);
643 	seq_put_decimal_ull(m, " ", delayacct_blkio_ticks(task));
644 	seq_put_decimal_ull(m, " ", nsec_to_clock_t(gtime));
645 	seq_put_decimal_ll(m, " ", nsec_to_clock_t(cgtime));
646 
647 	if (mm && permitted) {
648 		seq_put_decimal_ull(m, " ", mm->start_data);
649 		seq_put_decimal_ull(m, " ", mm->end_data);
650 		seq_put_decimal_ull(m, " ", mm->start_brk);
651 		seq_put_decimal_ull(m, " ", mm->arg_start);
652 		seq_put_decimal_ull(m, " ", mm->arg_end);
653 		seq_put_decimal_ull(m, " ", mm->env_start);
654 		seq_put_decimal_ull(m, " ", mm->env_end);
655 	} else
656 		seq_puts(m, " 0 0 0 0 0 0 0");
657 
658 	if (permitted)
659 		seq_put_decimal_ll(m, " ", exit_code);
660 	else
661 		seq_puts(m, " 0");
662 
663 	seq_putc(m, '\n');
664 	if (mm)
665 		mmput(mm);
666 	return 0;
667 }
668 
proc_tid_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)669 int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
670 			struct pid *pid, struct task_struct *task)
671 {
672 	return do_task_stat(m, ns, pid, task, 0);
673 }
674 
proc_tgid_stat(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)675 int proc_tgid_stat(struct seq_file *m, struct pid_namespace *ns,
676 			struct pid *pid, struct task_struct *task)
677 {
678 	return do_task_stat(m, ns, pid, task, 1);
679 }
680 
proc_pid_statm(struct seq_file * m,struct pid_namespace * ns,struct pid * pid,struct task_struct * task)681 int proc_pid_statm(struct seq_file *m, struct pid_namespace *ns,
682 			struct pid *pid, struct task_struct *task)
683 {
684 	struct mm_struct *mm = get_task_mm(task);
685 
686 	if (mm) {
687 		unsigned long size;
688 		unsigned long resident = 0;
689 		unsigned long shared = 0;
690 		unsigned long text = 0;
691 		unsigned long data = 0;
692 
693 		size = task_statm(mm, &shared, &text, &data, &resident);
694 		mmput(mm);
695 
696 		/*
697 		 * For quick read, open code by putting numbers directly
698 		 * expected format is
699 		 * seq_printf(m, "%lu %lu %lu %lu 0 %lu 0\n",
700 		 *               size, resident, shared, text, data);
701 		 */
702 		seq_put_decimal_ull(m, "", size);
703 		seq_put_decimal_ull(m, " ", resident);
704 		seq_put_decimal_ull(m, " ", shared);
705 		seq_put_decimal_ull(m, " ", text);
706 		seq_put_decimal_ull(m, " ", 0);
707 		seq_put_decimal_ull(m, " ", data);
708 		seq_put_decimal_ull(m, " ", 0);
709 		seq_putc(m, '\n');
710 	} else {
711 		seq_write(m, "0 0 0 0 0 0 0\n", 14);
712 	}
713 	return 0;
714 }
715 
716 #ifdef CONFIG_PROC_CHILDREN
717 static struct pid *
get_children_pid(struct inode * inode,struct pid * pid_prev,loff_t pos)718 get_children_pid(struct inode *inode, struct pid *pid_prev, loff_t pos)
719 {
720 	struct task_struct *start, *task;
721 	struct pid *pid = NULL;
722 
723 	read_lock(&tasklist_lock);
724 
725 	start = pid_task(proc_pid(inode), PIDTYPE_PID);
726 	if (!start)
727 		goto out;
728 
729 	/*
730 	 * Lets try to continue searching first, this gives
731 	 * us significant speedup on children-rich processes.
732 	 */
733 	if (pid_prev) {
734 		task = pid_task(pid_prev, PIDTYPE_PID);
735 		if (task && task->real_parent == start &&
736 		    !(list_empty(&task->sibling))) {
737 			if (list_is_last(&task->sibling, &start->children))
738 				goto out;
739 			task = list_first_entry(&task->sibling,
740 						struct task_struct, sibling);
741 			pid = get_pid(task_pid(task));
742 			goto out;
743 		}
744 	}
745 
746 	/*
747 	 * Slow search case.
748 	 *
749 	 * We might miss some children here if children
750 	 * are exited while we were not holding the lock,
751 	 * but it was never promised to be accurate that
752 	 * much.
753 	 *
754 	 * "Just suppose that the parent sleeps, but N children
755 	 *  exit after we printed their tids. Now the slow paths
756 	 *  skips N extra children, we miss N tasks." (c)
757 	 *
758 	 * So one need to stop or freeze the leader and all
759 	 * its children to get a precise result.
760 	 */
761 	list_for_each_entry(task, &start->children, sibling) {
762 		if (pos-- == 0) {
763 			pid = get_pid(task_pid(task));
764 			break;
765 		}
766 	}
767 
768 out:
769 	read_unlock(&tasklist_lock);
770 	return pid;
771 }
772 
children_seq_show(struct seq_file * seq,void * v)773 static int children_seq_show(struct seq_file *seq, void *v)
774 {
775 	struct inode *inode = file_inode(seq->file);
776 
777 	seq_printf(seq, "%d ", pid_nr_ns(v, proc_pid_ns(inode->i_sb)));
778 	return 0;
779 }
780 
children_seq_start(struct seq_file * seq,loff_t * pos)781 static void *children_seq_start(struct seq_file *seq, loff_t *pos)
782 {
783 	return get_children_pid(file_inode(seq->file), NULL, *pos);
784 }
785 
children_seq_next(struct seq_file * seq,void * v,loff_t * pos)786 static void *children_seq_next(struct seq_file *seq, void *v, loff_t *pos)
787 {
788 	struct pid *pid;
789 
790 	pid = get_children_pid(file_inode(seq->file), v, *pos + 1);
791 	put_pid(v);
792 
793 	++*pos;
794 	return pid;
795 }
796 
children_seq_stop(struct seq_file * seq,void * v)797 static void children_seq_stop(struct seq_file *seq, void *v)
798 {
799 	put_pid(v);
800 }
801 
802 static const struct seq_operations children_seq_ops = {
803 	.start	= children_seq_start,
804 	.next	= children_seq_next,
805 	.stop	= children_seq_stop,
806 	.show	= children_seq_show,
807 };
808 
children_seq_open(struct inode * inode,struct file * file)809 static int children_seq_open(struct inode *inode, struct file *file)
810 {
811 	return seq_open(file, &children_seq_ops);
812 }
813 
814 const struct file_operations proc_tid_children_operations = {
815 	.open    = children_seq_open,
816 	.read    = seq_read,
817 	.llseek  = seq_lseek,
818 	.release = seq_release,
819 };
820 #endif /* CONFIG_PROC_CHILDREN */
821