xref: /linux/fs/proc/base.c (revision 35219bc5c71f4197c8bd10297597de797c1eece5)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  linux/fs/proc/base.c
4  *
5  *  Copyright (C) 1991, 1992 Linus Torvalds
6  *
7  *  proc base directory handling functions
8  *
9  *  1999, Al Viro. Rewritten. Now it covers the whole per-process part.
10  *  Instead of using magical inumbers to determine the kind of object
11  *  we allocate and fill in-core inodes upon lookup. They don't even
12  *  go into icache. We cache the reference to task_struct upon lookup too.
13  *  Eventually it should become a filesystem in its own. We don't use the
14  *  rest of procfs anymore.
15  *
16  *
17  *  Changelog:
18  *  17-Jan-2005
19  *  Allan Bezerra
20  *  Bruna Moreira <bruna.moreira@indt.org.br>
21  *  Edjard Mota <edjard.mota@indt.org.br>
22  *  Ilias Biris <ilias.biris@indt.org.br>
23  *  Mauricio Lin <mauricio.lin@indt.org.br>
24  *
25  *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26  *
27  *  A new process specific entry (smaps) included in /proc. It shows the
28  *  size of rss for each memory area. The maps entry lacks information
29  *  about physical memory size (rss) for each mapped file, i.e.,
30  *  rss information for executables and library files.
31  *  This additional information is useful for any tools that need to know
32  *  about physical memory consumption for a process specific library.
33  *
34  *  Changelog:
35  *  21-Feb-2005
36  *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
37  *  Pud inclusion in the page table walking.
38  *
39  *  ChangeLog:
40  *  10-Mar-2005
41  *  10LE Instituto Nokia de Tecnologia - INdT:
42  *  A better way to walks through the page table as suggested by Hugh Dickins.
43  *
44  *  Simo Piiroinen <simo.piiroinen@nokia.com>:
45  *  Smaps information related to shared, private, clean and dirty pages.
46  *
47  *  Paul Mundt <paul.mundt@nokia.com>:
48  *  Overall revision about smaps.
49  */
50 
51 #include <linux/uaccess.h>
52 
53 #include <linux/errno.h>
54 #include <linux/time.h>
55 #include <linux/proc_fs.h>
56 #include <linux/stat.h>
57 #include <linux/task_io_accounting_ops.h>
58 #include <linux/init.h>
59 #include <linux/capability.h>
60 #include <linux/file.h>
61 #include <linux/fdtable.h>
62 #include <linux/generic-radix-tree.h>
63 #include <linux/string.h>
64 #include <linux/seq_file.h>
65 #include <linux/namei.h>
66 #include <linux/mnt_namespace.h>
67 #include <linux/mm.h>
68 #include <linux/swap.h>
69 #include <linux/rcupdate.h>
70 #include <linux/kallsyms.h>
71 #include <linux/stacktrace.h>
72 #include <linux/resource.h>
73 #include <linux/module.h>
74 #include <linux/mount.h>
75 #include <linux/security.h>
76 #include <linux/ptrace.h>
77 #include <linux/printk.h>
78 #include <linux/cache.h>
79 #include <linux/cgroup.h>
80 #include <linux/cpuset.h>
81 #include <linux/audit.h>
82 #include <linux/poll.h>
83 #include <linux/nsproxy.h>
84 #include <linux/oom.h>
85 #include <linux/elf.h>
86 #include <linux/pid_namespace.h>
87 #include <linux/user_namespace.h>
88 #include <linux/fs_parser.h>
89 #include <linux/fs_struct.h>
90 #include <linux/slab.h>
91 #include <linux/sched/autogroup.h>
92 #include <linux/sched/mm.h>
93 #include <linux/sched/coredump.h>
94 #include <linux/sched/debug.h>
95 #include <linux/sched/stat.h>
96 #include <linux/posix-timers.h>
97 #include <linux/time_namespace.h>
98 #include <linux/resctrl.h>
99 #include <linux/cn_proc.h>
100 #include <linux/ksm.h>
101 #include <uapi/linux/lsm.h>
102 #include <trace/events/oom.h>
103 #include "internal.h"
104 #include "fd.h"
105 
106 #include "../../lib/kstrtox.h"
107 
108 /* NOTE:
109  *	Implementing inode permission operations in /proc is almost
110  *	certainly an error.  Permission checks need to happen during
111  *	each system call not at open time.  The reason is that most of
112  *	what we wish to check for permissions in /proc varies at runtime.
113  *
114  *	The classic example of a problem is opening file descriptors
115  *	in /proc for a task before it execs a suid executable.
116  */
117 
118 static u8 nlink_tid __ro_after_init;
119 static u8 nlink_tgid __ro_after_init;
120 
121 enum proc_mem_force {
122 	PROC_MEM_FORCE_ALWAYS,
123 	PROC_MEM_FORCE_PTRACE,
124 	PROC_MEM_FORCE_NEVER
125 };
126 
127 static enum proc_mem_force proc_mem_force_override __ro_after_init =
128 	IS_ENABLED(CONFIG_PROC_MEM_NO_FORCE) ? PROC_MEM_FORCE_NEVER :
129 	IS_ENABLED(CONFIG_PROC_MEM_FORCE_PTRACE) ? PROC_MEM_FORCE_PTRACE :
130 	PROC_MEM_FORCE_ALWAYS;
131 
132 static const struct constant_table proc_mem_force_table[] __initconst = {
133 	{ "always", PROC_MEM_FORCE_ALWAYS },
134 	{ "ptrace", PROC_MEM_FORCE_PTRACE },
135 	{ "never", PROC_MEM_FORCE_NEVER },
136 	{ }
137 };
138 
139 static int __init early_proc_mem_force_override(char *buf)
140 {
141 	if (!buf)
142 		return -EINVAL;
143 
144 	/*
145 	 * lookup_constant() defaults to proc_mem_force_override to preseve
146 	 * the initial Kconfig choice in case an invalid param gets passed.
147 	 */
148 	proc_mem_force_override = lookup_constant(proc_mem_force_table,
149 						  buf, proc_mem_force_override);
150 
151 	return 0;
152 }
153 early_param("proc_mem.force_override", early_proc_mem_force_override);
154 
155 struct pid_entry {
156 	const char *name;
157 	unsigned int len;
158 	umode_t mode;
159 	const struct inode_operations *iop;
160 	const struct file_operations *fop;
161 	union proc_op op;
162 };
163 
164 #define NOD(NAME, MODE, IOP, FOP, OP) {			\
165 	.name = (NAME),					\
166 	.len  = sizeof(NAME) - 1,			\
167 	.mode = MODE,					\
168 	.iop  = IOP,					\
169 	.fop  = FOP,					\
170 	.op   = OP,					\
171 }
172 
173 #define DIR(NAME, MODE, iops, fops)	\
174 	NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
175 #define LNK(NAME, get_link)					\
176 	NOD(NAME, (S_IFLNK|S_IRWXUGO),				\
177 		&proc_pid_link_inode_operations, NULL,		\
178 		{ .proc_get_link = get_link } )
179 #define REG(NAME, MODE, fops)				\
180 	NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
181 #define ONE(NAME, MODE, show)				\
182 	NOD(NAME, (S_IFREG|(MODE)),			\
183 		NULL, &proc_single_file_operations,	\
184 		{ .proc_show = show } )
185 #define ATTR(LSMID, NAME, MODE)				\
186 	NOD(NAME, (S_IFREG|(MODE)),			\
187 		NULL, &proc_pid_attr_operations,	\
188 		{ .lsmid = LSMID })
189 
190 /*
191  * Count the number of hardlinks for the pid_entry table, excluding the .
192  * and .. links.
193  */
194 static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
195 	unsigned int n)
196 {
197 	unsigned int i;
198 	unsigned int count;
199 
200 	count = 2;
201 	for (i = 0; i < n; ++i) {
202 		if (S_ISDIR(entries[i].mode))
203 			++count;
204 	}
205 
206 	return count;
207 }
208 
209 static int get_task_root(struct task_struct *task, struct path *root)
210 {
211 	int result = -ENOENT;
212 
213 	task_lock(task);
214 	if (task->fs) {
215 		get_fs_root(task->fs, root);
216 		result = 0;
217 	}
218 	task_unlock(task);
219 	return result;
220 }
221 
222 static int proc_cwd_link(struct dentry *dentry, struct path *path)
223 {
224 	struct task_struct *task = get_proc_task(d_inode(dentry));
225 	int result = -ENOENT;
226 
227 	if (task) {
228 		task_lock(task);
229 		if (task->fs) {
230 			get_fs_pwd(task->fs, path);
231 			result = 0;
232 		}
233 		task_unlock(task);
234 		put_task_struct(task);
235 	}
236 	return result;
237 }
238 
239 static int proc_root_link(struct dentry *dentry, struct path *path)
240 {
241 	struct task_struct *task = get_proc_task(d_inode(dentry));
242 	int result = -ENOENT;
243 
244 	if (task) {
245 		result = get_task_root(task, path);
246 		put_task_struct(task);
247 	}
248 	return result;
249 }
250 
251 /*
252  * If the user used setproctitle(), we just get the string from
253  * user space at arg_start, and limit it to a maximum of one page.
254  */
255 static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf,
256 				size_t count, unsigned long pos,
257 				unsigned long arg_start)
258 {
259 	char *page;
260 	int ret, got;
261 
262 	if (pos >= PAGE_SIZE)
263 		return 0;
264 
265 	page = (char *)__get_free_page(GFP_KERNEL);
266 	if (!page)
267 		return -ENOMEM;
268 
269 	ret = 0;
270 	got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON);
271 	if (got > 0) {
272 		int len = strnlen(page, got);
273 
274 		/* Include the NUL character if it was found */
275 		if (len < got)
276 			len++;
277 
278 		if (len > pos) {
279 			len -= pos;
280 			if (len > count)
281 				len = count;
282 			len -= copy_to_user(buf, page+pos, len);
283 			if (!len)
284 				len = -EFAULT;
285 			ret = len;
286 		}
287 	}
288 	free_page((unsigned long)page);
289 	return ret;
290 }
291 
292 static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf,
293 			      size_t count, loff_t *ppos)
294 {
295 	unsigned long arg_start, arg_end, env_start, env_end;
296 	unsigned long pos, len;
297 	char *page, c;
298 
299 	/* Check if process spawned far enough to have cmdline. */
300 	if (!mm->env_end)
301 		return 0;
302 
303 	spin_lock(&mm->arg_lock);
304 	arg_start = mm->arg_start;
305 	arg_end = mm->arg_end;
306 	env_start = mm->env_start;
307 	env_end = mm->env_end;
308 	spin_unlock(&mm->arg_lock);
309 
310 	if (arg_start >= arg_end)
311 		return 0;
312 
313 	/*
314 	 * We allow setproctitle() to overwrite the argument
315 	 * strings, and overflow past the original end. But
316 	 * only when it overflows into the environment area.
317 	 */
318 	if (env_start != arg_end || env_end < env_start)
319 		env_start = env_end = arg_end;
320 	len = env_end - arg_start;
321 
322 	/* We're not going to care if "*ppos" has high bits set */
323 	pos = *ppos;
324 	if (pos >= len)
325 		return 0;
326 	if (count > len - pos)
327 		count = len - pos;
328 	if (!count)
329 		return 0;
330 
331 	/*
332 	 * Magical special case: if the argv[] end byte is not
333 	 * zero, the user has overwritten it with setproctitle(3).
334 	 *
335 	 * Possible future enhancement: do this only once when
336 	 * pos is 0, and set a flag in the 'struct file'.
337 	 */
338 	if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c)
339 		return get_mm_proctitle(mm, buf, count, pos, arg_start);
340 
341 	/*
342 	 * For the non-setproctitle() case we limit things strictly
343 	 * to the [arg_start, arg_end[ range.
344 	 */
345 	pos += arg_start;
346 	if (pos < arg_start || pos >= arg_end)
347 		return 0;
348 	if (count > arg_end - pos)
349 		count = arg_end - pos;
350 
351 	page = (char *)__get_free_page(GFP_KERNEL);
352 	if (!page)
353 		return -ENOMEM;
354 
355 	len = 0;
356 	while (count) {
357 		int got;
358 		size_t size = min_t(size_t, PAGE_SIZE, count);
359 
360 		got = access_remote_vm(mm, pos, page, size, FOLL_ANON);
361 		if (got <= 0)
362 			break;
363 		got -= copy_to_user(buf, page, got);
364 		if (unlikely(!got)) {
365 			if (!len)
366 				len = -EFAULT;
367 			break;
368 		}
369 		pos += got;
370 		buf += got;
371 		len += got;
372 		count -= got;
373 	}
374 
375 	free_page((unsigned long)page);
376 	return len;
377 }
378 
379 static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf,
380 				size_t count, loff_t *pos)
381 {
382 	struct mm_struct *mm;
383 	ssize_t ret;
384 
385 	mm = get_task_mm(tsk);
386 	if (!mm)
387 		return 0;
388 
389 	ret = get_mm_cmdline(mm, buf, count, pos);
390 	mmput(mm);
391 	return ret;
392 }
393 
394 static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
395 				     size_t count, loff_t *pos)
396 {
397 	struct task_struct *tsk;
398 	ssize_t ret;
399 
400 	BUG_ON(*pos < 0);
401 
402 	tsk = get_proc_task(file_inode(file));
403 	if (!tsk)
404 		return -ESRCH;
405 	ret = get_task_cmdline(tsk, buf, count, pos);
406 	put_task_struct(tsk);
407 	if (ret > 0)
408 		*pos += ret;
409 	return ret;
410 }
411 
412 static const struct file_operations proc_pid_cmdline_ops = {
413 	.read	= proc_pid_cmdline_read,
414 	.llseek	= generic_file_llseek,
415 };
416 
417 #ifdef CONFIG_KALLSYMS
418 /*
419  * Provides a wchan file via kallsyms in a proper one-value-per-file format.
420  * Returns the resolved symbol.  If that fails, simply return the address.
421  */
422 static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
423 			  struct pid *pid, struct task_struct *task)
424 {
425 	unsigned long wchan;
426 	char symname[KSYM_NAME_LEN];
427 
428 	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
429 		goto print0;
430 
431 	wchan = get_wchan(task);
432 	if (wchan && !lookup_symbol_name(wchan, symname)) {
433 		seq_puts(m, symname);
434 		return 0;
435 	}
436 
437 print0:
438 	seq_putc(m, '0');
439 	return 0;
440 }
441 #endif /* CONFIG_KALLSYMS */
442 
443 static int lock_trace(struct task_struct *task)
444 {
445 	int err = down_read_killable(&task->signal->exec_update_lock);
446 	if (err)
447 		return err;
448 	if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
449 		up_read(&task->signal->exec_update_lock);
450 		return -EPERM;
451 	}
452 	return 0;
453 }
454 
455 static void unlock_trace(struct task_struct *task)
456 {
457 	up_read(&task->signal->exec_update_lock);
458 }
459 
460 #ifdef CONFIG_STACKTRACE
461 
462 #define MAX_STACK_TRACE_DEPTH	64
463 
464 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
465 			  struct pid *pid, struct task_struct *task)
466 {
467 	unsigned long *entries;
468 	int err;
469 
470 	/*
471 	 * The ability to racily run the kernel stack unwinder on a running task
472 	 * and then observe the unwinder output is scary; while it is useful for
473 	 * debugging kernel issues, it can also allow an attacker to leak kernel
474 	 * stack contents.
475 	 * Doing this in a manner that is at least safe from races would require
476 	 * some work to ensure that the remote task can not be scheduled; and
477 	 * even then, this would still expose the unwinder as local attack
478 	 * surface.
479 	 * Therefore, this interface is restricted to root.
480 	 */
481 	if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN))
482 		return -EACCES;
483 
484 	entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries),
485 				GFP_KERNEL);
486 	if (!entries)
487 		return -ENOMEM;
488 
489 	err = lock_trace(task);
490 	if (!err) {
491 		unsigned int i, nr_entries;
492 
493 		nr_entries = stack_trace_save_tsk(task, entries,
494 						  MAX_STACK_TRACE_DEPTH, 0);
495 
496 		for (i = 0; i < nr_entries; i++) {
497 			seq_printf(m, "[<0>] %pB\n", (void *)entries[i]);
498 		}
499 
500 		unlock_trace(task);
501 	}
502 	kfree(entries);
503 
504 	return err;
505 }
506 #endif
507 
508 #ifdef CONFIG_SCHED_INFO
509 /*
510  * Provides /proc/PID/schedstat
511  */
512 static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
513 			      struct pid *pid, struct task_struct *task)
514 {
515 	if (unlikely(!sched_info_on()))
516 		seq_puts(m, "0 0 0\n");
517 	else
518 		seq_printf(m, "%llu %llu %lu\n",
519 		   (unsigned long long)task->se.sum_exec_runtime,
520 		   (unsigned long long)task->sched_info.run_delay,
521 		   task->sched_info.pcount);
522 
523 	return 0;
524 }
525 #endif
526 
527 #ifdef CONFIG_LATENCYTOP
528 static int lstats_show_proc(struct seq_file *m, void *v)
529 {
530 	int i;
531 	struct inode *inode = m->private;
532 	struct task_struct *task = get_proc_task(inode);
533 
534 	if (!task)
535 		return -ESRCH;
536 	seq_puts(m, "Latency Top version : v0.1\n");
537 	for (i = 0; i < LT_SAVECOUNT; i++) {
538 		struct latency_record *lr = &task->latency_record[i];
539 		if (lr->backtrace[0]) {
540 			int q;
541 			seq_printf(m, "%i %li %li",
542 				   lr->count, lr->time, lr->max);
543 			for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
544 				unsigned long bt = lr->backtrace[q];
545 
546 				if (!bt)
547 					break;
548 				seq_printf(m, " %ps", (void *)bt);
549 			}
550 			seq_putc(m, '\n');
551 		}
552 
553 	}
554 	put_task_struct(task);
555 	return 0;
556 }
557 
558 static int lstats_open(struct inode *inode, struct file *file)
559 {
560 	return single_open(file, lstats_show_proc, inode);
561 }
562 
563 static ssize_t lstats_write(struct file *file, const char __user *buf,
564 			    size_t count, loff_t *offs)
565 {
566 	struct task_struct *task = get_proc_task(file_inode(file));
567 
568 	if (!task)
569 		return -ESRCH;
570 	clear_tsk_latency_tracing(task);
571 	put_task_struct(task);
572 
573 	return count;
574 }
575 
576 static const struct file_operations proc_lstats_operations = {
577 	.open		= lstats_open,
578 	.read		= seq_read,
579 	.write		= lstats_write,
580 	.llseek		= seq_lseek,
581 	.release	= single_release,
582 };
583 
584 #endif
585 
586 static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
587 			  struct pid *pid, struct task_struct *task)
588 {
589 	unsigned long totalpages = totalram_pages() + total_swap_pages;
590 	unsigned long points = 0;
591 	long badness;
592 
593 	badness = oom_badness(task, totalpages);
594 	/*
595 	 * Special case OOM_SCORE_ADJ_MIN for all others scale the
596 	 * badness value into [0, 2000] range which we have been
597 	 * exporting for a long time so userspace might depend on it.
598 	 */
599 	if (badness != LONG_MIN)
600 		points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3;
601 
602 	seq_printf(m, "%lu\n", points);
603 
604 	return 0;
605 }
606 
607 struct limit_names {
608 	const char *name;
609 	const char *unit;
610 };
611 
612 static const struct limit_names lnames[RLIM_NLIMITS] = {
613 	[RLIMIT_CPU] = {"Max cpu time", "seconds"},
614 	[RLIMIT_FSIZE] = {"Max file size", "bytes"},
615 	[RLIMIT_DATA] = {"Max data size", "bytes"},
616 	[RLIMIT_STACK] = {"Max stack size", "bytes"},
617 	[RLIMIT_CORE] = {"Max core file size", "bytes"},
618 	[RLIMIT_RSS] = {"Max resident set", "bytes"},
619 	[RLIMIT_NPROC] = {"Max processes", "processes"},
620 	[RLIMIT_NOFILE] = {"Max open files", "files"},
621 	[RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
622 	[RLIMIT_AS] = {"Max address space", "bytes"},
623 	[RLIMIT_LOCKS] = {"Max file locks", "locks"},
624 	[RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
625 	[RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
626 	[RLIMIT_NICE] = {"Max nice priority", NULL},
627 	[RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
628 	[RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
629 };
630 
631 /* Display limits for a process */
632 static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
633 			   struct pid *pid, struct task_struct *task)
634 {
635 	unsigned int i;
636 	unsigned long flags;
637 
638 	struct rlimit rlim[RLIM_NLIMITS];
639 
640 	if (!lock_task_sighand(task, &flags))
641 		return 0;
642 	memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
643 	unlock_task_sighand(task, &flags);
644 
645 	/*
646 	 * print the file header
647 	 */
648 	seq_puts(m, "Limit                     "
649 		"Soft Limit           "
650 		"Hard Limit           "
651 		"Units     \n");
652 
653 	for (i = 0; i < RLIM_NLIMITS; i++) {
654 		if (rlim[i].rlim_cur == RLIM_INFINITY)
655 			seq_printf(m, "%-25s %-20s ",
656 				   lnames[i].name, "unlimited");
657 		else
658 			seq_printf(m, "%-25s %-20lu ",
659 				   lnames[i].name, rlim[i].rlim_cur);
660 
661 		if (rlim[i].rlim_max == RLIM_INFINITY)
662 			seq_printf(m, "%-20s ", "unlimited");
663 		else
664 			seq_printf(m, "%-20lu ", rlim[i].rlim_max);
665 
666 		if (lnames[i].unit)
667 			seq_printf(m, "%-10s\n", lnames[i].unit);
668 		else
669 			seq_putc(m, '\n');
670 	}
671 
672 	return 0;
673 }
674 
675 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
676 static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
677 			    struct pid *pid, struct task_struct *task)
678 {
679 	struct syscall_info info;
680 	u64 *args = &info.data.args[0];
681 	int res;
682 
683 	res = lock_trace(task);
684 	if (res)
685 		return res;
686 
687 	if (task_current_syscall(task, &info))
688 		seq_puts(m, "running\n");
689 	else if (info.data.nr < 0)
690 		seq_printf(m, "%d 0x%llx 0x%llx\n",
691 			   info.data.nr, info.sp, info.data.instruction_pointer);
692 	else
693 		seq_printf(m,
694 		       "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n",
695 		       info.data.nr,
696 		       args[0], args[1], args[2], args[3], args[4], args[5],
697 		       info.sp, info.data.instruction_pointer);
698 	unlock_trace(task);
699 
700 	return 0;
701 }
702 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
703 
704 /************************************************************************/
705 /*                       Here the fs part begins                        */
706 /************************************************************************/
707 
708 /* permission checks */
709 static bool proc_fd_access_allowed(struct inode *inode)
710 {
711 	struct task_struct *task;
712 	bool allowed = false;
713 	/* Allow access to a task's file descriptors if it is us or we
714 	 * may use ptrace attach to the process and find out that
715 	 * information.
716 	 */
717 	task = get_proc_task(inode);
718 	if (task) {
719 		allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
720 		put_task_struct(task);
721 	}
722 	return allowed;
723 }
724 
725 int proc_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
726 		 struct iattr *attr)
727 {
728 	int error;
729 	struct inode *inode = d_inode(dentry);
730 
731 	if (attr->ia_valid & ATTR_MODE)
732 		return -EPERM;
733 
734 	error = setattr_prepare(&nop_mnt_idmap, dentry, attr);
735 	if (error)
736 		return error;
737 
738 	setattr_copy(&nop_mnt_idmap, inode, attr);
739 	return 0;
740 }
741 
742 /*
743  * May current process learn task's sched/cmdline info (for hide_pid_min=1)
744  * or euid/egid (for hide_pid_min=2)?
745  */
746 static bool has_pid_permissions(struct proc_fs_info *fs_info,
747 				 struct task_struct *task,
748 				 enum proc_hidepid hide_pid_min)
749 {
750 	/*
751 	 * If 'hidpid' mount option is set force a ptrace check,
752 	 * we indicate that we are using a filesystem syscall
753 	 * by passing PTRACE_MODE_READ_FSCREDS
754 	 */
755 	if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE)
756 		return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
757 
758 	if (fs_info->hide_pid < hide_pid_min)
759 		return true;
760 	if (in_group_p(fs_info->pid_gid))
761 		return true;
762 	return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
763 }
764 
765 
766 static int proc_pid_permission(struct mnt_idmap *idmap,
767 			       struct inode *inode, int mask)
768 {
769 	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
770 	struct task_struct *task;
771 	bool has_perms;
772 
773 	task = get_proc_task(inode);
774 	if (!task)
775 		return -ESRCH;
776 	has_perms = has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS);
777 	put_task_struct(task);
778 
779 	if (!has_perms) {
780 		if (fs_info->hide_pid == HIDEPID_INVISIBLE) {
781 			/*
782 			 * Let's make getdents(), stat(), and open()
783 			 * consistent with each other.  If a process
784 			 * may not stat() a file, it shouldn't be seen
785 			 * in procfs at all.
786 			 */
787 			return -ENOENT;
788 		}
789 
790 		return -EPERM;
791 	}
792 	return generic_permission(&nop_mnt_idmap, inode, mask);
793 }
794 
795 
796 
797 static const struct inode_operations proc_def_inode_operations = {
798 	.setattr	= proc_setattr,
799 };
800 
801 static int proc_single_show(struct seq_file *m, void *v)
802 {
803 	struct inode *inode = m->private;
804 	struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
805 	struct pid *pid = proc_pid(inode);
806 	struct task_struct *task;
807 	int ret;
808 
809 	task = get_pid_task(pid, PIDTYPE_PID);
810 	if (!task)
811 		return -ESRCH;
812 
813 	ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
814 
815 	put_task_struct(task);
816 	return ret;
817 }
818 
819 static int proc_single_open(struct inode *inode, struct file *filp)
820 {
821 	return single_open(filp, proc_single_show, inode);
822 }
823 
824 static const struct file_operations proc_single_file_operations = {
825 	.open		= proc_single_open,
826 	.read		= seq_read,
827 	.llseek		= seq_lseek,
828 	.release	= single_release,
829 };
830 
831 
832 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
833 {
834 	struct task_struct *task = get_proc_task(inode);
835 	struct mm_struct *mm = ERR_PTR(-ESRCH);
836 
837 	if (task) {
838 		mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
839 		put_task_struct(task);
840 
841 		if (!IS_ERR_OR_NULL(mm)) {
842 			/* ensure this mm_struct can't be freed */
843 			mmgrab(mm);
844 			/* but do not pin its memory */
845 			mmput(mm);
846 		}
847 	}
848 
849 	return mm;
850 }
851 
852 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
853 {
854 	struct mm_struct *mm = proc_mem_open(inode, mode);
855 
856 	if (IS_ERR(mm))
857 		return PTR_ERR(mm);
858 
859 	file->private_data = mm;
860 	return 0;
861 }
862 
863 static int mem_open(struct inode *inode, struct file *file)
864 {
865 	if (WARN_ON_ONCE(!(file->f_op->fop_flags & FOP_UNSIGNED_OFFSET)))
866 		return -EINVAL;
867 	return __mem_open(inode, file, PTRACE_MODE_ATTACH);
868 }
869 
870 static bool proc_mem_foll_force(struct file *file, struct mm_struct *mm)
871 {
872 	struct task_struct *task;
873 	bool ptrace_active = false;
874 
875 	switch (proc_mem_force_override) {
876 	case PROC_MEM_FORCE_NEVER:
877 		return false;
878 	case PROC_MEM_FORCE_PTRACE:
879 		task = get_proc_task(file_inode(file));
880 		if (task) {
881 			ptrace_active =	READ_ONCE(task->ptrace) &&
882 					READ_ONCE(task->mm) == mm &&
883 					READ_ONCE(task->parent) == current;
884 			put_task_struct(task);
885 		}
886 		return ptrace_active;
887 	default:
888 		return true;
889 	}
890 }
891 
892 static ssize_t mem_rw(struct file *file, char __user *buf,
893 			size_t count, loff_t *ppos, int write)
894 {
895 	struct mm_struct *mm = file->private_data;
896 	unsigned long addr = *ppos;
897 	ssize_t copied;
898 	char *page;
899 	unsigned int flags;
900 
901 	if (!mm)
902 		return 0;
903 
904 	page = (char *)__get_free_page(GFP_KERNEL);
905 	if (!page)
906 		return -ENOMEM;
907 
908 	copied = 0;
909 	if (!mmget_not_zero(mm))
910 		goto free;
911 
912 	flags = write ? FOLL_WRITE : 0;
913 	if (proc_mem_foll_force(file, mm))
914 		flags |= FOLL_FORCE;
915 
916 	while (count > 0) {
917 		size_t this_len = min_t(size_t, count, PAGE_SIZE);
918 
919 		if (write && copy_from_user(page, buf, this_len)) {
920 			copied = -EFAULT;
921 			break;
922 		}
923 
924 		this_len = access_remote_vm(mm, addr, page, this_len, flags);
925 		if (!this_len) {
926 			if (!copied)
927 				copied = -EIO;
928 			break;
929 		}
930 
931 		if (!write && copy_to_user(buf, page, this_len)) {
932 			copied = -EFAULT;
933 			break;
934 		}
935 
936 		buf += this_len;
937 		addr += this_len;
938 		copied += this_len;
939 		count -= this_len;
940 	}
941 	*ppos = addr;
942 
943 	mmput(mm);
944 free:
945 	free_page((unsigned long) page);
946 	return copied;
947 }
948 
949 static ssize_t mem_read(struct file *file, char __user *buf,
950 			size_t count, loff_t *ppos)
951 {
952 	return mem_rw(file, buf, count, ppos, 0);
953 }
954 
955 static ssize_t mem_write(struct file *file, const char __user *buf,
956 			 size_t count, loff_t *ppos)
957 {
958 	return mem_rw(file, (char __user*)buf, count, ppos, 1);
959 }
960 
961 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
962 {
963 	switch (orig) {
964 	case 0:
965 		file->f_pos = offset;
966 		break;
967 	case 1:
968 		file->f_pos += offset;
969 		break;
970 	default:
971 		return -EINVAL;
972 	}
973 	force_successful_syscall_return();
974 	return file->f_pos;
975 }
976 
977 static int mem_release(struct inode *inode, struct file *file)
978 {
979 	struct mm_struct *mm = file->private_data;
980 	if (mm)
981 		mmdrop(mm);
982 	return 0;
983 }
984 
985 static const struct file_operations proc_mem_operations = {
986 	.llseek		= mem_lseek,
987 	.read		= mem_read,
988 	.write		= mem_write,
989 	.open		= mem_open,
990 	.release	= mem_release,
991 	.fop_flags	= FOP_UNSIGNED_OFFSET,
992 };
993 
994 static int environ_open(struct inode *inode, struct file *file)
995 {
996 	return __mem_open(inode, file, PTRACE_MODE_READ);
997 }
998 
999 static ssize_t environ_read(struct file *file, char __user *buf,
1000 			size_t count, loff_t *ppos)
1001 {
1002 	char *page;
1003 	unsigned long src = *ppos;
1004 	int ret = 0;
1005 	struct mm_struct *mm = file->private_data;
1006 	unsigned long env_start, env_end;
1007 
1008 	/* Ensure the process spawned far enough to have an environment. */
1009 	if (!mm || !mm->env_end)
1010 		return 0;
1011 
1012 	page = (char *)__get_free_page(GFP_KERNEL);
1013 	if (!page)
1014 		return -ENOMEM;
1015 
1016 	ret = 0;
1017 	if (!mmget_not_zero(mm))
1018 		goto free;
1019 
1020 	spin_lock(&mm->arg_lock);
1021 	env_start = mm->env_start;
1022 	env_end = mm->env_end;
1023 	spin_unlock(&mm->arg_lock);
1024 
1025 	while (count > 0) {
1026 		size_t this_len, max_len;
1027 		int retval;
1028 
1029 		if (src >= (env_end - env_start))
1030 			break;
1031 
1032 		this_len = env_end - (env_start + src);
1033 
1034 		max_len = min_t(size_t, PAGE_SIZE, count);
1035 		this_len = min(max_len, this_len);
1036 
1037 		retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON);
1038 
1039 		if (retval <= 0) {
1040 			ret = retval;
1041 			break;
1042 		}
1043 
1044 		if (copy_to_user(buf, page, retval)) {
1045 			ret = -EFAULT;
1046 			break;
1047 		}
1048 
1049 		ret += retval;
1050 		src += retval;
1051 		buf += retval;
1052 		count -= retval;
1053 	}
1054 	*ppos = src;
1055 	mmput(mm);
1056 
1057 free:
1058 	free_page((unsigned long) page);
1059 	return ret;
1060 }
1061 
1062 static const struct file_operations proc_environ_operations = {
1063 	.open		= environ_open,
1064 	.read		= environ_read,
1065 	.llseek		= generic_file_llseek,
1066 	.release	= mem_release,
1067 };
1068 
1069 static int auxv_open(struct inode *inode, struct file *file)
1070 {
1071 	return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
1072 }
1073 
1074 static ssize_t auxv_read(struct file *file, char __user *buf,
1075 			size_t count, loff_t *ppos)
1076 {
1077 	struct mm_struct *mm = file->private_data;
1078 	unsigned int nwords = 0;
1079 
1080 	if (!mm)
1081 		return 0;
1082 	do {
1083 		nwords += 2;
1084 	} while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1085 	return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
1086 				       nwords * sizeof(mm->saved_auxv[0]));
1087 }
1088 
1089 static const struct file_operations proc_auxv_operations = {
1090 	.open		= auxv_open,
1091 	.read		= auxv_read,
1092 	.llseek		= generic_file_llseek,
1093 	.release	= mem_release,
1094 };
1095 
1096 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1097 			    loff_t *ppos)
1098 {
1099 	struct task_struct *task = get_proc_task(file_inode(file));
1100 	char buffer[PROC_NUMBUF];
1101 	int oom_adj = OOM_ADJUST_MIN;
1102 	size_t len;
1103 
1104 	if (!task)
1105 		return -ESRCH;
1106 	if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1107 		oom_adj = OOM_ADJUST_MAX;
1108 	else
1109 		oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1110 			  OOM_SCORE_ADJ_MAX;
1111 	put_task_struct(task);
1112 	if (oom_adj > OOM_ADJUST_MAX)
1113 		oom_adj = OOM_ADJUST_MAX;
1114 	len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1115 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1116 }
1117 
1118 static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1119 {
1120 	struct mm_struct *mm = NULL;
1121 	struct task_struct *task;
1122 	int err = 0;
1123 
1124 	task = get_proc_task(file_inode(file));
1125 	if (!task)
1126 		return -ESRCH;
1127 
1128 	mutex_lock(&oom_adj_mutex);
1129 	if (legacy) {
1130 		if (oom_adj < task->signal->oom_score_adj &&
1131 				!capable(CAP_SYS_RESOURCE)) {
1132 			err = -EACCES;
1133 			goto err_unlock;
1134 		}
1135 		/*
1136 		 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1137 		 * /proc/pid/oom_score_adj instead.
1138 		 */
1139 		pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1140 			  current->comm, task_pid_nr(current), task_pid_nr(task),
1141 			  task_pid_nr(task));
1142 	} else {
1143 		if ((short)oom_adj < task->signal->oom_score_adj_min &&
1144 				!capable(CAP_SYS_RESOURCE)) {
1145 			err = -EACCES;
1146 			goto err_unlock;
1147 		}
1148 	}
1149 
1150 	/*
1151 	 * Make sure we will check other processes sharing the mm if this is
1152 	 * not vfrok which wants its own oom_score_adj.
1153 	 * pin the mm so it doesn't go away and get reused after task_unlock
1154 	 */
1155 	if (!task->vfork_done) {
1156 		struct task_struct *p = find_lock_task_mm(task);
1157 
1158 		if (p) {
1159 			if (test_bit(MMF_MULTIPROCESS, &p->mm->flags)) {
1160 				mm = p->mm;
1161 				mmgrab(mm);
1162 			}
1163 			task_unlock(p);
1164 		}
1165 	}
1166 
1167 	task->signal->oom_score_adj = oom_adj;
1168 	if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1169 		task->signal->oom_score_adj_min = (short)oom_adj;
1170 	trace_oom_score_adj_update(task);
1171 
1172 	if (mm) {
1173 		struct task_struct *p;
1174 
1175 		rcu_read_lock();
1176 		for_each_process(p) {
1177 			if (same_thread_group(task, p))
1178 				continue;
1179 
1180 			/* do not touch kernel threads or the global init */
1181 			if (p->flags & PF_KTHREAD || is_global_init(p))
1182 				continue;
1183 
1184 			task_lock(p);
1185 			if (!p->vfork_done && process_shares_mm(p, mm)) {
1186 				p->signal->oom_score_adj = oom_adj;
1187 				if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1188 					p->signal->oom_score_adj_min = (short)oom_adj;
1189 			}
1190 			task_unlock(p);
1191 		}
1192 		rcu_read_unlock();
1193 		mmdrop(mm);
1194 	}
1195 err_unlock:
1196 	mutex_unlock(&oom_adj_mutex);
1197 	put_task_struct(task);
1198 	return err;
1199 }
1200 
1201 /*
1202  * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1203  * kernels.  The effective policy is defined by oom_score_adj, which has a
1204  * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1205  * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1206  * Processes that become oom disabled via oom_adj will still be oom disabled
1207  * with this implementation.
1208  *
1209  * oom_adj cannot be removed since existing userspace binaries use it.
1210  */
1211 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1212 			     size_t count, loff_t *ppos)
1213 {
1214 	char buffer[PROC_NUMBUF] = {};
1215 	int oom_adj;
1216 	int err;
1217 
1218 	if (count > sizeof(buffer) - 1)
1219 		count = sizeof(buffer) - 1;
1220 	if (copy_from_user(buffer, buf, count)) {
1221 		err = -EFAULT;
1222 		goto out;
1223 	}
1224 
1225 	err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1226 	if (err)
1227 		goto out;
1228 	if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1229 	     oom_adj != OOM_DISABLE) {
1230 		err = -EINVAL;
1231 		goto out;
1232 	}
1233 
1234 	/*
1235 	 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1236 	 * value is always attainable.
1237 	 */
1238 	if (oom_adj == OOM_ADJUST_MAX)
1239 		oom_adj = OOM_SCORE_ADJ_MAX;
1240 	else
1241 		oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1242 
1243 	err = __set_oom_adj(file, oom_adj, true);
1244 out:
1245 	return err < 0 ? err : count;
1246 }
1247 
1248 static const struct file_operations proc_oom_adj_operations = {
1249 	.read		= oom_adj_read,
1250 	.write		= oom_adj_write,
1251 	.llseek		= generic_file_llseek,
1252 };
1253 
1254 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1255 					size_t count, loff_t *ppos)
1256 {
1257 	struct task_struct *task = get_proc_task(file_inode(file));
1258 	char buffer[PROC_NUMBUF];
1259 	short oom_score_adj = OOM_SCORE_ADJ_MIN;
1260 	size_t len;
1261 
1262 	if (!task)
1263 		return -ESRCH;
1264 	oom_score_adj = task->signal->oom_score_adj;
1265 	put_task_struct(task);
1266 	len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1267 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1268 }
1269 
1270 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1271 					size_t count, loff_t *ppos)
1272 {
1273 	char buffer[PROC_NUMBUF] = {};
1274 	int oom_score_adj;
1275 	int err;
1276 
1277 	if (count > sizeof(buffer) - 1)
1278 		count = sizeof(buffer) - 1;
1279 	if (copy_from_user(buffer, buf, count)) {
1280 		err = -EFAULT;
1281 		goto out;
1282 	}
1283 
1284 	err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1285 	if (err)
1286 		goto out;
1287 	if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1288 			oom_score_adj > OOM_SCORE_ADJ_MAX) {
1289 		err = -EINVAL;
1290 		goto out;
1291 	}
1292 
1293 	err = __set_oom_adj(file, oom_score_adj, false);
1294 out:
1295 	return err < 0 ? err : count;
1296 }
1297 
1298 static const struct file_operations proc_oom_score_adj_operations = {
1299 	.read		= oom_score_adj_read,
1300 	.write		= oom_score_adj_write,
1301 	.llseek		= default_llseek,
1302 };
1303 
1304 #ifdef CONFIG_AUDIT
1305 #define TMPBUFLEN 11
1306 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1307 				  size_t count, loff_t *ppos)
1308 {
1309 	struct inode * inode = file_inode(file);
1310 	struct task_struct *task = get_proc_task(inode);
1311 	ssize_t length;
1312 	char tmpbuf[TMPBUFLEN];
1313 
1314 	if (!task)
1315 		return -ESRCH;
1316 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1317 			   from_kuid(file->f_cred->user_ns,
1318 				     audit_get_loginuid(task)));
1319 	put_task_struct(task);
1320 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1321 }
1322 
1323 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1324 				   size_t count, loff_t *ppos)
1325 {
1326 	struct inode * inode = file_inode(file);
1327 	uid_t loginuid;
1328 	kuid_t kloginuid;
1329 	int rv;
1330 
1331 	/* Don't let kthreads write their own loginuid */
1332 	if (current->flags & PF_KTHREAD)
1333 		return -EPERM;
1334 
1335 	rcu_read_lock();
1336 	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1337 		rcu_read_unlock();
1338 		return -EPERM;
1339 	}
1340 	rcu_read_unlock();
1341 
1342 	if (*ppos != 0) {
1343 		/* No partial writes. */
1344 		return -EINVAL;
1345 	}
1346 
1347 	rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1348 	if (rv < 0)
1349 		return rv;
1350 
1351 	/* is userspace tring to explicitly UNSET the loginuid? */
1352 	if (loginuid == AUDIT_UID_UNSET) {
1353 		kloginuid = INVALID_UID;
1354 	} else {
1355 		kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1356 		if (!uid_valid(kloginuid))
1357 			return -EINVAL;
1358 	}
1359 
1360 	rv = audit_set_loginuid(kloginuid);
1361 	if (rv < 0)
1362 		return rv;
1363 	return count;
1364 }
1365 
1366 static const struct file_operations proc_loginuid_operations = {
1367 	.read		= proc_loginuid_read,
1368 	.write		= proc_loginuid_write,
1369 	.llseek		= generic_file_llseek,
1370 };
1371 
1372 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1373 				  size_t count, loff_t *ppos)
1374 {
1375 	struct inode * inode = file_inode(file);
1376 	struct task_struct *task = get_proc_task(inode);
1377 	ssize_t length;
1378 	char tmpbuf[TMPBUFLEN];
1379 
1380 	if (!task)
1381 		return -ESRCH;
1382 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1383 				audit_get_sessionid(task));
1384 	put_task_struct(task);
1385 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1386 }
1387 
1388 static const struct file_operations proc_sessionid_operations = {
1389 	.read		= proc_sessionid_read,
1390 	.llseek		= generic_file_llseek,
1391 };
1392 #endif
1393 
1394 #ifdef CONFIG_FAULT_INJECTION
1395 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1396 				      size_t count, loff_t *ppos)
1397 {
1398 	struct task_struct *task = get_proc_task(file_inode(file));
1399 	char buffer[PROC_NUMBUF];
1400 	size_t len;
1401 	int make_it_fail;
1402 
1403 	if (!task)
1404 		return -ESRCH;
1405 	make_it_fail = task->make_it_fail;
1406 	put_task_struct(task);
1407 
1408 	len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1409 
1410 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1411 }
1412 
1413 static ssize_t proc_fault_inject_write(struct file * file,
1414 			const char __user * buf, size_t count, loff_t *ppos)
1415 {
1416 	struct task_struct *task;
1417 	char buffer[PROC_NUMBUF] = {};
1418 	int make_it_fail;
1419 	int rv;
1420 
1421 	if (!capable(CAP_SYS_RESOURCE))
1422 		return -EPERM;
1423 
1424 	if (count > sizeof(buffer) - 1)
1425 		count = sizeof(buffer) - 1;
1426 	if (copy_from_user(buffer, buf, count))
1427 		return -EFAULT;
1428 	rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1429 	if (rv < 0)
1430 		return rv;
1431 	if (make_it_fail < 0 || make_it_fail > 1)
1432 		return -EINVAL;
1433 
1434 	task = get_proc_task(file_inode(file));
1435 	if (!task)
1436 		return -ESRCH;
1437 	task->make_it_fail = make_it_fail;
1438 	put_task_struct(task);
1439 
1440 	return count;
1441 }
1442 
1443 static const struct file_operations proc_fault_inject_operations = {
1444 	.read		= proc_fault_inject_read,
1445 	.write		= proc_fault_inject_write,
1446 	.llseek		= generic_file_llseek,
1447 };
1448 
1449 static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1450 				   size_t count, loff_t *ppos)
1451 {
1452 	struct task_struct *task;
1453 	int err;
1454 	unsigned int n;
1455 
1456 	err = kstrtouint_from_user(buf, count, 0, &n);
1457 	if (err)
1458 		return err;
1459 
1460 	task = get_proc_task(file_inode(file));
1461 	if (!task)
1462 		return -ESRCH;
1463 	task->fail_nth = n;
1464 	put_task_struct(task);
1465 
1466 	return count;
1467 }
1468 
1469 static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1470 				  size_t count, loff_t *ppos)
1471 {
1472 	struct task_struct *task;
1473 	char numbuf[PROC_NUMBUF];
1474 	ssize_t len;
1475 
1476 	task = get_proc_task(file_inode(file));
1477 	if (!task)
1478 		return -ESRCH;
1479 	len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth);
1480 	put_task_struct(task);
1481 	return simple_read_from_buffer(buf, count, ppos, numbuf, len);
1482 }
1483 
1484 static const struct file_operations proc_fail_nth_operations = {
1485 	.read		= proc_fail_nth_read,
1486 	.write		= proc_fail_nth_write,
1487 };
1488 #endif
1489 
1490 
1491 #ifdef CONFIG_SCHED_DEBUG
1492 /*
1493  * Print out various scheduling related per-task fields:
1494  */
1495 static int sched_show(struct seq_file *m, void *v)
1496 {
1497 	struct inode *inode = m->private;
1498 	struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
1499 	struct task_struct *p;
1500 
1501 	p = get_proc_task(inode);
1502 	if (!p)
1503 		return -ESRCH;
1504 	proc_sched_show_task(p, ns, m);
1505 
1506 	put_task_struct(p);
1507 
1508 	return 0;
1509 }
1510 
1511 static ssize_t
1512 sched_write(struct file *file, const char __user *buf,
1513 	    size_t count, loff_t *offset)
1514 {
1515 	struct inode *inode = file_inode(file);
1516 	struct task_struct *p;
1517 
1518 	p = get_proc_task(inode);
1519 	if (!p)
1520 		return -ESRCH;
1521 	proc_sched_set_task(p);
1522 
1523 	put_task_struct(p);
1524 
1525 	return count;
1526 }
1527 
1528 static int sched_open(struct inode *inode, struct file *filp)
1529 {
1530 	return single_open(filp, sched_show, inode);
1531 }
1532 
1533 static const struct file_operations proc_pid_sched_operations = {
1534 	.open		= sched_open,
1535 	.read		= seq_read,
1536 	.write		= sched_write,
1537 	.llseek		= seq_lseek,
1538 	.release	= single_release,
1539 };
1540 
1541 #endif
1542 
1543 #ifdef CONFIG_SCHED_AUTOGROUP
1544 /*
1545  * Print out autogroup related information:
1546  */
1547 static int sched_autogroup_show(struct seq_file *m, void *v)
1548 {
1549 	struct inode *inode = m->private;
1550 	struct task_struct *p;
1551 
1552 	p = get_proc_task(inode);
1553 	if (!p)
1554 		return -ESRCH;
1555 	proc_sched_autogroup_show_task(p, m);
1556 
1557 	put_task_struct(p);
1558 
1559 	return 0;
1560 }
1561 
1562 static ssize_t
1563 sched_autogroup_write(struct file *file, const char __user *buf,
1564 	    size_t count, loff_t *offset)
1565 {
1566 	struct inode *inode = file_inode(file);
1567 	struct task_struct *p;
1568 	char buffer[PROC_NUMBUF] = {};
1569 	int nice;
1570 	int err;
1571 
1572 	if (count > sizeof(buffer) - 1)
1573 		count = sizeof(buffer) - 1;
1574 	if (copy_from_user(buffer, buf, count))
1575 		return -EFAULT;
1576 
1577 	err = kstrtoint(strstrip(buffer), 0, &nice);
1578 	if (err < 0)
1579 		return err;
1580 
1581 	p = get_proc_task(inode);
1582 	if (!p)
1583 		return -ESRCH;
1584 
1585 	err = proc_sched_autogroup_set_nice(p, nice);
1586 	if (err)
1587 		count = err;
1588 
1589 	put_task_struct(p);
1590 
1591 	return count;
1592 }
1593 
1594 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1595 {
1596 	int ret;
1597 
1598 	ret = single_open(filp, sched_autogroup_show, NULL);
1599 	if (!ret) {
1600 		struct seq_file *m = filp->private_data;
1601 
1602 		m->private = inode;
1603 	}
1604 	return ret;
1605 }
1606 
1607 static const struct file_operations proc_pid_sched_autogroup_operations = {
1608 	.open		= sched_autogroup_open,
1609 	.read		= seq_read,
1610 	.write		= sched_autogroup_write,
1611 	.llseek		= seq_lseek,
1612 	.release	= single_release,
1613 };
1614 
1615 #endif /* CONFIG_SCHED_AUTOGROUP */
1616 
1617 #ifdef CONFIG_TIME_NS
1618 static int timens_offsets_show(struct seq_file *m, void *v)
1619 {
1620 	struct task_struct *p;
1621 
1622 	p = get_proc_task(file_inode(m->file));
1623 	if (!p)
1624 		return -ESRCH;
1625 	proc_timens_show_offsets(p, m);
1626 
1627 	put_task_struct(p);
1628 
1629 	return 0;
1630 }
1631 
1632 static ssize_t timens_offsets_write(struct file *file, const char __user *buf,
1633 				    size_t count, loff_t *ppos)
1634 {
1635 	struct inode *inode = file_inode(file);
1636 	struct proc_timens_offset offsets[2];
1637 	char *kbuf = NULL, *pos, *next_line;
1638 	struct task_struct *p;
1639 	int ret, noffsets;
1640 
1641 	/* Only allow < page size writes at the beginning of the file */
1642 	if ((*ppos != 0) || (count >= PAGE_SIZE))
1643 		return -EINVAL;
1644 
1645 	/* Slurp in the user data */
1646 	kbuf = memdup_user_nul(buf, count);
1647 	if (IS_ERR(kbuf))
1648 		return PTR_ERR(kbuf);
1649 
1650 	/* Parse the user data */
1651 	ret = -EINVAL;
1652 	noffsets = 0;
1653 	for (pos = kbuf; pos; pos = next_line) {
1654 		struct proc_timens_offset *off = &offsets[noffsets];
1655 		char clock[10];
1656 		int err;
1657 
1658 		/* Find the end of line and ensure we don't look past it */
1659 		next_line = strchr(pos, '\n');
1660 		if (next_line) {
1661 			*next_line = '\0';
1662 			next_line++;
1663 			if (*next_line == '\0')
1664 				next_line = NULL;
1665 		}
1666 
1667 		err = sscanf(pos, "%9s %lld %lu", clock,
1668 				&off->val.tv_sec, &off->val.tv_nsec);
1669 		if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC)
1670 			goto out;
1671 
1672 		clock[sizeof(clock) - 1] = 0;
1673 		if (strcmp(clock, "monotonic") == 0 ||
1674 		    strcmp(clock, __stringify(CLOCK_MONOTONIC)) == 0)
1675 			off->clockid = CLOCK_MONOTONIC;
1676 		else if (strcmp(clock, "boottime") == 0 ||
1677 			 strcmp(clock, __stringify(CLOCK_BOOTTIME)) == 0)
1678 			off->clockid = CLOCK_BOOTTIME;
1679 		else
1680 			goto out;
1681 
1682 		noffsets++;
1683 		if (noffsets == ARRAY_SIZE(offsets)) {
1684 			if (next_line)
1685 				count = next_line - kbuf;
1686 			break;
1687 		}
1688 	}
1689 
1690 	ret = -ESRCH;
1691 	p = get_proc_task(inode);
1692 	if (!p)
1693 		goto out;
1694 	ret = proc_timens_set_offset(file, p, offsets, noffsets);
1695 	put_task_struct(p);
1696 	if (ret)
1697 		goto out;
1698 
1699 	ret = count;
1700 out:
1701 	kfree(kbuf);
1702 	return ret;
1703 }
1704 
1705 static int timens_offsets_open(struct inode *inode, struct file *filp)
1706 {
1707 	return single_open(filp, timens_offsets_show, inode);
1708 }
1709 
1710 static const struct file_operations proc_timens_offsets_operations = {
1711 	.open		= timens_offsets_open,
1712 	.read		= seq_read,
1713 	.write		= timens_offsets_write,
1714 	.llseek		= seq_lseek,
1715 	.release	= single_release,
1716 };
1717 #endif /* CONFIG_TIME_NS */
1718 
1719 static ssize_t comm_write(struct file *file, const char __user *buf,
1720 				size_t count, loff_t *offset)
1721 {
1722 	struct inode *inode = file_inode(file);
1723 	struct task_struct *p;
1724 	char buffer[TASK_COMM_LEN] = {};
1725 	const size_t maxlen = sizeof(buffer) - 1;
1726 
1727 	if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1728 		return -EFAULT;
1729 
1730 	p = get_proc_task(inode);
1731 	if (!p)
1732 		return -ESRCH;
1733 
1734 	if (same_thread_group(current, p)) {
1735 		set_task_comm(p, buffer);
1736 		proc_comm_connector(p);
1737 	}
1738 	else
1739 		count = -EINVAL;
1740 
1741 	put_task_struct(p);
1742 
1743 	return count;
1744 }
1745 
1746 static int comm_show(struct seq_file *m, void *v)
1747 {
1748 	struct inode *inode = m->private;
1749 	struct task_struct *p;
1750 
1751 	p = get_proc_task(inode);
1752 	if (!p)
1753 		return -ESRCH;
1754 
1755 	proc_task_name(m, p, false);
1756 	seq_putc(m, '\n');
1757 
1758 	put_task_struct(p);
1759 
1760 	return 0;
1761 }
1762 
1763 static int comm_open(struct inode *inode, struct file *filp)
1764 {
1765 	return single_open(filp, comm_show, inode);
1766 }
1767 
1768 static const struct file_operations proc_pid_set_comm_operations = {
1769 	.open		= comm_open,
1770 	.read		= seq_read,
1771 	.write		= comm_write,
1772 	.llseek		= seq_lseek,
1773 	.release	= single_release,
1774 };
1775 
1776 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1777 {
1778 	struct task_struct *task;
1779 	struct file *exe_file;
1780 
1781 	task = get_proc_task(d_inode(dentry));
1782 	if (!task)
1783 		return -ENOENT;
1784 	exe_file = get_task_exe_file(task);
1785 	put_task_struct(task);
1786 	if (exe_file) {
1787 		*exe_path = exe_file->f_path;
1788 		path_get(&exe_file->f_path);
1789 		fput(exe_file);
1790 		return 0;
1791 	} else
1792 		return -ENOENT;
1793 }
1794 
1795 static const char *proc_pid_get_link(struct dentry *dentry,
1796 				     struct inode *inode,
1797 				     struct delayed_call *done)
1798 {
1799 	struct path path;
1800 	int error = -EACCES;
1801 
1802 	if (!dentry)
1803 		return ERR_PTR(-ECHILD);
1804 
1805 	/* Are we allowed to snoop on the tasks file descriptors? */
1806 	if (!proc_fd_access_allowed(inode))
1807 		goto out;
1808 
1809 	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1810 	if (error)
1811 		goto out;
1812 
1813 	error = nd_jump_link(&path);
1814 out:
1815 	return ERR_PTR(error);
1816 }
1817 
1818 static int do_proc_readlink(const struct path *path, char __user *buffer, int buflen)
1819 {
1820 	char *tmp = kmalloc(PATH_MAX, GFP_KERNEL);
1821 	char *pathname;
1822 	int len;
1823 
1824 	if (!tmp)
1825 		return -ENOMEM;
1826 
1827 	pathname = d_path(path, tmp, PATH_MAX);
1828 	len = PTR_ERR(pathname);
1829 	if (IS_ERR(pathname))
1830 		goto out;
1831 	len = tmp + PATH_MAX - 1 - pathname;
1832 
1833 	if (len > buflen)
1834 		len = buflen;
1835 	if (copy_to_user(buffer, pathname, len))
1836 		len = -EFAULT;
1837  out:
1838 	kfree(tmp);
1839 	return len;
1840 }
1841 
1842 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1843 {
1844 	int error = -EACCES;
1845 	struct inode *inode = d_inode(dentry);
1846 	struct path path;
1847 
1848 	/* Are we allowed to snoop on the tasks file descriptors? */
1849 	if (!proc_fd_access_allowed(inode))
1850 		goto out;
1851 
1852 	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1853 	if (error)
1854 		goto out;
1855 
1856 	error = do_proc_readlink(&path, buffer, buflen);
1857 	path_put(&path);
1858 out:
1859 	return error;
1860 }
1861 
1862 const struct inode_operations proc_pid_link_inode_operations = {
1863 	.readlink	= proc_pid_readlink,
1864 	.get_link	= proc_pid_get_link,
1865 	.setattr	= proc_setattr,
1866 };
1867 
1868 
1869 /* building an inode */
1870 
1871 void task_dump_owner(struct task_struct *task, umode_t mode,
1872 		     kuid_t *ruid, kgid_t *rgid)
1873 {
1874 	/* Depending on the state of dumpable compute who should own a
1875 	 * proc file for a task.
1876 	 */
1877 	const struct cred *cred;
1878 	kuid_t uid;
1879 	kgid_t gid;
1880 
1881 	if (unlikely(task->flags & PF_KTHREAD)) {
1882 		*ruid = GLOBAL_ROOT_UID;
1883 		*rgid = GLOBAL_ROOT_GID;
1884 		return;
1885 	}
1886 
1887 	/* Default to the tasks effective ownership */
1888 	rcu_read_lock();
1889 	cred = __task_cred(task);
1890 	uid = cred->euid;
1891 	gid = cred->egid;
1892 	rcu_read_unlock();
1893 
1894 	/*
1895 	 * Before the /proc/pid/status file was created the only way to read
1896 	 * the effective uid of a /process was to stat /proc/pid.  Reading
1897 	 * /proc/pid/status is slow enough that procps and other packages
1898 	 * kept stating /proc/pid.  To keep the rules in /proc simple I have
1899 	 * made this apply to all per process world readable and executable
1900 	 * directories.
1901 	 */
1902 	if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1903 		struct mm_struct *mm;
1904 		task_lock(task);
1905 		mm = task->mm;
1906 		/* Make non-dumpable tasks owned by some root */
1907 		if (mm) {
1908 			if (get_dumpable(mm) != SUID_DUMP_USER) {
1909 				struct user_namespace *user_ns = mm->user_ns;
1910 
1911 				uid = make_kuid(user_ns, 0);
1912 				if (!uid_valid(uid))
1913 					uid = GLOBAL_ROOT_UID;
1914 
1915 				gid = make_kgid(user_ns, 0);
1916 				if (!gid_valid(gid))
1917 					gid = GLOBAL_ROOT_GID;
1918 			}
1919 		} else {
1920 			uid = GLOBAL_ROOT_UID;
1921 			gid = GLOBAL_ROOT_GID;
1922 		}
1923 		task_unlock(task);
1924 	}
1925 	*ruid = uid;
1926 	*rgid = gid;
1927 }
1928 
1929 void proc_pid_evict_inode(struct proc_inode *ei)
1930 {
1931 	struct pid *pid = ei->pid;
1932 
1933 	if (S_ISDIR(ei->vfs_inode.i_mode)) {
1934 		spin_lock(&pid->lock);
1935 		hlist_del_init_rcu(&ei->sibling_inodes);
1936 		spin_unlock(&pid->lock);
1937 	}
1938 }
1939 
1940 struct inode *proc_pid_make_inode(struct super_block *sb,
1941 				  struct task_struct *task, umode_t mode)
1942 {
1943 	struct inode * inode;
1944 	struct proc_inode *ei;
1945 	struct pid *pid;
1946 
1947 	/* We need a new inode */
1948 
1949 	inode = new_inode(sb);
1950 	if (!inode)
1951 		goto out;
1952 
1953 	/* Common stuff */
1954 	ei = PROC_I(inode);
1955 	inode->i_mode = mode;
1956 	inode->i_ino = get_next_ino();
1957 	simple_inode_init_ts(inode);
1958 	inode->i_op = &proc_def_inode_operations;
1959 
1960 	/*
1961 	 * grab the reference to task.
1962 	 */
1963 	pid = get_task_pid(task, PIDTYPE_PID);
1964 	if (!pid)
1965 		goto out_unlock;
1966 
1967 	/* Let the pid remember us for quick removal */
1968 	ei->pid = pid;
1969 
1970 	task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1971 	security_task_to_inode(task, inode);
1972 
1973 out:
1974 	return inode;
1975 
1976 out_unlock:
1977 	iput(inode);
1978 	return NULL;
1979 }
1980 
1981 /*
1982  * Generating an inode and adding it into @pid->inodes, so that task will
1983  * invalidate inode's dentry before being released.
1984  *
1985  * This helper is used for creating dir-type entries under '/proc' and
1986  * '/proc/<tgid>/task'. Other entries(eg. fd, stat) under '/proc/<tgid>'
1987  * can be released by invalidating '/proc/<tgid>' dentry.
1988  * In theory, dentries under '/proc/<tgid>/task' can also be released by
1989  * invalidating '/proc/<tgid>' dentry, we reserve it to handle single
1990  * thread exiting situation: Any one of threads should invalidate its
1991  * '/proc/<tgid>/task/<pid>' dentry before released.
1992  */
1993 static struct inode *proc_pid_make_base_inode(struct super_block *sb,
1994 				struct task_struct *task, umode_t mode)
1995 {
1996 	struct inode *inode;
1997 	struct proc_inode *ei;
1998 	struct pid *pid;
1999 
2000 	inode = proc_pid_make_inode(sb, task, mode);
2001 	if (!inode)
2002 		return NULL;
2003 
2004 	/* Let proc_flush_pid find this directory inode */
2005 	ei = PROC_I(inode);
2006 	pid = ei->pid;
2007 	spin_lock(&pid->lock);
2008 	hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes);
2009 	spin_unlock(&pid->lock);
2010 
2011 	return inode;
2012 }
2013 
2014 int pid_getattr(struct mnt_idmap *idmap, const struct path *path,
2015 		struct kstat *stat, u32 request_mask, unsigned int query_flags)
2016 {
2017 	struct inode *inode = d_inode(path->dentry);
2018 	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
2019 	struct task_struct *task;
2020 
2021 	generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
2022 
2023 	stat->uid = GLOBAL_ROOT_UID;
2024 	stat->gid = GLOBAL_ROOT_GID;
2025 	rcu_read_lock();
2026 	task = pid_task(proc_pid(inode), PIDTYPE_PID);
2027 	if (task) {
2028 		if (!has_pid_permissions(fs_info, task, HIDEPID_INVISIBLE)) {
2029 			rcu_read_unlock();
2030 			/*
2031 			 * This doesn't prevent learning whether PID exists,
2032 			 * it only makes getattr() consistent with readdir().
2033 			 */
2034 			return -ENOENT;
2035 		}
2036 		task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
2037 	}
2038 	rcu_read_unlock();
2039 	return 0;
2040 }
2041 
2042 /* dentry stuff */
2043 
2044 /*
2045  * Set <pid>/... inode ownership (can change due to setuid(), etc.)
2046  */
2047 void pid_update_inode(struct task_struct *task, struct inode *inode)
2048 {
2049 	task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
2050 
2051 	inode->i_mode &= ~(S_ISUID | S_ISGID);
2052 	security_task_to_inode(task, inode);
2053 }
2054 
2055 /*
2056  * Rewrite the inode's ownerships here because the owning task may have
2057  * performed a setuid(), etc.
2058  *
2059  */
2060 static int pid_revalidate(struct dentry *dentry, unsigned int flags)
2061 {
2062 	struct inode *inode;
2063 	struct task_struct *task;
2064 	int ret = 0;
2065 
2066 	rcu_read_lock();
2067 	inode = d_inode_rcu(dentry);
2068 	if (!inode)
2069 		goto out;
2070 	task = pid_task(proc_pid(inode), PIDTYPE_PID);
2071 
2072 	if (task) {
2073 		pid_update_inode(task, inode);
2074 		ret = 1;
2075 	}
2076 out:
2077 	rcu_read_unlock();
2078 	return ret;
2079 }
2080 
2081 static inline bool proc_inode_is_dead(struct inode *inode)
2082 {
2083 	return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
2084 }
2085 
2086 int pid_delete_dentry(const struct dentry *dentry)
2087 {
2088 	/* Is the task we represent dead?
2089 	 * If so, then don't put the dentry on the lru list,
2090 	 * kill it immediately.
2091 	 */
2092 	return proc_inode_is_dead(d_inode(dentry));
2093 }
2094 
2095 const struct dentry_operations pid_dentry_operations =
2096 {
2097 	.d_revalidate	= pid_revalidate,
2098 	.d_delete	= pid_delete_dentry,
2099 };
2100 
2101 /* Lookups */
2102 
2103 /*
2104  * Fill a directory entry.
2105  *
2106  * If possible create the dcache entry and derive our inode number and
2107  * file type from dcache entry.
2108  *
2109  * Since all of the proc inode numbers are dynamically generated, the inode
2110  * numbers do not exist until the inode is cache.  This means creating
2111  * the dcache entry in readdir is necessary to keep the inode numbers
2112  * reported by readdir in sync with the inode numbers reported
2113  * by stat.
2114  */
2115 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
2116 	const char *name, unsigned int len,
2117 	instantiate_t instantiate, struct task_struct *task, const void *ptr)
2118 {
2119 	struct dentry *child, *dir = file->f_path.dentry;
2120 	struct qstr qname = QSTR_INIT(name, len);
2121 	struct inode *inode;
2122 	unsigned type = DT_UNKNOWN;
2123 	ino_t ino = 1;
2124 
2125 	child = d_hash_and_lookup(dir, &qname);
2126 	if (!child) {
2127 		DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
2128 		child = d_alloc_parallel(dir, &qname, &wq);
2129 		if (IS_ERR(child))
2130 			goto end_instantiate;
2131 		if (d_in_lookup(child)) {
2132 			struct dentry *res;
2133 			res = instantiate(child, task, ptr);
2134 			d_lookup_done(child);
2135 			if (unlikely(res)) {
2136 				dput(child);
2137 				child = res;
2138 				if (IS_ERR(child))
2139 					goto end_instantiate;
2140 			}
2141 		}
2142 	}
2143 	inode = d_inode(child);
2144 	ino = inode->i_ino;
2145 	type = inode->i_mode >> 12;
2146 	dput(child);
2147 end_instantiate:
2148 	return dir_emit(ctx, name, len, ino, type);
2149 }
2150 
2151 /*
2152  * dname_to_vma_addr - maps a dentry name into two unsigned longs
2153  * which represent vma start and end addresses.
2154  */
2155 static int dname_to_vma_addr(struct dentry *dentry,
2156 			     unsigned long *start, unsigned long *end)
2157 {
2158 	const char *str = dentry->d_name.name;
2159 	unsigned long long sval, eval;
2160 	unsigned int len;
2161 
2162 	if (str[0] == '0' && str[1] != '-')
2163 		return -EINVAL;
2164 	len = _parse_integer(str, 16, &sval);
2165 	if (len & KSTRTOX_OVERFLOW)
2166 		return -EINVAL;
2167 	if (sval != (unsigned long)sval)
2168 		return -EINVAL;
2169 	str += len;
2170 
2171 	if (*str != '-')
2172 		return -EINVAL;
2173 	str++;
2174 
2175 	if (str[0] == '0' && str[1])
2176 		return -EINVAL;
2177 	len = _parse_integer(str, 16, &eval);
2178 	if (len & KSTRTOX_OVERFLOW)
2179 		return -EINVAL;
2180 	if (eval != (unsigned long)eval)
2181 		return -EINVAL;
2182 	str += len;
2183 
2184 	if (*str != '\0')
2185 		return -EINVAL;
2186 
2187 	*start = sval;
2188 	*end = eval;
2189 
2190 	return 0;
2191 }
2192 
2193 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
2194 {
2195 	unsigned long vm_start, vm_end;
2196 	bool exact_vma_exists = false;
2197 	struct mm_struct *mm = NULL;
2198 	struct task_struct *task;
2199 	struct inode *inode;
2200 	int status = 0;
2201 
2202 	if (flags & LOOKUP_RCU)
2203 		return -ECHILD;
2204 
2205 	inode = d_inode(dentry);
2206 	task = get_proc_task(inode);
2207 	if (!task)
2208 		goto out_notask;
2209 
2210 	mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2211 	if (IS_ERR_OR_NULL(mm))
2212 		goto out;
2213 
2214 	if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2215 		status = mmap_read_lock_killable(mm);
2216 		if (!status) {
2217 			exact_vma_exists = !!find_exact_vma(mm, vm_start,
2218 							    vm_end);
2219 			mmap_read_unlock(mm);
2220 		}
2221 	}
2222 
2223 	mmput(mm);
2224 
2225 	if (exact_vma_exists) {
2226 		task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
2227 
2228 		security_task_to_inode(task, inode);
2229 		status = 1;
2230 	}
2231 
2232 out:
2233 	put_task_struct(task);
2234 
2235 out_notask:
2236 	return status;
2237 }
2238 
2239 static const struct dentry_operations tid_map_files_dentry_operations = {
2240 	.d_revalidate	= map_files_d_revalidate,
2241 	.d_delete	= pid_delete_dentry,
2242 };
2243 
2244 static int map_files_get_link(struct dentry *dentry, struct path *path)
2245 {
2246 	unsigned long vm_start, vm_end;
2247 	struct vm_area_struct *vma;
2248 	struct task_struct *task;
2249 	struct mm_struct *mm;
2250 	int rc;
2251 
2252 	rc = -ENOENT;
2253 	task = get_proc_task(d_inode(dentry));
2254 	if (!task)
2255 		goto out;
2256 
2257 	mm = get_task_mm(task);
2258 	put_task_struct(task);
2259 	if (!mm)
2260 		goto out;
2261 
2262 	rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2263 	if (rc)
2264 		goto out_mmput;
2265 
2266 	rc = mmap_read_lock_killable(mm);
2267 	if (rc)
2268 		goto out_mmput;
2269 
2270 	rc = -ENOENT;
2271 	vma = find_exact_vma(mm, vm_start, vm_end);
2272 	if (vma && vma->vm_file) {
2273 		*path = *file_user_path(vma->vm_file);
2274 		path_get(path);
2275 		rc = 0;
2276 	}
2277 	mmap_read_unlock(mm);
2278 
2279 out_mmput:
2280 	mmput(mm);
2281 out:
2282 	return rc;
2283 }
2284 
2285 struct map_files_info {
2286 	unsigned long	start;
2287 	unsigned long	end;
2288 	fmode_t		mode;
2289 };
2290 
2291 /*
2292  * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due
2293  * to concerns about how the symlinks may be used to bypass permissions on
2294  * ancestor directories in the path to the file in question.
2295  */
2296 static const char *
2297 proc_map_files_get_link(struct dentry *dentry,
2298 			struct inode *inode,
2299 		        struct delayed_call *done)
2300 {
2301 	if (!checkpoint_restore_ns_capable(&init_user_ns))
2302 		return ERR_PTR(-EPERM);
2303 
2304 	return proc_pid_get_link(dentry, inode, done);
2305 }
2306 
2307 /*
2308  * Identical to proc_pid_link_inode_operations except for get_link()
2309  */
2310 static const struct inode_operations proc_map_files_link_inode_operations = {
2311 	.readlink	= proc_pid_readlink,
2312 	.get_link	= proc_map_files_get_link,
2313 	.setattr	= proc_setattr,
2314 };
2315 
2316 static struct dentry *
2317 proc_map_files_instantiate(struct dentry *dentry,
2318 			   struct task_struct *task, const void *ptr)
2319 {
2320 	fmode_t mode = (fmode_t)(unsigned long)ptr;
2321 	struct proc_inode *ei;
2322 	struct inode *inode;
2323 
2324 	inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK |
2325 				    ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2326 				    ((mode & FMODE_WRITE) ? S_IWUSR : 0));
2327 	if (!inode)
2328 		return ERR_PTR(-ENOENT);
2329 
2330 	ei = PROC_I(inode);
2331 	ei->op.proc_get_link = map_files_get_link;
2332 
2333 	inode->i_op = &proc_map_files_link_inode_operations;
2334 	inode->i_size = 64;
2335 
2336 	return proc_splice_unmountable(inode, dentry,
2337 				       &tid_map_files_dentry_operations);
2338 }
2339 
2340 static struct dentry *proc_map_files_lookup(struct inode *dir,
2341 		struct dentry *dentry, unsigned int flags)
2342 {
2343 	unsigned long vm_start, vm_end;
2344 	struct vm_area_struct *vma;
2345 	struct task_struct *task;
2346 	struct dentry *result;
2347 	struct mm_struct *mm;
2348 
2349 	result = ERR_PTR(-ENOENT);
2350 	task = get_proc_task(dir);
2351 	if (!task)
2352 		goto out;
2353 
2354 	result = ERR_PTR(-EACCES);
2355 	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2356 		goto out_put_task;
2357 
2358 	result = ERR_PTR(-ENOENT);
2359 	if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2360 		goto out_put_task;
2361 
2362 	mm = get_task_mm(task);
2363 	if (!mm)
2364 		goto out_put_task;
2365 
2366 	result = ERR_PTR(-EINTR);
2367 	if (mmap_read_lock_killable(mm))
2368 		goto out_put_mm;
2369 
2370 	result = ERR_PTR(-ENOENT);
2371 	vma = find_exact_vma(mm, vm_start, vm_end);
2372 	if (!vma)
2373 		goto out_no_vma;
2374 
2375 	if (vma->vm_file)
2376 		result = proc_map_files_instantiate(dentry, task,
2377 				(void *)(unsigned long)vma->vm_file->f_mode);
2378 
2379 out_no_vma:
2380 	mmap_read_unlock(mm);
2381 out_put_mm:
2382 	mmput(mm);
2383 out_put_task:
2384 	put_task_struct(task);
2385 out:
2386 	return result;
2387 }
2388 
2389 static const struct inode_operations proc_map_files_inode_operations = {
2390 	.lookup		= proc_map_files_lookup,
2391 	.permission	= proc_fd_permission,
2392 	.setattr	= proc_setattr,
2393 };
2394 
2395 static int
2396 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2397 {
2398 	struct vm_area_struct *vma;
2399 	struct task_struct *task;
2400 	struct mm_struct *mm;
2401 	unsigned long nr_files, pos, i;
2402 	GENRADIX(struct map_files_info) fa;
2403 	struct map_files_info *p;
2404 	int ret;
2405 	struct vma_iterator vmi;
2406 
2407 	genradix_init(&fa);
2408 
2409 	ret = -ENOENT;
2410 	task = get_proc_task(file_inode(file));
2411 	if (!task)
2412 		goto out;
2413 
2414 	ret = -EACCES;
2415 	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2416 		goto out_put_task;
2417 
2418 	ret = 0;
2419 	if (!dir_emit_dots(file, ctx))
2420 		goto out_put_task;
2421 
2422 	mm = get_task_mm(task);
2423 	if (!mm)
2424 		goto out_put_task;
2425 
2426 	ret = mmap_read_lock_killable(mm);
2427 	if (ret) {
2428 		mmput(mm);
2429 		goto out_put_task;
2430 	}
2431 
2432 	nr_files = 0;
2433 
2434 	/*
2435 	 * We need two passes here:
2436 	 *
2437 	 *  1) Collect vmas of mapped files with mmap_lock taken
2438 	 *  2) Release mmap_lock and instantiate entries
2439 	 *
2440 	 * otherwise we get lockdep complained, since filldir()
2441 	 * routine might require mmap_lock taken in might_fault().
2442 	 */
2443 
2444 	pos = 2;
2445 	vma_iter_init(&vmi, mm, 0);
2446 	for_each_vma(vmi, vma) {
2447 		if (!vma->vm_file)
2448 			continue;
2449 		if (++pos <= ctx->pos)
2450 			continue;
2451 
2452 		p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL);
2453 		if (!p) {
2454 			ret = -ENOMEM;
2455 			mmap_read_unlock(mm);
2456 			mmput(mm);
2457 			goto out_put_task;
2458 		}
2459 
2460 		p->start = vma->vm_start;
2461 		p->end = vma->vm_end;
2462 		p->mode = vma->vm_file->f_mode;
2463 	}
2464 	mmap_read_unlock(mm);
2465 	mmput(mm);
2466 
2467 	for (i = 0; i < nr_files; i++) {
2468 		char buf[4 * sizeof(long) + 2];	/* max: %lx-%lx\0 */
2469 		unsigned int len;
2470 
2471 		p = genradix_ptr(&fa, i);
2472 		len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
2473 		if (!proc_fill_cache(file, ctx,
2474 				      buf, len,
2475 				      proc_map_files_instantiate,
2476 				      task,
2477 				      (void *)(unsigned long)p->mode))
2478 			break;
2479 		ctx->pos++;
2480 	}
2481 
2482 out_put_task:
2483 	put_task_struct(task);
2484 out:
2485 	genradix_free(&fa);
2486 	return ret;
2487 }
2488 
2489 static const struct file_operations proc_map_files_operations = {
2490 	.read		= generic_read_dir,
2491 	.iterate_shared	= proc_map_files_readdir,
2492 	.llseek		= generic_file_llseek,
2493 };
2494 
2495 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2496 struct timers_private {
2497 	struct pid *pid;
2498 	struct task_struct *task;
2499 	struct sighand_struct *sighand;
2500 	struct pid_namespace *ns;
2501 	unsigned long flags;
2502 };
2503 
2504 static void *timers_start(struct seq_file *m, loff_t *pos)
2505 {
2506 	struct timers_private *tp = m->private;
2507 
2508 	tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2509 	if (!tp->task)
2510 		return ERR_PTR(-ESRCH);
2511 
2512 	tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2513 	if (!tp->sighand)
2514 		return ERR_PTR(-ESRCH);
2515 
2516 	return seq_list_start(&tp->task->signal->posix_timers, *pos);
2517 }
2518 
2519 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2520 {
2521 	struct timers_private *tp = m->private;
2522 	return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2523 }
2524 
2525 static void timers_stop(struct seq_file *m, void *v)
2526 {
2527 	struct timers_private *tp = m->private;
2528 
2529 	if (tp->sighand) {
2530 		unlock_task_sighand(tp->task, &tp->flags);
2531 		tp->sighand = NULL;
2532 	}
2533 
2534 	if (tp->task) {
2535 		put_task_struct(tp->task);
2536 		tp->task = NULL;
2537 	}
2538 }
2539 
2540 static int show_timer(struct seq_file *m, void *v)
2541 {
2542 	struct k_itimer *timer;
2543 	struct timers_private *tp = m->private;
2544 	int notify;
2545 	static const char * const nstr[] = {
2546 		[SIGEV_SIGNAL] = "signal",
2547 		[SIGEV_NONE] = "none",
2548 		[SIGEV_THREAD] = "thread",
2549 	};
2550 
2551 	timer = list_entry((struct list_head *)v, struct k_itimer, list);
2552 	notify = timer->it_sigev_notify;
2553 
2554 	seq_printf(m, "ID: %d\n", timer->it_id);
2555 	seq_printf(m, "signal: %d/%px\n",
2556 		   timer->sigq->info.si_signo,
2557 		   timer->sigq->info.si_value.sival_ptr);
2558 	seq_printf(m, "notify: %s/%s.%d\n",
2559 		   nstr[notify & ~SIGEV_THREAD_ID],
2560 		   (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2561 		   pid_nr_ns(timer->it_pid, tp->ns));
2562 	seq_printf(m, "ClockID: %d\n", timer->it_clock);
2563 
2564 	return 0;
2565 }
2566 
2567 static const struct seq_operations proc_timers_seq_ops = {
2568 	.start	= timers_start,
2569 	.next	= timers_next,
2570 	.stop	= timers_stop,
2571 	.show	= show_timer,
2572 };
2573 
2574 static int proc_timers_open(struct inode *inode, struct file *file)
2575 {
2576 	struct timers_private *tp;
2577 
2578 	tp = __seq_open_private(file, &proc_timers_seq_ops,
2579 			sizeof(struct timers_private));
2580 	if (!tp)
2581 		return -ENOMEM;
2582 
2583 	tp->pid = proc_pid(inode);
2584 	tp->ns = proc_pid_ns(inode->i_sb);
2585 	return 0;
2586 }
2587 
2588 static const struct file_operations proc_timers_operations = {
2589 	.open		= proc_timers_open,
2590 	.read		= seq_read,
2591 	.llseek		= seq_lseek,
2592 	.release	= seq_release_private,
2593 };
2594 #endif
2595 
2596 static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2597 					size_t count, loff_t *offset)
2598 {
2599 	struct inode *inode = file_inode(file);
2600 	struct task_struct *p;
2601 	u64 slack_ns;
2602 	int err;
2603 
2604 	err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2605 	if (err < 0)
2606 		return err;
2607 
2608 	p = get_proc_task(inode);
2609 	if (!p)
2610 		return -ESRCH;
2611 
2612 	if (p != current) {
2613 		rcu_read_lock();
2614 		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2615 			rcu_read_unlock();
2616 			count = -EPERM;
2617 			goto out;
2618 		}
2619 		rcu_read_unlock();
2620 
2621 		err = security_task_setscheduler(p);
2622 		if (err) {
2623 			count = err;
2624 			goto out;
2625 		}
2626 	}
2627 
2628 	task_lock(p);
2629 	if (slack_ns == 0)
2630 		p->timer_slack_ns = p->default_timer_slack_ns;
2631 	else
2632 		p->timer_slack_ns = slack_ns;
2633 	task_unlock(p);
2634 
2635 out:
2636 	put_task_struct(p);
2637 
2638 	return count;
2639 }
2640 
2641 static int timerslack_ns_show(struct seq_file *m, void *v)
2642 {
2643 	struct inode *inode = m->private;
2644 	struct task_struct *p;
2645 	int err = 0;
2646 
2647 	p = get_proc_task(inode);
2648 	if (!p)
2649 		return -ESRCH;
2650 
2651 	if (p != current) {
2652 		rcu_read_lock();
2653 		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2654 			rcu_read_unlock();
2655 			err = -EPERM;
2656 			goto out;
2657 		}
2658 		rcu_read_unlock();
2659 
2660 		err = security_task_getscheduler(p);
2661 		if (err)
2662 			goto out;
2663 	}
2664 
2665 	task_lock(p);
2666 	seq_printf(m, "%llu\n", p->timer_slack_ns);
2667 	task_unlock(p);
2668 
2669 out:
2670 	put_task_struct(p);
2671 
2672 	return err;
2673 }
2674 
2675 static int timerslack_ns_open(struct inode *inode, struct file *filp)
2676 {
2677 	return single_open(filp, timerslack_ns_show, inode);
2678 }
2679 
2680 static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2681 	.open		= timerslack_ns_open,
2682 	.read		= seq_read,
2683 	.write		= timerslack_ns_write,
2684 	.llseek		= seq_lseek,
2685 	.release	= single_release,
2686 };
2687 
2688 static struct dentry *proc_pident_instantiate(struct dentry *dentry,
2689 	struct task_struct *task, const void *ptr)
2690 {
2691 	const struct pid_entry *p = ptr;
2692 	struct inode *inode;
2693 	struct proc_inode *ei;
2694 
2695 	inode = proc_pid_make_inode(dentry->d_sb, task, p->mode);
2696 	if (!inode)
2697 		return ERR_PTR(-ENOENT);
2698 
2699 	ei = PROC_I(inode);
2700 	if (S_ISDIR(inode->i_mode))
2701 		set_nlink(inode, 2);	/* Use getattr to fix if necessary */
2702 	if (p->iop)
2703 		inode->i_op = p->iop;
2704 	if (p->fop)
2705 		inode->i_fop = p->fop;
2706 	ei->op = p->op;
2707 	pid_update_inode(task, inode);
2708 	d_set_d_op(dentry, &pid_dentry_operations);
2709 	return d_splice_alias(inode, dentry);
2710 }
2711 
2712 static struct dentry *proc_pident_lookup(struct inode *dir,
2713 					 struct dentry *dentry,
2714 					 const struct pid_entry *p,
2715 					 const struct pid_entry *end)
2716 {
2717 	struct task_struct *task = get_proc_task(dir);
2718 	struct dentry *res = ERR_PTR(-ENOENT);
2719 
2720 	if (!task)
2721 		goto out_no_task;
2722 
2723 	/*
2724 	 * Yes, it does not scale. And it should not. Don't add
2725 	 * new entries into /proc/<tgid>/ without very good reasons.
2726 	 */
2727 	for (; p < end; p++) {
2728 		if (p->len != dentry->d_name.len)
2729 			continue;
2730 		if (!memcmp(dentry->d_name.name, p->name, p->len)) {
2731 			res = proc_pident_instantiate(dentry, task, p);
2732 			break;
2733 		}
2734 	}
2735 	put_task_struct(task);
2736 out_no_task:
2737 	return res;
2738 }
2739 
2740 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2741 		const struct pid_entry *ents, unsigned int nents)
2742 {
2743 	struct task_struct *task = get_proc_task(file_inode(file));
2744 	const struct pid_entry *p;
2745 
2746 	if (!task)
2747 		return -ENOENT;
2748 
2749 	if (!dir_emit_dots(file, ctx))
2750 		goto out;
2751 
2752 	if (ctx->pos >= nents + 2)
2753 		goto out;
2754 
2755 	for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
2756 		if (!proc_fill_cache(file, ctx, p->name, p->len,
2757 				proc_pident_instantiate, task, p))
2758 			break;
2759 		ctx->pos++;
2760 	}
2761 out:
2762 	put_task_struct(task);
2763 	return 0;
2764 }
2765 
2766 #ifdef CONFIG_SECURITY
2767 static int proc_pid_attr_open(struct inode *inode, struct file *file)
2768 {
2769 	file->private_data = NULL;
2770 	__mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
2771 	return 0;
2772 }
2773 
2774 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2775 				  size_t count, loff_t *ppos)
2776 {
2777 	struct inode * inode = file_inode(file);
2778 	char *p = NULL;
2779 	ssize_t length;
2780 	struct task_struct *task = get_proc_task(inode);
2781 
2782 	if (!task)
2783 		return -ESRCH;
2784 
2785 	length = security_getprocattr(task, PROC_I(inode)->op.lsmid,
2786 				      file->f_path.dentry->d_name.name,
2787 				      &p);
2788 	put_task_struct(task);
2789 	if (length > 0)
2790 		length = simple_read_from_buffer(buf, count, ppos, p, length);
2791 	kfree(p);
2792 	return length;
2793 }
2794 
2795 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2796 				   size_t count, loff_t *ppos)
2797 {
2798 	struct inode * inode = file_inode(file);
2799 	struct task_struct *task;
2800 	void *page;
2801 	int rv;
2802 
2803 	/* A task may only write when it was the opener. */
2804 	if (file->private_data != current->mm)
2805 		return -EPERM;
2806 
2807 	rcu_read_lock();
2808 	task = pid_task(proc_pid(inode), PIDTYPE_PID);
2809 	if (!task) {
2810 		rcu_read_unlock();
2811 		return -ESRCH;
2812 	}
2813 	/* A task may only write its own attributes. */
2814 	if (current != task) {
2815 		rcu_read_unlock();
2816 		return -EACCES;
2817 	}
2818 	/* Prevent changes to overridden credentials. */
2819 	if (current_cred() != current_real_cred()) {
2820 		rcu_read_unlock();
2821 		return -EBUSY;
2822 	}
2823 	rcu_read_unlock();
2824 
2825 	if (count > PAGE_SIZE)
2826 		count = PAGE_SIZE;
2827 
2828 	/* No partial writes. */
2829 	if (*ppos != 0)
2830 		return -EINVAL;
2831 
2832 	page = memdup_user(buf, count);
2833 	if (IS_ERR(page)) {
2834 		rv = PTR_ERR(page);
2835 		goto out;
2836 	}
2837 
2838 	/* Guard against adverse ptrace interaction */
2839 	rv = mutex_lock_interruptible(&current->signal->cred_guard_mutex);
2840 	if (rv < 0)
2841 		goto out_free;
2842 
2843 	rv = security_setprocattr(PROC_I(inode)->op.lsmid,
2844 				  file->f_path.dentry->d_name.name, page,
2845 				  count);
2846 	mutex_unlock(&current->signal->cred_guard_mutex);
2847 out_free:
2848 	kfree(page);
2849 out:
2850 	return rv;
2851 }
2852 
2853 static const struct file_operations proc_pid_attr_operations = {
2854 	.open		= proc_pid_attr_open,
2855 	.read		= proc_pid_attr_read,
2856 	.write		= proc_pid_attr_write,
2857 	.llseek		= generic_file_llseek,
2858 	.release	= mem_release,
2859 };
2860 
2861 #define LSM_DIR_OPS(LSM) \
2862 static int proc_##LSM##_attr_dir_iterate(struct file *filp, \
2863 			     struct dir_context *ctx) \
2864 { \
2865 	return proc_pident_readdir(filp, ctx, \
2866 				   LSM##_attr_dir_stuff, \
2867 				   ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2868 } \
2869 \
2870 static const struct file_operations proc_##LSM##_attr_dir_ops = { \
2871 	.read		= generic_read_dir, \
2872 	.iterate_shared	= proc_##LSM##_attr_dir_iterate, \
2873 	.llseek		= default_llseek, \
2874 }; \
2875 \
2876 static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \
2877 				struct dentry *dentry, unsigned int flags) \
2878 { \
2879 	return proc_pident_lookup(dir, dentry, \
2880 				  LSM##_attr_dir_stuff, \
2881 				  LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2882 } \
2883 \
2884 static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \
2885 	.lookup		= proc_##LSM##_attr_dir_lookup, \
2886 	.getattr	= pid_getattr, \
2887 	.setattr	= proc_setattr, \
2888 }
2889 
2890 #ifdef CONFIG_SECURITY_SMACK
2891 static const struct pid_entry smack_attr_dir_stuff[] = {
2892 	ATTR(LSM_ID_SMACK, "current",	0666),
2893 };
2894 LSM_DIR_OPS(smack);
2895 #endif
2896 
2897 #ifdef CONFIG_SECURITY_APPARMOR
2898 static const struct pid_entry apparmor_attr_dir_stuff[] = {
2899 	ATTR(LSM_ID_APPARMOR, "current",	0666),
2900 	ATTR(LSM_ID_APPARMOR, "prev",		0444),
2901 	ATTR(LSM_ID_APPARMOR, "exec",		0666),
2902 };
2903 LSM_DIR_OPS(apparmor);
2904 #endif
2905 
2906 static const struct pid_entry attr_dir_stuff[] = {
2907 	ATTR(LSM_ID_UNDEF, "current",	0666),
2908 	ATTR(LSM_ID_UNDEF, "prev",		0444),
2909 	ATTR(LSM_ID_UNDEF, "exec",		0666),
2910 	ATTR(LSM_ID_UNDEF, "fscreate",	0666),
2911 	ATTR(LSM_ID_UNDEF, "keycreate",	0666),
2912 	ATTR(LSM_ID_UNDEF, "sockcreate",	0666),
2913 #ifdef CONFIG_SECURITY_SMACK
2914 	DIR("smack",			0555,
2915 	    proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops),
2916 #endif
2917 #ifdef CONFIG_SECURITY_APPARMOR
2918 	DIR("apparmor",			0555,
2919 	    proc_apparmor_attr_dir_inode_ops, proc_apparmor_attr_dir_ops),
2920 #endif
2921 };
2922 
2923 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2924 {
2925 	return proc_pident_readdir(file, ctx,
2926 				   attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2927 }
2928 
2929 static const struct file_operations proc_attr_dir_operations = {
2930 	.read		= generic_read_dir,
2931 	.iterate_shared	= proc_attr_dir_readdir,
2932 	.llseek		= generic_file_llseek,
2933 };
2934 
2935 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2936 				struct dentry *dentry, unsigned int flags)
2937 {
2938 	return proc_pident_lookup(dir, dentry,
2939 				  attr_dir_stuff,
2940 				  attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff));
2941 }
2942 
2943 static const struct inode_operations proc_attr_dir_inode_operations = {
2944 	.lookup		= proc_attr_dir_lookup,
2945 	.getattr	= pid_getattr,
2946 	.setattr	= proc_setattr,
2947 };
2948 
2949 #endif
2950 
2951 #ifdef CONFIG_ELF_CORE
2952 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2953 					 size_t count, loff_t *ppos)
2954 {
2955 	struct task_struct *task = get_proc_task(file_inode(file));
2956 	struct mm_struct *mm;
2957 	char buffer[PROC_NUMBUF];
2958 	size_t len;
2959 	int ret;
2960 
2961 	if (!task)
2962 		return -ESRCH;
2963 
2964 	ret = 0;
2965 	mm = get_task_mm(task);
2966 	if (mm) {
2967 		len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2968 			       ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2969 				MMF_DUMP_FILTER_SHIFT));
2970 		mmput(mm);
2971 		ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2972 	}
2973 
2974 	put_task_struct(task);
2975 
2976 	return ret;
2977 }
2978 
2979 static ssize_t proc_coredump_filter_write(struct file *file,
2980 					  const char __user *buf,
2981 					  size_t count,
2982 					  loff_t *ppos)
2983 {
2984 	struct task_struct *task;
2985 	struct mm_struct *mm;
2986 	unsigned int val;
2987 	int ret;
2988 	int i;
2989 	unsigned long mask;
2990 
2991 	ret = kstrtouint_from_user(buf, count, 0, &val);
2992 	if (ret < 0)
2993 		return ret;
2994 
2995 	ret = -ESRCH;
2996 	task = get_proc_task(file_inode(file));
2997 	if (!task)
2998 		goto out_no_task;
2999 
3000 	mm = get_task_mm(task);
3001 	if (!mm)
3002 		goto out_no_mm;
3003 	ret = 0;
3004 
3005 	for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
3006 		if (val & mask)
3007 			set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
3008 		else
3009 			clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
3010 	}
3011 
3012 	mmput(mm);
3013  out_no_mm:
3014 	put_task_struct(task);
3015  out_no_task:
3016 	if (ret < 0)
3017 		return ret;
3018 	return count;
3019 }
3020 
3021 static const struct file_operations proc_coredump_filter_operations = {
3022 	.read		= proc_coredump_filter_read,
3023 	.write		= proc_coredump_filter_write,
3024 	.llseek		= generic_file_llseek,
3025 };
3026 #endif
3027 
3028 #ifdef CONFIG_TASK_IO_ACCOUNTING
3029 static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
3030 {
3031 	struct task_io_accounting acct;
3032 	int result;
3033 
3034 	result = down_read_killable(&task->signal->exec_update_lock);
3035 	if (result)
3036 		return result;
3037 
3038 	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
3039 		result = -EACCES;
3040 		goto out_unlock;
3041 	}
3042 
3043 	if (whole) {
3044 		struct signal_struct *sig = task->signal;
3045 		struct task_struct *t;
3046 		unsigned int seq = 1;
3047 		unsigned long flags;
3048 
3049 		rcu_read_lock();
3050 		do {
3051 			seq++; /* 2 on the 1st/lockless path, otherwise odd */
3052 			flags = read_seqbegin_or_lock_irqsave(&sig->stats_lock, &seq);
3053 
3054 			acct = sig->ioac;
3055 			__for_each_thread(sig, t)
3056 				task_io_accounting_add(&acct, &t->ioac);
3057 
3058 		} while (need_seqretry(&sig->stats_lock, seq));
3059 		done_seqretry_irqrestore(&sig->stats_lock, seq, flags);
3060 		rcu_read_unlock();
3061 	} else {
3062 		acct = task->ioac;
3063 	}
3064 
3065 	seq_printf(m,
3066 		   "rchar: %llu\n"
3067 		   "wchar: %llu\n"
3068 		   "syscr: %llu\n"
3069 		   "syscw: %llu\n"
3070 		   "read_bytes: %llu\n"
3071 		   "write_bytes: %llu\n"
3072 		   "cancelled_write_bytes: %llu\n",
3073 		   (unsigned long long)acct.rchar,
3074 		   (unsigned long long)acct.wchar,
3075 		   (unsigned long long)acct.syscr,
3076 		   (unsigned long long)acct.syscw,
3077 		   (unsigned long long)acct.read_bytes,
3078 		   (unsigned long long)acct.write_bytes,
3079 		   (unsigned long long)acct.cancelled_write_bytes);
3080 	result = 0;
3081 
3082 out_unlock:
3083 	up_read(&task->signal->exec_update_lock);
3084 	return result;
3085 }
3086 
3087 static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
3088 				  struct pid *pid, struct task_struct *task)
3089 {
3090 	return do_io_accounting(task, m, 0);
3091 }
3092 
3093 static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
3094 				   struct pid *pid, struct task_struct *task)
3095 {
3096 	return do_io_accounting(task, m, 1);
3097 }
3098 #endif /* CONFIG_TASK_IO_ACCOUNTING */
3099 
3100 #ifdef CONFIG_USER_NS
3101 static int proc_id_map_open(struct inode *inode, struct file *file,
3102 	const struct seq_operations *seq_ops)
3103 {
3104 	struct user_namespace *ns = NULL;
3105 	struct task_struct *task;
3106 	struct seq_file *seq;
3107 	int ret = -EINVAL;
3108 
3109 	task = get_proc_task(inode);
3110 	if (task) {
3111 		rcu_read_lock();
3112 		ns = get_user_ns(task_cred_xxx(task, user_ns));
3113 		rcu_read_unlock();
3114 		put_task_struct(task);
3115 	}
3116 	if (!ns)
3117 		goto err;
3118 
3119 	ret = seq_open(file, seq_ops);
3120 	if (ret)
3121 		goto err_put_ns;
3122 
3123 	seq = file->private_data;
3124 	seq->private = ns;
3125 
3126 	return 0;
3127 err_put_ns:
3128 	put_user_ns(ns);
3129 err:
3130 	return ret;
3131 }
3132 
3133 static int proc_id_map_release(struct inode *inode, struct file *file)
3134 {
3135 	struct seq_file *seq = file->private_data;
3136 	struct user_namespace *ns = seq->private;
3137 	put_user_ns(ns);
3138 	return seq_release(inode, file);
3139 }
3140 
3141 static int proc_uid_map_open(struct inode *inode, struct file *file)
3142 {
3143 	return proc_id_map_open(inode, file, &proc_uid_seq_operations);
3144 }
3145 
3146 static int proc_gid_map_open(struct inode *inode, struct file *file)
3147 {
3148 	return proc_id_map_open(inode, file, &proc_gid_seq_operations);
3149 }
3150 
3151 static int proc_projid_map_open(struct inode *inode, struct file *file)
3152 {
3153 	return proc_id_map_open(inode, file, &proc_projid_seq_operations);
3154 }
3155 
3156 static const struct file_operations proc_uid_map_operations = {
3157 	.open		= proc_uid_map_open,
3158 	.write		= proc_uid_map_write,
3159 	.read		= seq_read,
3160 	.llseek		= seq_lseek,
3161 	.release	= proc_id_map_release,
3162 };
3163 
3164 static const struct file_operations proc_gid_map_operations = {
3165 	.open		= proc_gid_map_open,
3166 	.write		= proc_gid_map_write,
3167 	.read		= seq_read,
3168 	.llseek		= seq_lseek,
3169 	.release	= proc_id_map_release,
3170 };
3171 
3172 static const struct file_operations proc_projid_map_operations = {
3173 	.open		= proc_projid_map_open,
3174 	.write		= proc_projid_map_write,
3175 	.read		= seq_read,
3176 	.llseek		= seq_lseek,
3177 	.release	= proc_id_map_release,
3178 };
3179 
3180 static int proc_setgroups_open(struct inode *inode, struct file *file)
3181 {
3182 	struct user_namespace *ns = NULL;
3183 	struct task_struct *task;
3184 	int ret;
3185 
3186 	ret = -ESRCH;
3187 	task = get_proc_task(inode);
3188 	if (task) {
3189 		rcu_read_lock();
3190 		ns = get_user_ns(task_cred_xxx(task, user_ns));
3191 		rcu_read_unlock();
3192 		put_task_struct(task);
3193 	}
3194 	if (!ns)
3195 		goto err;
3196 
3197 	if (file->f_mode & FMODE_WRITE) {
3198 		ret = -EACCES;
3199 		if (!ns_capable(ns, CAP_SYS_ADMIN))
3200 			goto err_put_ns;
3201 	}
3202 
3203 	ret = single_open(file, &proc_setgroups_show, ns);
3204 	if (ret)
3205 		goto err_put_ns;
3206 
3207 	return 0;
3208 err_put_ns:
3209 	put_user_ns(ns);
3210 err:
3211 	return ret;
3212 }
3213 
3214 static int proc_setgroups_release(struct inode *inode, struct file *file)
3215 {
3216 	struct seq_file *seq = file->private_data;
3217 	struct user_namespace *ns = seq->private;
3218 	int ret = single_release(inode, file);
3219 	put_user_ns(ns);
3220 	return ret;
3221 }
3222 
3223 static const struct file_operations proc_setgroups_operations = {
3224 	.open		= proc_setgroups_open,
3225 	.write		= proc_setgroups_write,
3226 	.read		= seq_read,
3227 	.llseek		= seq_lseek,
3228 	.release	= proc_setgroups_release,
3229 };
3230 #endif /* CONFIG_USER_NS */
3231 
3232 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3233 				struct pid *pid, struct task_struct *task)
3234 {
3235 	int err = lock_trace(task);
3236 	if (!err) {
3237 		seq_printf(m, "%08x\n", task->personality);
3238 		unlock_trace(task);
3239 	}
3240 	return err;
3241 }
3242 
3243 #ifdef CONFIG_LIVEPATCH
3244 static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
3245 				struct pid *pid, struct task_struct *task)
3246 {
3247 	seq_printf(m, "%d\n", task->patch_state);
3248 	return 0;
3249 }
3250 #endif /* CONFIG_LIVEPATCH */
3251 
3252 #ifdef CONFIG_KSM
3253 static int proc_pid_ksm_merging_pages(struct seq_file *m, struct pid_namespace *ns,
3254 				struct pid *pid, struct task_struct *task)
3255 {
3256 	struct mm_struct *mm;
3257 
3258 	mm = get_task_mm(task);
3259 	if (mm) {
3260 		seq_printf(m, "%lu\n", mm->ksm_merging_pages);
3261 		mmput(mm);
3262 	}
3263 
3264 	return 0;
3265 }
3266 static int proc_pid_ksm_stat(struct seq_file *m, struct pid_namespace *ns,
3267 				struct pid *pid, struct task_struct *task)
3268 {
3269 	struct mm_struct *mm;
3270 
3271 	mm = get_task_mm(task);
3272 	if (mm) {
3273 		seq_printf(m, "ksm_rmap_items %lu\n", mm->ksm_rmap_items);
3274 		seq_printf(m, "ksm_zero_pages %ld\n", mm_ksm_zero_pages(mm));
3275 		seq_printf(m, "ksm_merging_pages %lu\n", mm->ksm_merging_pages);
3276 		seq_printf(m, "ksm_process_profit %ld\n", ksm_process_profit(mm));
3277 		mmput(mm);
3278 	}
3279 
3280 	return 0;
3281 }
3282 #endif /* CONFIG_KSM */
3283 
3284 #ifdef CONFIG_STACKLEAK_METRICS
3285 static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns,
3286 				struct pid *pid, struct task_struct *task)
3287 {
3288 	unsigned long prev_depth = THREAD_SIZE -
3289 				(task->prev_lowest_stack & (THREAD_SIZE - 1));
3290 	unsigned long depth = THREAD_SIZE -
3291 				(task->lowest_stack & (THREAD_SIZE - 1));
3292 
3293 	seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n",
3294 							prev_depth, depth);
3295 	return 0;
3296 }
3297 #endif /* CONFIG_STACKLEAK_METRICS */
3298 
3299 /*
3300  * Thread groups
3301  */
3302 static const struct file_operations proc_task_operations;
3303 static const struct inode_operations proc_task_inode_operations;
3304 
3305 static const struct pid_entry tgid_base_stuff[] = {
3306 	DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3307 	DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3308 	DIR("map_files",  S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3309 	DIR("fdinfo",     S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3310 	DIR("ns",	  S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3311 #ifdef CONFIG_NET
3312 	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3313 #endif
3314 	REG("environ",    S_IRUSR, proc_environ_operations),
3315 	REG("auxv",       S_IRUSR, proc_auxv_operations),
3316 	ONE("status",     S_IRUGO, proc_pid_status),
3317 	ONE("personality", S_IRUSR, proc_pid_personality),
3318 	ONE("limits",	  S_IRUGO, proc_pid_limits),
3319 #ifdef CONFIG_SCHED_DEBUG
3320 	REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3321 #endif
3322 #ifdef CONFIG_SCHED_AUTOGROUP
3323 	REG("autogroup",  S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3324 #endif
3325 #ifdef CONFIG_TIME_NS
3326 	REG("timens_offsets",  S_IRUGO|S_IWUSR, proc_timens_offsets_operations),
3327 #endif
3328 	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3329 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3330 	ONE("syscall",    S_IRUSR, proc_pid_syscall),
3331 #endif
3332 	REG("cmdline",    S_IRUGO, proc_pid_cmdline_ops),
3333 	ONE("stat",       S_IRUGO, proc_tgid_stat),
3334 	ONE("statm",      S_IRUGO, proc_pid_statm),
3335 	REG("maps",       S_IRUGO, proc_pid_maps_operations),
3336 #ifdef CONFIG_NUMA
3337 	REG("numa_maps",  S_IRUGO, proc_pid_numa_maps_operations),
3338 #endif
3339 	REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
3340 	LNK("cwd",        proc_cwd_link),
3341 	LNK("root",       proc_root_link),
3342 	LNK("exe",        proc_exe_link),
3343 	REG("mounts",     S_IRUGO, proc_mounts_operations),
3344 	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3345 	REG("mountstats", S_IRUSR, proc_mountstats_operations),
3346 #ifdef CONFIG_PROC_PAGE_MONITOR
3347 	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3348 	REG("smaps",      S_IRUGO, proc_pid_smaps_operations),
3349 	REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3350 	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3351 #endif
3352 #ifdef CONFIG_SECURITY
3353 	DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3354 #endif
3355 #ifdef CONFIG_KALLSYMS
3356 	ONE("wchan",      S_IRUGO, proc_pid_wchan),
3357 #endif
3358 #ifdef CONFIG_STACKTRACE
3359 	ONE("stack",      S_IRUSR, proc_pid_stack),
3360 #endif
3361 #ifdef CONFIG_SCHED_INFO
3362 	ONE("schedstat",  S_IRUGO, proc_pid_schedstat),
3363 #endif
3364 #ifdef CONFIG_LATENCYTOP
3365 	REG("latency",  S_IRUGO, proc_lstats_operations),
3366 #endif
3367 #ifdef CONFIG_PROC_PID_CPUSET
3368 	ONE("cpuset",     S_IRUGO, proc_cpuset_show),
3369 #endif
3370 #ifdef CONFIG_CGROUPS
3371 	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3372 #endif
3373 #ifdef CONFIG_PROC_CPU_RESCTRL
3374 	ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3375 #endif
3376 	ONE("oom_score",  S_IRUGO, proc_oom_score),
3377 	REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3378 	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3379 #ifdef CONFIG_AUDIT
3380 	REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
3381 	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3382 #endif
3383 #ifdef CONFIG_FAULT_INJECTION
3384 	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3385 	REG("fail-nth", 0644, proc_fail_nth_operations),
3386 #endif
3387 #ifdef CONFIG_ELF_CORE
3388 	REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3389 #endif
3390 #ifdef CONFIG_TASK_IO_ACCOUNTING
3391 	ONE("io",	S_IRUSR, proc_tgid_io_accounting),
3392 #endif
3393 #ifdef CONFIG_USER_NS
3394 	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3395 	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3396 	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3397 	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3398 #endif
3399 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3400 	REG("timers",	  S_IRUGO, proc_timers_operations),
3401 #endif
3402 	REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
3403 #ifdef CONFIG_LIVEPATCH
3404 	ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3405 #endif
3406 #ifdef CONFIG_STACKLEAK_METRICS
3407 	ONE("stack_depth", S_IRUGO, proc_stack_depth),
3408 #endif
3409 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3410 	ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3411 #endif
3412 #ifdef CONFIG_SECCOMP_CACHE_DEBUG
3413 	ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache),
3414 #endif
3415 #ifdef CONFIG_KSM
3416 	ONE("ksm_merging_pages",  S_IRUSR, proc_pid_ksm_merging_pages),
3417 	ONE("ksm_stat",  S_IRUSR, proc_pid_ksm_stat),
3418 #endif
3419 };
3420 
3421 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
3422 {
3423 	return proc_pident_readdir(file, ctx,
3424 				   tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3425 }
3426 
3427 static const struct file_operations proc_tgid_base_operations = {
3428 	.read		= generic_read_dir,
3429 	.iterate_shared	= proc_tgid_base_readdir,
3430 	.llseek		= generic_file_llseek,
3431 };
3432 
3433 struct pid *tgid_pidfd_to_pid(const struct file *file)
3434 {
3435 	if (file->f_op != &proc_tgid_base_operations)
3436 		return ERR_PTR(-EBADF);
3437 
3438 	return proc_pid(file_inode(file));
3439 }
3440 
3441 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3442 {
3443 	return proc_pident_lookup(dir, dentry,
3444 				  tgid_base_stuff,
3445 				  tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff));
3446 }
3447 
3448 static const struct inode_operations proc_tgid_base_inode_operations = {
3449 	.lookup		= proc_tgid_base_lookup,
3450 	.getattr	= pid_getattr,
3451 	.setattr	= proc_setattr,
3452 	.permission	= proc_pid_permission,
3453 };
3454 
3455 /**
3456  * proc_flush_pid -  Remove dcache entries for @pid from the /proc dcache.
3457  * @pid: pid that should be flushed.
3458  *
3459  * This function walks a list of inodes (that belong to any proc
3460  * filesystem) that are attached to the pid and flushes them from
3461  * the dentry cache.
3462  *
3463  * It is safe and reasonable to cache /proc entries for a task until
3464  * that task exits.  After that they just clog up the dcache with
3465  * useless entries, possibly causing useful dcache entries to be
3466  * flushed instead.  This routine is provided to flush those useless
3467  * dcache entries when a process is reaped.
3468  *
3469  * NOTE: This routine is just an optimization so it does not guarantee
3470  *       that no dcache entries will exist after a process is reaped
3471  *       it just makes it very unlikely that any will persist.
3472  */
3473 
3474 void proc_flush_pid(struct pid *pid)
3475 {
3476 	proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock);
3477 }
3478 
3479 static struct dentry *proc_pid_instantiate(struct dentry * dentry,
3480 				   struct task_struct *task, const void *ptr)
3481 {
3482 	struct inode *inode;
3483 
3484 	inode = proc_pid_make_base_inode(dentry->d_sb, task,
3485 					 S_IFDIR | S_IRUGO | S_IXUGO);
3486 	if (!inode)
3487 		return ERR_PTR(-ENOENT);
3488 
3489 	inode->i_op = &proc_tgid_base_inode_operations;
3490 	inode->i_fop = &proc_tgid_base_operations;
3491 	inode->i_flags|=S_IMMUTABLE;
3492 
3493 	set_nlink(inode, nlink_tgid);
3494 	pid_update_inode(task, inode);
3495 
3496 	d_set_d_op(dentry, &pid_dentry_operations);
3497 	return d_splice_alias(inode, dentry);
3498 }
3499 
3500 struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags)
3501 {
3502 	struct task_struct *task;
3503 	unsigned tgid;
3504 	struct proc_fs_info *fs_info;
3505 	struct pid_namespace *ns;
3506 	struct dentry *result = ERR_PTR(-ENOENT);
3507 
3508 	tgid = name_to_int(&dentry->d_name);
3509 	if (tgid == ~0U)
3510 		goto out;
3511 
3512 	fs_info = proc_sb_info(dentry->d_sb);
3513 	ns = fs_info->pid_ns;
3514 	rcu_read_lock();
3515 	task = find_task_by_pid_ns(tgid, ns);
3516 	if (task)
3517 		get_task_struct(task);
3518 	rcu_read_unlock();
3519 	if (!task)
3520 		goto out;
3521 
3522 	/* Limit procfs to only ptraceable tasks */
3523 	if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) {
3524 		if (!has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS))
3525 			goto out_put_task;
3526 	}
3527 
3528 	result = proc_pid_instantiate(dentry, task, NULL);
3529 out_put_task:
3530 	put_task_struct(task);
3531 out:
3532 	return result;
3533 }
3534 
3535 /*
3536  * Find the first task with tgid >= tgid
3537  *
3538  */
3539 struct tgid_iter {
3540 	unsigned int tgid;
3541 	struct task_struct *task;
3542 };
3543 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3544 {
3545 	struct pid *pid;
3546 
3547 	if (iter.task)
3548 		put_task_struct(iter.task);
3549 	rcu_read_lock();
3550 retry:
3551 	iter.task = NULL;
3552 	pid = find_ge_pid(iter.tgid, ns);
3553 	if (pid) {
3554 		iter.tgid = pid_nr_ns(pid, ns);
3555 		iter.task = pid_task(pid, PIDTYPE_TGID);
3556 		if (!iter.task) {
3557 			iter.tgid += 1;
3558 			goto retry;
3559 		}
3560 		get_task_struct(iter.task);
3561 	}
3562 	rcu_read_unlock();
3563 	return iter;
3564 }
3565 
3566 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3567 
3568 /* for the /proc/ directory itself, after non-process stuff has been done */
3569 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3570 {
3571 	struct tgid_iter iter;
3572 	struct proc_fs_info *fs_info = proc_sb_info(file_inode(file)->i_sb);
3573 	struct pid_namespace *ns = proc_pid_ns(file_inode(file)->i_sb);
3574 	loff_t pos = ctx->pos;
3575 
3576 	if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3577 		return 0;
3578 
3579 	if (pos == TGID_OFFSET - 2) {
3580 		struct inode *inode = d_inode(fs_info->proc_self);
3581 		if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3582 			return 0;
3583 		ctx->pos = pos = pos + 1;
3584 	}
3585 	if (pos == TGID_OFFSET - 1) {
3586 		struct inode *inode = d_inode(fs_info->proc_thread_self);
3587 		if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3588 			return 0;
3589 		ctx->pos = pos = pos + 1;
3590 	}
3591 	iter.tgid = pos - TGID_OFFSET;
3592 	iter.task = NULL;
3593 	for (iter = next_tgid(ns, iter);
3594 	     iter.task;
3595 	     iter.tgid += 1, iter = next_tgid(ns, iter)) {
3596 		char name[10 + 1];
3597 		unsigned int len;
3598 
3599 		cond_resched();
3600 		if (!has_pid_permissions(fs_info, iter.task, HIDEPID_INVISIBLE))
3601 			continue;
3602 
3603 		len = snprintf(name, sizeof(name), "%u", iter.tgid);
3604 		ctx->pos = iter.tgid + TGID_OFFSET;
3605 		if (!proc_fill_cache(file, ctx, name, len,
3606 				     proc_pid_instantiate, iter.task, NULL)) {
3607 			put_task_struct(iter.task);
3608 			return 0;
3609 		}
3610 	}
3611 	ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3612 	return 0;
3613 }
3614 
3615 /*
3616  * proc_tid_comm_permission is a special permission function exclusively
3617  * used for the node /proc/<pid>/task/<tid>/comm.
3618  * It bypasses generic permission checks in the case where a task of the same
3619  * task group attempts to access the node.
3620  * The rationale behind this is that glibc and bionic access this node for
3621  * cross thread naming (pthread_set/getname_np(!self)). However, if
3622  * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3623  * which locks out the cross thread naming implementation.
3624  * This function makes sure that the node is always accessible for members of
3625  * same thread group.
3626  */
3627 static int proc_tid_comm_permission(struct mnt_idmap *idmap,
3628 				    struct inode *inode, int mask)
3629 {
3630 	bool is_same_tgroup;
3631 	struct task_struct *task;
3632 
3633 	task = get_proc_task(inode);
3634 	if (!task)
3635 		return -ESRCH;
3636 	is_same_tgroup = same_thread_group(current, task);
3637 	put_task_struct(task);
3638 
3639 	if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3640 		/* This file (/proc/<pid>/task/<tid>/comm) can always be
3641 		 * read or written by the members of the corresponding
3642 		 * thread group.
3643 		 */
3644 		return 0;
3645 	}
3646 
3647 	return generic_permission(&nop_mnt_idmap, inode, mask);
3648 }
3649 
3650 static const struct inode_operations proc_tid_comm_inode_operations = {
3651 		.setattr	= proc_setattr,
3652 		.permission	= proc_tid_comm_permission,
3653 };
3654 
3655 /*
3656  * Tasks
3657  */
3658 static const struct pid_entry tid_base_stuff[] = {
3659 	DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3660 	DIR("fdinfo",    S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3661 	DIR("ns",	 S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3662 #ifdef CONFIG_NET
3663 	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3664 #endif
3665 	REG("environ",   S_IRUSR, proc_environ_operations),
3666 	REG("auxv",      S_IRUSR, proc_auxv_operations),
3667 	ONE("status",    S_IRUGO, proc_pid_status),
3668 	ONE("personality", S_IRUSR, proc_pid_personality),
3669 	ONE("limits",	 S_IRUGO, proc_pid_limits),
3670 #ifdef CONFIG_SCHED_DEBUG
3671 	REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3672 #endif
3673 	NOD("comm",      S_IFREG|S_IRUGO|S_IWUSR,
3674 			 &proc_tid_comm_inode_operations,
3675 			 &proc_pid_set_comm_operations, {}),
3676 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3677 	ONE("syscall",   S_IRUSR, proc_pid_syscall),
3678 #endif
3679 	REG("cmdline",   S_IRUGO, proc_pid_cmdline_ops),
3680 	ONE("stat",      S_IRUGO, proc_tid_stat),
3681 	ONE("statm",     S_IRUGO, proc_pid_statm),
3682 	REG("maps",      S_IRUGO, proc_pid_maps_operations),
3683 #ifdef CONFIG_PROC_CHILDREN
3684 	REG("children",  S_IRUGO, proc_tid_children_operations),
3685 #endif
3686 #ifdef CONFIG_NUMA
3687 	REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3688 #endif
3689 	REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
3690 	LNK("cwd",       proc_cwd_link),
3691 	LNK("root",      proc_root_link),
3692 	LNK("exe",       proc_exe_link),
3693 	REG("mounts",    S_IRUGO, proc_mounts_operations),
3694 	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3695 #ifdef CONFIG_PROC_PAGE_MONITOR
3696 	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3697 	REG("smaps",     S_IRUGO, proc_pid_smaps_operations),
3698 	REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3699 	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3700 #endif
3701 #ifdef CONFIG_SECURITY
3702 	DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3703 #endif
3704 #ifdef CONFIG_KALLSYMS
3705 	ONE("wchan",     S_IRUGO, proc_pid_wchan),
3706 #endif
3707 #ifdef CONFIG_STACKTRACE
3708 	ONE("stack",      S_IRUSR, proc_pid_stack),
3709 #endif
3710 #ifdef CONFIG_SCHED_INFO
3711 	ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3712 #endif
3713 #ifdef CONFIG_LATENCYTOP
3714 	REG("latency",  S_IRUGO, proc_lstats_operations),
3715 #endif
3716 #ifdef CONFIG_PROC_PID_CPUSET
3717 	ONE("cpuset",    S_IRUGO, proc_cpuset_show),
3718 #endif
3719 #ifdef CONFIG_CGROUPS
3720 	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3721 #endif
3722 #ifdef CONFIG_PROC_CPU_RESCTRL
3723 	ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3724 #endif
3725 	ONE("oom_score", S_IRUGO, proc_oom_score),
3726 	REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3727 	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3728 #ifdef CONFIG_AUDIT
3729 	REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
3730 	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3731 #endif
3732 #ifdef CONFIG_FAULT_INJECTION
3733 	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3734 	REG("fail-nth", 0644, proc_fail_nth_operations),
3735 #endif
3736 #ifdef CONFIG_TASK_IO_ACCOUNTING
3737 	ONE("io",	S_IRUSR, proc_tid_io_accounting),
3738 #endif
3739 #ifdef CONFIG_USER_NS
3740 	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3741 	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3742 	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3743 	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3744 #endif
3745 #ifdef CONFIG_LIVEPATCH
3746 	ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3747 #endif
3748 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3749 	ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3750 #endif
3751 #ifdef CONFIG_SECCOMP_CACHE_DEBUG
3752 	ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache),
3753 #endif
3754 #ifdef CONFIG_KSM
3755 	ONE("ksm_merging_pages",  S_IRUSR, proc_pid_ksm_merging_pages),
3756 	ONE("ksm_stat",  S_IRUSR, proc_pid_ksm_stat),
3757 #endif
3758 };
3759 
3760 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3761 {
3762 	return proc_pident_readdir(file, ctx,
3763 				   tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3764 }
3765 
3766 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3767 {
3768 	return proc_pident_lookup(dir, dentry,
3769 				  tid_base_stuff,
3770 				  tid_base_stuff + ARRAY_SIZE(tid_base_stuff));
3771 }
3772 
3773 static const struct file_operations proc_tid_base_operations = {
3774 	.read		= generic_read_dir,
3775 	.iterate_shared	= proc_tid_base_readdir,
3776 	.llseek		= generic_file_llseek,
3777 };
3778 
3779 static const struct inode_operations proc_tid_base_inode_operations = {
3780 	.lookup		= proc_tid_base_lookup,
3781 	.getattr	= pid_getattr,
3782 	.setattr	= proc_setattr,
3783 };
3784 
3785 static struct dentry *proc_task_instantiate(struct dentry *dentry,
3786 	struct task_struct *task, const void *ptr)
3787 {
3788 	struct inode *inode;
3789 	inode = proc_pid_make_base_inode(dentry->d_sb, task,
3790 					 S_IFDIR | S_IRUGO | S_IXUGO);
3791 	if (!inode)
3792 		return ERR_PTR(-ENOENT);
3793 
3794 	inode->i_op = &proc_tid_base_inode_operations;
3795 	inode->i_fop = &proc_tid_base_operations;
3796 	inode->i_flags |= S_IMMUTABLE;
3797 
3798 	set_nlink(inode, nlink_tid);
3799 	pid_update_inode(task, inode);
3800 
3801 	d_set_d_op(dentry, &pid_dentry_operations);
3802 	return d_splice_alias(inode, dentry);
3803 }
3804 
3805 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3806 {
3807 	struct task_struct *task;
3808 	struct task_struct *leader = get_proc_task(dir);
3809 	unsigned tid;
3810 	struct proc_fs_info *fs_info;
3811 	struct pid_namespace *ns;
3812 	struct dentry *result = ERR_PTR(-ENOENT);
3813 
3814 	if (!leader)
3815 		goto out_no_task;
3816 
3817 	tid = name_to_int(&dentry->d_name);
3818 	if (tid == ~0U)
3819 		goto out;
3820 
3821 	fs_info = proc_sb_info(dentry->d_sb);
3822 	ns = fs_info->pid_ns;
3823 	rcu_read_lock();
3824 	task = find_task_by_pid_ns(tid, ns);
3825 	if (task)
3826 		get_task_struct(task);
3827 	rcu_read_unlock();
3828 	if (!task)
3829 		goto out;
3830 	if (!same_thread_group(leader, task))
3831 		goto out_drop_task;
3832 
3833 	result = proc_task_instantiate(dentry, task, NULL);
3834 out_drop_task:
3835 	put_task_struct(task);
3836 out:
3837 	put_task_struct(leader);
3838 out_no_task:
3839 	return result;
3840 }
3841 
3842 /*
3843  * Find the first tid of a thread group to return to user space.
3844  *
3845  * Usually this is just the thread group leader, but if the users
3846  * buffer was too small or there was a seek into the middle of the
3847  * directory we have more work todo.
3848  *
3849  * In the case of a short read we start with find_task_by_pid.
3850  *
3851  * In the case of a seek we start with the leader and walk nr
3852  * threads past it.
3853  */
3854 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3855 					struct pid_namespace *ns)
3856 {
3857 	struct task_struct *pos, *task;
3858 	unsigned long nr = f_pos;
3859 
3860 	if (nr != f_pos)	/* 32bit overflow? */
3861 		return NULL;
3862 
3863 	rcu_read_lock();
3864 	task = pid_task(pid, PIDTYPE_PID);
3865 	if (!task)
3866 		goto fail;
3867 
3868 	/* Attempt to start with the tid of a thread */
3869 	if (tid && nr) {
3870 		pos = find_task_by_pid_ns(tid, ns);
3871 		if (pos && same_thread_group(pos, task))
3872 			goto found;
3873 	}
3874 
3875 	/* If nr exceeds the number of threads there is nothing todo */
3876 	if (nr >= get_nr_threads(task))
3877 		goto fail;
3878 
3879 	/* If we haven't found our starting place yet start
3880 	 * with the leader and walk nr threads forward.
3881 	 */
3882 	for_each_thread(task, pos) {
3883 		if (!nr--)
3884 			goto found;
3885 	}
3886 fail:
3887 	pos = NULL;
3888 	goto out;
3889 found:
3890 	get_task_struct(pos);
3891 out:
3892 	rcu_read_unlock();
3893 	return pos;
3894 }
3895 
3896 /*
3897  * Find the next thread in the thread list.
3898  * Return NULL if there is an error or no next thread.
3899  *
3900  * The reference to the input task_struct is released.
3901  */
3902 static struct task_struct *next_tid(struct task_struct *start)
3903 {
3904 	struct task_struct *pos = NULL;
3905 	rcu_read_lock();
3906 	if (pid_alive(start)) {
3907 		pos = __next_thread(start);
3908 		if (pos)
3909 			get_task_struct(pos);
3910 	}
3911 	rcu_read_unlock();
3912 	put_task_struct(start);
3913 	return pos;
3914 }
3915 
3916 /* for the /proc/TGID/task/ directories */
3917 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3918 {
3919 	struct inode *inode = file_inode(file);
3920 	struct task_struct *task;
3921 	struct pid_namespace *ns;
3922 	int tid;
3923 
3924 	if (proc_inode_is_dead(inode))
3925 		return -ENOENT;
3926 
3927 	if (!dir_emit_dots(file, ctx))
3928 		return 0;
3929 
3930 	/* We cache the tgid value that the last readdir call couldn't
3931 	 * return and lseek resets it to 0.
3932 	 */
3933 	ns = proc_pid_ns(inode->i_sb);
3934 	tid = (int)(intptr_t)file->private_data;
3935 	file->private_data = NULL;
3936 	for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3937 	     task;
3938 	     task = next_tid(task), ctx->pos++) {
3939 		char name[10 + 1];
3940 		unsigned int len;
3941 
3942 		tid = task_pid_nr_ns(task, ns);
3943 		if (!tid)
3944 			continue;	/* The task has just exited. */
3945 		len = snprintf(name, sizeof(name), "%u", tid);
3946 		if (!proc_fill_cache(file, ctx, name, len,
3947 				proc_task_instantiate, task, NULL)) {
3948 			/* returning this tgid failed, save it as the first
3949 			 * pid for the next readir call */
3950 			file->private_data = (void *)(intptr_t)tid;
3951 			put_task_struct(task);
3952 			break;
3953 		}
3954 	}
3955 
3956 	return 0;
3957 }
3958 
3959 static int proc_task_getattr(struct mnt_idmap *idmap,
3960 			     const struct path *path, struct kstat *stat,
3961 			     u32 request_mask, unsigned int query_flags)
3962 {
3963 	struct inode *inode = d_inode(path->dentry);
3964 	struct task_struct *p = get_proc_task(inode);
3965 	generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
3966 
3967 	if (p) {
3968 		stat->nlink += get_nr_threads(p);
3969 		put_task_struct(p);
3970 	}
3971 
3972 	return 0;
3973 }
3974 
3975 /*
3976  * proc_task_readdir() set @file->private_data to a positive integer
3977  * value, so casting that to u64 is safe. generic_llseek_cookie() will
3978  * set @cookie to 0, so casting to an int is safe. The WARN_ON_ONCE() is
3979  * here to catch any unexpected change in behavior either in
3980  * proc_task_readdir() or generic_llseek_cookie().
3981  */
3982 static loff_t proc_dir_llseek(struct file *file, loff_t offset, int whence)
3983 {
3984 	u64 cookie = (u64)(intptr_t)file->private_data;
3985 	loff_t off;
3986 
3987 	off = generic_llseek_cookie(file, offset, whence, &cookie);
3988 	WARN_ON_ONCE(cookie > INT_MAX);
3989 	file->private_data = (void *)(intptr_t)cookie; /* serialized by f_pos_lock */
3990 	return off;
3991 }
3992 
3993 static const struct inode_operations proc_task_inode_operations = {
3994 	.lookup		= proc_task_lookup,
3995 	.getattr	= proc_task_getattr,
3996 	.setattr	= proc_setattr,
3997 	.permission	= proc_pid_permission,
3998 };
3999 
4000 static const struct file_operations proc_task_operations = {
4001 	.read		= generic_read_dir,
4002 	.iterate_shared	= proc_task_readdir,
4003 	.llseek		= proc_dir_llseek,
4004 };
4005 
4006 void __init set_proc_pid_nlink(void)
4007 {
4008 	nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
4009 	nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
4010 }
4011