xref: /linux/fs/proc/base.c (revision 9a379e77033f02c4a071891afdf0f0a01eff8ccb)
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/string.h>
63 #include <linux/seq_file.h>
64 #include <linux/namei.h>
65 #include <linux/mnt_namespace.h>
66 #include <linux/mm.h>
67 #include <linux/swap.h>
68 #include <linux/rcupdate.h>
69 #include <linux/kallsyms.h>
70 #include <linux/stacktrace.h>
71 #include <linux/resource.h>
72 #include <linux/module.h>
73 #include <linux/mount.h>
74 #include <linux/security.h>
75 #include <linux/ptrace.h>
76 #include <linux/tracehook.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_struct.h>
89 #include <linux/slab.h>
90 #include <linux/sched/autogroup.h>
91 #include <linux/sched/mm.h>
92 #include <linux/sched/coredump.h>
93 #include <linux/sched/debug.h>
94 #include <linux/sched/stat.h>
95 #include <linux/flex_array.h>
96 #include <linux/posix-timers.h>
97 #ifdef CONFIG_HARDWALL
98 #include <asm/hardwall.h>
99 #endif
100 #include <trace/events/oom.h>
101 #include "internal.h"
102 #include "fd.h"
103 
104 #include "../../lib/kstrtox.h"
105 
106 /* NOTE:
107  *	Implementing inode permission operations in /proc is almost
108  *	certainly an error.  Permission checks need to happen during
109  *	each system call not at open time.  The reason is that most of
110  *	what we wish to check for permissions in /proc varies at runtime.
111  *
112  *	The classic example of a problem is opening file descriptors
113  *	in /proc for a task before it execs a suid executable.
114  */
115 
116 static u8 nlink_tid __ro_after_init;
117 static u8 nlink_tgid __ro_after_init;
118 
119 struct pid_entry {
120 	const char *name;
121 	unsigned int len;
122 	umode_t mode;
123 	const struct inode_operations *iop;
124 	const struct file_operations *fop;
125 	union proc_op op;
126 };
127 
128 #define NOD(NAME, MODE, IOP, FOP, OP) {			\
129 	.name = (NAME),					\
130 	.len  = sizeof(NAME) - 1,			\
131 	.mode = MODE,					\
132 	.iop  = IOP,					\
133 	.fop  = FOP,					\
134 	.op   = OP,					\
135 }
136 
137 #define DIR(NAME, MODE, iops, fops)	\
138 	NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
139 #define LNK(NAME, get_link)					\
140 	NOD(NAME, (S_IFLNK|S_IRWXUGO),				\
141 		&proc_pid_link_inode_operations, NULL,		\
142 		{ .proc_get_link = get_link } )
143 #define REG(NAME, MODE, fops)				\
144 	NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
145 #define ONE(NAME, MODE, show)				\
146 	NOD(NAME, (S_IFREG|(MODE)), 			\
147 		NULL, &proc_single_file_operations,	\
148 		{ .proc_show = show } )
149 
150 /*
151  * Count the number of hardlinks for the pid_entry table, excluding the .
152  * and .. links.
153  */
154 static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
155 	unsigned int n)
156 {
157 	unsigned int i;
158 	unsigned int count;
159 
160 	count = 2;
161 	for (i = 0; i < n; ++i) {
162 		if (S_ISDIR(entries[i].mode))
163 			++count;
164 	}
165 
166 	return count;
167 }
168 
169 static int get_task_root(struct task_struct *task, struct path *root)
170 {
171 	int result = -ENOENT;
172 
173 	task_lock(task);
174 	if (task->fs) {
175 		get_fs_root(task->fs, root);
176 		result = 0;
177 	}
178 	task_unlock(task);
179 	return result;
180 }
181 
182 static int proc_cwd_link(struct dentry *dentry, struct path *path)
183 {
184 	struct task_struct *task = get_proc_task(d_inode(dentry));
185 	int result = -ENOENT;
186 
187 	if (task) {
188 		task_lock(task);
189 		if (task->fs) {
190 			get_fs_pwd(task->fs, path);
191 			result = 0;
192 		}
193 		task_unlock(task);
194 		put_task_struct(task);
195 	}
196 	return result;
197 }
198 
199 static int proc_root_link(struct dentry *dentry, struct path *path)
200 {
201 	struct task_struct *task = get_proc_task(d_inode(dentry));
202 	int result = -ENOENT;
203 
204 	if (task) {
205 		result = get_task_root(task, path);
206 		put_task_struct(task);
207 	}
208 	return result;
209 }
210 
211 static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
212 				     size_t _count, loff_t *pos)
213 {
214 	struct task_struct *tsk;
215 	struct mm_struct *mm;
216 	char *page;
217 	unsigned long count = _count;
218 	unsigned long arg_start, arg_end, env_start, env_end;
219 	unsigned long len1, len2, len;
220 	unsigned long p;
221 	char c;
222 	ssize_t rv;
223 
224 	BUG_ON(*pos < 0);
225 
226 	tsk = get_proc_task(file_inode(file));
227 	if (!tsk)
228 		return -ESRCH;
229 	mm = get_task_mm(tsk);
230 	put_task_struct(tsk);
231 	if (!mm)
232 		return 0;
233 	/* Check if process spawned far enough to have cmdline. */
234 	if (!mm->env_end) {
235 		rv = 0;
236 		goto out_mmput;
237 	}
238 
239 	page = (char *)__get_free_page(GFP_KERNEL);
240 	if (!page) {
241 		rv = -ENOMEM;
242 		goto out_mmput;
243 	}
244 
245 	down_read(&mm->mmap_sem);
246 	arg_start = mm->arg_start;
247 	arg_end = mm->arg_end;
248 	env_start = mm->env_start;
249 	env_end = mm->env_end;
250 	up_read(&mm->mmap_sem);
251 
252 	BUG_ON(arg_start > arg_end);
253 	BUG_ON(env_start > env_end);
254 
255 	len1 = arg_end - arg_start;
256 	len2 = env_end - env_start;
257 
258 	/* Empty ARGV. */
259 	if (len1 == 0) {
260 		rv = 0;
261 		goto out_free_page;
262 	}
263 	/*
264 	 * Inherently racy -- command line shares address space
265 	 * with code and data.
266 	 */
267 	rv = access_remote_vm(mm, arg_end - 1, &c, 1, 0);
268 	if (rv <= 0)
269 		goto out_free_page;
270 
271 	rv = 0;
272 
273 	if (c == '\0') {
274 		/* Command line (set of strings) occupies whole ARGV. */
275 		if (len1 <= *pos)
276 			goto out_free_page;
277 
278 		p = arg_start + *pos;
279 		len = len1 - *pos;
280 		while (count > 0 && len > 0) {
281 			unsigned int _count;
282 			int nr_read;
283 
284 			_count = min3(count, len, PAGE_SIZE);
285 			nr_read = access_remote_vm(mm, p, page, _count, 0);
286 			if (nr_read < 0)
287 				rv = nr_read;
288 			if (nr_read <= 0)
289 				goto out_free_page;
290 
291 			if (copy_to_user(buf, page, nr_read)) {
292 				rv = -EFAULT;
293 				goto out_free_page;
294 			}
295 
296 			p	+= nr_read;
297 			len	-= nr_read;
298 			buf	+= nr_read;
299 			count	-= nr_read;
300 			rv	+= nr_read;
301 		}
302 	} else {
303 		/*
304 		 * Command line (1 string) occupies ARGV and
305 		 * extends into ENVP.
306 		 */
307 		struct {
308 			unsigned long p;
309 			unsigned long len;
310 		} cmdline[2] = {
311 			{ .p = arg_start, .len = len1 },
312 			{ .p = env_start, .len = len2 },
313 		};
314 		loff_t pos1 = *pos;
315 		unsigned int i;
316 
317 		i = 0;
318 		while (i < 2 && pos1 >= cmdline[i].len) {
319 			pos1 -= cmdline[i].len;
320 			i++;
321 		}
322 		while (i < 2) {
323 			p = cmdline[i].p + pos1;
324 			len = cmdline[i].len - pos1;
325 			while (count > 0 && len > 0) {
326 				unsigned int _count, l;
327 				int nr_read;
328 				bool final;
329 
330 				_count = min3(count, len, PAGE_SIZE);
331 				nr_read = access_remote_vm(mm, p, page, _count, 0);
332 				if (nr_read < 0)
333 					rv = nr_read;
334 				if (nr_read <= 0)
335 					goto out_free_page;
336 
337 				/*
338 				 * Command line can be shorter than whole ARGV
339 				 * even if last "marker" byte says it is not.
340 				 */
341 				final = false;
342 				l = strnlen(page, nr_read);
343 				if (l < nr_read) {
344 					nr_read = l;
345 					final = true;
346 				}
347 
348 				if (copy_to_user(buf, page, nr_read)) {
349 					rv = -EFAULT;
350 					goto out_free_page;
351 				}
352 
353 				p	+= nr_read;
354 				len	-= nr_read;
355 				buf	+= nr_read;
356 				count	-= nr_read;
357 				rv	+= nr_read;
358 
359 				if (final)
360 					goto out_free_page;
361 			}
362 
363 			/* Only first chunk can be read partially. */
364 			pos1 = 0;
365 			i++;
366 		}
367 	}
368 
369 out_free_page:
370 	free_page((unsigned long)page);
371 out_mmput:
372 	mmput(mm);
373 	if (rv > 0)
374 		*pos += rv;
375 	return rv;
376 }
377 
378 static const struct file_operations proc_pid_cmdline_ops = {
379 	.read	= proc_pid_cmdline_read,
380 	.llseek	= generic_file_llseek,
381 };
382 
383 #ifdef CONFIG_KALLSYMS
384 /*
385  * Provides a wchan file via kallsyms in a proper one-value-per-file format.
386  * Returns the resolved symbol.  If that fails, simply return the address.
387  */
388 static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
389 			  struct pid *pid, struct task_struct *task)
390 {
391 	unsigned long wchan;
392 	char symname[KSYM_NAME_LEN];
393 
394 	wchan = get_wchan(task);
395 
396 	if (wchan && ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)
397 			&& !lookup_symbol_name(wchan, symname))
398 		seq_printf(m, "%s", symname);
399 	else
400 		seq_putc(m, '0');
401 
402 	return 0;
403 }
404 #endif /* CONFIG_KALLSYMS */
405 
406 static int lock_trace(struct task_struct *task)
407 {
408 	int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
409 	if (err)
410 		return err;
411 	if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
412 		mutex_unlock(&task->signal->cred_guard_mutex);
413 		return -EPERM;
414 	}
415 	return 0;
416 }
417 
418 static void unlock_trace(struct task_struct *task)
419 {
420 	mutex_unlock(&task->signal->cred_guard_mutex);
421 }
422 
423 #ifdef CONFIG_STACKTRACE
424 
425 #define MAX_STACK_TRACE_DEPTH	64
426 
427 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
428 			  struct pid *pid, struct task_struct *task)
429 {
430 	struct stack_trace trace;
431 	unsigned long *entries;
432 	int err;
433 	int i;
434 
435 	entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
436 	if (!entries)
437 		return -ENOMEM;
438 
439 	trace.nr_entries	= 0;
440 	trace.max_entries	= MAX_STACK_TRACE_DEPTH;
441 	trace.entries		= entries;
442 	trace.skip		= 0;
443 
444 	err = lock_trace(task);
445 	if (!err) {
446 		save_stack_trace_tsk(task, &trace);
447 
448 		for (i = 0; i < trace.nr_entries; i++) {
449 			seq_printf(m, "[<0>] %pB\n", (void *)entries[i]);
450 		}
451 		unlock_trace(task);
452 	}
453 	kfree(entries);
454 
455 	return err;
456 }
457 #endif
458 
459 #ifdef CONFIG_SCHED_INFO
460 /*
461  * Provides /proc/PID/schedstat
462  */
463 static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
464 			      struct pid *pid, struct task_struct *task)
465 {
466 	if (unlikely(!sched_info_on()))
467 		seq_printf(m, "0 0 0\n");
468 	else
469 		seq_printf(m, "%llu %llu %lu\n",
470 		   (unsigned long long)task->se.sum_exec_runtime,
471 		   (unsigned long long)task->sched_info.run_delay,
472 		   task->sched_info.pcount);
473 
474 	return 0;
475 }
476 #endif
477 
478 #ifdef CONFIG_LATENCYTOP
479 static int lstats_show_proc(struct seq_file *m, void *v)
480 {
481 	int i;
482 	struct inode *inode = m->private;
483 	struct task_struct *task = get_proc_task(inode);
484 
485 	if (!task)
486 		return -ESRCH;
487 	seq_puts(m, "Latency Top version : v0.1\n");
488 	for (i = 0; i < 32; i++) {
489 		struct latency_record *lr = &task->latency_record[i];
490 		if (lr->backtrace[0]) {
491 			int q;
492 			seq_printf(m, "%i %li %li",
493 				   lr->count, lr->time, lr->max);
494 			for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
495 				unsigned long bt = lr->backtrace[q];
496 				if (!bt)
497 					break;
498 				if (bt == ULONG_MAX)
499 					break;
500 				seq_printf(m, " %ps", (void *)bt);
501 			}
502 			seq_putc(m, '\n');
503 		}
504 
505 	}
506 	put_task_struct(task);
507 	return 0;
508 }
509 
510 static int lstats_open(struct inode *inode, struct file *file)
511 {
512 	return single_open(file, lstats_show_proc, inode);
513 }
514 
515 static ssize_t lstats_write(struct file *file, const char __user *buf,
516 			    size_t count, loff_t *offs)
517 {
518 	struct task_struct *task = get_proc_task(file_inode(file));
519 
520 	if (!task)
521 		return -ESRCH;
522 	clear_all_latency_tracing(task);
523 	put_task_struct(task);
524 
525 	return count;
526 }
527 
528 static const struct file_operations proc_lstats_operations = {
529 	.open		= lstats_open,
530 	.read		= seq_read,
531 	.write		= lstats_write,
532 	.llseek		= seq_lseek,
533 	.release	= single_release,
534 };
535 
536 #endif
537 
538 static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
539 			  struct pid *pid, struct task_struct *task)
540 {
541 	unsigned long totalpages = totalram_pages + total_swap_pages;
542 	unsigned long points = 0;
543 
544 	points = oom_badness(task, NULL, NULL, totalpages) *
545 					1000 / totalpages;
546 	seq_printf(m, "%lu\n", points);
547 
548 	return 0;
549 }
550 
551 struct limit_names {
552 	const char *name;
553 	const char *unit;
554 };
555 
556 static const struct limit_names lnames[RLIM_NLIMITS] = {
557 	[RLIMIT_CPU] = {"Max cpu time", "seconds"},
558 	[RLIMIT_FSIZE] = {"Max file size", "bytes"},
559 	[RLIMIT_DATA] = {"Max data size", "bytes"},
560 	[RLIMIT_STACK] = {"Max stack size", "bytes"},
561 	[RLIMIT_CORE] = {"Max core file size", "bytes"},
562 	[RLIMIT_RSS] = {"Max resident set", "bytes"},
563 	[RLIMIT_NPROC] = {"Max processes", "processes"},
564 	[RLIMIT_NOFILE] = {"Max open files", "files"},
565 	[RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
566 	[RLIMIT_AS] = {"Max address space", "bytes"},
567 	[RLIMIT_LOCKS] = {"Max file locks", "locks"},
568 	[RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
569 	[RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
570 	[RLIMIT_NICE] = {"Max nice priority", NULL},
571 	[RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
572 	[RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
573 };
574 
575 /* Display limits for a process */
576 static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
577 			   struct pid *pid, struct task_struct *task)
578 {
579 	unsigned int i;
580 	unsigned long flags;
581 
582 	struct rlimit rlim[RLIM_NLIMITS];
583 
584 	if (!lock_task_sighand(task, &flags))
585 		return 0;
586 	memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
587 	unlock_task_sighand(task, &flags);
588 
589 	/*
590 	 * print the file header
591 	 */
592        seq_printf(m, "%-25s %-20s %-20s %-10s\n",
593 		  "Limit", "Soft Limit", "Hard Limit", "Units");
594 
595 	for (i = 0; i < RLIM_NLIMITS; i++) {
596 		if (rlim[i].rlim_cur == RLIM_INFINITY)
597 			seq_printf(m, "%-25s %-20s ",
598 				   lnames[i].name, "unlimited");
599 		else
600 			seq_printf(m, "%-25s %-20lu ",
601 				   lnames[i].name, rlim[i].rlim_cur);
602 
603 		if (rlim[i].rlim_max == RLIM_INFINITY)
604 			seq_printf(m, "%-20s ", "unlimited");
605 		else
606 			seq_printf(m, "%-20lu ", rlim[i].rlim_max);
607 
608 		if (lnames[i].unit)
609 			seq_printf(m, "%-10s\n", lnames[i].unit);
610 		else
611 			seq_putc(m, '\n');
612 	}
613 
614 	return 0;
615 }
616 
617 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
618 static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
619 			    struct pid *pid, struct task_struct *task)
620 {
621 	long nr;
622 	unsigned long args[6], sp, pc;
623 	int res;
624 
625 	res = lock_trace(task);
626 	if (res)
627 		return res;
628 
629 	if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
630 		seq_puts(m, "running\n");
631 	else if (nr < 0)
632 		seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
633 	else
634 		seq_printf(m,
635 		       "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
636 		       nr,
637 		       args[0], args[1], args[2], args[3], args[4], args[5],
638 		       sp, pc);
639 	unlock_trace(task);
640 
641 	return 0;
642 }
643 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
644 
645 /************************************************************************/
646 /*                       Here the fs part begins                        */
647 /************************************************************************/
648 
649 /* permission checks */
650 static int proc_fd_access_allowed(struct inode *inode)
651 {
652 	struct task_struct *task;
653 	int allowed = 0;
654 	/* Allow access to a task's file descriptors if it is us or we
655 	 * may use ptrace attach to the process and find out that
656 	 * information.
657 	 */
658 	task = get_proc_task(inode);
659 	if (task) {
660 		allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
661 		put_task_struct(task);
662 	}
663 	return allowed;
664 }
665 
666 int proc_setattr(struct dentry *dentry, struct iattr *attr)
667 {
668 	int error;
669 	struct inode *inode = d_inode(dentry);
670 
671 	if (attr->ia_valid & ATTR_MODE)
672 		return -EPERM;
673 
674 	error = setattr_prepare(dentry, attr);
675 	if (error)
676 		return error;
677 
678 	setattr_copy(inode, attr);
679 	mark_inode_dirty(inode);
680 	return 0;
681 }
682 
683 /*
684  * May current process learn task's sched/cmdline info (for hide_pid_min=1)
685  * or euid/egid (for hide_pid_min=2)?
686  */
687 static bool has_pid_permissions(struct pid_namespace *pid,
688 				 struct task_struct *task,
689 				 int hide_pid_min)
690 {
691 	if (pid->hide_pid < hide_pid_min)
692 		return true;
693 	if (in_group_p(pid->pid_gid))
694 		return true;
695 	return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
696 }
697 
698 
699 static int proc_pid_permission(struct inode *inode, int mask)
700 {
701 	struct pid_namespace *pid = inode->i_sb->s_fs_info;
702 	struct task_struct *task;
703 	bool has_perms;
704 
705 	task = get_proc_task(inode);
706 	if (!task)
707 		return -ESRCH;
708 	has_perms = has_pid_permissions(pid, task, HIDEPID_NO_ACCESS);
709 	put_task_struct(task);
710 
711 	if (!has_perms) {
712 		if (pid->hide_pid == HIDEPID_INVISIBLE) {
713 			/*
714 			 * Let's make getdents(), stat(), and open()
715 			 * consistent with each other.  If a process
716 			 * may not stat() a file, it shouldn't be seen
717 			 * in procfs at all.
718 			 */
719 			return -ENOENT;
720 		}
721 
722 		return -EPERM;
723 	}
724 	return generic_permission(inode, mask);
725 }
726 
727 
728 
729 static const struct inode_operations proc_def_inode_operations = {
730 	.setattr	= proc_setattr,
731 };
732 
733 static int proc_single_show(struct seq_file *m, void *v)
734 {
735 	struct inode *inode = m->private;
736 	struct pid_namespace *ns;
737 	struct pid *pid;
738 	struct task_struct *task;
739 	int ret;
740 
741 	ns = inode->i_sb->s_fs_info;
742 	pid = proc_pid(inode);
743 	task = get_pid_task(pid, PIDTYPE_PID);
744 	if (!task)
745 		return -ESRCH;
746 
747 	ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
748 
749 	put_task_struct(task);
750 	return ret;
751 }
752 
753 static int proc_single_open(struct inode *inode, struct file *filp)
754 {
755 	return single_open(filp, proc_single_show, inode);
756 }
757 
758 static const struct file_operations proc_single_file_operations = {
759 	.open		= proc_single_open,
760 	.read		= seq_read,
761 	.llseek		= seq_lseek,
762 	.release	= single_release,
763 };
764 
765 
766 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
767 {
768 	struct task_struct *task = get_proc_task(inode);
769 	struct mm_struct *mm = ERR_PTR(-ESRCH);
770 
771 	if (task) {
772 		mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
773 		put_task_struct(task);
774 
775 		if (!IS_ERR_OR_NULL(mm)) {
776 			/* ensure this mm_struct can't be freed */
777 			mmgrab(mm);
778 			/* but do not pin its memory */
779 			mmput(mm);
780 		}
781 	}
782 
783 	return mm;
784 }
785 
786 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
787 {
788 	struct mm_struct *mm = proc_mem_open(inode, mode);
789 
790 	if (IS_ERR(mm))
791 		return PTR_ERR(mm);
792 
793 	file->private_data = mm;
794 	return 0;
795 }
796 
797 static int mem_open(struct inode *inode, struct file *file)
798 {
799 	int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
800 
801 	/* OK to pass negative loff_t, we can catch out-of-range */
802 	file->f_mode |= FMODE_UNSIGNED_OFFSET;
803 
804 	return ret;
805 }
806 
807 static ssize_t mem_rw(struct file *file, char __user *buf,
808 			size_t count, loff_t *ppos, int write)
809 {
810 	struct mm_struct *mm = file->private_data;
811 	unsigned long addr = *ppos;
812 	ssize_t copied;
813 	char *page;
814 	unsigned int flags;
815 
816 	if (!mm)
817 		return 0;
818 
819 	page = (char *)__get_free_page(GFP_KERNEL);
820 	if (!page)
821 		return -ENOMEM;
822 
823 	copied = 0;
824 	if (!mmget_not_zero(mm))
825 		goto free;
826 
827 	flags = FOLL_FORCE | (write ? FOLL_WRITE : 0);
828 
829 	while (count > 0) {
830 		int this_len = min_t(int, count, PAGE_SIZE);
831 
832 		if (write && copy_from_user(page, buf, this_len)) {
833 			copied = -EFAULT;
834 			break;
835 		}
836 
837 		this_len = access_remote_vm(mm, addr, page, this_len, flags);
838 		if (!this_len) {
839 			if (!copied)
840 				copied = -EIO;
841 			break;
842 		}
843 
844 		if (!write && copy_to_user(buf, page, this_len)) {
845 			copied = -EFAULT;
846 			break;
847 		}
848 
849 		buf += this_len;
850 		addr += this_len;
851 		copied += this_len;
852 		count -= this_len;
853 	}
854 	*ppos = addr;
855 
856 	mmput(mm);
857 free:
858 	free_page((unsigned long) page);
859 	return copied;
860 }
861 
862 static ssize_t mem_read(struct file *file, char __user *buf,
863 			size_t count, loff_t *ppos)
864 {
865 	return mem_rw(file, buf, count, ppos, 0);
866 }
867 
868 static ssize_t mem_write(struct file *file, const char __user *buf,
869 			 size_t count, loff_t *ppos)
870 {
871 	return mem_rw(file, (char __user*)buf, count, ppos, 1);
872 }
873 
874 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
875 {
876 	switch (orig) {
877 	case 0:
878 		file->f_pos = offset;
879 		break;
880 	case 1:
881 		file->f_pos += offset;
882 		break;
883 	default:
884 		return -EINVAL;
885 	}
886 	force_successful_syscall_return();
887 	return file->f_pos;
888 }
889 
890 static int mem_release(struct inode *inode, struct file *file)
891 {
892 	struct mm_struct *mm = file->private_data;
893 	if (mm)
894 		mmdrop(mm);
895 	return 0;
896 }
897 
898 static const struct file_operations proc_mem_operations = {
899 	.llseek		= mem_lseek,
900 	.read		= mem_read,
901 	.write		= mem_write,
902 	.open		= mem_open,
903 	.release	= mem_release,
904 };
905 
906 static int environ_open(struct inode *inode, struct file *file)
907 {
908 	return __mem_open(inode, file, PTRACE_MODE_READ);
909 }
910 
911 static ssize_t environ_read(struct file *file, char __user *buf,
912 			size_t count, loff_t *ppos)
913 {
914 	char *page;
915 	unsigned long src = *ppos;
916 	int ret = 0;
917 	struct mm_struct *mm = file->private_data;
918 	unsigned long env_start, env_end;
919 
920 	/* Ensure the process spawned far enough to have an environment. */
921 	if (!mm || !mm->env_end)
922 		return 0;
923 
924 	page = (char *)__get_free_page(GFP_KERNEL);
925 	if (!page)
926 		return -ENOMEM;
927 
928 	ret = 0;
929 	if (!mmget_not_zero(mm))
930 		goto free;
931 
932 	down_read(&mm->mmap_sem);
933 	env_start = mm->env_start;
934 	env_end = mm->env_end;
935 	up_read(&mm->mmap_sem);
936 
937 	while (count > 0) {
938 		size_t this_len, max_len;
939 		int retval;
940 
941 		if (src >= (env_end - env_start))
942 			break;
943 
944 		this_len = env_end - (env_start + src);
945 
946 		max_len = min_t(size_t, PAGE_SIZE, count);
947 		this_len = min(max_len, this_len);
948 
949 		retval = access_remote_vm(mm, (env_start + src), page, this_len, 0);
950 
951 		if (retval <= 0) {
952 			ret = retval;
953 			break;
954 		}
955 
956 		if (copy_to_user(buf, page, retval)) {
957 			ret = -EFAULT;
958 			break;
959 		}
960 
961 		ret += retval;
962 		src += retval;
963 		buf += retval;
964 		count -= retval;
965 	}
966 	*ppos = src;
967 	mmput(mm);
968 
969 free:
970 	free_page((unsigned long) page);
971 	return ret;
972 }
973 
974 static const struct file_operations proc_environ_operations = {
975 	.open		= environ_open,
976 	.read		= environ_read,
977 	.llseek		= generic_file_llseek,
978 	.release	= mem_release,
979 };
980 
981 static int auxv_open(struct inode *inode, struct file *file)
982 {
983 	return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
984 }
985 
986 static ssize_t auxv_read(struct file *file, char __user *buf,
987 			size_t count, loff_t *ppos)
988 {
989 	struct mm_struct *mm = file->private_data;
990 	unsigned int nwords = 0;
991 
992 	if (!mm)
993 		return 0;
994 	do {
995 		nwords += 2;
996 	} while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
997 	return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
998 				       nwords * sizeof(mm->saved_auxv[0]));
999 }
1000 
1001 static const struct file_operations proc_auxv_operations = {
1002 	.open		= auxv_open,
1003 	.read		= auxv_read,
1004 	.llseek		= generic_file_llseek,
1005 	.release	= mem_release,
1006 };
1007 
1008 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1009 			    loff_t *ppos)
1010 {
1011 	struct task_struct *task = get_proc_task(file_inode(file));
1012 	char buffer[PROC_NUMBUF];
1013 	int oom_adj = OOM_ADJUST_MIN;
1014 	size_t len;
1015 
1016 	if (!task)
1017 		return -ESRCH;
1018 	if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1019 		oom_adj = OOM_ADJUST_MAX;
1020 	else
1021 		oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1022 			  OOM_SCORE_ADJ_MAX;
1023 	put_task_struct(task);
1024 	len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1025 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1026 }
1027 
1028 static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1029 {
1030 	static DEFINE_MUTEX(oom_adj_mutex);
1031 	struct mm_struct *mm = NULL;
1032 	struct task_struct *task;
1033 	int err = 0;
1034 
1035 	task = get_proc_task(file_inode(file));
1036 	if (!task)
1037 		return -ESRCH;
1038 
1039 	mutex_lock(&oom_adj_mutex);
1040 	if (legacy) {
1041 		if (oom_adj < task->signal->oom_score_adj &&
1042 				!capable(CAP_SYS_RESOURCE)) {
1043 			err = -EACCES;
1044 			goto err_unlock;
1045 		}
1046 		/*
1047 		 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1048 		 * /proc/pid/oom_score_adj instead.
1049 		 */
1050 		pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1051 			  current->comm, task_pid_nr(current), task_pid_nr(task),
1052 			  task_pid_nr(task));
1053 	} else {
1054 		if ((short)oom_adj < task->signal->oom_score_adj_min &&
1055 				!capable(CAP_SYS_RESOURCE)) {
1056 			err = -EACCES;
1057 			goto err_unlock;
1058 		}
1059 	}
1060 
1061 	/*
1062 	 * Make sure we will check other processes sharing the mm if this is
1063 	 * not vfrok which wants its own oom_score_adj.
1064 	 * pin the mm so it doesn't go away and get reused after task_unlock
1065 	 */
1066 	if (!task->vfork_done) {
1067 		struct task_struct *p = find_lock_task_mm(task);
1068 
1069 		if (p) {
1070 			if (atomic_read(&p->mm->mm_users) > 1) {
1071 				mm = p->mm;
1072 				mmgrab(mm);
1073 			}
1074 			task_unlock(p);
1075 		}
1076 	}
1077 
1078 	task->signal->oom_score_adj = oom_adj;
1079 	if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1080 		task->signal->oom_score_adj_min = (short)oom_adj;
1081 	trace_oom_score_adj_update(task);
1082 
1083 	if (mm) {
1084 		struct task_struct *p;
1085 
1086 		rcu_read_lock();
1087 		for_each_process(p) {
1088 			if (same_thread_group(task, p))
1089 				continue;
1090 
1091 			/* do not touch kernel threads or the global init */
1092 			if (p->flags & PF_KTHREAD || is_global_init(p))
1093 				continue;
1094 
1095 			task_lock(p);
1096 			if (!p->vfork_done && process_shares_mm(p, mm)) {
1097 				pr_info("updating oom_score_adj for %d (%s) from %d to %d because it shares mm with %d (%s). Report if this is unexpected.\n",
1098 						task_pid_nr(p), p->comm,
1099 						p->signal->oom_score_adj, oom_adj,
1100 						task_pid_nr(task), task->comm);
1101 				p->signal->oom_score_adj = oom_adj;
1102 				if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1103 					p->signal->oom_score_adj_min = (short)oom_adj;
1104 			}
1105 			task_unlock(p);
1106 		}
1107 		rcu_read_unlock();
1108 		mmdrop(mm);
1109 	}
1110 err_unlock:
1111 	mutex_unlock(&oom_adj_mutex);
1112 	put_task_struct(task);
1113 	return err;
1114 }
1115 
1116 /*
1117  * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1118  * kernels.  The effective policy is defined by oom_score_adj, which has a
1119  * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1120  * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1121  * Processes that become oom disabled via oom_adj will still be oom disabled
1122  * with this implementation.
1123  *
1124  * oom_adj cannot be removed since existing userspace binaries use it.
1125  */
1126 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1127 			     size_t count, loff_t *ppos)
1128 {
1129 	char buffer[PROC_NUMBUF];
1130 	int oom_adj;
1131 	int err;
1132 
1133 	memset(buffer, 0, sizeof(buffer));
1134 	if (count > sizeof(buffer) - 1)
1135 		count = sizeof(buffer) - 1;
1136 	if (copy_from_user(buffer, buf, count)) {
1137 		err = -EFAULT;
1138 		goto out;
1139 	}
1140 
1141 	err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1142 	if (err)
1143 		goto out;
1144 	if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1145 	     oom_adj != OOM_DISABLE) {
1146 		err = -EINVAL;
1147 		goto out;
1148 	}
1149 
1150 	/*
1151 	 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1152 	 * value is always attainable.
1153 	 */
1154 	if (oom_adj == OOM_ADJUST_MAX)
1155 		oom_adj = OOM_SCORE_ADJ_MAX;
1156 	else
1157 		oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1158 
1159 	err = __set_oom_adj(file, oom_adj, true);
1160 out:
1161 	return err < 0 ? err : count;
1162 }
1163 
1164 static const struct file_operations proc_oom_adj_operations = {
1165 	.read		= oom_adj_read,
1166 	.write		= oom_adj_write,
1167 	.llseek		= generic_file_llseek,
1168 };
1169 
1170 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1171 					size_t count, loff_t *ppos)
1172 {
1173 	struct task_struct *task = get_proc_task(file_inode(file));
1174 	char buffer[PROC_NUMBUF];
1175 	short oom_score_adj = OOM_SCORE_ADJ_MIN;
1176 	size_t len;
1177 
1178 	if (!task)
1179 		return -ESRCH;
1180 	oom_score_adj = task->signal->oom_score_adj;
1181 	put_task_struct(task);
1182 	len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1183 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1184 }
1185 
1186 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1187 					size_t count, loff_t *ppos)
1188 {
1189 	char buffer[PROC_NUMBUF];
1190 	int oom_score_adj;
1191 	int err;
1192 
1193 	memset(buffer, 0, sizeof(buffer));
1194 	if (count > sizeof(buffer) - 1)
1195 		count = sizeof(buffer) - 1;
1196 	if (copy_from_user(buffer, buf, count)) {
1197 		err = -EFAULT;
1198 		goto out;
1199 	}
1200 
1201 	err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1202 	if (err)
1203 		goto out;
1204 	if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1205 			oom_score_adj > OOM_SCORE_ADJ_MAX) {
1206 		err = -EINVAL;
1207 		goto out;
1208 	}
1209 
1210 	err = __set_oom_adj(file, oom_score_adj, false);
1211 out:
1212 	return err < 0 ? err : count;
1213 }
1214 
1215 static const struct file_operations proc_oom_score_adj_operations = {
1216 	.read		= oom_score_adj_read,
1217 	.write		= oom_score_adj_write,
1218 	.llseek		= default_llseek,
1219 };
1220 
1221 #ifdef CONFIG_AUDITSYSCALL
1222 #define TMPBUFLEN 11
1223 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1224 				  size_t count, loff_t *ppos)
1225 {
1226 	struct inode * inode = file_inode(file);
1227 	struct task_struct *task = get_proc_task(inode);
1228 	ssize_t length;
1229 	char tmpbuf[TMPBUFLEN];
1230 
1231 	if (!task)
1232 		return -ESRCH;
1233 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1234 			   from_kuid(file->f_cred->user_ns,
1235 				     audit_get_loginuid(task)));
1236 	put_task_struct(task);
1237 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1238 }
1239 
1240 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1241 				   size_t count, loff_t *ppos)
1242 {
1243 	struct inode * inode = file_inode(file);
1244 	uid_t loginuid;
1245 	kuid_t kloginuid;
1246 	int rv;
1247 
1248 	rcu_read_lock();
1249 	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1250 		rcu_read_unlock();
1251 		return -EPERM;
1252 	}
1253 	rcu_read_unlock();
1254 
1255 	if (*ppos != 0) {
1256 		/* No partial writes. */
1257 		return -EINVAL;
1258 	}
1259 
1260 	rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1261 	if (rv < 0)
1262 		return rv;
1263 
1264 	/* is userspace tring to explicitly UNSET the loginuid? */
1265 	if (loginuid == AUDIT_UID_UNSET) {
1266 		kloginuid = INVALID_UID;
1267 	} else {
1268 		kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1269 		if (!uid_valid(kloginuid))
1270 			return -EINVAL;
1271 	}
1272 
1273 	rv = audit_set_loginuid(kloginuid);
1274 	if (rv < 0)
1275 		return rv;
1276 	return count;
1277 }
1278 
1279 static const struct file_operations proc_loginuid_operations = {
1280 	.read		= proc_loginuid_read,
1281 	.write		= proc_loginuid_write,
1282 	.llseek		= generic_file_llseek,
1283 };
1284 
1285 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1286 				  size_t count, loff_t *ppos)
1287 {
1288 	struct inode * inode = file_inode(file);
1289 	struct task_struct *task = get_proc_task(inode);
1290 	ssize_t length;
1291 	char tmpbuf[TMPBUFLEN];
1292 
1293 	if (!task)
1294 		return -ESRCH;
1295 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1296 				audit_get_sessionid(task));
1297 	put_task_struct(task);
1298 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1299 }
1300 
1301 static const struct file_operations proc_sessionid_operations = {
1302 	.read		= proc_sessionid_read,
1303 	.llseek		= generic_file_llseek,
1304 };
1305 #endif
1306 
1307 #ifdef CONFIG_FAULT_INJECTION
1308 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1309 				      size_t count, loff_t *ppos)
1310 {
1311 	struct task_struct *task = get_proc_task(file_inode(file));
1312 	char buffer[PROC_NUMBUF];
1313 	size_t len;
1314 	int make_it_fail;
1315 
1316 	if (!task)
1317 		return -ESRCH;
1318 	make_it_fail = task->make_it_fail;
1319 	put_task_struct(task);
1320 
1321 	len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1322 
1323 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1324 }
1325 
1326 static ssize_t proc_fault_inject_write(struct file * file,
1327 			const char __user * buf, size_t count, loff_t *ppos)
1328 {
1329 	struct task_struct *task;
1330 	char buffer[PROC_NUMBUF];
1331 	int make_it_fail;
1332 	int rv;
1333 
1334 	if (!capable(CAP_SYS_RESOURCE))
1335 		return -EPERM;
1336 	memset(buffer, 0, sizeof(buffer));
1337 	if (count > sizeof(buffer) - 1)
1338 		count = sizeof(buffer) - 1;
1339 	if (copy_from_user(buffer, buf, count))
1340 		return -EFAULT;
1341 	rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1342 	if (rv < 0)
1343 		return rv;
1344 	if (make_it_fail < 0 || make_it_fail > 1)
1345 		return -EINVAL;
1346 
1347 	task = get_proc_task(file_inode(file));
1348 	if (!task)
1349 		return -ESRCH;
1350 	task->make_it_fail = make_it_fail;
1351 	put_task_struct(task);
1352 
1353 	return count;
1354 }
1355 
1356 static const struct file_operations proc_fault_inject_operations = {
1357 	.read		= proc_fault_inject_read,
1358 	.write		= proc_fault_inject_write,
1359 	.llseek		= generic_file_llseek,
1360 };
1361 
1362 static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1363 				   size_t count, loff_t *ppos)
1364 {
1365 	struct task_struct *task;
1366 	int err;
1367 	unsigned int n;
1368 
1369 	err = kstrtouint_from_user(buf, count, 0, &n);
1370 	if (err)
1371 		return err;
1372 
1373 	task = get_proc_task(file_inode(file));
1374 	if (!task)
1375 		return -ESRCH;
1376 	task->fail_nth = n;
1377 	put_task_struct(task);
1378 
1379 	return count;
1380 }
1381 
1382 static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1383 				  size_t count, loff_t *ppos)
1384 {
1385 	struct task_struct *task;
1386 	char numbuf[PROC_NUMBUF];
1387 	ssize_t len;
1388 
1389 	task = get_proc_task(file_inode(file));
1390 	if (!task)
1391 		return -ESRCH;
1392 	len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth);
1393 	len = simple_read_from_buffer(buf, count, ppos, numbuf, len);
1394 	put_task_struct(task);
1395 
1396 	return len;
1397 }
1398 
1399 static const struct file_operations proc_fail_nth_operations = {
1400 	.read		= proc_fail_nth_read,
1401 	.write		= proc_fail_nth_write,
1402 };
1403 #endif
1404 
1405 
1406 #ifdef CONFIG_SCHED_DEBUG
1407 /*
1408  * Print out various scheduling related per-task fields:
1409  */
1410 static int sched_show(struct seq_file *m, void *v)
1411 {
1412 	struct inode *inode = m->private;
1413 	struct pid_namespace *ns = inode->i_sb->s_fs_info;
1414 	struct task_struct *p;
1415 
1416 	p = get_proc_task(inode);
1417 	if (!p)
1418 		return -ESRCH;
1419 	proc_sched_show_task(p, ns, m);
1420 
1421 	put_task_struct(p);
1422 
1423 	return 0;
1424 }
1425 
1426 static ssize_t
1427 sched_write(struct file *file, const char __user *buf,
1428 	    size_t count, loff_t *offset)
1429 {
1430 	struct inode *inode = file_inode(file);
1431 	struct task_struct *p;
1432 
1433 	p = get_proc_task(inode);
1434 	if (!p)
1435 		return -ESRCH;
1436 	proc_sched_set_task(p);
1437 
1438 	put_task_struct(p);
1439 
1440 	return count;
1441 }
1442 
1443 static int sched_open(struct inode *inode, struct file *filp)
1444 {
1445 	return single_open(filp, sched_show, inode);
1446 }
1447 
1448 static const struct file_operations proc_pid_sched_operations = {
1449 	.open		= sched_open,
1450 	.read		= seq_read,
1451 	.write		= sched_write,
1452 	.llseek		= seq_lseek,
1453 	.release	= single_release,
1454 };
1455 
1456 #endif
1457 
1458 #ifdef CONFIG_SCHED_AUTOGROUP
1459 /*
1460  * Print out autogroup related information:
1461  */
1462 static int sched_autogroup_show(struct seq_file *m, void *v)
1463 {
1464 	struct inode *inode = m->private;
1465 	struct task_struct *p;
1466 
1467 	p = get_proc_task(inode);
1468 	if (!p)
1469 		return -ESRCH;
1470 	proc_sched_autogroup_show_task(p, m);
1471 
1472 	put_task_struct(p);
1473 
1474 	return 0;
1475 }
1476 
1477 static ssize_t
1478 sched_autogroup_write(struct file *file, const char __user *buf,
1479 	    size_t count, loff_t *offset)
1480 {
1481 	struct inode *inode = file_inode(file);
1482 	struct task_struct *p;
1483 	char buffer[PROC_NUMBUF];
1484 	int nice;
1485 	int err;
1486 
1487 	memset(buffer, 0, sizeof(buffer));
1488 	if (count > sizeof(buffer) - 1)
1489 		count = sizeof(buffer) - 1;
1490 	if (copy_from_user(buffer, buf, count))
1491 		return -EFAULT;
1492 
1493 	err = kstrtoint(strstrip(buffer), 0, &nice);
1494 	if (err < 0)
1495 		return err;
1496 
1497 	p = get_proc_task(inode);
1498 	if (!p)
1499 		return -ESRCH;
1500 
1501 	err = proc_sched_autogroup_set_nice(p, nice);
1502 	if (err)
1503 		count = err;
1504 
1505 	put_task_struct(p);
1506 
1507 	return count;
1508 }
1509 
1510 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1511 {
1512 	int ret;
1513 
1514 	ret = single_open(filp, sched_autogroup_show, NULL);
1515 	if (!ret) {
1516 		struct seq_file *m = filp->private_data;
1517 
1518 		m->private = inode;
1519 	}
1520 	return ret;
1521 }
1522 
1523 static const struct file_operations proc_pid_sched_autogroup_operations = {
1524 	.open		= sched_autogroup_open,
1525 	.read		= seq_read,
1526 	.write		= sched_autogroup_write,
1527 	.llseek		= seq_lseek,
1528 	.release	= single_release,
1529 };
1530 
1531 #endif /* CONFIG_SCHED_AUTOGROUP */
1532 
1533 static ssize_t comm_write(struct file *file, const char __user *buf,
1534 				size_t count, loff_t *offset)
1535 {
1536 	struct inode *inode = file_inode(file);
1537 	struct task_struct *p;
1538 	char buffer[TASK_COMM_LEN];
1539 	const size_t maxlen = sizeof(buffer) - 1;
1540 
1541 	memset(buffer, 0, sizeof(buffer));
1542 	if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1543 		return -EFAULT;
1544 
1545 	p = get_proc_task(inode);
1546 	if (!p)
1547 		return -ESRCH;
1548 
1549 	if (same_thread_group(current, p))
1550 		set_task_comm(p, buffer);
1551 	else
1552 		count = -EINVAL;
1553 
1554 	put_task_struct(p);
1555 
1556 	return count;
1557 }
1558 
1559 static int comm_show(struct seq_file *m, void *v)
1560 {
1561 	struct inode *inode = m->private;
1562 	struct task_struct *p;
1563 
1564 	p = get_proc_task(inode);
1565 	if (!p)
1566 		return -ESRCH;
1567 
1568 	task_lock(p);
1569 	seq_printf(m, "%s\n", p->comm);
1570 	task_unlock(p);
1571 
1572 	put_task_struct(p);
1573 
1574 	return 0;
1575 }
1576 
1577 static int comm_open(struct inode *inode, struct file *filp)
1578 {
1579 	return single_open(filp, comm_show, inode);
1580 }
1581 
1582 static const struct file_operations proc_pid_set_comm_operations = {
1583 	.open		= comm_open,
1584 	.read		= seq_read,
1585 	.write		= comm_write,
1586 	.llseek		= seq_lseek,
1587 	.release	= single_release,
1588 };
1589 
1590 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1591 {
1592 	struct task_struct *task;
1593 	struct file *exe_file;
1594 
1595 	task = get_proc_task(d_inode(dentry));
1596 	if (!task)
1597 		return -ENOENT;
1598 	exe_file = get_task_exe_file(task);
1599 	put_task_struct(task);
1600 	if (exe_file) {
1601 		*exe_path = exe_file->f_path;
1602 		path_get(&exe_file->f_path);
1603 		fput(exe_file);
1604 		return 0;
1605 	} else
1606 		return -ENOENT;
1607 }
1608 
1609 static const char *proc_pid_get_link(struct dentry *dentry,
1610 				     struct inode *inode,
1611 				     struct delayed_call *done)
1612 {
1613 	struct path path;
1614 	int error = -EACCES;
1615 
1616 	if (!dentry)
1617 		return ERR_PTR(-ECHILD);
1618 
1619 	/* Are we allowed to snoop on the tasks file descriptors? */
1620 	if (!proc_fd_access_allowed(inode))
1621 		goto out;
1622 
1623 	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1624 	if (error)
1625 		goto out;
1626 
1627 	nd_jump_link(&path);
1628 	return NULL;
1629 out:
1630 	return ERR_PTR(error);
1631 }
1632 
1633 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1634 {
1635 	char *tmp = (char *)__get_free_page(GFP_KERNEL);
1636 	char *pathname;
1637 	int len;
1638 
1639 	if (!tmp)
1640 		return -ENOMEM;
1641 
1642 	pathname = d_path(path, tmp, PAGE_SIZE);
1643 	len = PTR_ERR(pathname);
1644 	if (IS_ERR(pathname))
1645 		goto out;
1646 	len = tmp + PAGE_SIZE - 1 - pathname;
1647 
1648 	if (len > buflen)
1649 		len = buflen;
1650 	if (copy_to_user(buffer, pathname, len))
1651 		len = -EFAULT;
1652  out:
1653 	free_page((unsigned long)tmp);
1654 	return len;
1655 }
1656 
1657 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1658 {
1659 	int error = -EACCES;
1660 	struct inode *inode = d_inode(dentry);
1661 	struct path path;
1662 
1663 	/* Are we allowed to snoop on the tasks file descriptors? */
1664 	if (!proc_fd_access_allowed(inode))
1665 		goto out;
1666 
1667 	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1668 	if (error)
1669 		goto out;
1670 
1671 	error = do_proc_readlink(&path, buffer, buflen);
1672 	path_put(&path);
1673 out:
1674 	return error;
1675 }
1676 
1677 const struct inode_operations proc_pid_link_inode_operations = {
1678 	.readlink	= proc_pid_readlink,
1679 	.get_link	= proc_pid_get_link,
1680 	.setattr	= proc_setattr,
1681 };
1682 
1683 
1684 /* building an inode */
1685 
1686 void task_dump_owner(struct task_struct *task, umode_t mode,
1687 		     kuid_t *ruid, kgid_t *rgid)
1688 {
1689 	/* Depending on the state of dumpable compute who should own a
1690 	 * proc file for a task.
1691 	 */
1692 	const struct cred *cred;
1693 	kuid_t uid;
1694 	kgid_t gid;
1695 
1696 	/* Default to the tasks effective ownership */
1697 	rcu_read_lock();
1698 	cred = __task_cred(task);
1699 	uid = cred->euid;
1700 	gid = cred->egid;
1701 	rcu_read_unlock();
1702 
1703 	/*
1704 	 * Before the /proc/pid/status file was created the only way to read
1705 	 * the effective uid of a /process was to stat /proc/pid.  Reading
1706 	 * /proc/pid/status is slow enough that procps and other packages
1707 	 * kept stating /proc/pid.  To keep the rules in /proc simple I have
1708 	 * made this apply to all per process world readable and executable
1709 	 * directories.
1710 	 */
1711 	if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1712 		struct mm_struct *mm;
1713 		task_lock(task);
1714 		mm = task->mm;
1715 		/* Make non-dumpable tasks owned by some root */
1716 		if (mm) {
1717 			if (get_dumpable(mm) != SUID_DUMP_USER) {
1718 				struct user_namespace *user_ns = mm->user_ns;
1719 
1720 				uid = make_kuid(user_ns, 0);
1721 				if (!uid_valid(uid))
1722 					uid = GLOBAL_ROOT_UID;
1723 
1724 				gid = make_kgid(user_ns, 0);
1725 				if (!gid_valid(gid))
1726 					gid = GLOBAL_ROOT_GID;
1727 			}
1728 		} else {
1729 			uid = GLOBAL_ROOT_UID;
1730 			gid = GLOBAL_ROOT_GID;
1731 		}
1732 		task_unlock(task);
1733 	}
1734 	*ruid = uid;
1735 	*rgid = gid;
1736 }
1737 
1738 struct inode *proc_pid_make_inode(struct super_block * sb,
1739 				  struct task_struct *task, umode_t mode)
1740 {
1741 	struct inode * inode;
1742 	struct proc_inode *ei;
1743 
1744 	/* We need a new inode */
1745 
1746 	inode = new_inode(sb);
1747 	if (!inode)
1748 		goto out;
1749 
1750 	/* Common stuff */
1751 	ei = PROC_I(inode);
1752 	inode->i_mode = mode;
1753 	inode->i_ino = get_next_ino();
1754 	inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
1755 	inode->i_op = &proc_def_inode_operations;
1756 
1757 	/*
1758 	 * grab the reference to task.
1759 	 */
1760 	ei->pid = get_task_pid(task, PIDTYPE_PID);
1761 	if (!ei->pid)
1762 		goto out_unlock;
1763 
1764 	task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1765 	security_task_to_inode(task, inode);
1766 
1767 out:
1768 	return inode;
1769 
1770 out_unlock:
1771 	iput(inode);
1772 	return NULL;
1773 }
1774 
1775 int pid_getattr(const struct path *path, struct kstat *stat,
1776 		u32 request_mask, unsigned int query_flags)
1777 {
1778 	struct inode *inode = d_inode(path->dentry);
1779 	struct task_struct *task;
1780 	struct pid_namespace *pid = path->dentry->d_sb->s_fs_info;
1781 
1782 	generic_fillattr(inode, stat);
1783 
1784 	rcu_read_lock();
1785 	stat->uid = GLOBAL_ROOT_UID;
1786 	stat->gid = GLOBAL_ROOT_GID;
1787 	task = pid_task(proc_pid(inode), PIDTYPE_PID);
1788 	if (task) {
1789 		if (!has_pid_permissions(pid, task, HIDEPID_INVISIBLE)) {
1790 			rcu_read_unlock();
1791 			/*
1792 			 * This doesn't prevent learning whether PID exists,
1793 			 * it only makes getattr() consistent with readdir().
1794 			 */
1795 			return -ENOENT;
1796 		}
1797 		task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
1798 	}
1799 	rcu_read_unlock();
1800 	return 0;
1801 }
1802 
1803 /* dentry stuff */
1804 
1805 /*
1806  *	Exceptional case: normally we are not allowed to unhash a busy
1807  * directory. In this case, however, we can do it - no aliasing problems
1808  * due to the way we treat inodes.
1809  *
1810  * Rewrite the inode's ownerships here because the owning task may have
1811  * performed a setuid(), etc.
1812  *
1813  */
1814 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1815 {
1816 	struct inode *inode;
1817 	struct task_struct *task;
1818 
1819 	if (flags & LOOKUP_RCU)
1820 		return -ECHILD;
1821 
1822 	inode = d_inode(dentry);
1823 	task = get_proc_task(inode);
1824 
1825 	if (task) {
1826 		task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
1827 
1828 		inode->i_mode &= ~(S_ISUID | S_ISGID);
1829 		security_task_to_inode(task, inode);
1830 		put_task_struct(task);
1831 		return 1;
1832 	}
1833 	return 0;
1834 }
1835 
1836 static inline bool proc_inode_is_dead(struct inode *inode)
1837 {
1838 	return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1839 }
1840 
1841 int pid_delete_dentry(const struct dentry *dentry)
1842 {
1843 	/* Is the task we represent dead?
1844 	 * If so, then don't put the dentry on the lru list,
1845 	 * kill it immediately.
1846 	 */
1847 	return proc_inode_is_dead(d_inode(dentry));
1848 }
1849 
1850 const struct dentry_operations pid_dentry_operations =
1851 {
1852 	.d_revalidate	= pid_revalidate,
1853 	.d_delete	= pid_delete_dentry,
1854 };
1855 
1856 /* Lookups */
1857 
1858 /*
1859  * Fill a directory entry.
1860  *
1861  * If possible create the dcache entry and derive our inode number and
1862  * file type from dcache entry.
1863  *
1864  * Since all of the proc inode numbers are dynamically generated, the inode
1865  * numbers do not exist until the inode is cache.  This means creating the
1866  * the dcache entry in readdir is necessary to keep the inode numbers
1867  * reported by readdir in sync with the inode numbers reported
1868  * by stat.
1869  */
1870 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
1871 	const char *name, int len,
1872 	instantiate_t instantiate, struct task_struct *task, const void *ptr)
1873 {
1874 	struct dentry *child, *dir = file->f_path.dentry;
1875 	struct qstr qname = QSTR_INIT(name, len);
1876 	struct inode *inode;
1877 	unsigned type;
1878 	ino_t ino;
1879 
1880 	child = d_hash_and_lookup(dir, &qname);
1881 	if (!child) {
1882 		DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
1883 		child = d_alloc_parallel(dir, &qname, &wq);
1884 		if (IS_ERR(child))
1885 			goto end_instantiate;
1886 		if (d_in_lookup(child)) {
1887 			int err = instantiate(d_inode(dir), child, task, ptr);
1888 			d_lookup_done(child);
1889 			if (err < 0) {
1890 				dput(child);
1891 				goto end_instantiate;
1892 			}
1893 		}
1894 	}
1895 	inode = d_inode(child);
1896 	ino = inode->i_ino;
1897 	type = inode->i_mode >> 12;
1898 	dput(child);
1899 	return dir_emit(ctx, name, len, ino, type);
1900 
1901 end_instantiate:
1902 	return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
1903 }
1904 
1905 /*
1906  * dname_to_vma_addr - maps a dentry name into two unsigned longs
1907  * which represent vma start and end addresses.
1908  */
1909 static int dname_to_vma_addr(struct dentry *dentry,
1910 			     unsigned long *start, unsigned long *end)
1911 {
1912 	const char *str = dentry->d_name.name;
1913 	unsigned long long sval, eval;
1914 	unsigned int len;
1915 
1916 	len = _parse_integer(str, 16, &sval);
1917 	if (len & KSTRTOX_OVERFLOW)
1918 		return -EINVAL;
1919 	if (sval != (unsigned long)sval)
1920 		return -EINVAL;
1921 	str += len;
1922 
1923 	if (*str != '-')
1924 		return -EINVAL;
1925 	str++;
1926 
1927 	len = _parse_integer(str, 16, &eval);
1928 	if (len & KSTRTOX_OVERFLOW)
1929 		return -EINVAL;
1930 	if (eval != (unsigned long)eval)
1931 		return -EINVAL;
1932 	str += len;
1933 
1934 	if (*str != '\0')
1935 		return -EINVAL;
1936 
1937 	*start = sval;
1938 	*end = eval;
1939 
1940 	return 0;
1941 }
1942 
1943 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1944 {
1945 	unsigned long vm_start, vm_end;
1946 	bool exact_vma_exists = false;
1947 	struct mm_struct *mm = NULL;
1948 	struct task_struct *task;
1949 	struct inode *inode;
1950 	int status = 0;
1951 
1952 	if (flags & LOOKUP_RCU)
1953 		return -ECHILD;
1954 
1955 	inode = d_inode(dentry);
1956 	task = get_proc_task(inode);
1957 	if (!task)
1958 		goto out_notask;
1959 
1960 	mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
1961 	if (IS_ERR_OR_NULL(mm))
1962 		goto out;
1963 
1964 	if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1965 		down_read(&mm->mmap_sem);
1966 		exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1967 		up_read(&mm->mmap_sem);
1968 	}
1969 
1970 	mmput(mm);
1971 
1972 	if (exact_vma_exists) {
1973 		task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1974 
1975 		security_task_to_inode(task, inode);
1976 		status = 1;
1977 	}
1978 
1979 out:
1980 	put_task_struct(task);
1981 
1982 out_notask:
1983 	return status;
1984 }
1985 
1986 static const struct dentry_operations tid_map_files_dentry_operations = {
1987 	.d_revalidate	= map_files_d_revalidate,
1988 	.d_delete	= pid_delete_dentry,
1989 };
1990 
1991 static int map_files_get_link(struct dentry *dentry, struct path *path)
1992 {
1993 	unsigned long vm_start, vm_end;
1994 	struct vm_area_struct *vma;
1995 	struct task_struct *task;
1996 	struct mm_struct *mm;
1997 	int rc;
1998 
1999 	rc = -ENOENT;
2000 	task = get_proc_task(d_inode(dentry));
2001 	if (!task)
2002 		goto out;
2003 
2004 	mm = get_task_mm(task);
2005 	put_task_struct(task);
2006 	if (!mm)
2007 		goto out;
2008 
2009 	rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2010 	if (rc)
2011 		goto out_mmput;
2012 
2013 	rc = -ENOENT;
2014 	down_read(&mm->mmap_sem);
2015 	vma = find_exact_vma(mm, vm_start, vm_end);
2016 	if (vma && vma->vm_file) {
2017 		*path = vma->vm_file->f_path;
2018 		path_get(path);
2019 		rc = 0;
2020 	}
2021 	up_read(&mm->mmap_sem);
2022 
2023 out_mmput:
2024 	mmput(mm);
2025 out:
2026 	return rc;
2027 }
2028 
2029 struct map_files_info {
2030 	unsigned long	start;
2031 	unsigned long	end;
2032 	fmode_t		mode;
2033 };
2034 
2035 /*
2036  * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
2037  * symlinks may be used to bypass permissions on ancestor directories in the
2038  * path to the file in question.
2039  */
2040 static const char *
2041 proc_map_files_get_link(struct dentry *dentry,
2042 			struct inode *inode,
2043 		        struct delayed_call *done)
2044 {
2045 	if (!capable(CAP_SYS_ADMIN))
2046 		return ERR_PTR(-EPERM);
2047 
2048 	return proc_pid_get_link(dentry, inode, done);
2049 }
2050 
2051 /*
2052  * Identical to proc_pid_link_inode_operations except for get_link()
2053  */
2054 static const struct inode_operations proc_map_files_link_inode_operations = {
2055 	.readlink	= proc_pid_readlink,
2056 	.get_link	= proc_map_files_get_link,
2057 	.setattr	= proc_setattr,
2058 };
2059 
2060 static int
2061 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2062 			   struct task_struct *task, const void *ptr)
2063 {
2064 	fmode_t mode = (fmode_t)(unsigned long)ptr;
2065 	struct proc_inode *ei;
2066 	struct inode *inode;
2067 
2068 	inode = proc_pid_make_inode(dir->i_sb, task, S_IFLNK |
2069 				    ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2070 				    ((mode & FMODE_WRITE) ? S_IWUSR : 0));
2071 	if (!inode)
2072 		return -ENOENT;
2073 
2074 	ei = PROC_I(inode);
2075 	ei->op.proc_get_link = map_files_get_link;
2076 
2077 	inode->i_op = &proc_map_files_link_inode_operations;
2078 	inode->i_size = 64;
2079 
2080 	d_set_d_op(dentry, &tid_map_files_dentry_operations);
2081 	d_add(dentry, inode);
2082 
2083 	return 0;
2084 }
2085 
2086 static struct dentry *proc_map_files_lookup(struct inode *dir,
2087 		struct dentry *dentry, unsigned int flags)
2088 {
2089 	unsigned long vm_start, vm_end;
2090 	struct vm_area_struct *vma;
2091 	struct task_struct *task;
2092 	int result;
2093 	struct mm_struct *mm;
2094 
2095 	result = -ENOENT;
2096 	task = get_proc_task(dir);
2097 	if (!task)
2098 		goto out;
2099 
2100 	result = -EACCES;
2101 	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2102 		goto out_put_task;
2103 
2104 	result = -ENOENT;
2105 	if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2106 		goto out_put_task;
2107 
2108 	mm = get_task_mm(task);
2109 	if (!mm)
2110 		goto out_put_task;
2111 
2112 	down_read(&mm->mmap_sem);
2113 	vma = find_exact_vma(mm, vm_start, vm_end);
2114 	if (!vma)
2115 		goto out_no_vma;
2116 
2117 	if (vma->vm_file)
2118 		result = proc_map_files_instantiate(dir, dentry, task,
2119 				(void *)(unsigned long)vma->vm_file->f_mode);
2120 
2121 out_no_vma:
2122 	up_read(&mm->mmap_sem);
2123 	mmput(mm);
2124 out_put_task:
2125 	put_task_struct(task);
2126 out:
2127 	return ERR_PTR(result);
2128 }
2129 
2130 static const struct inode_operations proc_map_files_inode_operations = {
2131 	.lookup		= proc_map_files_lookup,
2132 	.permission	= proc_fd_permission,
2133 	.setattr	= proc_setattr,
2134 };
2135 
2136 static int
2137 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2138 {
2139 	struct vm_area_struct *vma;
2140 	struct task_struct *task;
2141 	struct mm_struct *mm;
2142 	unsigned long nr_files, pos, i;
2143 	struct flex_array *fa = NULL;
2144 	struct map_files_info info;
2145 	struct map_files_info *p;
2146 	int ret;
2147 
2148 	ret = -ENOENT;
2149 	task = get_proc_task(file_inode(file));
2150 	if (!task)
2151 		goto out;
2152 
2153 	ret = -EACCES;
2154 	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2155 		goto out_put_task;
2156 
2157 	ret = 0;
2158 	if (!dir_emit_dots(file, ctx))
2159 		goto out_put_task;
2160 
2161 	mm = get_task_mm(task);
2162 	if (!mm)
2163 		goto out_put_task;
2164 	down_read(&mm->mmap_sem);
2165 
2166 	nr_files = 0;
2167 
2168 	/*
2169 	 * We need two passes here:
2170 	 *
2171 	 *  1) Collect vmas of mapped files with mmap_sem taken
2172 	 *  2) Release mmap_sem and instantiate entries
2173 	 *
2174 	 * otherwise we get lockdep complained, since filldir()
2175 	 * routine might require mmap_sem taken in might_fault().
2176 	 */
2177 
2178 	for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2179 		if (vma->vm_file && ++pos > ctx->pos)
2180 			nr_files++;
2181 	}
2182 
2183 	if (nr_files) {
2184 		fa = flex_array_alloc(sizeof(info), nr_files,
2185 					GFP_KERNEL);
2186 		if (!fa || flex_array_prealloc(fa, 0, nr_files,
2187 						GFP_KERNEL)) {
2188 			ret = -ENOMEM;
2189 			if (fa)
2190 				flex_array_free(fa);
2191 			up_read(&mm->mmap_sem);
2192 			mmput(mm);
2193 			goto out_put_task;
2194 		}
2195 		for (i = 0, vma = mm->mmap, pos = 2; vma;
2196 				vma = vma->vm_next) {
2197 			if (!vma->vm_file)
2198 				continue;
2199 			if (++pos <= ctx->pos)
2200 				continue;
2201 
2202 			info.start = vma->vm_start;
2203 			info.end = vma->vm_end;
2204 			info.mode = vma->vm_file->f_mode;
2205 			if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2206 				BUG();
2207 		}
2208 	}
2209 	up_read(&mm->mmap_sem);
2210 
2211 	for (i = 0; i < nr_files; i++) {
2212 		char buf[4 * sizeof(long) + 2];	/* max: %lx-%lx\0 */
2213 		unsigned int len;
2214 
2215 		p = flex_array_get(fa, i);
2216 		len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
2217 		if (!proc_fill_cache(file, ctx,
2218 				      buf, len,
2219 				      proc_map_files_instantiate,
2220 				      task,
2221 				      (void *)(unsigned long)p->mode))
2222 			break;
2223 		ctx->pos++;
2224 	}
2225 	if (fa)
2226 		flex_array_free(fa);
2227 	mmput(mm);
2228 
2229 out_put_task:
2230 	put_task_struct(task);
2231 out:
2232 	return ret;
2233 }
2234 
2235 static const struct file_operations proc_map_files_operations = {
2236 	.read		= generic_read_dir,
2237 	.iterate_shared	= proc_map_files_readdir,
2238 	.llseek		= generic_file_llseek,
2239 };
2240 
2241 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2242 struct timers_private {
2243 	struct pid *pid;
2244 	struct task_struct *task;
2245 	struct sighand_struct *sighand;
2246 	struct pid_namespace *ns;
2247 	unsigned long flags;
2248 };
2249 
2250 static void *timers_start(struct seq_file *m, loff_t *pos)
2251 {
2252 	struct timers_private *tp = m->private;
2253 
2254 	tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2255 	if (!tp->task)
2256 		return ERR_PTR(-ESRCH);
2257 
2258 	tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2259 	if (!tp->sighand)
2260 		return ERR_PTR(-ESRCH);
2261 
2262 	return seq_list_start(&tp->task->signal->posix_timers, *pos);
2263 }
2264 
2265 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2266 {
2267 	struct timers_private *tp = m->private;
2268 	return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2269 }
2270 
2271 static void timers_stop(struct seq_file *m, void *v)
2272 {
2273 	struct timers_private *tp = m->private;
2274 
2275 	if (tp->sighand) {
2276 		unlock_task_sighand(tp->task, &tp->flags);
2277 		tp->sighand = NULL;
2278 	}
2279 
2280 	if (tp->task) {
2281 		put_task_struct(tp->task);
2282 		tp->task = NULL;
2283 	}
2284 }
2285 
2286 static int show_timer(struct seq_file *m, void *v)
2287 {
2288 	struct k_itimer *timer;
2289 	struct timers_private *tp = m->private;
2290 	int notify;
2291 	static const char * const nstr[] = {
2292 		[SIGEV_SIGNAL] = "signal",
2293 		[SIGEV_NONE] = "none",
2294 		[SIGEV_THREAD] = "thread",
2295 	};
2296 
2297 	timer = list_entry((struct list_head *)v, struct k_itimer, list);
2298 	notify = timer->it_sigev_notify;
2299 
2300 	seq_printf(m, "ID: %d\n", timer->it_id);
2301 	seq_printf(m, "signal: %d/%px\n",
2302 		   timer->sigq->info.si_signo,
2303 		   timer->sigq->info.si_value.sival_ptr);
2304 	seq_printf(m, "notify: %s/%s.%d\n",
2305 		   nstr[notify & ~SIGEV_THREAD_ID],
2306 		   (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2307 		   pid_nr_ns(timer->it_pid, tp->ns));
2308 	seq_printf(m, "ClockID: %d\n", timer->it_clock);
2309 
2310 	return 0;
2311 }
2312 
2313 static const struct seq_operations proc_timers_seq_ops = {
2314 	.start	= timers_start,
2315 	.next	= timers_next,
2316 	.stop	= timers_stop,
2317 	.show	= show_timer,
2318 };
2319 
2320 static int proc_timers_open(struct inode *inode, struct file *file)
2321 {
2322 	struct timers_private *tp;
2323 
2324 	tp = __seq_open_private(file, &proc_timers_seq_ops,
2325 			sizeof(struct timers_private));
2326 	if (!tp)
2327 		return -ENOMEM;
2328 
2329 	tp->pid = proc_pid(inode);
2330 	tp->ns = inode->i_sb->s_fs_info;
2331 	return 0;
2332 }
2333 
2334 static const struct file_operations proc_timers_operations = {
2335 	.open		= proc_timers_open,
2336 	.read		= seq_read,
2337 	.llseek		= seq_lseek,
2338 	.release	= seq_release_private,
2339 };
2340 #endif
2341 
2342 static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2343 					size_t count, loff_t *offset)
2344 {
2345 	struct inode *inode = file_inode(file);
2346 	struct task_struct *p;
2347 	u64 slack_ns;
2348 	int err;
2349 
2350 	err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2351 	if (err < 0)
2352 		return err;
2353 
2354 	p = get_proc_task(inode);
2355 	if (!p)
2356 		return -ESRCH;
2357 
2358 	if (p != current) {
2359 		if (!capable(CAP_SYS_NICE)) {
2360 			count = -EPERM;
2361 			goto out;
2362 		}
2363 
2364 		err = security_task_setscheduler(p);
2365 		if (err) {
2366 			count = err;
2367 			goto out;
2368 		}
2369 	}
2370 
2371 	task_lock(p);
2372 	if (slack_ns == 0)
2373 		p->timer_slack_ns = p->default_timer_slack_ns;
2374 	else
2375 		p->timer_slack_ns = slack_ns;
2376 	task_unlock(p);
2377 
2378 out:
2379 	put_task_struct(p);
2380 
2381 	return count;
2382 }
2383 
2384 static int timerslack_ns_show(struct seq_file *m, void *v)
2385 {
2386 	struct inode *inode = m->private;
2387 	struct task_struct *p;
2388 	int err = 0;
2389 
2390 	p = get_proc_task(inode);
2391 	if (!p)
2392 		return -ESRCH;
2393 
2394 	if (p != current) {
2395 
2396 		if (!capable(CAP_SYS_NICE)) {
2397 			err = -EPERM;
2398 			goto out;
2399 		}
2400 		err = security_task_getscheduler(p);
2401 		if (err)
2402 			goto out;
2403 	}
2404 
2405 	task_lock(p);
2406 	seq_printf(m, "%llu\n", p->timer_slack_ns);
2407 	task_unlock(p);
2408 
2409 out:
2410 	put_task_struct(p);
2411 
2412 	return err;
2413 }
2414 
2415 static int timerslack_ns_open(struct inode *inode, struct file *filp)
2416 {
2417 	return single_open(filp, timerslack_ns_show, inode);
2418 }
2419 
2420 static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2421 	.open		= timerslack_ns_open,
2422 	.read		= seq_read,
2423 	.write		= timerslack_ns_write,
2424 	.llseek		= seq_lseek,
2425 	.release	= single_release,
2426 };
2427 
2428 static int proc_pident_instantiate(struct inode *dir,
2429 	struct dentry *dentry, struct task_struct *task, const void *ptr)
2430 {
2431 	const struct pid_entry *p = ptr;
2432 	struct inode *inode;
2433 	struct proc_inode *ei;
2434 
2435 	inode = proc_pid_make_inode(dir->i_sb, task, p->mode);
2436 	if (!inode)
2437 		goto out;
2438 
2439 	ei = PROC_I(inode);
2440 	if (S_ISDIR(inode->i_mode))
2441 		set_nlink(inode, 2);	/* Use getattr to fix if necessary */
2442 	if (p->iop)
2443 		inode->i_op = p->iop;
2444 	if (p->fop)
2445 		inode->i_fop = p->fop;
2446 	ei->op = p->op;
2447 	d_set_d_op(dentry, &pid_dentry_operations);
2448 	d_add(dentry, inode);
2449 	/* Close the race of the process dying before we return the dentry */
2450 	if (pid_revalidate(dentry, 0))
2451 		return 0;
2452 out:
2453 	return -ENOENT;
2454 }
2455 
2456 static struct dentry *proc_pident_lookup(struct inode *dir,
2457 					 struct dentry *dentry,
2458 					 const struct pid_entry *ents,
2459 					 unsigned int nents)
2460 {
2461 	int error;
2462 	struct task_struct *task = get_proc_task(dir);
2463 	const struct pid_entry *p, *last;
2464 
2465 	error = -ENOENT;
2466 
2467 	if (!task)
2468 		goto out_no_task;
2469 
2470 	/*
2471 	 * Yes, it does not scale. And it should not. Don't add
2472 	 * new entries into /proc/<tgid>/ without very good reasons.
2473 	 */
2474 	last = &ents[nents];
2475 	for (p = ents; p < last; p++) {
2476 		if (p->len != dentry->d_name.len)
2477 			continue;
2478 		if (!memcmp(dentry->d_name.name, p->name, p->len))
2479 			break;
2480 	}
2481 	if (p >= last)
2482 		goto out;
2483 
2484 	error = proc_pident_instantiate(dir, dentry, task, p);
2485 out:
2486 	put_task_struct(task);
2487 out_no_task:
2488 	return ERR_PTR(error);
2489 }
2490 
2491 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2492 		const struct pid_entry *ents, unsigned int nents)
2493 {
2494 	struct task_struct *task = get_proc_task(file_inode(file));
2495 	const struct pid_entry *p;
2496 
2497 	if (!task)
2498 		return -ENOENT;
2499 
2500 	if (!dir_emit_dots(file, ctx))
2501 		goto out;
2502 
2503 	if (ctx->pos >= nents + 2)
2504 		goto out;
2505 
2506 	for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
2507 		if (!proc_fill_cache(file, ctx, p->name, p->len,
2508 				proc_pident_instantiate, task, p))
2509 			break;
2510 		ctx->pos++;
2511 	}
2512 out:
2513 	put_task_struct(task);
2514 	return 0;
2515 }
2516 
2517 #ifdef CONFIG_SECURITY
2518 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2519 				  size_t count, loff_t *ppos)
2520 {
2521 	struct inode * inode = file_inode(file);
2522 	char *p = NULL;
2523 	ssize_t length;
2524 	struct task_struct *task = get_proc_task(inode);
2525 
2526 	if (!task)
2527 		return -ESRCH;
2528 
2529 	length = security_getprocattr(task,
2530 				      (char*)file->f_path.dentry->d_name.name,
2531 				      &p);
2532 	put_task_struct(task);
2533 	if (length > 0)
2534 		length = simple_read_from_buffer(buf, count, ppos, p, length);
2535 	kfree(p);
2536 	return length;
2537 }
2538 
2539 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2540 				   size_t count, loff_t *ppos)
2541 {
2542 	struct inode * inode = file_inode(file);
2543 	void *page;
2544 	ssize_t length;
2545 	struct task_struct *task = get_proc_task(inode);
2546 
2547 	length = -ESRCH;
2548 	if (!task)
2549 		goto out_no_task;
2550 
2551 	/* A task may only write its own attributes. */
2552 	length = -EACCES;
2553 	if (current != task)
2554 		goto out;
2555 
2556 	if (count > PAGE_SIZE)
2557 		count = PAGE_SIZE;
2558 
2559 	/* No partial writes. */
2560 	length = -EINVAL;
2561 	if (*ppos != 0)
2562 		goto out;
2563 
2564 	page = memdup_user(buf, count);
2565 	if (IS_ERR(page)) {
2566 		length = PTR_ERR(page);
2567 		goto out;
2568 	}
2569 
2570 	/* Guard against adverse ptrace interaction */
2571 	length = mutex_lock_interruptible(&current->signal->cred_guard_mutex);
2572 	if (length < 0)
2573 		goto out_free;
2574 
2575 	length = security_setprocattr(file->f_path.dentry->d_name.name,
2576 				      page, count);
2577 	mutex_unlock(&current->signal->cred_guard_mutex);
2578 out_free:
2579 	kfree(page);
2580 out:
2581 	put_task_struct(task);
2582 out_no_task:
2583 	return length;
2584 }
2585 
2586 static const struct file_operations proc_pid_attr_operations = {
2587 	.read		= proc_pid_attr_read,
2588 	.write		= proc_pid_attr_write,
2589 	.llseek		= generic_file_llseek,
2590 };
2591 
2592 static const struct pid_entry attr_dir_stuff[] = {
2593 	REG("current",    S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2594 	REG("prev",       S_IRUGO,	   proc_pid_attr_operations),
2595 	REG("exec",       S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2596 	REG("fscreate",   S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2597 	REG("keycreate",  S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2598 	REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2599 };
2600 
2601 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2602 {
2603 	return proc_pident_readdir(file, ctx,
2604 				   attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2605 }
2606 
2607 static const struct file_operations proc_attr_dir_operations = {
2608 	.read		= generic_read_dir,
2609 	.iterate_shared	= proc_attr_dir_readdir,
2610 	.llseek		= generic_file_llseek,
2611 };
2612 
2613 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2614 				struct dentry *dentry, unsigned int flags)
2615 {
2616 	return proc_pident_lookup(dir, dentry,
2617 				  attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2618 }
2619 
2620 static const struct inode_operations proc_attr_dir_inode_operations = {
2621 	.lookup		= proc_attr_dir_lookup,
2622 	.getattr	= pid_getattr,
2623 	.setattr	= proc_setattr,
2624 };
2625 
2626 #endif
2627 
2628 #ifdef CONFIG_ELF_CORE
2629 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2630 					 size_t count, loff_t *ppos)
2631 {
2632 	struct task_struct *task = get_proc_task(file_inode(file));
2633 	struct mm_struct *mm;
2634 	char buffer[PROC_NUMBUF];
2635 	size_t len;
2636 	int ret;
2637 
2638 	if (!task)
2639 		return -ESRCH;
2640 
2641 	ret = 0;
2642 	mm = get_task_mm(task);
2643 	if (mm) {
2644 		len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2645 			       ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2646 				MMF_DUMP_FILTER_SHIFT));
2647 		mmput(mm);
2648 		ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2649 	}
2650 
2651 	put_task_struct(task);
2652 
2653 	return ret;
2654 }
2655 
2656 static ssize_t proc_coredump_filter_write(struct file *file,
2657 					  const char __user *buf,
2658 					  size_t count,
2659 					  loff_t *ppos)
2660 {
2661 	struct task_struct *task;
2662 	struct mm_struct *mm;
2663 	unsigned int val;
2664 	int ret;
2665 	int i;
2666 	unsigned long mask;
2667 
2668 	ret = kstrtouint_from_user(buf, count, 0, &val);
2669 	if (ret < 0)
2670 		return ret;
2671 
2672 	ret = -ESRCH;
2673 	task = get_proc_task(file_inode(file));
2674 	if (!task)
2675 		goto out_no_task;
2676 
2677 	mm = get_task_mm(task);
2678 	if (!mm)
2679 		goto out_no_mm;
2680 	ret = 0;
2681 
2682 	for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2683 		if (val & mask)
2684 			set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2685 		else
2686 			clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2687 	}
2688 
2689 	mmput(mm);
2690  out_no_mm:
2691 	put_task_struct(task);
2692  out_no_task:
2693 	if (ret < 0)
2694 		return ret;
2695 	return count;
2696 }
2697 
2698 static const struct file_operations proc_coredump_filter_operations = {
2699 	.read		= proc_coredump_filter_read,
2700 	.write		= proc_coredump_filter_write,
2701 	.llseek		= generic_file_llseek,
2702 };
2703 #endif
2704 
2705 #ifdef CONFIG_TASK_IO_ACCOUNTING
2706 static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2707 {
2708 	struct task_io_accounting acct = task->ioac;
2709 	unsigned long flags;
2710 	int result;
2711 
2712 	result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2713 	if (result)
2714 		return result;
2715 
2716 	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
2717 		result = -EACCES;
2718 		goto out_unlock;
2719 	}
2720 
2721 	if (whole && lock_task_sighand(task, &flags)) {
2722 		struct task_struct *t = task;
2723 
2724 		task_io_accounting_add(&acct, &task->signal->ioac);
2725 		while_each_thread(task, t)
2726 			task_io_accounting_add(&acct, &t->ioac);
2727 
2728 		unlock_task_sighand(task, &flags);
2729 	}
2730 	seq_printf(m,
2731 		   "rchar: %llu\n"
2732 		   "wchar: %llu\n"
2733 		   "syscr: %llu\n"
2734 		   "syscw: %llu\n"
2735 		   "read_bytes: %llu\n"
2736 		   "write_bytes: %llu\n"
2737 		   "cancelled_write_bytes: %llu\n",
2738 		   (unsigned long long)acct.rchar,
2739 		   (unsigned long long)acct.wchar,
2740 		   (unsigned long long)acct.syscr,
2741 		   (unsigned long long)acct.syscw,
2742 		   (unsigned long long)acct.read_bytes,
2743 		   (unsigned long long)acct.write_bytes,
2744 		   (unsigned long long)acct.cancelled_write_bytes);
2745 	result = 0;
2746 
2747 out_unlock:
2748 	mutex_unlock(&task->signal->cred_guard_mutex);
2749 	return result;
2750 }
2751 
2752 static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2753 				  struct pid *pid, struct task_struct *task)
2754 {
2755 	return do_io_accounting(task, m, 0);
2756 }
2757 
2758 static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2759 				   struct pid *pid, struct task_struct *task)
2760 {
2761 	return do_io_accounting(task, m, 1);
2762 }
2763 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2764 
2765 #ifdef CONFIG_USER_NS
2766 static int proc_id_map_open(struct inode *inode, struct file *file,
2767 	const struct seq_operations *seq_ops)
2768 {
2769 	struct user_namespace *ns = NULL;
2770 	struct task_struct *task;
2771 	struct seq_file *seq;
2772 	int ret = -EINVAL;
2773 
2774 	task = get_proc_task(inode);
2775 	if (task) {
2776 		rcu_read_lock();
2777 		ns = get_user_ns(task_cred_xxx(task, user_ns));
2778 		rcu_read_unlock();
2779 		put_task_struct(task);
2780 	}
2781 	if (!ns)
2782 		goto err;
2783 
2784 	ret = seq_open(file, seq_ops);
2785 	if (ret)
2786 		goto err_put_ns;
2787 
2788 	seq = file->private_data;
2789 	seq->private = ns;
2790 
2791 	return 0;
2792 err_put_ns:
2793 	put_user_ns(ns);
2794 err:
2795 	return ret;
2796 }
2797 
2798 static int proc_id_map_release(struct inode *inode, struct file *file)
2799 {
2800 	struct seq_file *seq = file->private_data;
2801 	struct user_namespace *ns = seq->private;
2802 	put_user_ns(ns);
2803 	return seq_release(inode, file);
2804 }
2805 
2806 static int proc_uid_map_open(struct inode *inode, struct file *file)
2807 {
2808 	return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2809 }
2810 
2811 static int proc_gid_map_open(struct inode *inode, struct file *file)
2812 {
2813 	return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2814 }
2815 
2816 static int proc_projid_map_open(struct inode *inode, struct file *file)
2817 {
2818 	return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2819 }
2820 
2821 static const struct file_operations proc_uid_map_operations = {
2822 	.open		= proc_uid_map_open,
2823 	.write		= proc_uid_map_write,
2824 	.read		= seq_read,
2825 	.llseek		= seq_lseek,
2826 	.release	= proc_id_map_release,
2827 };
2828 
2829 static const struct file_operations proc_gid_map_operations = {
2830 	.open		= proc_gid_map_open,
2831 	.write		= proc_gid_map_write,
2832 	.read		= seq_read,
2833 	.llseek		= seq_lseek,
2834 	.release	= proc_id_map_release,
2835 };
2836 
2837 static const struct file_operations proc_projid_map_operations = {
2838 	.open		= proc_projid_map_open,
2839 	.write		= proc_projid_map_write,
2840 	.read		= seq_read,
2841 	.llseek		= seq_lseek,
2842 	.release	= proc_id_map_release,
2843 };
2844 
2845 static int proc_setgroups_open(struct inode *inode, struct file *file)
2846 {
2847 	struct user_namespace *ns = NULL;
2848 	struct task_struct *task;
2849 	int ret;
2850 
2851 	ret = -ESRCH;
2852 	task = get_proc_task(inode);
2853 	if (task) {
2854 		rcu_read_lock();
2855 		ns = get_user_ns(task_cred_xxx(task, user_ns));
2856 		rcu_read_unlock();
2857 		put_task_struct(task);
2858 	}
2859 	if (!ns)
2860 		goto err;
2861 
2862 	if (file->f_mode & FMODE_WRITE) {
2863 		ret = -EACCES;
2864 		if (!ns_capable(ns, CAP_SYS_ADMIN))
2865 			goto err_put_ns;
2866 	}
2867 
2868 	ret = single_open(file, &proc_setgroups_show, ns);
2869 	if (ret)
2870 		goto err_put_ns;
2871 
2872 	return 0;
2873 err_put_ns:
2874 	put_user_ns(ns);
2875 err:
2876 	return ret;
2877 }
2878 
2879 static int proc_setgroups_release(struct inode *inode, struct file *file)
2880 {
2881 	struct seq_file *seq = file->private_data;
2882 	struct user_namespace *ns = seq->private;
2883 	int ret = single_release(inode, file);
2884 	put_user_ns(ns);
2885 	return ret;
2886 }
2887 
2888 static const struct file_operations proc_setgroups_operations = {
2889 	.open		= proc_setgroups_open,
2890 	.write		= proc_setgroups_write,
2891 	.read		= seq_read,
2892 	.llseek		= seq_lseek,
2893 	.release	= proc_setgroups_release,
2894 };
2895 #endif /* CONFIG_USER_NS */
2896 
2897 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2898 				struct pid *pid, struct task_struct *task)
2899 {
2900 	int err = lock_trace(task);
2901 	if (!err) {
2902 		seq_printf(m, "%08x\n", task->personality);
2903 		unlock_trace(task);
2904 	}
2905 	return err;
2906 }
2907 
2908 #ifdef CONFIG_LIVEPATCH
2909 static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
2910 				struct pid *pid, struct task_struct *task)
2911 {
2912 	seq_printf(m, "%d\n", task->patch_state);
2913 	return 0;
2914 }
2915 #endif /* CONFIG_LIVEPATCH */
2916 
2917 /*
2918  * Thread groups
2919  */
2920 static const struct file_operations proc_task_operations;
2921 static const struct inode_operations proc_task_inode_operations;
2922 
2923 static const struct pid_entry tgid_base_stuff[] = {
2924 	DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2925 	DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2926 	DIR("map_files",  S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2927 	DIR("fdinfo",     S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2928 	DIR("ns",	  S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2929 #ifdef CONFIG_NET
2930 	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2931 #endif
2932 	REG("environ",    S_IRUSR, proc_environ_operations),
2933 	REG("auxv",       S_IRUSR, proc_auxv_operations),
2934 	ONE("status",     S_IRUGO, proc_pid_status),
2935 	ONE("personality", S_IRUSR, proc_pid_personality),
2936 	ONE("limits",	  S_IRUGO, proc_pid_limits),
2937 #ifdef CONFIG_SCHED_DEBUG
2938 	REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2939 #endif
2940 #ifdef CONFIG_SCHED_AUTOGROUP
2941 	REG("autogroup",  S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2942 #endif
2943 	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2944 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2945 	ONE("syscall",    S_IRUSR, proc_pid_syscall),
2946 #endif
2947 	REG("cmdline",    S_IRUGO, proc_pid_cmdline_ops),
2948 	ONE("stat",       S_IRUGO, proc_tgid_stat),
2949 	ONE("statm",      S_IRUGO, proc_pid_statm),
2950 	REG("maps",       S_IRUGO, proc_pid_maps_operations),
2951 #ifdef CONFIG_NUMA
2952 	REG("numa_maps",  S_IRUGO, proc_pid_numa_maps_operations),
2953 #endif
2954 	REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
2955 	LNK("cwd",        proc_cwd_link),
2956 	LNK("root",       proc_root_link),
2957 	LNK("exe",        proc_exe_link),
2958 	REG("mounts",     S_IRUGO, proc_mounts_operations),
2959 	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
2960 	REG("mountstats", S_IRUSR, proc_mountstats_operations),
2961 #ifdef CONFIG_PROC_PAGE_MONITOR
2962 	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2963 	REG("smaps",      S_IRUGO, proc_pid_smaps_operations),
2964 	REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
2965 	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
2966 #endif
2967 #ifdef CONFIG_SECURITY
2968 	DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2969 #endif
2970 #ifdef CONFIG_KALLSYMS
2971 	ONE("wchan",      S_IRUGO, proc_pid_wchan),
2972 #endif
2973 #ifdef CONFIG_STACKTRACE
2974 	ONE("stack",      S_IRUSR, proc_pid_stack),
2975 #endif
2976 #ifdef CONFIG_SCHED_INFO
2977 	ONE("schedstat",  S_IRUGO, proc_pid_schedstat),
2978 #endif
2979 #ifdef CONFIG_LATENCYTOP
2980 	REG("latency",  S_IRUGO, proc_lstats_operations),
2981 #endif
2982 #ifdef CONFIG_PROC_PID_CPUSET
2983 	ONE("cpuset",     S_IRUGO, proc_cpuset_show),
2984 #endif
2985 #ifdef CONFIG_CGROUPS
2986 	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
2987 #endif
2988 	ONE("oom_score",  S_IRUGO, proc_oom_score),
2989 	REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2990 	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2991 #ifdef CONFIG_AUDITSYSCALL
2992 	REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
2993 	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
2994 #endif
2995 #ifdef CONFIG_FAULT_INJECTION
2996 	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2997 	REG("fail-nth", 0644, proc_fail_nth_operations),
2998 #endif
2999 #ifdef CONFIG_ELF_CORE
3000 	REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3001 #endif
3002 #ifdef CONFIG_TASK_IO_ACCOUNTING
3003 	ONE("io",	S_IRUSR, proc_tgid_io_accounting),
3004 #endif
3005 #ifdef CONFIG_HARDWALL
3006 	ONE("hardwall",   S_IRUGO, proc_pid_hardwall),
3007 #endif
3008 #ifdef CONFIG_USER_NS
3009 	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3010 	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3011 	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3012 	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3013 #endif
3014 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3015 	REG("timers",	  S_IRUGO, proc_timers_operations),
3016 #endif
3017 	REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
3018 #ifdef CONFIG_LIVEPATCH
3019 	ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3020 #endif
3021 };
3022 
3023 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
3024 {
3025 	return proc_pident_readdir(file, ctx,
3026 				   tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3027 }
3028 
3029 static const struct file_operations proc_tgid_base_operations = {
3030 	.read		= generic_read_dir,
3031 	.iterate_shared	= proc_tgid_base_readdir,
3032 	.llseek		= generic_file_llseek,
3033 };
3034 
3035 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3036 {
3037 	return proc_pident_lookup(dir, dentry,
3038 				  tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3039 }
3040 
3041 static const struct inode_operations proc_tgid_base_inode_operations = {
3042 	.lookup		= proc_tgid_base_lookup,
3043 	.getattr	= pid_getattr,
3044 	.setattr	= proc_setattr,
3045 	.permission	= proc_pid_permission,
3046 };
3047 
3048 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
3049 {
3050 	struct dentry *dentry, *leader, *dir;
3051 	char buf[10 + 1];
3052 	struct qstr name;
3053 
3054 	name.name = buf;
3055 	name.len = snprintf(buf, sizeof(buf), "%u", pid);
3056 	/* no ->d_hash() rejects on procfs */
3057 	dentry = d_hash_and_lookup(mnt->mnt_root, &name);
3058 	if (dentry) {
3059 		d_invalidate(dentry);
3060 		dput(dentry);
3061 	}
3062 
3063 	if (pid == tgid)
3064 		return;
3065 
3066 	name.name = buf;
3067 	name.len = snprintf(buf, sizeof(buf), "%u", tgid);
3068 	leader = d_hash_and_lookup(mnt->mnt_root, &name);
3069 	if (!leader)
3070 		goto out;
3071 
3072 	name.name = "task";
3073 	name.len = strlen(name.name);
3074 	dir = d_hash_and_lookup(leader, &name);
3075 	if (!dir)
3076 		goto out_put_leader;
3077 
3078 	name.name = buf;
3079 	name.len = snprintf(buf, sizeof(buf), "%u", pid);
3080 	dentry = d_hash_and_lookup(dir, &name);
3081 	if (dentry) {
3082 		d_invalidate(dentry);
3083 		dput(dentry);
3084 	}
3085 
3086 	dput(dir);
3087 out_put_leader:
3088 	dput(leader);
3089 out:
3090 	return;
3091 }
3092 
3093 /**
3094  * proc_flush_task -  Remove dcache entries for @task from the /proc dcache.
3095  * @task: task that should be flushed.
3096  *
3097  * When flushing dentries from proc, one needs to flush them from global
3098  * proc (proc_mnt) and from all the namespaces' procs this task was seen
3099  * in. This call is supposed to do all of this job.
3100  *
3101  * Looks in the dcache for
3102  * /proc/@pid
3103  * /proc/@tgid/task/@pid
3104  * if either directory is present flushes it and all of it'ts children
3105  * from the dcache.
3106  *
3107  * It is safe and reasonable to cache /proc entries for a task until
3108  * that task exits.  After that they just clog up the dcache with
3109  * useless entries, possibly causing useful dcache entries to be
3110  * flushed instead.  This routine is proved to flush those useless
3111  * dcache entries at process exit time.
3112  *
3113  * NOTE: This routine is just an optimization so it does not guarantee
3114  *       that no dcache entries will exist at process exit time it
3115  *       just makes it very unlikely that any will persist.
3116  */
3117 
3118 void proc_flush_task(struct task_struct *task)
3119 {
3120 	int i;
3121 	struct pid *pid, *tgid;
3122 	struct upid *upid;
3123 
3124 	pid = task_pid(task);
3125 	tgid = task_tgid(task);
3126 
3127 	for (i = 0; i <= pid->level; i++) {
3128 		upid = &pid->numbers[i];
3129 		proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3130 					tgid->numbers[i].nr);
3131 	}
3132 }
3133 
3134 static int proc_pid_instantiate(struct inode *dir,
3135 				   struct dentry * dentry,
3136 				   struct task_struct *task, const void *ptr)
3137 {
3138 	struct inode *inode;
3139 
3140 	inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3141 	if (!inode)
3142 		goto out;
3143 
3144 	inode->i_op = &proc_tgid_base_inode_operations;
3145 	inode->i_fop = &proc_tgid_base_operations;
3146 	inode->i_flags|=S_IMMUTABLE;
3147 
3148 	set_nlink(inode, nlink_tgid);
3149 
3150 	d_set_d_op(dentry, &pid_dentry_operations);
3151 
3152 	d_add(dentry, inode);
3153 	/* Close the race of the process dying before we return the dentry */
3154 	if (pid_revalidate(dentry, 0))
3155 		return 0;
3156 out:
3157 	return -ENOENT;
3158 }
3159 
3160 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3161 {
3162 	int result = -ENOENT;
3163 	struct task_struct *task;
3164 	unsigned tgid;
3165 	struct pid_namespace *ns;
3166 
3167 	tgid = name_to_int(&dentry->d_name);
3168 	if (tgid == ~0U)
3169 		goto out;
3170 
3171 	ns = dentry->d_sb->s_fs_info;
3172 	rcu_read_lock();
3173 	task = find_task_by_pid_ns(tgid, ns);
3174 	if (task)
3175 		get_task_struct(task);
3176 	rcu_read_unlock();
3177 	if (!task)
3178 		goto out;
3179 
3180 	result = proc_pid_instantiate(dir, dentry, task, NULL);
3181 	put_task_struct(task);
3182 out:
3183 	return ERR_PTR(result);
3184 }
3185 
3186 /*
3187  * Find the first task with tgid >= tgid
3188  *
3189  */
3190 struct tgid_iter {
3191 	unsigned int tgid;
3192 	struct task_struct *task;
3193 };
3194 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3195 {
3196 	struct pid *pid;
3197 
3198 	if (iter.task)
3199 		put_task_struct(iter.task);
3200 	rcu_read_lock();
3201 retry:
3202 	iter.task = NULL;
3203 	pid = find_ge_pid(iter.tgid, ns);
3204 	if (pid) {
3205 		iter.tgid = pid_nr_ns(pid, ns);
3206 		iter.task = pid_task(pid, PIDTYPE_PID);
3207 		/* What we to know is if the pid we have find is the
3208 		 * pid of a thread_group_leader.  Testing for task
3209 		 * being a thread_group_leader is the obvious thing
3210 		 * todo but there is a window when it fails, due to
3211 		 * the pid transfer logic in de_thread.
3212 		 *
3213 		 * So we perform the straight forward test of seeing
3214 		 * if the pid we have found is the pid of a thread
3215 		 * group leader, and don't worry if the task we have
3216 		 * found doesn't happen to be a thread group leader.
3217 		 * As we don't care in the case of readdir.
3218 		 */
3219 		if (!iter.task || !has_group_leader_pid(iter.task)) {
3220 			iter.tgid += 1;
3221 			goto retry;
3222 		}
3223 		get_task_struct(iter.task);
3224 	}
3225 	rcu_read_unlock();
3226 	return iter;
3227 }
3228 
3229 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3230 
3231 /* for the /proc/ directory itself, after non-process stuff has been done */
3232 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3233 {
3234 	struct tgid_iter iter;
3235 	struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
3236 	loff_t pos = ctx->pos;
3237 
3238 	if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3239 		return 0;
3240 
3241 	if (pos == TGID_OFFSET - 2) {
3242 		struct inode *inode = d_inode(ns->proc_self);
3243 		if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3244 			return 0;
3245 		ctx->pos = pos = pos + 1;
3246 	}
3247 	if (pos == TGID_OFFSET - 1) {
3248 		struct inode *inode = d_inode(ns->proc_thread_self);
3249 		if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3250 			return 0;
3251 		ctx->pos = pos = pos + 1;
3252 	}
3253 	iter.tgid = pos - TGID_OFFSET;
3254 	iter.task = NULL;
3255 	for (iter = next_tgid(ns, iter);
3256 	     iter.task;
3257 	     iter.tgid += 1, iter = next_tgid(ns, iter)) {
3258 		char name[10 + 1];
3259 		int len;
3260 
3261 		cond_resched();
3262 		if (!has_pid_permissions(ns, iter.task, HIDEPID_INVISIBLE))
3263 			continue;
3264 
3265 		len = snprintf(name, sizeof(name), "%u", iter.tgid);
3266 		ctx->pos = iter.tgid + TGID_OFFSET;
3267 		if (!proc_fill_cache(file, ctx, name, len,
3268 				     proc_pid_instantiate, iter.task, NULL)) {
3269 			put_task_struct(iter.task);
3270 			return 0;
3271 		}
3272 	}
3273 	ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3274 	return 0;
3275 }
3276 
3277 /*
3278  * proc_tid_comm_permission is a special permission function exclusively
3279  * used for the node /proc/<pid>/task/<tid>/comm.
3280  * It bypasses generic permission checks in the case where a task of the same
3281  * task group attempts to access the node.
3282  * The rationale behind this is that glibc and bionic access this node for
3283  * cross thread naming (pthread_set/getname_np(!self)). However, if
3284  * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3285  * which locks out the cross thread naming implementation.
3286  * This function makes sure that the node is always accessible for members of
3287  * same thread group.
3288  */
3289 static int proc_tid_comm_permission(struct inode *inode, int mask)
3290 {
3291 	bool is_same_tgroup;
3292 	struct task_struct *task;
3293 
3294 	task = get_proc_task(inode);
3295 	if (!task)
3296 		return -ESRCH;
3297 	is_same_tgroup = same_thread_group(current, task);
3298 	put_task_struct(task);
3299 
3300 	if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3301 		/* This file (/proc/<pid>/task/<tid>/comm) can always be
3302 		 * read or written by the members of the corresponding
3303 		 * thread group.
3304 		 */
3305 		return 0;
3306 	}
3307 
3308 	return generic_permission(inode, mask);
3309 }
3310 
3311 static const struct inode_operations proc_tid_comm_inode_operations = {
3312 		.permission = proc_tid_comm_permission,
3313 };
3314 
3315 /*
3316  * Tasks
3317  */
3318 static const struct pid_entry tid_base_stuff[] = {
3319 	DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3320 	DIR("fdinfo",    S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3321 	DIR("ns",	 S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3322 #ifdef CONFIG_NET
3323 	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3324 #endif
3325 	REG("environ",   S_IRUSR, proc_environ_operations),
3326 	REG("auxv",      S_IRUSR, proc_auxv_operations),
3327 	ONE("status",    S_IRUGO, proc_pid_status),
3328 	ONE("personality", S_IRUSR, proc_pid_personality),
3329 	ONE("limits",	 S_IRUGO, proc_pid_limits),
3330 #ifdef CONFIG_SCHED_DEBUG
3331 	REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3332 #endif
3333 	NOD("comm",      S_IFREG|S_IRUGO|S_IWUSR,
3334 			 &proc_tid_comm_inode_operations,
3335 			 &proc_pid_set_comm_operations, {}),
3336 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3337 	ONE("syscall",   S_IRUSR, proc_pid_syscall),
3338 #endif
3339 	REG("cmdline",   S_IRUGO, proc_pid_cmdline_ops),
3340 	ONE("stat",      S_IRUGO, proc_tid_stat),
3341 	ONE("statm",     S_IRUGO, proc_pid_statm),
3342 	REG("maps",      S_IRUGO, proc_tid_maps_operations),
3343 #ifdef CONFIG_PROC_CHILDREN
3344 	REG("children",  S_IRUGO, proc_tid_children_operations),
3345 #endif
3346 #ifdef CONFIG_NUMA
3347 	REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3348 #endif
3349 	REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
3350 	LNK("cwd",       proc_cwd_link),
3351 	LNK("root",      proc_root_link),
3352 	LNK("exe",       proc_exe_link),
3353 	REG("mounts",    S_IRUGO, proc_mounts_operations),
3354 	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3355 #ifdef CONFIG_PROC_PAGE_MONITOR
3356 	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3357 	REG("smaps",     S_IRUGO, proc_tid_smaps_operations),
3358 	REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3359 	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3360 #endif
3361 #ifdef CONFIG_SECURITY
3362 	DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3363 #endif
3364 #ifdef CONFIG_KALLSYMS
3365 	ONE("wchan",     S_IRUGO, proc_pid_wchan),
3366 #endif
3367 #ifdef CONFIG_STACKTRACE
3368 	ONE("stack",      S_IRUSR, proc_pid_stack),
3369 #endif
3370 #ifdef CONFIG_SCHED_INFO
3371 	ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3372 #endif
3373 #ifdef CONFIG_LATENCYTOP
3374 	REG("latency",  S_IRUGO, proc_lstats_operations),
3375 #endif
3376 #ifdef CONFIG_PROC_PID_CPUSET
3377 	ONE("cpuset",    S_IRUGO, proc_cpuset_show),
3378 #endif
3379 #ifdef CONFIG_CGROUPS
3380 	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3381 #endif
3382 	ONE("oom_score", S_IRUGO, proc_oom_score),
3383 	REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3384 	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3385 #ifdef CONFIG_AUDITSYSCALL
3386 	REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
3387 	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3388 #endif
3389 #ifdef CONFIG_FAULT_INJECTION
3390 	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3391 	REG("fail-nth", 0644, proc_fail_nth_operations),
3392 #endif
3393 #ifdef CONFIG_TASK_IO_ACCOUNTING
3394 	ONE("io",	S_IRUSR, proc_tid_io_accounting),
3395 #endif
3396 #ifdef CONFIG_HARDWALL
3397 	ONE("hardwall",   S_IRUGO, proc_pid_hardwall),
3398 #endif
3399 #ifdef CONFIG_USER_NS
3400 	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3401 	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3402 	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3403 	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3404 #endif
3405 #ifdef CONFIG_LIVEPATCH
3406 	ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3407 #endif
3408 };
3409 
3410 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3411 {
3412 	return proc_pident_readdir(file, ctx,
3413 				   tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3414 }
3415 
3416 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3417 {
3418 	return proc_pident_lookup(dir, dentry,
3419 				  tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3420 }
3421 
3422 static const struct file_operations proc_tid_base_operations = {
3423 	.read		= generic_read_dir,
3424 	.iterate_shared	= proc_tid_base_readdir,
3425 	.llseek		= generic_file_llseek,
3426 };
3427 
3428 static const struct inode_operations proc_tid_base_inode_operations = {
3429 	.lookup		= proc_tid_base_lookup,
3430 	.getattr	= pid_getattr,
3431 	.setattr	= proc_setattr,
3432 };
3433 
3434 static int proc_task_instantiate(struct inode *dir,
3435 	struct dentry *dentry, struct task_struct *task, const void *ptr)
3436 {
3437 	struct inode *inode;
3438 	inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3439 
3440 	if (!inode)
3441 		goto out;
3442 	inode->i_op = &proc_tid_base_inode_operations;
3443 	inode->i_fop = &proc_tid_base_operations;
3444 	inode->i_flags|=S_IMMUTABLE;
3445 
3446 	set_nlink(inode, nlink_tid);
3447 
3448 	d_set_d_op(dentry, &pid_dentry_operations);
3449 
3450 	d_add(dentry, inode);
3451 	/* Close the race of the process dying before we return the dentry */
3452 	if (pid_revalidate(dentry, 0))
3453 		return 0;
3454 out:
3455 	return -ENOENT;
3456 }
3457 
3458 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3459 {
3460 	int result = -ENOENT;
3461 	struct task_struct *task;
3462 	struct task_struct *leader = get_proc_task(dir);
3463 	unsigned tid;
3464 	struct pid_namespace *ns;
3465 
3466 	if (!leader)
3467 		goto out_no_task;
3468 
3469 	tid = name_to_int(&dentry->d_name);
3470 	if (tid == ~0U)
3471 		goto out;
3472 
3473 	ns = dentry->d_sb->s_fs_info;
3474 	rcu_read_lock();
3475 	task = find_task_by_pid_ns(tid, ns);
3476 	if (task)
3477 		get_task_struct(task);
3478 	rcu_read_unlock();
3479 	if (!task)
3480 		goto out;
3481 	if (!same_thread_group(leader, task))
3482 		goto out_drop_task;
3483 
3484 	result = proc_task_instantiate(dir, dentry, task, NULL);
3485 out_drop_task:
3486 	put_task_struct(task);
3487 out:
3488 	put_task_struct(leader);
3489 out_no_task:
3490 	return ERR_PTR(result);
3491 }
3492 
3493 /*
3494  * Find the first tid of a thread group to return to user space.
3495  *
3496  * Usually this is just the thread group leader, but if the users
3497  * buffer was too small or there was a seek into the middle of the
3498  * directory we have more work todo.
3499  *
3500  * In the case of a short read we start with find_task_by_pid.
3501  *
3502  * In the case of a seek we start with the leader and walk nr
3503  * threads past it.
3504  */
3505 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3506 					struct pid_namespace *ns)
3507 {
3508 	struct task_struct *pos, *task;
3509 	unsigned long nr = f_pos;
3510 
3511 	if (nr != f_pos)	/* 32bit overflow? */
3512 		return NULL;
3513 
3514 	rcu_read_lock();
3515 	task = pid_task(pid, PIDTYPE_PID);
3516 	if (!task)
3517 		goto fail;
3518 
3519 	/* Attempt to start with the tid of a thread */
3520 	if (tid && nr) {
3521 		pos = find_task_by_pid_ns(tid, ns);
3522 		if (pos && same_thread_group(pos, task))
3523 			goto found;
3524 	}
3525 
3526 	/* If nr exceeds the number of threads there is nothing todo */
3527 	if (nr >= get_nr_threads(task))
3528 		goto fail;
3529 
3530 	/* If we haven't found our starting place yet start
3531 	 * with the leader and walk nr threads forward.
3532 	 */
3533 	pos = task = task->group_leader;
3534 	do {
3535 		if (!nr--)
3536 			goto found;
3537 	} while_each_thread(task, pos);
3538 fail:
3539 	pos = NULL;
3540 	goto out;
3541 found:
3542 	get_task_struct(pos);
3543 out:
3544 	rcu_read_unlock();
3545 	return pos;
3546 }
3547 
3548 /*
3549  * Find the next thread in the thread list.
3550  * Return NULL if there is an error or no next thread.
3551  *
3552  * The reference to the input task_struct is released.
3553  */
3554 static struct task_struct *next_tid(struct task_struct *start)
3555 {
3556 	struct task_struct *pos = NULL;
3557 	rcu_read_lock();
3558 	if (pid_alive(start)) {
3559 		pos = next_thread(start);
3560 		if (thread_group_leader(pos))
3561 			pos = NULL;
3562 		else
3563 			get_task_struct(pos);
3564 	}
3565 	rcu_read_unlock();
3566 	put_task_struct(start);
3567 	return pos;
3568 }
3569 
3570 /* for the /proc/TGID/task/ directories */
3571 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3572 {
3573 	struct inode *inode = file_inode(file);
3574 	struct task_struct *task;
3575 	struct pid_namespace *ns;
3576 	int tid;
3577 
3578 	if (proc_inode_is_dead(inode))
3579 		return -ENOENT;
3580 
3581 	if (!dir_emit_dots(file, ctx))
3582 		return 0;
3583 
3584 	/* f_version caches the tgid value that the last readdir call couldn't
3585 	 * return. lseek aka telldir automagically resets f_version to 0.
3586 	 */
3587 	ns = inode->i_sb->s_fs_info;
3588 	tid = (int)file->f_version;
3589 	file->f_version = 0;
3590 	for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3591 	     task;
3592 	     task = next_tid(task), ctx->pos++) {
3593 		char name[10 + 1];
3594 		int len;
3595 		tid = task_pid_nr_ns(task, ns);
3596 		len = snprintf(name, sizeof(name), "%u", tid);
3597 		if (!proc_fill_cache(file, ctx, name, len,
3598 				proc_task_instantiate, task, NULL)) {
3599 			/* returning this tgid failed, save it as the first
3600 			 * pid for the next readir call */
3601 			file->f_version = (u64)tid;
3602 			put_task_struct(task);
3603 			break;
3604 		}
3605 	}
3606 
3607 	return 0;
3608 }
3609 
3610 static int proc_task_getattr(const struct path *path, struct kstat *stat,
3611 			     u32 request_mask, unsigned int query_flags)
3612 {
3613 	struct inode *inode = d_inode(path->dentry);
3614 	struct task_struct *p = get_proc_task(inode);
3615 	generic_fillattr(inode, stat);
3616 
3617 	if (p) {
3618 		stat->nlink += get_nr_threads(p);
3619 		put_task_struct(p);
3620 	}
3621 
3622 	return 0;
3623 }
3624 
3625 static const struct inode_operations proc_task_inode_operations = {
3626 	.lookup		= proc_task_lookup,
3627 	.getattr	= proc_task_getattr,
3628 	.setattr	= proc_setattr,
3629 	.permission	= proc_pid_permission,
3630 };
3631 
3632 static const struct file_operations proc_task_operations = {
3633 	.read		= generic_read_dir,
3634 	.iterate_shared	= proc_task_readdir,
3635 	.llseek		= generic_file_llseek,
3636 };
3637 
3638 void __init set_proc_pid_nlink(void)
3639 {
3640 	nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3641 	nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3642 }
3643