xref: /openbmc/linux/fs/proc/base.c (revision 5968cecedd7a09f23e9fcb5f9fb4e893712f35ba)
1 /*
2  *  linux/fs/proc/base.c
3  *
4  *  Copyright (C) 1991, 1992 Linus Torvalds
5  *
6  *  proc base directory handling functions
7  *
8  *  1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9  *  Instead of using magical inumbers to determine the kind of object
10  *  we allocate and fill in-core inodes upon lookup. They don't even
11  *  go into icache. We cache the reference to task_struct upon lookup too.
12  *  Eventually it should become a filesystem in its own. We don't use the
13  *  rest of procfs anymore.
14  *
15  *
16  *  Changelog:
17  *  17-Jan-2005
18  *  Allan Bezerra
19  *  Bruna Moreira <bruna.moreira@indt.org.br>
20  *  Edjard Mota <edjard.mota@indt.org.br>
21  *  Ilias Biris <ilias.biris@indt.org.br>
22  *  Mauricio Lin <mauricio.lin@indt.org.br>
23  *
24  *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
25  *
26  *  A new process specific entry (smaps) included in /proc. It shows the
27  *  size of rss for each memory area. The maps entry lacks information
28  *  about physical memory size (rss) for each mapped file, i.e.,
29  *  rss information for executables and library files.
30  *  This additional information is useful for any tools that need to know
31  *  about physical memory consumption for a process specific library.
32  *
33  *  Changelog:
34  *  21-Feb-2005
35  *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36  *  Pud inclusion in the page table walking.
37  *
38  *  ChangeLog:
39  *  10-Mar-2005
40  *  10LE Instituto Nokia de Tecnologia - INdT:
41  *  A better way to walks through the page table as suggested by Hugh Dickins.
42  *
43  *  Simo Piiroinen <simo.piiroinen@nokia.com>:
44  *  Smaps information related to shared, private, clean and dirty pages.
45  *
46  *  Paul Mundt <paul.mundt@nokia.com>:
47  *  Overall revision about smaps.
48  */
49 
50 #include <asm/uaccess.h>
51 
52 #include <linux/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/task_io_accounting_ops.h>
57 #include <linux/init.h>
58 #include <linux/capability.h>
59 #include <linux/file.h>
60 #include <linux/fdtable.h>
61 #include <linux/string.h>
62 #include <linux/seq_file.h>
63 #include <linux/namei.h>
64 #include <linux/mnt_namespace.h>
65 #include <linux/mm.h>
66 #include <linux/swap.h>
67 #include <linux/rcupdate.h>
68 #include <linux/kallsyms.h>
69 #include <linux/stacktrace.h>
70 #include <linux/resource.h>
71 #include <linux/module.h>
72 #include <linux/mount.h>
73 #include <linux/security.h>
74 #include <linux/ptrace.h>
75 #include <linux/tracehook.h>
76 #include <linux/printk.h>
77 #include <linux/cgroup.h>
78 #include <linux/cpuset.h>
79 #include <linux/audit.h>
80 #include <linux/poll.h>
81 #include <linux/nsproxy.h>
82 #include <linux/oom.h>
83 #include <linux/elf.h>
84 #include <linux/pid_namespace.h>
85 #include <linux/user_namespace.h>
86 #include <linux/fs_struct.h>
87 #include <linux/slab.h>
88 #include <linux/flex_array.h>
89 #include <linux/posix-timers.h>
90 #ifdef CONFIG_HARDWALL
91 #include <asm/hardwall.h>
92 #endif
93 #include <trace/events/oom.h>
94 #include "internal.h"
95 #include "fd.h"
96 
97 /* NOTE:
98  *	Implementing inode permission operations in /proc is almost
99  *	certainly an error.  Permission checks need to happen during
100  *	each system call not at open time.  The reason is that most of
101  *	what we wish to check for permissions in /proc varies at runtime.
102  *
103  *	The classic example of a problem is opening file descriptors
104  *	in /proc for a task before it execs a suid executable.
105  */
106 
107 struct pid_entry {
108 	const char *name;
109 	int len;
110 	umode_t mode;
111 	const struct inode_operations *iop;
112 	const struct file_operations *fop;
113 	union proc_op op;
114 };
115 
116 #define NOD(NAME, MODE, IOP, FOP, OP) {			\
117 	.name = (NAME),					\
118 	.len  = sizeof(NAME) - 1,			\
119 	.mode = MODE,					\
120 	.iop  = IOP,					\
121 	.fop  = FOP,					\
122 	.op   = OP,					\
123 }
124 
125 #define DIR(NAME, MODE, iops, fops)	\
126 	NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
127 #define LNK(NAME, get_link)					\
128 	NOD(NAME, (S_IFLNK|S_IRWXUGO),				\
129 		&proc_pid_link_inode_operations, NULL,		\
130 		{ .proc_get_link = get_link } )
131 #define REG(NAME, MODE, fops)				\
132 	NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
133 #define ONE(NAME, MODE, show)				\
134 	NOD(NAME, (S_IFREG|(MODE)), 			\
135 		NULL, &proc_single_file_operations,	\
136 		{ .proc_show = show } )
137 
138 /*
139  * Count the number of hardlinks for the pid_entry table, excluding the .
140  * and .. links.
141  */
142 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
143 	unsigned int n)
144 {
145 	unsigned int i;
146 	unsigned int count;
147 
148 	count = 0;
149 	for (i = 0; i < n; ++i) {
150 		if (S_ISDIR(entries[i].mode))
151 			++count;
152 	}
153 
154 	return count;
155 }
156 
157 static int get_task_root(struct task_struct *task, struct path *root)
158 {
159 	int result = -ENOENT;
160 
161 	task_lock(task);
162 	if (task->fs) {
163 		get_fs_root(task->fs, root);
164 		result = 0;
165 	}
166 	task_unlock(task);
167 	return result;
168 }
169 
170 static int proc_cwd_link(struct dentry *dentry, struct path *path)
171 {
172 	struct task_struct *task = get_proc_task(d_inode(dentry));
173 	int result = -ENOENT;
174 
175 	if (task) {
176 		task_lock(task);
177 		if (task->fs) {
178 			get_fs_pwd(task->fs, path);
179 			result = 0;
180 		}
181 		task_unlock(task);
182 		put_task_struct(task);
183 	}
184 	return result;
185 }
186 
187 static int proc_root_link(struct dentry *dentry, struct path *path)
188 {
189 	struct task_struct *task = get_proc_task(d_inode(dentry));
190 	int result = -ENOENT;
191 
192 	if (task) {
193 		result = get_task_root(task, path);
194 		put_task_struct(task);
195 	}
196 	return result;
197 }
198 
199 static int proc_pid_cmdline(struct seq_file *m, struct pid_namespace *ns,
200 			    struct pid *pid, struct task_struct *task)
201 {
202 	/*
203 	 * Rely on struct seq_operations::show() being called once
204 	 * per internal buffer allocation. See single_open(), traverse().
205 	 */
206 	BUG_ON(m->size < PAGE_SIZE);
207 	m->count += get_cmdline(task, m->buf, PAGE_SIZE);
208 	return 0;
209 }
210 
211 static int proc_pid_auxv(struct seq_file *m, struct pid_namespace *ns,
212 			 struct pid *pid, struct task_struct *task)
213 {
214 	struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
215 	if (mm && !IS_ERR(mm)) {
216 		unsigned int nwords = 0;
217 		do {
218 			nwords += 2;
219 		} while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
220 		seq_write(m, mm->saved_auxv, nwords * sizeof(mm->saved_auxv[0]));
221 		mmput(mm);
222 		return 0;
223 	} else
224 		return PTR_ERR(mm);
225 }
226 
227 
228 #ifdef CONFIG_KALLSYMS
229 /*
230  * Provides a wchan file via kallsyms in a proper one-value-per-file format.
231  * Returns the resolved symbol.  If that fails, simply return the address.
232  */
233 static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
234 			  struct pid *pid, struct task_struct *task)
235 {
236 	unsigned long wchan;
237 	char symname[KSYM_NAME_LEN];
238 
239 	wchan = get_wchan(task);
240 
241 	if (lookup_symbol_name(wchan, symname) < 0) {
242 		if (!ptrace_may_access(task, PTRACE_MODE_READ))
243 			return 0;
244 		seq_printf(m, "%lu", wchan);
245 	} else {
246 		seq_printf(m, "%s", symname);
247 	}
248 
249 	return 0;
250 }
251 #endif /* CONFIG_KALLSYMS */
252 
253 static int lock_trace(struct task_struct *task)
254 {
255 	int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
256 	if (err)
257 		return err;
258 	if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
259 		mutex_unlock(&task->signal->cred_guard_mutex);
260 		return -EPERM;
261 	}
262 	return 0;
263 }
264 
265 static void unlock_trace(struct task_struct *task)
266 {
267 	mutex_unlock(&task->signal->cred_guard_mutex);
268 }
269 
270 #ifdef CONFIG_STACKTRACE
271 
272 #define MAX_STACK_TRACE_DEPTH	64
273 
274 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
275 			  struct pid *pid, struct task_struct *task)
276 {
277 	struct stack_trace trace;
278 	unsigned long *entries;
279 	int err;
280 	int i;
281 
282 	entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
283 	if (!entries)
284 		return -ENOMEM;
285 
286 	trace.nr_entries	= 0;
287 	trace.max_entries	= MAX_STACK_TRACE_DEPTH;
288 	trace.entries		= entries;
289 	trace.skip		= 0;
290 
291 	err = lock_trace(task);
292 	if (!err) {
293 		save_stack_trace_tsk(task, &trace);
294 
295 		for (i = 0; i < trace.nr_entries; i++) {
296 			seq_printf(m, "[<%pK>] %pS\n",
297 				   (void *)entries[i], (void *)entries[i]);
298 		}
299 		unlock_trace(task);
300 	}
301 	kfree(entries);
302 
303 	return err;
304 }
305 #endif
306 
307 #ifdef CONFIG_SCHED_INFO
308 /*
309  * Provides /proc/PID/schedstat
310  */
311 static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
312 			      struct pid *pid, struct task_struct *task)
313 {
314 	if (unlikely(!sched_info_on()))
315 		seq_printf(m, "0 0 0\n");
316 	else
317 		seq_printf(m, "%llu %llu %lu\n",
318 		   (unsigned long long)task->se.sum_exec_runtime,
319 		   (unsigned long long)task->sched_info.run_delay,
320 		   task->sched_info.pcount);
321 
322 	return 0;
323 }
324 #endif
325 
326 #ifdef CONFIG_LATENCYTOP
327 static int lstats_show_proc(struct seq_file *m, void *v)
328 {
329 	int i;
330 	struct inode *inode = m->private;
331 	struct task_struct *task = get_proc_task(inode);
332 
333 	if (!task)
334 		return -ESRCH;
335 	seq_puts(m, "Latency Top version : v0.1\n");
336 	for (i = 0; i < 32; i++) {
337 		struct latency_record *lr = &task->latency_record[i];
338 		if (lr->backtrace[0]) {
339 			int q;
340 			seq_printf(m, "%i %li %li",
341 				   lr->count, lr->time, lr->max);
342 			for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
343 				unsigned long bt = lr->backtrace[q];
344 				if (!bt)
345 					break;
346 				if (bt == ULONG_MAX)
347 					break;
348 				seq_printf(m, " %ps", (void *)bt);
349 			}
350 			seq_putc(m, '\n');
351 		}
352 
353 	}
354 	put_task_struct(task);
355 	return 0;
356 }
357 
358 static int lstats_open(struct inode *inode, struct file *file)
359 {
360 	return single_open(file, lstats_show_proc, inode);
361 }
362 
363 static ssize_t lstats_write(struct file *file, const char __user *buf,
364 			    size_t count, loff_t *offs)
365 {
366 	struct task_struct *task = get_proc_task(file_inode(file));
367 
368 	if (!task)
369 		return -ESRCH;
370 	clear_all_latency_tracing(task);
371 	put_task_struct(task);
372 
373 	return count;
374 }
375 
376 static const struct file_operations proc_lstats_operations = {
377 	.open		= lstats_open,
378 	.read		= seq_read,
379 	.write		= lstats_write,
380 	.llseek		= seq_lseek,
381 	.release	= single_release,
382 };
383 
384 #endif
385 
386 static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
387 			  struct pid *pid, struct task_struct *task)
388 {
389 	unsigned long totalpages = totalram_pages + total_swap_pages;
390 	unsigned long points = 0;
391 
392 	read_lock(&tasklist_lock);
393 	if (pid_alive(task))
394 		points = oom_badness(task, NULL, NULL, totalpages) *
395 						1000 / totalpages;
396 	read_unlock(&tasklist_lock);
397 	seq_printf(m, "%lu\n", points);
398 
399 	return 0;
400 }
401 
402 struct limit_names {
403 	const char *name;
404 	const char *unit;
405 };
406 
407 static const struct limit_names lnames[RLIM_NLIMITS] = {
408 	[RLIMIT_CPU] = {"Max cpu time", "seconds"},
409 	[RLIMIT_FSIZE] = {"Max file size", "bytes"},
410 	[RLIMIT_DATA] = {"Max data size", "bytes"},
411 	[RLIMIT_STACK] = {"Max stack size", "bytes"},
412 	[RLIMIT_CORE] = {"Max core file size", "bytes"},
413 	[RLIMIT_RSS] = {"Max resident set", "bytes"},
414 	[RLIMIT_NPROC] = {"Max processes", "processes"},
415 	[RLIMIT_NOFILE] = {"Max open files", "files"},
416 	[RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
417 	[RLIMIT_AS] = {"Max address space", "bytes"},
418 	[RLIMIT_LOCKS] = {"Max file locks", "locks"},
419 	[RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
420 	[RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
421 	[RLIMIT_NICE] = {"Max nice priority", NULL},
422 	[RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
423 	[RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
424 };
425 
426 /* Display limits for a process */
427 static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
428 			   struct pid *pid, struct task_struct *task)
429 {
430 	unsigned int i;
431 	unsigned long flags;
432 
433 	struct rlimit rlim[RLIM_NLIMITS];
434 
435 	if (!lock_task_sighand(task, &flags))
436 		return 0;
437 	memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
438 	unlock_task_sighand(task, &flags);
439 
440 	/*
441 	 * print the file header
442 	 */
443        seq_printf(m, "%-25s %-20s %-20s %-10s\n",
444 		  "Limit", "Soft Limit", "Hard Limit", "Units");
445 
446 	for (i = 0; i < RLIM_NLIMITS; i++) {
447 		if (rlim[i].rlim_cur == RLIM_INFINITY)
448 			seq_printf(m, "%-25s %-20s ",
449 				   lnames[i].name, "unlimited");
450 		else
451 			seq_printf(m, "%-25s %-20lu ",
452 				   lnames[i].name, rlim[i].rlim_cur);
453 
454 		if (rlim[i].rlim_max == RLIM_INFINITY)
455 			seq_printf(m, "%-20s ", "unlimited");
456 		else
457 			seq_printf(m, "%-20lu ", rlim[i].rlim_max);
458 
459 		if (lnames[i].unit)
460 			seq_printf(m, "%-10s\n", lnames[i].unit);
461 		else
462 			seq_putc(m, '\n');
463 	}
464 
465 	return 0;
466 }
467 
468 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
469 static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
470 			    struct pid *pid, struct task_struct *task)
471 {
472 	long nr;
473 	unsigned long args[6], sp, pc;
474 	int res;
475 
476 	res = lock_trace(task);
477 	if (res)
478 		return res;
479 
480 	if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
481 		seq_puts(m, "running\n");
482 	else if (nr < 0)
483 		seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
484 	else
485 		seq_printf(m,
486 		       "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
487 		       nr,
488 		       args[0], args[1], args[2], args[3], args[4], args[5],
489 		       sp, pc);
490 	unlock_trace(task);
491 
492 	return 0;
493 }
494 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
495 
496 /************************************************************************/
497 /*                       Here the fs part begins                        */
498 /************************************************************************/
499 
500 /* permission checks */
501 static int proc_fd_access_allowed(struct inode *inode)
502 {
503 	struct task_struct *task;
504 	int allowed = 0;
505 	/* Allow access to a task's file descriptors if it is us or we
506 	 * may use ptrace attach to the process and find out that
507 	 * information.
508 	 */
509 	task = get_proc_task(inode);
510 	if (task) {
511 		allowed = ptrace_may_access(task, PTRACE_MODE_READ);
512 		put_task_struct(task);
513 	}
514 	return allowed;
515 }
516 
517 int proc_setattr(struct dentry *dentry, struct iattr *attr)
518 {
519 	int error;
520 	struct inode *inode = d_inode(dentry);
521 
522 	if (attr->ia_valid & ATTR_MODE)
523 		return -EPERM;
524 
525 	error = inode_change_ok(inode, attr);
526 	if (error)
527 		return error;
528 
529 	setattr_copy(inode, attr);
530 	mark_inode_dirty(inode);
531 	return 0;
532 }
533 
534 /*
535  * May current process learn task's sched/cmdline info (for hide_pid_min=1)
536  * or euid/egid (for hide_pid_min=2)?
537  */
538 static bool has_pid_permissions(struct pid_namespace *pid,
539 				 struct task_struct *task,
540 				 int hide_pid_min)
541 {
542 	if (pid->hide_pid < hide_pid_min)
543 		return true;
544 	if (in_group_p(pid->pid_gid))
545 		return true;
546 	return ptrace_may_access(task, PTRACE_MODE_READ);
547 }
548 
549 
550 static int proc_pid_permission(struct inode *inode, int mask)
551 {
552 	struct pid_namespace *pid = inode->i_sb->s_fs_info;
553 	struct task_struct *task;
554 	bool has_perms;
555 
556 	task = get_proc_task(inode);
557 	if (!task)
558 		return -ESRCH;
559 	has_perms = has_pid_permissions(pid, task, 1);
560 	put_task_struct(task);
561 
562 	if (!has_perms) {
563 		if (pid->hide_pid == 2) {
564 			/*
565 			 * Let's make getdents(), stat(), and open()
566 			 * consistent with each other.  If a process
567 			 * may not stat() a file, it shouldn't be seen
568 			 * in procfs at all.
569 			 */
570 			return -ENOENT;
571 		}
572 
573 		return -EPERM;
574 	}
575 	return generic_permission(inode, mask);
576 }
577 
578 
579 
580 static const struct inode_operations proc_def_inode_operations = {
581 	.setattr	= proc_setattr,
582 };
583 
584 static int proc_single_show(struct seq_file *m, void *v)
585 {
586 	struct inode *inode = m->private;
587 	struct pid_namespace *ns;
588 	struct pid *pid;
589 	struct task_struct *task;
590 	int ret;
591 
592 	ns = inode->i_sb->s_fs_info;
593 	pid = proc_pid(inode);
594 	task = get_pid_task(pid, PIDTYPE_PID);
595 	if (!task)
596 		return -ESRCH;
597 
598 	ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
599 
600 	put_task_struct(task);
601 	return ret;
602 }
603 
604 static int proc_single_open(struct inode *inode, struct file *filp)
605 {
606 	return single_open(filp, proc_single_show, inode);
607 }
608 
609 static const struct file_operations proc_single_file_operations = {
610 	.open		= proc_single_open,
611 	.read		= seq_read,
612 	.llseek		= seq_lseek,
613 	.release	= single_release,
614 };
615 
616 
617 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
618 {
619 	struct task_struct *task = get_proc_task(inode);
620 	struct mm_struct *mm = ERR_PTR(-ESRCH);
621 
622 	if (task) {
623 		mm = mm_access(task, mode);
624 		put_task_struct(task);
625 
626 		if (!IS_ERR_OR_NULL(mm)) {
627 			/* ensure this mm_struct can't be freed */
628 			atomic_inc(&mm->mm_count);
629 			/* but do not pin its memory */
630 			mmput(mm);
631 		}
632 	}
633 
634 	return mm;
635 }
636 
637 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
638 {
639 	struct mm_struct *mm = proc_mem_open(inode, mode);
640 
641 	if (IS_ERR(mm))
642 		return PTR_ERR(mm);
643 
644 	file->private_data = mm;
645 	return 0;
646 }
647 
648 static int mem_open(struct inode *inode, struct file *file)
649 {
650 	int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
651 
652 	/* OK to pass negative loff_t, we can catch out-of-range */
653 	file->f_mode |= FMODE_UNSIGNED_OFFSET;
654 
655 	return ret;
656 }
657 
658 static ssize_t mem_rw(struct file *file, char __user *buf,
659 			size_t count, loff_t *ppos, int write)
660 {
661 	struct mm_struct *mm = file->private_data;
662 	unsigned long addr = *ppos;
663 	ssize_t copied;
664 	char *page;
665 
666 	if (!mm)
667 		return 0;
668 
669 	page = (char *)__get_free_page(GFP_TEMPORARY);
670 	if (!page)
671 		return -ENOMEM;
672 
673 	copied = 0;
674 	if (!atomic_inc_not_zero(&mm->mm_users))
675 		goto free;
676 
677 	while (count > 0) {
678 		int this_len = min_t(int, count, PAGE_SIZE);
679 
680 		if (write && copy_from_user(page, buf, this_len)) {
681 			copied = -EFAULT;
682 			break;
683 		}
684 
685 		this_len = access_remote_vm(mm, addr, page, this_len, write);
686 		if (!this_len) {
687 			if (!copied)
688 				copied = -EIO;
689 			break;
690 		}
691 
692 		if (!write && copy_to_user(buf, page, this_len)) {
693 			copied = -EFAULT;
694 			break;
695 		}
696 
697 		buf += this_len;
698 		addr += this_len;
699 		copied += this_len;
700 		count -= this_len;
701 	}
702 	*ppos = addr;
703 
704 	mmput(mm);
705 free:
706 	free_page((unsigned long) page);
707 	return copied;
708 }
709 
710 static ssize_t mem_read(struct file *file, char __user *buf,
711 			size_t count, loff_t *ppos)
712 {
713 	return mem_rw(file, buf, count, ppos, 0);
714 }
715 
716 static ssize_t mem_write(struct file *file, const char __user *buf,
717 			 size_t count, loff_t *ppos)
718 {
719 	return mem_rw(file, (char __user*)buf, count, ppos, 1);
720 }
721 
722 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
723 {
724 	switch (orig) {
725 	case 0:
726 		file->f_pos = offset;
727 		break;
728 	case 1:
729 		file->f_pos += offset;
730 		break;
731 	default:
732 		return -EINVAL;
733 	}
734 	force_successful_syscall_return();
735 	return file->f_pos;
736 }
737 
738 static int mem_release(struct inode *inode, struct file *file)
739 {
740 	struct mm_struct *mm = file->private_data;
741 	if (mm)
742 		mmdrop(mm);
743 	return 0;
744 }
745 
746 static const struct file_operations proc_mem_operations = {
747 	.llseek		= mem_lseek,
748 	.read		= mem_read,
749 	.write		= mem_write,
750 	.open		= mem_open,
751 	.release	= mem_release,
752 };
753 
754 static int environ_open(struct inode *inode, struct file *file)
755 {
756 	return __mem_open(inode, file, PTRACE_MODE_READ);
757 }
758 
759 static ssize_t environ_read(struct file *file, char __user *buf,
760 			size_t count, loff_t *ppos)
761 {
762 	char *page;
763 	unsigned long src = *ppos;
764 	int ret = 0;
765 	struct mm_struct *mm = file->private_data;
766 
767 	if (!mm)
768 		return 0;
769 
770 	page = (char *)__get_free_page(GFP_TEMPORARY);
771 	if (!page)
772 		return -ENOMEM;
773 
774 	ret = 0;
775 	if (!atomic_inc_not_zero(&mm->mm_users))
776 		goto free;
777 	while (count > 0) {
778 		size_t this_len, max_len;
779 		int retval;
780 
781 		if (src >= (mm->env_end - mm->env_start))
782 			break;
783 
784 		this_len = mm->env_end - (mm->env_start + src);
785 
786 		max_len = min_t(size_t, PAGE_SIZE, count);
787 		this_len = min(max_len, this_len);
788 
789 		retval = access_remote_vm(mm, (mm->env_start + src),
790 			page, this_len, 0);
791 
792 		if (retval <= 0) {
793 			ret = retval;
794 			break;
795 		}
796 
797 		if (copy_to_user(buf, page, retval)) {
798 			ret = -EFAULT;
799 			break;
800 		}
801 
802 		ret += retval;
803 		src += retval;
804 		buf += retval;
805 		count -= retval;
806 	}
807 	*ppos = src;
808 	mmput(mm);
809 
810 free:
811 	free_page((unsigned long) page);
812 	return ret;
813 }
814 
815 static const struct file_operations proc_environ_operations = {
816 	.open		= environ_open,
817 	.read		= environ_read,
818 	.llseek		= generic_file_llseek,
819 	.release	= mem_release,
820 };
821 
822 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
823 			    loff_t *ppos)
824 {
825 	struct task_struct *task = get_proc_task(file_inode(file));
826 	char buffer[PROC_NUMBUF];
827 	int oom_adj = OOM_ADJUST_MIN;
828 	size_t len;
829 	unsigned long flags;
830 
831 	if (!task)
832 		return -ESRCH;
833 	if (lock_task_sighand(task, &flags)) {
834 		if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
835 			oom_adj = OOM_ADJUST_MAX;
836 		else
837 			oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
838 				  OOM_SCORE_ADJ_MAX;
839 		unlock_task_sighand(task, &flags);
840 	}
841 	put_task_struct(task);
842 	len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
843 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
844 }
845 
846 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
847 			     size_t count, loff_t *ppos)
848 {
849 	struct task_struct *task;
850 	char buffer[PROC_NUMBUF];
851 	int oom_adj;
852 	unsigned long flags;
853 	int err;
854 
855 	memset(buffer, 0, sizeof(buffer));
856 	if (count > sizeof(buffer) - 1)
857 		count = sizeof(buffer) - 1;
858 	if (copy_from_user(buffer, buf, count)) {
859 		err = -EFAULT;
860 		goto out;
861 	}
862 
863 	err = kstrtoint(strstrip(buffer), 0, &oom_adj);
864 	if (err)
865 		goto out;
866 	if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
867 	     oom_adj != OOM_DISABLE) {
868 		err = -EINVAL;
869 		goto out;
870 	}
871 
872 	task = get_proc_task(file_inode(file));
873 	if (!task) {
874 		err = -ESRCH;
875 		goto out;
876 	}
877 
878 	task_lock(task);
879 	if (!task->mm) {
880 		err = -EINVAL;
881 		goto err_task_lock;
882 	}
883 
884 	if (!lock_task_sighand(task, &flags)) {
885 		err = -ESRCH;
886 		goto err_task_lock;
887 	}
888 
889 	/*
890 	 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
891 	 * value is always attainable.
892 	 */
893 	if (oom_adj == OOM_ADJUST_MAX)
894 		oom_adj = OOM_SCORE_ADJ_MAX;
895 	else
896 		oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
897 
898 	if (oom_adj < task->signal->oom_score_adj &&
899 	    !capable(CAP_SYS_RESOURCE)) {
900 		err = -EACCES;
901 		goto err_sighand;
902 	}
903 
904 	/*
905 	 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
906 	 * /proc/pid/oom_score_adj instead.
907 	 */
908 	pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
909 		  current->comm, task_pid_nr(current), task_pid_nr(task),
910 		  task_pid_nr(task));
911 
912 	task->signal->oom_score_adj = oom_adj;
913 	trace_oom_score_adj_update(task);
914 err_sighand:
915 	unlock_task_sighand(task, &flags);
916 err_task_lock:
917 	task_unlock(task);
918 	put_task_struct(task);
919 out:
920 	return err < 0 ? err : count;
921 }
922 
923 static const struct file_operations proc_oom_adj_operations = {
924 	.read		= oom_adj_read,
925 	.write		= oom_adj_write,
926 	.llseek		= generic_file_llseek,
927 };
928 
929 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
930 					size_t count, loff_t *ppos)
931 {
932 	struct task_struct *task = get_proc_task(file_inode(file));
933 	char buffer[PROC_NUMBUF];
934 	short oom_score_adj = OOM_SCORE_ADJ_MIN;
935 	unsigned long flags;
936 	size_t len;
937 
938 	if (!task)
939 		return -ESRCH;
940 	if (lock_task_sighand(task, &flags)) {
941 		oom_score_adj = task->signal->oom_score_adj;
942 		unlock_task_sighand(task, &flags);
943 	}
944 	put_task_struct(task);
945 	len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
946 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
947 }
948 
949 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
950 					size_t count, loff_t *ppos)
951 {
952 	struct task_struct *task;
953 	char buffer[PROC_NUMBUF];
954 	unsigned long flags;
955 	int oom_score_adj;
956 	int err;
957 
958 	memset(buffer, 0, sizeof(buffer));
959 	if (count > sizeof(buffer) - 1)
960 		count = sizeof(buffer) - 1;
961 	if (copy_from_user(buffer, buf, count)) {
962 		err = -EFAULT;
963 		goto out;
964 	}
965 
966 	err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
967 	if (err)
968 		goto out;
969 	if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
970 			oom_score_adj > OOM_SCORE_ADJ_MAX) {
971 		err = -EINVAL;
972 		goto out;
973 	}
974 
975 	task = get_proc_task(file_inode(file));
976 	if (!task) {
977 		err = -ESRCH;
978 		goto out;
979 	}
980 
981 	task_lock(task);
982 	if (!task->mm) {
983 		err = -EINVAL;
984 		goto err_task_lock;
985 	}
986 
987 	if (!lock_task_sighand(task, &flags)) {
988 		err = -ESRCH;
989 		goto err_task_lock;
990 	}
991 
992 	if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
993 			!capable(CAP_SYS_RESOURCE)) {
994 		err = -EACCES;
995 		goto err_sighand;
996 	}
997 
998 	task->signal->oom_score_adj = (short)oom_score_adj;
999 	if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1000 		task->signal->oom_score_adj_min = (short)oom_score_adj;
1001 	trace_oom_score_adj_update(task);
1002 
1003 err_sighand:
1004 	unlock_task_sighand(task, &flags);
1005 err_task_lock:
1006 	task_unlock(task);
1007 	put_task_struct(task);
1008 out:
1009 	return err < 0 ? err : count;
1010 }
1011 
1012 static const struct file_operations proc_oom_score_adj_operations = {
1013 	.read		= oom_score_adj_read,
1014 	.write		= oom_score_adj_write,
1015 	.llseek		= default_llseek,
1016 };
1017 
1018 #ifdef CONFIG_AUDITSYSCALL
1019 #define TMPBUFLEN 21
1020 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1021 				  size_t count, loff_t *ppos)
1022 {
1023 	struct inode * inode = file_inode(file);
1024 	struct task_struct *task = get_proc_task(inode);
1025 	ssize_t length;
1026 	char tmpbuf[TMPBUFLEN];
1027 
1028 	if (!task)
1029 		return -ESRCH;
1030 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1031 			   from_kuid(file->f_cred->user_ns,
1032 				     audit_get_loginuid(task)));
1033 	put_task_struct(task);
1034 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1035 }
1036 
1037 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1038 				   size_t count, loff_t *ppos)
1039 {
1040 	struct inode * inode = file_inode(file);
1041 	char *page, *tmp;
1042 	ssize_t length;
1043 	uid_t loginuid;
1044 	kuid_t kloginuid;
1045 
1046 	rcu_read_lock();
1047 	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1048 		rcu_read_unlock();
1049 		return -EPERM;
1050 	}
1051 	rcu_read_unlock();
1052 
1053 	if (count >= PAGE_SIZE)
1054 		count = PAGE_SIZE - 1;
1055 
1056 	if (*ppos != 0) {
1057 		/* No partial writes. */
1058 		return -EINVAL;
1059 	}
1060 	page = (char*)__get_free_page(GFP_TEMPORARY);
1061 	if (!page)
1062 		return -ENOMEM;
1063 	length = -EFAULT;
1064 	if (copy_from_user(page, buf, count))
1065 		goto out_free_page;
1066 
1067 	page[count] = '\0';
1068 	loginuid = simple_strtoul(page, &tmp, 10);
1069 	if (tmp == page) {
1070 		length = -EINVAL;
1071 		goto out_free_page;
1072 
1073 	}
1074 
1075 	/* is userspace tring to explicitly UNSET the loginuid? */
1076 	if (loginuid == AUDIT_UID_UNSET) {
1077 		kloginuid = INVALID_UID;
1078 	} else {
1079 		kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1080 		if (!uid_valid(kloginuid)) {
1081 			length = -EINVAL;
1082 			goto out_free_page;
1083 		}
1084 	}
1085 
1086 	length = audit_set_loginuid(kloginuid);
1087 	if (likely(length == 0))
1088 		length = count;
1089 
1090 out_free_page:
1091 	free_page((unsigned long) page);
1092 	return length;
1093 }
1094 
1095 static const struct file_operations proc_loginuid_operations = {
1096 	.read		= proc_loginuid_read,
1097 	.write		= proc_loginuid_write,
1098 	.llseek		= generic_file_llseek,
1099 };
1100 
1101 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1102 				  size_t count, loff_t *ppos)
1103 {
1104 	struct inode * inode = file_inode(file);
1105 	struct task_struct *task = get_proc_task(inode);
1106 	ssize_t length;
1107 	char tmpbuf[TMPBUFLEN];
1108 
1109 	if (!task)
1110 		return -ESRCH;
1111 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1112 				audit_get_sessionid(task));
1113 	put_task_struct(task);
1114 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1115 }
1116 
1117 static const struct file_operations proc_sessionid_operations = {
1118 	.read		= proc_sessionid_read,
1119 	.llseek		= generic_file_llseek,
1120 };
1121 #endif
1122 
1123 #ifdef CONFIG_FAULT_INJECTION
1124 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1125 				      size_t count, loff_t *ppos)
1126 {
1127 	struct task_struct *task = get_proc_task(file_inode(file));
1128 	char buffer[PROC_NUMBUF];
1129 	size_t len;
1130 	int make_it_fail;
1131 
1132 	if (!task)
1133 		return -ESRCH;
1134 	make_it_fail = task->make_it_fail;
1135 	put_task_struct(task);
1136 
1137 	len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1138 
1139 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1140 }
1141 
1142 static ssize_t proc_fault_inject_write(struct file * file,
1143 			const char __user * buf, size_t count, loff_t *ppos)
1144 {
1145 	struct task_struct *task;
1146 	char buffer[PROC_NUMBUF], *end;
1147 	int make_it_fail;
1148 
1149 	if (!capable(CAP_SYS_RESOURCE))
1150 		return -EPERM;
1151 	memset(buffer, 0, sizeof(buffer));
1152 	if (count > sizeof(buffer) - 1)
1153 		count = sizeof(buffer) - 1;
1154 	if (copy_from_user(buffer, buf, count))
1155 		return -EFAULT;
1156 	make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1157 	if (*end)
1158 		return -EINVAL;
1159 	if (make_it_fail < 0 || make_it_fail > 1)
1160 		return -EINVAL;
1161 
1162 	task = get_proc_task(file_inode(file));
1163 	if (!task)
1164 		return -ESRCH;
1165 	task->make_it_fail = make_it_fail;
1166 	put_task_struct(task);
1167 
1168 	return count;
1169 }
1170 
1171 static const struct file_operations proc_fault_inject_operations = {
1172 	.read		= proc_fault_inject_read,
1173 	.write		= proc_fault_inject_write,
1174 	.llseek		= generic_file_llseek,
1175 };
1176 #endif
1177 
1178 
1179 #ifdef CONFIG_SCHED_DEBUG
1180 /*
1181  * Print out various scheduling related per-task fields:
1182  */
1183 static int sched_show(struct seq_file *m, void *v)
1184 {
1185 	struct inode *inode = m->private;
1186 	struct task_struct *p;
1187 
1188 	p = get_proc_task(inode);
1189 	if (!p)
1190 		return -ESRCH;
1191 	proc_sched_show_task(p, m);
1192 
1193 	put_task_struct(p);
1194 
1195 	return 0;
1196 }
1197 
1198 static ssize_t
1199 sched_write(struct file *file, const char __user *buf,
1200 	    size_t count, loff_t *offset)
1201 {
1202 	struct inode *inode = file_inode(file);
1203 	struct task_struct *p;
1204 
1205 	p = get_proc_task(inode);
1206 	if (!p)
1207 		return -ESRCH;
1208 	proc_sched_set_task(p);
1209 
1210 	put_task_struct(p);
1211 
1212 	return count;
1213 }
1214 
1215 static int sched_open(struct inode *inode, struct file *filp)
1216 {
1217 	return single_open(filp, sched_show, inode);
1218 }
1219 
1220 static const struct file_operations proc_pid_sched_operations = {
1221 	.open		= sched_open,
1222 	.read		= seq_read,
1223 	.write		= sched_write,
1224 	.llseek		= seq_lseek,
1225 	.release	= single_release,
1226 };
1227 
1228 #endif
1229 
1230 #ifdef CONFIG_SCHED_AUTOGROUP
1231 /*
1232  * Print out autogroup related information:
1233  */
1234 static int sched_autogroup_show(struct seq_file *m, void *v)
1235 {
1236 	struct inode *inode = m->private;
1237 	struct task_struct *p;
1238 
1239 	p = get_proc_task(inode);
1240 	if (!p)
1241 		return -ESRCH;
1242 	proc_sched_autogroup_show_task(p, m);
1243 
1244 	put_task_struct(p);
1245 
1246 	return 0;
1247 }
1248 
1249 static ssize_t
1250 sched_autogroup_write(struct file *file, const char __user *buf,
1251 	    size_t count, loff_t *offset)
1252 {
1253 	struct inode *inode = file_inode(file);
1254 	struct task_struct *p;
1255 	char buffer[PROC_NUMBUF];
1256 	int nice;
1257 	int err;
1258 
1259 	memset(buffer, 0, sizeof(buffer));
1260 	if (count > sizeof(buffer) - 1)
1261 		count = sizeof(buffer) - 1;
1262 	if (copy_from_user(buffer, buf, count))
1263 		return -EFAULT;
1264 
1265 	err = kstrtoint(strstrip(buffer), 0, &nice);
1266 	if (err < 0)
1267 		return err;
1268 
1269 	p = get_proc_task(inode);
1270 	if (!p)
1271 		return -ESRCH;
1272 
1273 	err = proc_sched_autogroup_set_nice(p, nice);
1274 	if (err)
1275 		count = err;
1276 
1277 	put_task_struct(p);
1278 
1279 	return count;
1280 }
1281 
1282 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1283 {
1284 	int ret;
1285 
1286 	ret = single_open(filp, sched_autogroup_show, NULL);
1287 	if (!ret) {
1288 		struct seq_file *m = filp->private_data;
1289 
1290 		m->private = inode;
1291 	}
1292 	return ret;
1293 }
1294 
1295 static const struct file_operations proc_pid_sched_autogroup_operations = {
1296 	.open		= sched_autogroup_open,
1297 	.read		= seq_read,
1298 	.write		= sched_autogroup_write,
1299 	.llseek		= seq_lseek,
1300 	.release	= single_release,
1301 };
1302 
1303 #endif /* CONFIG_SCHED_AUTOGROUP */
1304 
1305 static ssize_t comm_write(struct file *file, const char __user *buf,
1306 				size_t count, loff_t *offset)
1307 {
1308 	struct inode *inode = file_inode(file);
1309 	struct task_struct *p;
1310 	char buffer[TASK_COMM_LEN];
1311 	const size_t maxlen = sizeof(buffer) - 1;
1312 
1313 	memset(buffer, 0, sizeof(buffer));
1314 	if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1315 		return -EFAULT;
1316 
1317 	p = get_proc_task(inode);
1318 	if (!p)
1319 		return -ESRCH;
1320 
1321 	if (same_thread_group(current, p))
1322 		set_task_comm(p, buffer);
1323 	else
1324 		count = -EINVAL;
1325 
1326 	put_task_struct(p);
1327 
1328 	return count;
1329 }
1330 
1331 static int comm_show(struct seq_file *m, void *v)
1332 {
1333 	struct inode *inode = m->private;
1334 	struct task_struct *p;
1335 
1336 	p = get_proc_task(inode);
1337 	if (!p)
1338 		return -ESRCH;
1339 
1340 	task_lock(p);
1341 	seq_printf(m, "%s\n", p->comm);
1342 	task_unlock(p);
1343 
1344 	put_task_struct(p);
1345 
1346 	return 0;
1347 }
1348 
1349 static int comm_open(struct inode *inode, struct file *filp)
1350 {
1351 	return single_open(filp, comm_show, inode);
1352 }
1353 
1354 static const struct file_operations proc_pid_set_comm_operations = {
1355 	.open		= comm_open,
1356 	.read		= seq_read,
1357 	.write		= comm_write,
1358 	.llseek		= seq_lseek,
1359 	.release	= single_release,
1360 };
1361 
1362 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1363 {
1364 	struct task_struct *task;
1365 	struct mm_struct *mm;
1366 	struct file *exe_file;
1367 
1368 	task = get_proc_task(d_inode(dentry));
1369 	if (!task)
1370 		return -ENOENT;
1371 	mm = get_task_mm(task);
1372 	put_task_struct(task);
1373 	if (!mm)
1374 		return -ENOENT;
1375 	exe_file = get_mm_exe_file(mm);
1376 	mmput(mm);
1377 	if (exe_file) {
1378 		*exe_path = exe_file->f_path;
1379 		path_get(&exe_file->f_path);
1380 		fput(exe_file);
1381 		return 0;
1382 	} else
1383 		return -ENOENT;
1384 }
1385 
1386 static const char *proc_pid_follow_link(struct dentry *dentry, void **cookie)
1387 {
1388 	struct inode *inode = d_inode(dentry);
1389 	struct path path;
1390 	int error = -EACCES;
1391 
1392 	/* Are we allowed to snoop on the tasks file descriptors? */
1393 	if (!proc_fd_access_allowed(inode))
1394 		goto out;
1395 
1396 	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1397 	if (error)
1398 		goto out;
1399 
1400 	nd_jump_link(&path);
1401 	return NULL;
1402 out:
1403 	return ERR_PTR(error);
1404 }
1405 
1406 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1407 {
1408 	char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1409 	char *pathname;
1410 	int len;
1411 
1412 	if (!tmp)
1413 		return -ENOMEM;
1414 
1415 	pathname = d_path(path, tmp, PAGE_SIZE);
1416 	len = PTR_ERR(pathname);
1417 	if (IS_ERR(pathname))
1418 		goto out;
1419 	len = tmp + PAGE_SIZE - 1 - pathname;
1420 
1421 	if (len > buflen)
1422 		len = buflen;
1423 	if (copy_to_user(buffer, pathname, len))
1424 		len = -EFAULT;
1425  out:
1426 	free_page((unsigned long)tmp);
1427 	return len;
1428 }
1429 
1430 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1431 {
1432 	int error = -EACCES;
1433 	struct inode *inode = d_inode(dentry);
1434 	struct path path;
1435 
1436 	/* Are we allowed to snoop on the tasks file descriptors? */
1437 	if (!proc_fd_access_allowed(inode))
1438 		goto out;
1439 
1440 	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1441 	if (error)
1442 		goto out;
1443 
1444 	error = do_proc_readlink(&path, buffer, buflen);
1445 	path_put(&path);
1446 out:
1447 	return error;
1448 }
1449 
1450 const struct inode_operations proc_pid_link_inode_operations = {
1451 	.readlink	= proc_pid_readlink,
1452 	.follow_link	= proc_pid_follow_link,
1453 	.setattr	= proc_setattr,
1454 };
1455 
1456 
1457 /* building an inode */
1458 
1459 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1460 {
1461 	struct inode * inode;
1462 	struct proc_inode *ei;
1463 	const struct cred *cred;
1464 
1465 	/* We need a new inode */
1466 
1467 	inode = new_inode(sb);
1468 	if (!inode)
1469 		goto out;
1470 
1471 	/* Common stuff */
1472 	ei = PROC_I(inode);
1473 	inode->i_ino = get_next_ino();
1474 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1475 	inode->i_op = &proc_def_inode_operations;
1476 
1477 	/*
1478 	 * grab the reference to task.
1479 	 */
1480 	ei->pid = get_task_pid(task, PIDTYPE_PID);
1481 	if (!ei->pid)
1482 		goto out_unlock;
1483 
1484 	if (task_dumpable(task)) {
1485 		rcu_read_lock();
1486 		cred = __task_cred(task);
1487 		inode->i_uid = cred->euid;
1488 		inode->i_gid = cred->egid;
1489 		rcu_read_unlock();
1490 	}
1491 	security_task_to_inode(task, inode);
1492 
1493 out:
1494 	return inode;
1495 
1496 out_unlock:
1497 	iput(inode);
1498 	return NULL;
1499 }
1500 
1501 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1502 {
1503 	struct inode *inode = d_inode(dentry);
1504 	struct task_struct *task;
1505 	const struct cred *cred;
1506 	struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1507 
1508 	generic_fillattr(inode, stat);
1509 
1510 	rcu_read_lock();
1511 	stat->uid = GLOBAL_ROOT_UID;
1512 	stat->gid = GLOBAL_ROOT_GID;
1513 	task = pid_task(proc_pid(inode), PIDTYPE_PID);
1514 	if (task) {
1515 		if (!has_pid_permissions(pid, task, 2)) {
1516 			rcu_read_unlock();
1517 			/*
1518 			 * This doesn't prevent learning whether PID exists,
1519 			 * it only makes getattr() consistent with readdir().
1520 			 */
1521 			return -ENOENT;
1522 		}
1523 		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1524 		    task_dumpable(task)) {
1525 			cred = __task_cred(task);
1526 			stat->uid = cred->euid;
1527 			stat->gid = cred->egid;
1528 		}
1529 	}
1530 	rcu_read_unlock();
1531 	return 0;
1532 }
1533 
1534 /* dentry stuff */
1535 
1536 /*
1537  *	Exceptional case: normally we are not allowed to unhash a busy
1538  * directory. In this case, however, we can do it - no aliasing problems
1539  * due to the way we treat inodes.
1540  *
1541  * Rewrite the inode's ownerships here because the owning task may have
1542  * performed a setuid(), etc.
1543  *
1544  * Before the /proc/pid/status file was created the only way to read
1545  * the effective uid of a /process was to stat /proc/pid.  Reading
1546  * /proc/pid/status is slow enough that procps and other packages
1547  * kept stating /proc/pid.  To keep the rules in /proc simple I have
1548  * made this apply to all per process world readable and executable
1549  * directories.
1550  */
1551 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1552 {
1553 	struct inode *inode;
1554 	struct task_struct *task;
1555 	const struct cred *cred;
1556 
1557 	if (flags & LOOKUP_RCU)
1558 		return -ECHILD;
1559 
1560 	inode = d_inode(dentry);
1561 	task = get_proc_task(inode);
1562 
1563 	if (task) {
1564 		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1565 		    task_dumpable(task)) {
1566 			rcu_read_lock();
1567 			cred = __task_cred(task);
1568 			inode->i_uid = cred->euid;
1569 			inode->i_gid = cred->egid;
1570 			rcu_read_unlock();
1571 		} else {
1572 			inode->i_uid = GLOBAL_ROOT_UID;
1573 			inode->i_gid = GLOBAL_ROOT_GID;
1574 		}
1575 		inode->i_mode &= ~(S_ISUID | S_ISGID);
1576 		security_task_to_inode(task, inode);
1577 		put_task_struct(task);
1578 		return 1;
1579 	}
1580 	return 0;
1581 }
1582 
1583 static inline bool proc_inode_is_dead(struct inode *inode)
1584 {
1585 	return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1586 }
1587 
1588 int pid_delete_dentry(const struct dentry *dentry)
1589 {
1590 	/* Is the task we represent dead?
1591 	 * If so, then don't put the dentry on the lru list,
1592 	 * kill it immediately.
1593 	 */
1594 	return proc_inode_is_dead(d_inode(dentry));
1595 }
1596 
1597 const struct dentry_operations pid_dentry_operations =
1598 {
1599 	.d_revalidate	= pid_revalidate,
1600 	.d_delete	= pid_delete_dentry,
1601 };
1602 
1603 /* Lookups */
1604 
1605 /*
1606  * Fill a directory entry.
1607  *
1608  * If possible create the dcache entry and derive our inode number and
1609  * file type from dcache entry.
1610  *
1611  * Since all of the proc inode numbers are dynamically generated, the inode
1612  * numbers do not exist until the inode is cache.  This means creating the
1613  * the dcache entry in readdir is necessary to keep the inode numbers
1614  * reported by readdir in sync with the inode numbers reported
1615  * by stat.
1616  */
1617 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
1618 	const char *name, int len,
1619 	instantiate_t instantiate, struct task_struct *task, const void *ptr)
1620 {
1621 	struct dentry *child, *dir = file->f_path.dentry;
1622 	struct qstr qname = QSTR_INIT(name, len);
1623 	struct inode *inode;
1624 	unsigned type;
1625 	ino_t ino;
1626 
1627 	child = d_hash_and_lookup(dir, &qname);
1628 	if (!child) {
1629 		child = d_alloc(dir, &qname);
1630 		if (!child)
1631 			goto end_instantiate;
1632 		if (instantiate(d_inode(dir), child, task, ptr) < 0) {
1633 			dput(child);
1634 			goto end_instantiate;
1635 		}
1636 	}
1637 	inode = d_inode(child);
1638 	ino = inode->i_ino;
1639 	type = inode->i_mode >> 12;
1640 	dput(child);
1641 	return dir_emit(ctx, name, len, ino, type);
1642 
1643 end_instantiate:
1644 	return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
1645 }
1646 
1647 #ifdef CONFIG_CHECKPOINT_RESTORE
1648 
1649 /*
1650  * dname_to_vma_addr - maps a dentry name into two unsigned longs
1651  * which represent vma start and end addresses.
1652  */
1653 static int dname_to_vma_addr(struct dentry *dentry,
1654 			     unsigned long *start, unsigned long *end)
1655 {
1656 	if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1657 		return -EINVAL;
1658 
1659 	return 0;
1660 }
1661 
1662 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1663 {
1664 	unsigned long vm_start, vm_end;
1665 	bool exact_vma_exists = false;
1666 	struct mm_struct *mm = NULL;
1667 	struct task_struct *task;
1668 	const struct cred *cred;
1669 	struct inode *inode;
1670 	int status = 0;
1671 
1672 	if (flags & LOOKUP_RCU)
1673 		return -ECHILD;
1674 
1675 	if (!capable(CAP_SYS_ADMIN)) {
1676 		status = -EPERM;
1677 		goto out_notask;
1678 	}
1679 
1680 	inode = d_inode(dentry);
1681 	task = get_proc_task(inode);
1682 	if (!task)
1683 		goto out_notask;
1684 
1685 	mm = mm_access(task, PTRACE_MODE_READ);
1686 	if (IS_ERR_OR_NULL(mm))
1687 		goto out;
1688 
1689 	if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1690 		down_read(&mm->mmap_sem);
1691 		exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1692 		up_read(&mm->mmap_sem);
1693 	}
1694 
1695 	mmput(mm);
1696 
1697 	if (exact_vma_exists) {
1698 		if (task_dumpable(task)) {
1699 			rcu_read_lock();
1700 			cred = __task_cred(task);
1701 			inode->i_uid = cred->euid;
1702 			inode->i_gid = cred->egid;
1703 			rcu_read_unlock();
1704 		} else {
1705 			inode->i_uid = GLOBAL_ROOT_UID;
1706 			inode->i_gid = GLOBAL_ROOT_GID;
1707 		}
1708 		security_task_to_inode(task, inode);
1709 		status = 1;
1710 	}
1711 
1712 out:
1713 	put_task_struct(task);
1714 
1715 out_notask:
1716 	return status;
1717 }
1718 
1719 static const struct dentry_operations tid_map_files_dentry_operations = {
1720 	.d_revalidate	= map_files_d_revalidate,
1721 	.d_delete	= pid_delete_dentry,
1722 };
1723 
1724 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1725 {
1726 	unsigned long vm_start, vm_end;
1727 	struct vm_area_struct *vma;
1728 	struct task_struct *task;
1729 	struct mm_struct *mm;
1730 	int rc;
1731 
1732 	rc = -ENOENT;
1733 	task = get_proc_task(d_inode(dentry));
1734 	if (!task)
1735 		goto out;
1736 
1737 	mm = get_task_mm(task);
1738 	put_task_struct(task);
1739 	if (!mm)
1740 		goto out;
1741 
1742 	rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1743 	if (rc)
1744 		goto out_mmput;
1745 
1746 	rc = -ENOENT;
1747 	down_read(&mm->mmap_sem);
1748 	vma = find_exact_vma(mm, vm_start, vm_end);
1749 	if (vma && vma->vm_file) {
1750 		*path = vma->vm_file->f_path;
1751 		path_get(path);
1752 		rc = 0;
1753 	}
1754 	up_read(&mm->mmap_sem);
1755 
1756 out_mmput:
1757 	mmput(mm);
1758 out:
1759 	return rc;
1760 }
1761 
1762 struct map_files_info {
1763 	fmode_t		mode;
1764 	unsigned long	len;
1765 	unsigned char	name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1766 };
1767 
1768 static int
1769 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1770 			   struct task_struct *task, const void *ptr)
1771 {
1772 	fmode_t mode = (fmode_t)(unsigned long)ptr;
1773 	struct proc_inode *ei;
1774 	struct inode *inode;
1775 
1776 	inode = proc_pid_make_inode(dir->i_sb, task);
1777 	if (!inode)
1778 		return -ENOENT;
1779 
1780 	ei = PROC_I(inode);
1781 	ei->op.proc_get_link = proc_map_files_get_link;
1782 
1783 	inode->i_op = &proc_pid_link_inode_operations;
1784 	inode->i_size = 64;
1785 	inode->i_mode = S_IFLNK;
1786 
1787 	if (mode & FMODE_READ)
1788 		inode->i_mode |= S_IRUSR;
1789 	if (mode & FMODE_WRITE)
1790 		inode->i_mode |= S_IWUSR;
1791 
1792 	d_set_d_op(dentry, &tid_map_files_dentry_operations);
1793 	d_add(dentry, inode);
1794 
1795 	return 0;
1796 }
1797 
1798 static struct dentry *proc_map_files_lookup(struct inode *dir,
1799 		struct dentry *dentry, unsigned int flags)
1800 {
1801 	unsigned long vm_start, vm_end;
1802 	struct vm_area_struct *vma;
1803 	struct task_struct *task;
1804 	int result;
1805 	struct mm_struct *mm;
1806 
1807 	result = -EPERM;
1808 	if (!capable(CAP_SYS_ADMIN))
1809 		goto out;
1810 
1811 	result = -ENOENT;
1812 	task = get_proc_task(dir);
1813 	if (!task)
1814 		goto out;
1815 
1816 	result = -EACCES;
1817 	if (!ptrace_may_access(task, PTRACE_MODE_READ))
1818 		goto out_put_task;
1819 
1820 	result = -ENOENT;
1821 	if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
1822 		goto out_put_task;
1823 
1824 	mm = get_task_mm(task);
1825 	if (!mm)
1826 		goto out_put_task;
1827 
1828 	down_read(&mm->mmap_sem);
1829 	vma = find_exact_vma(mm, vm_start, vm_end);
1830 	if (!vma)
1831 		goto out_no_vma;
1832 
1833 	if (vma->vm_file)
1834 		result = proc_map_files_instantiate(dir, dentry, task,
1835 				(void *)(unsigned long)vma->vm_file->f_mode);
1836 
1837 out_no_vma:
1838 	up_read(&mm->mmap_sem);
1839 	mmput(mm);
1840 out_put_task:
1841 	put_task_struct(task);
1842 out:
1843 	return ERR_PTR(result);
1844 }
1845 
1846 static const struct inode_operations proc_map_files_inode_operations = {
1847 	.lookup		= proc_map_files_lookup,
1848 	.permission	= proc_fd_permission,
1849 	.setattr	= proc_setattr,
1850 };
1851 
1852 static int
1853 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
1854 {
1855 	struct vm_area_struct *vma;
1856 	struct task_struct *task;
1857 	struct mm_struct *mm;
1858 	unsigned long nr_files, pos, i;
1859 	struct flex_array *fa = NULL;
1860 	struct map_files_info info;
1861 	struct map_files_info *p;
1862 	int ret;
1863 
1864 	ret = -EPERM;
1865 	if (!capable(CAP_SYS_ADMIN))
1866 		goto out;
1867 
1868 	ret = -ENOENT;
1869 	task = get_proc_task(file_inode(file));
1870 	if (!task)
1871 		goto out;
1872 
1873 	ret = -EACCES;
1874 	if (!ptrace_may_access(task, PTRACE_MODE_READ))
1875 		goto out_put_task;
1876 
1877 	ret = 0;
1878 	if (!dir_emit_dots(file, ctx))
1879 		goto out_put_task;
1880 
1881 	mm = get_task_mm(task);
1882 	if (!mm)
1883 		goto out_put_task;
1884 	down_read(&mm->mmap_sem);
1885 
1886 	nr_files = 0;
1887 
1888 	/*
1889 	 * We need two passes here:
1890 	 *
1891 	 *  1) Collect vmas of mapped files with mmap_sem taken
1892 	 *  2) Release mmap_sem and instantiate entries
1893 	 *
1894 	 * otherwise we get lockdep complained, since filldir()
1895 	 * routine might require mmap_sem taken in might_fault().
1896 	 */
1897 
1898 	for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
1899 		if (vma->vm_file && ++pos > ctx->pos)
1900 			nr_files++;
1901 	}
1902 
1903 	if (nr_files) {
1904 		fa = flex_array_alloc(sizeof(info), nr_files,
1905 					GFP_KERNEL);
1906 		if (!fa || flex_array_prealloc(fa, 0, nr_files,
1907 						GFP_KERNEL)) {
1908 			ret = -ENOMEM;
1909 			if (fa)
1910 				flex_array_free(fa);
1911 			up_read(&mm->mmap_sem);
1912 			mmput(mm);
1913 			goto out_put_task;
1914 		}
1915 		for (i = 0, vma = mm->mmap, pos = 2; vma;
1916 				vma = vma->vm_next) {
1917 			if (!vma->vm_file)
1918 				continue;
1919 			if (++pos <= ctx->pos)
1920 				continue;
1921 
1922 			info.mode = vma->vm_file->f_mode;
1923 			info.len = snprintf(info.name,
1924 					sizeof(info.name), "%lx-%lx",
1925 					vma->vm_start, vma->vm_end);
1926 			if (flex_array_put(fa, i++, &info, GFP_KERNEL))
1927 				BUG();
1928 		}
1929 	}
1930 	up_read(&mm->mmap_sem);
1931 
1932 	for (i = 0; i < nr_files; i++) {
1933 		p = flex_array_get(fa, i);
1934 		if (!proc_fill_cache(file, ctx,
1935 				      p->name, p->len,
1936 				      proc_map_files_instantiate,
1937 				      task,
1938 				      (void *)(unsigned long)p->mode))
1939 			break;
1940 		ctx->pos++;
1941 	}
1942 	if (fa)
1943 		flex_array_free(fa);
1944 	mmput(mm);
1945 
1946 out_put_task:
1947 	put_task_struct(task);
1948 out:
1949 	return ret;
1950 }
1951 
1952 static const struct file_operations proc_map_files_operations = {
1953 	.read		= generic_read_dir,
1954 	.iterate	= proc_map_files_readdir,
1955 	.llseek		= default_llseek,
1956 };
1957 
1958 struct timers_private {
1959 	struct pid *pid;
1960 	struct task_struct *task;
1961 	struct sighand_struct *sighand;
1962 	struct pid_namespace *ns;
1963 	unsigned long flags;
1964 };
1965 
1966 static void *timers_start(struct seq_file *m, loff_t *pos)
1967 {
1968 	struct timers_private *tp = m->private;
1969 
1970 	tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
1971 	if (!tp->task)
1972 		return ERR_PTR(-ESRCH);
1973 
1974 	tp->sighand = lock_task_sighand(tp->task, &tp->flags);
1975 	if (!tp->sighand)
1976 		return ERR_PTR(-ESRCH);
1977 
1978 	return seq_list_start(&tp->task->signal->posix_timers, *pos);
1979 }
1980 
1981 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
1982 {
1983 	struct timers_private *tp = m->private;
1984 	return seq_list_next(v, &tp->task->signal->posix_timers, pos);
1985 }
1986 
1987 static void timers_stop(struct seq_file *m, void *v)
1988 {
1989 	struct timers_private *tp = m->private;
1990 
1991 	if (tp->sighand) {
1992 		unlock_task_sighand(tp->task, &tp->flags);
1993 		tp->sighand = NULL;
1994 	}
1995 
1996 	if (tp->task) {
1997 		put_task_struct(tp->task);
1998 		tp->task = NULL;
1999 	}
2000 }
2001 
2002 static int show_timer(struct seq_file *m, void *v)
2003 {
2004 	struct k_itimer *timer;
2005 	struct timers_private *tp = m->private;
2006 	int notify;
2007 	static const char * const nstr[] = {
2008 		[SIGEV_SIGNAL] = "signal",
2009 		[SIGEV_NONE] = "none",
2010 		[SIGEV_THREAD] = "thread",
2011 	};
2012 
2013 	timer = list_entry((struct list_head *)v, struct k_itimer, list);
2014 	notify = timer->it_sigev_notify;
2015 
2016 	seq_printf(m, "ID: %d\n", timer->it_id);
2017 	seq_printf(m, "signal: %d/%p\n",
2018 		   timer->sigq->info.si_signo,
2019 		   timer->sigq->info.si_value.sival_ptr);
2020 	seq_printf(m, "notify: %s/%s.%d\n",
2021 		   nstr[notify & ~SIGEV_THREAD_ID],
2022 		   (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2023 		   pid_nr_ns(timer->it_pid, tp->ns));
2024 	seq_printf(m, "ClockID: %d\n", timer->it_clock);
2025 
2026 	return 0;
2027 }
2028 
2029 static const struct seq_operations proc_timers_seq_ops = {
2030 	.start	= timers_start,
2031 	.next	= timers_next,
2032 	.stop	= timers_stop,
2033 	.show	= show_timer,
2034 };
2035 
2036 static int proc_timers_open(struct inode *inode, struct file *file)
2037 {
2038 	struct timers_private *tp;
2039 
2040 	tp = __seq_open_private(file, &proc_timers_seq_ops,
2041 			sizeof(struct timers_private));
2042 	if (!tp)
2043 		return -ENOMEM;
2044 
2045 	tp->pid = proc_pid(inode);
2046 	tp->ns = inode->i_sb->s_fs_info;
2047 	return 0;
2048 }
2049 
2050 static const struct file_operations proc_timers_operations = {
2051 	.open		= proc_timers_open,
2052 	.read		= seq_read,
2053 	.llseek		= seq_lseek,
2054 	.release	= seq_release_private,
2055 };
2056 #endif /* CONFIG_CHECKPOINT_RESTORE */
2057 
2058 static int proc_pident_instantiate(struct inode *dir,
2059 	struct dentry *dentry, struct task_struct *task, const void *ptr)
2060 {
2061 	const struct pid_entry *p = ptr;
2062 	struct inode *inode;
2063 	struct proc_inode *ei;
2064 
2065 	inode = proc_pid_make_inode(dir->i_sb, task);
2066 	if (!inode)
2067 		goto out;
2068 
2069 	ei = PROC_I(inode);
2070 	inode->i_mode = p->mode;
2071 	if (S_ISDIR(inode->i_mode))
2072 		set_nlink(inode, 2);	/* Use getattr to fix if necessary */
2073 	if (p->iop)
2074 		inode->i_op = p->iop;
2075 	if (p->fop)
2076 		inode->i_fop = p->fop;
2077 	ei->op = p->op;
2078 	d_set_d_op(dentry, &pid_dentry_operations);
2079 	d_add(dentry, inode);
2080 	/* Close the race of the process dying before we return the dentry */
2081 	if (pid_revalidate(dentry, 0))
2082 		return 0;
2083 out:
2084 	return -ENOENT;
2085 }
2086 
2087 static struct dentry *proc_pident_lookup(struct inode *dir,
2088 					 struct dentry *dentry,
2089 					 const struct pid_entry *ents,
2090 					 unsigned int nents)
2091 {
2092 	int error;
2093 	struct task_struct *task = get_proc_task(dir);
2094 	const struct pid_entry *p, *last;
2095 
2096 	error = -ENOENT;
2097 
2098 	if (!task)
2099 		goto out_no_task;
2100 
2101 	/*
2102 	 * Yes, it does not scale. And it should not. Don't add
2103 	 * new entries into /proc/<tgid>/ without very good reasons.
2104 	 */
2105 	last = &ents[nents - 1];
2106 	for (p = ents; p <= last; p++) {
2107 		if (p->len != dentry->d_name.len)
2108 			continue;
2109 		if (!memcmp(dentry->d_name.name, p->name, p->len))
2110 			break;
2111 	}
2112 	if (p > last)
2113 		goto out;
2114 
2115 	error = proc_pident_instantiate(dir, dentry, task, p);
2116 out:
2117 	put_task_struct(task);
2118 out_no_task:
2119 	return ERR_PTR(error);
2120 }
2121 
2122 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2123 		const struct pid_entry *ents, unsigned int nents)
2124 {
2125 	struct task_struct *task = get_proc_task(file_inode(file));
2126 	const struct pid_entry *p;
2127 
2128 	if (!task)
2129 		return -ENOENT;
2130 
2131 	if (!dir_emit_dots(file, ctx))
2132 		goto out;
2133 
2134 	if (ctx->pos >= nents + 2)
2135 		goto out;
2136 
2137 	for (p = ents + (ctx->pos - 2); p <= ents + nents - 1; p++) {
2138 		if (!proc_fill_cache(file, ctx, p->name, p->len,
2139 				proc_pident_instantiate, task, p))
2140 			break;
2141 		ctx->pos++;
2142 	}
2143 out:
2144 	put_task_struct(task);
2145 	return 0;
2146 }
2147 
2148 #ifdef CONFIG_SECURITY
2149 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2150 				  size_t count, loff_t *ppos)
2151 {
2152 	struct inode * inode = file_inode(file);
2153 	char *p = NULL;
2154 	ssize_t length;
2155 	struct task_struct *task = get_proc_task(inode);
2156 
2157 	if (!task)
2158 		return -ESRCH;
2159 
2160 	length = security_getprocattr(task,
2161 				      (char*)file->f_path.dentry->d_name.name,
2162 				      &p);
2163 	put_task_struct(task);
2164 	if (length > 0)
2165 		length = simple_read_from_buffer(buf, count, ppos, p, length);
2166 	kfree(p);
2167 	return length;
2168 }
2169 
2170 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2171 				   size_t count, loff_t *ppos)
2172 {
2173 	struct inode * inode = file_inode(file);
2174 	char *page;
2175 	ssize_t length;
2176 	struct task_struct *task = get_proc_task(inode);
2177 
2178 	length = -ESRCH;
2179 	if (!task)
2180 		goto out_no_task;
2181 	if (count > PAGE_SIZE)
2182 		count = PAGE_SIZE;
2183 
2184 	/* No partial writes. */
2185 	length = -EINVAL;
2186 	if (*ppos != 0)
2187 		goto out;
2188 
2189 	length = -ENOMEM;
2190 	page = (char*)__get_free_page(GFP_TEMPORARY);
2191 	if (!page)
2192 		goto out;
2193 
2194 	length = -EFAULT;
2195 	if (copy_from_user(page, buf, count))
2196 		goto out_free;
2197 
2198 	/* Guard against adverse ptrace interaction */
2199 	length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2200 	if (length < 0)
2201 		goto out_free;
2202 
2203 	length = security_setprocattr(task,
2204 				      (char*)file->f_path.dentry->d_name.name,
2205 				      (void*)page, count);
2206 	mutex_unlock(&task->signal->cred_guard_mutex);
2207 out_free:
2208 	free_page((unsigned long) page);
2209 out:
2210 	put_task_struct(task);
2211 out_no_task:
2212 	return length;
2213 }
2214 
2215 static const struct file_operations proc_pid_attr_operations = {
2216 	.read		= proc_pid_attr_read,
2217 	.write		= proc_pid_attr_write,
2218 	.llseek		= generic_file_llseek,
2219 };
2220 
2221 static const struct pid_entry attr_dir_stuff[] = {
2222 	REG("current",    S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2223 	REG("prev",       S_IRUGO,	   proc_pid_attr_operations),
2224 	REG("exec",       S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2225 	REG("fscreate",   S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2226 	REG("keycreate",  S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2227 	REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2228 };
2229 
2230 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2231 {
2232 	return proc_pident_readdir(file, ctx,
2233 				   attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2234 }
2235 
2236 static const struct file_operations proc_attr_dir_operations = {
2237 	.read		= generic_read_dir,
2238 	.iterate	= proc_attr_dir_readdir,
2239 	.llseek		= default_llseek,
2240 };
2241 
2242 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2243 				struct dentry *dentry, unsigned int flags)
2244 {
2245 	return proc_pident_lookup(dir, dentry,
2246 				  attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2247 }
2248 
2249 static const struct inode_operations proc_attr_dir_inode_operations = {
2250 	.lookup		= proc_attr_dir_lookup,
2251 	.getattr	= pid_getattr,
2252 	.setattr	= proc_setattr,
2253 };
2254 
2255 #endif
2256 
2257 #ifdef CONFIG_ELF_CORE
2258 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2259 					 size_t count, loff_t *ppos)
2260 {
2261 	struct task_struct *task = get_proc_task(file_inode(file));
2262 	struct mm_struct *mm;
2263 	char buffer[PROC_NUMBUF];
2264 	size_t len;
2265 	int ret;
2266 
2267 	if (!task)
2268 		return -ESRCH;
2269 
2270 	ret = 0;
2271 	mm = get_task_mm(task);
2272 	if (mm) {
2273 		len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2274 			       ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2275 				MMF_DUMP_FILTER_SHIFT));
2276 		mmput(mm);
2277 		ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2278 	}
2279 
2280 	put_task_struct(task);
2281 
2282 	return ret;
2283 }
2284 
2285 static ssize_t proc_coredump_filter_write(struct file *file,
2286 					  const char __user *buf,
2287 					  size_t count,
2288 					  loff_t *ppos)
2289 {
2290 	struct task_struct *task;
2291 	struct mm_struct *mm;
2292 	char buffer[PROC_NUMBUF], *end;
2293 	unsigned int val;
2294 	int ret;
2295 	int i;
2296 	unsigned long mask;
2297 
2298 	ret = -EFAULT;
2299 	memset(buffer, 0, sizeof(buffer));
2300 	if (count > sizeof(buffer) - 1)
2301 		count = sizeof(buffer) - 1;
2302 	if (copy_from_user(buffer, buf, count))
2303 		goto out_no_task;
2304 
2305 	ret = -EINVAL;
2306 	val = (unsigned int)simple_strtoul(buffer, &end, 0);
2307 	if (*end == '\n')
2308 		end++;
2309 	if (end - buffer == 0)
2310 		goto out_no_task;
2311 
2312 	ret = -ESRCH;
2313 	task = get_proc_task(file_inode(file));
2314 	if (!task)
2315 		goto out_no_task;
2316 
2317 	ret = end - buffer;
2318 	mm = get_task_mm(task);
2319 	if (!mm)
2320 		goto out_no_mm;
2321 
2322 	for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2323 		if (val & mask)
2324 			set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2325 		else
2326 			clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2327 	}
2328 
2329 	mmput(mm);
2330  out_no_mm:
2331 	put_task_struct(task);
2332  out_no_task:
2333 	return ret;
2334 }
2335 
2336 static const struct file_operations proc_coredump_filter_operations = {
2337 	.read		= proc_coredump_filter_read,
2338 	.write		= proc_coredump_filter_write,
2339 	.llseek		= generic_file_llseek,
2340 };
2341 #endif
2342 
2343 #ifdef CONFIG_TASK_IO_ACCOUNTING
2344 static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2345 {
2346 	struct task_io_accounting acct = task->ioac;
2347 	unsigned long flags;
2348 	int result;
2349 
2350 	result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2351 	if (result)
2352 		return result;
2353 
2354 	if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2355 		result = -EACCES;
2356 		goto out_unlock;
2357 	}
2358 
2359 	if (whole && lock_task_sighand(task, &flags)) {
2360 		struct task_struct *t = task;
2361 
2362 		task_io_accounting_add(&acct, &task->signal->ioac);
2363 		while_each_thread(task, t)
2364 			task_io_accounting_add(&acct, &t->ioac);
2365 
2366 		unlock_task_sighand(task, &flags);
2367 	}
2368 	seq_printf(m,
2369 		   "rchar: %llu\n"
2370 		   "wchar: %llu\n"
2371 		   "syscr: %llu\n"
2372 		   "syscw: %llu\n"
2373 		   "read_bytes: %llu\n"
2374 		   "write_bytes: %llu\n"
2375 		   "cancelled_write_bytes: %llu\n",
2376 		   (unsigned long long)acct.rchar,
2377 		   (unsigned long long)acct.wchar,
2378 		   (unsigned long long)acct.syscr,
2379 		   (unsigned long long)acct.syscw,
2380 		   (unsigned long long)acct.read_bytes,
2381 		   (unsigned long long)acct.write_bytes,
2382 		   (unsigned long long)acct.cancelled_write_bytes);
2383 	result = 0;
2384 
2385 out_unlock:
2386 	mutex_unlock(&task->signal->cred_guard_mutex);
2387 	return result;
2388 }
2389 
2390 static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2391 				  struct pid *pid, struct task_struct *task)
2392 {
2393 	return do_io_accounting(task, m, 0);
2394 }
2395 
2396 static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2397 				   struct pid *pid, struct task_struct *task)
2398 {
2399 	return do_io_accounting(task, m, 1);
2400 }
2401 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2402 
2403 #ifdef CONFIG_USER_NS
2404 static int proc_id_map_open(struct inode *inode, struct file *file,
2405 	const struct seq_operations *seq_ops)
2406 {
2407 	struct user_namespace *ns = NULL;
2408 	struct task_struct *task;
2409 	struct seq_file *seq;
2410 	int ret = -EINVAL;
2411 
2412 	task = get_proc_task(inode);
2413 	if (task) {
2414 		rcu_read_lock();
2415 		ns = get_user_ns(task_cred_xxx(task, user_ns));
2416 		rcu_read_unlock();
2417 		put_task_struct(task);
2418 	}
2419 	if (!ns)
2420 		goto err;
2421 
2422 	ret = seq_open(file, seq_ops);
2423 	if (ret)
2424 		goto err_put_ns;
2425 
2426 	seq = file->private_data;
2427 	seq->private = ns;
2428 
2429 	return 0;
2430 err_put_ns:
2431 	put_user_ns(ns);
2432 err:
2433 	return ret;
2434 }
2435 
2436 static int proc_id_map_release(struct inode *inode, struct file *file)
2437 {
2438 	struct seq_file *seq = file->private_data;
2439 	struct user_namespace *ns = seq->private;
2440 	put_user_ns(ns);
2441 	return seq_release(inode, file);
2442 }
2443 
2444 static int proc_uid_map_open(struct inode *inode, struct file *file)
2445 {
2446 	return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2447 }
2448 
2449 static int proc_gid_map_open(struct inode *inode, struct file *file)
2450 {
2451 	return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2452 }
2453 
2454 static int proc_projid_map_open(struct inode *inode, struct file *file)
2455 {
2456 	return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2457 }
2458 
2459 static const struct file_operations proc_uid_map_operations = {
2460 	.open		= proc_uid_map_open,
2461 	.write		= proc_uid_map_write,
2462 	.read		= seq_read,
2463 	.llseek		= seq_lseek,
2464 	.release	= proc_id_map_release,
2465 };
2466 
2467 static const struct file_operations proc_gid_map_operations = {
2468 	.open		= proc_gid_map_open,
2469 	.write		= proc_gid_map_write,
2470 	.read		= seq_read,
2471 	.llseek		= seq_lseek,
2472 	.release	= proc_id_map_release,
2473 };
2474 
2475 static const struct file_operations proc_projid_map_operations = {
2476 	.open		= proc_projid_map_open,
2477 	.write		= proc_projid_map_write,
2478 	.read		= seq_read,
2479 	.llseek		= seq_lseek,
2480 	.release	= proc_id_map_release,
2481 };
2482 
2483 static int proc_setgroups_open(struct inode *inode, struct file *file)
2484 {
2485 	struct user_namespace *ns = NULL;
2486 	struct task_struct *task;
2487 	int ret;
2488 
2489 	ret = -ESRCH;
2490 	task = get_proc_task(inode);
2491 	if (task) {
2492 		rcu_read_lock();
2493 		ns = get_user_ns(task_cred_xxx(task, user_ns));
2494 		rcu_read_unlock();
2495 		put_task_struct(task);
2496 	}
2497 	if (!ns)
2498 		goto err;
2499 
2500 	if (file->f_mode & FMODE_WRITE) {
2501 		ret = -EACCES;
2502 		if (!ns_capable(ns, CAP_SYS_ADMIN))
2503 			goto err_put_ns;
2504 	}
2505 
2506 	ret = single_open(file, &proc_setgroups_show, ns);
2507 	if (ret)
2508 		goto err_put_ns;
2509 
2510 	return 0;
2511 err_put_ns:
2512 	put_user_ns(ns);
2513 err:
2514 	return ret;
2515 }
2516 
2517 static int proc_setgroups_release(struct inode *inode, struct file *file)
2518 {
2519 	struct seq_file *seq = file->private_data;
2520 	struct user_namespace *ns = seq->private;
2521 	int ret = single_release(inode, file);
2522 	put_user_ns(ns);
2523 	return ret;
2524 }
2525 
2526 static const struct file_operations proc_setgroups_operations = {
2527 	.open		= proc_setgroups_open,
2528 	.write		= proc_setgroups_write,
2529 	.read		= seq_read,
2530 	.llseek		= seq_lseek,
2531 	.release	= proc_setgroups_release,
2532 };
2533 #endif /* CONFIG_USER_NS */
2534 
2535 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2536 				struct pid *pid, struct task_struct *task)
2537 {
2538 	int err = lock_trace(task);
2539 	if (!err) {
2540 		seq_printf(m, "%08x\n", task->personality);
2541 		unlock_trace(task);
2542 	}
2543 	return err;
2544 }
2545 
2546 /*
2547  * Thread groups
2548  */
2549 static const struct file_operations proc_task_operations;
2550 static const struct inode_operations proc_task_inode_operations;
2551 
2552 static const struct pid_entry tgid_base_stuff[] = {
2553 	DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2554 	DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2555 #ifdef CONFIG_CHECKPOINT_RESTORE
2556 	DIR("map_files",  S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2557 #endif
2558 	DIR("fdinfo",     S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2559 	DIR("ns",	  S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2560 #ifdef CONFIG_NET
2561 	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2562 #endif
2563 	REG("environ",    S_IRUSR, proc_environ_operations),
2564 	ONE("auxv",       S_IRUSR, proc_pid_auxv),
2565 	ONE("status",     S_IRUGO, proc_pid_status),
2566 	ONE("personality", S_IRUSR, proc_pid_personality),
2567 	ONE("limits",	  S_IRUGO, proc_pid_limits),
2568 #ifdef CONFIG_SCHED_DEBUG
2569 	REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2570 #endif
2571 #ifdef CONFIG_SCHED_AUTOGROUP
2572 	REG("autogroup",  S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2573 #endif
2574 	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2575 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2576 	ONE("syscall",    S_IRUSR, proc_pid_syscall),
2577 #endif
2578 	ONE("cmdline",    S_IRUGO, proc_pid_cmdline),
2579 	ONE("stat",       S_IRUGO, proc_tgid_stat),
2580 	ONE("statm",      S_IRUGO, proc_pid_statm),
2581 	REG("maps",       S_IRUGO, proc_pid_maps_operations),
2582 #ifdef CONFIG_NUMA
2583 	REG("numa_maps",  S_IRUGO, proc_pid_numa_maps_operations),
2584 #endif
2585 	REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
2586 	LNK("cwd",        proc_cwd_link),
2587 	LNK("root",       proc_root_link),
2588 	LNK("exe",        proc_exe_link),
2589 	REG("mounts",     S_IRUGO, proc_mounts_operations),
2590 	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
2591 	REG("mountstats", S_IRUSR, proc_mountstats_operations),
2592 #ifdef CONFIG_PROC_PAGE_MONITOR
2593 	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2594 	REG("smaps",      S_IRUGO, proc_pid_smaps_operations),
2595 	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
2596 #endif
2597 #ifdef CONFIG_SECURITY
2598 	DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2599 #endif
2600 #ifdef CONFIG_KALLSYMS
2601 	ONE("wchan",      S_IRUGO, proc_pid_wchan),
2602 #endif
2603 #ifdef CONFIG_STACKTRACE
2604 	ONE("stack",      S_IRUSR, proc_pid_stack),
2605 #endif
2606 #ifdef CONFIG_SCHED_INFO
2607 	ONE("schedstat",  S_IRUGO, proc_pid_schedstat),
2608 #endif
2609 #ifdef CONFIG_LATENCYTOP
2610 	REG("latency",  S_IRUGO, proc_lstats_operations),
2611 #endif
2612 #ifdef CONFIG_PROC_PID_CPUSET
2613 	ONE("cpuset",     S_IRUGO, proc_cpuset_show),
2614 #endif
2615 #ifdef CONFIG_CGROUPS
2616 	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
2617 #endif
2618 	ONE("oom_score",  S_IRUGO, proc_oom_score),
2619 	REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2620 	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2621 #ifdef CONFIG_AUDITSYSCALL
2622 	REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
2623 	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
2624 #endif
2625 #ifdef CONFIG_FAULT_INJECTION
2626 	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2627 #endif
2628 #ifdef CONFIG_ELF_CORE
2629 	REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2630 #endif
2631 #ifdef CONFIG_TASK_IO_ACCOUNTING
2632 	ONE("io",	S_IRUSR, proc_tgid_io_accounting),
2633 #endif
2634 #ifdef CONFIG_HARDWALL
2635 	ONE("hardwall",   S_IRUGO, proc_pid_hardwall),
2636 #endif
2637 #ifdef CONFIG_USER_NS
2638 	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
2639 	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
2640 	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2641 	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
2642 #endif
2643 #ifdef CONFIG_CHECKPOINT_RESTORE
2644 	REG("timers",	  S_IRUGO, proc_timers_operations),
2645 #endif
2646 };
2647 
2648 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
2649 {
2650 	return proc_pident_readdir(file, ctx,
2651 				   tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2652 }
2653 
2654 static const struct file_operations proc_tgid_base_operations = {
2655 	.read		= generic_read_dir,
2656 	.iterate	= proc_tgid_base_readdir,
2657 	.llseek		= default_llseek,
2658 };
2659 
2660 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2661 {
2662 	return proc_pident_lookup(dir, dentry,
2663 				  tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2664 }
2665 
2666 static const struct inode_operations proc_tgid_base_inode_operations = {
2667 	.lookup		= proc_tgid_base_lookup,
2668 	.getattr	= pid_getattr,
2669 	.setattr	= proc_setattr,
2670 	.permission	= proc_pid_permission,
2671 };
2672 
2673 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2674 {
2675 	struct dentry *dentry, *leader, *dir;
2676 	char buf[PROC_NUMBUF];
2677 	struct qstr name;
2678 
2679 	name.name = buf;
2680 	name.len = snprintf(buf, sizeof(buf), "%d", pid);
2681 	/* no ->d_hash() rejects on procfs */
2682 	dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2683 	if (dentry) {
2684 		d_invalidate(dentry);
2685 		dput(dentry);
2686 	}
2687 
2688 	if (pid == tgid)
2689 		return;
2690 
2691 	name.name = buf;
2692 	name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2693 	leader = d_hash_and_lookup(mnt->mnt_root, &name);
2694 	if (!leader)
2695 		goto out;
2696 
2697 	name.name = "task";
2698 	name.len = strlen(name.name);
2699 	dir = d_hash_and_lookup(leader, &name);
2700 	if (!dir)
2701 		goto out_put_leader;
2702 
2703 	name.name = buf;
2704 	name.len = snprintf(buf, sizeof(buf), "%d", pid);
2705 	dentry = d_hash_and_lookup(dir, &name);
2706 	if (dentry) {
2707 		d_invalidate(dentry);
2708 		dput(dentry);
2709 	}
2710 
2711 	dput(dir);
2712 out_put_leader:
2713 	dput(leader);
2714 out:
2715 	return;
2716 }
2717 
2718 /**
2719  * proc_flush_task -  Remove dcache entries for @task from the /proc dcache.
2720  * @task: task that should be flushed.
2721  *
2722  * When flushing dentries from proc, one needs to flush them from global
2723  * proc (proc_mnt) and from all the namespaces' procs this task was seen
2724  * in. This call is supposed to do all of this job.
2725  *
2726  * Looks in the dcache for
2727  * /proc/@pid
2728  * /proc/@tgid/task/@pid
2729  * if either directory is present flushes it and all of it'ts children
2730  * from the dcache.
2731  *
2732  * It is safe and reasonable to cache /proc entries for a task until
2733  * that task exits.  After that they just clog up the dcache with
2734  * useless entries, possibly causing useful dcache entries to be
2735  * flushed instead.  This routine is proved to flush those useless
2736  * dcache entries at process exit time.
2737  *
2738  * NOTE: This routine is just an optimization so it does not guarantee
2739  *       that no dcache entries will exist at process exit time it
2740  *       just makes it very unlikely that any will persist.
2741  */
2742 
2743 void proc_flush_task(struct task_struct *task)
2744 {
2745 	int i;
2746 	struct pid *pid, *tgid;
2747 	struct upid *upid;
2748 
2749 	pid = task_pid(task);
2750 	tgid = task_tgid(task);
2751 
2752 	for (i = 0; i <= pid->level; i++) {
2753 		upid = &pid->numbers[i];
2754 		proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2755 					tgid->numbers[i].nr);
2756 	}
2757 }
2758 
2759 static int proc_pid_instantiate(struct inode *dir,
2760 				   struct dentry * dentry,
2761 				   struct task_struct *task, const void *ptr)
2762 {
2763 	struct inode *inode;
2764 
2765 	inode = proc_pid_make_inode(dir->i_sb, task);
2766 	if (!inode)
2767 		goto out;
2768 
2769 	inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2770 	inode->i_op = &proc_tgid_base_inode_operations;
2771 	inode->i_fop = &proc_tgid_base_operations;
2772 	inode->i_flags|=S_IMMUTABLE;
2773 
2774 	set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2775 						  ARRAY_SIZE(tgid_base_stuff)));
2776 
2777 	d_set_d_op(dentry, &pid_dentry_operations);
2778 
2779 	d_add(dentry, inode);
2780 	/* Close the race of the process dying before we return the dentry */
2781 	if (pid_revalidate(dentry, 0))
2782 		return 0;
2783 out:
2784 	return -ENOENT;
2785 }
2786 
2787 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
2788 {
2789 	int result = -ENOENT;
2790 	struct task_struct *task;
2791 	unsigned tgid;
2792 	struct pid_namespace *ns;
2793 
2794 	tgid = name_to_int(&dentry->d_name);
2795 	if (tgid == ~0U)
2796 		goto out;
2797 
2798 	ns = dentry->d_sb->s_fs_info;
2799 	rcu_read_lock();
2800 	task = find_task_by_pid_ns(tgid, ns);
2801 	if (task)
2802 		get_task_struct(task);
2803 	rcu_read_unlock();
2804 	if (!task)
2805 		goto out;
2806 
2807 	result = proc_pid_instantiate(dir, dentry, task, NULL);
2808 	put_task_struct(task);
2809 out:
2810 	return ERR_PTR(result);
2811 }
2812 
2813 /*
2814  * Find the first task with tgid >= tgid
2815  *
2816  */
2817 struct tgid_iter {
2818 	unsigned int tgid;
2819 	struct task_struct *task;
2820 };
2821 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2822 {
2823 	struct pid *pid;
2824 
2825 	if (iter.task)
2826 		put_task_struct(iter.task);
2827 	rcu_read_lock();
2828 retry:
2829 	iter.task = NULL;
2830 	pid = find_ge_pid(iter.tgid, ns);
2831 	if (pid) {
2832 		iter.tgid = pid_nr_ns(pid, ns);
2833 		iter.task = pid_task(pid, PIDTYPE_PID);
2834 		/* What we to know is if the pid we have find is the
2835 		 * pid of a thread_group_leader.  Testing for task
2836 		 * being a thread_group_leader is the obvious thing
2837 		 * todo but there is a window when it fails, due to
2838 		 * the pid transfer logic in de_thread.
2839 		 *
2840 		 * So we perform the straight forward test of seeing
2841 		 * if the pid we have found is the pid of a thread
2842 		 * group leader, and don't worry if the task we have
2843 		 * found doesn't happen to be a thread group leader.
2844 		 * As we don't care in the case of readdir.
2845 		 */
2846 		if (!iter.task || !has_group_leader_pid(iter.task)) {
2847 			iter.tgid += 1;
2848 			goto retry;
2849 		}
2850 		get_task_struct(iter.task);
2851 	}
2852 	rcu_read_unlock();
2853 	return iter;
2854 }
2855 
2856 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
2857 
2858 /* for the /proc/ directory itself, after non-process stuff has been done */
2859 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
2860 {
2861 	struct tgid_iter iter;
2862 	struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
2863 	loff_t pos = ctx->pos;
2864 
2865 	if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
2866 		return 0;
2867 
2868 	if (pos == TGID_OFFSET - 2) {
2869 		struct inode *inode = d_inode(ns->proc_self);
2870 		if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
2871 			return 0;
2872 		ctx->pos = pos = pos + 1;
2873 	}
2874 	if (pos == TGID_OFFSET - 1) {
2875 		struct inode *inode = d_inode(ns->proc_thread_self);
2876 		if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
2877 			return 0;
2878 		ctx->pos = pos = pos + 1;
2879 	}
2880 	iter.tgid = pos - TGID_OFFSET;
2881 	iter.task = NULL;
2882 	for (iter = next_tgid(ns, iter);
2883 	     iter.task;
2884 	     iter.tgid += 1, iter = next_tgid(ns, iter)) {
2885 		char name[PROC_NUMBUF];
2886 		int len;
2887 		if (!has_pid_permissions(ns, iter.task, 2))
2888 			continue;
2889 
2890 		len = snprintf(name, sizeof(name), "%d", iter.tgid);
2891 		ctx->pos = iter.tgid + TGID_OFFSET;
2892 		if (!proc_fill_cache(file, ctx, name, len,
2893 				     proc_pid_instantiate, iter.task, NULL)) {
2894 			put_task_struct(iter.task);
2895 			return 0;
2896 		}
2897 	}
2898 	ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
2899 	return 0;
2900 }
2901 
2902 /*
2903  * Tasks
2904  */
2905 static const struct pid_entry tid_base_stuff[] = {
2906 	DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2907 	DIR("fdinfo",    S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2908 	DIR("ns",	 S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2909 #ifdef CONFIG_NET
2910 	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2911 #endif
2912 	REG("environ",   S_IRUSR, proc_environ_operations),
2913 	ONE("auxv",      S_IRUSR, proc_pid_auxv),
2914 	ONE("status",    S_IRUGO, proc_pid_status),
2915 	ONE("personality", S_IRUSR, proc_pid_personality),
2916 	ONE("limits",	 S_IRUGO, proc_pid_limits),
2917 #ifdef CONFIG_SCHED_DEBUG
2918 	REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2919 #endif
2920 	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2921 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2922 	ONE("syscall",   S_IRUSR, proc_pid_syscall),
2923 #endif
2924 	ONE("cmdline",   S_IRUGO, proc_pid_cmdline),
2925 	ONE("stat",      S_IRUGO, proc_tid_stat),
2926 	ONE("statm",     S_IRUGO, proc_pid_statm),
2927 	REG("maps",      S_IRUGO, proc_tid_maps_operations),
2928 #ifdef CONFIG_CHECKPOINT_RESTORE
2929 	REG("children",  S_IRUGO, proc_tid_children_operations),
2930 #endif
2931 #ifdef CONFIG_NUMA
2932 	REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
2933 #endif
2934 	REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
2935 	LNK("cwd",       proc_cwd_link),
2936 	LNK("root",      proc_root_link),
2937 	LNK("exe",       proc_exe_link),
2938 	REG("mounts",    S_IRUGO, proc_mounts_operations),
2939 	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
2940 #ifdef CONFIG_PROC_PAGE_MONITOR
2941 	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2942 	REG("smaps",     S_IRUGO, proc_tid_smaps_operations),
2943 	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
2944 #endif
2945 #ifdef CONFIG_SECURITY
2946 	DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2947 #endif
2948 #ifdef CONFIG_KALLSYMS
2949 	ONE("wchan",     S_IRUGO, proc_pid_wchan),
2950 #endif
2951 #ifdef CONFIG_STACKTRACE
2952 	ONE("stack",      S_IRUSR, proc_pid_stack),
2953 #endif
2954 #ifdef CONFIG_SCHED_INFO
2955 	ONE("schedstat", S_IRUGO, proc_pid_schedstat),
2956 #endif
2957 #ifdef CONFIG_LATENCYTOP
2958 	REG("latency",  S_IRUGO, proc_lstats_operations),
2959 #endif
2960 #ifdef CONFIG_PROC_PID_CPUSET
2961 	ONE("cpuset",    S_IRUGO, proc_cpuset_show),
2962 #endif
2963 #ifdef CONFIG_CGROUPS
2964 	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
2965 #endif
2966 	ONE("oom_score", S_IRUGO, proc_oom_score),
2967 	REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2968 	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2969 #ifdef CONFIG_AUDITSYSCALL
2970 	REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
2971 	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
2972 #endif
2973 #ifdef CONFIG_FAULT_INJECTION
2974 	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2975 #endif
2976 #ifdef CONFIG_TASK_IO_ACCOUNTING
2977 	ONE("io",	S_IRUSR, proc_tid_io_accounting),
2978 #endif
2979 #ifdef CONFIG_HARDWALL
2980 	ONE("hardwall",   S_IRUGO, proc_pid_hardwall),
2981 #endif
2982 #ifdef CONFIG_USER_NS
2983 	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
2984 	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
2985 	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2986 	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
2987 #endif
2988 };
2989 
2990 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
2991 {
2992 	return proc_pident_readdir(file, ctx,
2993 				   tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2994 }
2995 
2996 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2997 {
2998 	return proc_pident_lookup(dir, dentry,
2999 				  tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3000 }
3001 
3002 static const struct file_operations proc_tid_base_operations = {
3003 	.read		= generic_read_dir,
3004 	.iterate	= proc_tid_base_readdir,
3005 	.llseek		= default_llseek,
3006 };
3007 
3008 static const struct inode_operations proc_tid_base_inode_operations = {
3009 	.lookup		= proc_tid_base_lookup,
3010 	.getattr	= pid_getattr,
3011 	.setattr	= proc_setattr,
3012 };
3013 
3014 static int proc_task_instantiate(struct inode *dir,
3015 	struct dentry *dentry, struct task_struct *task, const void *ptr)
3016 {
3017 	struct inode *inode;
3018 	inode = proc_pid_make_inode(dir->i_sb, task);
3019 
3020 	if (!inode)
3021 		goto out;
3022 	inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3023 	inode->i_op = &proc_tid_base_inode_operations;
3024 	inode->i_fop = &proc_tid_base_operations;
3025 	inode->i_flags|=S_IMMUTABLE;
3026 
3027 	set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3028 						  ARRAY_SIZE(tid_base_stuff)));
3029 
3030 	d_set_d_op(dentry, &pid_dentry_operations);
3031 
3032 	d_add(dentry, inode);
3033 	/* Close the race of the process dying before we return the dentry */
3034 	if (pid_revalidate(dentry, 0))
3035 		return 0;
3036 out:
3037 	return -ENOENT;
3038 }
3039 
3040 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3041 {
3042 	int result = -ENOENT;
3043 	struct task_struct *task;
3044 	struct task_struct *leader = get_proc_task(dir);
3045 	unsigned tid;
3046 	struct pid_namespace *ns;
3047 
3048 	if (!leader)
3049 		goto out_no_task;
3050 
3051 	tid = name_to_int(&dentry->d_name);
3052 	if (tid == ~0U)
3053 		goto out;
3054 
3055 	ns = dentry->d_sb->s_fs_info;
3056 	rcu_read_lock();
3057 	task = find_task_by_pid_ns(tid, ns);
3058 	if (task)
3059 		get_task_struct(task);
3060 	rcu_read_unlock();
3061 	if (!task)
3062 		goto out;
3063 	if (!same_thread_group(leader, task))
3064 		goto out_drop_task;
3065 
3066 	result = proc_task_instantiate(dir, dentry, task, NULL);
3067 out_drop_task:
3068 	put_task_struct(task);
3069 out:
3070 	put_task_struct(leader);
3071 out_no_task:
3072 	return ERR_PTR(result);
3073 }
3074 
3075 /*
3076  * Find the first tid of a thread group to return to user space.
3077  *
3078  * Usually this is just the thread group leader, but if the users
3079  * buffer was too small or there was a seek into the middle of the
3080  * directory we have more work todo.
3081  *
3082  * In the case of a short read we start with find_task_by_pid.
3083  *
3084  * In the case of a seek we start with the leader and walk nr
3085  * threads past it.
3086  */
3087 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3088 					struct pid_namespace *ns)
3089 {
3090 	struct task_struct *pos, *task;
3091 	unsigned long nr = f_pos;
3092 
3093 	if (nr != f_pos)	/* 32bit overflow? */
3094 		return NULL;
3095 
3096 	rcu_read_lock();
3097 	task = pid_task(pid, PIDTYPE_PID);
3098 	if (!task)
3099 		goto fail;
3100 
3101 	/* Attempt to start with the tid of a thread */
3102 	if (tid && nr) {
3103 		pos = find_task_by_pid_ns(tid, ns);
3104 		if (pos && same_thread_group(pos, task))
3105 			goto found;
3106 	}
3107 
3108 	/* If nr exceeds the number of threads there is nothing todo */
3109 	if (nr >= get_nr_threads(task))
3110 		goto fail;
3111 
3112 	/* If we haven't found our starting place yet start
3113 	 * with the leader and walk nr threads forward.
3114 	 */
3115 	pos = task = task->group_leader;
3116 	do {
3117 		if (!nr--)
3118 			goto found;
3119 	} while_each_thread(task, pos);
3120 fail:
3121 	pos = NULL;
3122 	goto out;
3123 found:
3124 	get_task_struct(pos);
3125 out:
3126 	rcu_read_unlock();
3127 	return pos;
3128 }
3129 
3130 /*
3131  * Find the next thread in the thread list.
3132  * Return NULL if there is an error or no next thread.
3133  *
3134  * The reference to the input task_struct is released.
3135  */
3136 static struct task_struct *next_tid(struct task_struct *start)
3137 {
3138 	struct task_struct *pos = NULL;
3139 	rcu_read_lock();
3140 	if (pid_alive(start)) {
3141 		pos = next_thread(start);
3142 		if (thread_group_leader(pos))
3143 			pos = NULL;
3144 		else
3145 			get_task_struct(pos);
3146 	}
3147 	rcu_read_unlock();
3148 	put_task_struct(start);
3149 	return pos;
3150 }
3151 
3152 /* for the /proc/TGID/task/ directories */
3153 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3154 {
3155 	struct inode *inode = file_inode(file);
3156 	struct task_struct *task;
3157 	struct pid_namespace *ns;
3158 	int tid;
3159 
3160 	if (proc_inode_is_dead(inode))
3161 		return -ENOENT;
3162 
3163 	if (!dir_emit_dots(file, ctx))
3164 		return 0;
3165 
3166 	/* f_version caches the tgid value that the last readdir call couldn't
3167 	 * return. lseek aka telldir automagically resets f_version to 0.
3168 	 */
3169 	ns = inode->i_sb->s_fs_info;
3170 	tid = (int)file->f_version;
3171 	file->f_version = 0;
3172 	for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3173 	     task;
3174 	     task = next_tid(task), ctx->pos++) {
3175 		char name[PROC_NUMBUF];
3176 		int len;
3177 		tid = task_pid_nr_ns(task, ns);
3178 		len = snprintf(name, sizeof(name), "%d", tid);
3179 		if (!proc_fill_cache(file, ctx, name, len,
3180 				proc_task_instantiate, task, NULL)) {
3181 			/* returning this tgid failed, save it as the first
3182 			 * pid for the next readir call */
3183 			file->f_version = (u64)tid;
3184 			put_task_struct(task);
3185 			break;
3186 		}
3187 	}
3188 
3189 	return 0;
3190 }
3191 
3192 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3193 {
3194 	struct inode *inode = d_inode(dentry);
3195 	struct task_struct *p = get_proc_task(inode);
3196 	generic_fillattr(inode, stat);
3197 
3198 	if (p) {
3199 		stat->nlink += get_nr_threads(p);
3200 		put_task_struct(p);
3201 	}
3202 
3203 	return 0;
3204 }
3205 
3206 static const struct inode_operations proc_task_inode_operations = {
3207 	.lookup		= proc_task_lookup,
3208 	.getattr	= proc_task_getattr,
3209 	.setattr	= proc_setattr,
3210 	.permission	= proc_pid_permission,
3211 };
3212 
3213 static const struct file_operations proc_task_operations = {
3214 	.read		= generic_read_dir,
3215 	.iterate	= proc_task_readdir,
3216 	.llseek		= default_llseek,
3217 };
3218