xref: /openbmc/linux/fs/proc/base.c (revision ab73b751)
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/cgroup.h>
77 #include <linux/cpuset.h>
78 #include <linux/audit.h>
79 #include <linux/poll.h>
80 #include <linux/nsproxy.h>
81 #include <linux/oom.h>
82 #include <linux/elf.h>
83 #include <linux/pid_namespace.h>
84 #include <linux/user_namespace.h>
85 #include <linux/fs_struct.h>
86 #include <linux/slab.h>
87 #include <linux/flex_array.h>
88 #ifdef CONFIG_HARDWALL
89 #include <asm/hardwall.h>
90 #endif
91 #include <trace/events/oom.h>
92 #include "internal.h"
93 
94 /* NOTE:
95  *	Implementing inode permission operations in /proc is almost
96  *	certainly an error.  Permission checks need to happen during
97  *	each system call not at open time.  The reason is that most of
98  *	what we wish to check for permissions in /proc varies at runtime.
99  *
100  *	The classic example of a problem is opening file descriptors
101  *	in /proc for a task before it execs a suid executable.
102  */
103 
104 struct pid_entry {
105 	char *name;
106 	int len;
107 	umode_t mode;
108 	const struct inode_operations *iop;
109 	const struct file_operations *fop;
110 	union proc_op op;
111 };
112 
113 #define NOD(NAME, MODE, IOP, FOP, OP) {			\
114 	.name = (NAME),					\
115 	.len  = sizeof(NAME) - 1,			\
116 	.mode = MODE,					\
117 	.iop  = IOP,					\
118 	.fop  = FOP,					\
119 	.op   = OP,					\
120 }
121 
122 #define DIR(NAME, MODE, iops, fops)	\
123 	NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
124 #define LNK(NAME, get_link)					\
125 	NOD(NAME, (S_IFLNK|S_IRWXUGO),				\
126 		&proc_pid_link_inode_operations, NULL,		\
127 		{ .proc_get_link = get_link } )
128 #define REG(NAME, MODE, fops)				\
129 	NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
130 #define INF(NAME, MODE, read)				\
131 	NOD(NAME, (S_IFREG|(MODE)), 			\
132 		NULL, &proc_info_file_operations,	\
133 		{ .proc_read = read } )
134 #define ONE(NAME, MODE, show)				\
135 	NOD(NAME, (S_IFREG|(MODE)), 			\
136 		NULL, &proc_single_file_operations,	\
137 		{ .proc_show = show } )
138 
139 static int proc_fd_permission(struct inode *inode, int mask);
140 
141 /*
142  * Count the number of hardlinks for the pid_entry table, excluding the .
143  * and .. links.
144  */
145 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
146 	unsigned int n)
147 {
148 	unsigned int i;
149 	unsigned int count;
150 
151 	count = 0;
152 	for (i = 0; i < n; ++i) {
153 		if (S_ISDIR(entries[i].mode))
154 			++count;
155 	}
156 
157 	return count;
158 }
159 
160 static int get_task_root(struct task_struct *task, struct path *root)
161 {
162 	int result = -ENOENT;
163 
164 	task_lock(task);
165 	if (task->fs) {
166 		get_fs_root(task->fs, root);
167 		result = 0;
168 	}
169 	task_unlock(task);
170 	return result;
171 }
172 
173 static int proc_cwd_link(struct dentry *dentry, struct path *path)
174 {
175 	struct task_struct *task = get_proc_task(dentry->d_inode);
176 	int result = -ENOENT;
177 
178 	if (task) {
179 		task_lock(task);
180 		if (task->fs) {
181 			get_fs_pwd(task->fs, path);
182 			result = 0;
183 		}
184 		task_unlock(task);
185 		put_task_struct(task);
186 	}
187 	return result;
188 }
189 
190 static int proc_root_link(struct dentry *dentry, struct path *path)
191 {
192 	struct task_struct *task = get_proc_task(dentry->d_inode);
193 	int result = -ENOENT;
194 
195 	if (task) {
196 		result = get_task_root(task, path);
197 		put_task_struct(task);
198 	}
199 	return result;
200 }
201 
202 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
203 {
204 	int res = 0;
205 	unsigned int len;
206 	struct mm_struct *mm = get_task_mm(task);
207 	if (!mm)
208 		goto out;
209 	if (!mm->arg_end)
210 		goto out_mm;	/* Shh! No looking before we're done */
211 
212  	len = mm->arg_end - mm->arg_start;
213 
214 	if (len > PAGE_SIZE)
215 		len = PAGE_SIZE;
216 
217 	res = access_process_vm(task, mm->arg_start, buffer, len, 0);
218 
219 	// If the nul at the end of args has been overwritten, then
220 	// assume application is using setproctitle(3).
221 	if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
222 		len = strnlen(buffer, res);
223 		if (len < res) {
224 		    res = len;
225 		} else {
226 			len = mm->env_end - mm->env_start;
227 			if (len > PAGE_SIZE - res)
228 				len = PAGE_SIZE - res;
229 			res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
230 			res = strnlen(buffer, res);
231 		}
232 	}
233 out_mm:
234 	mmput(mm);
235 out:
236 	return res;
237 }
238 
239 static int proc_pid_auxv(struct task_struct *task, char *buffer)
240 {
241 	struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
242 	int res = PTR_ERR(mm);
243 	if (mm && !IS_ERR(mm)) {
244 		unsigned int nwords = 0;
245 		do {
246 			nwords += 2;
247 		} while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
248 		res = nwords * sizeof(mm->saved_auxv[0]);
249 		if (res > PAGE_SIZE)
250 			res = PAGE_SIZE;
251 		memcpy(buffer, mm->saved_auxv, res);
252 		mmput(mm);
253 	}
254 	return res;
255 }
256 
257 
258 #ifdef CONFIG_KALLSYMS
259 /*
260  * Provides a wchan file via kallsyms in a proper one-value-per-file format.
261  * Returns the resolved symbol.  If that fails, simply return the address.
262  */
263 static int proc_pid_wchan(struct task_struct *task, char *buffer)
264 {
265 	unsigned long wchan;
266 	char symname[KSYM_NAME_LEN];
267 
268 	wchan = get_wchan(task);
269 
270 	if (lookup_symbol_name(wchan, symname) < 0)
271 		if (!ptrace_may_access(task, PTRACE_MODE_READ))
272 			return 0;
273 		else
274 			return sprintf(buffer, "%lu", wchan);
275 	else
276 		return sprintf(buffer, "%s", symname);
277 }
278 #endif /* CONFIG_KALLSYMS */
279 
280 static int lock_trace(struct task_struct *task)
281 {
282 	int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
283 	if (err)
284 		return err;
285 	if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
286 		mutex_unlock(&task->signal->cred_guard_mutex);
287 		return -EPERM;
288 	}
289 	return 0;
290 }
291 
292 static void unlock_trace(struct task_struct *task)
293 {
294 	mutex_unlock(&task->signal->cred_guard_mutex);
295 }
296 
297 #ifdef CONFIG_STACKTRACE
298 
299 #define MAX_STACK_TRACE_DEPTH	64
300 
301 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
302 			  struct pid *pid, struct task_struct *task)
303 {
304 	struct stack_trace trace;
305 	unsigned long *entries;
306 	int err;
307 	int i;
308 
309 	entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
310 	if (!entries)
311 		return -ENOMEM;
312 
313 	trace.nr_entries	= 0;
314 	trace.max_entries	= MAX_STACK_TRACE_DEPTH;
315 	trace.entries		= entries;
316 	trace.skip		= 0;
317 
318 	err = lock_trace(task);
319 	if (!err) {
320 		save_stack_trace_tsk(task, &trace);
321 
322 		for (i = 0; i < trace.nr_entries; i++) {
323 			seq_printf(m, "[<%pK>] %pS\n",
324 				   (void *)entries[i], (void *)entries[i]);
325 		}
326 		unlock_trace(task);
327 	}
328 	kfree(entries);
329 
330 	return err;
331 }
332 #endif
333 
334 #ifdef CONFIG_SCHEDSTATS
335 /*
336  * Provides /proc/PID/schedstat
337  */
338 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
339 {
340 	return sprintf(buffer, "%llu %llu %lu\n",
341 			(unsigned long long)task->se.sum_exec_runtime,
342 			(unsigned long long)task->sched_info.run_delay,
343 			task->sched_info.pcount);
344 }
345 #endif
346 
347 #ifdef CONFIG_LATENCYTOP
348 static int lstats_show_proc(struct seq_file *m, void *v)
349 {
350 	int i;
351 	struct inode *inode = m->private;
352 	struct task_struct *task = get_proc_task(inode);
353 
354 	if (!task)
355 		return -ESRCH;
356 	seq_puts(m, "Latency Top version : v0.1\n");
357 	for (i = 0; i < 32; i++) {
358 		struct latency_record *lr = &task->latency_record[i];
359 		if (lr->backtrace[0]) {
360 			int q;
361 			seq_printf(m, "%i %li %li",
362 				   lr->count, lr->time, lr->max);
363 			for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
364 				unsigned long bt = lr->backtrace[q];
365 				if (!bt)
366 					break;
367 				if (bt == ULONG_MAX)
368 					break;
369 				seq_printf(m, " %ps", (void *)bt);
370 			}
371 			seq_putc(m, '\n');
372 		}
373 
374 	}
375 	put_task_struct(task);
376 	return 0;
377 }
378 
379 static int lstats_open(struct inode *inode, struct file *file)
380 {
381 	return single_open(file, lstats_show_proc, inode);
382 }
383 
384 static ssize_t lstats_write(struct file *file, const char __user *buf,
385 			    size_t count, loff_t *offs)
386 {
387 	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
388 
389 	if (!task)
390 		return -ESRCH;
391 	clear_all_latency_tracing(task);
392 	put_task_struct(task);
393 
394 	return count;
395 }
396 
397 static const struct file_operations proc_lstats_operations = {
398 	.open		= lstats_open,
399 	.read		= seq_read,
400 	.write		= lstats_write,
401 	.llseek		= seq_lseek,
402 	.release	= single_release,
403 };
404 
405 #endif
406 
407 static int proc_oom_score(struct task_struct *task, char *buffer)
408 {
409 	unsigned long totalpages = totalram_pages + total_swap_pages;
410 	unsigned long points = 0;
411 
412 	read_lock(&tasklist_lock);
413 	if (pid_alive(task))
414 		points = oom_badness(task, NULL, NULL, totalpages) *
415 						1000 / totalpages;
416 	read_unlock(&tasklist_lock);
417 	return sprintf(buffer, "%lu\n", points);
418 }
419 
420 struct limit_names {
421 	char *name;
422 	char *unit;
423 };
424 
425 static const struct limit_names lnames[RLIM_NLIMITS] = {
426 	[RLIMIT_CPU] = {"Max cpu time", "seconds"},
427 	[RLIMIT_FSIZE] = {"Max file size", "bytes"},
428 	[RLIMIT_DATA] = {"Max data size", "bytes"},
429 	[RLIMIT_STACK] = {"Max stack size", "bytes"},
430 	[RLIMIT_CORE] = {"Max core file size", "bytes"},
431 	[RLIMIT_RSS] = {"Max resident set", "bytes"},
432 	[RLIMIT_NPROC] = {"Max processes", "processes"},
433 	[RLIMIT_NOFILE] = {"Max open files", "files"},
434 	[RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
435 	[RLIMIT_AS] = {"Max address space", "bytes"},
436 	[RLIMIT_LOCKS] = {"Max file locks", "locks"},
437 	[RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
438 	[RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
439 	[RLIMIT_NICE] = {"Max nice priority", NULL},
440 	[RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
441 	[RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
442 };
443 
444 /* Display limits for a process */
445 static int proc_pid_limits(struct task_struct *task, char *buffer)
446 {
447 	unsigned int i;
448 	int count = 0;
449 	unsigned long flags;
450 	char *bufptr = buffer;
451 
452 	struct rlimit rlim[RLIM_NLIMITS];
453 
454 	if (!lock_task_sighand(task, &flags))
455 		return 0;
456 	memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
457 	unlock_task_sighand(task, &flags);
458 
459 	/*
460 	 * print the file header
461 	 */
462 	count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
463 			"Limit", "Soft Limit", "Hard Limit", "Units");
464 
465 	for (i = 0; i < RLIM_NLIMITS; i++) {
466 		if (rlim[i].rlim_cur == RLIM_INFINITY)
467 			count += sprintf(&bufptr[count], "%-25s %-20s ",
468 					 lnames[i].name, "unlimited");
469 		else
470 			count += sprintf(&bufptr[count], "%-25s %-20lu ",
471 					 lnames[i].name, rlim[i].rlim_cur);
472 
473 		if (rlim[i].rlim_max == RLIM_INFINITY)
474 			count += sprintf(&bufptr[count], "%-20s ", "unlimited");
475 		else
476 			count += sprintf(&bufptr[count], "%-20lu ",
477 					 rlim[i].rlim_max);
478 
479 		if (lnames[i].unit)
480 			count += sprintf(&bufptr[count], "%-10s\n",
481 					 lnames[i].unit);
482 		else
483 			count += sprintf(&bufptr[count], "\n");
484 	}
485 
486 	return count;
487 }
488 
489 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
490 static int proc_pid_syscall(struct task_struct *task, char *buffer)
491 {
492 	long nr;
493 	unsigned long args[6], sp, pc;
494 	int res = lock_trace(task);
495 	if (res)
496 		return res;
497 
498 	if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
499 		res = sprintf(buffer, "running\n");
500 	else if (nr < 0)
501 		res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
502 	else
503 		res = sprintf(buffer,
504 		       "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
505 		       nr,
506 		       args[0], args[1], args[2], args[3], args[4], args[5],
507 		       sp, pc);
508 	unlock_trace(task);
509 	return res;
510 }
511 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
512 
513 /************************************************************************/
514 /*                       Here the fs part begins                        */
515 /************************************************************************/
516 
517 /* permission checks */
518 static int proc_fd_access_allowed(struct inode *inode)
519 {
520 	struct task_struct *task;
521 	int allowed = 0;
522 	/* Allow access to a task's file descriptors if it is us or we
523 	 * may use ptrace attach to the process and find out that
524 	 * information.
525 	 */
526 	task = get_proc_task(inode);
527 	if (task) {
528 		allowed = ptrace_may_access(task, PTRACE_MODE_READ);
529 		put_task_struct(task);
530 	}
531 	return allowed;
532 }
533 
534 int proc_setattr(struct dentry *dentry, struct iattr *attr)
535 {
536 	int error;
537 	struct inode *inode = dentry->d_inode;
538 
539 	if (attr->ia_valid & ATTR_MODE)
540 		return -EPERM;
541 
542 	error = inode_change_ok(inode, attr);
543 	if (error)
544 		return error;
545 
546 	if ((attr->ia_valid & ATTR_SIZE) &&
547 	    attr->ia_size != i_size_read(inode)) {
548 		error = vmtruncate(inode, attr->ia_size);
549 		if (error)
550 			return error;
551 	}
552 
553 	setattr_copy(inode, attr);
554 	mark_inode_dirty(inode);
555 	return 0;
556 }
557 
558 /*
559  * May current process learn task's sched/cmdline info (for hide_pid_min=1)
560  * or euid/egid (for hide_pid_min=2)?
561  */
562 static bool has_pid_permissions(struct pid_namespace *pid,
563 				 struct task_struct *task,
564 				 int hide_pid_min)
565 {
566 	if (pid->hide_pid < hide_pid_min)
567 		return true;
568 	if (in_group_p(pid->pid_gid))
569 		return true;
570 	return ptrace_may_access(task, PTRACE_MODE_READ);
571 }
572 
573 
574 static int proc_pid_permission(struct inode *inode, int mask)
575 {
576 	struct pid_namespace *pid = inode->i_sb->s_fs_info;
577 	struct task_struct *task;
578 	bool has_perms;
579 
580 	task = get_proc_task(inode);
581 	if (!task)
582 		return -ESRCH;
583 	has_perms = has_pid_permissions(pid, task, 1);
584 	put_task_struct(task);
585 
586 	if (!has_perms) {
587 		if (pid->hide_pid == 2) {
588 			/*
589 			 * Let's make getdents(), stat(), and open()
590 			 * consistent with each other.  If a process
591 			 * may not stat() a file, it shouldn't be seen
592 			 * in procfs at all.
593 			 */
594 			return -ENOENT;
595 		}
596 
597 		return -EPERM;
598 	}
599 	return generic_permission(inode, mask);
600 }
601 
602 
603 
604 static const struct inode_operations proc_def_inode_operations = {
605 	.setattr	= proc_setattr,
606 };
607 
608 #define PROC_BLOCK_SIZE	(3*1024)		/* 4K page size but our output routines use some slack for overruns */
609 
610 static ssize_t proc_info_read(struct file * file, char __user * buf,
611 			  size_t count, loff_t *ppos)
612 {
613 	struct inode * inode = file->f_path.dentry->d_inode;
614 	unsigned long page;
615 	ssize_t length;
616 	struct task_struct *task = get_proc_task(inode);
617 
618 	length = -ESRCH;
619 	if (!task)
620 		goto out_no_task;
621 
622 	if (count > PROC_BLOCK_SIZE)
623 		count = PROC_BLOCK_SIZE;
624 
625 	length = -ENOMEM;
626 	if (!(page = __get_free_page(GFP_TEMPORARY)))
627 		goto out;
628 
629 	length = PROC_I(inode)->op.proc_read(task, (char*)page);
630 
631 	if (length >= 0)
632 		length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
633 	free_page(page);
634 out:
635 	put_task_struct(task);
636 out_no_task:
637 	return length;
638 }
639 
640 static const struct file_operations proc_info_file_operations = {
641 	.read		= proc_info_read,
642 	.llseek		= generic_file_llseek,
643 };
644 
645 static int proc_single_show(struct seq_file *m, void *v)
646 {
647 	struct inode *inode = m->private;
648 	struct pid_namespace *ns;
649 	struct pid *pid;
650 	struct task_struct *task;
651 	int ret;
652 
653 	ns = inode->i_sb->s_fs_info;
654 	pid = proc_pid(inode);
655 	task = get_pid_task(pid, PIDTYPE_PID);
656 	if (!task)
657 		return -ESRCH;
658 
659 	ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
660 
661 	put_task_struct(task);
662 	return ret;
663 }
664 
665 static int proc_single_open(struct inode *inode, struct file *filp)
666 {
667 	return single_open(filp, proc_single_show, inode);
668 }
669 
670 static const struct file_operations proc_single_file_operations = {
671 	.open		= proc_single_open,
672 	.read		= seq_read,
673 	.llseek		= seq_lseek,
674 	.release	= single_release,
675 };
676 
677 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
678 {
679 	struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
680 	struct mm_struct *mm;
681 
682 	if (!task)
683 		return -ESRCH;
684 
685 	mm = mm_access(task, mode);
686 	put_task_struct(task);
687 
688 	if (IS_ERR(mm))
689 		return PTR_ERR(mm);
690 
691 	if (mm) {
692 		/* ensure this mm_struct can't be freed */
693 		atomic_inc(&mm->mm_count);
694 		/* but do not pin its memory */
695 		mmput(mm);
696 	}
697 
698 	/* OK to pass negative loff_t, we can catch out-of-range */
699 	file->f_mode |= FMODE_UNSIGNED_OFFSET;
700 	file->private_data = mm;
701 
702 	return 0;
703 }
704 
705 static int mem_open(struct inode *inode, struct file *file)
706 {
707 	return __mem_open(inode, file, PTRACE_MODE_ATTACH);
708 }
709 
710 static ssize_t mem_rw(struct file *file, char __user *buf,
711 			size_t count, loff_t *ppos, int write)
712 {
713 	struct mm_struct *mm = file->private_data;
714 	unsigned long addr = *ppos;
715 	ssize_t copied;
716 	char *page;
717 
718 	if (!mm)
719 		return 0;
720 
721 	page = (char *)__get_free_page(GFP_TEMPORARY);
722 	if (!page)
723 		return -ENOMEM;
724 
725 	copied = 0;
726 	if (!atomic_inc_not_zero(&mm->mm_users))
727 		goto free;
728 
729 	while (count > 0) {
730 		int this_len = min_t(int, count, PAGE_SIZE);
731 
732 		if (write && copy_from_user(page, buf, this_len)) {
733 			copied = -EFAULT;
734 			break;
735 		}
736 
737 		this_len = access_remote_vm(mm, addr, page, this_len, write);
738 		if (!this_len) {
739 			if (!copied)
740 				copied = -EIO;
741 			break;
742 		}
743 
744 		if (!write && copy_to_user(buf, page, this_len)) {
745 			copied = -EFAULT;
746 			break;
747 		}
748 
749 		buf += this_len;
750 		addr += this_len;
751 		copied += this_len;
752 		count -= this_len;
753 	}
754 	*ppos = addr;
755 
756 	mmput(mm);
757 free:
758 	free_page((unsigned long) page);
759 	return copied;
760 }
761 
762 static ssize_t mem_read(struct file *file, char __user *buf,
763 			size_t count, loff_t *ppos)
764 {
765 	return mem_rw(file, buf, count, ppos, 0);
766 }
767 
768 static ssize_t mem_write(struct file *file, const char __user *buf,
769 			 size_t count, loff_t *ppos)
770 {
771 	return mem_rw(file, (char __user*)buf, count, ppos, 1);
772 }
773 
774 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
775 {
776 	switch (orig) {
777 	case 0:
778 		file->f_pos = offset;
779 		break;
780 	case 1:
781 		file->f_pos += offset;
782 		break;
783 	default:
784 		return -EINVAL;
785 	}
786 	force_successful_syscall_return();
787 	return file->f_pos;
788 }
789 
790 static int mem_release(struct inode *inode, struct file *file)
791 {
792 	struct mm_struct *mm = file->private_data;
793 	if (mm)
794 		mmdrop(mm);
795 	return 0;
796 }
797 
798 static const struct file_operations proc_mem_operations = {
799 	.llseek		= mem_lseek,
800 	.read		= mem_read,
801 	.write		= mem_write,
802 	.open		= mem_open,
803 	.release	= mem_release,
804 };
805 
806 static int environ_open(struct inode *inode, struct file *file)
807 {
808 	return __mem_open(inode, file, PTRACE_MODE_READ);
809 }
810 
811 static ssize_t environ_read(struct file *file, char __user *buf,
812 			size_t count, loff_t *ppos)
813 {
814 	char *page;
815 	unsigned long src = *ppos;
816 	int ret = 0;
817 	struct mm_struct *mm = file->private_data;
818 
819 	if (!mm)
820 		return 0;
821 
822 	page = (char *)__get_free_page(GFP_TEMPORARY);
823 	if (!page)
824 		return -ENOMEM;
825 
826 	ret = 0;
827 	if (!atomic_inc_not_zero(&mm->mm_users))
828 		goto free;
829 	while (count > 0) {
830 		int this_len, retval, max_len;
831 
832 		this_len = mm->env_end - (mm->env_start + src);
833 
834 		if (this_len <= 0)
835 			break;
836 
837 		max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
838 		this_len = (this_len > max_len) ? max_len : this_len;
839 
840 		retval = access_remote_vm(mm, (mm->env_start + src),
841 			page, this_len, 0);
842 
843 		if (retval <= 0) {
844 			ret = retval;
845 			break;
846 		}
847 
848 		if (copy_to_user(buf, page, retval)) {
849 			ret = -EFAULT;
850 			break;
851 		}
852 
853 		ret += retval;
854 		src += retval;
855 		buf += retval;
856 		count -= retval;
857 	}
858 	*ppos = src;
859 	mmput(mm);
860 
861 free:
862 	free_page((unsigned long) page);
863 	return ret;
864 }
865 
866 static const struct file_operations proc_environ_operations = {
867 	.open		= environ_open,
868 	.read		= environ_read,
869 	.llseek		= generic_file_llseek,
870 	.release	= mem_release,
871 };
872 
873 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
874 				size_t count, loff_t *ppos)
875 {
876 	struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
877 	char buffer[PROC_NUMBUF];
878 	size_t len;
879 	int oom_adjust = OOM_DISABLE;
880 	unsigned long flags;
881 
882 	if (!task)
883 		return -ESRCH;
884 
885 	if (lock_task_sighand(task, &flags)) {
886 		oom_adjust = task->signal->oom_adj;
887 		unlock_task_sighand(task, &flags);
888 	}
889 
890 	put_task_struct(task);
891 
892 	len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
893 
894 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
895 }
896 
897 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
898 				size_t count, loff_t *ppos)
899 {
900 	struct task_struct *task;
901 	char buffer[PROC_NUMBUF];
902 	int oom_adjust;
903 	unsigned long flags;
904 	int err;
905 
906 	memset(buffer, 0, sizeof(buffer));
907 	if (count > sizeof(buffer) - 1)
908 		count = sizeof(buffer) - 1;
909 	if (copy_from_user(buffer, buf, count)) {
910 		err = -EFAULT;
911 		goto out;
912 	}
913 
914 	err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
915 	if (err)
916 		goto out;
917 	if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
918 	     oom_adjust != OOM_DISABLE) {
919 		err = -EINVAL;
920 		goto out;
921 	}
922 
923 	task = get_proc_task(file->f_path.dentry->d_inode);
924 	if (!task) {
925 		err = -ESRCH;
926 		goto out;
927 	}
928 
929 	task_lock(task);
930 	if (!task->mm) {
931 		err = -EINVAL;
932 		goto err_task_lock;
933 	}
934 
935 	if (!lock_task_sighand(task, &flags)) {
936 		err = -ESRCH;
937 		goto err_task_lock;
938 	}
939 
940 	if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
941 		err = -EACCES;
942 		goto err_sighand;
943 	}
944 
945 	/*
946 	 * Warn that /proc/pid/oom_adj is deprecated, see
947 	 * Documentation/feature-removal-schedule.txt.
948 	 */
949 	printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
950 		  current->comm, task_pid_nr(current), task_pid_nr(task),
951 		  task_pid_nr(task));
952 	task->signal->oom_adj = oom_adjust;
953 	/*
954 	 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
955 	 * value is always attainable.
956 	 */
957 	if (task->signal->oom_adj == OOM_ADJUST_MAX)
958 		task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
959 	else
960 		task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
961 								-OOM_DISABLE;
962 	trace_oom_score_adj_update(task);
963 err_sighand:
964 	unlock_task_sighand(task, &flags);
965 err_task_lock:
966 	task_unlock(task);
967 	put_task_struct(task);
968 out:
969 	return err < 0 ? err : count;
970 }
971 
972 static const struct file_operations proc_oom_adjust_operations = {
973 	.read		= oom_adjust_read,
974 	.write		= oom_adjust_write,
975 	.llseek		= generic_file_llseek,
976 };
977 
978 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
979 					size_t count, loff_t *ppos)
980 {
981 	struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
982 	char buffer[PROC_NUMBUF];
983 	int oom_score_adj = OOM_SCORE_ADJ_MIN;
984 	unsigned long flags;
985 	size_t len;
986 
987 	if (!task)
988 		return -ESRCH;
989 	if (lock_task_sighand(task, &flags)) {
990 		oom_score_adj = task->signal->oom_score_adj;
991 		unlock_task_sighand(task, &flags);
992 	}
993 	put_task_struct(task);
994 	len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
995 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
996 }
997 
998 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
999 					size_t count, loff_t *ppos)
1000 {
1001 	struct task_struct *task;
1002 	char buffer[PROC_NUMBUF];
1003 	unsigned long flags;
1004 	int oom_score_adj;
1005 	int err;
1006 
1007 	memset(buffer, 0, sizeof(buffer));
1008 	if (count > sizeof(buffer) - 1)
1009 		count = sizeof(buffer) - 1;
1010 	if (copy_from_user(buffer, buf, count)) {
1011 		err = -EFAULT;
1012 		goto out;
1013 	}
1014 
1015 	err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1016 	if (err)
1017 		goto out;
1018 	if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1019 			oom_score_adj > OOM_SCORE_ADJ_MAX) {
1020 		err = -EINVAL;
1021 		goto out;
1022 	}
1023 
1024 	task = get_proc_task(file->f_path.dentry->d_inode);
1025 	if (!task) {
1026 		err = -ESRCH;
1027 		goto out;
1028 	}
1029 
1030 	task_lock(task);
1031 	if (!task->mm) {
1032 		err = -EINVAL;
1033 		goto err_task_lock;
1034 	}
1035 
1036 	if (!lock_task_sighand(task, &flags)) {
1037 		err = -ESRCH;
1038 		goto err_task_lock;
1039 	}
1040 
1041 	if (oom_score_adj < task->signal->oom_score_adj_min &&
1042 			!capable(CAP_SYS_RESOURCE)) {
1043 		err = -EACCES;
1044 		goto err_sighand;
1045 	}
1046 
1047 	task->signal->oom_score_adj = oom_score_adj;
1048 	if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1049 		task->signal->oom_score_adj_min = oom_score_adj;
1050 	trace_oom_score_adj_update(task);
1051 	/*
1052 	 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1053 	 * always attainable.
1054 	 */
1055 	if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1056 		task->signal->oom_adj = OOM_DISABLE;
1057 	else
1058 		task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1059 							OOM_SCORE_ADJ_MAX;
1060 err_sighand:
1061 	unlock_task_sighand(task, &flags);
1062 err_task_lock:
1063 	task_unlock(task);
1064 	put_task_struct(task);
1065 out:
1066 	return err < 0 ? err : count;
1067 }
1068 
1069 static const struct file_operations proc_oom_score_adj_operations = {
1070 	.read		= oom_score_adj_read,
1071 	.write		= oom_score_adj_write,
1072 	.llseek		= default_llseek,
1073 };
1074 
1075 #ifdef CONFIG_AUDITSYSCALL
1076 #define TMPBUFLEN 21
1077 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1078 				  size_t count, loff_t *ppos)
1079 {
1080 	struct inode * inode = file->f_path.dentry->d_inode;
1081 	struct task_struct *task = get_proc_task(inode);
1082 	ssize_t length;
1083 	char tmpbuf[TMPBUFLEN];
1084 
1085 	if (!task)
1086 		return -ESRCH;
1087 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1088 				audit_get_loginuid(task));
1089 	put_task_struct(task);
1090 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1091 }
1092 
1093 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1094 				   size_t count, loff_t *ppos)
1095 {
1096 	struct inode * inode = file->f_path.dentry->d_inode;
1097 	char *page, *tmp;
1098 	ssize_t length;
1099 	uid_t loginuid;
1100 
1101 	rcu_read_lock();
1102 	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1103 		rcu_read_unlock();
1104 		return -EPERM;
1105 	}
1106 	rcu_read_unlock();
1107 
1108 	if (count >= PAGE_SIZE)
1109 		count = PAGE_SIZE - 1;
1110 
1111 	if (*ppos != 0) {
1112 		/* No partial writes. */
1113 		return -EINVAL;
1114 	}
1115 	page = (char*)__get_free_page(GFP_TEMPORARY);
1116 	if (!page)
1117 		return -ENOMEM;
1118 	length = -EFAULT;
1119 	if (copy_from_user(page, buf, count))
1120 		goto out_free_page;
1121 
1122 	page[count] = '\0';
1123 	loginuid = simple_strtoul(page, &tmp, 10);
1124 	if (tmp == page) {
1125 		length = -EINVAL;
1126 		goto out_free_page;
1127 
1128 	}
1129 	length = audit_set_loginuid(loginuid);
1130 	if (likely(length == 0))
1131 		length = count;
1132 
1133 out_free_page:
1134 	free_page((unsigned long) page);
1135 	return length;
1136 }
1137 
1138 static const struct file_operations proc_loginuid_operations = {
1139 	.read		= proc_loginuid_read,
1140 	.write		= proc_loginuid_write,
1141 	.llseek		= generic_file_llseek,
1142 };
1143 
1144 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1145 				  size_t count, loff_t *ppos)
1146 {
1147 	struct inode * inode = file->f_path.dentry->d_inode;
1148 	struct task_struct *task = get_proc_task(inode);
1149 	ssize_t length;
1150 	char tmpbuf[TMPBUFLEN];
1151 
1152 	if (!task)
1153 		return -ESRCH;
1154 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1155 				audit_get_sessionid(task));
1156 	put_task_struct(task);
1157 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1158 }
1159 
1160 static const struct file_operations proc_sessionid_operations = {
1161 	.read		= proc_sessionid_read,
1162 	.llseek		= generic_file_llseek,
1163 };
1164 #endif
1165 
1166 #ifdef CONFIG_FAULT_INJECTION
1167 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1168 				      size_t count, loff_t *ppos)
1169 {
1170 	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1171 	char buffer[PROC_NUMBUF];
1172 	size_t len;
1173 	int make_it_fail;
1174 
1175 	if (!task)
1176 		return -ESRCH;
1177 	make_it_fail = task->make_it_fail;
1178 	put_task_struct(task);
1179 
1180 	len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1181 
1182 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1183 }
1184 
1185 static ssize_t proc_fault_inject_write(struct file * file,
1186 			const char __user * buf, size_t count, loff_t *ppos)
1187 {
1188 	struct task_struct *task;
1189 	char buffer[PROC_NUMBUF], *end;
1190 	int make_it_fail;
1191 
1192 	if (!capable(CAP_SYS_RESOURCE))
1193 		return -EPERM;
1194 	memset(buffer, 0, sizeof(buffer));
1195 	if (count > sizeof(buffer) - 1)
1196 		count = sizeof(buffer) - 1;
1197 	if (copy_from_user(buffer, buf, count))
1198 		return -EFAULT;
1199 	make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1200 	if (*end)
1201 		return -EINVAL;
1202 	task = get_proc_task(file->f_dentry->d_inode);
1203 	if (!task)
1204 		return -ESRCH;
1205 	task->make_it_fail = make_it_fail;
1206 	put_task_struct(task);
1207 
1208 	return count;
1209 }
1210 
1211 static const struct file_operations proc_fault_inject_operations = {
1212 	.read		= proc_fault_inject_read,
1213 	.write		= proc_fault_inject_write,
1214 	.llseek		= generic_file_llseek,
1215 };
1216 #endif
1217 
1218 
1219 #ifdef CONFIG_SCHED_DEBUG
1220 /*
1221  * Print out various scheduling related per-task fields:
1222  */
1223 static int sched_show(struct seq_file *m, void *v)
1224 {
1225 	struct inode *inode = m->private;
1226 	struct task_struct *p;
1227 
1228 	p = get_proc_task(inode);
1229 	if (!p)
1230 		return -ESRCH;
1231 	proc_sched_show_task(p, m);
1232 
1233 	put_task_struct(p);
1234 
1235 	return 0;
1236 }
1237 
1238 static ssize_t
1239 sched_write(struct file *file, const char __user *buf,
1240 	    size_t count, loff_t *offset)
1241 {
1242 	struct inode *inode = file->f_path.dentry->d_inode;
1243 	struct task_struct *p;
1244 
1245 	p = get_proc_task(inode);
1246 	if (!p)
1247 		return -ESRCH;
1248 	proc_sched_set_task(p);
1249 
1250 	put_task_struct(p);
1251 
1252 	return count;
1253 }
1254 
1255 static int sched_open(struct inode *inode, struct file *filp)
1256 {
1257 	return single_open(filp, sched_show, inode);
1258 }
1259 
1260 static const struct file_operations proc_pid_sched_operations = {
1261 	.open		= sched_open,
1262 	.read		= seq_read,
1263 	.write		= sched_write,
1264 	.llseek		= seq_lseek,
1265 	.release	= single_release,
1266 };
1267 
1268 #endif
1269 
1270 #ifdef CONFIG_SCHED_AUTOGROUP
1271 /*
1272  * Print out autogroup related information:
1273  */
1274 static int sched_autogroup_show(struct seq_file *m, void *v)
1275 {
1276 	struct inode *inode = m->private;
1277 	struct task_struct *p;
1278 
1279 	p = get_proc_task(inode);
1280 	if (!p)
1281 		return -ESRCH;
1282 	proc_sched_autogroup_show_task(p, m);
1283 
1284 	put_task_struct(p);
1285 
1286 	return 0;
1287 }
1288 
1289 static ssize_t
1290 sched_autogroup_write(struct file *file, const char __user *buf,
1291 	    size_t count, loff_t *offset)
1292 {
1293 	struct inode *inode = file->f_path.dentry->d_inode;
1294 	struct task_struct *p;
1295 	char buffer[PROC_NUMBUF];
1296 	int nice;
1297 	int err;
1298 
1299 	memset(buffer, 0, sizeof(buffer));
1300 	if (count > sizeof(buffer) - 1)
1301 		count = sizeof(buffer) - 1;
1302 	if (copy_from_user(buffer, buf, count))
1303 		return -EFAULT;
1304 
1305 	err = kstrtoint(strstrip(buffer), 0, &nice);
1306 	if (err < 0)
1307 		return err;
1308 
1309 	p = get_proc_task(inode);
1310 	if (!p)
1311 		return -ESRCH;
1312 
1313 	err = proc_sched_autogroup_set_nice(p, nice);
1314 	if (err)
1315 		count = err;
1316 
1317 	put_task_struct(p);
1318 
1319 	return count;
1320 }
1321 
1322 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1323 {
1324 	int ret;
1325 
1326 	ret = single_open(filp, sched_autogroup_show, NULL);
1327 	if (!ret) {
1328 		struct seq_file *m = filp->private_data;
1329 
1330 		m->private = inode;
1331 	}
1332 	return ret;
1333 }
1334 
1335 static const struct file_operations proc_pid_sched_autogroup_operations = {
1336 	.open		= sched_autogroup_open,
1337 	.read		= seq_read,
1338 	.write		= sched_autogroup_write,
1339 	.llseek		= seq_lseek,
1340 	.release	= single_release,
1341 };
1342 
1343 #endif /* CONFIG_SCHED_AUTOGROUP */
1344 
1345 static ssize_t comm_write(struct file *file, const char __user *buf,
1346 				size_t count, loff_t *offset)
1347 {
1348 	struct inode *inode = file->f_path.dentry->d_inode;
1349 	struct task_struct *p;
1350 	char buffer[TASK_COMM_LEN];
1351 
1352 	memset(buffer, 0, sizeof(buffer));
1353 	if (count > sizeof(buffer) - 1)
1354 		count = sizeof(buffer) - 1;
1355 	if (copy_from_user(buffer, buf, count))
1356 		return -EFAULT;
1357 
1358 	p = get_proc_task(inode);
1359 	if (!p)
1360 		return -ESRCH;
1361 
1362 	if (same_thread_group(current, p))
1363 		set_task_comm(p, buffer);
1364 	else
1365 		count = -EINVAL;
1366 
1367 	put_task_struct(p);
1368 
1369 	return count;
1370 }
1371 
1372 static int comm_show(struct seq_file *m, void *v)
1373 {
1374 	struct inode *inode = m->private;
1375 	struct task_struct *p;
1376 
1377 	p = get_proc_task(inode);
1378 	if (!p)
1379 		return -ESRCH;
1380 
1381 	task_lock(p);
1382 	seq_printf(m, "%s\n", p->comm);
1383 	task_unlock(p);
1384 
1385 	put_task_struct(p);
1386 
1387 	return 0;
1388 }
1389 
1390 static int comm_open(struct inode *inode, struct file *filp)
1391 {
1392 	return single_open(filp, comm_show, inode);
1393 }
1394 
1395 static const struct file_operations proc_pid_set_comm_operations = {
1396 	.open		= comm_open,
1397 	.read		= seq_read,
1398 	.write		= comm_write,
1399 	.llseek		= seq_lseek,
1400 	.release	= single_release,
1401 };
1402 
1403 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1404 {
1405 	struct task_struct *task;
1406 	struct mm_struct *mm;
1407 	struct file *exe_file;
1408 
1409 	task = get_proc_task(dentry->d_inode);
1410 	if (!task)
1411 		return -ENOENT;
1412 	mm = get_task_mm(task);
1413 	put_task_struct(task);
1414 	if (!mm)
1415 		return -ENOENT;
1416 	exe_file = get_mm_exe_file(mm);
1417 	mmput(mm);
1418 	if (exe_file) {
1419 		*exe_path = exe_file->f_path;
1420 		path_get(&exe_file->f_path);
1421 		fput(exe_file);
1422 		return 0;
1423 	} else
1424 		return -ENOENT;
1425 }
1426 
1427 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1428 {
1429 	struct inode *inode = dentry->d_inode;
1430 	int error = -EACCES;
1431 
1432 	/* We don't need a base pointer in the /proc filesystem */
1433 	path_put(&nd->path);
1434 
1435 	/* Are we allowed to snoop on the tasks file descriptors? */
1436 	if (!proc_fd_access_allowed(inode))
1437 		goto out;
1438 
1439 	error = PROC_I(inode)->op.proc_get_link(dentry, &nd->path);
1440 out:
1441 	return ERR_PTR(error);
1442 }
1443 
1444 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1445 {
1446 	char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1447 	char *pathname;
1448 	int len;
1449 
1450 	if (!tmp)
1451 		return -ENOMEM;
1452 
1453 	pathname = d_path(path, tmp, PAGE_SIZE);
1454 	len = PTR_ERR(pathname);
1455 	if (IS_ERR(pathname))
1456 		goto out;
1457 	len = tmp + PAGE_SIZE - 1 - pathname;
1458 
1459 	if (len > buflen)
1460 		len = buflen;
1461 	if (copy_to_user(buffer, pathname, len))
1462 		len = -EFAULT;
1463  out:
1464 	free_page((unsigned long)tmp);
1465 	return len;
1466 }
1467 
1468 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1469 {
1470 	int error = -EACCES;
1471 	struct inode *inode = dentry->d_inode;
1472 	struct path path;
1473 
1474 	/* Are we allowed to snoop on the tasks file descriptors? */
1475 	if (!proc_fd_access_allowed(inode))
1476 		goto out;
1477 
1478 	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1479 	if (error)
1480 		goto out;
1481 
1482 	error = do_proc_readlink(&path, buffer, buflen);
1483 	path_put(&path);
1484 out:
1485 	return error;
1486 }
1487 
1488 static const struct inode_operations proc_pid_link_inode_operations = {
1489 	.readlink	= proc_pid_readlink,
1490 	.follow_link	= proc_pid_follow_link,
1491 	.setattr	= proc_setattr,
1492 };
1493 
1494 
1495 /* building an inode */
1496 
1497 static int task_dumpable(struct task_struct *task)
1498 {
1499 	int dumpable = 0;
1500 	struct mm_struct *mm;
1501 
1502 	task_lock(task);
1503 	mm = task->mm;
1504 	if (mm)
1505 		dumpable = get_dumpable(mm);
1506 	task_unlock(task);
1507 	if(dumpable == 1)
1508 		return 1;
1509 	return 0;
1510 }
1511 
1512 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1513 {
1514 	struct inode * inode;
1515 	struct proc_inode *ei;
1516 	const struct cred *cred;
1517 
1518 	/* We need a new inode */
1519 
1520 	inode = new_inode(sb);
1521 	if (!inode)
1522 		goto out;
1523 
1524 	/* Common stuff */
1525 	ei = PROC_I(inode);
1526 	inode->i_ino = get_next_ino();
1527 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1528 	inode->i_op = &proc_def_inode_operations;
1529 
1530 	/*
1531 	 * grab the reference to task.
1532 	 */
1533 	ei->pid = get_task_pid(task, PIDTYPE_PID);
1534 	if (!ei->pid)
1535 		goto out_unlock;
1536 
1537 	if (task_dumpable(task)) {
1538 		rcu_read_lock();
1539 		cred = __task_cred(task);
1540 		inode->i_uid = cred->euid;
1541 		inode->i_gid = cred->egid;
1542 		rcu_read_unlock();
1543 	}
1544 	security_task_to_inode(task, inode);
1545 
1546 out:
1547 	return inode;
1548 
1549 out_unlock:
1550 	iput(inode);
1551 	return NULL;
1552 }
1553 
1554 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1555 {
1556 	struct inode *inode = dentry->d_inode;
1557 	struct task_struct *task;
1558 	const struct cred *cred;
1559 	struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1560 
1561 	generic_fillattr(inode, stat);
1562 
1563 	rcu_read_lock();
1564 	stat->uid = GLOBAL_ROOT_UID;
1565 	stat->gid = GLOBAL_ROOT_GID;
1566 	task = pid_task(proc_pid(inode), PIDTYPE_PID);
1567 	if (task) {
1568 		if (!has_pid_permissions(pid, task, 2)) {
1569 			rcu_read_unlock();
1570 			/*
1571 			 * This doesn't prevent learning whether PID exists,
1572 			 * it only makes getattr() consistent with readdir().
1573 			 */
1574 			return -ENOENT;
1575 		}
1576 		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1577 		    task_dumpable(task)) {
1578 			cred = __task_cred(task);
1579 			stat->uid = cred->euid;
1580 			stat->gid = cred->egid;
1581 		}
1582 	}
1583 	rcu_read_unlock();
1584 	return 0;
1585 }
1586 
1587 /* dentry stuff */
1588 
1589 /*
1590  *	Exceptional case: normally we are not allowed to unhash a busy
1591  * directory. In this case, however, we can do it - no aliasing problems
1592  * due to the way we treat inodes.
1593  *
1594  * Rewrite the inode's ownerships here because the owning task may have
1595  * performed a setuid(), etc.
1596  *
1597  * Before the /proc/pid/status file was created the only way to read
1598  * the effective uid of a /process was to stat /proc/pid.  Reading
1599  * /proc/pid/status is slow enough that procps and other packages
1600  * kept stating /proc/pid.  To keep the rules in /proc simple I have
1601  * made this apply to all per process world readable and executable
1602  * directories.
1603  */
1604 int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1605 {
1606 	struct inode *inode;
1607 	struct task_struct *task;
1608 	const struct cred *cred;
1609 
1610 	if (nd && nd->flags & LOOKUP_RCU)
1611 		return -ECHILD;
1612 
1613 	inode = dentry->d_inode;
1614 	task = get_proc_task(inode);
1615 
1616 	if (task) {
1617 		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1618 		    task_dumpable(task)) {
1619 			rcu_read_lock();
1620 			cred = __task_cred(task);
1621 			inode->i_uid = cred->euid;
1622 			inode->i_gid = cred->egid;
1623 			rcu_read_unlock();
1624 		} else {
1625 			inode->i_uid = GLOBAL_ROOT_UID;
1626 			inode->i_gid = GLOBAL_ROOT_GID;
1627 		}
1628 		inode->i_mode &= ~(S_ISUID | S_ISGID);
1629 		security_task_to_inode(task, inode);
1630 		put_task_struct(task);
1631 		return 1;
1632 	}
1633 	d_drop(dentry);
1634 	return 0;
1635 }
1636 
1637 static int pid_delete_dentry(const struct dentry * dentry)
1638 {
1639 	/* Is the task we represent dead?
1640 	 * If so, then don't put the dentry on the lru list,
1641 	 * kill it immediately.
1642 	 */
1643 	return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1644 }
1645 
1646 const struct dentry_operations pid_dentry_operations =
1647 {
1648 	.d_revalidate	= pid_revalidate,
1649 	.d_delete	= pid_delete_dentry,
1650 };
1651 
1652 /* Lookups */
1653 
1654 /*
1655  * Fill a directory entry.
1656  *
1657  * If possible create the dcache entry and derive our inode number and
1658  * file type from dcache entry.
1659  *
1660  * Since all of the proc inode numbers are dynamically generated, the inode
1661  * numbers do not exist until the inode is cache.  This means creating the
1662  * the dcache entry in readdir is necessary to keep the inode numbers
1663  * reported by readdir in sync with the inode numbers reported
1664  * by stat.
1665  */
1666 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1667 	const char *name, int len,
1668 	instantiate_t instantiate, struct task_struct *task, const void *ptr)
1669 {
1670 	struct dentry *child, *dir = filp->f_path.dentry;
1671 	struct inode *inode;
1672 	struct qstr qname;
1673 	ino_t ino = 0;
1674 	unsigned type = DT_UNKNOWN;
1675 
1676 	qname.name = name;
1677 	qname.len  = len;
1678 	qname.hash = full_name_hash(name, len);
1679 
1680 	child = d_lookup(dir, &qname);
1681 	if (!child) {
1682 		struct dentry *new;
1683 		new = d_alloc(dir, &qname);
1684 		if (new) {
1685 			child = instantiate(dir->d_inode, new, task, ptr);
1686 			if (child)
1687 				dput(new);
1688 			else
1689 				child = new;
1690 		}
1691 	}
1692 	if (!child || IS_ERR(child) || !child->d_inode)
1693 		goto end_instantiate;
1694 	inode = child->d_inode;
1695 	if (inode) {
1696 		ino = inode->i_ino;
1697 		type = inode->i_mode >> 12;
1698 	}
1699 	dput(child);
1700 end_instantiate:
1701 	if (!ino)
1702 		ino = find_inode_number(dir, &qname);
1703 	if (!ino)
1704 		ino = 1;
1705 	return filldir(dirent, name, len, filp->f_pos, ino, type);
1706 }
1707 
1708 static unsigned name_to_int(struct dentry *dentry)
1709 {
1710 	const char *name = dentry->d_name.name;
1711 	int len = dentry->d_name.len;
1712 	unsigned n = 0;
1713 
1714 	if (len > 1 && *name == '0')
1715 		goto out;
1716 	while (len-- > 0) {
1717 		unsigned c = *name++ - '0';
1718 		if (c > 9)
1719 			goto out;
1720 		if (n >= (~0U-9)/10)
1721 			goto out;
1722 		n *= 10;
1723 		n += c;
1724 	}
1725 	return n;
1726 out:
1727 	return ~0U;
1728 }
1729 
1730 #define PROC_FDINFO_MAX 64
1731 
1732 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1733 {
1734 	struct task_struct *task = get_proc_task(inode);
1735 	struct files_struct *files = NULL;
1736 	struct file *file;
1737 	int fd = proc_fd(inode);
1738 
1739 	if (task) {
1740 		files = get_files_struct(task);
1741 		put_task_struct(task);
1742 	}
1743 	if (files) {
1744 		/*
1745 		 * We are not taking a ref to the file structure, so we must
1746 		 * hold ->file_lock.
1747 		 */
1748 		spin_lock(&files->file_lock);
1749 		file = fcheck_files(files, fd);
1750 		if (file) {
1751 			unsigned int f_flags;
1752 			struct fdtable *fdt;
1753 
1754 			fdt = files_fdtable(files);
1755 			f_flags = file->f_flags & ~O_CLOEXEC;
1756 			if (close_on_exec(fd, fdt))
1757 				f_flags |= O_CLOEXEC;
1758 
1759 			if (path) {
1760 				*path = file->f_path;
1761 				path_get(&file->f_path);
1762 			}
1763 			if (info)
1764 				snprintf(info, PROC_FDINFO_MAX,
1765 					 "pos:\t%lli\n"
1766 					 "flags:\t0%o\n",
1767 					 (long long) file->f_pos,
1768 					 f_flags);
1769 			spin_unlock(&files->file_lock);
1770 			put_files_struct(files);
1771 			return 0;
1772 		}
1773 		spin_unlock(&files->file_lock);
1774 		put_files_struct(files);
1775 	}
1776 	return -ENOENT;
1777 }
1778 
1779 static int proc_fd_link(struct dentry *dentry, struct path *path)
1780 {
1781 	return proc_fd_info(dentry->d_inode, path, NULL);
1782 }
1783 
1784 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1785 {
1786 	struct inode *inode;
1787 	struct task_struct *task;
1788 	int fd;
1789 	struct files_struct *files;
1790 	const struct cred *cred;
1791 
1792 	if (nd && nd->flags & LOOKUP_RCU)
1793 		return -ECHILD;
1794 
1795 	inode = dentry->d_inode;
1796 	task = get_proc_task(inode);
1797 	fd = proc_fd(inode);
1798 
1799 	if (task) {
1800 		files = get_files_struct(task);
1801 		if (files) {
1802 			struct file *file;
1803 			rcu_read_lock();
1804 			file = fcheck_files(files, fd);
1805 			if (file) {
1806 				unsigned i_mode, f_mode = file->f_mode;
1807 
1808 				rcu_read_unlock();
1809 				put_files_struct(files);
1810 
1811 				if (task_dumpable(task)) {
1812 					rcu_read_lock();
1813 					cred = __task_cred(task);
1814 					inode->i_uid = cred->euid;
1815 					inode->i_gid = cred->egid;
1816 					rcu_read_unlock();
1817 				} else {
1818 					inode->i_uid = GLOBAL_ROOT_UID;
1819 					inode->i_gid = GLOBAL_ROOT_GID;
1820 				}
1821 
1822 				i_mode = S_IFLNK;
1823 				if (f_mode & FMODE_READ)
1824 					i_mode |= S_IRUSR | S_IXUSR;
1825 				if (f_mode & FMODE_WRITE)
1826 					i_mode |= S_IWUSR | S_IXUSR;
1827 				inode->i_mode = i_mode;
1828 
1829 				security_task_to_inode(task, inode);
1830 				put_task_struct(task);
1831 				return 1;
1832 			}
1833 			rcu_read_unlock();
1834 			put_files_struct(files);
1835 		}
1836 		put_task_struct(task);
1837 	}
1838 	d_drop(dentry);
1839 	return 0;
1840 }
1841 
1842 static const struct dentry_operations tid_fd_dentry_operations =
1843 {
1844 	.d_revalidate	= tid_fd_revalidate,
1845 	.d_delete	= pid_delete_dentry,
1846 };
1847 
1848 static struct dentry *proc_fd_instantiate(struct inode *dir,
1849 	struct dentry *dentry, struct task_struct *task, const void *ptr)
1850 {
1851 	unsigned fd = (unsigned long)ptr;
1852  	struct inode *inode;
1853  	struct proc_inode *ei;
1854 	struct dentry *error = ERR_PTR(-ENOENT);
1855 
1856 	inode = proc_pid_make_inode(dir->i_sb, task);
1857 	if (!inode)
1858 		goto out;
1859 	ei = PROC_I(inode);
1860 	ei->fd = fd;
1861 
1862 	inode->i_op = &proc_pid_link_inode_operations;
1863 	inode->i_size = 64;
1864 	ei->op.proc_get_link = proc_fd_link;
1865 	d_set_d_op(dentry, &tid_fd_dentry_operations);
1866 	d_add(dentry, inode);
1867 	/* Close the race of the process dying before we return the dentry */
1868 	if (tid_fd_revalidate(dentry, NULL))
1869 		error = NULL;
1870 
1871  out:
1872 	return error;
1873 }
1874 
1875 static struct dentry *proc_lookupfd_common(struct inode *dir,
1876 					   struct dentry *dentry,
1877 					   instantiate_t instantiate)
1878 {
1879 	struct task_struct *task = get_proc_task(dir);
1880 	unsigned fd = name_to_int(dentry);
1881 	struct dentry *result = ERR_PTR(-ENOENT);
1882 
1883 	if (!task)
1884 		goto out_no_task;
1885 	if (fd == ~0U)
1886 		goto out;
1887 
1888 	result = instantiate(dir, dentry, task, (void *)(unsigned long)fd);
1889 out:
1890 	put_task_struct(task);
1891 out_no_task:
1892 	return result;
1893 }
1894 
1895 static int proc_readfd_common(struct file * filp, void * dirent,
1896 			      filldir_t filldir, instantiate_t instantiate)
1897 {
1898 	struct dentry *dentry = filp->f_path.dentry;
1899 	struct inode *inode = dentry->d_inode;
1900 	struct task_struct *p = get_proc_task(inode);
1901 	unsigned int fd, ino;
1902 	int retval;
1903 	struct files_struct * files;
1904 
1905 	retval = -ENOENT;
1906 	if (!p)
1907 		goto out_no_task;
1908 	retval = 0;
1909 
1910 	fd = filp->f_pos;
1911 	switch (fd) {
1912 		case 0:
1913 			if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1914 				goto out;
1915 			filp->f_pos++;
1916 		case 1:
1917 			ino = parent_ino(dentry);
1918 			if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1919 				goto out;
1920 			filp->f_pos++;
1921 		default:
1922 			files = get_files_struct(p);
1923 			if (!files)
1924 				goto out;
1925 			rcu_read_lock();
1926 			for (fd = filp->f_pos-2;
1927 			     fd < files_fdtable(files)->max_fds;
1928 			     fd++, filp->f_pos++) {
1929 				char name[PROC_NUMBUF];
1930 				int len;
1931 				int rv;
1932 
1933 				if (!fcheck_files(files, fd))
1934 					continue;
1935 				rcu_read_unlock();
1936 
1937 				len = snprintf(name, sizeof(name), "%d", fd);
1938 				rv = proc_fill_cache(filp, dirent, filldir,
1939 						     name, len, instantiate, p,
1940 						     (void *)(unsigned long)fd);
1941 				if (rv < 0)
1942 					goto out_fd_loop;
1943 				rcu_read_lock();
1944 			}
1945 			rcu_read_unlock();
1946 out_fd_loop:
1947 			put_files_struct(files);
1948 	}
1949 out:
1950 	put_task_struct(p);
1951 out_no_task:
1952 	return retval;
1953 }
1954 
1955 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1956 				    struct nameidata *nd)
1957 {
1958 	return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1959 }
1960 
1961 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1962 {
1963 	return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1964 }
1965 
1966 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1967 				      size_t len, loff_t *ppos)
1968 {
1969 	char tmp[PROC_FDINFO_MAX];
1970 	int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1971 	if (!err)
1972 		err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1973 	return err;
1974 }
1975 
1976 static const struct file_operations proc_fdinfo_file_operations = {
1977 	.open           = nonseekable_open,
1978 	.read		= proc_fdinfo_read,
1979 	.llseek		= no_llseek,
1980 };
1981 
1982 static const struct file_operations proc_fd_operations = {
1983 	.read		= generic_read_dir,
1984 	.readdir	= proc_readfd,
1985 	.llseek		= default_llseek,
1986 };
1987 
1988 #ifdef CONFIG_CHECKPOINT_RESTORE
1989 
1990 /*
1991  * dname_to_vma_addr - maps a dentry name into two unsigned longs
1992  * which represent vma start and end addresses.
1993  */
1994 static int dname_to_vma_addr(struct dentry *dentry,
1995 			     unsigned long *start, unsigned long *end)
1996 {
1997 	if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1998 		return -EINVAL;
1999 
2000 	return 0;
2001 }
2002 
2003 static int map_files_d_revalidate(struct dentry *dentry, struct nameidata *nd)
2004 {
2005 	unsigned long vm_start, vm_end;
2006 	bool exact_vma_exists = false;
2007 	struct mm_struct *mm = NULL;
2008 	struct task_struct *task;
2009 	const struct cred *cred;
2010 	struct inode *inode;
2011 	int status = 0;
2012 
2013 	if (nd && nd->flags & LOOKUP_RCU)
2014 		return -ECHILD;
2015 
2016 	if (!capable(CAP_SYS_ADMIN)) {
2017 		status = -EACCES;
2018 		goto out_notask;
2019 	}
2020 
2021 	inode = dentry->d_inode;
2022 	task = get_proc_task(inode);
2023 	if (!task)
2024 		goto out_notask;
2025 
2026 	mm = mm_access(task, PTRACE_MODE_READ);
2027 	if (IS_ERR_OR_NULL(mm))
2028 		goto out;
2029 
2030 	if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2031 		down_read(&mm->mmap_sem);
2032 		exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
2033 		up_read(&mm->mmap_sem);
2034 	}
2035 
2036 	mmput(mm);
2037 
2038 	if (exact_vma_exists) {
2039 		if (task_dumpable(task)) {
2040 			rcu_read_lock();
2041 			cred = __task_cred(task);
2042 			inode->i_uid = cred->euid;
2043 			inode->i_gid = cred->egid;
2044 			rcu_read_unlock();
2045 		} else {
2046 			inode->i_uid = GLOBAL_ROOT_UID;
2047 			inode->i_gid = GLOBAL_ROOT_GID;
2048 		}
2049 		security_task_to_inode(task, inode);
2050 		status = 1;
2051 	}
2052 
2053 out:
2054 	put_task_struct(task);
2055 
2056 out_notask:
2057 	if (status <= 0)
2058 		d_drop(dentry);
2059 
2060 	return status;
2061 }
2062 
2063 static const struct dentry_operations tid_map_files_dentry_operations = {
2064 	.d_revalidate	= map_files_d_revalidate,
2065 	.d_delete	= pid_delete_dentry,
2066 };
2067 
2068 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
2069 {
2070 	unsigned long vm_start, vm_end;
2071 	struct vm_area_struct *vma;
2072 	struct task_struct *task;
2073 	struct mm_struct *mm;
2074 	int rc;
2075 
2076 	rc = -ENOENT;
2077 	task = get_proc_task(dentry->d_inode);
2078 	if (!task)
2079 		goto out;
2080 
2081 	mm = get_task_mm(task);
2082 	put_task_struct(task);
2083 	if (!mm)
2084 		goto out;
2085 
2086 	rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2087 	if (rc)
2088 		goto out_mmput;
2089 
2090 	down_read(&mm->mmap_sem);
2091 	vma = find_exact_vma(mm, vm_start, vm_end);
2092 	if (vma && vma->vm_file) {
2093 		*path = vma->vm_file->f_path;
2094 		path_get(path);
2095 		rc = 0;
2096 	}
2097 	up_read(&mm->mmap_sem);
2098 
2099 out_mmput:
2100 	mmput(mm);
2101 out:
2102 	return rc;
2103 }
2104 
2105 struct map_files_info {
2106 	struct file	*file;
2107 	unsigned long	len;
2108 	unsigned char	name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2109 };
2110 
2111 static struct dentry *
2112 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2113 			   struct task_struct *task, const void *ptr)
2114 {
2115 	const struct file *file = ptr;
2116 	struct proc_inode *ei;
2117 	struct inode *inode;
2118 
2119 	if (!file)
2120 		return ERR_PTR(-ENOENT);
2121 
2122 	inode = proc_pid_make_inode(dir->i_sb, task);
2123 	if (!inode)
2124 		return ERR_PTR(-ENOENT);
2125 
2126 	ei = PROC_I(inode);
2127 	ei->op.proc_get_link = proc_map_files_get_link;
2128 
2129 	inode->i_op = &proc_pid_link_inode_operations;
2130 	inode->i_size = 64;
2131 	inode->i_mode = S_IFLNK;
2132 
2133 	if (file->f_mode & FMODE_READ)
2134 		inode->i_mode |= S_IRUSR;
2135 	if (file->f_mode & FMODE_WRITE)
2136 		inode->i_mode |= S_IWUSR;
2137 
2138 	d_set_d_op(dentry, &tid_map_files_dentry_operations);
2139 	d_add(dentry, inode);
2140 
2141 	return NULL;
2142 }
2143 
2144 static struct dentry *proc_map_files_lookup(struct inode *dir,
2145 		struct dentry *dentry, struct nameidata *nd)
2146 {
2147 	unsigned long vm_start, vm_end;
2148 	struct vm_area_struct *vma;
2149 	struct task_struct *task;
2150 	struct dentry *result;
2151 	struct mm_struct *mm;
2152 
2153 	result = ERR_PTR(-EACCES);
2154 	if (!capable(CAP_SYS_ADMIN))
2155 		goto out;
2156 
2157 	result = ERR_PTR(-ENOENT);
2158 	task = get_proc_task(dir);
2159 	if (!task)
2160 		goto out;
2161 
2162 	result = ERR_PTR(-EACCES);
2163 	if (!ptrace_may_access(task, PTRACE_MODE_READ))
2164 		goto out_put_task;
2165 
2166 	result = ERR_PTR(-ENOENT);
2167 	if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2168 		goto out_put_task;
2169 
2170 	mm = get_task_mm(task);
2171 	if (!mm)
2172 		goto out_put_task;
2173 
2174 	down_read(&mm->mmap_sem);
2175 	vma = find_exact_vma(mm, vm_start, vm_end);
2176 	if (!vma)
2177 		goto out_no_vma;
2178 
2179 	result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
2180 
2181 out_no_vma:
2182 	up_read(&mm->mmap_sem);
2183 	mmput(mm);
2184 out_put_task:
2185 	put_task_struct(task);
2186 out:
2187 	return result;
2188 }
2189 
2190 static const struct inode_operations proc_map_files_inode_operations = {
2191 	.lookup		= proc_map_files_lookup,
2192 	.permission	= proc_fd_permission,
2193 	.setattr	= proc_setattr,
2194 };
2195 
2196 static int
2197 proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2198 {
2199 	struct dentry *dentry = filp->f_path.dentry;
2200 	struct inode *inode = dentry->d_inode;
2201 	struct vm_area_struct *vma;
2202 	struct task_struct *task;
2203 	struct mm_struct *mm;
2204 	ino_t ino;
2205 	int ret;
2206 
2207 	ret = -EACCES;
2208 	if (!capable(CAP_SYS_ADMIN))
2209 		goto out;
2210 
2211 	ret = -ENOENT;
2212 	task = get_proc_task(inode);
2213 	if (!task)
2214 		goto out;
2215 
2216 	ret = -EACCES;
2217 	if (!ptrace_may_access(task, PTRACE_MODE_READ))
2218 		goto out_put_task;
2219 
2220 	ret = 0;
2221 	switch (filp->f_pos) {
2222 	case 0:
2223 		ino = inode->i_ino;
2224 		if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
2225 			goto out_put_task;
2226 		filp->f_pos++;
2227 	case 1:
2228 		ino = parent_ino(dentry);
2229 		if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2230 			goto out_put_task;
2231 		filp->f_pos++;
2232 	default:
2233 	{
2234 		unsigned long nr_files, pos, i;
2235 		struct flex_array *fa = NULL;
2236 		struct map_files_info info;
2237 		struct map_files_info *p;
2238 
2239 		mm = get_task_mm(task);
2240 		if (!mm)
2241 			goto out_put_task;
2242 		down_read(&mm->mmap_sem);
2243 
2244 		nr_files = 0;
2245 
2246 		/*
2247 		 * We need two passes here:
2248 		 *
2249 		 *  1) Collect vmas of mapped files with mmap_sem taken
2250 		 *  2) Release mmap_sem and instantiate entries
2251 		 *
2252 		 * otherwise we get lockdep complained, since filldir()
2253 		 * routine might require mmap_sem taken in might_fault().
2254 		 */
2255 
2256 		for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2257 			if (vma->vm_file && ++pos > filp->f_pos)
2258 				nr_files++;
2259 		}
2260 
2261 		if (nr_files) {
2262 			fa = flex_array_alloc(sizeof(info), nr_files,
2263 						GFP_KERNEL);
2264 			if (!fa || flex_array_prealloc(fa, 0, nr_files,
2265 							GFP_KERNEL)) {
2266 				ret = -ENOMEM;
2267 				if (fa)
2268 					flex_array_free(fa);
2269 				up_read(&mm->mmap_sem);
2270 				mmput(mm);
2271 				goto out_put_task;
2272 			}
2273 			for (i = 0, vma = mm->mmap, pos = 2; vma;
2274 					vma = vma->vm_next) {
2275 				if (!vma->vm_file)
2276 					continue;
2277 				if (++pos <= filp->f_pos)
2278 					continue;
2279 
2280 				get_file(vma->vm_file);
2281 				info.file = vma->vm_file;
2282 				info.len = snprintf(info.name,
2283 						sizeof(info.name), "%lx-%lx",
2284 						vma->vm_start, vma->vm_end);
2285 				if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2286 					BUG();
2287 			}
2288 		}
2289 		up_read(&mm->mmap_sem);
2290 
2291 		for (i = 0; i < nr_files; i++) {
2292 			p = flex_array_get(fa, i);
2293 			ret = proc_fill_cache(filp, dirent, filldir,
2294 					      p->name, p->len,
2295 					      proc_map_files_instantiate,
2296 					      task, p->file);
2297 			if (ret)
2298 				break;
2299 			filp->f_pos++;
2300 			fput(p->file);
2301 		}
2302 		for (; i < nr_files; i++) {
2303 			/*
2304 			 * In case of error don't forget
2305 			 * to put rest of file refs.
2306 			 */
2307 			p = flex_array_get(fa, i);
2308 			fput(p->file);
2309 		}
2310 		if (fa)
2311 			flex_array_free(fa);
2312 		mmput(mm);
2313 	}
2314 	}
2315 
2316 out_put_task:
2317 	put_task_struct(task);
2318 out:
2319 	return ret;
2320 }
2321 
2322 static const struct file_operations proc_map_files_operations = {
2323 	.read		= generic_read_dir,
2324 	.readdir	= proc_map_files_readdir,
2325 	.llseek		= default_llseek,
2326 };
2327 
2328 #endif /* CONFIG_CHECKPOINT_RESTORE */
2329 
2330 /*
2331  * /proc/pid/fd needs a special permission handler so that a process can still
2332  * access /proc/self/fd after it has executed a setuid().
2333  */
2334 static int proc_fd_permission(struct inode *inode, int mask)
2335 {
2336 	int rv = generic_permission(inode, mask);
2337 	if (rv == 0)
2338 		return 0;
2339 	if (task_pid(current) == proc_pid(inode))
2340 		rv = 0;
2341 	return rv;
2342 }
2343 
2344 /*
2345  * proc directories can do almost nothing..
2346  */
2347 static const struct inode_operations proc_fd_inode_operations = {
2348 	.lookup		= proc_lookupfd,
2349 	.permission	= proc_fd_permission,
2350 	.setattr	= proc_setattr,
2351 };
2352 
2353 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2354 	struct dentry *dentry, struct task_struct *task, const void *ptr)
2355 {
2356 	unsigned fd = (unsigned long)ptr;
2357  	struct inode *inode;
2358  	struct proc_inode *ei;
2359 	struct dentry *error = ERR_PTR(-ENOENT);
2360 
2361 	inode = proc_pid_make_inode(dir->i_sb, task);
2362 	if (!inode)
2363 		goto out;
2364 	ei = PROC_I(inode);
2365 	ei->fd = fd;
2366 	inode->i_mode = S_IFREG | S_IRUSR;
2367 	inode->i_fop = &proc_fdinfo_file_operations;
2368 	d_set_d_op(dentry, &tid_fd_dentry_operations);
2369 	d_add(dentry, inode);
2370 	/* Close the race of the process dying before we return the dentry */
2371 	if (tid_fd_revalidate(dentry, NULL))
2372 		error = NULL;
2373 
2374  out:
2375 	return error;
2376 }
2377 
2378 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2379 					struct dentry *dentry,
2380 					struct nameidata *nd)
2381 {
2382 	return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2383 }
2384 
2385 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2386 {
2387 	return proc_readfd_common(filp, dirent, filldir,
2388 				  proc_fdinfo_instantiate);
2389 }
2390 
2391 static const struct file_operations proc_fdinfo_operations = {
2392 	.read		= generic_read_dir,
2393 	.readdir	= proc_readfdinfo,
2394 	.llseek		= default_llseek,
2395 };
2396 
2397 /*
2398  * proc directories can do almost nothing..
2399  */
2400 static const struct inode_operations proc_fdinfo_inode_operations = {
2401 	.lookup		= proc_lookupfdinfo,
2402 	.setattr	= proc_setattr,
2403 };
2404 
2405 
2406 static struct dentry *proc_pident_instantiate(struct inode *dir,
2407 	struct dentry *dentry, struct task_struct *task, const void *ptr)
2408 {
2409 	const struct pid_entry *p = ptr;
2410 	struct inode *inode;
2411 	struct proc_inode *ei;
2412 	struct dentry *error = ERR_PTR(-ENOENT);
2413 
2414 	inode = proc_pid_make_inode(dir->i_sb, task);
2415 	if (!inode)
2416 		goto out;
2417 
2418 	ei = PROC_I(inode);
2419 	inode->i_mode = p->mode;
2420 	if (S_ISDIR(inode->i_mode))
2421 		set_nlink(inode, 2);	/* Use getattr to fix if necessary */
2422 	if (p->iop)
2423 		inode->i_op = p->iop;
2424 	if (p->fop)
2425 		inode->i_fop = p->fop;
2426 	ei->op = p->op;
2427 	d_set_d_op(dentry, &pid_dentry_operations);
2428 	d_add(dentry, inode);
2429 	/* Close the race of the process dying before we return the dentry */
2430 	if (pid_revalidate(dentry, NULL))
2431 		error = NULL;
2432 out:
2433 	return error;
2434 }
2435 
2436 static struct dentry *proc_pident_lookup(struct inode *dir,
2437 					 struct dentry *dentry,
2438 					 const struct pid_entry *ents,
2439 					 unsigned int nents)
2440 {
2441 	struct dentry *error;
2442 	struct task_struct *task = get_proc_task(dir);
2443 	const struct pid_entry *p, *last;
2444 
2445 	error = ERR_PTR(-ENOENT);
2446 
2447 	if (!task)
2448 		goto out_no_task;
2449 
2450 	/*
2451 	 * Yes, it does not scale. And it should not. Don't add
2452 	 * new entries into /proc/<tgid>/ without very good reasons.
2453 	 */
2454 	last = &ents[nents - 1];
2455 	for (p = ents; p <= last; p++) {
2456 		if (p->len != dentry->d_name.len)
2457 			continue;
2458 		if (!memcmp(dentry->d_name.name, p->name, p->len))
2459 			break;
2460 	}
2461 	if (p > last)
2462 		goto out;
2463 
2464 	error = proc_pident_instantiate(dir, dentry, task, p);
2465 out:
2466 	put_task_struct(task);
2467 out_no_task:
2468 	return error;
2469 }
2470 
2471 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2472 	filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2473 {
2474 	return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2475 				proc_pident_instantiate, task, p);
2476 }
2477 
2478 static int proc_pident_readdir(struct file *filp,
2479 		void *dirent, filldir_t filldir,
2480 		const struct pid_entry *ents, unsigned int nents)
2481 {
2482 	int i;
2483 	struct dentry *dentry = filp->f_path.dentry;
2484 	struct inode *inode = dentry->d_inode;
2485 	struct task_struct *task = get_proc_task(inode);
2486 	const struct pid_entry *p, *last;
2487 	ino_t ino;
2488 	int ret;
2489 
2490 	ret = -ENOENT;
2491 	if (!task)
2492 		goto out_no_task;
2493 
2494 	ret = 0;
2495 	i = filp->f_pos;
2496 	switch (i) {
2497 	case 0:
2498 		ino = inode->i_ino;
2499 		if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2500 			goto out;
2501 		i++;
2502 		filp->f_pos++;
2503 		/* fall through */
2504 	case 1:
2505 		ino = parent_ino(dentry);
2506 		if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2507 			goto out;
2508 		i++;
2509 		filp->f_pos++;
2510 		/* fall through */
2511 	default:
2512 		i -= 2;
2513 		if (i >= nents) {
2514 			ret = 1;
2515 			goto out;
2516 		}
2517 		p = ents + i;
2518 		last = &ents[nents - 1];
2519 		while (p <= last) {
2520 			if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2521 				goto out;
2522 			filp->f_pos++;
2523 			p++;
2524 		}
2525 	}
2526 
2527 	ret = 1;
2528 out:
2529 	put_task_struct(task);
2530 out_no_task:
2531 	return ret;
2532 }
2533 
2534 #ifdef CONFIG_SECURITY
2535 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2536 				  size_t count, loff_t *ppos)
2537 {
2538 	struct inode * inode = file->f_path.dentry->d_inode;
2539 	char *p = NULL;
2540 	ssize_t length;
2541 	struct task_struct *task = get_proc_task(inode);
2542 
2543 	if (!task)
2544 		return -ESRCH;
2545 
2546 	length = security_getprocattr(task,
2547 				      (char*)file->f_path.dentry->d_name.name,
2548 				      &p);
2549 	put_task_struct(task);
2550 	if (length > 0)
2551 		length = simple_read_from_buffer(buf, count, ppos, p, length);
2552 	kfree(p);
2553 	return length;
2554 }
2555 
2556 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2557 				   size_t count, loff_t *ppos)
2558 {
2559 	struct inode * inode = file->f_path.dentry->d_inode;
2560 	char *page;
2561 	ssize_t length;
2562 	struct task_struct *task = get_proc_task(inode);
2563 
2564 	length = -ESRCH;
2565 	if (!task)
2566 		goto out_no_task;
2567 	if (count > PAGE_SIZE)
2568 		count = PAGE_SIZE;
2569 
2570 	/* No partial writes. */
2571 	length = -EINVAL;
2572 	if (*ppos != 0)
2573 		goto out;
2574 
2575 	length = -ENOMEM;
2576 	page = (char*)__get_free_page(GFP_TEMPORARY);
2577 	if (!page)
2578 		goto out;
2579 
2580 	length = -EFAULT;
2581 	if (copy_from_user(page, buf, count))
2582 		goto out_free;
2583 
2584 	/* Guard against adverse ptrace interaction */
2585 	length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2586 	if (length < 0)
2587 		goto out_free;
2588 
2589 	length = security_setprocattr(task,
2590 				      (char*)file->f_path.dentry->d_name.name,
2591 				      (void*)page, count);
2592 	mutex_unlock(&task->signal->cred_guard_mutex);
2593 out_free:
2594 	free_page((unsigned long) page);
2595 out:
2596 	put_task_struct(task);
2597 out_no_task:
2598 	return length;
2599 }
2600 
2601 static const struct file_operations proc_pid_attr_operations = {
2602 	.read		= proc_pid_attr_read,
2603 	.write		= proc_pid_attr_write,
2604 	.llseek		= generic_file_llseek,
2605 };
2606 
2607 static const struct pid_entry attr_dir_stuff[] = {
2608 	REG("current",    S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2609 	REG("prev",       S_IRUGO,	   proc_pid_attr_operations),
2610 	REG("exec",       S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2611 	REG("fscreate",   S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2612 	REG("keycreate",  S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2613 	REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2614 };
2615 
2616 static int proc_attr_dir_readdir(struct file * filp,
2617 			     void * dirent, filldir_t filldir)
2618 {
2619 	return proc_pident_readdir(filp,dirent,filldir,
2620 				   attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2621 }
2622 
2623 static const struct file_operations proc_attr_dir_operations = {
2624 	.read		= generic_read_dir,
2625 	.readdir	= proc_attr_dir_readdir,
2626 	.llseek		= default_llseek,
2627 };
2628 
2629 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2630 				struct dentry *dentry, struct nameidata *nd)
2631 {
2632 	return proc_pident_lookup(dir, dentry,
2633 				  attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2634 }
2635 
2636 static const struct inode_operations proc_attr_dir_inode_operations = {
2637 	.lookup		= proc_attr_dir_lookup,
2638 	.getattr	= pid_getattr,
2639 	.setattr	= proc_setattr,
2640 };
2641 
2642 #endif
2643 
2644 #ifdef CONFIG_ELF_CORE
2645 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2646 					 size_t count, loff_t *ppos)
2647 {
2648 	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2649 	struct mm_struct *mm;
2650 	char buffer[PROC_NUMBUF];
2651 	size_t len;
2652 	int ret;
2653 
2654 	if (!task)
2655 		return -ESRCH;
2656 
2657 	ret = 0;
2658 	mm = get_task_mm(task);
2659 	if (mm) {
2660 		len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2661 			       ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2662 				MMF_DUMP_FILTER_SHIFT));
2663 		mmput(mm);
2664 		ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2665 	}
2666 
2667 	put_task_struct(task);
2668 
2669 	return ret;
2670 }
2671 
2672 static ssize_t proc_coredump_filter_write(struct file *file,
2673 					  const char __user *buf,
2674 					  size_t count,
2675 					  loff_t *ppos)
2676 {
2677 	struct task_struct *task;
2678 	struct mm_struct *mm;
2679 	char buffer[PROC_NUMBUF], *end;
2680 	unsigned int val;
2681 	int ret;
2682 	int i;
2683 	unsigned long mask;
2684 
2685 	ret = -EFAULT;
2686 	memset(buffer, 0, sizeof(buffer));
2687 	if (count > sizeof(buffer) - 1)
2688 		count = sizeof(buffer) - 1;
2689 	if (copy_from_user(buffer, buf, count))
2690 		goto out_no_task;
2691 
2692 	ret = -EINVAL;
2693 	val = (unsigned int)simple_strtoul(buffer, &end, 0);
2694 	if (*end == '\n')
2695 		end++;
2696 	if (end - buffer == 0)
2697 		goto out_no_task;
2698 
2699 	ret = -ESRCH;
2700 	task = get_proc_task(file->f_dentry->d_inode);
2701 	if (!task)
2702 		goto out_no_task;
2703 
2704 	ret = end - buffer;
2705 	mm = get_task_mm(task);
2706 	if (!mm)
2707 		goto out_no_mm;
2708 
2709 	for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2710 		if (val & mask)
2711 			set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2712 		else
2713 			clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2714 	}
2715 
2716 	mmput(mm);
2717  out_no_mm:
2718 	put_task_struct(task);
2719  out_no_task:
2720 	return ret;
2721 }
2722 
2723 static const struct file_operations proc_coredump_filter_operations = {
2724 	.read		= proc_coredump_filter_read,
2725 	.write		= proc_coredump_filter_write,
2726 	.llseek		= generic_file_llseek,
2727 };
2728 #endif
2729 
2730 /*
2731  * /proc/self:
2732  */
2733 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2734 			      int buflen)
2735 {
2736 	struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2737 	pid_t tgid = task_tgid_nr_ns(current, ns);
2738 	char tmp[PROC_NUMBUF];
2739 	if (!tgid)
2740 		return -ENOENT;
2741 	sprintf(tmp, "%d", tgid);
2742 	return vfs_readlink(dentry,buffer,buflen,tmp);
2743 }
2744 
2745 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2746 {
2747 	struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2748 	pid_t tgid = task_tgid_nr_ns(current, ns);
2749 	char *name = ERR_PTR(-ENOENT);
2750 	if (tgid) {
2751 		name = __getname();
2752 		if (!name)
2753 			name = ERR_PTR(-ENOMEM);
2754 		else
2755 			sprintf(name, "%d", tgid);
2756 	}
2757 	nd_set_link(nd, name);
2758 	return NULL;
2759 }
2760 
2761 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2762 				void *cookie)
2763 {
2764 	char *s = nd_get_link(nd);
2765 	if (!IS_ERR(s))
2766 		__putname(s);
2767 }
2768 
2769 static const struct inode_operations proc_self_inode_operations = {
2770 	.readlink	= proc_self_readlink,
2771 	.follow_link	= proc_self_follow_link,
2772 	.put_link	= proc_self_put_link,
2773 };
2774 
2775 /*
2776  * proc base
2777  *
2778  * These are the directory entries in the root directory of /proc
2779  * that properly belong to the /proc filesystem, as they describe
2780  * describe something that is process related.
2781  */
2782 static const struct pid_entry proc_base_stuff[] = {
2783 	NOD("self", S_IFLNK|S_IRWXUGO,
2784 		&proc_self_inode_operations, NULL, {}),
2785 };
2786 
2787 static struct dentry *proc_base_instantiate(struct inode *dir,
2788 	struct dentry *dentry, struct task_struct *task, const void *ptr)
2789 {
2790 	const struct pid_entry *p = ptr;
2791 	struct inode *inode;
2792 	struct proc_inode *ei;
2793 	struct dentry *error;
2794 
2795 	/* Allocate the inode */
2796 	error = ERR_PTR(-ENOMEM);
2797 	inode = new_inode(dir->i_sb);
2798 	if (!inode)
2799 		goto out;
2800 
2801 	/* Initialize the inode */
2802 	ei = PROC_I(inode);
2803 	inode->i_ino = get_next_ino();
2804 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2805 
2806 	/*
2807 	 * grab the reference to the task.
2808 	 */
2809 	ei->pid = get_task_pid(task, PIDTYPE_PID);
2810 	if (!ei->pid)
2811 		goto out_iput;
2812 
2813 	inode->i_mode = p->mode;
2814 	if (S_ISDIR(inode->i_mode))
2815 		set_nlink(inode, 2);
2816 	if (S_ISLNK(inode->i_mode))
2817 		inode->i_size = 64;
2818 	if (p->iop)
2819 		inode->i_op = p->iop;
2820 	if (p->fop)
2821 		inode->i_fop = p->fop;
2822 	ei->op = p->op;
2823 	d_add(dentry, inode);
2824 	error = NULL;
2825 out:
2826 	return error;
2827 out_iput:
2828 	iput(inode);
2829 	goto out;
2830 }
2831 
2832 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2833 {
2834 	struct dentry *error;
2835 	struct task_struct *task = get_proc_task(dir);
2836 	const struct pid_entry *p, *last;
2837 
2838 	error = ERR_PTR(-ENOENT);
2839 
2840 	if (!task)
2841 		goto out_no_task;
2842 
2843 	/* Lookup the directory entry */
2844 	last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2845 	for (p = proc_base_stuff; p <= last; p++) {
2846 		if (p->len != dentry->d_name.len)
2847 			continue;
2848 		if (!memcmp(dentry->d_name.name, p->name, p->len))
2849 			break;
2850 	}
2851 	if (p > last)
2852 		goto out;
2853 
2854 	error = proc_base_instantiate(dir, dentry, task, p);
2855 
2856 out:
2857 	put_task_struct(task);
2858 out_no_task:
2859 	return error;
2860 }
2861 
2862 static int proc_base_fill_cache(struct file *filp, void *dirent,
2863 	filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2864 {
2865 	return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2866 				proc_base_instantiate, task, p);
2867 }
2868 
2869 #ifdef CONFIG_TASK_IO_ACCOUNTING
2870 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2871 {
2872 	struct task_io_accounting acct = task->ioac;
2873 	unsigned long flags;
2874 	int result;
2875 
2876 	result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2877 	if (result)
2878 		return result;
2879 
2880 	if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2881 		result = -EACCES;
2882 		goto out_unlock;
2883 	}
2884 
2885 	if (whole && lock_task_sighand(task, &flags)) {
2886 		struct task_struct *t = task;
2887 
2888 		task_io_accounting_add(&acct, &task->signal->ioac);
2889 		while_each_thread(task, t)
2890 			task_io_accounting_add(&acct, &t->ioac);
2891 
2892 		unlock_task_sighand(task, &flags);
2893 	}
2894 	result = sprintf(buffer,
2895 			"rchar: %llu\n"
2896 			"wchar: %llu\n"
2897 			"syscr: %llu\n"
2898 			"syscw: %llu\n"
2899 			"read_bytes: %llu\n"
2900 			"write_bytes: %llu\n"
2901 			"cancelled_write_bytes: %llu\n",
2902 			(unsigned long long)acct.rchar,
2903 			(unsigned long long)acct.wchar,
2904 			(unsigned long long)acct.syscr,
2905 			(unsigned long long)acct.syscw,
2906 			(unsigned long long)acct.read_bytes,
2907 			(unsigned long long)acct.write_bytes,
2908 			(unsigned long long)acct.cancelled_write_bytes);
2909 out_unlock:
2910 	mutex_unlock(&task->signal->cred_guard_mutex);
2911 	return result;
2912 }
2913 
2914 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2915 {
2916 	return do_io_accounting(task, buffer, 0);
2917 }
2918 
2919 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2920 {
2921 	return do_io_accounting(task, buffer, 1);
2922 }
2923 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2924 
2925 #ifdef CONFIG_USER_NS
2926 static int proc_id_map_open(struct inode *inode, struct file *file,
2927 	struct seq_operations *seq_ops)
2928 {
2929 	struct user_namespace *ns = NULL;
2930 	struct task_struct *task;
2931 	struct seq_file *seq;
2932 	int ret = -EINVAL;
2933 
2934 	task = get_proc_task(inode);
2935 	if (task) {
2936 		rcu_read_lock();
2937 		ns = get_user_ns(task_cred_xxx(task, user_ns));
2938 		rcu_read_unlock();
2939 		put_task_struct(task);
2940 	}
2941 	if (!ns)
2942 		goto err;
2943 
2944 	ret = seq_open(file, seq_ops);
2945 	if (ret)
2946 		goto err_put_ns;
2947 
2948 	seq = file->private_data;
2949 	seq->private = ns;
2950 
2951 	return 0;
2952 err_put_ns:
2953 	put_user_ns(ns);
2954 err:
2955 	return ret;
2956 }
2957 
2958 static int proc_id_map_release(struct inode *inode, struct file *file)
2959 {
2960 	struct seq_file *seq = file->private_data;
2961 	struct user_namespace *ns = seq->private;
2962 	put_user_ns(ns);
2963 	return seq_release(inode, file);
2964 }
2965 
2966 static int proc_uid_map_open(struct inode *inode, struct file *file)
2967 {
2968 	return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2969 }
2970 
2971 static int proc_gid_map_open(struct inode *inode, struct file *file)
2972 {
2973 	return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2974 }
2975 
2976 static const struct file_operations proc_uid_map_operations = {
2977 	.open		= proc_uid_map_open,
2978 	.write		= proc_uid_map_write,
2979 	.read		= seq_read,
2980 	.llseek		= seq_lseek,
2981 	.release	= proc_id_map_release,
2982 };
2983 
2984 static const struct file_operations proc_gid_map_operations = {
2985 	.open		= proc_gid_map_open,
2986 	.write		= proc_gid_map_write,
2987 	.read		= seq_read,
2988 	.llseek		= seq_lseek,
2989 	.release	= proc_id_map_release,
2990 };
2991 #endif /* CONFIG_USER_NS */
2992 
2993 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2994 				struct pid *pid, struct task_struct *task)
2995 {
2996 	int err = lock_trace(task);
2997 	if (!err) {
2998 		seq_printf(m, "%08x\n", task->personality);
2999 		unlock_trace(task);
3000 	}
3001 	return err;
3002 }
3003 
3004 /*
3005  * Thread groups
3006  */
3007 static const struct file_operations proc_task_operations;
3008 static const struct inode_operations proc_task_inode_operations;
3009 
3010 static const struct pid_entry tgid_base_stuff[] = {
3011 	DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3012 	DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3013 #ifdef CONFIG_CHECKPOINT_RESTORE
3014 	DIR("map_files",  S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3015 #endif
3016 	DIR("fdinfo",     S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3017 	DIR("ns",	  S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3018 #ifdef CONFIG_NET
3019 	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3020 #endif
3021 	REG("environ",    S_IRUSR, proc_environ_operations),
3022 	INF("auxv",       S_IRUSR, proc_pid_auxv),
3023 	ONE("status",     S_IRUGO, proc_pid_status),
3024 	ONE("personality", S_IRUGO, proc_pid_personality),
3025 	INF("limits",	  S_IRUGO, proc_pid_limits),
3026 #ifdef CONFIG_SCHED_DEBUG
3027 	REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3028 #endif
3029 #ifdef CONFIG_SCHED_AUTOGROUP
3030 	REG("autogroup",  S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3031 #endif
3032 	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3033 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3034 	INF("syscall",    S_IRUGO, proc_pid_syscall),
3035 #endif
3036 	INF("cmdline",    S_IRUGO, proc_pid_cmdline),
3037 	ONE("stat",       S_IRUGO, proc_tgid_stat),
3038 	ONE("statm",      S_IRUGO, proc_pid_statm),
3039 	REG("maps",       S_IRUGO, proc_pid_maps_operations),
3040 #ifdef CONFIG_NUMA
3041 	REG("numa_maps",  S_IRUGO, proc_pid_numa_maps_operations),
3042 #endif
3043 	REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
3044 	LNK("cwd",        proc_cwd_link),
3045 	LNK("root",       proc_root_link),
3046 	LNK("exe",        proc_exe_link),
3047 	REG("mounts",     S_IRUGO, proc_mounts_operations),
3048 	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3049 	REG("mountstats", S_IRUSR, proc_mountstats_operations),
3050 #ifdef CONFIG_PROC_PAGE_MONITOR
3051 	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3052 	REG("smaps",      S_IRUGO, proc_pid_smaps_operations),
3053 	REG("pagemap",    S_IRUGO, proc_pagemap_operations),
3054 #endif
3055 #ifdef CONFIG_SECURITY
3056 	DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3057 #endif
3058 #ifdef CONFIG_KALLSYMS
3059 	INF("wchan",      S_IRUGO, proc_pid_wchan),
3060 #endif
3061 #ifdef CONFIG_STACKTRACE
3062 	ONE("stack",      S_IRUGO, proc_pid_stack),
3063 #endif
3064 #ifdef CONFIG_SCHEDSTATS
3065 	INF("schedstat",  S_IRUGO, proc_pid_schedstat),
3066 #endif
3067 #ifdef CONFIG_LATENCYTOP
3068 	REG("latency",  S_IRUGO, proc_lstats_operations),
3069 #endif
3070 #ifdef CONFIG_PROC_PID_CPUSET
3071 	REG("cpuset",     S_IRUGO, proc_cpuset_operations),
3072 #endif
3073 #ifdef CONFIG_CGROUPS
3074 	REG("cgroup",  S_IRUGO, proc_cgroup_operations),
3075 #endif
3076 	INF("oom_score",  S_IRUGO, proc_oom_score),
3077 	REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3078 	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3079 #ifdef CONFIG_AUDITSYSCALL
3080 	REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
3081 	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3082 #endif
3083 #ifdef CONFIG_FAULT_INJECTION
3084 	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3085 #endif
3086 #ifdef CONFIG_ELF_CORE
3087 	REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3088 #endif
3089 #ifdef CONFIG_TASK_IO_ACCOUNTING
3090 	INF("io",	S_IRUSR, proc_tgid_io_accounting),
3091 #endif
3092 #ifdef CONFIG_HARDWALL
3093 	INF("hardwall",   S_IRUGO, proc_pid_hardwall),
3094 #endif
3095 #ifdef CONFIG_USER_NS
3096 	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3097 	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3098 #endif
3099 };
3100 
3101 static int proc_tgid_base_readdir(struct file * filp,
3102 			     void * dirent, filldir_t filldir)
3103 {
3104 	return proc_pident_readdir(filp,dirent,filldir,
3105 				   tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
3106 }
3107 
3108 static const struct file_operations proc_tgid_base_operations = {
3109 	.read		= generic_read_dir,
3110 	.readdir	= proc_tgid_base_readdir,
3111 	.llseek		= default_llseek,
3112 };
3113 
3114 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3115 	return proc_pident_lookup(dir, dentry,
3116 				  tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3117 }
3118 
3119 static const struct inode_operations proc_tgid_base_inode_operations = {
3120 	.lookup		= proc_tgid_base_lookup,
3121 	.getattr	= pid_getattr,
3122 	.setattr	= proc_setattr,
3123 	.permission	= proc_pid_permission,
3124 };
3125 
3126 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
3127 {
3128 	struct dentry *dentry, *leader, *dir;
3129 	char buf[PROC_NUMBUF];
3130 	struct qstr name;
3131 
3132 	name.name = buf;
3133 	name.len = snprintf(buf, sizeof(buf), "%d", pid);
3134 	dentry = d_hash_and_lookup(mnt->mnt_root, &name);
3135 	if (dentry) {
3136 		shrink_dcache_parent(dentry);
3137 		d_drop(dentry);
3138 		dput(dentry);
3139 	}
3140 
3141 	name.name = buf;
3142 	name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3143 	leader = d_hash_and_lookup(mnt->mnt_root, &name);
3144 	if (!leader)
3145 		goto out;
3146 
3147 	name.name = "task";
3148 	name.len = strlen(name.name);
3149 	dir = d_hash_and_lookup(leader, &name);
3150 	if (!dir)
3151 		goto out_put_leader;
3152 
3153 	name.name = buf;
3154 	name.len = snprintf(buf, sizeof(buf), "%d", pid);
3155 	dentry = d_hash_and_lookup(dir, &name);
3156 	if (dentry) {
3157 		shrink_dcache_parent(dentry);
3158 		d_drop(dentry);
3159 		dput(dentry);
3160 	}
3161 
3162 	dput(dir);
3163 out_put_leader:
3164 	dput(leader);
3165 out:
3166 	return;
3167 }
3168 
3169 /**
3170  * proc_flush_task -  Remove dcache entries for @task from the /proc dcache.
3171  * @task: task that should be flushed.
3172  *
3173  * When flushing dentries from proc, one needs to flush them from global
3174  * proc (proc_mnt) and from all the namespaces' procs this task was seen
3175  * in. This call is supposed to do all of this job.
3176  *
3177  * Looks in the dcache for
3178  * /proc/@pid
3179  * /proc/@tgid/task/@pid
3180  * if either directory is present flushes it and all of it'ts children
3181  * from the dcache.
3182  *
3183  * It is safe and reasonable to cache /proc entries for a task until
3184  * that task exits.  After that they just clog up the dcache with
3185  * useless entries, possibly causing useful dcache entries to be
3186  * flushed instead.  This routine is proved to flush those useless
3187  * dcache entries at process exit time.
3188  *
3189  * NOTE: This routine is just an optimization so it does not guarantee
3190  *       that no dcache entries will exist at process exit time it
3191  *       just makes it very unlikely that any will persist.
3192  */
3193 
3194 void proc_flush_task(struct task_struct *task)
3195 {
3196 	int i;
3197 	struct pid *pid, *tgid;
3198 	struct upid *upid;
3199 
3200 	pid = task_pid(task);
3201 	tgid = task_tgid(task);
3202 
3203 	for (i = 0; i <= pid->level; i++) {
3204 		upid = &pid->numbers[i];
3205 		proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3206 					tgid->numbers[i].nr);
3207 	}
3208 
3209 	upid = &pid->numbers[pid->level];
3210 	if (upid->nr == 1)
3211 		pid_ns_release_proc(upid->ns);
3212 }
3213 
3214 static struct dentry *proc_pid_instantiate(struct inode *dir,
3215 					   struct dentry * dentry,
3216 					   struct task_struct *task, const void *ptr)
3217 {
3218 	struct dentry *error = ERR_PTR(-ENOENT);
3219 	struct inode *inode;
3220 
3221 	inode = proc_pid_make_inode(dir->i_sb, task);
3222 	if (!inode)
3223 		goto out;
3224 
3225 	inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3226 	inode->i_op = &proc_tgid_base_inode_operations;
3227 	inode->i_fop = &proc_tgid_base_operations;
3228 	inode->i_flags|=S_IMMUTABLE;
3229 
3230 	set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3231 						  ARRAY_SIZE(tgid_base_stuff)));
3232 
3233 	d_set_d_op(dentry, &pid_dentry_operations);
3234 
3235 	d_add(dentry, inode);
3236 	/* Close the race of the process dying before we return the dentry */
3237 	if (pid_revalidate(dentry, NULL))
3238 		error = NULL;
3239 out:
3240 	return error;
3241 }
3242 
3243 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3244 {
3245 	struct dentry *result;
3246 	struct task_struct *task;
3247 	unsigned tgid;
3248 	struct pid_namespace *ns;
3249 
3250 	result = proc_base_lookup(dir, dentry);
3251 	if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3252 		goto out;
3253 
3254 	tgid = name_to_int(dentry);
3255 	if (tgid == ~0U)
3256 		goto out;
3257 
3258 	ns = dentry->d_sb->s_fs_info;
3259 	rcu_read_lock();
3260 	task = find_task_by_pid_ns(tgid, ns);
3261 	if (task)
3262 		get_task_struct(task);
3263 	rcu_read_unlock();
3264 	if (!task)
3265 		goto out;
3266 
3267 	result = proc_pid_instantiate(dir, dentry, task, NULL);
3268 	put_task_struct(task);
3269 out:
3270 	return result;
3271 }
3272 
3273 /*
3274  * Find the first task with tgid >= tgid
3275  *
3276  */
3277 struct tgid_iter {
3278 	unsigned int tgid;
3279 	struct task_struct *task;
3280 };
3281 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3282 {
3283 	struct pid *pid;
3284 
3285 	if (iter.task)
3286 		put_task_struct(iter.task);
3287 	rcu_read_lock();
3288 retry:
3289 	iter.task = NULL;
3290 	pid = find_ge_pid(iter.tgid, ns);
3291 	if (pid) {
3292 		iter.tgid = pid_nr_ns(pid, ns);
3293 		iter.task = pid_task(pid, PIDTYPE_PID);
3294 		/* What we to know is if the pid we have find is the
3295 		 * pid of a thread_group_leader.  Testing for task
3296 		 * being a thread_group_leader is the obvious thing
3297 		 * todo but there is a window when it fails, due to
3298 		 * the pid transfer logic in de_thread.
3299 		 *
3300 		 * So we perform the straight forward test of seeing
3301 		 * if the pid we have found is the pid of a thread
3302 		 * group leader, and don't worry if the task we have
3303 		 * found doesn't happen to be a thread group leader.
3304 		 * As we don't care in the case of readdir.
3305 		 */
3306 		if (!iter.task || !has_group_leader_pid(iter.task)) {
3307 			iter.tgid += 1;
3308 			goto retry;
3309 		}
3310 		get_task_struct(iter.task);
3311 	}
3312 	rcu_read_unlock();
3313 	return iter;
3314 }
3315 
3316 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3317 
3318 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3319 	struct tgid_iter iter)
3320 {
3321 	char name[PROC_NUMBUF];
3322 	int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3323 	return proc_fill_cache(filp, dirent, filldir, name, len,
3324 				proc_pid_instantiate, iter.task, NULL);
3325 }
3326 
3327 static int fake_filldir(void *buf, const char *name, int namelen,
3328 			loff_t offset, u64 ino, unsigned d_type)
3329 {
3330 	return 0;
3331 }
3332 
3333 /* for the /proc/ directory itself, after non-process stuff has been done */
3334 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3335 {
3336 	unsigned int nr;
3337 	struct task_struct *reaper;
3338 	struct tgid_iter iter;
3339 	struct pid_namespace *ns;
3340 	filldir_t __filldir;
3341 
3342 	if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3343 		goto out_no_task;
3344 	nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3345 
3346 	reaper = get_proc_task(filp->f_path.dentry->d_inode);
3347 	if (!reaper)
3348 		goto out_no_task;
3349 
3350 	for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3351 		const struct pid_entry *p = &proc_base_stuff[nr];
3352 		if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3353 			goto out;
3354 	}
3355 
3356 	ns = filp->f_dentry->d_sb->s_fs_info;
3357 	iter.task = NULL;
3358 	iter.tgid = filp->f_pos - TGID_OFFSET;
3359 	for (iter = next_tgid(ns, iter);
3360 	     iter.task;
3361 	     iter.tgid += 1, iter = next_tgid(ns, iter)) {
3362 		if (has_pid_permissions(ns, iter.task, 2))
3363 			__filldir = filldir;
3364 		else
3365 			__filldir = fake_filldir;
3366 
3367 		filp->f_pos = iter.tgid + TGID_OFFSET;
3368 		if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
3369 			put_task_struct(iter.task);
3370 			goto out;
3371 		}
3372 	}
3373 	filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3374 out:
3375 	put_task_struct(reaper);
3376 out_no_task:
3377 	return 0;
3378 }
3379 
3380 /*
3381  * Tasks
3382  */
3383 static const struct pid_entry tid_base_stuff[] = {
3384 	DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3385 	DIR("fdinfo",    S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3386 	DIR("ns",	 S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3387 	REG("environ",   S_IRUSR, proc_environ_operations),
3388 	INF("auxv",      S_IRUSR, proc_pid_auxv),
3389 	ONE("status",    S_IRUGO, proc_pid_status),
3390 	ONE("personality", S_IRUGO, proc_pid_personality),
3391 	INF("limits",	 S_IRUGO, proc_pid_limits),
3392 #ifdef CONFIG_SCHED_DEBUG
3393 	REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3394 #endif
3395 	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3396 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3397 	INF("syscall",   S_IRUGO, proc_pid_syscall),
3398 #endif
3399 	INF("cmdline",   S_IRUGO, proc_pid_cmdline),
3400 	ONE("stat",      S_IRUGO, proc_tid_stat),
3401 	ONE("statm",     S_IRUGO, proc_pid_statm),
3402 	REG("maps",      S_IRUGO, proc_tid_maps_operations),
3403 #ifdef CONFIG_CHECKPOINT_RESTORE
3404 	REG("children",  S_IRUGO, proc_tid_children_operations),
3405 #endif
3406 #ifdef CONFIG_NUMA
3407 	REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3408 #endif
3409 	REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
3410 	LNK("cwd",       proc_cwd_link),
3411 	LNK("root",      proc_root_link),
3412 	LNK("exe",       proc_exe_link),
3413 	REG("mounts",    S_IRUGO, proc_mounts_operations),
3414 	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3415 #ifdef CONFIG_PROC_PAGE_MONITOR
3416 	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3417 	REG("smaps",     S_IRUGO, proc_tid_smaps_operations),
3418 	REG("pagemap",    S_IRUGO, proc_pagemap_operations),
3419 #endif
3420 #ifdef CONFIG_SECURITY
3421 	DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3422 #endif
3423 #ifdef CONFIG_KALLSYMS
3424 	INF("wchan",     S_IRUGO, proc_pid_wchan),
3425 #endif
3426 #ifdef CONFIG_STACKTRACE
3427 	ONE("stack",      S_IRUGO, proc_pid_stack),
3428 #endif
3429 #ifdef CONFIG_SCHEDSTATS
3430 	INF("schedstat", S_IRUGO, proc_pid_schedstat),
3431 #endif
3432 #ifdef CONFIG_LATENCYTOP
3433 	REG("latency",  S_IRUGO, proc_lstats_operations),
3434 #endif
3435 #ifdef CONFIG_PROC_PID_CPUSET
3436 	REG("cpuset",    S_IRUGO, proc_cpuset_operations),
3437 #endif
3438 #ifdef CONFIG_CGROUPS
3439 	REG("cgroup",  S_IRUGO, proc_cgroup_operations),
3440 #endif
3441 	INF("oom_score", S_IRUGO, proc_oom_score),
3442 	REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3443 	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3444 #ifdef CONFIG_AUDITSYSCALL
3445 	REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
3446 	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3447 #endif
3448 #ifdef CONFIG_FAULT_INJECTION
3449 	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3450 #endif
3451 #ifdef CONFIG_TASK_IO_ACCOUNTING
3452 	INF("io",	S_IRUSR, proc_tid_io_accounting),
3453 #endif
3454 #ifdef CONFIG_HARDWALL
3455 	INF("hardwall",   S_IRUGO, proc_pid_hardwall),
3456 #endif
3457 #ifdef CONFIG_USER_NS
3458 	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3459 	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3460 #endif
3461 };
3462 
3463 static int proc_tid_base_readdir(struct file * filp,
3464 			     void * dirent, filldir_t filldir)
3465 {
3466 	return proc_pident_readdir(filp,dirent,filldir,
3467 				   tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3468 }
3469 
3470 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3471 	return proc_pident_lookup(dir, dentry,
3472 				  tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3473 }
3474 
3475 static const struct file_operations proc_tid_base_operations = {
3476 	.read		= generic_read_dir,
3477 	.readdir	= proc_tid_base_readdir,
3478 	.llseek		= default_llseek,
3479 };
3480 
3481 static const struct inode_operations proc_tid_base_inode_operations = {
3482 	.lookup		= proc_tid_base_lookup,
3483 	.getattr	= pid_getattr,
3484 	.setattr	= proc_setattr,
3485 };
3486 
3487 static struct dentry *proc_task_instantiate(struct inode *dir,
3488 	struct dentry *dentry, struct task_struct *task, const void *ptr)
3489 {
3490 	struct dentry *error = ERR_PTR(-ENOENT);
3491 	struct inode *inode;
3492 	inode = proc_pid_make_inode(dir->i_sb, task);
3493 
3494 	if (!inode)
3495 		goto out;
3496 	inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3497 	inode->i_op = &proc_tid_base_inode_operations;
3498 	inode->i_fop = &proc_tid_base_operations;
3499 	inode->i_flags|=S_IMMUTABLE;
3500 
3501 	set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3502 						  ARRAY_SIZE(tid_base_stuff)));
3503 
3504 	d_set_d_op(dentry, &pid_dentry_operations);
3505 
3506 	d_add(dentry, inode);
3507 	/* Close the race of the process dying before we return the dentry */
3508 	if (pid_revalidate(dentry, NULL))
3509 		error = NULL;
3510 out:
3511 	return error;
3512 }
3513 
3514 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3515 {
3516 	struct dentry *result = ERR_PTR(-ENOENT);
3517 	struct task_struct *task;
3518 	struct task_struct *leader = get_proc_task(dir);
3519 	unsigned tid;
3520 	struct pid_namespace *ns;
3521 
3522 	if (!leader)
3523 		goto out_no_task;
3524 
3525 	tid = name_to_int(dentry);
3526 	if (tid == ~0U)
3527 		goto out;
3528 
3529 	ns = dentry->d_sb->s_fs_info;
3530 	rcu_read_lock();
3531 	task = find_task_by_pid_ns(tid, ns);
3532 	if (task)
3533 		get_task_struct(task);
3534 	rcu_read_unlock();
3535 	if (!task)
3536 		goto out;
3537 	if (!same_thread_group(leader, task))
3538 		goto out_drop_task;
3539 
3540 	result = proc_task_instantiate(dir, dentry, task, NULL);
3541 out_drop_task:
3542 	put_task_struct(task);
3543 out:
3544 	put_task_struct(leader);
3545 out_no_task:
3546 	return result;
3547 }
3548 
3549 /*
3550  * Find the first tid of a thread group to return to user space.
3551  *
3552  * Usually this is just the thread group leader, but if the users
3553  * buffer was too small or there was a seek into the middle of the
3554  * directory we have more work todo.
3555  *
3556  * In the case of a short read we start with find_task_by_pid.
3557  *
3558  * In the case of a seek we start with the leader and walk nr
3559  * threads past it.
3560  */
3561 static struct task_struct *first_tid(struct task_struct *leader,
3562 		int tid, int nr, struct pid_namespace *ns)
3563 {
3564 	struct task_struct *pos;
3565 
3566 	rcu_read_lock();
3567 	/* Attempt to start with the pid of a thread */
3568 	if (tid && (nr > 0)) {
3569 		pos = find_task_by_pid_ns(tid, ns);
3570 		if (pos && (pos->group_leader == leader))
3571 			goto found;
3572 	}
3573 
3574 	/* If nr exceeds the number of threads there is nothing todo */
3575 	pos = NULL;
3576 	if (nr && nr >= get_nr_threads(leader))
3577 		goto out;
3578 
3579 	/* If we haven't found our starting place yet start
3580 	 * with the leader and walk nr threads forward.
3581 	 */
3582 	for (pos = leader; nr > 0; --nr) {
3583 		pos = next_thread(pos);
3584 		if (pos == leader) {
3585 			pos = NULL;
3586 			goto out;
3587 		}
3588 	}
3589 found:
3590 	get_task_struct(pos);
3591 out:
3592 	rcu_read_unlock();
3593 	return pos;
3594 }
3595 
3596 /*
3597  * Find the next thread in the thread list.
3598  * Return NULL if there is an error or no next thread.
3599  *
3600  * The reference to the input task_struct is released.
3601  */
3602 static struct task_struct *next_tid(struct task_struct *start)
3603 {
3604 	struct task_struct *pos = NULL;
3605 	rcu_read_lock();
3606 	if (pid_alive(start)) {
3607 		pos = next_thread(start);
3608 		if (thread_group_leader(pos))
3609 			pos = NULL;
3610 		else
3611 			get_task_struct(pos);
3612 	}
3613 	rcu_read_unlock();
3614 	put_task_struct(start);
3615 	return pos;
3616 }
3617 
3618 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3619 	struct task_struct *task, int tid)
3620 {
3621 	char name[PROC_NUMBUF];
3622 	int len = snprintf(name, sizeof(name), "%d", tid);
3623 	return proc_fill_cache(filp, dirent, filldir, name, len,
3624 				proc_task_instantiate, task, NULL);
3625 }
3626 
3627 /* for the /proc/TGID/task/ directories */
3628 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3629 {
3630 	struct dentry *dentry = filp->f_path.dentry;
3631 	struct inode *inode = dentry->d_inode;
3632 	struct task_struct *leader = NULL;
3633 	struct task_struct *task;
3634 	int retval = -ENOENT;
3635 	ino_t ino;
3636 	int tid;
3637 	struct pid_namespace *ns;
3638 
3639 	task = get_proc_task(inode);
3640 	if (!task)
3641 		goto out_no_task;
3642 	rcu_read_lock();
3643 	if (pid_alive(task)) {
3644 		leader = task->group_leader;
3645 		get_task_struct(leader);
3646 	}
3647 	rcu_read_unlock();
3648 	put_task_struct(task);
3649 	if (!leader)
3650 		goto out_no_task;
3651 	retval = 0;
3652 
3653 	switch ((unsigned long)filp->f_pos) {
3654 	case 0:
3655 		ino = inode->i_ino;
3656 		if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3657 			goto out;
3658 		filp->f_pos++;
3659 		/* fall through */
3660 	case 1:
3661 		ino = parent_ino(dentry);
3662 		if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3663 			goto out;
3664 		filp->f_pos++;
3665 		/* fall through */
3666 	}
3667 
3668 	/* f_version caches the tgid value that the last readdir call couldn't
3669 	 * return. lseek aka telldir automagically resets f_version to 0.
3670 	 */
3671 	ns = filp->f_dentry->d_sb->s_fs_info;
3672 	tid = (int)filp->f_version;
3673 	filp->f_version = 0;
3674 	for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3675 	     task;
3676 	     task = next_tid(task), filp->f_pos++) {
3677 		tid = task_pid_nr_ns(task, ns);
3678 		if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3679 			/* returning this tgid failed, save it as the first
3680 			 * pid for the next readir call */
3681 			filp->f_version = (u64)tid;
3682 			put_task_struct(task);
3683 			break;
3684 		}
3685 	}
3686 out:
3687 	put_task_struct(leader);
3688 out_no_task:
3689 	return retval;
3690 }
3691 
3692 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3693 {
3694 	struct inode *inode = dentry->d_inode;
3695 	struct task_struct *p = get_proc_task(inode);
3696 	generic_fillattr(inode, stat);
3697 
3698 	if (p) {
3699 		stat->nlink += get_nr_threads(p);
3700 		put_task_struct(p);
3701 	}
3702 
3703 	return 0;
3704 }
3705 
3706 static const struct inode_operations proc_task_inode_operations = {
3707 	.lookup		= proc_task_lookup,
3708 	.getattr	= proc_task_getattr,
3709 	.setattr	= proc_setattr,
3710 	.permission	= proc_pid_permission,
3711 };
3712 
3713 static const struct file_operations proc_task_operations = {
3714 	.read		= generic_read_dir,
3715 	.readdir	= proc_task_readdir,
3716 	.llseek		= default_llseek,
3717 };
3718