xref: /linux/fs/proc/base.c (revision 2b8232ce512105e28453f301d1510de8363bccd1)
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/init.h>
57 #include <linux/capability.h>
58 #include <linux/file.h>
59 #include <linux/string.h>
60 #include <linux/seq_file.h>
61 #include <linux/namei.h>
62 #include <linux/mnt_namespace.h>
63 #include <linux/mm.h>
64 #include <linux/rcupdate.h>
65 #include <linux/kallsyms.h>
66 #include <linux/module.h>
67 #include <linux/mount.h>
68 #include <linux/security.h>
69 #include <linux/ptrace.h>
70 #include <linux/cpuset.h>
71 #include <linux/audit.h>
72 #include <linux/poll.h>
73 #include <linux/nsproxy.h>
74 #include <linux/oom.h>
75 #include <linux/elf.h>
76 #include "internal.h"
77 
78 /* NOTE:
79  *	Implementing inode permission operations in /proc is almost
80  *	certainly an error.  Permission checks need to happen during
81  *	each system call not at open time.  The reason is that most of
82  *	what we wish to check for permissions in /proc varies at runtime.
83  *
84  *	The classic example of a problem is opening file descriptors
85  *	in /proc for a task before it execs a suid executable.
86  */
87 
88 
89 /* Worst case buffer size needed for holding an integer. */
90 #define PROC_NUMBUF 13
91 
92 struct pid_entry {
93 	char *name;
94 	int len;
95 	mode_t mode;
96 	const struct inode_operations *iop;
97 	const struct file_operations *fop;
98 	union proc_op op;
99 };
100 
101 #define NOD(NAME, MODE, IOP, FOP, OP) {			\
102 	.name = (NAME),					\
103 	.len  = sizeof(NAME) - 1,			\
104 	.mode = MODE,					\
105 	.iop  = IOP,					\
106 	.fop  = FOP,					\
107 	.op   = OP,					\
108 }
109 
110 #define DIR(NAME, MODE, OTYPE)							\
111 	NOD(NAME, (S_IFDIR|(MODE)),						\
112 		&proc_##OTYPE##_inode_operations, &proc_##OTYPE##_operations,	\
113 		{} )
114 #define LNK(NAME, OTYPE)					\
115 	NOD(NAME, (S_IFLNK|S_IRWXUGO),				\
116 		&proc_pid_link_inode_operations, NULL,		\
117 		{ .proc_get_link = &proc_##OTYPE##_link } )
118 #define REG(NAME, MODE, OTYPE)				\
119 	NOD(NAME, (S_IFREG|(MODE)), NULL,		\
120 		&proc_##OTYPE##_operations, {})
121 #define INF(NAME, MODE, OTYPE)				\
122 	NOD(NAME, (S_IFREG|(MODE)), 			\
123 		NULL, &proc_info_file_operations,	\
124 		{ .proc_read = &proc_##OTYPE } )
125 
126 int maps_protect;
127 EXPORT_SYMBOL(maps_protect);
128 
129 static struct fs_struct *get_fs_struct(struct task_struct *task)
130 {
131 	struct fs_struct *fs;
132 	task_lock(task);
133 	fs = task->fs;
134 	if(fs)
135 		atomic_inc(&fs->count);
136 	task_unlock(task);
137 	return fs;
138 }
139 
140 static int get_nr_threads(struct task_struct *tsk)
141 {
142 	/* Must be called with the rcu_read_lock held */
143 	unsigned long flags;
144 	int count = 0;
145 
146 	if (lock_task_sighand(tsk, &flags)) {
147 		count = atomic_read(&tsk->signal->count);
148 		unlock_task_sighand(tsk, &flags);
149 	}
150 	return count;
151 }
152 
153 static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
154 {
155 	struct task_struct *task = get_proc_task(inode);
156 	struct fs_struct *fs = NULL;
157 	int result = -ENOENT;
158 
159 	if (task) {
160 		fs = get_fs_struct(task);
161 		put_task_struct(task);
162 	}
163 	if (fs) {
164 		read_lock(&fs->lock);
165 		*mnt = mntget(fs->pwdmnt);
166 		*dentry = dget(fs->pwd);
167 		read_unlock(&fs->lock);
168 		result = 0;
169 		put_fs_struct(fs);
170 	}
171 	return result;
172 }
173 
174 static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
175 {
176 	struct task_struct *task = get_proc_task(inode);
177 	struct fs_struct *fs = NULL;
178 	int result = -ENOENT;
179 
180 	if (task) {
181 		fs = get_fs_struct(task);
182 		put_task_struct(task);
183 	}
184 	if (fs) {
185 		read_lock(&fs->lock);
186 		*mnt = mntget(fs->rootmnt);
187 		*dentry = dget(fs->root);
188 		read_unlock(&fs->lock);
189 		result = 0;
190 		put_fs_struct(fs);
191 	}
192 	return result;
193 }
194 
195 #define MAY_PTRACE(task) \
196 	(task == current || \
197 	(task->parent == current && \
198 	(task->ptrace & PT_PTRACED) && \
199 	 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
200 	 security_ptrace(current,task) == 0))
201 
202 static int proc_pid_environ(struct task_struct *task, char * buffer)
203 {
204 	int res = 0;
205 	struct mm_struct *mm = get_task_mm(task);
206 	if (mm) {
207 		unsigned int len;
208 
209 		res = -ESRCH;
210 		if (!ptrace_may_attach(task))
211 			goto out;
212 
213 		len  = mm->env_end - mm->env_start;
214 		if (len > PAGE_SIZE)
215 			len = PAGE_SIZE;
216 		res = access_process_vm(task, mm->env_start, buffer, len, 0);
217 out:
218 		mmput(mm);
219 	}
220 	return res;
221 }
222 
223 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
224 {
225 	int res = 0;
226 	unsigned int len;
227 	struct mm_struct *mm = get_task_mm(task);
228 	if (!mm)
229 		goto out;
230 	if (!mm->arg_end)
231 		goto out_mm;	/* Shh! No looking before we're done */
232 
233  	len = mm->arg_end - mm->arg_start;
234 
235 	if (len > PAGE_SIZE)
236 		len = PAGE_SIZE;
237 
238 	res = access_process_vm(task, mm->arg_start, buffer, len, 0);
239 
240 	// If the nul at the end of args has been overwritten, then
241 	// assume application is using setproctitle(3).
242 	if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
243 		len = strnlen(buffer, res);
244 		if (len < res) {
245 		    res = len;
246 		} else {
247 			len = mm->env_end - mm->env_start;
248 			if (len > PAGE_SIZE - res)
249 				len = PAGE_SIZE - res;
250 			res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
251 			res = strnlen(buffer, res);
252 		}
253 	}
254 out_mm:
255 	mmput(mm);
256 out:
257 	return res;
258 }
259 
260 static int proc_pid_auxv(struct task_struct *task, char *buffer)
261 {
262 	int res = 0;
263 	struct mm_struct *mm = get_task_mm(task);
264 	if (mm) {
265 		unsigned int nwords = 0;
266 		do
267 			nwords += 2;
268 		while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
269 		res = nwords * sizeof(mm->saved_auxv[0]);
270 		if (res > PAGE_SIZE)
271 			res = PAGE_SIZE;
272 		memcpy(buffer, mm->saved_auxv, res);
273 		mmput(mm);
274 	}
275 	return res;
276 }
277 
278 
279 #ifdef CONFIG_KALLSYMS
280 /*
281  * Provides a wchan file via kallsyms in a proper one-value-per-file format.
282  * Returns the resolved symbol.  If that fails, simply return the address.
283  */
284 static int proc_pid_wchan(struct task_struct *task, char *buffer)
285 {
286 	unsigned long wchan;
287 	char symname[KSYM_NAME_LEN];
288 
289 	wchan = get_wchan(task);
290 
291 	if (lookup_symbol_name(wchan, symname) < 0)
292 		return sprintf(buffer, "%lu", wchan);
293 	else
294 		return sprintf(buffer, "%s", symname);
295 }
296 #endif /* CONFIG_KALLSYMS */
297 
298 #ifdef CONFIG_SCHEDSTATS
299 /*
300  * Provides /proc/PID/schedstat
301  */
302 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
303 {
304 	return sprintf(buffer, "%llu %llu %lu\n",
305 			task->sched_info.cpu_time,
306 			task->sched_info.run_delay,
307 			task->sched_info.pcnt);
308 }
309 #endif
310 
311 /* The badness from the OOM killer */
312 unsigned long badness(struct task_struct *p, unsigned long uptime);
313 static int proc_oom_score(struct task_struct *task, char *buffer)
314 {
315 	unsigned long points;
316 	struct timespec uptime;
317 
318 	do_posix_clock_monotonic_gettime(&uptime);
319 	read_lock(&tasklist_lock);
320 	points = badness(task, uptime.tv_sec);
321 	read_unlock(&tasklist_lock);
322 	return sprintf(buffer, "%lu\n", points);
323 }
324 
325 /************************************************************************/
326 /*                       Here the fs part begins                        */
327 /************************************************************************/
328 
329 /* permission checks */
330 static int proc_fd_access_allowed(struct inode *inode)
331 {
332 	struct task_struct *task;
333 	int allowed = 0;
334 	/* Allow access to a task's file descriptors if it is us or we
335 	 * may use ptrace attach to the process and find out that
336 	 * information.
337 	 */
338 	task = get_proc_task(inode);
339 	if (task) {
340 		allowed = ptrace_may_attach(task);
341 		put_task_struct(task);
342 	}
343 	return allowed;
344 }
345 
346 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
347 {
348 	int error;
349 	struct inode *inode = dentry->d_inode;
350 
351 	if (attr->ia_valid & ATTR_MODE)
352 		return -EPERM;
353 
354 	error = inode_change_ok(inode, attr);
355 	if (!error)
356 		error = inode_setattr(inode, attr);
357 	return error;
358 }
359 
360 static const struct inode_operations proc_def_inode_operations = {
361 	.setattr	= proc_setattr,
362 };
363 
364 extern struct seq_operations mounts_op;
365 struct proc_mounts {
366 	struct seq_file m;
367 	int event;
368 };
369 
370 static int mounts_open(struct inode *inode, struct file *file)
371 {
372 	struct task_struct *task = get_proc_task(inode);
373 	struct mnt_namespace *ns = NULL;
374 	struct proc_mounts *p;
375 	int ret = -EINVAL;
376 
377 	if (task) {
378 		task_lock(task);
379 		if (task->nsproxy) {
380 			ns = task->nsproxy->mnt_ns;
381 			if (ns)
382 				get_mnt_ns(ns);
383 		}
384 		task_unlock(task);
385 		put_task_struct(task);
386 	}
387 
388 	if (ns) {
389 		ret = -ENOMEM;
390 		p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
391 		if (p) {
392 			file->private_data = &p->m;
393 			ret = seq_open(file, &mounts_op);
394 			if (!ret) {
395 				p->m.private = ns;
396 				p->event = ns->event;
397 				return 0;
398 			}
399 			kfree(p);
400 		}
401 		put_mnt_ns(ns);
402 	}
403 	return ret;
404 }
405 
406 static int mounts_release(struct inode *inode, struct file *file)
407 {
408 	struct seq_file *m = file->private_data;
409 	struct mnt_namespace *ns = m->private;
410 	put_mnt_ns(ns);
411 	return seq_release(inode, file);
412 }
413 
414 static unsigned mounts_poll(struct file *file, poll_table *wait)
415 {
416 	struct proc_mounts *p = file->private_data;
417 	struct mnt_namespace *ns = p->m.private;
418 	unsigned res = 0;
419 
420 	poll_wait(file, &ns->poll, wait);
421 
422 	spin_lock(&vfsmount_lock);
423 	if (p->event != ns->event) {
424 		p->event = ns->event;
425 		res = POLLERR;
426 	}
427 	spin_unlock(&vfsmount_lock);
428 
429 	return res;
430 }
431 
432 static const struct file_operations proc_mounts_operations = {
433 	.open		= mounts_open,
434 	.read		= seq_read,
435 	.llseek		= seq_lseek,
436 	.release	= mounts_release,
437 	.poll		= mounts_poll,
438 };
439 
440 extern struct seq_operations mountstats_op;
441 static int mountstats_open(struct inode *inode, struct file *file)
442 {
443 	int ret = seq_open(file, &mountstats_op);
444 
445 	if (!ret) {
446 		struct seq_file *m = file->private_data;
447 		struct mnt_namespace *mnt_ns = NULL;
448 		struct task_struct *task = get_proc_task(inode);
449 
450 		if (task) {
451 			task_lock(task);
452 			if (task->nsproxy)
453 				mnt_ns = task->nsproxy->mnt_ns;
454 			if (mnt_ns)
455 				get_mnt_ns(mnt_ns);
456 			task_unlock(task);
457 			put_task_struct(task);
458 		}
459 
460 		if (mnt_ns)
461 			m->private = mnt_ns;
462 		else {
463 			seq_release(inode, file);
464 			ret = -EINVAL;
465 		}
466 	}
467 	return ret;
468 }
469 
470 static const struct file_operations proc_mountstats_operations = {
471 	.open		= mountstats_open,
472 	.read		= seq_read,
473 	.llseek		= seq_lseek,
474 	.release	= mounts_release,
475 };
476 
477 #define PROC_BLOCK_SIZE	(3*1024)		/* 4K page size but our output routines use some slack for overruns */
478 
479 static ssize_t proc_info_read(struct file * file, char __user * buf,
480 			  size_t count, loff_t *ppos)
481 {
482 	struct inode * inode = file->f_path.dentry->d_inode;
483 	unsigned long page;
484 	ssize_t length;
485 	struct task_struct *task = get_proc_task(inode);
486 
487 	length = -ESRCH;
488 	if (!task)
489 		goto out_no_task;
490 
491 	if (count > PROC_BLOCK_SIZE)
492 		count = PROC_BLOCK_SIZE;
493 
494 	length = -ENOMEM;
495 	if (!(page = __get_free_page(GFP_KERNEL)))
496 		goto out;
497 
498 	length = PROC_I(inode)->op.proc_read(task, (char*)page);
499 
500 	if (length >= 0)
501 		length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
502 	free_page(page);
503 out:
504 	put_task_struct(task);
505 out_no_task:
506 	return length;
507 }
508 
509 static const struct file_operations proc_info_file_operations = {
510 	.read		= proc_info_read,
511 };
512 
513 static int mem_open(struct inode* inode, struct file* file)
514 {
515 	file->private_data = (void*)((long)current->self_exec_id);
516 	return 0;
517 }
518 
519 static ssize_t mem_read(struct file * file, char __user * buf,
520 			size_t count, loff_t *ppos)
521 {
522 	struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
523 	char *page;
524 	unsigned long src = *ppos;
525 	int ret = -ESRCH;
526 	struct mm_struct *mm;
527 
528 	if (!task)
529 		goto out_no_task;
530 
531 	if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
532 		goto out;
533 
534 	ret = -ENOMEM;
535 	page = (char *)__get_free_page(GFP_USER);
536 	if (!page)
537 		goto out;
538 
539 	ret = 0;
540 
541 	mm = get_task_mm(task);
542 	if (!mm)
543 		goto out_free;
544 
545 	ret = -EIO;
546 
547 	if (file->private_data != (void*)((long)current->self_exec_id))
548 		goto out_put;
549 
550 	ret = 0;
551 
552 	while (count > 0) {
553 		int this_len, retval;
554 
555 		this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
556 		retval = access_process_vm(task, src, page, this_len, 0);
557 		if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
558 			if (!ret)
559 				ret = -EIO;
560 			break;
561 		}
562 
563 		if (copy_to_user(buf, page, retval)) {
564 			ret = -EFAULT;
565 			break;
566 		}
567 
568 		ret += retval;
569 		src += retval;
570 		buf += retval;
571 		count -= retval;
572 	}
573 	*ppos = src;
574 
575 out_put:
576 	mmput(mm);
577 out_free:
578 	free_page((unsigned long) page);
579 out:
580 	put_task_struct(task);
581 out_no_task:
582 	return ret;
583 }
584 
585 #define mem_write NULL
586 
587 #ifndef mem_write
588 /* This is a security hazard */
589 static ssize_t mem_write(struct file * file, const char __user *buf,
590 			 size_t count, loff_t *ppos)
591 {
592 	int copied;
593 	char *page;
594 	struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
595 	unsigned long dst = *ppos;
596 
597 	copied = -ESRCH;
598 	if (!task)
599 		goto out_no_task;
600 
601 	if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
602 		goto out;
603 
604 	copied = -ENOMEM;
605 	page = (char *)__get_free_page(GFP_USER);
606 	if (!page)
607 		goto out;
608 
609 	copied = 0;
610 	while (count > 0) {
611 		int this_len, retval;
612 
613 		this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
614 		if (copy_from_user(page, buf, this_len)) {
615 			copied = -EFAULT;
616 			break;
617 		}
618 		retval = access_process_vm(task, dst, page, this_len, 1);
619 		if (!retval) {
620 			if (!copied)
621 				copied = -EIO;
622 			break;
623 		}
624 		copied += retval;
625 		buf += retval;
626 		dst += retval;
627 		count -= retval;
628 	}
629 	*ppos = dst;
630 	free_page((unsigned long) page);
631 out:
632 	put_task_struct(task);
633 out_no_task:
634 	return copied;
635 }
636 #endif
637 
638 static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
639 {
640 	switch (orig) {
641 	case 0:
642 		file->f_pos = offset;
643 		break;
644 	case 1:
645 		file->f_pos += offset;
646 		break;
647 	default:
648 		return -EINVAL;
649 	}
650 	force_successful_syscall_return();
651 	return file->f_pos;
652 }
653 
654 static const struct file_operations proc_mem_operations = {
655 	.llseek		= mem_lseek,
656 	.read		= mem_read,
657 	.write		= mem_write,
658 	.open		= mem_open,
659 };
660 
661 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
662 				size_t count, loff_t *ppos)
663 {
664 	struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
665 	char buffer[PROC_NUMBUF];
666 	size_t len;
667 	int oom_adjust;
668 
669 	if (!task)
670 		return -ESRCH;
671 	oom_adjust = task->oomkilladj;
672 	put_task_struct(task);
673 
674 	len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
675 
676 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
677 }
678 
679 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
680 				size_t count, loff_t *ppos)
681 {
682 	struct task_struct *task;
683 	char buffer[PROC_NUMBUF], *end;
684 	int oom_adjust;
685 
686 	memset(buffer, 0, sizeof(buffer));
687 	if (count > sizeof(buffer) - 1)
688 		count = sizeof(buffer) - 1;
689 	if (copy_from_user(buffer, buf, count))
690 		return -EFAULT;
691 	oom_adjust = simple_strtol(buffer, &end, 0);
692 	if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
693 	     oom_adjust != OOM_DISABLE)
694 		return -EINVAL;
695 	if (*end == '\n')
696 		end++;
697 	task = get_proc_task(file->f_path.dentry->d_inode);
698 	if (!task)
699 		return -ESRCH;
700 	if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) {
701 		put_task_struct(task);
702 		return -EACCES;
703 	}
704 	task->oomkilladj = oom_adjust;
705 	put_task_struct(task);
706 	if (end - buffer == 0)
707 		return -EIO;
708 	return end - buffer;
709 }
710 
711 static const struct file_operations proc_oom_adjust_operations = {
712 	.read		= oom_adjust_read,
713 	.write		= oom_adjust_write,
714 };
715 
716 #ifdef CONFIG_MMU
717 static ssize_t clear_refs_write(struct file *file, const char __user *buf,
718 				size_t count, loff_t *ppos)
719 {
720 	struct task_struct *task;
721 	char buffer[PROC_NUMBUF], *end;
722 	struct mm_struct *mm;
723 
724 	memset(buffer, 0, sizeof(buffer));
725 	if (count > sizeof(buffer) - 1)
726 		count = sizeof(buffer) - 1;
727 	if (copy_from_user(buffer, buf, count))
728 		return -EFAULT;
729 	if (!simple_strtol(buffer, &end, 0))
730 		return -EINVAL;
731 	if (*end == '\n')
732 		end++;
733 	task = get_proc_task(file->f_path.dentry->d_inode);
734 	if (!task)
735 		return -ESRCH;
736 	mm = get_task_mm(task);
737 	if (mm) {
738 		clear_refs_smap(mm);
739 		mmput(mm);
740 	}
741 	put_task_struct(task);
742 	if (end - buffer == 0)
743 		return -EIO;
744 	return end - buffer;
745 }
746 
747 static struct file_operations proc_clear_refs_operations = {
748 	.write		= clear_refs_write,
749 };
750 #endif
751 
752 #ifdef CONFIG_AUDITSYSCALL
753 #define TMPBUFLEN 21
754 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
755 				  size_t count, loff_t *ppos)
756 {
757 	struct inode * inode = file->f_path.dentry->d_inode;
758 	struct task_struct *task = get_proc_task(inode);
759 	ssize_t length;
760 	char tmpbuf[TMPBUFLEN];
761 
762 	if (!task)
763 		return -ESRCH;
764 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
765 				audit_get_loginuid(task->audit_context));
766 	put_task_struct(task);
767 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
768 }
769 
770 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
771 				   size_t count, loff_t *ppos)
772 {
773 	struct inode * inode = file->f_path.dentry->d_inode;
774 	char *page, *tmp;
775 	ssize_t length;
776 	uid_t loginuid;
777 
778 	if (!capable(CAP_AUDIT_CONTROL))
779 		return -EPERM;
780 
781 	if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
782 		return -EPERM;
783 
784 	if (count >= PAGE_SIZE)
785 		count = PAGE_SIZE - 1;
786 
787 	if (*ppos != 0) {
788 		/* No partial writes. */
789 		return -EINVAL;
790 	}
791 	page = (char*)__get_free_page(GFP_USER);
792 	if (!page)
793 		return -ENOMEM;
794 	length = -EFAULT;
795 	if (copy_from_user(page, buf, count))
796 		goto out_free_page;
797 
798 	page[count] = '\0';
799 	loginuid = simple_strtoul(page, &tmp, 10);
800 	if (tmp == page) {
801 		length = -EINVAL;
802 		goto out_free_page;
803 
804 	}
805 	length = audit_set_loginuid(current, loginuid);
806 	if (likely(length == 0))
807 		length = count;
808 
809 out_free_page:
810 	free_page((unsigned long) page);
811 	return length;
812 }
813 
814 static const struct file_operations proc_loginuid_operations = {
815 	.read		= proc_loginuid_read,
816 	.write		= proc_loginuid_write,
817 };
818 #endif
819 
820 #ifdef CONFIG_FAULT_INJECTION
821 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
822 				      size_t count, loff_t *ppos)
823 {
824 	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
825 	char buffer[PROC_NUMBUF];
826 	size_t len;
827 	int make_it_fail;
828 
829 	if (!task)
830 		return -ESRCH;
831 	make_it_fail = task->make_it_fail;
832 	put_task_struct(task);
833 
834 	len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
835 
836 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
837 }
838 
839 static ssize_t proc_fault_inject_write(struct file * file,
840 			const char __user * buf, size_t count, loff_t *ppos)
841 {
842 	struct task_struct *task;
843 	char buffer[PROC_NUMBUF], *end;
844 	int make_it_fail;
845 
846 	if (!capable(CAP_SYS_RESOURCE))
847 		return -EPERM;
848 	memset(buffer, 0, sizeof(buffer));
849 	if (count > sizeof(buffer) - 1)
850 		count = sizeof(buffer) - 1;
851 	if (copy_from_user(buffer, buf, count))
852 		return -EFAULT;
853 	make_it_fail = simple_strtol(buffer, &end, 0);
854 	if (*end == '\n')
855 		end++;
856 	task = get_proc_task(file->f_dentry->d_inode);
857 	if (!task)
858 		return -ESRCH;
859 	task->make_it_fail = make_it_fail;
860 	put_task_struct(task);
861 	if (end - buffer == 0)
862 		return -EIO;
863 	return end - buffer;
864 }
865 
866 static const struct file_operations proc_fault_inject_operations = {
867 	.read		= proc_fault_inject_read,
868 	.write		= proc_fault_inject_write,
869 };
870 #endif
871 
872 #ifdef CONFIG_SCHED_DEBUG
873 /*
874  * Print out various scheduling related per-task fields:
875  */
876 static int sched_show(struct seq_file *m, void *v)
877 {
878 	struct inode *inode = m->private;
879 	struct task_struct *p;
880 
881 	WARN_ON(!inode);
882 
883 	p = get_proc_task(inode);
884 	if (!p)
885 		return -ESRCH;
886 	proc_sched_show_task(p, m);
887 
888 	put_task_struct(p);
889 
890 	return 0;
891 }
892 
893 static ssize_t
894 sched_write(struct file *file, const char __user *buf,
895 	    size_t count, loff_t *offset)
896 {
897 	struct inode *inode = file->f_path.dentry->d_inode;
898 	struct task_struct *p;
899 
900 	WARN_ON(!inode);
901 
902 	p = get_proc_task(inode);
903 	if (!p)
904 		return -ESRCH;
905 	proc_sched_set_task(p);
906 
907 	put_task_struct(p);
908 
909 	return count;
910 }
911 
912 static int sched_open(struct inode *inode, struct file *filp)
913 {
914 	int ret;
915 
916 	ret = single_open(filp, sched_show, NULL);
917 	if (!ret) {
918 		struct seq_file *m = filp->private_data;
919 
920 		m->private = inode;
921 	}
922 	return ret;
923 }
924 
925 static const struct file_operations proc_pid_sched_operations = {
926 	.open		= sched_open,
927 	.read		= seq_read,
928 	.write		= sched_write,
929 	.llseek		= seq_lseek,
930 	.release	= single_release,
931 };
932 
933 #endif
934 
935 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
936 {
937 	struct inode *inode = dentry->d_inode;
938 	int error = -EACCES;
939 
940 	/* We don't need a base pointer in the /proc filesystem */
941 	path_release(nd);
942 
943 	/* Are we allowed to snoop on the tasks file descriptors? */
944 	if (!proc_fd_access_allowed(inode))
945 		goto out;
946 
947 	error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
948 	nd->last_type = LAST_BIND;
949 out:
950 	return ERR_PTR(error);
951 }
952 
953 static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
954 			    char __user *buffer, int buflen)
955 {
956 	struct inode * inode;
957 	char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
958 	int len;
959 
960 	if (!tmp)
961 		return -ENOMEM;
962 
963 	inode = dentry->d_inode;
964 	path = d_path(dentry, mnt, tmp, PAGE_SIZE);
965 	len = PTR_ERR(path);
966 	if (IS_ERR(path))
967 		goto out;
968 	len = tmp + PAGE_SIZE - 1 - path;
969 
970 	if (len > buflen)
971 		len = buflen;
972 	if (copy_to_user(buffer, path, len))
973 		len = -EFAULT;
974  out:
975 	free_page((unsigned long)tmp);
976 	return len;
977 }
978 
979 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
980 {
981 	int error = -EACCES;
982 	struct inode *inode = dentry->d_inode;
983 	struct dentry *de;
984 	struct vfsmount *mnt = NULL;
985 
986 	/* Are we allowed to snoop on the tasks file descriptors? */
987 	if (!proc_fd_access_allowed(inode))
988 		goto out;
989 
990 	error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
991 	if (error)
992 		goto out;
993 
994 	error = do_proc_readlink(de, mnt, buffer, buflen);
995 	dput(de);
996 	mntput(mnt);
997 out:
998 	return error;
999 }
1000 
1001 static const struct inode_operations proc_pid_link_inode_operations = {
1002 	.readlink	= proc_pid_readlink,
1003 	.follow_link	= proc_pid_follow_link,
1004 	.setattr	= proc_setattr,
1005 };
1006 
1007 
1008 /* building an inode */
1009 
1010 static int task_dumpable(struct task_struct *task)
1011 {
1012 	int dumpable = 0;
1013 	struct mm_struct *mm;
1014 
1015 	task_lock(task);
1016 	mm = task->mm;
1017 	if (mm)
1018 		dumpable = get_dumpable(mm);
1019 	task_unlock(task);
1020 	if(dumpable == 1)
1021 		return 1;
1022 	return 0;
1023 }
1024 
1025 
1026 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1027 {
1028 	struct inode * inode;
1029 	struct proc_inode *ei;
1030 
1031 	/* We need a new inode */
1032 
1033 	inode = new_inode(sb);
1034 	if (!inode)
1035 		goto out;
1036 
1037 	/* Common stuff */
1038 	ei = PROC_I(inode);
1039 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1040 	inode->i_op = &proc_def_inode_operations;
1041 
1042 	/*
1043 	 * grab the reference to task.
1044 	 */
1045 	ei->pid = get_task_pid(task, PIDTYPE_PID);
1046 	if (!ei->pid)
1047 		goto out_unlock;
1048 
1049 	inode->i_uid = 0;
1050 	inode->i_gid = 0;
1051 	if (task_dumpable(task)) {
1052 		inode->i_uid = task->euid;
1053 		inode->i_gid = task->egid;
1054 	}
1055 	security_task_to_inode(task, inode);
1056 
1057 out:
1058 	return inode;
1059 
1060 out_unlock:
1061 	iput(inode);
1062 	return NULL;
1063 }
1064 
1065 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1066 {
1067 	struct inode *inode = dentry->d_inode;
1068 	struct task_struct *task;
1069 	generic_fillattr(inode, stat);
1070 
1071 	rcu_read_lock();
1072 	stat->uid = 0;
1073 	stat->gid = 0;
1074 	task = pid_task(proc_pid(inode), PIDTYPE_PID);
1075 	if (task) {
1076 		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1077 		    task_dumpable(task)) {
1078 			stat->uid = task->euid;
1079 			stat->gid = task->egid;
1080 		}
1081 	}
1082 	rcu_read_unlock();
1083 	return 0;
1084 }
1085 
1086 /* dentry stuff */
1087 
1088 /*
1089  *	Exceptional case: normally we are not allowed to unhash a busy
1090  * directory. In this case, however, we can do it - no aliasing problems
1091  * due to the way we treat inodes.
1092  *
1093  * Rewrite the inode's ownerships here because the owning task may have
1094  * performed a setuid(), etc.
1095  *
1096  * Before the /proc/pid/status file was created the only way to read
1097  * the effective uid of a /process was to stat /proc/pid.  Reading
1098  * /proc/pid/status is slow enough that procps and other packages
1099  * kept stating /proc/pid.  To keep the rules in /proc simple I have
1100  * made this apply to all per process world readable and executable
1101  * directories.
1102  */
1103 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1104 {
1105 	struct inode *inode = dentry->d_inode;
1106 	struct task_struct *task = get_proc_task(inode);
1107 	if (task) {
1108 		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1109 		    task_dumpable(task)) {
1110 			inode->i_uid = task->euid;
1111 			inode->i_gid = task->egid;
1112 		} else {
1113 			inode->i_uid = 0;
1114 			inode->i_gid = 0;
1115 		}
1116 		inode->i_mode &= ~(S_ISUID | S_ISGID);
1117 		security_task_to_inode(task, inode);
1118 		put_task_struct(task);
1119 		return 1;
1120 	}
1121 	d_drop(dentry);
1122 	return 0;
1123 }
1124 
1125 static int pid_delete_dentry(struct dentry * dentry)
1126 {
1127 	/* Is the task we represent dead?
1128 	 * If so, then don't put the dentry on the lru list,
1129 	 * kill it immediately.
1130 	 */
1131 	return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1132 }
1133 
1134 static struct dentry_operations pid_dentry_operations =
1135 {
1136 	.d_revalidate	= pid_revalidate,
1137 	.d_delete	= pid_delete_dentry,
1138 };
1139 
1140 /* Lookups */
1141 
1142 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1143 				struct task_struct *, const void *);
1144 
1145 /*
1146  * Fill a directory entry.
1147  *
1148  * If possible create the dcache entry and derive our inode number and
1149  * file type from dcache entry.
1150  *
1151  * Since all of the proc inode numbers are dynamically generated, the inode
1152  * numbers do not exist until the inode is cache.  This means creating the
1153  * the dcache entry in readdir is necessary to keep the inode numbers
1154  * reported by readdir in sync with the inode numbers reported
1155  * by stat.
1156  */
1157 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1158 	char *name, int len,
1159 	instantiate_t instantiate, struct task_struct *task, const void *ptr)
1160 {
1161 	struct dentry *child, *dir = filp->f_path.dentry;
1162 	struct inode *inode;
1163 	struct qstr qname;
1164 	ino_t ino = 0;
1165 	unsigned type = DT_UNKNOWN;
1166 
1167 	qname.name = name;
1168 	qname.len  = len;
1169 	qname.hash = full_name_hash(name, len);
1170 
1171 	child = d_lookup(dir, &qname);
1172 	if (!child) {
1173 		struct dentry *new;
1174 		new = d_alloc(dir, &qname);
1175 		if (new) {
1176 			child = instantiate(dir->d_inode, new, task, ptr);
1177 			if (child)
1178 				dput(new);
1179 			else
1180 				child = new;
1181 		}
1182 	}
1183 	if (!child || IS_ERR(child) || !child->d_inode)
1184 		goto end_instantiate;
1185 	inode = child->d_inode;
1186 	if (inode) {
1187 		ino = inode->i_ino;
1188 		type = inode->i_mode >> 12;
1189 	}
1190 	dput(child);
1191 end_instantiate:
1192 	if (!ino)
1193 		ino = find_inode_number(dir, &qname);
1194 	if (!ino)
1195 		ino = 1;
1196 	return filldir(dirent, name, len, filp->f_pos, ino, type);
1197 }
1198 
1199 static unsigned name_to_int(struct dentry *dentry)
1200 {
1201 	const char *name = dentry->d_name.name;
1202 	int len = dentry->d_name.len;
1203 	unsigned n = 0;
1204 
1205 	if (len > 1 && *name == '0')
1206 		goto out;
1207 	while (len-- > 0) {
1208 		unsigned c = *name++ - '0';
1209 		if (c > 9)
1210 			goto out;
1211 		if (n >= (~0U-9)/10)
1212 			goto out;
1213 		n *= 10;
1214 		n += c;
1215 	}
1216 	return n;
1217 out:
1218 	return ~0U;
1219 }
1220 
1221 #define PROC_FDINFO_MAX 64
1222 
1223 static int proc_fd_info(struct inode *inode, struct dentry **dentry,
1224 			struct vfsmount **mnt, char *info)
1225 {
1226 	struct task_struct *task = get_proc_task(inode);
1227 	struct files_struct *files = NULL;
1228 	struct file *file;
1229 	int fd = proc_fd(inode);
1230 
1231 	if (task) {
1232 		files = get_files_struct(task);
1233 		put_task_struct(task);
1234 	}
1235 	if (files) {
1236 		/*
1237 		 * We are not taking a ref to the file structure, so we must
1238 		 * hold ->file_lock.
1239 		 */
1240 		spin_lock(&files->file_lock);
1241 		file = fcheck_files(files, fd);
1242 		if (file) {
1243 			if (mnt)
1244 				*mnt = mntget(file->f_path.mnt);
1245 			if (dentry)
1246 				*dentry = dget(file->f_path.dentry);
1247 			if (info)
1248 				snprintf(info, PROC_FDINFO_MAX,
1249 					 "pos:\t%lli\n"
1250 					 "flags:\t0%o\n",
1251 					 (long long) file->f_pos,
1252 					 file->f_flags);
1253 			spin_unlock(&files->file_lock);
1254 			put_files_struct(files);
1255 			return 0;
1256 		}
1257 		spin_unlock(&files->file_lock);
1258 		put_files_struct(files);
1259 	}
1260 	return -ENOENT;
1261 }
1262 
1263 static int proc_fd_link(struct inode *inode, struct dentry **dentry,
1264 			struct vfsmount **mnt)
1265 {
1266 	return proc_fd_info(inode, dentry, mnt, NULL);
1267 }
1268 
1269 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1270 {
1271 	struct inode *inode = dentry->d_inode;
1272 	struct task_struct *task = get_proc_task(inode);
1273 	int fd = proc_fd(inode);
1274 	struct files_struct *files;
1275 
1276 	if (task) {
1277 		files = get_files_struct(task);
1278 		if (files) {
1279 			rcu_read_lock();
1280 			if (fcheck_files(files, fd)) {
1281 				rcu_read_unlock();
1282 				put_files_struct(files);
1283 				if (task_dumpable(task)) {
1284 					inode->i_uid = task->euid;
1285 					inode->i_gid = task->egid;
1286 				} else {
1287 					inode->i_uid = 0;
1288 					inode->i_gid = 0;
1289 				}
1290 				inode->i_mode &= ~(S_ISUID | S_ISGID);
1291 				security_task_to_inode(task, inode);
1292 				put_task_struct(task);
1293 				return 1;
1294 			}
1295 			rcu_read_unlock();
1296 			put_files_struct(files);
1297 		}
1298 		put_task_struct(task);
1299 	}
1300 	d_drop(dentry);
1301 	return 0;
1302 }
1303 
1304 static struct dentry_operations tid_fd_dentry_operations =
1305 {
1306 	.d_revalidate	= tid_fd_revalidate,
1307 	.d_delete	= pid_delete_dentry,
1308 };
1309 
1310 static struct dentry *proc_fd_instantiate(struct inode *dir,
1311 	struct dentry *dentry, struct task_struct *task, const void *ptr)
1312 {
1313 	unsigned fd = *(const unsigned *)ptr;
1314 	struct file *file;
1315 	struct files_struct *files;
1316  	struct inode *inode;
1317  	struct proc_inode *ei;
1318 	struct dentry *error = ERR_PTR(-ENOENT);
1319 
1320 	inode = proc_pid_make_inode(dir->i_sb, task);
1321 	if (!inode)
1322 		goto out;
1323 	ei = PROC_I(inode);
1324 	ei->fd = fd;
1325 	files = get_files_struct(task);
1326 	if (!files)
1327 		goto out_iput;
1328 	inode->i_mode = S_IFLNK;
1329 
1330 	/*
1331 	 * We are not taking a ref to the file structure, so we must
1332 	 * hold ->file_lock.
1333 	 */
1334 	spin_lock(&files->file_lock);
1335 	file = fcheck_files(files, fd);
1336 	if (!file)
1337 		goto out_unlock;
1338 	if (file->f_mode & 1)
1339 		inode->i_mode |= S_IRUSR | S_IXUSR;
1340 	if (file->f_mode & 2)
1341 		inode->i_mode |= S_IWUSR | S_IXUSR;
1342 	spin_unlock(&files->file_lock);
1343 	put_files_struct(files);
1344 
1345 	inode->i_op = &proc_pid_link_inode_operations;
1346 	inode->i_size = 64;
1347 	ei->op.proc_get_link = proc_fd_link;
1348 	dentry->d_op = &tid_fd_dentry_operations;
1349 	d_add(dentry, inode);
1350 	/* Close the race of the process dying before we return the dentry */
1351 	if (tid_fd_revalidate(dentry, NULL))
1352 		error = NULL;
1353 
1354  out:
1355 	return error;
1356 out_unlock:
1357 	spin_unlock(&files->file_lock);
1358 	put_files_struct(files);
1359 out_iput:
1360 	iput(inode);
1361 	goto out;
1362 }
1363 
1364 static struct dentry *proc_lookupfd_common(struct inode *dir,
1365 					   struct dentry *dentry,
1366 					   instantiate_t instantiate)
1367 {
1368 	struct task_struct *task = get_proc_task(dir);
1369 	unsigned fd = name_to_int(dentry);
1370 	struct dentry *result = ERR_PTR(-ENOENT);
1371 
1372 	if (!task)
1373 		goto out_no_task;
1374 	if (fd == ~0U)
1375 		goto out;
1376 
1377 	result = instantiate(dir, dentry, task, &fd);
1378 out:
1379 	put_task_struct(task);
1380 out_no_task:
1381 	return result;
1382 }
1383 
1384 static int proc_readfd_common(struct file * filp, void * dirent,
1385 			      filldir_t filldir, instantiate_t instantiate)
1386 {
1387 	struct dentry *dentry = filp->f_path.dentry;
1388 	struct inode *inode = dentry->d_inode;
1389 	struct task_struct *p = get_proc_task(inode);
1390 	unsigned int fd, tid, ino;
1391 	int retval;
1392 	struct files_struct * files;
1393 	struct fdtable *fdt;
1394 
1395 	retval = -ENOENT;
1396 	if (!p)
1397 		goto out_no_task;
1398 	retval = 0;
1399 	tid = p->pid;
1400 
1401 	fd = filp->f_pos;
1402 	switch (fd) {
1403 		case 0:
1404 			if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1405 				goto out;
1406 			filp->f_pos++;
1407 		case 1:
1408 			ino = parent_ino(dentry);
1409 			if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1410 				goto out;
1411 			filp->f_pos++;
1412 		default:
1413 			files = get_files_struct(p);
1414 			if (!files)
1415 				goto out;
1416 			rcu_read_lock();
1417 			fdt = files_fdtable(files);
1418 			for (fd = filp->f_pos-2;
1419 			     fd < fdt->max_fds;
1420 			     fd++, filp->f_pos++) {
1421 				char name[PROC_NUMBUF];
1422 				int len;
1423 
1424 				if (!fcheck_files(files, fd))
1425 					continue;
1426 				rcu_read_unlock();
1427 
1428 				len = snprintf(name, sizeof(name), "%d", fd);
1429 				if (proc_fill_cache(filp, dirent, filldir,
1430 						    name, len, instantiate,
1431 						    p, &fd) < 0) {
1432 					rcu_read_lock();
1433 					break;
1434 				}
1435 				rcu_read_lock();
1436 			}
1437 			rcu_read_unlock();
1438 			put_files_struct(files);
1439 	}
1440 out:
1441 	put_task_struct(p);
1442 out_no_task:
1443 	return retval;
1444 }
1445 
1446 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1447 				    struct nameidata *nd)
1448 {
1449 	return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1450 }
1451 
1452 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1453 {
1454 	return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1455 }
1456 
1457 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1458 				      size_t len, loff_t *ppos)
1459 {
1460 	char tmp[PROC_FDINFO_MAX];
1461 	int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, NULL, tmp);
1462 	if (!err)
1463 		err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1464 	return err;
1465 }
1466 
1467 static const struct file_operations proc_fdinfo_file_operations = {
1468 	.open		= nonseekable_open,
1469 	.read		= proc_fdinfo_read,
1470 };
1471 
1472 static const struct file_operations proc_fd_operations = {
1473 	.read		= generic_read_dir,
1474 	.readdir	= proc_readfd,
1475 };
1476 
1477 /*
1478  * /proc/pid/fd needs a special permission handler so that a process can still
1479  * access /proc/self/fd after it has executed a setuid().
1480  */
1481 static int proc_fd_permission(struct inode *inode, int mask,
1482 				struct nameidata *nd)
1483 {
1484 	int rv;
1485 
1486 	rv = generic_permission(inode, mask, NULL);
1487 	if (rv == 0)
1488 		return 0;
1489 	if (task_pid(current) == proc_pid(inode))
1490 		rv = 0;
1491 	return rv;
1492 }
1493 
1494 /*
1495  * proc directories can do almost nothing..
1496  */
1497 static const struct inode_operations proc_fd_inode_operations = {
1498 	.lookup		= proc_lookupfd,
1499 	.permission	= proc_fd_permission,
1500 	.setattr	= proc_setattr,
1501 };
1502 
1503 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1504 	struct dentry *dentry, struct task_struct *task, const void *ptr)
1505 {
1506 	unsigned fd = *(unsigned *)ptr;
1507  	struct inode *inode;
1508  	struct proc_inode *ei;
1509 	struct dentry *error = ERR_PTR(-ENOENT);
1510 
1511 	inode = proc_pid_make_inode(dir->i_sb, task);
1512 	if (!inode)
1513 		goto out;
1514 	ei = PROC_I(inode);
1515 	ei->fd = fd;
1516 	inode->i_mode = S_IFREG | S_IRUSR;
1517 	inode->i_fop = &proc_fdinfo_file_operations;
1518 	dentry->d_op = &tid_fd_dentry_operations;
1519 	d_add(dentry, inode);
1520 	/* Close the race of the process dying before we return the dentry */
1521 	if (tid_fd_revalidate(dentry, NULL))
1522 		error = NULL;
1523 
1524  out:
1525 	return error;
1526 }
1527 
1528 static struct dentry *proc_lookupfdinfo(struct inode *dir,
1529 					struct dentry *dentry,
1530 					struct nameidata *nd)
1531 {
1532 	return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
1533 }
1534 
1535 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
1536 {
1537 	return proc_readfd_common(filp, dirent, filldir,
1538 				  proc_fdinfo_instantiate);
1539 }
1540 
1541 static const struct file_operations proc_fdinfo_operations = {
1542 	.read		= generic_read_dir,
1543 	.readdir	= proc_readfdinfo,
1544 };
1545 
1546 /*
1547  * proc directories can do almost nothing..
1548  */
1549 static const struct inode_operations proc_fdinfo_inode_operations = {
1550 	.lookup		= proc_lookupfdinfo,
1551 	.setattr	= proc_setattr,
1552 };
1553 
1554 
1555 static struct dentry *proc_pident_instantiate(struct inode *dir,
1556 	struct dentry *dentry, struct task_struct *task, const void *ptr)
1557 {
1558 	const struct pid_entry *p = ptr;
1559 	struct inode *inode;
1560 	struct proc_inode *ei;
1561 	struct dentry *error = ERR_PTR(-EINVAL);
1562 
1563 	inode = proc_pid_make_inode(dir->i_sb, task);
1564 	if (!inode)
1565 		goto out;
1566 
1567 	ei = PROC_I(inode);
1568 	inode->i_mode = p->mode;
1569 	if (S_ISDIR(inode->i_mode))
1570 		inode->i_nlink = 2;	/* Use getattr to fix if necessary */
1571 	if (p->iop)
1572 		inode->i_op = p->iop;
1573 	if (p->fop)
1574 		inode->i_fop = p->fop;
1575 	ei->op = p->op;
1576 	dentry->d_op = &pid_dentry_operations;
1577 	d_add(dentry, inode);
1578 	/* Close the race of the process dying before we return the dentry */
1579 	if (pid_revalidate(dentry, NULL))
1580 		error = NULL;
1581 out:
1582 	return error;
1583 }
1584 
1585 static struct dentry *proc_pident_lookup(struct inode *dir,
1586 					 struct dentry *dentry,
1587 					 const struct pid_entry *ents,
1588 					 unsigned int nents)
1589 {
1590 	struct inode *inode;
1591 	struct dentry *error;
1592 	struct task_struct *task = get_proc_task(dir);
1593 	const struct pid_entry *p, *last;
1594 
1595 	error = ERR_PTR(-ENOENT);
1596 	inode = NULL;
1597 
1598 	if (!task)
1599 		goto out_no_task;
1600 
1601 	/*
1602 	 * Yes, it does not scale. And it should not. Don't add
1603 	 * new entries into /proc/<tgid>/ without very good reasons.
1604 	 */
1605 	last = &ents[nents - 1];
1606 	for (p = ents; p <= last; p++) {
1607 		if (p->len != dentry->d_name.len)
1608 			continue;
1609 		if (!memcmp(dentry->d_name.name, p->name, p->len))
1610 			break;
1611 	}
1612 	if (p > last)
1613 		goto out;
1614 
1615 	error = proc_pident_instantiate(dir, dentry, task, p);
1616 out:
1617 	put_task_struct(task);
1618 out_no_task:
1619 	return error;
1620 }
1621 
1622 static int proc_pident_fill_cache(struct file *filp, void *dirent,
1623 	filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
1624 {
1625 	return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1626 				proc_pident_instantiate, task, p);
1627 }
1628 
1629 static int proc_pident_readdir(struct file *filp,
1630 		void *dirent, filldir_t filldir,
1631 		const struct pid_entry *ents, unsigned int nents)
1632 {
1633 	int i;
1634 	int pid;
1635 	struct dentry *dentry = filp->f_path.dentry;
1636 	struct inode *inode = dentry->d_inode;
1637 	struct task_struct *task = get_proc_task(inode);
1638 	const struct pid_entry *p, *last;
1639 	ino_t ino;
1640 	int ret;
1641 
1642 	ret = -ENOENT;
1643 	if (!task)
1644 		goto out_no_task;
1645 
1646 	ret = 0;
1647 	pid = task->pid;
1648 	i = filp->f_pos;
1649 	switch (i) {
1650 	case 0:
1651 		ino = inode->i_ino;
1652 		if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1653 			goto out;
1654 		i++;
1655 		filp->f_pos++;
1656 		/* fall through */
1657 	case 1:
1658 		ino = parent_ino(dentry);
1659 		if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1660 			goto out;
1661 		i++;
1662 		filp->f_pos++;
1663 		/* fall through */
1664 	default:
1665 		i -= 2;
1666 		if (i >= nents) {
1667 			ret = 1;
1668 			goto out;
1669 		}
1670 		p = ents + i;
1671 		last = &ents[nents - 1];
1672 		while (p <= last) {
1673 			if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
1674 				goto out;
1675 			filp->f_pos++;
1676 			p++;
1677 		}
1678 	}
1679 
1680 	ret = 1;
1681 out:
1682 	put_task_struct(task);
1683 out_no_task:
1684 	return ret;
1685 }
1686 
1687 #ifdef CONFIG_SECURITY
1688 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1689 				  size_t count, loff_t *ppos)
1690 {
1691 	struct inode * inode = file->f_path.dentry->d_inode;
1692 	char *p = NULL;
1693 	ssize_t length;
1694 	struct task_struct *task = get_proc_task(inode);
1695 
1696 	if (!task)
1697 		return -ESRCH;
1698 
1699 	length = security_getprocattr(task,
1700 				      (char*)file->f_path.dentry->d_name.name,
1701 				      &p);
1702 	put_task_struct(task);
1703 	if (length > 0)
1704 		length = simple_read_from_buffer(buf, count, ppos, p, length);
1705 	kfree(p);
1706 	return length;
1707 }
1708 
1709 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1710 				   size_t count, loff_t *ppos)
1711 {
1712 	struct inode * inode = file->f_path.dentry->d_inode;
1713 	char *page;
1714 	ssize_t length;
1715 	struct task_struct *task = get_proc_task(inode);
1716 
1717 	length = -ESRCH;
1718 	if (!task)
1719 		goto out_no_task;
1720 	if (count > PAGE_SIZE)
1721 		count = PAGE_SIZE;
1722 
1723 	/* No partial writes. */
1724 	length = -EINVAL;
1725 	if (*ppos != 0)
1726 		goto out;
1727 
1728 	length = -ENOMEM;
1729 	page = (char*)__get_free_page(GFP_USER);
1730 	if (!page)
1731 		goto out;
1732 
1733 	length = -EFAULT;
1734 	if (copy_from_user(page, buf, count))
1735 		goto out_free;
1736 
1737 	length = security_setprocattr(task,
1738 				      (char*)file->f_path.dentry->d_name.name,
1739 				      (void*)page, count);
1740 out_free:
1741 	free_page((unsigned long) page);
1742 out:
1743 	put_task_struct(task);
1744 out_no_task:
1745 	return length;
1746 }
1747 
1748 static const struct file_operations proc_pid_attr_operations = {
1749 	.read		= proc_pid_attr_read,
1750 	.write		= proc_pid_attr_write,
1751 };
1752 
1753 static const struct pid_entry attr_dir_stuff[] = {
1754 	REG("current",    S_IRUGO|S_IWUGO, pid_attr),
1755 	REG("prev",       S_IRUGO,	   pid_attr),
1756 	REG("exec",       S_IRUGO|S_IWUGO, pid_attr),
1757 	REG("fscreate",   S_IRUGO|S_IWUGO, pid_attr),
1758 	REG("keycreate",  S_IRUGO|S_IWUGO, pid_attr),
1759 	REG("sockcreate", S_IRUGO|S_IWUGO, pid_attr),
1760 };
1761 
1762 static int proc_attr_dir_readdir(struct file * filp,
1763 			     void * dirent, filldir_t filldir)
1764 {
1765 	return proc_pident_readdir(filp,dirent,filldir,
1766 				   attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
1767 }
1768 
1769 static const struct file_operations proc_attr_dir_operations = {
1770 	.read		= generic_read_dir,
1771 	.readdir	= proc_attr_dir_readdir,
1772 };
1773 
1774 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
1775 				struct dentry *dentry, struct nameidata *nd)
1776 {
1777 	return proc_pident_lookup(dir, dentry,
1778 				  attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
1779 }
1780 
1781 static const struct inode_operations proc_attr_dir_inode_operations = {
1782 	.lookup		= proc_attr_dir_lookup,
1783 	.getattr	= pid_getattr,
1784 	.setattr	= proc_setattr,
1785 };
1786 
1787 #endif
1788 
1789 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
1790 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
1791 					 size_t count, loff_t *ppos)
1792 {
1793 	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1794 	struct mm_struct *mm;
1795 	char buffer[PROC_NUMBUF];
1796 	size_t len;
1797 	int ret;
1798 
1799 	if (!task)
1800 		return -ESRCH;
1801 
1802 	ret = 0;
1803 	mm = get_task_mm(task);
1804 	if (mm) {
1805 		len = snprintf(buffer, sizeof(buffer), "%08lx\n",
1806 			       ((mm->flags & MMF_DUMP_FILTER_MASK) >>
1807 				MMF_DUMP_FILTER_SHIFT));
1808 		mmput(mm);
1809 		ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
1810 	}
1811 
1812 	put_task_struct(task);
1813 
1814 	return ret;
1815 }
1816 
1817 static ssize_t proc_coredump_filter_write(struct file *file,
1818 					  const char __user *buf,
1819 					  size_t count,
1820 					  loff_t *ppos)
1821 {
1822 	struct task_struct *task;
1823 	struct mm_struct *mm;
1824 	char buffer[PROC_NUMBUF], *end;
1825 	unsigned int val;
1826 	int ret;
1827 	int i;
1828 	unsigned long mask;
1829 
1830 	ret = -EFAULT;
1831 	memset(buffer, 0, sizeof(buffer));
1832 	if (count > sizeof(buffer) - 1)
1833 		count = sizeof(buffer) - 1;
1834 	if (copy_from_user(buffer, buf, count))
1835 		goto out_no_task;
1836 
1837 	ret = -EINVAL;
1838 	val = (unsigned int)simple_strtoul(buffer, &end, 0);
1839 	if (*end == '\n')
1840 		end++;
1841 	if (end - buffer == 0)
1842 		goto out_no_task;
1843 
1844 	ret = -ESRCH;
1845 	task = get_proc_task(file->f_dentry->d_inode);
1846 	if (!task)
1847 		goto out_no_task;
1848 
1849 	ret = end - buffer;
1850 	mm = get_task_mm(task);
1851 	if (!mm)
1852 		goto out_no_mm;
1853 
1854 	for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
1855 		if (val & mask)
1856 			set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
1857 		else
1858 			clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
1859 	}
1860 
1861 	mmput(mm);
1862  out_no_mm:
1863 	put_task_struct(task);
1864  out_no_task:
1865 	return ret;
1866 }
1867 
1868 static const struct file_operations proc_coredump_filter_operations = {
1869 	.read		= proc_coredump_filter_read,
1870 	.write		= proc_coredump_filter_write,
1871 };
1872 #endif
1873 
1874 /*
1875  * /proc/self:
1876  */
1877 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1878 			      int buflen)
1879 {
1880 	char tmp[PROC_NUMBUF];
1881 	sprintf(tmp, "%d", current->tgid);
1882 	return vfs_readlink(dentry,buffer,buflen,tmp);
1883 }
1884 
1885 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
1886 {
1887 	char tmp[PROC_NUMBUF];
1888 	sprintf(tmp, "%d", current->tgid);
1889 	return ERR_PTR(vfs_follow_link(nd,tmp));
1890 }
1891 
1892 static const struct inode_operations proc_self_inode_operations = {
1893 	.readlink	= proc_self_readlink,
1894 	.follow_link	= proc_self_follow_link,
1895 };
1896 
1897 /*
1898  * proc base
1899  *
1900  * These are the directory entries in the root directory of /proc
1901  * that properly belong to the /proc filesystem, as they describe
1902  * describe something that is process related.
1903  */
1904 static const struct pid_entry proc_base_stuff[] = {
1905 	NOD("self", S_IFLNK|S_IRWXUGO,
1906 		&proc_self_inode_operations, NULL, {}),
1907 };
1908 
1909 /*
1910  *	Exceptional case: normally we are not allowed to unhash a busy
1911  * directory. In this case, however, we can do it - no aliasing problems
1912  * due to the way we treat inodes.
1913  */
1914 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
1915 {
1916 	struct inode *inode = dentry->d_inode;
1917 	struct task_struct *task = get_proc_task(inode);
1918 	if (task) {
1919 		put_task_struct(task);
1920 		return 1;
1921 	}
1922 	d_drop(dentry);
1923 	return 0;
1924 }
1925 
1926 static struct dentry_operations proc_base_dentry_operations =
1927 {
1928 	.d_revalidate	= proc_base_revalidate,
1929 	.d_delete	= pid_delete_dentry,
1930 };
1931 
1932 static struct dentry *proc_base_instantiate(struct inode *dir,
1933 	struct dentry *dentry, struct task_struct *task, const void *ptr)
1934 {
1935 	const struct pid_entry *p = ptr;
1936 	struct inode *inode;
1937 	struct proc_inode *ei;
1938 	struct dentry *error = ERR_PTR(-EINVAL);
1939 
1940 	/* Allocate the inode */
1941 	error = ERR_PTR(-ENOMEM);
1942 	inode = new_inode(dir->i_sb);
1943 	if (!inode)
1944 		goto out;
1945 
1946 	/* Initialize the inode */
1947 	ei = PROC_I(inode);
1948 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1949 
1950 	/*
1951 	 * grab the reference to the task.
1952 	 */
1953 	ei->pid = get_task_pid(task, PIDTYPE_PID);
1954 	if (!ei->pid)
1955 		goto out_iput;
1956 
1957 	inode->i_uid = 0;
1958 	inode->i_gid = 0;
1959 	inode->i_mode = p->mode;
1960 	if (S_ISDIR(inode->i_mode))
1961 		inode->i_nlink = 2;
1962 	if (S_ISLNK(inode->i_mode))
1963 		inode->i_size = 64;
1964 	if (p->iop)
1965 		inode->i_op = p->iop;
1966 	if (p->fop)
1967 		inode->i_fop = p->fop;
1968 	ei->op = p->op;
1969 	dentry->d_op = &proc_base_dentry_operations;
1970 	d_add(dentry, inode);
1971 	error = NULL;
1972 out:
1973 	return error;
1974 out_iput:
1975 	iput(inode);
1976 	goto out;
1977 }
1978 
1979 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
1980 {
1981 	struct dentry *error;
1982 	struct task_struct *task = get_proc_task(dir);
1983 	const struct pid_entry *p, *last;
1984 
1985 	error = ERR_PTR(-ENOENT);
1986 
1987 	if (!task)
1988 		goto out_no_task;
1989 
1990 	/* Lookup the directory entry */
1991 	last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
1992 	for (p = proc_base_stuff; p <= last; p++) {
1993 		if (p->len != dentry->d_name.len)
1994 			continue;
1995 		if (!memcmp(dentry->d_name.name, p->name, p->len))
1996 			break;
1997 	}
1998 	if (p > last)
1999 		goto out;
2000 
2001 	error = proc_base_instantiate(dir, dentry, task, p);
2002 
2003 out:
2004 	put_task_struct(task);
2005 out_no_task:
2006 	return error;
2007 }
2008 
2009 static int proc_base_fill_cache(struct file *filp, void *dirent,
2010 	filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2011 {
2012 	return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2013 				proc_base_instantiate, task, p);
2014 }
2015 
2016 #ifdef CONFIG_TASK_IO_ACCOUNTING
2017 static int proc_pid_io_accounting(struct task_struct *task, char *buffer)
2018 {
2019 	return sprintf(buffer,
2020 #ifdef CONFIG_TASK_XACCT
2021 			"rchar: %llu\n"
2022 			"wchar: %llu\n"
2023 			"syscr: %llu\n"
2024 			"syscw: %llu\n"
2025 #endif
2026 			"read_bytes: %llu\n"
2027 			"write_bytes: %llu\n"
2028 			"cancelled_write_bytes: %llu\n",
2029 #ifdef CONFIG_TASK_XACCT
2030 			(unsigned long long)task->rchar,
2031 			(unsigned long long)task->wchar,
2032 			(unsigned long long)task->syscr,
2033 			(unsigned long long)task->syscw,
2034 #endif
2035 			(unsigned long long)task->ioac.read_bytes,
2036 			(unsigned long long)task->ioac.write_bytes,
2037 			(unsigned long long)task->ioac.cancelled_write_bytes);
2038 }
2039 #endif
2040 
2041 /*
2042  * Thread groups
2043  */
2044 static const struct file_operations proc_task_operations;
2045 static const struct inode_operations proc_task_inode_operations;
2046 
2047 static const struct pid_entry tgid_base_stuff[] = {
2048 	DIR("task",       S_IRUGO|S_IXUGO, task),
2049 	DIR("fd",         S_IRUSR|S_IXUSR, fd),
2050 	DIR("fdinfo",     S_IRUSR|S_IXUSR, fdinfo),
2051 	INF("environ",    S_IRUSR, pid_environ),
2052 	INF("auxv",       S_IRUSR, pid_auxv),
2053 	INF("status",     S_IRUGO, pid_status),
2054 #ifdef CONFIG_SCHED_DEBUG
2055 	REG("sched",      S_IRUGO|S_IWUSR, pid_sched),
2056 #endif
2057 	INF("cmdline",    S_IRUGO, pid_cmdline),
2058 	INF("stat",       S_IRUGO, tgid_stat),
2059 	INF("statm",      S_IRUGO, pid_statm),
2060 	REG("maps",       S_IRUGO, maps),
2061 #ifdef CONFIG_NUMA
2062 	REG("numa_maps",  S_IRUGO, numa_maps),
2063 #endif
2064 	REG("mem",        S_IRUSR|S_IWUSR, mem),
2065 	LNK("cwd",        cwd),
2066 	LNK("root",       root),
2067 	LNK("exe",        exe),
2068 	REG("mounts",     S_IRUGO, mounts),
2069 	REG("mountstats", S_IRUSR, mountstats),
2070 #ifdef CONFIG_MMU
2071 	REG("clear_refs", S_IWUSR, clear_refs),
2072 	REG("smaps",      S_IRUGO, smaps),
2073 #endif
2074 #ifdef CONFIG_SECURITY
2075 	DIR("attr",       S_IRUGO|S_IXUGO, attr_dir),
2076 #endif
2077 #ifdef CONFIG_KALLSYMS
2078 	INF("wchan",      S_IRUGO, pid_wchan),
2079 #endif
2080 #ifdef CONFIG_SCHEDSTATS
2081 	INF("schedstat",  S_IRUGO, pid_schedstat),
2082 #endif
2083 #ifdef CONFIG_CPUSETS
2084 	REG("cpuset",     S_IRUGO, cpuset),
2085 #endif
2086 	INF("oom_score",  S_IRUGO, oom_score),
2087 	REG("oom_adj",    S_IRUGO|S_IWUSR, oom_adjust),
2088 #ifdef CONFIG_AUDITSYSCALL
2089 	REG("loginuid",   S_IWUSR|S_IRUGO, loginuid),
2090 #endif
2091 #ifdef CONFIG_FAULT_INJECTION
2092 	REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2093 #endif
2094 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2095 	REG("coredump_filter", S_IRUGO|S_IWUSR, coredump_filter),
2096 #endif
2097 #ifdef CONFIG_TASK_IO_ACCOUNTING
2098 	INF("io",	S_IRUGO, pid_io_accounting),
2099 #endif
2100 };
2101 
2102 static int proc_tgid_base_readdir(struct file * filp,
2103 			     void * dirent, filldir_t filldir)
2104 {
2105 	return proc_pident_readdir(filp,dirent,filldir,
2106 				   tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2107 }
2108 
2109 static const struct file_operations proc_tgid_base_operations = {
2110 	.read		= generic_read_dir,
2111 	.readdir	= proc_tgid_base_readdir,
2112 };
2113 
2114 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2115 	return proc_pident_lookup(dir, dentry,
2116 				  tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2117 }
2118 
2119 static const struct inode_operations proc_tgid_base_inode_operations = {
2120 	.lookup		= proc_tgid_base_lookup,
2121 	.getattr	= pid_getattr,
2122 	.setattr	= proc_setattr,
2123 };
2124 
2125 /**
2126  * proc_flush_task -  Remove dcache entries for @task from the /proc dcache.
2127  *
2128  * @task: task that should be flushed.
2129  *
2130  * Looks in the dcache for
2131  * /proc/@pid
2132  * /proc/@tgid/task/@pid
2133  * if either directory is present flushes it and all of it'ts children
2134  * from the dcache.
2135  *
2136  * It is safe and reasonable to cache /proc entries for a task until
2137  * that task exits.  After that they just clog up the dcache with
2138  * useless entries, possibly causing useful dcache entries to be
2139  * flushed instead.  This routine is proved to flush those useless
2140  * dcache entries at process exit time.
2141  *
2142  * NOTE: This routine is just an optimization so it does not guarantee
2143  *       that no dcache entries will exist at process exit time it
2144  *       just makes it very unlikely that any will persist.
2145  */
2146 void proc_flush_task(struct task_struct *task)
2147 {
2148 	struct dentry *dentry, *leader, *dir;
2149 	char buf[PROC_NUMBUF];
2150 	struct qstr name;
2151 
2152 	name.name = buf;
2153 	name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2154 	dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2155 	if (dentry) {
2156 		shrink_dcache_parent(dentry);
2157 		d_drop(dentry);
2158 		dput(dentry);
2159 	}
2160 
2161 	if (thread_group_leader(task))
2162 		goto out;
2163 
2164 	name.name = buf;
2165 	name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
2166 	leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
2167 	if (!leader)
2168 		goto out;
2169 
2170 	name.name = "task";
2171 	name.len = strlen(name.name);
2172 	dir = d_hash_and_lookup(leader, &name);
2173 	if (!dir)
2174 		goto out_put_leader;
2175 
2176 	name.name = buf;
2177 	name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
2178 	dentry = d_hash_and_lookup(dir, &name);
2179 	if (dentry) {
2180 		shrink_dcache_parent(dentry);
2181 		d_drop(dentry);
2182 		dput(dentry);
2183 	}
2184 
2185 	dput(dir);
2186 out_put_leader:
2187 	dput(leader);
2188 out:
2189 	return;
2190 }
2191 
2192 static struct dentry *proc_pid_instantiate(struct inode *dir,
2193 					   struct dentry * dentry,
2194 					   struct task_struct *task, const void *ptr)
2195 {
2196 	struct dentry *error = ERR_PTR(-ENOENT);
2197 	struct inode *inode;
2198 
2199 	inode = proc_pid_make_inode(dir->i_sb, task);
2200 	if (!inode)
2201 		goto out;
2202 
2203 	inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2204 	inode->i_op = &proc_tgid_base_inode_operations;
2205 	inode->i_fop = &proc_tgid_base_operations;
2206 	inode->i_flags|=S_IMMUTABLE;
2207 	inode->i_nlink = 5;
2208 #ifdef CONFIG_SECURITY
2209 	inode->i_nlink += 1;
2210 #endif
2211 
2212 	dentry->d_op = &pid_dentry_operations;
2213 
2214 	d_add(dentry, inode);
2215 	/* Close the race of the process dying before we return the dentry */
2216 	if (pid_revalidate(dentry, NULL))
2217 		error = NULL;
2218 out:
2219 	return error;
2220 }
2221 
2222 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2223 {
2224 	struct dentry *result = ERR_PTR(-ENOENT);
2225 	struct task_struct *task;
2226 	unsigned tgid;
2227 
2228 	result = proc_base_lookup(dir, dentry);
2229 	if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2230 		goto out;
2231 
2232 	tgid = name_to_int(dentry);
2233 	if (tgid == ~0U)
2234 		goto out;
2235 
2236 	rcu_read_lock();
2237 	task = find_task_by_pid(tgid);
2238 	if (task)
2239 		get_task_struct(task);
2240 	rcu_read_unlock();
2241 	if (!task)
2242 		goto out;
2243 
2244 	result = proc_pid_instantiate(dir, dentry, task, NULL);
2245 	put_task_struct(task);
2246 out:
2247 	return result;
2248 }
2249 
2250 /*
2251  * Find the first task with tgid >= tgid
2252  *
2253  */
2254 static struct task_struct *next_tgid(unsigned int tgid)
2255 {
2256 	struct task_struct *task;
2257 	struct pid *pid;
2258 
2259 	rcu_read_lock();
2260 retry:
2261 	task = NULL;
2262 	pid = find_ge_pid(tgid);
2263 	if (pid) {
2264 		tgid = pid->nr + 1;
2265 		task = pid_task(pid, PIDTYPE_PID);
2266 		/* What we to know is if the pid we have find is the
2267 		 * pid of a thread_group_leader.  Testing for task
2268 		 * being a thread_group_leader is the obvious thing
2269 		 * todo but there is a window when it fails, due to
2270 		 * the pid transfer logic in de_thread.
2271 		 *
2272 		 * So we perform the straight forward test of seeing
2273 		 * if the pid we have found is the pid of a thread
2274 		 * group leader, and don't worry if the task we have
2275 		 * found doesn't happen to be a thread group leader.
2276 		 * As we don't care in the case of readdir.
2277 		 */
2278 		if (!task || !has_group_leader_pid(task))
2279 			goto retry;
2280 		get_task_struct(task);
2281 	}
2282 	rcu_read_unlock();
2283 	return task;
2284 }
2285 
2286 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
2287 
2288 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2289 	struct task_struct *task, int tgid)
2290 {
2291 	char name[PROC_NUMBUF];
2292 	int len = snprintf(name, sizeof(name), "%d", tgid);
2293 	return proc_fill_cache(filp, dirent, filldir, name, len,
2294 				proc_pid_instantiate, task, NULL);
2295 }
2296 
2297 /* for the /proc/ directory itself, after non-process stuff has been done */
2298 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2299 {
2300 	unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
2301 	struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
2302 	struct task_struct *task;
2303 	int tgid;
2304 
2305 	if (!reaper)
2306 		goto out_no_task;
2307 
2308 	for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
2309 		const struct pid_entry *p = &proc_base_stuff[nr];
2310 		if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
2311 			goto out;
2312 	}
2313 
2314 	tgid = filp->f_pos - TGID_OFFSET;
2315 	for (task = next_tgid(tgid);
2316 	     task;
2317 	     put_task_struct(task), task = next_tgid(tgid + 1)) {
2318 		tgid = task->pid;
2319 		filp->f_pos = tgid + TGID_OFFSET;
2320 		if (proc_pid_fill_cache(filp, dirent, filldir, task, tgid) < 0) {
2321 			put_task_struct(task);
2322 			goto out;
2323 		}
2324 	}
2325 	filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2326 out:
2327 	put_task_struct(reaper);
2328 out_no_task:
2329 	return 0;
2330 }
2331 
2332 /*
2333  * Tasks
2334  */
2335 static const struct pid_entry tid_base_stuff[] = {
2336 	DIR("fd",        S_IRUSR|S_IXUSR, fd),
2337 	DIR("fdinfo",    S_IRUSR|S_IXUSR, fdinfo),
2338 	INF("environ",   S_IRUSR, pid_environ),
2339 	INF("auxv",      S_IRUSR, pid_auxv),
2340 	INF("status",    S_IRUGO, pid_status),
2341 #ifdef CONFIG_SCHED_DEBUG
2342 	REG("sched",     S_IRUGO|S_IWUSR, pid_sched),
2343 #endif
2344 	INF("cmdline",   S_IRUGO, pid_cmdline),
2345 	INF("stat",      S_IRUGO, tid_stat),
2346 	INF("statm",     S_IRUGO, pid_statm),
2347 	REG("maps",      S_IRUGO, maps),
2348 #ifdef CONFIG_NUMA
2349 	REG("numa_maps", S_IRUGO, numa_maps),
2350 #endif
2351 	REG("mem",       S_IRUSR|S_IWUSR, mem),
2352 	LNK("cwd",       cwd),
2353 	LNK("root",      root),
2354 	LNK("exe",       exe),
2355 	REG("mounts",    S_IRUGO, mounts),
2356 #ifdef CONFIG_MMU
2357 	REG("clear_refs", S_IWUSR, clear_refs),
2358 	REG("smaps",     S_IRUGO, smaps),
2359 #endif
2360 #ifdef CONFIG_SECURITY
2361 	DIR("attr",      S_IRUGO|S_IXUGO, attr_dir),
2362 #endif
2363 #ifdef CONFIG_KALLSYMS
2364 	INF("wchan",     S_IRUGO, pid_wchan),
2365 #endif
2366 #ifdef CONFIG_SCHEDSTATS
2367 	INF("schedstat", S_IRUGO, pid_schedstat),
2368 #endif
2369 #ifdef CONFIG_CPUSETS
2370 	REG("cpuset",    S_IRUGO, cpuset),
2371 #endif
2372 	INF("oom_score", S_IRUGO, oom_score),
2373 	REG("oom_adj",   S_IRUGO|S_IWUSR, oom_adjust),
2374 #ifdef CONFIG_AUDITSYSCALL
2375 	REG("loginuid",  S_IWUSR|S_IRUGO, loginuid),
2376 #endif
2377 #ifdef CONFIG_FAULT_INJECTION
2378 	REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2379 #endif
2380 };
2381 
2382 static int proc_tid_base_readdir(struct file * filp,
2383 			     void * dirent, filldir_t filldir)
2384 {
2385 	return proc_pident_readdir(filp,dirent,filldir,
2386 				   tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2387 }
2388 
2389 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2390 	return proc_pident_lookup(dir, dentry,
2391 				  tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
2392 }
2393 
2394 static const struct file_operations proc_tid_base_operations = {
2395 	.read		= generic_read_dir,
2396 	.readdir	= proc_tid_base_readdir,
2397 };
2398 
2399 static const struct inode_operations proc_tid_base_inode_operations = {
2400 	.lookup		= proc_tid_base_lookup,
2401 	.getattr	= pid_getattr,
2402 	.setattr	= proc_setattr,
2403 };
2404 
2405 static struct dentry *proc_task_instantiate(struct inode *dir,
2406 	struct dentry *dentry, struct task_struct *task, const void *ptr)
2407 {
2408 	struct dentry *error = ERR_PTR(-ENOENT);
2409 	struct inode *inode;
2410 	inode = proc_pid_make_inode(dir->i_sb, task);
2411 
2412 	if (!inode)
2413 		goto out;
2414 	inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2415 	inode->i_op = &proc_tid_base_inode_operations;
2416 	inode->i_fop = &proc_tid_base_operations;
2417 	inode->i_flags|=S_IMMUTABLE;
2418 	inode->i_nlink = 4;
2419 #ifdef CONFIG_SECURITY
2420 	inode->i_nlink += 1;
2421 #endif
2422 
2423 	dentry->d_op = &pid_dentry_operations;
2424 
2425 	d_add(dentry, inode);
2426 	/* Close the race of the process dying before we return the dentry */
2427 	if (pid_revalidate(dentry, NULL))
2428 		error = NULL;
2429 out:
2430 	return error;
2431 }
2432 
2433 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2434 {
2435 	struct dentry *result = ERR_PTR(-ENOENT);
2436 	struct task_struct *task;
2437 	struct task_struct *leader = get_proc_task(dir);
2438 	unsigned tid;
2439 
2440 	if (!leader)
2441 		goto out_no_task;
2442 
2443 	tid = name_to_int(dentry);
2444 	if (tid == ~0U)
2445 		goto out;
2446 
2447 	rcu_read_lock();
2448 	task = find_task_by_pid(tid);
2449 	if (task)
2450 		get_task_struct(task);
2451 	rcu_read_unlock();
2452 	if (!task)
2453 		goto out;
2454 	if (leader->tgid != task->tgid)
2455 		goto out_drop_task;
2456 
2457 	result = proc_task_instantiate(dir, dentry, task, NULL);
2458 out_drop_task:
2459 	put_task_struct(task);
2460 out:
2461 	put_task_struct(leader);
2462 out_no_task:
2463 	return result;
2464 }
2465 
2466 /*
2467  * Find the first tid of a thread group to return to user space.
2468  *
2469  * Usually this is just the thread group leader, but if the users
2470  * buffer was too small or there was a seek into the middle of the
2471  * directory we have more work todo.
2472  *
2473  * In the case of a short read we start with find_task_by_pid.
2474  *
2475  * In the case of a seek we start with the leader and walk nr
2476  * threads past it.
2477  */
2478 static struct task_struct *first_tid(struct task_struct *leader,
2479 					int tid, int nr)
2480 {
2481 	struct task_struct *pos;
2482 
2483 	rcu_read_lock();
2484 	/* Attempt to start with the pid of a thread */
2485 	if (tid && (nr > 0)) {
2486 		pos = find_task_by_pid(tid);
2487 		if (pos && (pos->group_leader == leader))
2488 			goto found;
2489 	}
2490 
2491 	/* If nr exceeds the number of threads there is nothing todo */
2492 	pos = NULL;
2493 	if (nr && nr >= get_nr_threads(leader))
2494 		goto out;
2495 
2496 	/* If we haven't found our starting place yet start
2497 	 * with the leader and walk nr threads forward.
2498 	 */
2499 	for (pos = leader; nr > 0; --nr) {
2500 		pos = next_thread(pos);
2501 		if (pos == leader) {
2502 			pos = NULL;
2503 			goto out;
2504 		}
2505 	}
2506 found:
2507 	get_task_struct(pos);
2508 out:
2509 	rcu_read_unlock();
2510 	return pos;
2511 }
2512 
2513 /*
2514  * Find the next thread in the thread list.
2515  * Return NULL if there is an error or no next thread.
2516  *
2517  * The reference to the input task_struct is released.
2518  */
2519 static struct task_struct *next_tid(struct task_struct *start)
2520 {
2521 	struct task_struct *pos = NULL;
2522 	rcu_read_lock();
2523 	if (pid_alive(start)) {
2524 		pos = next_thread(start);
2525 		if (thread_group_leader(pos))
2526 			pos = NULL;
2527 		else
2528 			get_task_struct(pos);
2529 	}
2530 	rcu_read_unlock();
2531 	put_task_struct(start);
2532 	return pos;
2533 }
2534 
2535 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2536 	struct task_struct *task, int tid)
2537 {
2538 	char name[PROC_NUMBUF];
2539 	int len = snprintf(name, sizeof(name), "%d", tid);
2540 	return proc_fill_cache(filp, dirent, filldir, name, len,
2541 				proc_task_instantiate, task, NULL);
2542 }
2543 
2544 /* for the /proc/TGID/task/ directories */
2545 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2546 {
2547 	struct dentry *dentry = filp->f_path.dentry;
2548 	struct inode *inode = dentry->d_inode;
2549 	struct task_struct *leader = NULL;
2550 	struct task_struct *task;
2551 	int retval = -ENOENT;
2552 	ino_t ino;
2553 	int tid;
2554 	unsigned long pos = filp->f_pos;  /* avoiding "long long" filp->f_pos */
2555 
2556 	task = get_proc_task(inode);
2557 	if (!task)
2558 		goto out_no_task;
2559 	rcu_read_lock();
2560 	if (pid_alive(task)) {
2561 		leader = task->group_leader;
2562 		get_task_struct(leader);
2563 	}
2564 	rcu_read_unlock();
2565 	put_task_struct(task);
2566 	if (!leader)
2567 		goto out_no_task;
2568 	retval = 0;
2569 
2570 	switch (pos) {
2571 	case 0:
2572 		ino = inode->i_ino;
2573 		if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2574 			goto out;
2575 		pos++;
2576 		/* fall through */
2577 	case 1:
2578 		ino = parent_ino(dentry);
2579 		if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2580 			goto out;
2581 		pos++;
2582 		/* fall through */
2583 	}
2584 
2585 	/* f_version caches the tgid value that the last readdir call couldn't
2586 	 * return. lseek aka telldir automagically resets f_version to 0.
2587 	 */
2588 	tid = filp->f_version;
2589 	filp->f_version = 0;
2590 	for (task = first_tid(leader, tid, pos - 2);
2591 	     task;
2592 	     task = next_tid(task), pos++) {
2593 		tid = task->pid;
2594 		if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
2595 			/* returning this tgid failed, save it as the first
2596 			 * pid for the next readir call */
2597 			filp->f_version = tid;
2598 			put_task_struct(task);
2599 			break;
2600 		}
2601 	}
2602 out:
2603 	filp->f_pos = pos;
2604 	put_task_struct(leader);
2605 out_no_task:
2606 	return retval;
2607 }
2608 
2609 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2610 {
2611 	struct inode *inode = dentry->d_inode;
2612 	struct task_struct *p = get_proc_task(inode);
2613 	generic_fillattr(inode, stat);
2614 
2615 	if (p) {
2616 		rcu_read_lock();
2617 		stat->nlink += get_nr_threads(p);
2618 		rcu_read_unlock();
2619 		put_task_struct(p);
2620 	}
2621 
2622 	return 0;
2623 }
2624 
2625 static const struct inode_operations proc_task_inode_operations = {
2626 	.lookup		= proc_task_lookup,
2627 	.getattr	= proc_task_getattr,
2628 	.setattr	= proc_setattr,
2629 };
2630 
2631 static const struct file_operations proc_task_operations = {
2632 	.read		= generic_read_dir,
2633 	.readdir	= proc_task_readdir,
2634 };
2635