xref: /linux/fs/proc/base.c (revision cb299ba8b5ef2239429484072fea394cd7581bd7)
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/fs_struct.h>
85 #include <linux/slab.h>
86 #include "internal.h"
87 
88 /* NOTE:
89  *	Implementing inode permission operations in /proc is almost
90  *	certainly an error.  Permission checks need to happen during
91  *	each system call not at open time.  The reason is that most of
92  *	what we wish to check for permissions in /proc varies at runtime.
93  *
94  *	The classic example of a problem is opening file descriptors
95  *	in /proc for a task before it execs a suid executable.
96  */
97 
98 struct pid_entry {
99 	char *name;
100 	int len;
101 	mode_t mode;
102 	const struct inode_operations *iop;
103 	const struct file_operations *fop;
104 	union proc_op op;
105 };
106 
107 #define NOD(NAME, MODE, IOP, FOP, OP) {			\
108 	.name = (NAME),					\
109 	.len  = sizeof(NAME) - 1,			\
110 	.mode = MODE,					\
111 	.iop  = IOP,					\
112 	.fop  = FOP,					\
113 	.op   = OP,					\
114 }
115 
116 #define DIR(NAME, MODE, iops, fops)	\
117 	NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
118 #define LNK(NAME, get_link)					\
119 	NOD(NAME, (S_IFLNK|S_IRWXUGO),				\
120 		&proc_pid_link_inode_operations, NULL,		\
121 		{ .proc_get_link = get_link } )
122 #define REG(NAME, MODE, fops)				\
123 	NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
124 #define INF(NAME, MODE, read)				\
125 	NOD(NAME, (S_IFREG|(MODE)), 			\
126 		NULL, &proc_info_file_operations,	\
127 		{ .proc_read = read } )
128 #define ONE(NAME, MODE, show)				\
129 	NOD(NAME, (S_IFREG|(MODE)), 			\
130 		NULL, &proc_single_file_operations,	\
131 		{ .proc_show = show } )
132 
133 /*
134  * Count the number of hardlinks for the pid_entry table, excluding the .
135  * and .. links.
136  */
137 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
138 	unsigned int n)
139 {
140 	unsigned int i;
141 	unsigned int count;
142 
143 	count = 0;
144 	for (i = 0; i < n; ++i) {
145 		if (S_ISDIR(entries[i].mode))
146 			++count;
147 	}
148 
149 	return count;
150 }
151 
152 static int get_task_root(struct task_struct *task, struct path *root)
153 {
154 	int result = -ENOENT;
155 
156 	task_lock(task);
157 	if (task->fs) {
158 		get_fs_root(task->fs, root);
159 		result = 0;
160 	}
161 	task_unlock(task);
162 	return result;
163 }
164 
165 static int proc_cwd_link(struct inode *inode, struct path *path)
166 {
167 	struct task_struct *task = get_proc_task(inode);
168 	int result = -ENOENT;
169 
170 	if (task) {
171 		task_lock(task);
172 		if (task->fs) {
173 			get_fs_pwd(task->fs, path);
174 			result = 0;
175 		}
176 		task_unlock(task);
177 		put_task_struct(task);
178 	}
179 	return result;
180 }
181 
182 static int proc_root_link(struct inode *inode, struct path *path)
183 {
184 	struct task_struct *task = get_proc_task(inode);
185 	int result = -ENOENT;
186 
187 	if (task) {
188 		result = get_task_root(task, path);
189 		put_task_struct(task);
190 	}
191 	return result;
192 }
193 
194 /*
195  * Return zero if current may access user memory in @task, -error if not.
196  */
197 static int check_mem_permission(struct task_struct *task)
198 {
199 	/*
200 	 * A task can always look at itself, in case it chooses
201 	 * to use system calls instead of load instructions.
202 	 */
203 	if (task == current)
204 		return 0;
205 
206 	/*
207 	 * If current is actively ptrace'ing, and would also be
208 	 * permitted to freshly attach with ptrace now, permit it.
209 	 */
210 	if (task_is_stopped_or_traced(task)) {
211 		int match;
212 		rcu_read_lock();
213 		match = (tracehook_tracer_task(task) == current);
214 		rcu_read_unlock();
215 		if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
216 			return 0;
217 	}
218 
219 	/*
220 	 * Noone else is allowed.
221 	 */
222 	return -EPERM;
223 }
224 
225 struct mm_struct *mm_for_maps(struct task_struct *task)
226 {
227 	struct mm_struct *mm;
228 
229 	if (mutex_lock_killable(&task->signal->cred_guard_mutex))
230 		return NULL;
231 
232 	mm = get_task_mm(task);
233 	if (mm && mm != current->mm &&
234 			!ptrace_may_access(task, PTRACE_MODE_READ)) {
235 		mmput(mm);
236 		mm = NULL;
237 	}
238 	mutex_unlock(&task->signal->cred_guard_mutex);
239 
240 	return mm;
241 }
242 
243 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
244 {
245 	int res = 0;
246 	unsigned int len;
247 	struct mm_struct *mm = get_task_mm(task);
248 	if (!mm)
249 		goto out;
250 	if (!mm->arg_end)
251 		goto out_mm;	/* Shh! No looking before we're done */
252 
253  	len = mm->arg_end - mm->arg_start;
254 
255 	if (len > PAGE_SIZE)
256 		len = PAGE_SIZE;
257 
258 	res = access_process_vm(task, mm->arg_start, buffer, len, 0);
259 
260 	// If the nul at the end of args has been overwritten, then
261 	// assume application is using setproctitle(3).
262 	if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
263 		len = strnlen(buffer, res);
264 		if (len < res) {
265 		    res = len;
266 		} else {
267 			len = mm->env_end - mm->env_start;
268 			if (len > PAGE_SIZE - res)
269 				len = PAGE_SIZE - res;
270 			res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
271 			res = strnlen(buffer, res);
272 		}
273 	}
274 out_mm:
275 	mmput(mm);
276 out:
277 	return res;
278 }
279 
280 static int proc_pid_auxv(struct task_struct *task, char *buffer)
281 {
282 	int res = 0;
283 	struct mm_struct *mm = get_task_mm(task);
284 	if (mm) {
285 		unsigned int nwords = 0;
286 		do {
287 			nwords += 2;
288 		} while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
289 		res = nwords * sizeof(mm->saved_auxv[0]);
290 		if (res > PAGE_SIZE)
291 			res = PAGE_SIZE;
292 		memcpy(buffer, mm->saved_auxv, res);
293 		mmput(mm);
294 	}
295 	return res;
296 }
297 
298 
299 #ifdef CONFIG_KALLSYMS
300 /*
301  * Provides a wchan file via kallsyms in a proper one-value-per-file format.
302  * Returns the resolved symbol.  If that fails, simply return the address.
303  */
304 static int proc_pid_wchan(struct task_struct *task, char *buffer)
305 {
306 	unsigned long wchan;
307 	char symname[KSYM_NAME_LEN];
308 
309 	wchan = get_wchan(task);
310 
311 	if (lookup_symbol_name(wchan, symname) < 0)
312 		if (!ptrace_may_access(task, PTRACE_MODE_READ))
313 			return 0;
314 		else
315 			return sprintf(buffer, "%lu", wchan);
316 	else
317 		return sprintf(buffer, "%s", symname);
318 }
319 #endif /* CONFIG_KALLSYMS */
320 
321 #ifdef CONFIG_STACKTRACE
322 
323 #define MAX_STACK_TRACE_DEPTH	64
324 
325 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
326 			  struct pid *pid, struct task_struct *task)
327 {
328 	struct stack_trace trace;
329 	unsigned long *entries;
330 	int i;
331 
332 	entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
333 	if (!entries)
334 		return -ENOMEM;
335 
336 	trace.nr_entries	= 0;
337 	trace.max_entries	= MAX_STACK_TRACE_DEPTH;
338 	trace.entries		= entries;
339 	trace.skip		= 0;
340 	save_stack_trace_tsk(task, &trace);
341 
342 	for (i = 0; i < trace.nr_entries; i++) {
343 		seq_printf(m, "[<%p>] %pS\n",
344 			   (void *)entries[i], (void *)entries[i]);
345 	}
346 	kfree(entries);
347 
348 	return 0;
349 }
350 #endif
351 
352 #ifdef CONFIG_SCHEDSTATS
353 /*
354  * Provides /proc/PID/schedstat
355  */
356 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
357 {
358 	return sprintf(buffer, "%llu %llu %lu\n",
359 			(unsigned long long)task->se.sum_exec_runtime,
360 			(unsigned long long)task->sched_info.run_delay,
361 			task->sched_info.pcount);
362 }
363 #endif
364 
365 #ifdef CONFIG_LATENCYTOP
366 static int lstats_show_proc(struct seq_file *m, void *v)
367 {
368 	int i;
369 	struct inode *inode = m->private;
370 	struct task_struct *task = get_proc_task(inode);
371 
372 	if (!task)
373 		return -ESRCH;
374 	seq_puts(m, "Latency Top version : v0.1\n");
375 	for (i = 0; i < 32; i++) {
376 		if (task->latency_record[i].backtrace[0]) {
377 			int q;
378 			seq_printf(m, "%i %li %li ",
379 				task->latency_record[i].count,
380 				task->latency_record[i].time,
381 				task->latency_record[i].max);
382 			for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
383 				char sym[KSYM_SYMBOL_LEN];
384 				char *c;
385 				if (!task->latency_record[i].backtrace[q])
386 					break;
387 				if (task->latency_record[i].backtrace[q] == ULONG_MAX)
388 					break;
389 				sprint_symbol(sym, task->latency_record[i].backtrace[q]);
390 				c = strchr(sym, '+');
391 				if (c)
392 					*c = 0;
393 				seq_printf(m, "%s ", sym);
394 			}
395 			seq_printf(m, "\n");
396 		}
397 
398 	}
399 	put_task_struct(task);
400 	return 0;
401 }
402 
403 static int lstats_open(struct inode *inode, struct file *file)
404 {
405 	return single_open(file, lstats_show_proc, inode);
406 }
407 
408 static ssize_t lstats_write(struct file *file, const char __user *buf,
409 			    size_t count, loff_t *offs)
410 {
411 	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
412 
413 	if (!task)
414 		return -ESRCH;
415 	clear_all_latency_tracing(task);
416 	put_task_struct(task);
417 
418 	return count;
419 }
420 
421 static const struct file_operations proc_lstats_operations = {
422 	.open		= lstats_open,
423 	.read		= seq_read,
424 	.write		= lstats_write,
425 	.llseek		= seq_lseek,
426 	.release	= single_release,
427 };
428 
429 #endif
430 
431 static int proc_oom_score(struct task_struct *task, char *buffer)
432 {
433 	unsigned long points = 0;
434 
435 	read_lock(&tasklist_lock);
436 	if (pid_alive(task))
437 		points = oom_badness(task, NULL, NULL,
438 					totalram_pages + total_swap_pages);
439 	read_unlock(&tasklist_lock);
440 	return sprintf(buffer, "%lu\n", points);
441 }
442 
443 struct limit_names {
444 	char *name;
445 	char *unit;
446 };
447 
448 static const struct limit_names lnames[RLIM_NLIMITS] = {
449 	[RLIMIT_CPU] = {"Max cpu time", "seconds"},
450 	[RLIMIT_FSIZE] = {"Max file size", "bytes"},
451 	[RLIMIT_DATA] = {"Max data size", "bytes"},
452 	[RLIMIT_STACK] = {"Max stack size", "bytes"},
453 	[RLIMIT_CORE] = {"Max core file size", "bytes"},
454 	[RLIMIT_RSS] = {"Max resident set", "bytes"},
455 	[RLIMIT_NPROC] = {"Max processes", "processes"},
456 	[RLIMIT_NOFILE] = {"Max open files", "files"},
457 	[RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
458 	[RLIMIT_AS] = {"Max address space", "bytes"},
459 	[RLIMIT_LOCKS] = {"Max file locks", "locks"},
460 	[RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
461 	[RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
462 	[RLIMIT_NICE] = {"Max nice priority", NULL},
463 	[RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
464 	[RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
465 };
466 
467 /* Display limits for a process */
468 static int proc_pid_limits(struct task_struct *task, char *buffer)
469 {
470 	unsigned int i;
471 	int count = 0;
472 	unsigned long flags;
473 	char *bufptr = buffer;
474 
475 	struct rlimit rlim[RLIM_NLIMITS];
476 
477 	if (!lock_task_sighand(task, &flags))
478 		return 0;
479 	memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
480 	unlock_task_sighand(task, &flags);
481 
482 	/*
483 	 * print the file header
484 	 */
485 	count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
486 			"Limit", "Soft Limit", "Hard Limit", "Units");
487 
488 	for (i = 0; i < RLIM_NLIMITS; i++) {
489 		if (rlim[i].rlim_cur == RLIM_INFINITY)
490 			count += sprintf(&bufptr[count], "%-25s %-20s ",
491 					 lnames[i].name, "unlimited");
492 		else
493 			count += sprintf(&bufptr[count], "%-25s %-20lu ",
494 					 lnames[i].name, rlim[i].rlim_cur);
495 
496 		if (rlim[i].rlim_max == RLIM_INFINITY)
497 			count += sprintf(&bufptr[count], "%-20s ", "unlimited");
498 		else
499 			count += sprintf(&bufptr[count], "%-20lu ",
500 					 rlim[i].rlim_max);
501 
502 		if (lnames[i].unit)
503 			count += sprintf(&bufptr[count], "%-10s\n",
504 					 lnames[i].unit);
505 		else
506 			count += sprintf(&bufptr[count], "\n");
507 	}
508 
509 	return count;
510 }
511 
512 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
513 static int proc_pid_syscall(struct task_struct *task, char *buffer)
514 {
515 	long nr;
516 	unsigned long args[6], sp, pc;
517 
518 	if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
519 		return sprintf(buffer, "running\n");
520 
521 	if (nr < 0)
522 		return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
523 
524 	return sprintf(buffer,
525 		       "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
526 		       nr,
527 		       args[0], args[1], args[2], args[3], args[4], args[5],
528 		       sp, pc);
529 }
530 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
531 
532 /************************************************************************/
533 /*                       Here the fs part begins                        */
534 /************************************************************************/
535 
536 /* permission checks */
537 static int proc_fd_access_allowed(struct inode *inode)
538 {
539 	struct task_struct *task;
540 	int allowed = 0;
541 	/* Allow access to a task's file descriptors if it is us or we
542 	 * may use ptrace attach to the process and find out that
543 	 * information.
544 	 */
545 	task = get_proc_task(inode);
546 	if (task) {
547 		allowed = ptrace_may_access(task, PTRACE_MODE_READ);
548 		put_task_struct(task);
549 	}
550 	return allowed;
551 }
552 
553 static int proc_setattr(struct dentry *dentry, struct iattr *attr)
554 {
555 	int error;
556 	struct inode *inode = dentry->d_inode;
557 
558 	if (attr->ia_valid & ATTR_MODE)
559 		return -EPERM;
560 
561 	error = inode_change_ok(inode, attr);
562 	if (error)
563 		return error;
564 
565 	if ((attr->ia_valid & ATTR_SIZE) &&
566 	    attr->ia_size != i_size_read(inode)) {
567 		error = vmtruncate(inode, attr->ia_size);
568 		if (error)
569 			return error;
570 	}
571 
572 	setattr_copy(inode, attr);
573 	mark_inode_dirty(inode);
574 	return 0;
575 }
576 
577 static const struct inode_operations proc_def_inode_operations = {
578 	.setattr	= proc_setattr,
579 };
580 
581 static int mounts_open_common(struct inode *inode, struct file *file,
582 			      const struct seq_operations *op)
583 {
584 	struct task_struct *task = get_proc_task(inode);
585 	struct nsproxy *nsp;
586 	struct mnt_namespace *ns = NULL;
587 	struct path root;
588 	struct proc_mounts *p;
589 	int ret = -EINVAL;
590 
591 	if (task) {
592 		rcu_read_lock();
593 		nsp = task_nsproxy(task);
594 		if (nsp) {
595 			ns = nsp->mnt_ns;
596 			if (ns)
597 				get_mnt_ns(ns);
598 		}
599 		rcu_read_unlock();
600 		if (ns && get_task_root(task, &root) == 0)
601 			ret = 0;
602 		put_task_struct(task);
603 	}
604 
605 	if (!ns)
606 		goto err;
607 	if (ret)
608 		goto err_put_ns;
609 
610 	ret = -ENOMEM;
611 	p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
612 	if (!p)
613 		goto err_put_path;
614 
615 	file->private_data = &p->m;
616 	ret = seq_open(file, op);
617 	if (ret)
618 		goto err_free;
619 
620 	p->m.private = p;
621 	p->ns = ns;
622 	p->root = root;
623 	p->event = ns->event;
624 
625 	return 0;
626 
627  err_free:
628 	kfree(p);
629  err_put_path:
630 	path_put(&root);
631  err_put_ns:
632 	put_mnt_ns(ns);
633  err:
634 	return ret;
635 }
636 
637 static int mounts_release(struct inode *inode, struct file *file)
638 {
639 	struct proc_mounts *p = file->private_data;
640 	path_put(&p->root);
641 	put_mnt_ns(p->ns);
642 	return seq_release(inode, file);
643 }
644 
645 static unsigned mounts_poll(struct file *file, poll_table *wait)
646 {
647 	struct proc_mounts *p = file->private_data;
648 	unsigned res = POLLIN | POLLRDNORM;
649 
650 	poll_wait(file, &p->ns->poll, wait);
651 	if (mnt_had_events(p))
652 		res |= POLLERR | POLLPRI;
653 
654 	return res;
655 }
656 
657 static int mounts_open(struct inode *inode, struct file *file)
658 {
659 	return mounts_open_common(inode, file, &mounts_op);
660 }
661 
662 static const struct file_operations proc_mounts_operations = {
663 	.open		= mounts_open,
664 	.read		= seq_read,
665 	.llseek		= seq_lseek,
666 	.release	= mounts_release,
667 	.poll		= mounts_poll,
668 };
669 
670 static int mountinfo_open(struct inode *inode, struct file *file)
671 {
672 	return mounts_open_common(inode, file, &mountinfo_op);
673 }
674 
675 static const struct file_operations proc_mountinfo_operations = {
676 	.open		= mountinfo_open,
677 	.read		= seq_read,
678 	.llseek		= seq_lseek,
679 	.release	= mounts_release,
680 	.poll		= mounts_poll,
681 };
682 
683 static int mountstats_open(struct inode *inode, struct file *file)
684 {
685 	return mounts_open_common(inode, file, &mountstats_op);
686 }
687 
688 static const struct file_operations proc_mountstats_operations = {
689 	.open		= mountstats_open,
690 	.read		= seq_read,
691 	.llseek		= seq_lseek,
692 	.release	= mounts_release,
693 };
694 
695 #define PROC_BLOCK_SIZE	(3*1024)		/* 4K page size but our output routines use some slack for overruns */
696 
697 static ssize_t proc_info_read(struct file * file, char __user * buf,
698 			  size_t count, loff_t *ppos)
699 {
700 	struct inode * inode = file->f_path.dentry->d_inode;
701 	unsigned long page;
702 	ssize_t length;
703 	struct task_struct *task = get_proc_task(inode);
704 
705 	length = -ESRCH;
706 	if (!task)
707 		goto out_no_task;
708 
709 	if (count > PROC_BLOCK_SIZE)
710 		count = PROC_BLOCK_SIZE;
711 
712 	length = -ENOMEM;
713 	if (!(page = __get_free_page(GFP_TEMPORARY)))
714 		goto out;
715 
716 	length = PROC_I(inode)->op.proc_read(task, (char*)page);
717 
718 	if (length >= 0)
719 		length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
720 	free_page(page);
721 out:
722 	put_task_struct(task);
723 out_no_task:
724 	return length;
725 }
726 
727 static const struct file_operations proc_info_file_operations = {
728 	.read		= proc_info_read,
729 	.llseek		= generic_file_llseek,
730 };
731 
732 static int proc_single_show(struct seq_file *m, void *v)
733 {
734 	struct inode *inode = m->private;
735 	struct pid_namespace *ns;
736 	struct pid *pid;
737 	struct task_struct *task;
738 	int ret;
739 
740 	ns = inode->i_sb->s_fs_info;
741 	pid = proc_pid(inode);
742 	task = get_pid_task(pid, PIDTYPE_PID);
743 	if (!task)
744 		return -ESRCH;
745 
746 	ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
747 
748 	put_task_struct(task);
749 	return ret;
750 }
751 
752 static int proc_single_open(struct inode *inode, struct file *filp)
753 {
754 	int ret;
755 	ret = single_open(filp, proc_single_show, NULL);
756 	if (!ret) {
757 		struct seq_file *m = filp->private_data;
758 
759 		m->private = inode;
760 	}
761 	return ret;
762 }
763 
764 static const struct file_operations proc_single_file_operations = {
765 	.open		= proc_single_open,
766 	.read		= seq_read,
767 	.llseek		= seq_lseek,
768 	.release	= single_release,
769 };
770 
771 static int mem_open(struct inode* inode, struct file* file)
772 {
773 	file->private_data = (void*)((long)current->self_exec_id);
774 	/* OK to pass negative loff_t, we can catch out-of-range */
775 	file->f_mode |= FMODE_UNSIGNED_OFFSET;
776 	return 0;
777 }
778 
779 static ssize_t mem_read(struct file * file, char __user * buf,
780 			size_t count, loff_t *ppos)
781 {
782 	struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
783 	char *page;
784 	unsigned long src = *ppos;
785 	int ret = -ESRCH;
786 	struct mm_struct *mm;
787 
788 	if (!task)
789 		goto out_no_task;
790 
791 	if (check_mem_permission(task))
792 		goto out;
793 
794 	ret = -ENOMEM;
795 	page = (char *)__get_free_page(GFP_TEMPORARY);
796 	if (!page)
797 		goto out;
798 
799 	ret = 0;
800 
801 	mm = get_task_mm(task);
802 	if (!mm)
803 		goto out_free;
804 
805 	ret = -EIO;
806 
807 	if (file->private_data != (void*)((long)current->self_exec_id))
808 		goto out_put;
809 
810 	ret = 0;
811 
812 	while (count > 0) {
813 		int this_len, retval;
814 
815 		this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
816 		retval = access_process_vm(task, src, page, this_len, 0);
817 		if (!retval || check_mem_permission(task)) {
818 			if (!ret)
819 				ret = -EIO;
820 			break;
821 		}
822 
823 		if (copy_to_user(buf, page, retval)) {
824 			ret = -EFAULT;
825 			break;
826 		}
827 
828 		ret += retval;
829 		src += retval;
830 		buf += retval;
831 		count -= retval;
832 	}
833 	*ppos = src;
834 
835 out_put:
836 	mmput(mm);
837 out_free:
838 	free_page((unsigned long) page);
839 out:
840 	put_task_struct(task);
841 out_no_task:
842 	return ret;
843 }
844 
845 #define mem_write NULL
846 
847 #ifndef mem_write
848 /* This is a security hazard */
849 static ssize_t mem_write(struct file * file, const char __user *buf,
850 			 size_t count, loff_t *ppos)
851 {
852 	int copied;
853 	char *page;
854 	struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
855 	unsigned long dst = *ppos;
856 
857 	copied = -ESRCH;
858 	if (!task)
859 		goto out_no_task;
860 
861 	if (check_mem_permission(task))
862 		goto out;
863 
864 	copied = -ENOMEM;
865 	page = (char *)__get_free_page(GFP_TEMPORARY);
866 	if (!page)
867 		goto out;
868 
869 	copied = 0;
870 	while (count > 0) {
871 		int this_len, retval;
872 
873 		this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
874 		if (copy_from_user(page, buf, this_len)) {
875 			copied = -EFAULT;
876 			break;
877 		}
878 		retval = access_process_vm(task, dst, page, this_len, 1);
879 		if (!retval) {
880 			if (!copied)
881 				copied = -EIO;
882 			break;
883 		}
884 		copied += retval;
885 		buf += retval;
886 		dst += retval;
887 		count -= retval;
888 	}
889 	*ppos = dst;
890 	free_page((unsigned long) page);
891 out:
892 	put_task_struct(task);
893 out_no_task:
894 	return copied;
895 }
896 #endif
897 
898 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
899 {
900 	switch (orig) {
901 	case 0:
902 		file->f_pos = offset;
903 		break;
904 	case 1:
905 		file->f_pos += offset;
906 		break;
907 	default:
908 		return -EINVAL;
909 	}
910 	force_successful_syscall_return();
911 	return file->f_pos;
912 }
913 
914 static const struct file_operations proc_mem_operations = {
915 	.llseek		= mem_lseek,
916 	.read		= mem_read,
917 	.write		= mem_write,
918 	.open		= mem_open,
919 };
920 
921 static ssize_t environ_read(struct file *file, char __user *buf,
922 			size_t count, loff_t *ppos)
923 {
924 	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
925 	char *page;
926 	unsigned long src = *ppos;
927 	int ret = -ESRCH;
928 	struct mm_struct *mm;
929 
930 	if (!task)
931 		goto out_no_task;
932 
933 	if (!ptrace_may_access(task, PTRACE_MODE_READ))
934 		goto out;
935 
936 	ret = -ENOMEM;
937 	page = (char *)__get_free_page(GFP_TEMPORARY);
938 	if (!page)
939 		goto out;
940 
941 	ret = 0;
942 
943 	mm = get_task_mm(task);
944 	if (!mm)
945 		goto out_free;
946 
947 	while (count > 0) {
948 		int this_len, retval, max_len;
949 
950 		this_len = mm->env_end - (mm->env_start + src);
951 
952 		if (this_len <= 0)
953 			break;
954 
955 		max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
956 		this_len = (this_len > max_len) ? max_len : this_len;
957 
958 		retval = access_process_vm(task, (mm->env_start + src),
959 			page, this_len, 0);
960 
961 		if (retval <= 0) {
962 			ret = retval;
963 			break;
964 		}
965 
966 		if (copy_to_user(buf, page, retval)) {
967 			ret = -EFAULT;
968 			break;
969 		}
970 
971 		ret += retval;
972 		src += retval;
973 		buf += retval;
974 		count -= retval;
975 	}
976 	*ppos = src;
977 
978 	mmput(mm);
979 out_free:
980 	free_page((unsigned long) page);
981 out:
982 	put_task_struct(task);
983 out_no_task:
984 	return ret;
985 }
986 
987 static const struct file_operations proc_environ_operations = {
988 	.read		= environ_read,
989 	.llseek		= generic_file_llseek,
990 };
991 
992 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
993 				size_t count, loff_t *ppos)
994 {
995 	struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
996 	char buffer[PROC_NUMBUF];
997 	size_t len;
998 	int oom_adjust = OOM_DISABLE;
999 	unsigned long flags;
1000 
1001 	if (!task)
1002 		return -ESRCH;
1003 
1004 	if (lock_task_sighand(task, &flags)) {
1005 		oom_adjust = task->signal->oom_adj;
1006 		unlock_task_sighand(task, &flags);
1007 	}
1008 
1009 	put_task_struct(task);
1010 
1011 	len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
1012 
1013 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1014 }
1015 
1016 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1017 				size_t count, loff_t *ppos)
1018 {
1019 	struct task_struct *task;
1020 	char buffer[PROC_NUMBUF];
1021 	long oom_adjust;
1022 	unsigned long flags;
1023 	int err;
1024 
1025 	memset(buffer, 0, sizeof(buffer));
1026 	if (count > sizeof(buffer) - 1)
1027 		count = sizeof(buffer) - 1;
1028 	if (copy_from_user(buffer, buf, count)) {
1029 		err = -EFAULT;
1030 		goto out;
1031 	}
1032 
1033 	err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1034 	if (err)
1035 		goto out;
1036 	if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1037 	     oom_adjust != OOM_DISABLE) {
1038 		err = -EINVAL;
1039 		goto out;
1040 	}
1041 
1042 	task = get_proc_task(file->f_path.dentry->d_inode);
1043 	if (!task) {
1044 		err = -ESRCH;
1045 		goto out;
1046 	}
1047 
1048 	task_lock(task);
1049 	if (!task->mm) {
1050 		err = -EINVAL;
1051 		goto err_task_lock;
1052 	}
1053 
1054 	if (!lock_task_sighand(task, &flags)) {
1055 		err = -ESRCH;
1056 		goto err_task_lock;
1057 	}
1058 
1059 	if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1060 		err = -EACCES;
1061 		goto err_sighand;
1062 	}
1063 
1064 	if (oom_adjust != task->signal->oom_adj) {
1065 		if (oom_adjust == OOM_DISABLE)
1066 			atomic_inc(&task->mm->oom_disable_count);
1067 		if (task->signal->oom_adj == OOM_DISABLE)
1068 			atomic_dec(&task->mm->oom_disable_count);
1069 	}
1070 
1071 	/*
1072 	 * Warn that /proc/pid/oom_adj is deprecated, see
1073 	 * Documentation/feature-removal-schedule.txt.
1074 	 */
1075 	printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, "
1076 			"please use /proc/%d/oom_score_adj instead.\n",
1077 			current->comm, task_pid_nr(current),
1078 			task_pid_nr(task), task_pid_nr(task));
1079 	task->signal->oom_adj = oom_adjust;
1080 	/*
1081 	 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1082 	 * value is always attainable.
1083 	 */
1084 	if (task->signal->oom_adj == OOM_ADJUST_MAX)
1085 		task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1086 	else
1087 		task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1088 								-OOM_DISABLE;
1089 err_sighand:
1090 	unlock_task_sighand(task, &flags);
1091 err_task_lock:
1092 	task_unlock(task);
1093 	put_task_struct(task);
1094 out:
1095 	return err < 0 ? err : count;
1096 }
1097 
1098 static const struct file_operations proc_oom_adjust_operations = {
1099 	.read		= oom_adjust_read,
1100 	.write		= oom_adjust_write,
1101 	.llseek		= generic_file_llseek,
1102 };
1103 
1104 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1105 					size_t count, loff_t *ppos)
1106 {
1107 	struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1108 	char buffer[PROC_NUMBUF];
1109 	int oom_score_adj = OOM_SCORE_ADJ_MIN;
1110 	unsigned long flags;
1111 	size_t len;
1112 
1113 	if (!task)
1114 		return -ESRCH;
1115 	if (lock_task_sighand(task, &flags)) {
1116 		oom_score_adj = task->signal->oom_score_adj;
1117 		unlock_task_sighand(task, &flags);
1118 	}
1119 	put_task_struct(task);
1120 	len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1121 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1122 }
1123 
1124 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1125 					size_t count, loff_t *ppos)
1126 {
1127 	struct task_struct *task;
1128 	char buffer[PROC_NUMBUF];
1129 	unsigned long flags;
1130 	long oom_score_adj;
1131 	int err;
1132 
1133 	memset(buffer, 0, sizeof(buffer));
1134 	if (count > sizeof(buffer) - 1)
1135 		count = sizeof(buffer) - 1;
1136 	if (copy_from_user(buffer, buf, count)) {
1137 		err = -EFAULT;
1138 		goto out;
1139 	}
1140 
1141 	err = strict_strtol(strstrip(buffer), 0, &oom_score_adj);
1142 	if (err)
1143 		goto out;
1144 	if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1145 			oom_score_adj > OOM_SCORE_ADJ_MAX) {
1146 		err = -EINVAL;
1147 		goto out;
1148 	}
1149 
1150 	task = get_proc_task(file->f_path.dentry->d_inode);
1151 	if (!task) {
1152 		err = -ESRCH;
1153 		goto out;
1154 	}
1155 
1156 	task_lock(task);
1157 	if (!task->mm) {
1158 		err = -EINVAL;
1159 		goto err_task_lock;
1160 	}
1161 
1162 	if (!lock_task_sighand(task, &flags)) {
1163 		err = -ESRCH;
1164 		goto err_task_lock;
1165 	}
1166 
1167 	if (oom_score_adj < task->signal->oom_score_adj &&
1168 			!capable(CAP_SYS_RESOURCE)) {
1169 		err = -EACCES;
1170 		goto err_sighand;
1171 	}
1172 
1173 	if (oom_score_adj != task->signal->oom_score_adj) {
1174 		if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1175 			atomic_inc(&task->mm->oom_disable_count);
1176 		if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1177 			atomic_dec(&task->mm->oom_disable_count);
1178 	}
1179 	task->signal->oom_score_adj = oom_score_adj;
1180 	/*
1181 	 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1182 	 * always attainable.
1183 	 */
1184 	if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1185 		task->signal->oom_adj = OOM_DISABLE;
1186 	else
1187 		task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1188 							OOM_SCORE_ADJ_MAX;
1189 err_sighand:
1190 	unlock_task_sighand(task, &flags);
1191 err_task_lock:
1192 	task_unlock(task);
1193 	put_task_struct(task);
1194 out:
1195 	return err < 0 ? err : count;
1196 }
1197 
1198 static const struct file_operations proc_oom_score_adj_operations = {
1199 	.read		= oom_score_adj_read,
1200 	.write		= oom_score_adj_write,
1201 	.llseek		= default_llseek,
1202 };
1203 
1204 #ifdef CONFIG_AUDITSYSCALL
1205 #define TMPBUFLEN 21
1206 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1207 				  size_t count, loff_t *ppos)
1208 {
1209 	struct inode * inode = file->f_path.dentry->d_inode;
1210 	struct task_struct *task = get_proc_task(inode);
1211 	ssize_t length;
1212 	char tmpbuf[TMPBUFLEN];
1213 
1214 	if (!task)
1215 		return -ESRCH;
1216 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1217 				audit_get_loginuid(task));
1218 	put_task_struct(task);
1219 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1220 }
1221 
1222 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1223 				   size_t count, loff_t *ppos)
1224 {
1225 	struct inode * inode = file->f_path.dentry->d_inode;
1226 	char *page, *tmp;
1227 	ssize_t length;
1228 	uid_t loginuid;
1229 
1230 	if (!capable(CAP_AUDIT_CONTROL))
1231 		return -EPERM;
1232 
1233 	rcu_read_lock();
1234 	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1235 		rcu_read_unlock();
1236 		return -EPERM;
1237 	}
1238 	rcu_read_unlock();
1239 
1240 	if (count >= PAGE_SIZE)
1241 		count = PAGE_SIZE - 1;
1242 
1243 	if (*ppos != 0) {
1244 		/* No partial writes. */
1245 		return -EINVAL;
1246 	}
1247 	page = (char*)__get_free_page(GFP_TEMPORARY);
1248 	if (!page)
1249 		return -ENOMEM;
1250 	length = -EFAULT;
1251 	if (copy_from_user(page, buf, count))
1252 		goto out_free_page;
1253 
1254 	page[count] = '\0';
1255 	loginuid = simple_strtoul(page, &tmp, 10);
1256 	if (tmp == page) {
1257 		length = -EINVAL;
1258 		goto out_free_page;
1259 
1260 	}
1261 	length = audit_set_loginuid(current, loginuid);
1262 	if (likely(length == 0))
1263 		length = count;
1264 
1265 out_free_page:
1266 	free_page((unsigned long) page);
1267 	return length;
1268 }
1269 
1270 static const struct file_operations proc_loginuid_operations = {
1271 	.read		= proc_loginuid_read,
1272 	.write		= proc_loginuid_write,
1273 	.llseek		= generic_file_llseek,
1274 };
1275 
1276 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1277 				  size_t count, loff_t *ppos)
1278 {
1279 	struct inode * inode = file->f_path.dentry->d_inode;
1280 	struct task_struct *task = get_proc_task(inode);
1281 	ssize_t length;
1282 	char tmpbuf[TMPBUFLEN];
1283 
1284 	if (!task)
1285 		return -ESRCH;
1286 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1287 				audit_get_sessionid(task));
1288 	put_task_struct(task);
1289 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1290 }
1291 
1292 static const struct file_operations proc_sessionid_operations = {
1293 	.read		= proc_sessionid_read,
1294 	.llseek		= generic_file_llseek,
1295 };
1296 #endif
1297 
1298 #ifdef CONFIG_FAULT_INJECTION
1299 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1300 				      size_t count, loff_t *ppos)
1301 {
1302 	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1303 	char buffer[PROC_NUMBUF];
1304 	size_t len;
1305 	int make_it_fail;
1306 
1307 	if (!task)
1308 		return -ESRCH;
1309 	make_it_fail = task->make_it_fail;
1310 	put_task_struct(task);
1311 
1312 	len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1313 
1314 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1315 }
1316 
1317 static ssize_t proc_fault_inject_write(struct file * file,
1318 			const char __user * buf, size_t count, loff_t *ppos)
1319 {
1320 	struct task_struct *task;
1321 	char buffer[PROC_NUMBUF], *end;
1322 	int make_it_fail;
1323 
1324 	if (!capable(CAP_SYS_RESOURCE))
1325 		return -EPERM;
1326 	memset(buffer, 0, sizeof(buffer));
1327 	if (count > sizeof(buffer) - 1)
1328 		count = sizeof(buffer) - 1;
1329 	if (copy_from_user(buffer, buf, count))
1330 		return -EFAULT;
1331 	make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1332 	if (*end)
1333 		return -EINVAL;
1334 	task = get_proc_task(file->f_dentry->d_inode);
1335 	if (!task)
1336 		return -ESRCH;
1337 	task->make_it_fail = make_it_fail;
1338 	put_task_struct(task);
1339 
1340 	return count;
1341 }
1342 
1343 static const struct file_operations proc_fault_inject_operations = {
1344 	.read		= proc_fault_inject_read,
1345 	.write		= proc_fault_inject_write,
1346 	.llseek		= generic_file_llseek,
1347 };
1348 #endif
1349 
1350 
1351 #ifdef CONFIG_SCHED_DEBUG
1352 /*
1353  * Print out various scheduling related per-task fields:
1354  */
1355 static int sched_show(struct seq_file *m, void *v)
1356 {
1357 	struct inode *inode = m->private;
1358 	struct task_struct *p;
1359 
1360 	p = get_proc_task(inode);
1361 	if (!p)
1362 		return -ESRCH;
1363 	proc_sched_show_task(p, m);
1364 
1365 	put_task_struct(p);
1366 
1367 	return 0;
1368 }
1369 
1370 static ssize_t
1371 sched_write(struct file *file, const char __user *buf,
1372 	    size_t count, loff_t *offset)
1373 {
1374 	struct inode *inode = file->f_path.dentry->d_inode;
1375 	struct task_struct *p;
1376 
1377 	p = get_proc_task(inode);
1378 	if (!p)
1379 		return -ESRCH;
1380 	proc_sched_set_task(p);
1381 
1382 	put_task_struct(p);
1383 
1384 	return count;
1385 }
1386 
1387 static int sched_open(struct inode *inode, struct file *filp)
1388 {
1389 	int ret;
1390 
1391 	ret = single_open(filp, sched_show, NULL);
1392 	if (!ret) {
1393 		struct seq_file *m = filp->private_data;
1394 
1395 		m->private = inode;
1396 	}
1397 	return ret;
1398 }
1399 
1400 static const struct file_operations proc_pid_sched_operations = {
1401 	.open		= sched_open,
1402 	.read		= seq_read,
1403 	.write		= sched_write,
1404 	.llseek		= seq_lseek,
1405 	.release	= single_release,
1406 };
1407 
1408 #endif
1409 
1410 static ssize_t comm_write(struct file *file, const char __user *buf,
1411 				size_t count, loff_t *offset)
1412 {
1413 	struct inode *inode = file->f_path.dentry->d_inode;
1414 	struct task_struct *p;
1415 	char buffer[TASK_COMM_LEN];
1416 
1417 	memset(buffer, 0, sizeof(buffer));
1418 	if (count > sizeof(buffer) - 1)
1419 		count = sizeof(buffer) - 1;
1420 	if (copy_from_user(buffer, buf, count))
1421 		return -EFAULT;
1422 
1423 	p = get_proc_task(inode);
1424 	if (!p)
1425 		return -ESRCH;
1426 
1427 	if (same_thread_group(current, p))
1428 		set_task_comm(p, buffer);
1429 	else
1430 		count = -EINVAL;
1431 
1432 	put_task_struct(p);
1433 
1434 	return count;
1435 }
1436 
1437 static int comm_show(struct seq_file *m, void *v)
1438 {
1439 	struct inode *inode = m->private;
1440 	struct task_struct *p;
1441 
1442 	p = get_proc_task(inode);
1443 	if (!p)
1444 		return -ESRCH;
1445 
1446 	task_lock(p);
1447 	seq_printf(m, "%s\n", p->comm);
1448 	task_unlock(p);
1449 
1450 	put_task_struct(p);
1451 
1452 	return 0;
1453 }
1454 
1455 static int comm_open(struct inode *inode, struct file *filp)
1456 {
1457 	int ret;
1458 
1459 	ret = single_open(filp, comm_show, NULL);
1460 	if (!ret) {
1461 		struct seq_file *m = filp->private_data;
1462 
1463 		m->private = inode;
1464 	}
1465 	return ret;
1466 }
1467 
1468 static const struct file_operations proc_pid_set_comm_operations = {
1469 	.open		= comm_open,
1470 	.read		= seq_read,
1471 	.write		= comm_write,
1472 	.llseek		= seq_lseek,
1473 	.release	= single_release,
1474 };
1475 
1476 /*
1477  * We added or removed a vma mapping the executable. The vmas are only mapped
1478  * during exec and are not mapped with the mmap system call.
1479  * Callers must hold down_write() on the mm's mmap_sem for these
1480  */
1481 void added_exe_file_vma(struct mm_struct *mm)
1482 {
1483 	mm->num_exe_file_vmas++;
1484 }
1485 
1486 void removed_exe_file_vma(struct mm_struct *mm)
1487 {
1488 	mm->num_exe_file_vmas--;
1489 	if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1490 		fput(mm->exe_file);
1491 		mm->exe_file = NULL;
1492 	}
1493 
1494 }
1495 
1496 void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1497 {
1498 	if (new_exe_file)
1499 		get_file(new_exe_file);
1500 	if (mm->exe_file)
1501 		fput(mm->exe_file);
1502 	mm->exe_file = new_exe_file;
1503 	mm->num_exe_file_vmas = 0;
1504 }
1505 
1506 struct file *get_mm_exe_file(struct mm_struct *mm)
1507 {
1508 	struct file *exe_file;
1509 
1510 	/* We need mmap_sem to protect against races with removal of
1511 	 * VM_EXECUTABLE vmas */
1512 	down_read(&mm->mmap_sem);
1513 	exe_file = mm->exe_file;
1514 	if (exe_file)
1515 		get_file(exe_file);
1516 	up_read(&mm->mmap_sem);
1517 	return exe_file;
1518 }
1519 
1520 void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1521 {
1522 	/* It's safe to write the exe_file pointer without exe_file_lock because
1523 	 * this is called during fork when the task is not yet in /proc */
1524 	newmm->exe_file = get_mm_exe_file(oldmm);
1525 }
1526 
1527 static int proc_exe_link(struct inode *inode, struct path *exe_path)
1528 {
1529 	struct task_struct *task;
1530 	struct mm_struct *mm;
1531 	struct file *exe_file;
1532 
1533 	task = get_proc_task(inode);
1534 	if (!task)
1535 		return -ENOENT;
1536 	mm = get_task_mm(task);
1537 	put_task_struct(task);
1538 	if (!mm)
1539 		return -ENOENT;
1540 	exe_file = get_mm_exe_file(mm);
1541 	mmput(mm);
1542 	if (exe_file) {
1543 		*exe_path = exe_file->f_path;
1544 		path_get(&exe_file->f_path);
1545 		fput(exe_file);
1546 		return 0;
1547 	} else
1548 		return -ENOENT;
1549 }
1550 
1551 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1552 {
1553 	struct inode *inode = dentry->d_inode;
1554 	int error = -EACCES;
1555 
1556 	/* We don't need a base pointer in the /proc filesystem */
1557 	path_put(&nd->path);
1558 
1559 	/* Are we allowed to snoop on the tasks file descriptors? */
1560 	if (!proc_fd_access_allowed(inode))
1561 		goto out;
1562 
1563 	error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
1564 out:
1565 	return ERR_PTR(error);
1566 }
1567 
1568 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1569 {
1570 	char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1571 	char *pathname;
1572 	int len;
1573 
1574 	if (!tmp)
1575 		return -ENOMEM;
1576 
1577 	pathname = d_path_with_unreachable(path, tmp, PAGE_SIZE);
1578 	len = PTR_ERR(pathname);
1579 	if (IS_ERR(pathname))
1580 		goto out;
1581 	len = tmp + PAGE_SIZE - 1 - pathname;
1582 
1583 	if (len > buflen)
1584 		len = buflen;
1585 	if (copy_to_user(buffer, pathname, len))
1586 		len = -EFAULT;
1587  out:
1588 	free_page((unsigned long)tmp);
1589 	return len;
1590 }
1591 
1592 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1593 {
1594 	int error = -EACCES;
1595 	struct inode *inode = dentry->d_inode;
1596 	struct path path;
1597 
1598 	/* Are we allowed to snoop on the tasks file descriptors? */
1599 	if (!proc_fd_access_allowed(inode))
1600 		goto out;
1601 
1602 	error = PROC_I(inode)->op.proc_get_link(inode, &path);
1603 	if (error)
1604 		goto out;
1605 
1606 	error = do_proc_readlink(&path, buffer, buflen);
1607 	path_put(&path);
1608 out:
1609 	return error;
1610 }
1611 
1612 static const struct inode_operations proc_pid_link_inode_operations = {
1613 	.readlink	= proc_pid_readlink,
1614 	.follow_link	= proc_pid_follow_link,
1615 	.setattr	= proc_setattr,
1616 };
1617 
1618 
1619 /* building an inode */
1620 
1621 static int task_dumpable(struct task_struct *task)
1622 {
1623 	int dumpable = 0;
1624 	struct mm_struct *mm;
1625 
1626 	task_lock(task);
1627 	mm = task->mm;
1628 	if (mm)
1629 		dumpable = get_dumpable(mm);
1630 	task_unlock(task);
1631 	if(dumpable == 1)
1632 		return 1;
1633 	return 0;
1634 }
1635 
1636 
1637 static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1638 {
1639 	struct inode * inode;
1640 	struct proc_inode *ei;
1641 	const struct cred *cred;
1642 
1643 	/* We need a new inode */
1644 
1645 	inode = new_inode(sb);
1646 	if (!inode)
1647 		goto out;
1648 
1649 	/* Common stuff */
1650 	ei = PROC_I(inode);
1651 	inode->i_ino = get_next_ino();
1652 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1653 	inode->i_op = &proc_def_inode_operations;
1654 
1655 	/*
1656 	 * grab the reference to task.
1657 	 */
1658 	ei->pid = get_task_pid(task, PIDTYPE_PID);
1659 	if (!ei->pid)
1660 		goto out_unlock;
1661 
1662 	if (task_dumpable(task)) {
1663 		rcu_read_lock();
1664 		cred = __task_cred(task);
1665 		inode->i_uid = cred->euid;
1666 		inode->i_gid = cred->egid;
1667 		rcu_read_unlock();
1668 	}
1669 	security_task_to_inode(task, inode);
1670 
1671 out:
1672 	return inode;
1673 
1674 out_unlock:
1675 	iput(inode);
1676 	return NULL;
1677 }
1678 
1679 static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1680 {
1681 	struct inode *inode = dentry->d_inode;
1682 	struct task_struct *task;
1683 	const struct cred *cred;
1684 
1685 	generic_fillattr(inode, stat);
1686 
1687 	rcu_read_lock();
1688 	stat->uid = 0;
1689 	stat->gid = 0;
1690 	task = pid_task(proc_pid(inode), PIDTYPE_PID);
1691 	if (task) {
1692 		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1693 		    task_dumpable(task)) {
1694 			cred = __task_cred(task);
1695 			stat->uid = cred->euid;
1696 			stat->gid = cred->egid;
1697 		}
1698 	}
1699 	rcu_read_unlock();
1700 	return 0;
1701 }
1702 
1703 /* dentry stuff */
1704 
1705 /*
1706  *	Exceptional case: normally we are not allowed to unhash a busy
1707  * directory. In this case, however, we can do it - no aliasing problems
1708  * due to the way we treat inodes.
1709  *
1710  * Rewrite the inode's ownerships here because the owning task may have
1711  * performed a setuid(), etc.
1712  *
1713  * Before the /proc/pid/status file was created the only way to read
1714  * the effective uid of a /process was to stat /proc/pid.  Reading
1715  * /proc/pid/status is slow enough that procps and other packages
1716  * kept stating /proc/pid.  To keep the rules in /proc simple I have
1717  * made this apply to all per process world readable and executable
1718  * directories.
1719  */
1720 static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1721 {
1722 	struct inode *inode = dentry->d_inode;
1723 	struct task_struct *task = get_proc_task(inode);
1724 	const struct cred *cred;
1725 
1726 	if (task) {
1727 		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1728 		    task_dumpable(task)) {
1729 			rcu_read_lock();
1730 			cred = __task_cred(task);
1731 			inode->i_uid = cred->euid;
1732 			inode->i_gid = cred->egid;
1733 			rcu_read_unlock();
1734 		} else {
1735 			inode->i_uid = 0;
1736 			inode->i_gid = 0;
1737 		}
1738 		inode->i_mode &= ~(S_ISUID | S_ISGID);
1739 		security_task_to_inode(task, inode);
1740 		put_task_struct(task);
1741 		return 1;
1742 	}
1743 	d_drop(dentry);
1744 	return 0;
1745 }
1746 
1747 static int pid_delete_dentry(struct dentry * dentry)
1748 {
1749 	/* Is the task we represent dead?
1750 	 * If so, then don't put the dentry on the lru list,
1751 	 * kill it immediately.
1752 	 */
1753 	return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1754 }
1755 
1756 static const struct dentry_operations pid_dentry_operations =
1757 {
1758 	.d_revalidate	= pid_revalidate,
1759 	.d_delete	= pid_delete_dentry,
1760 };
1761 
1762 /* Lookups */
1763 
1764 typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1765 				struct task_struct *, const void *);
1766 
1767 /*
1768  * Fill a directory entry.
1769  *
1770  * If possible create the dcache entry and derive our inode number and
1771  * file type from dcache entry.
1772  *
1773  * Since all of the proc inode numbers are dynamically generated, the inode
1774  * numbers do not exist until the inode is cache.  This means creating the
1775  * the dcache entry in readdir is necessary to keep the inode numbers
1776  * reported by readdir in sync with the inode numbers reported
1777  * by stat.
1778  */
1779 static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1780 	char *name, int len,
1781 	instantiate_t instantiate, struct task_struct *task, const void *ptr)
1782 {
1783 	struct dentry *child, *dir = filp->f_path.dentry;
1784 	struct inode *inode;
1785 	struct qstr qname;
1786 	ino_t ino = 0;
1787 	unsigned type = DT_UNKNOWN;
1788 
1789 	qname.name = name;
1790 	qname.len  = len;
1791 	qname.hash = full_name_hash(name, len);
1792 
1793 	child = d_lookup(dir, &qname);
1794 	if (!child) {
1795 		struct dentry *new;
1796 		new = d_alloc(dir, &qname);
1797 		if (new) {
1798 			child = instantiate(dir->d_inode, new, task, ptr);
1799 			if (child)
1800 				dput(new);
1801 			else
1802 				child = new;
1803 		}
1804 	}
1805 	if (!child || IS_ERR(child) || !child->d_inode)
1806 		goto end_instantiate;
1807 	inode = child->d_inode;
1808 	if (inode) {
1809 		ino = inode->i_ino;
1810 		type = inode->i_mode >> 12;
1811 	}
1812 	dput(child);
1813 end_instantiate:
1814 	if (!ino)
1815 		ino = find_inode_number(dir, &qname);
1816 	if (!ino)
1817 		ino = 1;
1818 	return filldir(dirent, name, len, filp->f_pos, ino, type);
1819 }
1820 
1821 static unsigned name_to_int(struct dentry *dentry)
1822 {
1823 	const char *name = dentry->d_name.name;
1824 	int len = dentry->d_name.len;
1825 	unsigned n = 0;
1826 
1827 	if (len > 1 && *name == '0')
1828 		goto out;
1829 	while (len-- > 0) {
1830 		unsigned c = *name++ - '0';
1831 		if (c > 9)
1832 			goto out;
1833 		if (n >= (~0U-9)/10)
1834 			goto out;
1835 		n *= 10;
1836 		n += c;
1837 	}
1838 	return n;
1839 out:
1840 	return ~0U;
1841 }
1842 
1843 #define PROC_FDINFO_MAX 64
1844 
1845 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1846 {
1847 	struct task_struct *task = get_proc_task(inode);
1848 	struct files_struct *files = NULL;
1849 	struct file *file;
1850 	int fd = proc_fd(inode);
1851 
1852 	if (task) {
1853 		files = get_files_struct(task);
1854 		put_task_struct(task);
1855 	}
1856 	if (files) {
1857 		/*
1858 		 * We are not taking a ref to the file structure, so we must
1859 		 * hold ->file_lock.
1860 		 */
1861 		spin_lock(&files->file_lock);
1862 		file = fcheck_files(files, fd);
1863 		if (file) {
1864 			if (path) {
1865 				*path = file->f_path;
1866 				path_get(&file->f_path);
1867 			}
1868 			if (info)
1869 				snprintf(info, PROC_FDINFO_MAX,
1870 					 "pos:\t%lli\n"
1871 					 "flags:\t0%o\n",
1872 					 (long long) file->f_pos,
1873 					 file->f_flags);
1874 			spin_unlock(&files->file_lock);
1875 			put_files_struct(files);
1876 			return 0;
1877 		}
1878 		spin_unlock(&files->file_lock);
1879 		put_files_struct(files);
1880 	}
1881 	return -ENOENT;
1882 }
1883 
1884 static int proc_fd_link(struct inode *inode, struct path *path)
1885 {
1886 	return proc_fd_info(inode, path, NULL);
1887 }
1888 
1889 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1890 {
1891 	struct inode *inode = dentry->d_inode;
1892 	struct task_struct *task = get_proc_task(inode);
1893 	int fd = proc_fd(inode);
1894 	struct files_struct *files;
1895 	const struct cred *cred;
1896 
1897 	if (task) {
1898 		files = get_files_struct(task);
1899 		if (files) {
1900 			rcu_read_lock();
1901 			if (fcheck_files(files, fd)) {
1902 				rcu_read_unlock();
1903 				put_files_struct(files);
1904 				if (task_dumpable(task)) {
1905 					rcu_read_lock();
1906 					cred = __task_cred(task);
1907 					inode->i_uid = cred->euid;
1908 					inode->i_gid = cred->egid;
1909 					rcu_read_unlock();
1910 				} else {
1911 					inode->i_uid = 0;
1912 					inode->i_gid = 0;
1913 				}
1914 				inode->i_mode &= ~(S_ISUID | S_ISGID);
1915 				security_task_to_inode(task, inode);
1916 				put_task_struct(task);
1917 				return 1;
1918 			}
1919 			rcu_read_unlock();
1920 			put_files_struct(files);
1921 		}
1922 		put_task_struct(task);
1923 	}
1924 	d_drop(dentry);
1925 	return 0;
1926 }
1927 
1928 static const struct dentry_operations tid_fd_dentry_operations =
1929 {
1930 	.d_revalidate	= tid_fd_revalidate,
1931 	.d_delete	= pid_delete_dentry,
1932 };
1933 
1934 static struct dentry *proc_fd_instantiate(struct inode *dir,
1935 	struct dentry *dentry, struct task_struct *task, const void *ptr)
1936 {
1937 	unsigned fd = *(const unsigned *)ptr;
1938 	struct file *file;
1939 	struct files_struct *files;
1940  	struct inode *inode;
1941  	struct proc_inode *ei;
1942 	struct dentry *error = ERR_PTR(-ENOENT);
1943 
1944 	inode = proc_pid_make_inode(dir->i_sb, task);
1945 	if (!inode)
1946 		goto out;
1947 	ei = PROC_I(inode);
1948 	ei->fd = fd;
1949 	files = get_files_struct(task);
1950 	if (!files)
1951 		goto out_iput;
1952 	inode->i_mode = S_IFLNK;
1953 
1954 	/*
1955 	 * We are not taking a ref to the file structure, so we must
1956 	 * hold ->file_lock.
1957 	 */
1958 	spin_lock(&files->file_lock);
1959 	file = fcheck_files(files, fd);
1960 	if (!file)
1961 		goto out_unlock;
1962 	if (file->f_mode & FMODE_READ)
1963 		inode->i_mode |= S_IRUSR | S_IXUSR;
1964 	if (file->f_mode & FMODE_WRITE)
1965 		inode->i_mode |= S_IWUSR | S_IXUSR;
1966 	spin_unlock(&files->file_lock);
1967 	put_files_struct(files);
1968 
1969 	inode->i_op = &proc_pid_link_inode_operations;
1970 	inode->i_size = 64;
1971 	ei->op.proc_get_link = proc_fd_link;
1972 	dentry->d_op = &tid_fd_dentry_operations;
1973 	d_add(dentry, inode);
1974 	/* Close the race of the process dying before we return the dentry */
1975 	if (tid_fd_revalidate(dentry, NULL))
1976 		error = NULL;
1977 
1978  out:
1979 	return error;
1980 out_unlock:
1981 	spin_unlock(&files->file_lock);
1982 	put_files_struct(files);
1983 out_iput:
1984 	iput(inode);
1985 	goto out;
1986 }
1987 
1988 static struct dentry *proc_lookupfd_common(struct inode *dir,
1989 					   struct dentry *dentry,
1990 					   instantiate_t instantiate)
1991 {
1992 	struct task_struct *task = get_proc_task(dir);
1993 	unsigned fd = name_to_int(dentry);
1994 	struct dentry *result = ERR_PTR(-ENOENT);
1995 
1996 	if (!task)
1997 		goto out_no_task;
1998 	if (fd == ~0U)
1999 		goto out;
2000 
2001 	result = instantiate(dir, dentry, task, &fd);
2002 out:
2003 	put_task_struct(task);
2004 out_no_task:
2005 	return result;
2006 }
2007 
2008 static int proc_readfd_common(struct file * filp, void * dirent,
2009 			      filldir_t filldir, instantiate_t instantiate)
2010 {
2011 	struct dentry *dentry = filp->f_path.dentry;
2012 	struct inode *inode = dentry->d_inode;
2013 	struct task_struct *p = get_proc_task(inode);
2014 	unsigned int fd, ino;
2015 	int retval;
2016 	struct files_struct * files;
2017 
2018 	retval = -ENOENT;
2019 	if (!p)
2020 		goto out_no_task;
2021 	retval = 0;
2022 
2023 	fd = filp->f_pos;
2024 	switch (fd) {
2025 		case 0:
2026 			if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2027 				goto out;
2028 			filp->f_pos++;
2029 		case 1:
2030 			ino = parent_ino(dentry);
2031 			if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2032 				goto out;
2033 			filp->f_pos++;
2034 		default:
2035 			files = get_files_struct(p);
2036 			if (!files)
2037 				goto out;
2038 			rcu_read_lock();
2039 			for (fd = filp->f_pos-2;
2040 			     fd < files_fdtable(files)->max_fds;
2041 			     fd++, filp->f_pos++) {
2042 				char name[PROC_NUMBUF];
2043 				int len;
2044 
2045 				if (!fcheck_files(files, fd))
2046 					continue;
2047 				rcu_read_unlock();
2048 
2049 				len = snprintf(name, sizeof(name), "%d", fd);
2050 				if (proc_fill_cache(filp, dirent, filldir,
2051 						    name, len, instantiate,
2052 						    p, &fd) < 0) {
2053 					rcu_read_lock();
2054 					break;
2055 				}
2056 				rcu_read_lock();
2057 			}
2058 			rcu_read_unlock();
2059 			put_files_struct(files);
2060 	}
2061 out:
2062 	put_task_struct(p);
2063 out_no_task:
2064 	return retval;
2065 }
2066 
2067 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2068 				    struct nameidata *nd)
2069 {
2070 	return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2071 }
2072 
2073 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2074 {
2075 	return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2076 }
2077 
2078 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2079 				      size_t len, loff_t *ppos)
2080 {
2081 	char tmp[PROC_FDINFO_MAX];
2082 	int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
2083 	if (!err)
2084 		err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2085 	return err;
2086 }
2087 
2088 static const struct file_operations proc_fdinfo_file_operations = {
2089 	.open           = nonseekable_open,
2090 	.read		= proc_fdinfo_read,
2091 	.llseek		= no_llseek,
2092 };
2093 
2094 static const struct file_operations proc_fd_operations = {
2095 	.read		= generic_read_dir,
2096 	.readdir	= proc_readfd,
2097 	.llseek		= default_llseek,
2098 };
2099 
2100 /*
2101  * /proc/pid/fd needs a special permission handler so that a process can still
2102  * access /proc/self/fd after it has executed a setuid().
2103  */
2104 static int proc_fd_permission(struct inode *inode, int mask)
2105 {
2106 	int rv;
2107 
2108 	rv = generic_permission(inode, mask, NULL);
2109 	if (rv == 0)
2110 		return 0;
2111 	if (task_pid(current) == proc_pid(inode))
2112 		rv = 0;
2113 	return rv;
2114 }
2115 
2116 /*
2117  * proc directories can do almost nothing..
2118  */
2119 static const struct inode_operations proc_fd_inode_operations = {
2120 	.lookup		= proc_lookupfd,
2121 	.permission	= proc_fd_permission,
2122 	.setattr	= proc_setattr,
2123 };
2124 
2125 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2126 	struct dentry *dentry, struct task_struct *task, const void *ptr)
2127 {
2128 	unsigned fd = *(unsigned *)ptr;
2129  	struct inode *inode;
2130  	struct proc_inode *ei;
2131 	struct dentry *error = ERR_PTR(-ENOENT);
2132 
2133 	inode = proc_pid_make_inode(dir->i_sb, task);
2134 	if (!inode)
2135 		goto out;
2136 	ei = PROC_I(inode);
2137 	ei->fd = fd;
2138 	inode->i_mode = S_IFREG | S_IRUSR;
2139 	inode->i_fop = &proc_fdinfo_file_operations;
2140 	dentry->d_op = &tid_fd_dentry_operations;
2141 	d_add(dentry, inode);
2142 	/* Close the race of the process dying before we return the dentry */
2143 	if (tid_fd_revalidate(dentry, NULL))
2144 		error = NULL;
2145 
2146  out:
2147 	return error;
2148 }
2149 
2150 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2151 					struct dentry *dentry,
2152 					struct nameidata *nd)
2153 {
2154 	return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2155 }
2156 
2157 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2158 {
2159 	return proc_readfd_common(filp, dirent, filldir,
2160 				  proc_fdinfo_instantiate);
2161 }
2162 
2163 static const struct file_operations proc_fdinfo_operations = {
2164 	.read		= generic_read_dir,
2165 	.readdir	= proc_readfdinfo,
2166 	.llseek		= default_llseek,
2167 };
2168 
2169 /*
2170  * proc directories can do almost nothing..
2171  */
2172 static const struct inode_operations proc_fdinfo_inode_operations = {
2173 	.lookup		= proc_lookupfdinfo,
2174 	.setattr	= proc_setattr,
2175 };
2176 
2177 
2178 static struct dentry *proc_pident_instantiate(struct inode *dir,
2179 	struct dentry *dentry, struct task_struct *task, const void *ptr)
2180 {
2181 	const struct pid_entry *p = ptr;
2182 	struct inode *inode;
2183 	struct proc_inode *ei;
2184 	struct dentry *error = ERR_PTR(-ENOENT);
2185 
2186 	inode = proc_pid_make_inode(dir->i_sb, task);
2187 	if (!inode)
2188 		goto out;
2189 
2190 	ei = PROC_I(inode);
2191 	inode->i_mode = p->mode;
2192 	if (S_ISDIR(inode->i_mode))
2193 		inode->i_nlink = 2;	/* Use getattr to fix if necessary */
2194 	if (p->iop)
2195 		inode->i_op = p->iop;
2196 	if (p->fop)
2197 		inode->i_fop = p->fop;
2198 	ei->op = p->op;
2199 	dentry->d_op = &pid_dentry_operations;
2200 	d_add(dentry, inode);
2201 	/* Close the race of the process dying before we return the dentry */
2202 	if (pid_revalidate(dentry, NULL))
2203 		error = NULL;
2204 out:
2205 	return error;
2206 }
2207 
2208 static struct dentry *proc_pident_lookup(struct inode *dir,
2209 					 struct dentry *dentry,
2210 					 const struct pid_entry *ents,
2211 					 unsigned int nents)
2212 {
2213 	struct dentry *error;
2214 	struct task_struct *task = get_proc_task(dir);
2215 	const struct pid_entry *p, *last;
2216 
2217 	error = ERR_PTR(-ENOENT);
2218 
2219 	if (!task)
2220 		goto out_no_task;
2221 
2222 	/*
2223 	 * Yes, it does not scale. And it should not. Don't add
2224 	 * new entries into /proc/<tgid>/ without very good reasons.
2225 	 */
2226 	last = &ents[nents - 1];
2227 	for (p = ents; p <= last; p++) {
2228 		if (p->len != dentry->d_name.len)
2229 			continue;
2230 		if (!memcmp(dentry->d_name.name, p->name, p->len))
2231 			break;
2232 	}
2233 	if (p > last)
2234 		goto out;
2235 
2236 	error = proc_pident_instantiate(dir, dentry, task, p);
2237 out:
2238 	put_task_struct(task);
2239 out_no_task:
2240 	return error;
2241 }
2242 
2243 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2244 	filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2245 {
2246 	return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2247 				proc_pident_instantiate, task, p);
2248 }
2249 
2250 static int proc_pident_readdir(struct file *filp,
2251 		void *dirent, filldir_t filldir,
2252 		const struct pid_entry *ents, unsigned int nents)
2253 {
2254 	int i;
2255 	struct dentry *dentry = filp->f_path.dentry;
2256 	struct inode *inode = dentry->d_inode;
2257 	struct task_struct *task = get_proc_task(inode);
2258 	const struct pid_entry *p, *last;
2259 	ino_t ino;
2260 	int ret;
2261 
2262 	ret = -ENOENT;
2263 	if (!task)
2264 		goto out_no_task;
2265 
2266 	ret = 0;
2267 	i = filp->f_pos;
2268 	switch (i) {
2269 	case 0:
2270 		ino = inode->i_ino;
2271 		if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2272 			goto out;
2273 		i++;
2274 		filp->f_pos++;
2275 		/* fall through */
2276 	case 1:
2277 		ino = parent_ino(dentry);
2278 		if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2279 			goto out;
2280 		i++;
2281 		filp->f_pos++;
2282 		/* fall through */
2283 	default:
2284 		i -= 2;
2285 		if (i >= nents) {
2286 			ret = 1;
2287 			goto out;
2288 		}
2289 		p = ents + i;
2290 		last = &ents[nents - 1];
2291 		while (p <= last) {
2292 			if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2293 				goto out;
2294 			filp->f_pos++;
2295 			p++;
2296 		}
2297 	}
2298 
2299 	ret = 1;
2300 out:
2301 	put_task_struct(task);
2302 out_no_task:
2303 	return ret;
2304 }
2305 
2306 #ifdef CONFIG_SECURITY
2307 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2308 				  size_t count, loff_t *ppos)
2309 {
2310 	struct inode * inode = file->f_path.dentry->d_inode;
2311 	char *p = NULL;
2312 	ssize_t length;
2313 	struct task_struct *task = get_proc_task(inode);
2314 
2315 	if (!task)
2316 		return -ESRCH;
2317 
2318 	length = security_getprocattr(task,
2319 				      (char*)file->f_path.dentry->d_name.name,
2320 				      &p);
2321 	put_task_struct(task);
2322 	if (length > 0)
2323 		length = simple_read_from_buffer(buf, count, ppos, p, length);
2324 	kfree(p);
2325 	return length;
2326 }
2327 
2328 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2329 				   size_t count, loff_t *ppos)
2330 {
2331 	struct inode * inode = file->f_path.dentry->d_inode;
2332 	char *page;
2333 	ssize_t length;
2334 	struct task_struct *task = get_proc_task(inode);
2335 
2336 	length = -ESRCH;
2337 	if (!task)
2338 		goto out_no_task;
2339 	if (count > PAGE_SIZE)
2340 		count = PAGE_SIZE;
2341 
2342 	/* No partial writes. */
2343 	length = -EINVAL;
2344 	if (*ppos != 0)
2345 		goto out;
2346 
2347 	length = -ENOMEM;
2348 	page = (char*)__get_free_page(GFP_TEMPORARY);
2349 	if (!page)
2350 		goto out;
2351 
2352 	length = -EFAULT;
2353 	if (copy_from_user(page, buf, count))
2354 		goto out_free;
2355 
2356 	/* Guard against adverse ptrace interaction */
2357 	length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2358 	if (length < 0)
2359 		goto out_free;
2360 
2361 	length = security_setprocattr(task,
2362 				      (char*)file->f_path.dentry->d_name.name,
2363 				      (void*)page, count);
2364 	mutex_unlock(&task->signal->cred_guard_mutex);
2365 out_free:
2366 	free_page((unsigned long) page);
2367 out:
2368 	put_task_struct(task);
2369 out_no_task:
2370 	return length;
2371 }
2372 
2373 static const struct file_operations proc_pid_attr_operations = {
2374 	.read		= proc_pid_attr_read,
2375 	.write		= proc_pid_attr_write,
2376 	.llseek		= generic_file_llseek,
2377 };
2378 
2379 static const struct pid_entry attr_dir_stuff[] = {
2380 	REG("current",    S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2381 	REG("prev",       S_IRUGO,	   proc_pid_attr_operations),
2382 	REG("exec",       S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2383 	REG("fscreate",   S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2384 	REG("keycreate",  S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2385 	REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2386 };
2387 
2388 static int proc_attr_dir_readdir(struct file * filp,
2389 			     void * dirent, filldir_t filldir)
2390 {
2391 	return proc_pident_readdir(filp,dirent,filldir,
2392 				   attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2393 }
2394 
2395 static const struct file_operations proc_attr_dir_operations = {
2396 	.read		= generic_read_dir,
2397 	.readdir	= proc_attr_dir_readdir,
2398 	.llseek		= default_llseek,
2399 };
2400 
2401 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2402 				struct dentry *dentry, struct nameidata *nd)
2403 {
2404 	return proc_pident_lookup(dir, dentry,
2405 				  attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2406 }
2407 
2408 static const struct inode_operations proc_attr_dir_inode_operations = {
2409 	.lookup		= proc_attr_dir_lookup,
2410 	.getattr	= pid_getattr,
2411 	.setattr	= proc_setattr,
2412 };
2413 
2414 #endif
2415 
2416 #ifdef CONFIG_ELF_CORE
2417 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2418 					 size_t count, loff_t *ppos)
2419 {
2420 	struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2421 	struct mm_struct *mm;
2422 	char buffer[PROC_NUMBUF];
2423 	size_t len;
2424 	int ret;
2425 
2426 	if (!task)
2427 		return -ESRCH;
2428 
2429 	ret = 0;
2430 	mm = get_task_mm(task);
2431 	if (mm) {
2432 		len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2433 			       ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2434 				MMF_DUMP_FILTER_SHIFT));
2435 		mmput(mm);
2436 		ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2437 	}
2438 
2439 	put_task_struct(task);
2440 
2441 	return ret;
2442 }
2443 
2444 static ssize_t proc_coredump_filter_write(struct file *file,
2445 					  const char __user *buf,
2446 					  size_t count,
2447 					  loff_t *ppos)
2448 {
2449 	struct task_struct *task;
2450 	struct mm_struct *mm;
2451 	char buffer[PROC_NUMBUF], *end;
2452 	unsigned int val;
2453 	int ret;
2454 	int i;
2455 	unsigned long mask;
2456 
2457 	ret = -EFAULT;
2458 	memset(buffer, 0, sizeof(buffer));
2459 	if (count > sizeof(buffer) - 1)
2460 		count = sizeof(buffer) - 1;
2461 	if (copy_from_user(buffer, buf, count))
2462 		goto out_no_task;
2463 
2464 	ret = -EINVAL;
2465 	val = (unsigned int)simple_strtoul(buffer, &end, 0);
2466 	if (*end == '\n')
2467 		end++;
2468 	if (end - buffer == 0)
2469 		goto out_no_task;
2470 
2471 	ret = -ESRCH;
2472 	task = get_proc_task(file->f_dentry->d_inode);
2473 	if (!task)
2474 		goto out_no_task;
2475 
2476 	ret = end - buffer;
2477 	mm = get_task_mm(task);
2478 	if (!mm)
2479 		goto out_no_mm;
2480 
2481 	for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2482 		if (val & mask)
2483 			set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2484 		else
2485 			clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2486 	}
2487 
2488 	mmput(mm);
2489  out_no_mm:
2490 	put_task_struct(task);
2491  out_no_task:
2492 	return ret;
2493 }
2494 
2495 static const struct file_operations proc_coredump_filter_operations = {
2496 	.read		= proc_coredump_filter_read,
2497 	.write		= proc_coredump_filter_write,
2498 	.llseek		= generic_file_llseek,
2499 };
2500 #endif
2501 
2502 /*
2503  * /proc/self:
2504  */
2505 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2506 			      int buflen)
2507 {
2508 	struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2509 	pid_t tgid = task_tgid_nr_ns(current, ns);
2510 	char tmp[PROC_NUMBUF];
2511 	if (!tgid)
2512 		return -ENOENT;
2513 	sprintf(tmp, "%d", tgid);
2514 	return vfs_readlink(dentry,buffer,buflen,tmp);
2515 }
2516 
2517 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2518 {
2519 	struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2520 	pid_t tgid = task_tgid_nr_ns(current, ns);
2521 	char *name = ERR_PTR(-ENOENT);
2522 	if (tgid) {
2523 		name = __getname();
2524 		if (!name)
2525 			name = ERR_PTR(-ENOMEM);
2526 		else
2527 			sprintf(name, "%d", tgid);
2528 	}
2529 	nd_set_link(nd, name);
2530 	return NULL;
2531 }
2532 
2533 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2534 				void *cookie)
2535 {
2536 	char *s = nd_get_link(nd);
2537 	if (!IS_ERR(s))
2538 		__putname(s);
2539 }
2540 
2541 static const struct inode_operations proc_self_inode_operations = {
2542 	.readlink	= proc_self_readlink,
2543 	.follow_link	= proc_self_follow_link,
2544 	.put_link	= proc_self_put_link,
2545 };
2546 
2547 /*
2548  * proc base
2549  *
2550  * These are the directory entries in the root directory of /proc
2551  * that properly belong to the /proc filesystem, as they describe
2552  * describe something that is process related.
2553  */
2554 static const struct pid_entry proc_base_stuff[] = {
2555 	NOD("self", S_IFLNK|S_IRWXUGO,
2556 		&proc_self_inode_operations, NULL, {}),
2557 };
2558 
2559 /*
2560  *	Exceptional case: normally we are not allowed to unhash a busy
2561  * directory. In this case, however, we can do it - no aliasing problems
2562  * due to the way we treat inodes.
2563  */
2564 static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2565 {
2566 	struct inode *inode = dentry->d_inode;
2567 	struct task_struct *task = get_proc_task(inode);
2568 	if (task) {
2569 		put_task_struct(task);
2570 		return 1;
2571 	}
2572 	d_drop(dentry);
2573 	return 0;
2574 }
2575 
2576 static const struct dentry_operations proc_base_dentry_operations =
2577 {
2578 	.d_revalidate	= proc_base_revalidate,
2579 	.d_delete	= pid_delete_dentry,
2580 };
2581 
2582 static struct dentry *proc_base_instantiate(struct inode *dir,
2583 	struct dentry *dentry, struct task_struct *task, const void *ptr)
2584 {
2585 	const struct pid_entry *p = ptr;
2586 	struct inode *inode;
2587 	struct proc_inode *ei;
2588 	struct dentry *error;
2589 
2590 	/* Allocate the inode */
2591 	error = ERR_PTR(-ENOMEM);
2592 	inode = new_inode(dir->i_sb);
2593 	if (!inode)
2594 		goto out;
2595 
2596 	/* Initialize the inode */
2597 	ei = PROC_I(inode);
2598 	inode->i_ino = get_next_ino();
2599 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2600 
2601 	/*
2602 	 * grab the reference to the task.
2603 	 */
2604 	ei->pid = get_task_pid(task, PIDTYPE_PID);
2605 	if (!ei->pid)
2606 		goto out_iput;
2607 
2608 	inode->i_mode = p->mode;
2609 	if (S_ISDIR(inode->i_mode))
2610 		inode->i_nlink = 2;
2611 	if (S_ISLNK(inode->i_mode))
2612 		inode->i_size = 64;
2613 	if (p->iop)
2614 		inode->i_op = p->iop;
2615 	if (p->fop)
2616 		inode->i_fop = p->fop;
2617 	ei->op = p->op;
2618 	dentry->d_op = &proc_base_dentry_operations;
2619 	d_add(dentry, inode);
2620 	error = NULL;
2621 out:
2622 	return error;
2623 out_iput:
2624 	iput(inode);
2625 	goto out;
2626 }
2627 
2628 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2629 {
2630 	struct dentry *error;
2631 	struct task_struct *task = get_proc_task(dir);
2632 	const struct pid_entry *p, *last;
2633 
2634 	error = ERR_PTR(-ENOENT);
2635 
2636 	if (!task)
2637 		goto out_no_task;
2638 
2639 	/* Lookup the directory entry */
2640 	last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2641 	for (p = proc_base_stuff; p <= last; p++) {
2642 		if (p->len != dentry->d_name.len)
2643 			continue;
2644 		if (!memcmp(dentry->d_name.name, p->name, p->len))
2645 			break;
2646 	}
2647 	if (p > last)
2648 		goto out;
2649 
2650 	error = proc_base_instantiate(dir, dentry, task, p);
2651 
2652 out:
2653 	put_task_struct(task);
2654 out_no_task:
2655 	return error;
2656 }
2657 
2658 static int proc_base_fill_cache(struct file *filp, void *dirent,
2659 	filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2660 {
2661 	return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2662 				proc_base_instantiate, task, p);
2663 }
2664 
2665 #ifdef CONFIG_TASK_IO_ACCOUNTING
2666 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2667 {
2668 	struct task_io_accounting acct = task->ioac;
2669 	unsigned long flags;
2670 
2671 	if (whole && lock_task_sighand(task, &flags)) {
2672 		struct task_struct *t = task;
2673 
2674 		task_io_accounting_add(&acct, &task->signal->ioac);
2675 		while_each_thread(task, t)
2676 			task_io_accounting_add(&acct, &t->ioac);
2677 
2678 		unlock_task_sighand(task, &flags);
2679 	}
2680 	return sprintf(buffer,
2681 			"rchar: %llu\n"
2682 			"wchar: %llu\n"
2683 			"syscr: %llu\n"
2684 			"syscw: %llu\n"
2685 			"read_bytes: %llu\n"
2686 			"write_bytes: %llu\n"
2687 			"cancelled_write_bytes: %llu\n",
2688 			(unsigned long long)acct.rchar,
2689 			(unsigned long long)acct.wchar,
2690 			(unsigned long long)acct.syscr,
2691 			(unsigned long long)acct.syscw,
2692 			(unsigned long long)acct.read_bytes,
2693 			(unsigned long long)acct.write_bytes,
2694 			(unsigned long long)acct.cancelled_write_bytes);
2695 }
2696 
2697 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2698 {
2699 	return do_io_accounting(task, buffer, 0);
2700 }
2701 
2702 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2703 {
2704 	return do_io_accounting(task, buffer, 1);
2705 }
2706 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2707 
2708 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2709 				struct pid *pid, struct task_struct *task)
2710 {
2711 	seq_printf(m, "%08x\n", task->personality);
2712 	return 0;
2713 }
2714 
2715 /*
2716  * Thread groups
2717  */
2718 static const struct file_operations proc_task_operations;
2719 static const struct inode_operations proc_task_inode_operations;
2720 
2721 static const struct pid_entry tgid_base_stuff[] = {
2722 	DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2723 	DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2724 	DIR("fdinfo",     S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2725 #ifdef CONFIG_NET
2726 	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2727 #endif
2728 	REG("environ",    S_IRUSR, proc_environ_operations),
2729 	INF("auxv",       S_IRUSR, proc_pid_auxv),
2730 	ONE("status",     S_IRUGO, proc_pid_status),
2731 	ONE("personality", S_IRUSR, proc_pid_personality),
2732 	INF("limits",	  S_IRUGO, proc_pid_limits),
2733 #ifdef CONFIG_SCHED_DEBUG
2734 	REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2735 #endif
2736 	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2737 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2738 	INF("syscall",    S_IRUSR, proc_pid_syscall),
2739 #endif
2740 	INF("cmdline",    S_IRUGO, proc_pid_cmdline),
2741 	ONE("stat",       S_IRUGO, proc_tgid_stat),
2742 	ONE("statm",      S_IRUGO, proc_pid_statm),
2743 	REG("maps",       S_IRUGO, proc_maps_operations),
2744 #ifdef CONFIG_NUMA
2745 	REG("numa_maps",  S_IRUGO, proc_numa_maps_operations),
2746 #endif
2747 	REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
2748 	LNK("cwd",        proc_cwd_link),
2749 	LNK("root",       proc_root_link),
2750 	LNK("exe",        proc_exe_link),
2751 	REG("mounts",     S_IRUGO, proc_mounts_operations),
2752 	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
2753 	REG("mountstats", S_IRUSR, proc_mountstats_operations),
2754 #ifdef CONFIG_PROC_PAGE_MONITOR
2755 	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2756 	REG("smaps",      S_IRUGO, proc_smaps_operations),
2757 	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
2758 #endif
2759 #ifdef CONFIG_SECURITY
2760 	DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2761 #endif
2762 #ifdef CONFIG_KALLSYMS
2763 	INF("wchan",      S_IRUGO, proc_pid_wchan),
2764 #endif
2765 #ifdef CONFIG_STACKTRACE
2766 	ONE("stack",      S_IRUSR, proc_pid_stack),
2767 #endif
2768 #ifdef CONFIG_SCHEDSTATS
2769 	INF("schedstat",  S_IRUGO, proc_pid_schedstat),
2770 #endif
2771 #ifdef CONFIG_LATENCYTOP
2772 	REG("latency",  S_IRUGO, proc_lstats_operations),
2773 #endif
2774 #ifdef CONFIG_PROC_PID_CPUSET
2775 	REG("cpuset",     S_IRUGO, proc_cpuset_operations),
2776 #endif
2777 #ifdef CONFIG_CGROUPS
2778 	REG("cgroup",  S_IRUGO, proc_cgroup_operations),
2779 #endif
2780 	INF("oom_score",  S_IRUGO, proc_oom_score),
2781 	REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
2782 	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2783 #ifdef CONFIG_AUDITSYSCALL
2784 	REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
2785 	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
2786 #endif
2787 #ifdef CONFIG_FAULT_INJECTION
2788 	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2789 #endif
2790 #ifdef CONFIG_ELF_CORE
2791 	REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2792 #endif
2793 #ifdef CONFIG_TASK_IO_ACCOUNTING
2794 	INF("io",	S_IRUGO, proc_tgid_io_accounting),
2795 #endif
2796 };
2797 
2798 static int proc_tgid_base_readdir(struct file * filp,
2799 			     void * dirent, filldir_t filldir)
2800 {
2801 	return proc_pident_readdir(filp,dirent,filldir,
2802 				   tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2803 }
2804 
2805 static const struct file_operations proc_tgid_base_operations = {
2806 	.read		= generic_read_dir,
2807 	.readdir	= proc_tgid_base_readdir,
2808 	.llseek		= default_llseek,
2809 };
2810 
2811 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
2812 	return proc_pident_lookup(dir, dentry,
2813 				  tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2814 }
2815 
2816 static const struct inode_operations proc_tgid_base_inode_operations = {
2817 	.lookup		= proc_tgid_base_lookup,
2818 	.getattr	= pid_getattr,
2819 	.setattr	= proc_setattr,
2820 };
2821 
2822 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2823 {
2824 	struct dentry *dentry, *leader, *dir;
2825 	char buf[PROC_NUMBUF];
2826 	struct qstr name;
2827 
2828 	name.name = buf;
2829 	name.len = snprintf(buf, sizeof(buf), "%d", pid);
2830 	dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2831 	if (dentry) {
2832 		shrink_dcache_parent(dentry);
2833 		d_drop(dentry);
2834 		dput(dentry);
2835 	}
2836 
2837 	name.name = buf;
2838 	name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2839 	leader = d_hash_and_lookup(mnt->mnt_root, &name);
2840 	if (!leader)
2841 		goto out;
2842 
2843 	name.name = "task";
2844 	name.len = strlen(name.name);
2845 	dir = d_hash_and_lookup(leader, &name);
2846 	if (!dir)
2847 		goto out_put_leader;
2848 
2849 	name.name = buf;
2850 	name.len = snprintf(buf, sizeof(buf), "%d", pid);
2851 	dentry = d_hash_and_lookup(dir, &name);
2852 	if (dentry) {
2853 		shrink_dcache_parent(dentry);
2854 		d_drop(dentry);
2855 		dput(dentry);
2856 	}
2857 
2858 	dput(dir);
2859 out_put_leader:
2860 	dput(leader);
2861 out:
2862 	return;
2863 }
2864 
2865 /**
2866  * proc_flush_task -  Remove dcache entries for @task from the /proc dcache.
2867  * @task: task that should be flushed.
2868  *
2869  * When flushing dentries from proc, one needs to flush them from global
2870  * proc (proc_mnt) and from all the namespaces' procs this task was seen
2871  * in. This call is supposed to do all of this job.
2872  *
2873  * Looks in the dcache for
2874  * /proc/@pid
2875  * /proc/@tgid/task/@pid
2876  * if either directory is present flushes it and all of it'ts children
2877  * from the dcache.
2878  *
2879  * It is safe and reasonable to cache /proc entries for a task until
2880  * that task exits.  After that they just clog up the dcache with
2881  * useless entries, possibly causing useful dcache entries to be
2882  * flushed instead.  This routine is proved to flush those useless
2883  * dcache entries at process exit time.
2884  *
2885  * NOTE: This routine is just an optimization so it does not guarantee
2886  *       that no dcache entries will exist at process exit time it
2887  *       just makes it very unlikely that any will persist.
2888  */
2889 
2890 void proc_flush_task(struct task_struct *task)
2891 {
2892 	int i;
2893 	struct pid *pid, *tgid;
2894 	struct upid *upid;
2895 
2896 	pid = task_pid(task);
2897 	tgid = task_tgid(task);
2898 
2899 	for (i = 0; i <= pid->level; i++) {
2900 		upid = &pid->numbers[i];
2901 		proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2902 					tgid->numbers[i].nr);
2903 	}
2904 
2905 	upid = &pid->numbers[pid->level];
2906 	if (upid->nr == 1)
2907 		pid_ns_release_proc(upid->ns);
2908 }
2909 
2910 static struct dentry *proc_pid_instantiate(struct inode *dir,
2911 					   struct dentry * dentry,
2912 					   struct task_struct *task, const void *ptr)
2913 {
2914 	struct dentry *error = ERR_PTR(-ENOENT);
2915 	struct inode *inode;
2916 
2917 	inode = proc_pid_make_inode(dir->i_sb, task);
2918 	if (!inode)
2919 		goto out;
2920 
2921 	inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2922 	inode->i_op = &proc_tgid_base_inode_operations;
2923 	inode->i_fop = &proc_tgid_base_operations;
2924 	inode->i_flags|=S_IMMUTABLE;
2925 
2926 	inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2927 		ARRAY_SIZE(tgid_base_stuff));
2928 
2929 	dentry->d_op = &pid_dentry_operations;
2930 
2931 	d_add(dentry, inode);
2932 	/* Close the race of the process dying before we return the dentry */
2933 	if (pid_revalidate(dentry, NULL))
2934 		error = NULL;
2935 out:
2936 	return error;
2937 }
2938 
2939 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2940 {
2941 	struct dentry *result;
2942 	struct task_struct *task;
2943 	unsigned tgid;
2944 	struct pid_namespace *ns;
2945 
2946 	result = proc_base_lookup(dir, dentry);
2947 	if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2948 		goto out;
2949 
2950 	tgid = name_to_int(dentry);
2951 	if (tgid == ~0U)
2952 		goto out;
2953 
2954 	ns = dentry->d_sb->s_fs_info;
2955 	rcu_read_lock();
2956 	task = find_task_by_pid_ns(tgid, ns);
2957 	if (task)
2958 		get_task_struct(task);
2959 	rcu_read_unlock();
2960 	if (!task)
2961 		goto out;
2962 
2963 	result = proc_pid_instantiate(dir, dentry, task, NULL);
2964 	put_task_struct(task);
2965 out:
2966 	return result;
2967 }
2968 
2969 /*
2970  * Find the first task with tgid >= tgid
2971  *
2972  */
2973 struct tgid_iter {
2974 	unsigned int tgid;
2975 	struct task_struct *task;
2976 };
2977 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2978 {
2979 	struct pid *pid;
2980 
2981 	if (iter.task)
2982 		put_task_struct(iter.task);
2983 	rcu_read_lock();
2984 retry:
2985 	iter.task = NULL;
2986 	pid = find_ge_pid(iter.tgid, ns);
2987 	if (pid) {
2988 		iter.tgid = pid_nr_ns(pid, ns);
2989 		iter.task = pid_task(pid, PIDTYPE_PID);
2990 		/* What we to know is if the pid we have find is the
2991 		 * pid of a thread_group_leader.  Testing for task
2992 		 * being a thread_group_leader is the obvious thing
2993 		 * todo but there is a window when it fails, due to
2994 		 * the pid transfer logic in de_thread.
2995 		 *
2996 		 * So we perform the straight forward test of seeing
2997 		 * if the pid we have found is the pid of a thread
2998 		 * group leader, and don't worry if the task we have
2999 		 * found doesn't happen to be a thread group leader.
3000 		 * As we don't care in the case of readdir.
3001 		 */
3002 		if (!iter.task || !has_group_leader_pid(iter.task)) {
3003 			iter.tgid += 1;
3004 			goto retry;
3005 		}
3006 		get_task_struct(iter.task);
3007 	}
3008 	rcu_read_unlock();
3009 	return iter;
3010 }
3011 
3012 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3013 
3014 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3015 	struct tgid_iter iter)
3016 {
3017 	char name[PROC_NUMBUF];
3018 	int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3019 	return proc_fill_cache(filp, dirent, filldir, name, len,
3020 				proc_pid_instantiate, iter.task, NULL);
3021 }
3022 
3023 /* for the /proc/ directory itself, after non-process stuff has been done */
3024 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3025 {
3026 	unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3027 	struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
3028 	struct tgid_iter iter;
3029 	struct pid_namespace *ns;
3030 
3031 	if (!reaper)
3032 		goto out_no_task;
3033 
3034 	for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3035 		const struct pid_entry *p = &proc_base_stuff[nr];
3036 		if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3037 			goto out;
3038 	}
3039 
3040 	ns = filp->f_dentry->d_sb->s_fs_info;
3041 	iter.task = NULL;
3042 	iter.tgid = filp->f_pos - TGID_OFFSET;
3043 	for (iter = next_tgid(ns, iter);
3044 	     iter.task;
3045 	     iter.tgid += 1, iter = next_tgid(ns, iter)) {
3046 		filp->f_pos = iter.tgid + TGID_OFFSET;
3047 		if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3048 			put_task_struct(iter.task);
3049 			goto out;
3050 		}
3051 	}
3052 	filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3053 out:
3054 	put_task_struct(reaper);
3055 out_no_task:
3056 	return 0;
3057 }
3058 
3059 /*
3060  * Tasks
3061  */
3062 static const struct pid_entry tid_base_stuff[] = {
3063 	DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3064 	DIR("fdinfo",    S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3065 	REG("environ",   S_IRUSR, proc_environ_operations),
3066 	INF("auxv",      S_IRUSR, proc_pid_auxv),
3067 	ONE("status",    S_IRUGO, proc_pid_status),
3068 	ONE("personality", S_IRUSR, proc_pid_personality),
3069 	INF("limits",	 S_IRUGO, proc_pid_limits),
3070 #ifdef CONFIG_SCHED_DEBUG
3071 	REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3072 #endif
3073 	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3074 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3075 	INF("syscall",   S_IRUSR, proc_pid_syscall),
3076 #endif
3077 	INF("cmdline",   S_IRUGO, proc_pid_cmdline),
3078 	ONE("stat",      S_IRUGO, proc_tid_stat),
3079 	ONE("statm",     S_IRUGO, proc_pid_statm),
3080 	REG("maps",      S_IRUGO, proc_maps_operations),
3081 #ifdef CONFIG_NUMA
3082 	REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
3083 #endif
3084 	REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
3085 	LNK("cwd",       proc_cwd_link),
3086 	LNK("root",      proc_root_link),
3087 	LNK("exe",       proc_exe_link),
3088 	REG("mounts",    S_IRUGO, proc_mounts_operations),
3089 	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3090 #ifdef CONFIG_PROC_PAGE_MONITOR
3091 	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3092 	REG("smaps",     S_IRUGO, proc_smaps_operations),
3093 	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3094 #endif
3095 #ifdef CONFIG_SECURITY
3096 	DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3097 #endif
3098 #ifdef CONFIG_KALLSYMS
3099 	INF("wchan",     S_IRUGO, proc_pid_wchan),
3100 #endif
3101 #ifdef CONFIG_STACKTRACE
3102 	ONE("stack",      S_IRUSR, proc_pid_stack),
3103 #endif
3104 #ifdef CONFIG_SCHEDSTATS
3105 	INF("schedstat", S_IRUGO, proc_pid_schedstat),
3106 #endif
3107 #ifdef CONFIG_LATENCYTOP
3108 	REG("latency",  S_IRUGO, proc_lstats_operations),
3109 #endif
3110 #ifdef CONFIG_PROC_PID_CPUSET
3111 	REG("cpuset",    S_IRUGO, proc_cpuset_operations),
3112 #endif
3113 #ifdef CONFIG_CGROUPS
3114 	REG("cgroup",  S_IRUGO, proc_cgroup_operations),
3115 #endif
3116 	INF("oom_score", S_IRUGO, proc_oom_score),
3117 	REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3118 	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3119 #ifdef CONFIG_AUDITSYSCALL
3120 	REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
3121 	REG("sessionid",  S_IRUSR, proc_sessionid_operations),
3122 #endif
3123 #ifdef CONFIG_FAULT_INJECTION
3124 	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3125 #endif
3126 #ifdef CONFIG_TASK_IO_ACCOUNTING
3127 	INF("io",	S_IRUGO, proc_tid_io_accounting),
3128 #endif
3129 };
3130 
3131 static int proc_tid_base_readdir(struct file * filp,
3132 			     void * dirent, filldir_t filldir)
3133 {
3134 	return proc_pident_readdir(filp,dirent,filldir,
3135 				   tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3136 }
3137 
3138 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3139 	return proc_pident_lookup(dir, dentry,
3140 				  tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3141 }
3142 
3143 static const struct file_operations proc_tid_base_operations = {
3144 	.read		= generic_read_dir,
3145 	.readdir	= proc_tid_base_readdir,
3146 	.llseek		= default_llseek,
3147 };
3148 
3149 static const struct inode_operations proc_tid_base_inode_operations = {
3150 	.lookup		= proc_tid_base_lookup,
3151 	.getattr	= pid_getattr,
3152 	.setattr	= proc_setattr,
3153 };
3154 
3155 static struct dentry *proc_task_instantiate(struct inode *dir,
3156 	struct dentry *dentry, struct task_struct *task, const void *ptr)
3157 {
3158 	struct dentry *error = ERR_PTR(-ENOENT);
3159 	struct inode *inode;
3160 	inode = proc_pid_make_inode(dir->i_sb, task);
3161 
3162 	if (!inode)
3163 		goto out;
3164 	inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3165 	inode->i_op = &proc_tid_base_inode_operations;
3166 	inode->i_fop = &proc_tid_base_operations;
3167 	inode->i_flags|=S_IMMUTABLE;
3168 
3169 	inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3170 		ARRAY_SIZE(tid_base_stuff));
3171 
3172 	dentry->d_op = &pid_dentry_operations;
3173 
3174 	d_add(dentry, inode);
3175 	/* Close the race of the process dying before we return the dentry */
3176 	if (pid_revalidate(dentry, NULL))
3177 		error = NULL;
3178 out:
3179 	return error;
3180 }
3181 
3182 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3183 {
3184 	struct dentry *result = ERR_PTR(-ENOENT);
3185 	struct task_struct *task;
3186 	struct task_struct *leader = get_proc_task(dir);
3187 	unsigned tid;
3188 	struct pid_namespace *ns;
3189 
3190 	if (!leader)
3191 		goto out_no_task;
3192 
3193 	tid = name_to_int(dentry);
3194 	if (tid == ~0U)
3195 		goto out;
3196 
3197 	ns = dentry->d_sb->s_fs_info;
3198 	rcu_read_lock();
3199 	task = find_task_by_pid_ns(tid, ns);
3200 	if (task)
3201 		get_task_struct(task);
3202 	rcu_read_unlock();
3203 	if (!task)
3204 		goto out;
3205 	if (!same_thread_group(leader, task))
3206 		goto out_drop_task;
3207 
3208 	result = proc_task_instantiate(dir, dentry, task, NULL);
3209 out_drop_task:
3210 	put_task_struct(task);
3211 out:
3212 	put_task_struct(leader);
3213 out_no_task:
3214 	return result;
3215 }
3216 
3217 /*
3218  * Find the first tid of a thread group to return to user space.
3219  *
3220  * Usually this is just the thread group leader, but if the users
3221  * buffer was too small or there was a seek into the middle of the
3222  * directory we have more work todo.
3223  *
3224  * In the case of a short read we start with find_task_by_pid.
3225  *
3226  * In the case of a seek we start with the leader and walk nr
3227  * threads past it.
3228  */
3229 static struct task_struct *first_tid(struct task_struct *leader,
3230 		int tid, int nr, struct pid_namespace *ns)
3231 {
3232 	struct task_struct *pos;
3233 
3234 	rcu_read_lock();
3235 	/* Attempt to start with the pid of a thread */
3236 	if (tid && (nr > 0)) {
3237 		pos = find_task_by_pid_ns(tid, ns);
3238 		if (pos && (pos->group_leader == leader))
3239 			goto found;
3240 	}
3241 
3242 	/* If nr exceeds the number of threads there is nothing todo */
3243 	pos = NULL;
3244 	if (nr && nr >= get_nr_threads(leader))
3245 		goto out;
3246 
3247 	/* If we haven't found our starting place yet start
3248 	 * with the leader and walk nr threads forward.
3249 	 */
3250 	for (pos = leader; nr > 0; --nr) {
3251 		pos = next_thread(pos);
3252 		if (pos == leader) {
3253 			pos = NULL;
3254 			goto out;
3255 		}
3256 	}
3257 found:
3258 	get_task_struct(pos);
3259 out:
3260 	rcu_read_unlock();
3261 	return pos;
3262 }
3263 
3264 /*
3265  * Find the next thread in the thread list.
3266  * Return NULL if there is an error or no next thread.
3267  *
3268  * The reference to the input task_struct is released.
3269  */
3270 static struct task_struct *next_tid(struct task_struct *start)
3271 {
3272 	struct task_struct *pos = NULL;
3273 	rcu_read_lock();
3274 	if (pid_alive(start)) {
3275 		pos = next_thread(start);
3276 		if (thread_group_leader(pos))
3277 			pos = NULL;
3278 		else
3279 			get_task_struct(pos);
3280 	}
3281 	rcu_read_unlock();
3282 	put_task_struct(start);
3283 	return pos;
3284 }
3285 
3286 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3287 	struct task_struct *task, int tid)
3288 {
3289 	char name[PROC_NUMBUF];
3290 	int len = snprintf(name, sizeof(name), "%d", tid);
3291 	return proc_fill_cache(filp, dirent, filldir, name, len,
3292 				proc_task_instantiate, task, NULL);
3293 }
3294 
3295 /* for the /proc/TGID/task/ directories */
3296 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3297 {
3298 	struct dentry *dentry = filp->f_path.dentry;
3299 	struct inode *inode = dentry->d_inode;
3300 	struct task_struct *leader = NULL;
3301 	struct task_struct *task;
3302 	int retval = -ENOENT;
3303 	ino_t ino;
3304 	int tid;
3305 	struct pid_namespace *ns;
3306 
3307 	task = get_proc_task(inode);
3308 	if (!task)
3309 		goto out_no_task;
3310 	rcu_read_lock();
3311 	if (pid_alive(task)) {
3312 		leader = task->group_leader;
3313 		get_task_struct(leader);
3314 	}
3315 	rcu_read_unlock();
3316 	put_task_struct(task);
3317 	if (!leader)
3318 		goto out_no_task;
3319 	retval = 0;
3320 
3321 	switch ((unsigned long)filp->f_pos) {
3322 	case 0:
3323 		ino = inode->i_ino;
3324 		if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3325 			goto out;
3326 		filp->f_pos++;
3327 		/* fall through */
3328 	case 1:
3329 		ino = parent_ino(dentry);
3330 		if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3331 			goto out;
3332 		filp->f_pos++;
3333 		/* fall through */
3334 	}
3335 
3336 	/* f_version caches the tgid value that the last readdir call couldn't
3337 	 * return. lseek aka telldir automagically resets f_version to 0.
3338 	 */
3339 	ns = filp->f_dentry->d_sb->s_fs_info;
3340 	tid = (int)filp->f_version;
3341 	filp->f_version = 0;
3342 	for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3343 	     task;
3344 	     task = next_tid(task), filp->f_pos++) {
3345 		tid = task_pid_nr_ns(task, ns);
3346 		if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3347 			/* returning this tgid failed, save it as the first
3348 			 * pid for the next readir call */
3349 			filp->f_version = (u64)tid;
3350 			put_task_struct(task);
3351 			break;
3352 		}
3353 	}
3354 out:
3355 	put_task_struct(leader);
3356 out_no_task:
3357 	return retval;
3358 }
3359 
3360 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3361 {
3362 	struct inode *inode = dentry->d_inode;
3363 	struct task_struct *p = get_proc_task(inode);
3364 	generic_fillattr(inode, stat);
3365 
3366 	if (p) {
3367 		stat->nlink += get_nr_threads(p);
3368 		put_task_struct(p);
3369 	}
3370 
3371 	return 0;
3372 }
3373 
3374 static const struct inode_operations proc_task_inode_operations = {
3375 	.lookup		= proc_task_lookup,
3376 	.getattr	= proc_task_getattr,
3377 	.setattr	= proc_setattr,
3378 };
3379 
3380 static const struct file_operations proc_task_operations = {
3381 	.read		= generic_read_dir,
3382 	.readdir	= proc_task_readdir,
3383 	.llseek		= default_llseek,
3384 };
3385