xref: /linux/fs/proc/base.c (revision 17afab1de42236ee2f6235f4383cc6f3f13f8a10)
1 /*
2  *  linux/fs/proc/base.c
3  *
4  *  Copyright (C) 1991, 1992 Linus Torvalds
5  *
6  *  proc base directory handling functions
7  *
8  *  1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9  *  Instead of using magical inumbers to determine the kind of object
10  *  we allocate and fill in-core inodes upon lookup. They don't even
11  *  go into icache. We cache the reference to task_struct upon lookup too.
12  *  Eventually it should become a filesystem in its own. We don't use the
13  *  rest of procfs anymore.
14  *
15  *
16  *  Changelog:
17  *  17-Jan-2005
18  *  Allan Bezerra
19  *  Bruna Moreira <bruna.moreira@indt.org.br>
20  *  Edjard Mota <edjard.mota@indt.org.br>
21  *  Ilias Biris <ilias.biris@indt.org.br>
22  *  Mauricio Lin <mauricio.lin@indt.org.br>
23  *
24  *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
25  *
26  *  A new process specific entry (smaps) included in /proc. It shows the
27  *  size of rss for each memory area. The maps entry lacks information
28  *  about physical memory size (rss) for each mapped file, i.e.,
29  *  rss information for executables and library files.
30  *  This additional information is useful for any tools that need to know
31  *  about physical memory consumption for a process specific library.
32  *
33  *  Changelog:
34  *  21-Feb-2005
35  *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36  *  Pud inclusion in the page table walking.
37  *
38  *  ChangeLog:
39  *  10-Mar-2005
40  *  10LE Instituto Nokia de Tecnologia - INdT:
41  *  A better way to walks through the page table as suggested by Hugh Dickins.
42  *
43  *  Simo Piiroinen <simo.piiroinen@nokia.com>:
44  *  Smaps information related to shared, private, clean and dirty pages.
45  *
46  *  Paul Mundt <paul.mundt@nokia.com>:
47  *  Overall revision about smaps.
48  */
49 
50 #include <asm/uaccess.h>
51 
52 #include <linux/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/task_io_accounting_ops.h>
57 #include <linux/init.h>
58 #include <linux/capability.h>
59 #include <linux/file.h>
60 #include <linux/fdtable.h>
61 #include <linux/string.h>
62 #include <linux/seq_file.h>
63 #include <linux/namei.h>
64 #include <linux/mnt_namespace.h>
65 #include <linux/mm.h>
66 #include <linux/swap.h>
67 #include <linux/rcupdate.h>
68 #include <linux/kallsyms.h>
69 #include <linux/stacktrace.h>
70 #include <linux/resource.h>
71 #include <linux/module.h>
72 #include <linux/mount.h>
73 #include <linux/security.h>
74 #include <linux/ptrace.h>
75 #include <linux/tracehook.h>
76 #include <linux/printk.h>
77 #include <linux/cgroup.h>
78 #include <linux/cpuset.h>
79 #include <linux/audit.h>
80 #include <linux/poll.h>
81 #include <linux/nsproxy.h>
82 #include <linux/oom.h>
83 #include <linux/elf.h>
84 #include <linux/pid_namespace.h>
85 #include <linux/user_namespace.h>
86 #include <linux/fs_struct.h>
87 #include <linux/slab.h>
88 #include <linux/flex_array.h>
89 #include <linux/posix-timers.h>
90 #ifdef CONFIG_HARDWALL
91 #include <asm/hardwall.h>
92 #endif
93 #include <trace/events/oom.h>
94 #include "internal.h"
95 #include "fd.h"
96 
97 /* NOTE:
98  *	Implementing inode permission operations in /proc is almost
99  *	certainly an error.  Permission checks need to happen during
100  *	each system call not at open time.  The reason is that most of
101  *	what we wish to check for permissions in /proc varies at runtime.
102  *
103  *	The classic example of a problem is opening file descriptors
104  *	in /proc for a task before it execs a suid executable.
105  */
106 
107 struct pid_entry {
108 	char *name;
109 	int len;
110 	umode_t mode;
111 	const struct inode_operations *iop;
112 	const struct file_operations *fop;
113 	union proc_op op;
114 };
115 
116 #define NOD(NAME, MODE, IOP, FOP, OP) {			\
117 	.name = (NAME),					\
118 	.len  = sizeof(NAME) - 1,			\
119 	.mode = MODE,					\
120 	.iop  = IOP,					\
121 	.fop  = FOP,					\
122 	.op   = OP,					\
123 }
124 
125 #define DIR(NAME, MODE, iops, fops)	\
126 	NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
127 #define LNK(NAME, get_link)					\
128 	NOD(NAME, (S_IFLNK|S_IRWXUGO),				\
129 		&proc_pid_link_inode_operations, NULL,		\
130 		{ .proc_get_link = get_link } )
131 #define REG(NAME, MODE, fops)				\
132 	NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
133 #define INF(NAME, MODE, read)				\
134 	NOD(NAME, (S_IFREG|(MODE)), 			\
135 		NULL, &proc_info_file_operations,	\
136 		{ .proc_read = read } )
137 #define ONE(NAME, MODE, show)				\
138 	NOD(NAME, (S_IFREG|(MODE)), 			\
139 		NULL, &proc_single_file_operations,	\
140 		{ .proc_show = show } )
141 
142 /*
143  * Count the number of hardlinks for the pid_entry table, excluding the .
144  * and .. links.
145  */
146 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
147 	unsigned int n)
148 {
149 	unsigned int i;
150 	unsigned int count;
151 
152 	count = 0;
153 	for (i = 0; i < n; ++i) {
154 		if (S_ISDIR(entries[i].mode))
155 			++count;
156 	}
157 
158 	return count;
159 }
160 
161 static int get_task_root(struct task_struct *task, struct path *root)
162 {
163 	int result = -ENOENT;
164 
165 	task_lock(task);
166 	if (task->fs) {
167 		get_fs_root(task->fs, root);
168 		result = 0;
169 	}
170 	task_unlock(task);
171 	return result;
172 }
173 
174 static int proc_cwd_link(struct dentry *dentry, struct path *path)
175 {
176 	struct task_struct *task = get_proc_task(dentry->d_inode);
177 	int result = -ENOENT;
178 
179 	if (task) {
180 		task_lock(task);
181 		if (task->fs) {
182 			get_fs_pwd(task->fs, path);
183 			result = 0;
184 		}
185 		task_unlock(task);
186 		put_task_struct(task);
187 	}
188 	return result;
189 }
190 
191 static int proc_root_link(struct dentry *dentry, struct path *path)
192 {
193 	struct task_struct *task = get_proc_task(dentry->d_inode);
194 	int result = -ENOENT;
195 
196 	if (task) {
197 		result = get_task_root(task, path);
198 		put_task_struct(task);
199 	}
200 	return result;
201 }
202 
203 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
204 {
205 	int res = 0;
206 	unsigned int len;
207 	struct mm_struct *mm = get_task_mm(task);
208 	if (!mm)
209 		goto out;
210 	if (!mm->arg_end)
211 		goto out_mm;	/* Shh! No looking before we're done */
212 
213  	len = mm->arg_end - mm->arg_start;
214 
215 	if (len > PAGE_SIZE)
216 		len = PAGE_SIZE;
217 
218 	res = access_process_vm(task, mm->arg_start, buffer, len, 0);
219 
220 	// If the nul at the end of args has been overwritten, then
221 	// assume application is using setproctitle(3).
222 	if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
223 		len = strnlen(buffer, res);
224 		if (len < res) {
225 		    res = len;
226 		} else {
227 			len = mm->env_end - mm->env_start;
228 			if (len > PAGE_SIZE - res)
229 				len = PAGE_SIZE - res;
230 			res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
231 			res = strnlen(buffer, res);
232 		}
233 	}
234 out_mm:
235 	mmput(mm);
236 out:
237 	return res;
238 }
239 
240 static int proc_pid_auxv(struct task_struct *task, char *buffer)
241 {
242 	struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
243 	int res = PTR_ERR(mm);
244 	if (mm && !IS_ERR(mm)) {
245 		unsigned int nwords = 0;
246 		do {
247 			nwords += 2;
248 		} while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
249 		res = nwords * sizeof(mm->saved_auxv[0]);
250 		if (res > PAGE_SIZE)
251 			res = PAGE_SIZE;
252 		memcpy(buffer, mm->saved_auxv, res);
253 		mmput(mm);
254 	}
255 	return res;
256 }
257 
258 
259 #ifdef CONFIG_KALLSYMS
260 /*
261  * Provides a wchan file via kallsyms in a proper one-value-per-file format.
262  * Returns the resolved symbol.  If that fails, simply return the address.
263  */
264 static int proc_pid_wchan(struct task_struct *task, char *buffer)
265 {
266 	unsigned long wchan;
267 	char symname[KSYM_NAME_LEN];
268 
269 	wchan = get_wchan(task);
270 
271 	if (lookup_symbol_name(wchan, symname) < 0)
272 		if (!ptrace_may_access(task, PTRACE_MODE_READ))
273 			return 0;
274 		else
275 			return sprintf(buffer, "%lu", wchan);
276 	else
277 		return sprintf(buffer, "%s", symname);
278 }
279 #endif /* CONFIG_KALLSYMS */
280 
281 static int lock_trace(struct task_struct *task)
282 {
283 	int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
284 	if (err)
285 		return err;
286 	if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
287 		mutex_unlock(&task->signal->cred_guard_mutex);
288 		return -EPERM;
289 	}
290 	return 0;
291 }
292 
293 static void unlock_trace(struct task_struct *task)
294 {
295 	mutex_unlock(&task->signal->cred_guard_mutex);
296 }
297 
298 #ifdef CONFIG_STACKTRACE
299 
300 #define MAX_STACK_TRACE_DEPTH	64
301 
302 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
303 			  struct pid *pid, struct task_struct *task)
304 {
305 	struct stack_trace trace;
306 	unsigned long *entries;
307 	int err;
308 	int i;
309 
310 	entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
311 	if (!entries)
312 		return -ENOMEM;
313 
314 	trace.nr_entries	= 0;
315 	trace.max_entries	= MAX_STACK_TRACE_DEPTH;
316 	trace.entries		= entries;
317 	trace.skip		= 0;
318 
319 	err = lock_trace(task);
320 	if (!err) {
321 		save_stack_trace_tsk(task, &trace);
322 
323 		for (i = 0; i < trace.nr_entries; i++) {
324 			seq_printf(m, "[<%pK>] %pS\n",
325 				   (void *)entries[i], (void *)entries[i]);
326 		}
327 		unlock_trace(task);
328 	}
329 	kfree(entries);
330 
331 	return err;
332 }
333 #endif
334 
335 #ifdef CONFIG_SCHEDSTATS
336 /*
337  * Provides /proc/PID/schedstat
338  */
339 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
340 {
341 	return sprintf(buffer, "%llu %llu %lu\n",
342 			(unsigned long long)task->se.sum_exec_runtime,
343 			(unsigned long long)task->sched_info.run_delay,
344 			task->sched_info.pcount);
345 }
346 #endif
347 
348 #ifdef CONFIG_LATENCYTOP
349 static int lstats_show_proc(struct seq_file *m, void *v)
350 {
351 	int i;
352 	struct inode *inode = m->private;
353 	struct task_struct *task = get_proc_task(inode);
354 
355 	if (!task)
356 		return -ESRCH;
357 	seq_puts(m, "Latency Top version : v0.1\n");
358 	for (i = 0; i < 32; i++) {
359 		struct latency_record *lr = &task->latency_record[i];
360 		if (lr->backtrace[0]) {
361 			int q;
362 			seq_printf(m, "%i %li %li",
363 				   lr->count, lr->time, lr->max);
364 			for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
365 				unsigned long bt = lr->backtrace[q];
366 				if (!bt)
367 					break;
368 				if (bt == ULONG_MAX)
369 					break;
370 				seq_printf(m, " %ps", (void *)bt);
371 			}
372 			seq_putc(m, '\n');
373 		}
374 
375 	}
376 	put_task_struct(task);
377 	return 0;
378 }
379 
380 static int lstats_open(struct inode *inode, struct file *file)
381 {
382 	return single_open(file, lstats_show_proc, inode);
383 }
384 
385 static ssize_t lstats_write(struct file *file, const char __user *buf,
386 			    size_t count, loff_t *offs)
387 {
388 	struct task_struct *task = get_proc_task(file_inode(file));
389 
390 	if (!task)
391 		return -ESRCH;
392 	clear_all_latency_tracing(task);
393 	put_task_struct(task);
394 
395 	return count;
396 }
397 
398 static const struct file_operations proc_lstats_operations = {
399 	.open		= lstats_open,
400 	.read		= seq_read,
401 	.write		= lstats_write,
402 	.llseek		= seq_lseek,
403 	.release	= single_release,
404 };
405 
406 #endif
407 
408 static int proc_oom_score(struct task_struct *task, char *buffer)
409 {
410 	unsigned long totalpages = totalram_pages + total_swap_pages;
411 	unsigned long points = 0;
412 
413 	read_lock(&tasklist_lock);
414 	if (pid_alive(task))
415 		points = oom_badness(task, NULL, NULL, totalpages) *
416 						1000 / totalpages;
417 	read_unlock(&tasklist_lock);
418 	return sprintf(buffer, "%lu\n", points);
419 }
420 
421 struct limit_names {
422 	char *name;
423 	char *unit;
424 };
425 
426 static const struct limit_names lnames[RLIM_NLIMITS] = {
427 	[RLIMIT_CPU] = {"Max cpu time", "seconds"},
428 	[RLIMIT_FSIZE] = {"Max file size", "bytes"},
429 	[RLIMIT_DATA] = {"Max data size", "bytes"},
430 	[RLIMIT_STACK] = {"Max stack size", "bytes"},
431 	[RLIMIT_CORE] = {"Max core file size", "bytes"},
432 	[RLIMIT_RSS] = {"Max resident set", "bytes"},
433 	[RLIMIT_NPROC] = {"Max processes", "processes"},
434 	[RLIMIT_NOFILE] = {"Max open files", "files"},
435 	[RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
436 	[RLIMIT_AS] = {"Max address space", "bytes"},
437 	[RLIMIT_LOCKS] = {"Max file locks", "locks"},
438 	[RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
439 	[RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
440 	[RLIMIT_NICE] = {"Max nice priority", NULL},
441 	[RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
442 	[RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
443 };
444 
445 /* Display limits for a process */
446 static int proc_pid_limits(struct task_struct *task, char *buffer)
447 {
448 	unsigned int i;
449 	int count = 0;
450 	unsigned long flags;
451 	char *bufptr = buffer;
452 
453 	struct rlimit rlim[RLIM_NLIMITS];
454 
455 	if (!lock_task_sighand(task, &flags))
456 		return 0;
457 	memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
458 	unlock_task_sighand(task, &flags);
459 
460 	/*
461 	 * print the file header
462 	 */
463 	count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
464 			"Limit", "Soft Limit", "Hard Limit", "Units");
465 
466 	for (i = 0; i < RLIM_NLIMITS; i++) {
467 		if (rlim[i].rlim_cur == RLIM_INFINITY)
468 			count += sprintf(&bufptr[count], "%-25s %-20s ",
469 					 lnames[i].name, "unlimited");
470 		else
471 			count += sprintf(&bufptr[count], "%-25s %-20lu ",
472 					 lnames[i].name, rlim[i].rlim_cur);
473 
474 		if (rlim[i].rlim_max == RLIM_INFINITY)
475 			count += sprintf(&bufptr[count], "%-20s ", "unlimited");
476 		else
477 			count += sprintf(&bufptr[count], "%-20lu ",
478 					 rlim[i].rlim_max);
479 
480 		if (lnames[i].unit)
481 			count += sprintf(&bufptr[count], "%-10s\n",
482 					 lnames[i].unit);
483 		else
484 			count += sprintf(&bufptr[count], "\n");
485 	}
486 
487 	return count;
488 }
489 
490 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
491 static int proc_pid_syscall(struct task_struct *task, char *buffer)
492 {
493 	long nr;
494 	unsigned long args[6], sp, pc;
495 	int res = lock_trace(task);
496 	if (res)
497 		return res;
498 
499 	if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
500 		res = sprintf(buffer, "running\n");
501 	else if (nr < 0)
502 		res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
503 	else
504 		res = sprintf(buffer,
505 		       "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
506 		       nr,
507 		       args[0], args[1], args[2], args[3], args[4], args[5],
508 		       sp, pc);
509 	unlock_trace(task);
510 	return res;
511 }
512 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
513 
514 /************************************************************************/
515 /*                       Here the fs part begins                        */
516 /************************************************************************/
517 
518 /* permission checks */
519 static int proc_fd_access_allowed(struct inode *inode)
520 {
521 	struct task_struct *task;
522 	int allowed = 0;
523 	/* Allow access to a task's file descriptors if it is us or we
524 	 * may use ptrace attach to the process and find out that
525 	 * information.
526 	 */
527 	task = get_proc_task(inode);
528 	if (task) {
529 		allowed = ptrace_may_access(task, PTRACE_MODE_READ);
530 		put_task_struct(task);
531 	}
532 	return allowed;
533 }
534 
535 int proc_setattr(struct dentry *dentry, struct iattr *attr)
536 {
537 	int error;
538 	struct inode *inode = dentry->d_inode;
539 
540 	if (attr->ia_valid & ATTR_MODE)
541 		return -EPERM;
542 
543 	error = inode_change_ok(inode, attr);
544 	if (error)
545 		return error;
546 
547 	setattr_copy(inode, attr);
548 	mark_inode_dirty(inode);
549 	return 0;
550 }
551 
552 /*
553  * May current process learn task's sched/cmdline info (for hide_pid_min=1)
554  * or euid/egid (for hide_pid_min=2)?
555  */
556 static bool has_pid_permissions(struct pid_namespace *pid,
557 				 struct task_struct *task,
558 				 int hide_pid_min)
559 {
560 	if (pid->hide_pid < hide_pid_min)
561 		return true;
562 	if (in_group_p(pid->pid_gid))
563 		return true;
564 	return ptrace_may_access(task, PTRACE_MODE_READ);
565 }
566 
567 
568 static int proc_pid_permission(struct inode *inode, int mask)
569 {
570 	struct pid_namespace *pid = inode->i_sb->s_fs_info;
571 	struct task_struct *task;
572 	bool has_perms;
573 
574 	task = get_proc_task(inode);
575 	if (!task)
576 		return -ESRCH;
577 	has_perms = has_pid_permissions(pid, task, 1);
578 	put_task_struct(task);
579 
580 	if (!has_perms) {
581 		if (pid->hide_pid == 2) {
582 			/*
583 			 * Let's make getdents(), stat(), and open()
584 			 * consistent with each other.  If a process
585 			 * may not stat() a file, it shouldn't be seen
586 			 * in procfs at all.
587 			 */
588 			return -ENOENT;
589 		}
590 
591 		return -EPERM;
592 	}
593 	return generic_permission(inode, mask);
594 }
595 
596 
597 
598 static const struct inode_operations proc_def_inode_operations = {
599 	.setattr	= proc_setattr,
600 };
601 
602 #define PROC_BLOCK_SIZE	(3*1024)		/* 4K page size but our output routines use some slack for overruns */
603 
604 static ssize_t proc_info_read(struct file * file, char __user * buf,
605 			  size_t count, loff_t *ppos)
606 {
607 	struct inode * inode = file_inode(file);
608 	unsigned long page;
609 	ssize_t length;
610 	struct task_struct *task = get_proc_task(inode);
611 
612 	length = -ESRCH;
613 	if (!task)
614 		goto out_no_task;
615 
616 	if (count > PROC_BLOCK_SIZE)
617 		count = PROC_BLOCK_SIZE;
618 
619 	length = -ENOMEM;
620 	if (!(page = __get_free_page(GFP_TEMPORARY)))
621 		goto out;
622 
623 	length = PROC_I(inode)->op.proc_read(task, (char*)page);
624 
625 	if (length >= 0)
626 		length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
627 	free_page(page);
628 out:
629 	put_task_struct(task);
630 out_no_task:
631 	return length;
632 }
633 
634 static const struct file_operations proc_info_file_operations = {
635 	.read		= proc_info_read,
636 	.llseek		= generic_file_llseek,
637 };
638 
639 static int proc_single_show(struct seq_file *m, void *v)
640 {
641 	struct inode *inode = m->private;
642 	struct pid_namespace *ns;
643 	struct pid *pid;
644 	struct task_struct *task;
645 	int ret;
646 
647 	ns = inode->i_sb->s_fs_info;
648 	pid = proc_pid(inode);
649 	task = get_pid_task(pid, PIDTYPE_PID);
650 	if (!task)
651 		return -ESRCH;
652 
653 	ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
654 
655 	put_task_struct(task);
656 	return ret;
657 }
658 
659 static int proc_single_open(struct inode *inode, struct file *filp)
660 {
661 	return single_open(filp, proc_single_show, inode);
662 }
663 
664 static const struct file_operations proc_single_file_operations = {
665 	.open		= proc_single_open,
666 	.read		= seq_read,
667 	.llseek		= seq_lseek,
668 	.release	= single_release,
669 };
670 
671 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
672 {
673 	struct task_struct *task = get_proc_task(file_inode(file));
674 	struct mm_struct *mm;
675 
676 	if (!task)
677 		return -ESRCH;
678 
679 	mm = mm_access(task, mode);
680 	put_task_struct(task);
681 
682 	if (IS_ERR(mm))
683 		return PTR_ERR(mm);
684 
685 	if (mm) {
686 		/* ensure this mm_struct can't be freed */
687 		atomic_inc(&mm->mm_count);
688 		/* but do not pin its memory */
689 		mmput(mm);
690 	}
691 
692 	file->private_data = mm;
693 
694 	return 0;
695 }
696 
697 static int mem_open(struct inode *inode, struct file *file)
698 {
699 	int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
700 
701 	/* OK to pass negative loff_t, we can catch out-of-range */
702 	file->f_mode |= FMODE_UNSIGNED_OFFSET;
703 
704 	return ret;
705 }
706 
707 static ssize_t mem_rw(struct file *file, char __user *buf,
708 			size_t count, loff_t *ppos, int write)
709 {
710 	struct mm_struct *mm = file->private_data;
711 	unsigned long addr = *ppos;
712 	ssize_t copied;
713 	char *page;
714 
715 	if (!mm)
716 		return 0;
717 
718 	page = (char *)__get_free_page(GFP_TEMPORARY);
719 	if (!page)
720 		return -ENOMEM;
721 
722 	copied = 0;
723 	if (!atomic_inc_not_zero(&mm->mm_users))
724 		goto free;
725 
726 	while (count > 0) {
727 		int this_len = min_t(int, count, PAGE_SIZE);
728 
729 		if (write && copy_from_user(page, buf, this_len)) {
730 			copied = -EFAULT;
731 			break;
732 		}
733 
734 		this_len = access_remote_vm(mm, addr, page, this_len, write);
735 		if (!this_len) {
736 			if (!copied)
737 				copied = -EIO;
738 			break;
739 		}
740 
741 		if (!write && copy_to_user(buf, page, this_len)) {
742 			copied = -EFAULT;
743 			break;
744 		}
745 
746 		buf += this_len;
747 		addr += this_len;
748 		copied += this_len;
749 		count -= this_len;
750 	}
751 	*ppos = addr;
752 
753 	mmput(mm);
754 free:
755 	free_page((unsigned long) page);
756 	return copied;
757 }
758 
759 static ssize_t mem_read(struct file *file, char __user *buf,
760 			size_t count, loff_t *ppos)
761 {
762 	return mem_rw(file, buf, count, ppos, 0);
763 }
764 
765 static ssize_t mem_write(struct file *file, const char __user *buf,
766 			 size_t count, loff_t *ppos)
767 {
768 	return mem_rw(file, (char __user*)buf, count, ppos, 1);
769 }
770 
771 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
772 {
773 	switch (orig) {
774 	case 0:
775 		file->f_pos = offset;
776 		break;
777 	case 1:
778 		file->f_pos += offset;
779 		break;
780 	default:
781 		return -EINVAL;
782 	}
783 	force_successful_syscall_return();
784 	return file->f_pos;
785 }
786 
787 static int mem_release(struct inode *inode, struct file *file)
788 {
789 	struct mm_struct *mm = file->private_data;
790 	if (mm)
791 		mmdrop(mm);
792 	return 0;
793 }
794 
795 static const struct file_operations proc_mem_operations = {
796 	.llseek		= mem_lseek,
797 	.read		= mem_read,
798 	.write		= mem_write,
799 	.open		= mem_open,
800 	.release	= mem_release,
801 };
802 
803 static int environ_open(struct inode *inode, struct file *file)
804 {
805 	return __mem_open(inode, file, PTRACE_MODE_READ);
806 }
807 
808 static ssize_t environ_read(struct file *file, char __user *buf,
809 			size_t count, loff_t *ppos)
810 {
811 	char *page;
812 	unsigned long src = *ppos;
813 	int ret = 0;
814 	struct mm_struct *mm = file->private_data;
815 
816 	if (!mm)
817 		return 0;
818 
819 	page = (char *)__get_free_page(GFP_TEMPORARY);
820 	if (!page)
821 		return -ENOMEM;
822 
823 	ret = 0;
824 	if (!atomic_inc_not_zero(&mm->mm_users))
825 		goto free;
826 	while (count > 0) {
827 		size_t this_len, max_len;
828 		int retval;
829 
830 		if (src >= (mm->env_end - mm->env_start))
831 			break;
832 
833 		this_len = mm->env_end - (mm->env_start + src);
834 
835 		max_len = min_t(size_t, PAGE_SIZE, count);
836 		this_len = min(max_len, this_len);
837 
838 		retval = access_remote_vm(mm, (mm->env_start + src),
839 			page, this_len, 0);
840 
841 		if (retval <= 0) {
842 			ret = retval;
843 			break;
844 		}
845 
846 		if (copy_to_user(buf, page, retval)) {
847 			ret = -EFAULT;
848 			break;
849 		}
850 
851 		ret += retval;
852 		src += retval;
853 		buf += retval;
854 		count -= retval;
855 	}
856 	*ppos = src;
857 	mmput(mm);
858 
859 free:
860 	free_page((unsigned long) page);
861 	return ret;
862 }
863 
864 static const struct file_operations proc_environ_operations = {
865 	.open		= environ_open,
866 	.read		= environ_read,
867 	.llseek		= generic_file_llseek,
868 	.release	= mem_release,
869 };
870 
871 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
872 			    loff_t *ppos)
873 {
874 	struct task_struct *task = get_proc_task(file_inode(file));
875 	char buffer[PROC_NUMBUF];
876 	int oom_adj = OOM_ADJUST_MIN;
877 	size_t len;
878 	unsigned long flags;
879 
880 	if (!task)
881 		return -ESRCH;
882 	if (lock_task_sighand(task, &flags)) {
883 		if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
884 			oom_adj = OOM_ADJUST_MAX;
885 		else
886 			oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
887 				  OOM_SCORE_ADJ_MAX;
888 		unlock_task_sighand(task, &flags);
889 	}
890 	put_task_struct(task);
891 	len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
892 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
893 }
894 
895 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
896 			     size_t count, loff_t *ppos)
897 {
898 	struct task_struct *task;
899 	char buffer[PROC_NUMBUF];
900 	int oom_adj;
901 	unsigned long flags;
902 	int err;
903 
904 	memset(buffer, 0, sizeof(buffer));
905 	if (count > sizeof(buffer) - 1)
906 		count = sizeof(buffer) - 1;
907 	if (copy_from_user(buffer, buf, count)) {
908 		err = -EFAULT;
909 		goto out;
910 	}
911 
912 	err = kstrtoint(strstrip(buffer), 0, &oom_adj);
913 	if (err)
914 		goto out;
915 	if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
916 	     oom_adj != OOM_DISABLE) {
917 		err = -EINVAL;
918 		goto out;
919 	}
920 
921 	task = get_proc_task(file_inode(file));
922 	if (!task) {
923 		err = -ESRCH;
924 		goto out;
925 	}
926 
927 	task_lock(task);
928 	if (!task->mm) {
929 		err = -EINVAL;
930 		goto err_task_lock;
931 	}
932 
933 	if (!lock_task_sighand(task, &flags)) {
934 		err = -ESRCH;
935 		goto err_task_lock;
936 	}
937 
938 	/*
939 	 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
940 	 * value is always attainable.
941 	 */
942 	if (oom_adj == OOM_ADJUST_MAX)
943 		oom_adj = OOM_SCORE_ADJ_MAX;
944 	else
945 		oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
946 
947 	if (oom_adj < task->signal->oom_score_adj &&
948 	    !capable(CAP_SYS_RESOURCE)) {
949 		err = -EACCES;
950 		goto err_sighand;
951 	}
952 
953 	/*
954 	 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
955 	 * /proc/pid/oom_score_adj instead.
956 	 */
957 	pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
958 		  current->comm, task_pid_nr(current), task_pid_nr(task),
959 		  task_pid_nr(task));
960 
961 	task->signal->oom_score_adj = oom_adj;
962 	trace_oom_score_adj_update(task);
963 err_sighand:
964 	unlock_task_sighand(task, &flags);
965 err_task_lock:
966 	task_unlock(task);
967 	put_task_struct(task);
968 out:
969 	return err < 0 ? err : count;
970 }
971 
972 static const struct file_operations proc_oom_adj_operations = {
973 	.read		= oom_adj_read,
974 	.write		= oom_adj_write,
975 	.llseek		= generic_file_llseek,
976 };
977 
978 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
979 					size_t count, loff_t *ppos)
980 {
981 	struct task_struct *task = get_proc_task(file_inode(file));
982 	char buffer[PROC_NUMBUF];
983 	short oom_score_adj = OOM_SCORE_ADJ_MIN;
984 	unsigned long flags;
985 	size_t len;
986 
987 	if (!task)
988 		return -ESRCH;
989 	if (lock_task_sighand(task, &flags)) {
990 		oom_score_adj = task->signal->oom_score_adj;
991 		unlock_task_sighand(task, &flags);
992 	}
993 	put_task_struct(task);
994 	len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
995 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
996 }
997 
998 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
999 					size_t count, loff_t *ppos)
1000 {
1001 	struct task_struct *task;
1002 	char buffer[PROC_NUMBUF];
1003 	unsigned long flags;
1004 	int oom_score_adj;
1005 	int err;
1006 
1007 	memset(buffer, 0, sizeof(buffer));
1008 	if (count > sizeof(buffer) - 1)
1009 		count = sizeof(buffer) - 1;
1010 	if (copy_from_user(buffer, buf, count)) {
1011 		err = -EFAULT;
1012 		goto out;
1013 	}
1014 
1015 	err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1016 	if (err)
1017 		goto out;
1018 	if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1019 			oom_score_adj > OOM_SCORE_ADJ_MAX) {
1020 		err = -EINVAL;
1021 		goto out;
1022 	}
1023 
1024 	task = get_proc_task(file_inode(file));
1025 	if (!task) {
1026 		err = -ESRCH;
1027 		goto out;
1028 	}
1029 
1030 	task_lock(task);
1031 	if (!task->mm) {
1032 		err = -EINVAL;
1033 		goto err_task_lock;
1034 	}
1035 
1036 	if (!lock_task_sighand(task, &flags)) {
1037 		err = -ESRCH;
1038 		goto err_task_lock;
1039 	}
1040 
1041 	if ((short)oom_score_adj < task->signal->oom_score_adj_min &&
1042 			!capable(CAP_SYS_RESOURCE)) {
1043 		err = -EACCES;
1044 		goto err_sighand;
1045 	}
1046 
1047 	task->signal->oom_score_adj = (short)oom_score_adj;
1048 	if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1049 		task->signal->oom_score_adj_min = (short)oom_score_adj;
1050 	trace_oom_score_adj_update(task);
1051 
1052 err_sighand:
1053 	unlock_task_sighand(task, &flags);
1054 err_task_lock:
1055 	task_unlock(task);
1056 	put_task_struct(task);
1057 out:
1058 	return err < 0 ? err : count;
1059 }
1060 
1061 static const struct file_operations proc_oom_score_adj_operations = {
1062 	.read		= oom_score_adj_read,
1063 	.write		= oom_score_adj_write,
1064 	.llseek		= default_llseek,
1065 };
1066 
1067 #ifdef CONFIG_AUDITSYSCALL
1068 #define TMPBUFLEN 21
1069 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1070 				  size_t count, loff_t *ppos)
1071 {
1072 	struct inode * inode = file_inode(file);
1073 	struct task_struct *task = get_proc_task(inode);
1074 	ssize_t length;
1075 	char tmpbuf[TMPBUFLEN];
1076 
1077 	if (!task)
1078 		return -ESRCH;
1079 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1080 			   from_kuid(file->f_cred->user_ns,
1081 				     audit_get_loginuid(task)));
1082 	put_task_struct(task);
1083 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1084 }
1085 
1086 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1087 				   size_t count, loff_t *ppos)
1088 {
1089 	struct inode * inode = file_inode(file);
1090 	char *page, *tmp;
1091 	ssize_t length;
1092 	uid_t loginuid;
1093 	kuid_t kloginuid;
1094 
1095 	rcu_read_lock();
1096 	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1097 		rcu_read_unlock();
1098 		return -EPERM;
1099 	}
1100 	rcu_read_unlock();
1101 
1102 	if (count >= PAGE_SIZE)
1103 		count = PAGE_SIZE - 1;
1104 
1105 	if (*ppos != 0) {
1106 		/* No partial writes. */
1107 		return -EINVAL;
1108 	}
1109 	page = (char*)__get_free_page(GFP_TEMPORARY);
1110 	if (!page)
1111 		return -ENOMEM;
1112 	length = -EFAULT;
1113 	if (copy_from_user(page, buf, count))
1114 		goto out_free_page;
1115 
1116 	page[count] = '\0';
1117 	loginuid = simple_strtoul(page, &tmp, 10);
1118 	if (tmp == page) {
1119 		length = -EINVAL;
1120 		goto out_free_page;
1121 
1122 	}
1123 	kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1124 	if (!uid_valid(kloginuid)) {
1125 		length = -EINVAL;
1126 		goto out_free_page;
1127 	}
1128 
1129 	length = audit_set_loginuid(kloginuid);
1130 	if (likely(length == 0))
1131 		length = count;
1132 
1133 out_free_page:
1134 	free_page((unsigned long) page);
1135 	return length;
1136 }
1137 
1138 static const struct file_operations proc_loginuid_operations = {
1139 	.read		= proc_loginuid_read,
1140 	.write		= proc_loginuid_write,
1141 	.llseek		= generic_file_llseek,
1142 };
1143 
1144 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1145 				  size_t count, loff_t *ppos)
1146 {
1147 	struct inode * inode = file_inode(file);
1148 	struct task_struct *task = get_proc_task(inode);
1149 	ssize_t length;
1150 	char tmpbuf[TMPBUFLEN];
1151 
1152 	if (!task)
1153 		return -ESRCH;
1154 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1155 				audit_get_sessionid(task));
1156 	put_task_struct(task);
1157 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1158 }
1159 
1160 static const struct file_operations proc_sessionid_operations = {
1161 	.read		= proc_sessionid_read,
1162 	.llseek		= generic_file_llseek,
1163 };
1164 #endif
1165 
1166 #ifdef CONFIG_FAULT_INJECTION
1167 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1168 				      size_t count, loff_t *ppos)
1169 {
1170 	struct task_struct *task = get_proc_task(file_inode(file));
1171 	char buffer[PROC_NUMBUF];
1172 	size_t len;
1173 	int make_it_fail;
1174 
1175 	if (!task)
1176 		return -ESRCH;
1177 	make_it_fail = task->make_it_fail;
1178 	put_task_struct(task);
1179 
1180 	len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1181 
1182 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1183 }
1184 
1185 static ssize_t proc_fault_inject_write(struct file * file,
1186 			const char __user * buf, size_t count, loff_t *ppos)
1187 {
1188 	struct task_struct *task;
1189 	char buffer[PROC_NUMBUF], *end;
1190 	int make_it_fail;
1191 
1192 	if (!capable(CAP_SYS_RESOURCE))
1193 		return -EPERM;
1194 	memset(buffer, 0, sizeof(buffer));
1195 	if (count > sizeof(buffer) - 1)
1196 		count = sizeof(buffer) - 1;
1197 	if (copy_from_user(buffer, buf, count))
1198 		return -EFAULT;
1199 	make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1200 	if (*end)
1201 		return -EINVAL;
1202 	task = get_proc_task(file_inode(file));
1203 	if (!task)
1204 		return -ESRCH;
1205 	task->make_it_fail = make_it_fail;
1206 	put_task_struct(task);
1207 
1208 	return count;
1209 }
1210 
1211 static const struct file_operations proc_fault_inject_operations = {
1212 	.read		= proc_fault_inject_read,
1213 	.write		= proc_fault_inject_write,
1214 	.llseek		= generic_file_llseek,
1215 };
1216 #endif
1217 
1218 
1219 #ifdef CONFIG_SCHED_DEBUG
1220 /*
1221  * Print out various scheduling related per-task fields:
1222  */
1223 static int sched_show(struct seq_file *m, void *v)
1224 {
1225 	struct inode *inode = m->private;
1226 	struct task_struct *p;
1227 
1228 	p = get_proc_task(inode);
1229 	if (!p)
1230 		return -ESRCH;
1231 	proc_sched_show_task(p, m);
1232 
1233 	put_task_struct(p);
1234 
1235 	return 0;
1236 }
1237 
1238 static ssize_t
1239 sched_write(struct file *file, const char __user *buf,
1240 	    size_t count, loff_t *offset)
1241 {
1242 	struct inode *inode = file_inode(file);
1243 	struct task_struct *p;
1244 
1245 	p = get_proc_task(inode);
1246 	if (!p)
1247 		return -ESRCH;
1248 	proc_sched_set_task(p);
1249 
1250 	put_task_struct(p);
1251 
1252 	return count;
1253 }
1254 
1255 static int sched_open(struct inode *inode, struct file *filp)
1256 {
1257 	return single_open(filp, sched_show, inode);
1258 }
1259 
1260 static const struct file_operations proc_pid_sched_operations = {
1261 	.open		= sched_open,
1262 	.read		= seq_read,
1263 	.write		= sched_write,
1264 	.llseek		= seq_lseek,
1265 	.release	= single_release,
1266 };
1267 
1268 #endif
1269 
1270 #ifdef CONFIG_SCHED_AUTOGROUP
1271 /*
1272  * Print out autogroup related information:
1273  */
1274 static int sched_autogroup_show(struct seq_file *m, void *v)
1275 {
1276 	struct inode *inode = m->private;
1277 	struct task_struct *p;
1278 
1279 	p = get_proc_task(inode);
1280 	if (!p)
1281 		return -ESRCH;
1282 	proc_sched_autogroup_show_task(p, m);
1283 
1284 	put_task_struct(p);
1285 
1286 	return 0;
1287 }
1288 
1289 static ssize_t
1290 sched_autogroup_write(struct file *file, const char __user *buf,
1291 	    size_t count, loff_t *offset)
1292 {
1293 	struct inode *inode = file_inode(file);
1294 	struct task_struct *p;
1295 	char buffer[PROC_NUMBUF];
1296 	int nice;
1297 	int err;
1298 
1299 	memset(buffer, 0, sizeof(buffer));
1300 	if (count > sizeof(buffer) - 1)
1301 		count = sizeof(buffer) - 1;
1302 	if (copy_from_user(buffer, buf, count))
1303 		return -EFAULT;
1304 
1305 	err = kstrtoint(strstrip(buffer), 0, &nice);
1306 	if (err < 0)
1307 		return err;
1308 
1309 	p = get_proc_task(inode);
1310 	if (!p)
1311 		return -ESRCH;
1312 
1313 	err = proc_sched_autogroup_set_nice(p, nice);
1314 	if (err)
1315 		count = err;
1316 
1317 	put_task_struct(p);
1318 
1319 	return count;
1320 }
1321 
1322 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1323 {
1324 	int ret;
1325 
1326 	ret = single_open(filp, sched_autogroup_show, NULL);
1327 	if (!ret) {
1328 		struct seq_file *m = filp->private_data;
1329 
1330 		m->private = inode;
1331 	}
1332 	return ret;
1333 }
1334 
1335 static const struct file_operations proc_pid_sched_autogroup_operations = {
1336 	.open		= sched_autogroup_open,
1337 	.read		= seq_read,
1338 	.write		= sched_autogroup_write,
1339 	.llseek		= seq_lseek,
1340 	.release	= single_release,
1341 };
1342 
1343 #endif /* CONFIG_SCHED_AUTOGROUP */
1344 
1345 static ssize_t comm_write(struct file *file, const char __user *buf,
1346 				size_t count, loff_t *offset)
1347 {
1348 	struct inode *inode = file_inode(file);
1349 	struct task_struct *p;
1350 	char buffer[TASK_COMM_LEN];
1351 
1352 	memset(buffer, 0, sizeof(buffer));
1353 	if (count > sizeof(buffer) - 1)
1354 		count = sizeof(buffer) - 1;
1355 	if (copy_from_user(buffer, buf, count))
1356 		return -EFAULT;
1357 
1358 	p = get_proc_task(inode);
1359 	if (!p)
1360 		return -ESRCH;
1361 
1362 	if (same_thread_group(current, p))
1363 		set_task_comm(p, buffer);
1364 	else
1365 		count = -EINVAL;
1366 
1367 	put_task_struct(p);
1368 
1369 	return count;
1370 }
1371 
1372 static int comm_show(struct seq_file *m, void *v)
1373 {
1374 	struct inode *inode = m->private;
1375 	struct task_struct *p;
1376 
1377 	p = get_proc_task(inode);
1378 	if (!p)
1379 		return -ESRCH;
1380 
1381 	task_lock(p);
1382 	seq_printf(m, "%s\n", p->comm);
1383 	task_unlock(p);
1384 
1385 	put_task_struct(p);
1386 
1387 	return 0;
1388 }
1389 
1390 static int comm_open(struct inode *inode, struct file *filp)
1391 {
1392 	return single_open(filp, comm_show, inode);
1393 }
1394 
1395 static const struct file_operations proc_pid_set_comm_operations = {
1396 	.open		= comm_open,
1397 	.read		= seq_read,
1398 	.write		= comm_write,
1399 	.llseek		= seq_lseek,
1400 	.release	= single_release,
1401 };
1402 
1403 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1404 {
1405 	struct task_struct *task;
1406 	struct mm_struct *mm;
1407 	struct file *exe_file;
1408 
1409 	task = get_proc_task(dentry->d_inode);
1410 	if (!task)
1411 		return -ENOENT;
1412 	mm = get_task_mm(task);
1413 	put_task_struct(task);
1414 	if (!mm)
1415 		return -ENOENT;
1416 	exe_file = get_mm_exe_file(mm);
1417 	mmput(mm);
1418 	if (exe_file) {
1419 		*exe_path = exe_file->f_path;
1420 		path_get(&exe_file->f_path);
1421 		fput(exe_file);
1422 		return 0;
1423 	} else
1424 		return -ENOENT;
1425 }
1426 
1427 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1428 {
1429 	struct inode *inode = dentry->d_inode;
1430 	struct path path;
1431 	int error = -EACCES;
1432 
1433 	/* Are we allowed to snoop on the tasks file descriptors? */
1434 	if (!proc_fd_access_allowed(inode))
1435 		goto out;
1436 
1437 	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1438 	if (error)
1439 		goto out;
1440 
1441 	nd_jump_link(nd, &path);
1442 	return NULL;
1443 out:
1444 	return ERR_PTR(error);
1445 }
1446 
1447 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1448 {
1449 	char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1450 	char *pathname;
1451 	int len;
1452 
1453 	if (!tmp)
1454 		return -ENOMEM;
1455 
1456 	pathname = d_path(path, tmp, PAGE_SIZE);
1457 	len = PTR_ERR(pathname);
1458 	if (IS_ERR(pathname))
1459 		goto out;
1460 	len = tmp + PAGE_SIZE - 1 - pathname;
1461 
1462 	if (len > buflen)
1463 		len = buflen;
1464 	if (copy_to_user(buffer, pathname, len))
1465 		len = -EFAULT;
1466  out:
1467 	free_page((unsigned long)tmp);
1468 	return len;
1469 }
1470 
1471 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1472 {
1473 	int error = -EACCES;
1474 	struct inode *inode = dentry->d_inode;
1475 	struct path path;
1476 
1477 	/* Are we allowed to snoop on the tasks file descriptors? */
1478 	if (!proc_fd_access_allowed(inode))
1479 		goto out;
1480 
1481 	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1482 	if (error)
1483 		goto out;
1484 
1485 	error = do_proc_readlink(&path, buffer, buflen);
1486 	path_put(&path);
1487 out:
1488 	return error;
1489 }
1490 
1491 const struct inode_operations proc_pid_link_inode_operations = {
1492 	.readlink	= proc_pid_readlink,
1493 	.follow_link	= proc_pid_follow_link,
1494 	.setattr	= proc_setattr,
1495 };
1496 
1497 
1498 /* building an inode */
1499 
1500 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1501 {
1502 	struct inode * inode;
1503 	struct proc_inode *ei;
1504 	const struct cred *cred;
1505 
1506 	/* We need a new inode */
1507 
1508 	inode = new_inode(sb);
1509 	if (!inode)
1510 		goto out;
1511 
1512 	/* Common stuff */
1513 	ei = PROC_I(inode);
1514 	inode->i_ino = get_next_ino();
1515 	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1516 	inode->i_op = &proc_def_inode_operations;
1517 
1518 	/*
1519 	 * grab the reference to task.
1520 	 */
1521 	ei->pid = get_task_pid(task, PIDTYPE_PID);
1522 	if (!ei->pid)
1523 		goto out_unlock;
1524 
1525 	if (task_dumpable(task)) {
1526 		rcu_read_lock();
1527 		cred = __task_cred(task);
1528 		inode->i_uid = cred->euid;
1529 		inode->i_gid = cred->egid;
1530 		rcu_read_unlock();
1531 	}
1532 	security_task_to_inode(task, inode);
1533 
1534 out:
1535 	return inode;
1536 
1537 out_unlock:
1538 	iput(inode);
1539 	return NULL;
1540 }
1541 
1542 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1543 {
1544 	struct inode *inode = dentry->d_inode;
1545 	struct task_struct *task;
1546 	const struct cred *cred;
1547 	struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1548 
1549 	generic_fillattr(inode, stat);
1550 
1551 	rcu_read_lock();
1552 	stat->uid = GLOBAL_ROOT_UID;
1553 	stat->gid = GLOBAL_ROOT_GID;
1554 	task = pid_task(proc_pid(inode), PIDTYPE_PID);
1555 	if (task) {
1556 		if (!has_pid_permissions(pid, task, 2)) {
1557 			rcu_read_unlock();
1558 			/*
1559 			 * This doesn't prevent learning whether PID exists,
1560 			 * it only makes getattr() consistent with readdir().
1561 			 */
1562 			return -ENOENT;
1563 		}
1564 		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1565 		    task_dumpable(task)) {
1566 			cred = __task_cred(task);
1567 			stat->uid = cred->euid;
1568 			stat->gid = cred->egid;
1569 		}
1570 	}
1571 	rcu_read_unlock();
1572 	return 0;
1573 }
1574 
1575 /* dentry stuff */
1576 
1577 /*
1578  *	Exceptional case: normally we are not allowed to unhash a busy
1579  * directory. In this case, however, we can do it - no aliasing problems
1580  * due to the way we treat inodes.
1581  *
1582  * Rewrite the inode's ownerships here because the owning task may have
1583  * performed a setuid(), etc.
1584  *
1585  * Before the /proc/pid/status file was created the only way to read
1586  * the effective uid of a /process was to stat /proc/pid.  Reading
1587  * /proc/pid/status is slow enough that procps and other packages
1588  * kept stating /proc/pid.  To keep the rules in /proc simple I have
1589  * made this apply to all per process world readable and executable
1590  * directories.
1591  */
1592 int pid_revalidate(struct dentry *dentry, unsigned int flags)
1593 {
1594 	struct inode *inode;
1595 	struct task_struct *task;
1596 	const struct cred *cred;
1597 
1598 	if (flags & LOOKUP_RCU)
1599 		return -ECHILD;
1600 
1601 	inode = dentry->d_inode;
1602 	task = get_proc_task(inode);
1603 
1604 	if (task) {
1605 		if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1606 		    task_dumpable(task)) {
1607 			rcu_read_lock();
1608 			cred = __task_cred(task);
1609 			inode->i_uid = cred->euid;
1610 			inode->i_gid = cred->egid;
1611 			rcu_read_unlock();
1612 		} else {
1613 			inode->i_uid = GLOBAL_ROOT_UID;
1614 			inode->i_gid = GLOBAL_ROOT_GID;
1615 		}
1616 		inode->i_mode &= ~(S_ISUID | S_ISGID);
1617 		security_task_to_inode(task, inode);
1618 		put_task_struct(task);
1619 		return 1;
1620 	}
1621 	d_drop(dentry);
1622 	return 0;
1623 }
1624 
1625 const struct dentry_operations pid_dentry_operations =
1626 {
1627 	.d_revalidate	= pid_revalidate,
1628 	.d_delete	= pid_delete_dentry,
1629 };
1630 
1631 /* Lookups */
1632 
1633 /*
1634  * Fill a directory entry.
1635  *
1636  * If possible create the dcache entry and derive our inode number and
1637  * file type from dcache entry.
1638  *
1639  * Since all of the proc inode numbers are dynamically generated, the inode
1640  * numbers do not exist until the inode is cache.  This means creating the
1641  * the dcache entry in readdir is necessary to keep the inode numbers
1642  * reported by readdir in sync with the inode numbers reported
1643  * by stat.
1644  */
1645 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1646 	const char *name, int len,
1647 	instantiate_t instantiate, struct task_struct *task, const void *ptr)
1648 {
1649 	struct dentry *child, *dir = filp->f_path.dentry;
1650 	struct inode *inode;
1651 	struct qstr qname;
1652 	ino_t ino = 0;
1653 	unsigned type = DT_UNKNOWN;
1654 
1655 	qname.name = name;
1656 	qname.len  = len;
1657 	qname.hash = full_name_hash(name, len);
1658 
1659 	child = d_lookup(dir, &qname);
1660 	if (!child) {
1661 		struct dentry *new;
1662 		new = d_alloc(dir, &qname);
1663 		if (new) {
1664 			child = instantiate(dir->d_inode, new, task, ptr);
1665 			if (child)
1666 				dput(new);
1667 			else
1668 				child = new;
1669 		}
1670 	}
1671 	if (!child || IS_ERR(child) || !child->d_inode)
1672 		goto end_instantiate;
1673 	inode = child->d_inode;
1674 	if (inode) {
1675 		ino = inode->i_ino;
1676 		type = inode->i_mode >> 12;
1677 	}
1678 	dput(child);
1679 end_instantiate:
1680 	if (!ino)
1681 		ino = find_inode_number(dir, &qname);
1682 	if (!ino)
1683 		ino = 1;
1684 	return filldir(dirent, name, len, filp->f_pos, ino, type);
1685 }
1686 
1687 #ifdef CONFIG_CHECKPOINT_RESTORE
1688 
1689 /*
1690  * dname_to_vma_addr - maps a dentry name into two unsigned longs
1691  * which represent vma start and end addresses.
1692  */
1693 static int dname_to_vma_addr(struct dentry *dentry,
1694 			     unsigned long *start, unsigned long *end)
1695 {
1696 	if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1697 		return -EINVAL;
1698 
1699 	return 0;
1700 }
1701 
1702 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
1703 {
1704 	unsigned long vm_start, vm_end;
1705 	bool exact_vma_exists = false;
1706 	struct mm_struct *mm = NULL;
1707 	struct task_struct *task;
1708 	const struct cred *cred;
1709 	struct inode *inode;
1710 	int status = 0;
1711 
1712 	if (flags & LOOKUP_RCU)
1713 		return -ECHILD;
1714 
1715 	if (!capable(CAP_SYS_ADMIN)) {
1716 		status = -EPERM;
1717 		goto out_notask;
1718 	}
1719 
1720 	inode = dentry->d_inode;
1721 	task = get_proc_task(inode);
1722 	if (!task)
1723 		goto out_notask;
1724 
1725 	mm = mm_access(task, PTRACE_MODE_READ);
1726 	if (IS_ERR_OR_NULL(mm))
1727 		goto out;
1728 
1729 	if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1730 		down_read(&mm->mmap_sem);
1731 		exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1732 		up_read(&mm->mmap_sem);
1733 	}
1734 
1735 	mmput(mm);
1736 
1737 	if (exact_vma_exists) {
1738 		if (task_dumpable(task)) {
1739 			rcu_read_lock();
1740 			cred = __task_cred(task);
1741 			inode->i_uid = cred->euid;
1742 			inode->i_gid = cred->egid;
1743 			rcu_read_unlock();
1744 		} else {
1745 			inode->i_uid = GLOBAL_ROOT_UID;
1746 			inode->i_gid = GLOBAL_ROOT_GID;
1747 		}
1748 		security_task_to_inode(task, inode);
1749 		status = 1;
1750 	}
1751 
1752 out:
1753 	put_task_struct(task);
1754 
1755 out_notask:
1756 	if (status <= 0)
1757 		d_drop(dentry);
1758 
1759 	return status;
1760 }
1761 
1762 static const struct dentry_operations tid_map_files_dentry_operations = {
1763 	.d_revalidate	= map_files_d_revalidate,
1764 	.d_delete	= pid_delete_dentry,
1765 };
1766 
1767 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
1768 {
1769 	unsigned long vm_start, vm_end;
1770 	struct vm_area_struct *vma;
1771 	struct task_struct *task;
1772 	struct mm_struct *mm;
1773 	int rc;
1774 
1775 	rc = -ENOENT;
1776 	task = get_proc_task(dentry->d_inode);
1777 	if (!task)
1778 		goto out;
1779 
1780 	mm = get_task_mm(task);
1781 	put_task_struct(task);
1782 	if (!mm)
1783 		goto out;
1784 
1785 	rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1786 	if (rc)
1787 		goto out_mmput;
1788 
1789 	down_read(&mm->mmap_sem);
1790 	vma = find_exact_vma(mm, vm_start, vm_end);
1791 	if (vma && vma->vm_file) {
1792 		*path = vma->vm_file->f_path;
1793 		path_get(path);
1794 		rc = 0;
1795 	}
1796 	up_read(&mm->mmap_sem);
1797 
1798 out_mmput:
1799 	mmput(mm);
1800 out:
1801 	return rc;
1802 }
1803 
1804 struct map_files_info {
1805 	fmode_t		mode;
1806 	unsigned long	len;
1807 	unsigned char	name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
1808 };
1809 
1810 static struct dentry *
1811 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
1812 			   struct task_struct *task, const void *ptr)
1813 {
1814 	fmode_t mode = (fmode_t)(unsigned long)ptr;
1815 	struct proc_inode *ei;
1816 	struct inode *inode;
1817 
1818 	inode = proc_pid_make_inode(dir->i_sb, task);
1819 	if (!inode)
1820 		return ERR_PTR(-ENOENT);
1821 
1822 	ei = PROC_I(inode);
1823 	ei->op.proc_get_link = proc_map_files_get_link;
1824 
1825 	inode->i_op = &proc_pid_link_inode_operations;
1826 	inode->i_size = 64;
1827 	inode->i_mode = S_IFLNK;
1828 
1829 	if (mode & FMODE_READ)
1830 		inode->i_mode |= S_IRUSR;
1831 	if (mode & FMODE_WRITE)
1832 		inode->i_mode |= S_IWUSR;
1833 
1834 	d_set_d_op(dentry, &tid_map_files_dentry_operations);
1835 	d_add(dentry, inode);
1836 
1837 	return NULL;
1838 }
1839 
1840 static struct dentry *proc_map_files_lookup(struct inode *dir,
1841 		struct dentry *dentry, unsigned int flags)
1842 {
1843 	unsigned long vm_start, vm_end;
1844 	struct vm_area_struct *vma;
1845 	struct task_struct *task;
1846 	struct dentry *result;
1847 	struct mm_struct *mm;
1848 
1849 	result = ERR_PTR(-EPERM);
1850 	if (!capable(CAP_SYS_ADMIN))
1851 		goto out;
1852 
1853 	result = ERR_PTR(-ENOENT);
1854 	task = get_proc_task(dir);
1855 	if (!task)
1856 		goto out;
1857 
1858 	result = ERR_PTR(-EACCES);
1859 	if (!ptrace_may_access(task, PTRACE_MODE_READ))
1860 		goto out_put_task;
1861 
1862 	result = ERR_PTR(-ENOENT);
1863 	if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
1864 		goto out_put_task;
1865 
1866 	mm = get_task_mm(task);
1867 	if (!mm)
1868 		goto out_put_task;
1869 
1870 	down_read(&mm->mmap_sem);
1871 	vma = find_exact_vma(mm, vm_start, vm_end);
1872 	if (!vma)
1873 		goto out_no_vma;
1874 
1875 	if (vma->vm_file)
1876 		result = proc_map_files_instantiate(dir, dentry, task,
1877 				(void *)(unsigned long)vma->vm_file->f_mode);
1878 
1879 out_no_vma:
1880 	up_read(&mm->mmap_sem);
1881 	mmput(mm);
1882 out_put_task:
1883 	put_task_struct(task);
1884 out:
1885 	return result;
1886 }
1887 
1888 static const struct inode_operations proc_map_files_inode_operations = {
1889 	.lookup		= proc_map_files_lookup,
1890 	.permission	= proc_fd_permission,
1891 	.setattr	= proc_setattr,
1892 };
1893 
1894 static int
1895 proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
1896 {
1897 	struct dentry *dentry = filp->f_path.dentry;
1898 	struct inode *inode = dentry->d_inode;
1899 	struct vm_area_struct *vma;
1900 	struct task_struct *task;
1901 	struct mm_struct *mm;
1902 	ino_t ino;
1903 	int ret;
1904 
1905 	ret = -EPERM;
1906 	if (!capable(CAP_SYS_ADMIN))
1907 		goto out;
1908 
1909 	ret = -ENOENT;
1910 	task = get_proc_task(inode);
1911 	if (!task)
1912 		goto out;
1913 
1914 	ret = -EACCES;
1915 	if (!ptrace_may_access(task, PTRACE_MODE_READ))
1916 		goto out_put_task;
1917 
1918 	ret = 0;
1919 	switch (filp->f_pos) {
1920 	case 0:
1921 		ino = inode->i_ino;
1922 		if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
1923 			goto out_put_task;
1924 		filp->f_pos++;
1925 	case 1:
1926 		ino = parent_ino(dentry);
1927 		if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1928 			goto out_put_task;
1929 		filp->f_pos++;
1930 	default:
1931 	{
1932 		unsigned long nr_files, pos, i;
1933 		struct flex_array *fa = NULL;
1934 		struct map_files_info info;
1935 		struct map_files_info *p;
1936 
1937 		mm = get_task_mm(task);
1938 		if (!mm)
1939 			goto out_put_task;
1940 		down_read(&mm->mmap_sem);
1941 
1942 		nr_files = 0;
1943 
1944 		/*
1945 		 * We need two passes here:
1946 		 *
1947 		 *  1) Collect vmas of mapped files with mmap_sem taken
1948 		 *  2) Release mmap_sem and instantiate entries
1949 		 *
1950 		 * otherwise we get lockdep complained, since filldir()
1951 		 * routine might require mmap_sem taken in might_fault().
1952 		 */
1953 
1954 		for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
1955 			if (vma->vm_file && ++pos > filp->f_pos)
1956 				nr_files++;
1957 		}
1958 
1959 		if (nr_files) {
1960 			fa = flex_array_alloc(sizeof(info), nr_files,
1961 						GFP_KERNEL);
1962 			if (!fa || flex_array_prealloc(fa, 0, nr_files,
1963 							GFP_KERNEL)) {
1964 				ret = -ENOMEM;
1965 				if (fa)
1966 					flex_array_free(fa);
1967 				up_read(&mm->mmap_sem);
1968 				mmput(mm);
1969 				goto out_put_task;
1970 			}
1971 			for (i = 0, vma = mm->mmap, pos = 2; vma;
1972 					vma = vma->vm_next) {
1973 				if (!vma->vm_file)
1974 					continue;
1975 				if (++pos <= filp->f_pos)
1976 					continue;
1977 
1978 				info.mode = vma->vm_file->f_mode;
1979 				info.len = snprintf(info.name,
1980 						sizeof(info.name), "%lx-%lx",
1981 						vma->vm_start, vma->vm_end);
1982 				if (flex_array_put(fa, i++, &info, GFP_KERNEL))
1983 					BUG();
1984 			}
1985 		}
1986 		up_read(&mm->mmap_sem);
1987 
1988 		for (i = 0; i < nr_files; i++) {
1989 			p = flex_array_get(fa, i);
1990 			ret = proc_fill_cache(filp, dirent, filldir,
1991 					      p->name, p->len,
1992 					      proc_map_files_instantiate,
1993 					      task,
1994 					      (void *)(unsigned long)p->mode);
1995 			if (ret)
1996 				break;
1997 			filp->f_pos++;
1998 		}
1999 		if (fa)
2000 			flex_array_free(fa);
2001 		mmput(mm);
2002 	}
2003 	}
2004 
2005 out_put_task:
2006 	put_task_struct(task);
2007 out:
2008 	return ret;
2009 }
2010 
2011 static const struct file_operations proc_map_files_operations = {
2012 	.read		= generic_read_dir,
2013 	.readdir	= proc_map_files_readdir,
2014 	.llseek		= default_llseek,
2015 };
2016 
2017 struct timers_private {
2018 	struct pid *pid;
2019 	struct task_struct *task;
2020 	struct sighand_struct *sighand;
2021 	struct pid_namespace *ns;
2022 	unsigned long flags;
2023 };
2024 
2025 static void *timers_start(struct seq_file *m, loff_t *pos)
2026 {
2027 	struct timers_private *tp = m->private;
2028 
2029 	tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2030 	if (!tp->task)
2031 		return ERR_PTR(-ESRCH);
2032 
2033 	tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2034 	if (!tp->sighand)
2035 		return ERR_PTR(-ESRCH);
2036 
2037 	return seq_list_start(&tp->task->signal->posix_timers, *pos);
2038 }
2039 
2040 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2041 {
2042 	struct timers_private *tp = m->private;
2043 	return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2044 }
2045 
2046 static void timers_stop(struct seq_file *m, void *v)
2047 {
2048 	struct timers_private *tp = m->private;
2049 
2050 	if (tp->sighand) {
2051 		unlock_task_sighand(tp->task, &tp->flags);
2052 		tp->sighand = NULL;
2053 	}
2054 
2055 	if (tp->task) {
2056 		put_task_struct(tp->task);
2057 		tp->task = NULL;
2058 	}
2059 }
2060 
2061 static int show_timer(struct seq_file *m, void *v)
2062 {
2063 	struct k_itimer *timer;
2064 	struct timers_private *tp = m->private;
2065 	int notify;
2066 	static char *nstr[] = {
2067 		[SIGEV_SIGNAL] = "signal",
2068 		[SIGEV_NONE] = "none",
2069 		[SIGEV_THREAD] = "thread",
2070 	};
2071 
2072 	timer = list_entry((struct list_head *)v, struct k_itimer, list);
2073 	notify = timer->it_sigev_notify;
2074 
2075 	seq_printf(m, "ID: %d\n", timer->it_id);
2076 	seq_printf(m, "signal: %d/%p\n", timer->sigq->info.si_signo,
2077 			timer->sigq->info.si_value.sival_ptr);
2078 	seq_printf(m, "notify: %s/%s.%d\n",
2079 		nstr[notify & ~SIGEV_THREAD_ID],
2080 		(notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2081 		pid_nr_ns(timer->it_pid, tp->ns));
2082 
2083 	return 0;
2084 }
2085 
2086 static const struct seq_operations proc_timers_seq_ops = {
2087 	.start	= timers_start,
2088 	.next	= timers_next,
2089 	.stop	= timers_stop,
2090 	.show	= show_timer,
2091 };
2092 
2093 static int proc_timers_open(struct inode *inode, struct file *file)
2094 {
2095 	struct timers_private *tp;
2096 
2097 	tp = __seq_open_private(file, &proc_timers_seq_ops,
2098 			sizeof(struct timers_private));
2099 	if (!tp)
2100 		return -ENOMEM;
2101 
2102 	tp->pid = proc_pid(inode);
2103 	tp->ns = inode->i_sb->s_fs_info;
2104 	return 0;
2105 }
2106 
2107 static const struct file_operations proc_timers_operations = {
2108 	.open		= proc_timers_open,
2109 	.read		= seq_read,
2110 	.llseek		= seq_lseek,
2111 	.release	= seq_release_private,
2112 };
2113 #endif /* CONFIG_CHECKPOINT_RESTORE */
2114 
2115 static struct dentry *proc_pident_instantiate(struct inode *dir,
2116 	struct dentry *dentry, struct task_struct *task, const void *ptr)
2117 {
2118 	const struct pid_entry *p = ptr;
2119 	struct inode *inode;
2120 	struct proc_inode *ei;
2121 	struct dentry *error = ERR_PTR(-ENOENT);
2122 
2123 	inode = proc_pid_make_inode(dir->i_sb, task);
2124 	if (!inode)
2125 		goto out;
2126 
2127 	ei = PROC_I(inode);
2128 	inode->i_mode = p->mode;
2129 	if (S_ISDIR(inode->i_mode))
2130 		set_nlink(inode, 2);	/* Use getattr to fix if necessary */
2131 	if (p->iop)
2132 		inode->i_op = p->iop;
2133 	if (p->fop)
2134 		inode->i_fop = p->fop;
2135 	ei->op = p->op;
2136 	d_set_d_op(dentry, &pid_dentry_operations);
2137 	d_add(dentry, inode);
2138 	/* Close the race of the process dying before we return the dentry */
2139 	if (pid_revalidate(dentry, 0))
2140 		error = NULL;
2141 out:
2142 	return error;
2143 }
2144 
2145 static struct dentry *proc_pident_lookup(struct inode *dir,
2146 					 struct dentry *dentry,
2147 					 const struct pid_entry *ents,
2148 					 unsigned int nents)
2149 {
2150 	struct dentry *error;
2151 	struct task_struct *task = get_proc_task(dir);
2152 	const struct pid_entry *p, *last;
2153 
2154 	error = ERR_PTR(-ENOENT);
2155 
2156 	if (!task)
2157 		goto out_no_task;
2158 
2159 	/*
2160 	 * Yes, it does not scale. And it should not. Don't add
2161 	 * new entries into /proc/<tgid>/ without very good reasons.
2162 	 */
2163 	last = &ents[nents - 1];
2164 	for (p = ents; p <= last; p++) {
2165 		if (p->len != dentry->d_name.len)
2166 			continue;
2167 		if (!memcmp(dentry->d_name.name, p->name, p->len))
2168 			break;
2169 	}
2170 	if (p > last)
2171 		goto out;
2172 
2173 	error = proc_pident_instantiate(dir, dentry, task, p);
2174 out:
2175 	put_task_struct(task);
2176 out_no_task:
2177 	return error;
2178 }
2179 
2180 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2181 	filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2182 {
2183 	return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2184 				proc_pident_instantiate, task, p);
2185 }
2186 
2187 static int proc_pident_readdir(struct file *filp,
2188 		void *dirent, filldir_t filldir,
2189 		const struct pid_entry *ents, unsigned int nents)
2190 {
2191 	int i;
2192 	struct dentry *dentry = filp->f_path.dentry;
2193 	struct inode *inode = dentry->d_inode;
2194 	struct task_struct *task = get_proc_task(inode);
2195 	const struct pid_entry *p, *last;
2196 	ino_t ino;
2197 	int ret;
2198 
2199 	ret = -ENOENT;
2200 	if (!task)
2201 		goto out_no_task;
2202 
2203 	ret = 0;
2204 	i = filp->f_pos;
2205 	switch (i) {
2206 	case 0:
2207 		ino = inode->i_ino;
2208 		if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2209 			goto out;
2210 		i++;
2211 		filp->f_pos++;
2212 		/* fall through */
2213 	case 1:
2214 		ino = parent_ino(dentry);
2215 		if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2216 			goto out;
2217 		i++;
2218 		filp->f_pos++;
2219 		/* fall through */
2220 	default:
2221 		i -= 2;
2222 		if (i >= nents) {
2223 			ret = 1;
2224 			goto out;
2225 		}
2226 		p = ents + i;
2227 		last = &ents[nents - 1];
2228 		while (p <= last) {
2229 			if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2230 				goto out;
2231 			filp->f_pos++;
2232 			p++;
2233 		}
2234 	}
2235 
2236 	ret = 1;
2237 out:
2238 	put_task_struct(task);
2239 out_no_task:
2240 	return ret;
2241 }
2242 
2243 #ifdef CONFIG_SECURITY
2244 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2245 				  size_t count, loff_t *ppos)
2246 {
2247 	struct inode * inode = file_inode(file);
2248 	char *p = NULL;
2249 	ssize_t length;
2250 	struct task_struct *task = get_proc_task(inode);
2251 
2252 	if (!task)
2253 		return -ESRCH;
2254 
2255 	length = security_getprocattr(task,
2256 				      (char*)file->f_path.dentry->d_name.name,
2257 				      &p);
2258 	put_task_struct(task);
2259 	if (length > 0)
2260 		length = simple_read_from_buffer(buf, count, ppos, p, length);
2261 	kfree(p);
2262 	return length;
2263 }
2264 
2265 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2266 				   size_t count, loff_t *ppos)
2267 {
2268 	struct inode * inode = file_inode(file);
2269 	char *page;
2270 	ssize_t length;
2271 	struct task_struct *task = get_proc_task(inode);
2272 
2273 	length = -ESRCH;
2274 	if (!task)
2275 		goto out_no_task;
2276 	if (count > PAGE_SIZE)
2277 		count = PAGE_SIZE;
2278 
2279 	/* No partial writes. */
2280 	length = -EINVAL;
2281 	if (*ppos != 0)
2282 		goto out;
2283 
2284 	length = -ENOMEM;
2285 	page = (char*)__get_free_page(GFP_TEMPORARY);
2286 	if (!page)
2287 		goto out;
2288 
2289 	length = -EFAULT;
2290 	if (copy_from_user(page, buf, count))
2291 		goto out_free;
2292 
2293 	/* Guard against adverse ptrace interaction */
2294 	length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2295 	if (length < 0)
2296 		goto out_free;
2297 
2298 	length = security_setprocattr(task,
2299 				      (char*)file->f_path.dentry->d_name.name,
2300 				      (void*)page, count);
2301 	mutex_unlock(&task->signal->cred_guard_mutex);
2302 out_free:
2303 	free_page((unsigned long) page);
2304 out:
2305 	put_task_struct(task);
2306 out_no_task:
2307 	return length;
2308 }
2309 
2310 static const struct file_operations proc_pid_attr_operations = {
2311 	.read		= proc_pid_attr_read,
2312 	.write		= proc_pid_attr_write,
2313 	.llseek		= generic_file_llseek,
2314 };
2315 
2316 static const struct pid_entry attr_dir_stuff[] = {
2317 	REG("current",    S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2318 	REG("prev",       S_IRUGO,	   proc_pid_attr_operations),
2319 	REG("exec",       S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2320 	REG("fscreate",   S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2321 	REG("keycreate",  S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2322 	REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2323 };
2324 
2325 static int proc_attr_dir_readdir(struct file * filp,
2326 			     void * dirent, filldir_t filldir)
2327 {
2328 	return proc_pident_readdir(filp,dirent,filldir,
2329 				   attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2330 }
2331 
2332 static const struct file_operations proc_attr_dir_operations = {
2333 	.read		= generic_read_dir,
2334 	.readdir	= proc_attr_dir_readdir,
2335 	.llseek		= default_llseek,
2336 };
2337 
2338 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2339 				struct dentry *dentry, unsigned int flags)
2340 {
2341 	return proc_pident_lookup(dir, dentry,
2342 				  attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2343 }
2344 
2345 static const struct inode_operations proc_attr_dir_inode_operations = {
2346 	.lookup		= proc_attr_dir_lookup,
2347 	.getattr	= pid_getattr,
2348 	.setattr	= proc_setattr,
2349 };
2350 
2351 #endif
2352 
2353 #ifdef CONFIG_ELF_CORE
2354 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2355 					 size_t count, loff_t *ppos)
2356 {
2357 	struct task_struct *task = get_proc_task(file_inode(file));
2358 	struct mm_struct *mm;
2359 	char buffer[PROC_NUMBUF];
2360 	size_t len;
2361 	int ret;
2362 
2363 	if (!task)
2364 		return -ESRCH;
2365 
2366 	ret = 0;
2367 	mm = get_task_mm(task);
2368 	if (mm) {
2369 		len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2370 			       ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2371 				MMF_DUMP_FILTER_SHIFT));
2372 		mmput(mm);
2373 		ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2374 	}
2375 
2376 	put_task_struct(task);
2377 
2378 	return ret;
2379 }
2380 
2381 static ssize_t proc_coredump_filter_write(struct file *file,
2382 					  const char __user *buf,
2383 					  size_t count,
2384 					  loff_t *ppos)
2385 {
2386 	struct task_struct *task;
2387 	struct mm_struct *mm;
2388 	char buffer[PROC_NUMBUF], *end;
2389 	unsigned int val;
2390 	int ret;
2391 	int i;
2392 	unsigned long mask;
2393 
2394 	ret = -EFAULT;
2395 	memset(buffer, 0, sizeof(buffer));
2396 	if (count > sizeof(buffer) - 1)
2397 		count = sizeof(buffer) - 1;
2398 	if (copy_from_user(buffer, buf, count))
2399 		goto out_no_task;
2400 
2401 	ret = -EINVAL;
2402 	val = (unsigned int)simple_strtoul(buffer, &end, 0);
2403 	if (*end == '\n')
2404 		end++;
2405 	if (end - buffer == 0)
2406 		goto out_no_task;
2407 
2408 	ret = -ESRCH;
2409 	task = get_proc_task(file_inode(file));
2410 	if (!task)
2411 		goto out_no_task;
2412 
2413 	ret = end - buffer;
2414 	mm = get_task_mm(task);
2415 	if (!mm)
2416 		goto out_no_mm;
2417 
2418 	for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2419 		if (val & mask)
2420 			set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2421 		else
2422 			clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2423 	}
2424 
2425 	mmput(mm);
2426  out_no_mm:
2427 	put_task_struct(task);
2428  out_no_task:
2429 	return ret;
2430 }
2431 
2432 static const struct file_operations proc_coredump_filter_operations = {
2433 	.read		= proc_coredump_filter_read,
2434 	.write		= proc_coredump_filter_write,
2435 	.llseek		= generic_file_llseek,
2436 };
2437 #endif
2438 
2439 #ifdef CONFIG_TASK_IO_ACCOUNTING
2440 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2441 {
2442 	struct task_io_accounting acct = task->ioac;
2443 	unsigned long flags;
2444 	int result;
2445 
2446 	result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2447 	if (result)
2448 		return result;
2449 
2450 	if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2451 		result = -EACCES;
2452 		goto out_unlock;
2453 	}
2454 
2455 	if (whole && lock_task_sighand(task, &flags)) {
2456 		struct task_struct *t = task;
2457 
2458 		task_io_accounting_add(&acct, &task->signal->ioac);
2459 		while_each_thread(task, t)
2460 			task_io_accounting_add(&acct, &t->ioac);
2461 
2462 		unlock_task_sighand(task, &flags);
2463 	}
2464 	result = sprintf(buffer,
2465 			"rchar: %llu\n"
2466 			"wchar: %llu\n"
2467 			"syscr: %llu\n"
2468 			"syscw: %llu\n"
2469 			"read_bytes: %llu\n"
2470 			"write_bytes: %llu\n"
2471 			"cancelled_write_bytes: %llu\n",
2472 			(unsigned long long)acct.rchar,
2473 			(unsigned long long)acct.wchar,
2474 			(unsigned long long)acct.syscr,
2475 			(unsigned long long)acct.syscw,
2476 			(unsigned long long)acct.read_bytes,
2477 			(unsigned long long)acct.write_bytes,
2478 			(unsigned long long)acct.cancelled_write_bytes);
2479 out_unlock:
2480 	mutex_unlock(&task->signal->cred_guard_mutex);
2481 	return result;
2482 }
2483 
2484 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2485 {
2486 	return do_io_accounting(task, buffer, 0);
2487 }
2488 
2489 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2490 {
2491 	return do_io_accounting(task, buffer, 1);
2492 }
2493 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2494 
2495 #ifdef CONFIG_USER_NS
2496 static int proc_id_map_open(struct inode *inode, struct file *file,
2497 	struct seq_operations *seq_ops)
2498 {
2499 	struct user_namespace *ns = NULL;
2500 	struct task_struct *task;
2501 	struct seq_file *seq;
2502 	int ret = -EINVAL;
2503 
2504 	task = get_proc_task(inode);
2505 	if (task) {
2506 		rcu_read_lock();
2507 		ns = get_user_ns(task_cred_xxx(task, user_ns));
2508 		rcu_read_unlock();
2509 		put_task_struct(task);
2510 	}
2511 	if (!ns)
2512 		goto err;
2513 
2514 	ret = seq_open(file, seq_ops);
2515 	if (ret)
2516 		goto err_put_ns;
2517 
2518 	seq = file->private_data;
2519 	seq->private = ns;
2520 
2521 	return 0;
2522 err_put_ns:
2523 	put_user_ns(ns);
2524 err:
2525 	return ret;
2526 }
2527 
2528 static int proc_id_map_release(struct inode *inode, struct file *file)
2529 {
2530 	struct seq_file *seq = file->private_data;
2531 	struct user_namespace *ns = seq->private;
2532 	put_user_ns(ns);
2533 	return seq_release(inode, file);
2534 }
2535 
2536 static int proc_uid_map_open(struct inode *inode, struct file *file)
2537 {
2538 	return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2539 }
2540 
2541 static int proc_gid_map_open(struct inode *inode, struct file *file)
2542 {
2543 	return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2544 }
2545 
2546 static int proc_projid_map_open(struct inode *inode, struct file *file)
2547 {
2548 	return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2549 }
2550 
2551 static const struct file_operations proc_uid_map_operations = {
2552 	.open		= proc_uid_map_open,
2553 	.write		= proc_uid_map_write,
2554 	.read		= seq_read,
2555 	.llseek		= seq_lseek,
2556 	.release	= proc_id_map_release,
2557 };
2558 
2559 static const struct file_operations proc_gid_map_operations = {
2560 	.open		= proc_gid_map_open,
2561 	.write		= proc_gid_map_write,
2562 	.read		= seq_read,
2563 	.llseek		= seq_lseek,
2564 	.release	= proc_id_map_release,
2565 };
2566 
2567 static const struct file_operations proc_projid_map_operations = {
2568 	.open		= proc_projid_map_open,
2569 	.write		= proc_projid_map_write,
2570 	.read		= seq_read,
2571 	.llseek		= seq_lseek,
2572 	.release	= proc_id_map_release,
2573 };
2574 #endif /* CONFIG_USER_NS */
2575 
2576 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2577 				struct pid *pid, struct task_struct *task)
2578 {
2579 	int err = lock_trace(task);
2580 	if (!err) {
2581 		seq_printf(m, "%08x\n", task->personality);
2582 		unlock_trace(task);
2583 	}
2584 	return err;
2585 }
2586 
2587 /*
2588  * Thread groups
2589  */
2590 static const struct file_operations proc_task_operations;
2591 static const struct inode_operations proc_task_inode_operations;
2592 
2593 static const struct pid_entry tgid_base_stuff[] = {
2594 	DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2595 	DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2596 #ifdef CONFIG_CHECKPOINT_RESTORE
2597 	DIR("map_files",  S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2598 #endif
2599 	DIR("fdinfo",     S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2600 	DIR("ns",	  S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2601 #ifdef CONFIG_NET
2602 	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
2603 #endif
2604 	REG("environ",    S_IRUSR, proc_environ_operations),
2605 	INF("auxv",       S_IRUSR, proc_pid_auxv),
2606 	ONE("status",     S_IRUGO, proc_pid_status),
2607 	ONE("personality", S_IRUGO, proc_pid_personality),
2608 	INF("limits",	  S_IRUGO, proc_pid_limits),
2609 #ifdef CONFIG_SCHED_DEBUG
2610 	REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2611 #endif
2612 #ifdef CONFIG_SCHED_AUTOGROUP
2613 	REG("autogroup",  S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2614 #endif
2615 	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2616 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2617 	INF("syscall",    S_IRUGO, proc_pid_syscall),
2618 #endif
2619 	INF("cmdline",    S_IRUGO, proc_pid_cmdline),
2620 	ONE("stat",       S_IRUGO, proc_tgid_stat),
2621 	ONE("statm",      S_IRUGO, proc_pid_statm),
2622 	REG("maps",       S_IRUGO, proc_pid_maps_operations),
2623 #ifdef CONFIG_NUMA
2624 	REG("numa_maps",  S_IRUGO, proc_pid_numa_maps_operations),
2625 #endif
2626 	REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
2627 	LNK("cwd",        proc_cwd_link),
2628 	LNK("root",       proc_root_link),
2629 	LNK("exe",        proc_exe_link),
2630 	REG("mounts",     S_IRUGO, proc_mounts_operations),
2631 	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
2632 	REG("mountstats", S_IRUSR, proc_mountstats_operations),
2633 #ifdef CONFIG_PROC_PAGE_MONITOR
2634 	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2635 	REG("smaps",      S_IRUGO, proc_pid_smaps_operations),
2636 	REG("pagemap",    S_IRUGO, proc_pagemap_operations),
2637 #endif
2638 #ifdef CONFIG_SECURITY
2639 	DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2640 #endif
2641 #ifdef CONFIG_KALLSYMS
2642 	INF("wchan",      S_IRUGO, proc_pid_wchan),
2643 #endif
2644 #ifdef CONFIG_STACKTRACE
2645 	ONE("stack",      S_IRUGO, proc_pid_stack),
2646 #endif
2647 #ifdef CONFIG_SCHEDSTATS
2648 	INF("schedstat",  S_IRUGO, proc_pid_schedstat),
2649 #endif
2650 #ifdef CONFIG_LATENCYTOP
2651 	REG("latency",  S_IRUGO, proc_lstats_operations),
2652 #endif
2653 #ifdef CONFIG_PROC_PID_CPUSET
2654 	REG("cpuset",     S_IRUGO, proc_cpuset_operations),
2655 #endif
2656 #ifdef CONFIG_CGROUPS
2657 	REG("cgroup",  S_IRUGO, proc_cgroup_operations),
2658 #endif
2659 	INF("oom_score",  S_IRUGO, proc_oom_score),
2660 	REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adj_operations),
2661 	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
2662 #ifdef CONFIG_AUDITSYSCALL
2663 	REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
2664 	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
2665 #endif
2666 #ifdef CONFIG_FAULT_INJECTION
2667 	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
2668 #endif
2669 #ifdef CONFIG_ELF_CORE
2670 	REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
2671 #endif
2672 #ifdef CONFIG_TASK_IO_ACCOUNTING
2673 	INF("io",	S_IRUSR, proc_tgid_io_accounting),
2674 #endif
2675 #ifdef CONFIG_HARDWALL
2676 	INF("hardwall",   S_IRUGO, proc_pid_hardwall),
2677 #endif
2678 #ifdef CONFIG_USER_NS
2679 	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
2680 	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
2681 	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
2682 #endif
2683 #ifdef CONFIG_CHECKPOINT_RESTORE
2684 	REG("timers",	  S_IRUGO, proc_timers_operations),
2685 #endif
2686 };
2687 
2688 static int proc_tgid_base_readdir(struct file * filp,
2689 			     void * dirent, filldir_t filldir)
2690 {
2691 	return proc_pident_readdir(filp,dirent,filldir,
2692 				   tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2693 }
2694 
2695 static const struct file_operations proc_tgid_base_operations = {
2696 	.read		= generic_read_dir,
2697 	.readdir	= proc_tgid_base_readdir,
2698 	.llseek		= default_llseek,
2699 };
2700 
2701 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2702 {
2703 	return proc_pident_lookup(dir, dentry,
2704 				  tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
2705 }
2706 
2707 static const struct inode_operations proc_tgid_base_inode_operations = {
2708 	.lookup		= proc_tgid_base_lookup,
2709 	.getattr	= pid_getattr,
2710 	.setattr	= proc_setattr,
2711 	.permission	= proc_pid_permission,
2712 };
2713 
2714 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
2715 {
2716 	struct dentry *dentry, *leader, *dir;
2717 	char buf[PROC_NUMBUF];
2718 	struct qstr name;
2719 
2720 	name.name = buf;
2721 	name.len = snprintf(buf, sizeof(buf), "%d", pid);
2722 	/* no ->d_hash() rejects on procfs */
2723 	dentry = d_hash_and_lookup(mnt->mnt_root, &name);
2724 	if (dentry) {
2725 		shrink_dcache_parent(dentry);
2726 		d_drop(dentry);
2727 		dput(dentry);
2728 	}
2729 
2730 	name.name = buf;
2731 	name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2732 	leader = d_hash_and_lookup(mnt->mnt_root, &name);
2733 	if (!leader)
2734 		goto out;
2735 
2736 	name.name = "task";
2737 	name.len = strlen(name.name);
2738 	dir = d_hash_and_lookup(leader, &name);
2739 	if (!dir)
2740 		goto out_put_leader;
2741 
2742 	name.name = buf;
2743 	name.len = snprintf(buf, sizeof(buf), "%d", pid);
2744 	dentry = d_hash_and_lookup(dir, &name);
2745 	if (dentry) {
2746 		shrink_dcache_parent(dentry);
2747 		d_drop(dentry);
2748 		dput(dentry);
2749 	}
2750 
2751 	dput(dir);
2752 out_put_leader:
2753 	dput(leader);
2754 out:
2755 	return;
2756 }
2757 
2758 /**
2759  * proc_flush_task -  Remove dcache entries for @task from the /proc dcache.
2760  * @task: task that should be flushed.
2761  *
2762  * When flushing dentries from proc, one needs to flush them from global
2763  * proc (proc_mnt) and from all the namespaces' procs this task was seen
2764  * in. This call is supposed to do all of this job.
2765  *
2766  * Looks in the dcache for
2767  * /proc/@pid
2768  * /proc/@tgid/task/@pid
2769  * if either directory is present flushes it and all of it'ts children
2770  * from the dcache.
2771  *
2772  * It is safe and reasonable to cache /proc entries for a task until
2773  * that task exits.  After that they just clog up the dcache with
2774  * useless entries, possibly causing useful dcache entries to be
2775  * flushed instead.  This routine is proved to flush those useless
2776  * dcache entries at process exit time.
2777  *
2778  * NOTE: This routine is just an optimization so it does not guarantee
2779  *       that no dcache entries will exist at process exit time it
2780  *       just makes it very unlikely that any will persist.
2781  */
2782 
2783 void proc_flush_task(struct task_struct *task)
2784 {
2785 	int i;
2786 	struct pid *pid, *tgid;
2787 	struct upid *upid;
2788 
2789 	pid = task_pid(task);
2790 	tgid = task_tgid(task);
2791 
2792 	for (i = 0; i <= pid->level; i++) {
2793 		upid = &pid->numbers[i];
2794 		proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
2795 					tgid->numbers[i].nr);
2796 	}
2797 }
2798 
2799 static struct dentry *proc_pid_instantiate(struct inode *dir,
2800 					   struct dentry * dentry,
2801 					   struct task_struct *task, const void *ptr)
2802 {
2803 	struct dentry *error = ERR_PTR(-ENOENT);
2804 	struct inode *inode;
2805 
2806 	inode = proc_pid_make_inode(dir->i_sb, task);
2807 	if (!inode)
2808 		goto out;
2809 
2810 	inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2811 	inode->i_op = &proc_tgid_base_inode_operations;
2812 	inode->i_fop = &proc_tgid_base_operations;
2813 	inode->i_flags|=S_IMMUTABLE;
2814 
2815 	set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2816 						  ARRAY_SIZE(tgid_base_stuff)));
2817 
2818 	d_set_d_op(dentry, &pid_dentry_operations);
2819 
2820 	d_add(dentry, inode);
2821 	/* Close the race of the process dying before we return the dentry */
2822 	if (pid_revalidate(dentry, 0))
2823 		error = NULL;
2824 out:
2825 	return error;
2826 }
2827 
2828 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
2829 {
2830 	struct dentry *result = NULL;
2831 	struct task_struct *task;
2832 	unsigned tgid;
2833 	struct pid_namespace *ns;
2834 
2835 	tgid = name_to_int(dentry);
2836 	if (tgid == ~0U)
2837 		goto out;
2838 
2839 	ns = dentry->d_sb->s_fs_info;
2840 	rcu_read_lock();
2841 	task = find_task_by_pid_ns(tgid, ns);
2842 	if (task)
2843 		get_task_struct(task);
2844 	rcu_read_unlock();
2845 	if (!task)
2846 		goto out;
2847 
2848 	result = proc_pid_instantiate(dir, dentry, task, NULL);
2849 	put_task_struct(task);
2850 out:
2851 	return result;
2852 }
2853 
2854 /*
2855  * Find the first task with tgid >= tgid
2856  *
2857  */
2858 struct tgid_iter {
2859 	unsigned int tgid;
2860 	struct task_struct *task;
2861 };
2862 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2863 {
2864 	struct pid *pid;
2865 
2866 	if (iter.task)
2867 		put_task_struct(iter.task);
2868 	rcu_read_lock();
2869 retry:
2870 	iter.task = NULL;
2871 	pid = find_ge_pid(iter.tgid, ns);
2872 	if (pid) {
2873 		iter.tgid = pid_nr_ns(pid, ns);
2874 		iter.task = pid_task(pid, PIDTYPE_PID);
2875 		/* What we to know is if the pid we have find is the
2876 		 * pid of a thread_group_leader.  Testing for task
2877 		 * being a thread_group_leader is the obvious thing
2878 		 * todo but there is a window when it fails, due to
2879 		 * the pid transfer logic in de_thread.
2880 		 *
2881 		 * So we perform the straight forward test of seeing
2882 		 * if the pid we have found is the pid of a thread
2883 		 * group leader, and don't worry if the task we have
2884 		 * found doesn't happen to be a thread group leader.
2885 		 * As we don't care in the case of readdir.
2886 		 */
2887 		if (!iter.task || !has_group_leader_pid(iter.task)) {
2888 			iter.tgid += 1;
2889 			goto retry;
2890 		}
2891 		get_task_struct(iter.task);
2892 	}
2893 	rcu_read_unlock();
2894 	return iter;
2895 }
2896 
2897 #define TGID_OFFSET (FIRST_PROCESS_ENTRY)
2898 
2899 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2900 	struct tgid_iter iter)
2901 {
2902 	char name[PROC_NUMBUF];
2903 	int len = snprintf(name, sizeof(name), "%d", iter.tgid);
2904 	return proc_fill_cache(filp, dirent, filldir, name, len,
2905 				proc_pid_instantiate, iter.task, NULL);
2906 }
2907 
2908 static int fake_filldir(void *buf, const char *name, int namelen,
2909 			loff_t offset, u64 ino, unsigned d_type)
2910 {
2911 	return 0;
2912 }
2913 
2914 /* for the /proc/ directory itself, after non-process stuff has been done */
2915 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2916 {
2917 	struct tgid_iter iter;
2918 	struct pid_namespace *ns;
2919 	filldir_t __filldir;
2920 
2921 	if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
2922 		goto out;
2923 
2924 	ns = filp->f_dentry->d_sb->s_fs_info;
2925 	iter.task = NULL;
2926 	iter.tgid = filp->f_pos - TGID_OFFSET;
2927 	for (iter = next_tgid(ns, iter);
2928 	     iter.task;
2929 	     iter.tgid += 1, iter = next_tgid(ns, iter)) {
2930 		if (has_pid_permissions(ns, iter.task, 2))
2931 			__filldir = filldir;
2932 		else
2933 			__filldir = fake_filldir;
2934 
2935 		filp->f_pos = iter.tgid + TGID_OFFSET;
2936 		if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
2937 			put_task_struct(iter.task);
2938 			goto out;
2939 		}
2940 	}
2941 	filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2942 out:
2943 	return 0;
2944 }
2945 
2946 /*
2947  * Tasks
2948  */
2949 static const struct pid_entry tid_base_stuff[] = {
2950 	DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2951 	DIR("fdinfo",    S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
2952 	DIR("ns",	 S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
2953 	REG("environ",   S_IRUSR, proc_environ_operations),
2954 	INF("auxv",      S_IRUSR, proc_pid_auxv),
2955 	ONE("status",    S_IRUGO, proc_pid_status),
2956 	ONE("personality", S_IRUGO, proc_pid_personality),
2957 	INF("limits",	 S_IRUGO, proc_pid_limits),
2958 #ifdef CONFIG_SCHED_DEBUG
2959 	REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
2960 #endif
2961 	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
2962 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
2963 	INF("syscall",   S_IRUGO, proc_pid_syscall),
2964 #endif
2965 	INF("cmdline",   S_IRUGO, proc_pid_cmdline),
2966 	ONE("stat",      S_IRUGO, proc_tid_stat),
2967 	ONE("statm",     S_IRUGO, proc_pid_statm),
2968 	REG("maps",      S_IRUGO, proc_tid_maps_operations),
2969 #ifdef CONFIG_CHECKPOINT_RESTORE
2970 	REG("children",  S_IRUGO, proc_tid_children_operations),
2971 #endif
2972 #ifdef CONFIG_NUMA
2973 	REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
2974 #endif
2975 	REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
2976 	LNK("cwd",       proc_cwd_link),
2977 	LNK("root",      proc_root_link),
2978 	LNK("exe",       proc_exe_link),
2979 	REG("mounts",    S_IRUGO, proc_mounts_operations),
2980 	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
2981 #ifdef CONFIG_PROC_PAGE_MONITOR
2982 	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2983 	REG("smaps",     S_IRUGO, proc_tid_smaps_operations),
2984 	REG("pagemap",    S_IRUGO, proc_pagemap_operations),
2985 #endif
2986 #ifdef CONFIG_SECURITY
2987 	DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
2988 #endif
2989 #ifdef CONFIG_KALLSYMS
2990 	INF("wchan",     S_IRUGO, proc_pid_wchan),
2991 #endif
2992 #ifdef CONFIG_STACKTRACE
2993 	ONE("stack",      S_IRUGO, proc_pid_stack),
2994 #endif
2995 #ifdef CONFIG_SCHEDSTATS
2996 	INF("schedstat", S_IRUGO, proc_pid_schedstat),
2997 #endif
2998 #ifdef CONFIG_LATENCYTOP
2999 	REG("latency",  S_IRUGO, proc_lstats_operations),
3000 #endif
3001 #ifdef CONFIG_PROC_PID_CPUSET
3002 	REG("cpuset",    S_IRUGO, proc_cpuset_operations),
3003 #endif
3004 #ifdef CONFIG_CGROUPS
3005 	REG("cgroup",  S_IRUGO, proc_cgroup_operations),
3006 #endif
3007 	INF("oom_score", S_IRUGO, proc_oom_score),
3008 	REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3009 	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3010 #ifdef CONFIG_AUDITSYSCALL
3011 	REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
3012 	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3013 #endif
3014 #ifdef CONFIG_FAULT_INJECTION
3015 	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3016 #endif
3017 #ifdef CONFIG_TASK_IO_ACCOUNTING
3018 	INF("io",	S_IRUSR, proc_tid_io_accounting),
3019 #endif
3020 #ifdef CONFIG_HARDWALL
3021 	INF("hardwall",   S_IRUGO, proc_pid_hardwall),
3022 #endif
3023 #ifdef CONFIG_USER_NS
3024 	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3025 	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3026 	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3027 #endif
3028 };
3029 
3030 static int proc_tid_base_readdir(struct file * filp,
3031 			     void * dirent, filldir_t filldir)
3032 {
3033 	return proc_pident_readdir(filp,dirent,filldir,
3034 				   tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3035 }
3036 
3037 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3038 {
3039 	return proc_pident_lookup(dir, dentry,
3040 				  tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3041 }
3042 
3043 static const struct file_operations proc_tid_base_operations = {
3044 	.read		= generic_read_dir,
3045 	.readdir	= proc_tid_base_readdir,
3046 	.llseek		= default_llseek,
3047 };
3048 
3049 static const struct inode_operations proc_tid_base_inode_operations = {
3050 	.lookup		= proc_tid_base_lookup,
3051 	.getattr	= pid_getattr,
3052 	.setattr	= proc_setattr,
3053 };
3054 
3055 static struct dentry *proc_task_instantiate(struct inode *dir,
3056 	struct dentry *dentry, struct task_struct *task, const void *ptr)
3057 {
3058 	struct dentry *error = ERR_PTR(-ENOENT);
3059 	struct inode *inode;
3060 	inode = proc_pid_make_inode(dir->i_sb, task);
3061 
3062 	if (!inode)
3063 		goto out;
3064 	inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3065 	inode->i_op = &proc_tid_base_inode_operations;
3066 	inode->i_fop = &proc_tid_base_operations;
3067 	inode->i_flags|=S_IMMUTABLE;
3068 
3069 	set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3070 						  ARRAY_SIZE(tid_base_stuff)));
3071 
3072 	d_set_d_op(dentry, &pid_dentry_operations);
3073 
3074 	d_add(dentry, inode);
3075 	/* Close the race of the process dying before we return the dentry */
3076 	if (pid_revalidate(dentry, 0))
3077 		error = NULL;
3078 out:
3079 	return error;
3080 }
3081 
3082 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3083 {
3084 	struct dentry *result = ERR_PTR(-ENOENT);
3085 	struct task_struct *task;
3086 	struct task_struct *leader = get_proc_task(dir);
3087 	unsigned tid;
3088 	struct pid_namespace *ns;
3089 
3090 	if (!leader)
3091 		goto out_no_task;
3092 
3093 	tid = name_to_int(dentry);
3094 	if (tid == ~0U)
3095 		goto out;
3096 
3097 	ns = dentry->d_sb->s_fs_info;
3098 	rcu_read_lock();
3099 	task = find_task_by_pid_ns(tid, ns);
3100 	if (task)
3101 		get_task_struct(task);
3102 	rcu_read_unlock();
3103 	if (!task)
3104 		goto out;
3105 	if (!same_thread_group(leader, task))
3106 		goto out_drop_task;
3107 
3108 	result = proc_task_instantiate(dir, dentry, task, NULL);
3109 out_drop_task:
3110 	put_task_struct(task);
3111 out:
3112 	put_task_struct(leader);
3113 out_no_task:
3114 	return result;
3115 }
3116 
3117 /*
3118  * Find the first tid of a thread group to return to user space.
3119  *
3120  * Usually this is just the thread group leader, but if the users
3121  * buffer was too small or there was a seek into the middle of the
3122  * directory we have more work todo.
3123  *
3124  * In the case of a short read we start with find_task_by_pid.
3125  *
3126  * In the case of a seek we start with the leader and walk nr
3127  * threads past it.
3128  */
3129 static struct task_struct *first_tid(struct task_struct *leader,
3130 		int tid, int nr, struct pid_namespace *ns)
3131 {
3132 	struct task_struct *pos;
3133 
3134 	rcu_read_lock();
3135 	/* Attempt to start with the pid of a thread */
3136 	if (tid && (nr > 0)) {
3137 		pos = find_task_by_pid_ns(tid, ns);
3138 		if (pos && (pos->group_leader == leader))
3139 			goto found;
3140 	}
3141 
3142 	/* If nr exceeds the number of threads there is nothing todo */
3143 	pos = NULL;
3144 	if (nr && nr >= get_nr_threads(leader))
3145 		goto out;
3146 
3147 	/* If we haven't found our starting place yet start
3148 	 * with the leader and walk nr threads forward.
3149 	 */
3150 	for (pos = leader; nr > 0; --nr) {
3151 		pos = next_thread(pos);
3152 		if (pos == leader) {
3153 			pos = NULL;
3154 			goto out;
3155 		}
3156 	}
3157 found:
3158 	get_task_struct(pos);
3159 out:
3160 	rcu_read_unlock();
3161 	return pos;
3162 }
3163 
3164 /*
3165  * Find the next thread in the thread list.
3166  * Return NULL if there is an error or no next thread.
3167  *
3168  * The reference to the input task_struct is released.
3169  */
3170 static struct task_struct *next_tid(struct task_struct *start)
3171 {
3172 	struct task_struct *pos = NULL;
3173 	rcu_read_lock();
3174 	if (pid_alive(start)) {
3175 		pos = next_thread(start);
3176 		if (thread_group_leader(pos))
3177 			pos = NULL;
3178 		else
3179 			get_task_struct(pos);
3180 	}
3181 	rcu_read_unlock();
3182 	put_task_struct(start);
3183 	return pos;
3184 }
3185 
3186 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3187 	struct task_struct *task, int tid)
3188 {
3189 	char name[PROC_NUMBUF];
3190 	int len = snprintf(name, sizeof(name), "%d", tid);
3191 	return proc_fill_cache(filp, dirent, filldir, name, len,
3192 				proc_task_instantiate, task, NULL);
3193 }
3194 
3195 /* for the /proc/TGID/task/ directories */
3196 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3197 {
3198 	struct dentry *dentry = filp->f_path.dentry;
3199 	struct inode *inode = dentry->d_inode;
3200 	struct task_struct *leader = NULL;
3201 	struct task_struct *task;
3202 	int retval = -ENOENT;
3203 	ino_t ino;
3204 	int tid;
3205 	struct pid_namespace *ns;
3206 
3207 	task = get_proc_task(inode);
3208 	if (!task)
3209 		goto out_no_task;
3210 	rcu_read_lock();
3211 	if (pid_alive(task)) {
3212 		leader = task->group_leader;
3213 		get_task_struct(leader);
3214 	}
3215 	rcu_read_unlock();
3216 	put_task_struct(task);
3217 	if (!leader)
3218 		goto out_no_task;
3219 	retval = 0;
3220 
3221 	switch ((unsigned long)filp->f_pos) {
3222 	case 0:
3223 		ino = inode->i_ino;
3224 		if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3225 			goto out;
3226 		filp->f_pos++;
3227 		/* fall through */
3228 	case 1:
3229 		ino = parent_ino(dentry);
3230 		if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3231 			goto out;
3232 		filp->f_pos++;
3233 		/* fall through */
3234 	}
3235 
3236 	/* f_version caches the tgid value that the last readdir call couldn't
3237 	 * return. lseek aka telldir automagically resets f_version to 0.
3238 	 */
3239 	ns = filp->f_dentry->d_sb->s_fs_info;
3240 	tid = (int)filp->f_version;
3241 	filp->f_version = 0;
3242 	for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3243 	     task;
3244 	     task = next_tid(task), filp->f_pos++) {
3245 		tid = task_pid_nr_ns(task, ns);
3246 		if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3247 			/* returning this tgid failed, save it as the first
3248 			 * pid for the next readir call */
3249 			filp->f_version = (u64)tid;
3250 			put_task_struct(task);
3251 			break;
3252 		}
3253 	}
3254 out:
3255 	put_task_struct(leader);
3256 out_no_task:
3257 	return retval;
3258 }
3259 
3260 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3261 {
3262 	struct inode *inode = dentry->d_inode;
3263 	struct task_struct *p = get_proc_task(inode);
3264 	generic_fillattr(inode, stat);
3265 
3266 	if (p) {
3267 		stat->nlink += get_nr_threads(p);
3268 		put_task_struct(p);
3269 	}
3270 
3271 	return 0;
3272 }
3273 
3274 static const struct inode_operations proc_task_inode_operations = {
3275 	.lookup		= proc_task_lookup,
3276 	.getattr	= proc_task_getattr,
3277 	.setattr	= proc_setattr,
3278 	.permission	= proc_pid_permission,
3279 };
3280 
3281 static const struct file_operations proc_task_operations = {
3282 	.read		= generic_read_dir,
3283 	.readdir	= proc_task_readdir,
3284 	.llseek		= default_llseek,
3285 };
3286