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