xref: /kernel/linux/linux-6.6/fs/proc/base.c (revision 62306a36)
1// SPDX-License-Identifier: GPL-2.0
2/*
3 *  linux/fs/proc/base.c
4 *
5 *  Copyright (C) 1991, 1992 Linus Torvalds
6 *
7 *  proc base directory handling functions
8 *
9 *  1999, Al Viro. Rewritten. Now it covers the whole per-process part.
10 *  Instead of using magical inumbers to determine the kind of object
11 *  we allocate and fill in-core inodes upon lookup. They don't even
12 *  go into icache. We cache the reference to task_struct upon lookup too.
13 *  Eventually it should become a filesystem in its own. We don't use the
14 *  rest of procfs anymore.
15 *
16 *
17 *  Changelog:
18 *  17-Jan-2005
19 *  Allan Bezerra
20 *  Bruna Moreira <bruna.moreira@indt.org.br>
21 *  Edjard Mota <edjard.mota@indt.org.br>
22 *  Ilias Biris <ilias.biris@indt.org.br>
23 *  Mauricio Lin <mauricio.lin@indt.org.br>
24 *
25 *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 *
27 *  A new process specific entry (smaps) included in /proc. It shows the
28 *  size of rss for each memory area. The maps entry lacks information
29 *  about physical memory size (rss) for each mapped file, i.e.,
30 *  rss information for executables and library files.
31 *  This additional information is useful for any tools that need to know
32 *  about physical memory consumption for a process specific library.
33 *
34 *  Changelog:
35 *  21-Feb-2005
36 *  Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
37 *  Pud inclusion in the page table walking.
38 *
39 *  ChangeLog:
40 *  10-Mar-2005
41 *  10LE Instituto Nokia de Tecnologia - INdT:
42 *  A better way to walks through the page table as suggested by Hugh Dickins.
43 *
44 *  Simo Piiroinen <simo.piiroinen@nokia.com>:
45 *  Smaps information related to shared, private, clean and dirty pages.
46 *
47 *  Paul Mundt <paul.mundt@nokia.com>:
48 *  Overall revision about smaps.
49 */
50
51#include <linux/uaccess.h>
52
53#include <linux/errno.h>
54#include <linux/time.h>
55#include <linux/proc_fs.h>
56#include <linux/stat.h>
57
58#ifdef CONFIG_QOS_CTRL
59#include <linux/sched/qos_ctrl.h>
60#endif
61
62#include <linux/task_io_accounting_ops.h>
63#include <linux/init.h>
64#include <linux/capability.h>
65#include <linux/file.h>
66#include <linux/fdtable.h>
67#include <linux/generic-radix-tree.h>
68#include <linux/string.h>
69#include <linux/seq_file.h>
70#include <linux/namei.h>
71#include <linux/mnt_namespace.h>
72#include <linux/mm.h>
73#include <linux/swap.h>
74#include <linux/rcupdate.h>
75#include <linux/kallsyms.h>
76#include <linux/stacktrace.h>
77#include <linux/resource.h>
78#include <linux/module.h>
79#include <linux/mount.h>
80#include <linux/security.h>
81#include <linux/ptrace.h>
82#include <linux/printk.h>
83#include <linux/cache.h>
84#include <linux/cgroup.h>
85#include <linux/cpuset.h>
86#include <linux/audit.h>
87#include <linux/poll.h>
88#include <linux/nsproxy.h>
89#include <linux/oom.h>
90#include <linux/elf.h>
91#include <linux/pid_namespace.h>
92#include <linux/user_namespace.h>
93#include <linux/fs_struct.h>
94#include <linux/slab.h>
95#include <linux/sched.h>
96#ifdef CONFIG_SCHED_RTG
97#include <linux/sched/rtg_ctrl.h>
98#endif
99#include <linux/sched/autogroup.h>
100#include <linux/sched/mm.h>
101#include <linux/sched/coredump.h>
102#include <linux/sched/debug.h>
103#include <linux/sched/stat.h>
104#include <linux/posix-timers.h>
105#include <linux/time_namespace.h>
106#include <linux/resctrl.h>
107#include <linux/cn_proc.h>
108#include <linux/ksm.h>
109#include <trace/events/oom.h>
110#ifdef CONFIG_SCHED_RTG
111#include <linux/sched/rtg.h>
112#endif
113#include "internal.h"
114#include "fd.h"
115
116#include "../../lib/kstrtox.h"
117
118/* NOTE:
119 *	Implementing inode permission operations in /proc is almost
120 *	certainly an error.  Permission checks need to happen during
121 *	each system call not at open time.  The reason is that most of
122 *	what we wish to check for permissions in /proc varies at runtime.
123 *
124 *	The classic example of a problem is opening file descriptors
125 *	in /proc for a task before it execs a suid executable.
126 */
127
128static u8 nlink_tid __ro_after_init;
129static u8 nlink_tgid __ro_after_init;
130
131struct pid_entry {
132	const char *name;
133	unsigned int len;
134	umode_t mode;
135	const struct inode_operations *iop;
136	const struct file_operations *fop;
137	union proc_op op;
138};
139
140#define NOD(NAME, MODE, IOP, FOP, OP) {			\
141	.name = (NAME),					\
142	.len  = sizeof(NAME) - 1,			\
143	.mode = MODE,					\
144	.iop  = IOP,					\
145	.fop  = FOP,					\
146	.op   = OP,					\
147}
148
149#define DIR(NAME, MODE, iops, fops)	\
150	NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
151#define LNK(NAME, get_link)					\
152	NOD(NAME, (S_IFLNK|S_IRWXUGO),				\
153		&proc_pid_link_inode_operations, NULL,		\
154		{ .proc_get_link = get_link } )
155#define REG(NAME, MODE, fops)				\
156	NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
157#define ONE(NAME, MODE, show)				\
158	NOD(NAME, (S_IFREG|(MODE)),			\
159		NULL, &proc_single_file_operations,	\
160		{ .proc_show = show } )
161#define ATTR(LSM, NAME, MODE)				\
162	NOD(NAME, (S_IFREG|(MODE)),			\
163		NULL, &proc_pid_attr_operations,	\
164		{ .lsm = LSM })
165
166/*
167 * Count the number of hardlinks for the pid_entry table, excluding the .
168 * and .. links.
169 */
170static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
171	unsigned int n)
172{
173	unsigned int i;
174	unsigned int count;
175
176	count = 2;
177	for (i = 0; i < n; ++i) {
178		if (S_ISDIR(entries[i].mode))
179			++count;
180	}
181
182	return count;
183}
184
185static int get_task_root(struct task_struct *task, struct path *root)
186{
187	int result = -ENOENT;
188
189	task_lock(task);
190	if (task->fs) {
191		get_fs_root(task->fs, root);
192		result = 0;
193	}
194	task_unlock(task);
195	return result;
196}
197
198static int proc_cwd_link(struct dentry *dentry, struct path *path)
199{
200	struct task_struct *task = get_proc_task(d_inode(dentry));
201	int result = -ENOENT;
202
203	if (task) {
204		task_lock(task);
205		if (task->fs) {
206			get_fs_pwd(task->fs, path);
207			result = 0;
208		}
209		task_unlock(task);
210		put_task_struct(task);
211	}
212	return result;
213}
214
215static int proc_root_link(struct dentry *dentry, struct path *path)
216{
217	struct task_struct *task = get_proc_task(d_inode(dentry));
218	int result = -ENOENT;
219
220	if (task) {
221		result = get_task_root(task, path);
222		put_task_struct(task);
223	}
224	return result;
225}
226
227/*
228 * If the user used setproctitle(), we just get the string from
229 * user space at arg_start, and limit it to a maximum of one page.
230 */
231static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf,
232				size_t count, unsigned long pos,
233				unsigned long arg_start)
234{
235	char *page;
236	int ret, got;
237
238	if (pos >= PAGE_SIZE)
239		return 0;
240
241	page = (char *)__get_free_page(GFP_KERNEL);
242	if (!page)
243		return -ENOMEM;
244
245	ret = 0;
246	got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON);
247	if (got > 0) {
248		int len = strnlen(page, got);
249
250		/* Include the NUL character if it was found */
251		if (len < got)
252			len++;
253
254		if (len > pos) {
255			len -= pos;
256			if (len > count)
257				len = count;
258			len -= copy_to_user(buf, page+pos, len);
259			if (!len)
260				len = -EFAULT;
261			ret = len;
262		}
263	}
264	free_page((unsigned long)page);
265	return ret;
266}
267
268static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf,
269			      size_t count, loff_t *ppos)
270{
271	unsigned long arg_start, arg_end, env_start, env_end;
272	unsigned long pos, len;
273	char *page, c;
274
275	/* Check if process spawned far enough to have cmdline. */
276	if (!mm->env_end)
277		return 0;
278
279	spin_lock(&mm->arg_lock);
280	arg_start = mm->arg_start;
281	arg_end = mm->arg_end;
282	env_start = mm->env_start;
283	env_end = mm->env_end;
284	spin_unlock(&mm->arg_lock);
285
286	if (arg_start >= arg_end)
287		return 0;
288
289	/*
290	 * We allow setproctitle() to overwrite the argument
291	 * strings, and overflow past the original end. But
292	 * only when it overflows into the environment area.
293	 */
294	if (env_start != arg_end || env_end < env_start)
295		env_start = env_end = arg_end;
296	len = env_end - arg_start;
297
298	/* We're not going to care if "*ppos" has high bits set */
299	pos = *ppos;
300	if (pos >= len)
301		return 0;
302	if (count > len - pos)
303		count = len - pos;
304	if (!count)
305		return 0;
306
307	/*
308	 * Magical special case: if the argv[] end byte is not
309	 * zero, the user has overwritten it with setproctitle(3).
310	 *
311	 * Possible future enhancement: do this only once when
312	 * pos is 0, and set a flag in the 'struct file'.
313	 */
314	if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c)
315		return get_mm_proctitle(mm, buf, count, pos, arg_start);
316
317	/*
318	 * For the non-setproctitle() case we limit things strictly
319	 * to the [arg_start, arg_end[ range.
320	 */
321	pos += arg_start;
322	if (pos < arg_start || pos >= arg_end)
323		return 0;
324	if (count > arg_end - pos)
325		count = arg_end - pos;
326
327	page = (char *)__get_free_page(GFP_KERNEL);
328	if (!page)
329		return -ENOMEM;
330
331	len = 0;
332	while (count) {
333		int got;
334		size_t size = min_t(size_t, PAGE_SIZE, count);
335
336		got = access_remote_vm(mm, pos, page, size, FOLL_ANON);
337		if (got <= 0)
338			break;
339		got -= copy_to_user(buf, page, got);
340		if (unlikely(!got)) {
341			if (!len)
342				len = -EFAULT;
343			break;
344		}
345		pos += got;
346		buf += got;
347		len += got;
348		count -= got;
349	}
350
351	free_page((unsigned long)page);
352	return len;
353}
354
355static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf,
356				size_t count, loff_t *pos)
357{
358	struct mm_struct *mm;
359	ssize_t ret;
360
361	mm = get_task_mm(tsk);
362	if (!mm)
363		return 0;
364
365	ret = get_mm_cmdline(mm, buf, count, pos);
366	mmput(mm);
367	return ret;
368}
369
370static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
371				     size_t count, loff_t *pos)
372{
373	struct task_struct *tsk;
374	ssize_t ret;
375
376	BUG_ON(*pos < 0);
377
378	tsk = get_proc_task(file_inode(file));
379	if (!tsk)
380		return -ESRCH;
381	ret = get_task_cmdline(tsk, buf, count, pos);
382	put_task_struct(tsk);
383	if (ret > 0)
384		*pos += ret;
385	return ret;
386}
387
388static const struct file_operations proc_pid_cmdline_ops = {
389	.read	= proc_pid_cmdline_read,
390	.llseek	= generic_file_llseek,
391};
392
393#ifdef CONFIG_KALLSYMS
394/*
395 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
396 * Returns the resolved symbol.  If that fails, simply return the address.
397 */
398static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
399			  struct pid *pid, struct task_struct *task)
400{
401	unsigned long wchan;
402	char symname[KSYM_NAME_LEN];
403
404	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
405		goto print0;
406
407	wchan = get_wchan(task);
408	if (wchan && !lookup_symbol_name(wchan, symname)) {
409		seq_puts(m, symname);
410		return 0;
411	}
412
413print0:
414	seq_putc(m, '0');
415	return 0;
416}
417#endif /* CONFIG_KALLSYMS */
418
419static int lock_trace(struct task_struct *task)
420{
421	int err = down_read_killable(&task->signal->exec_update_lock);
422	if (err)
423		return err;
424	if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
425		up_read(&task->signal->exec_update_lock);
426		return -EPERM;
427	}
428	return 0;
429}
430
431static void unlock_trace(struct task_struct *task)
432{
433	up_read(&task->signal->exec_update_lock);
434}
435
436#ifdef CONFIG_STACKTRACE
437
438#define MAX_STACK_TRACE_DEPTH	64
439
440static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
441			  struct pid *pid, struct task_struct *task)
442{
443	unsigned long *entries;
444	int err;
445
446	/*
447	 * The ability to racily run the kernel stack unwinder on a running task
448	 * and then observe the unwinder output is scary; while it is useful for
449	 * debugging kernel issues, it can also allow an attacker to leak kernel
450	 * stack contents.
451	 * Doing this in a manner that is at least safe from races would require
452	 * some work to ensure that the remote task can not be scheduled; and
453	 * even then, this would still expose the unwinder as local attack
454	 * surface.
455	 * Therefore, this interface is restricted to root.
456	 */
457	if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN))
458		return -EACCES;
459
460	entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries),
461				GFP_KERNEL);
462	if (!entries)
463		return -ENOMEM;
464
465	err = lock_trace(task);
466	if (!err) {
467		unsigned int i, nr_entries;
468
469		nr_entries = stack_trace_save_tsk(task, entries,
470						  MAX_STACK_TRACE_DEPTH, 0);
471
472		for (i = 0; i < nr_entries; i++) {
473			seq_printf(m, "[<0>] %pB\n", (void *)entries[i]);
474		}
475
476		unlock_trace(task);
477	}
478	kfree(entries);
479
480	return err;
481}
482#endif
483
484#ifdef CONFIG_SCHED_INFO
485/*
486 * Provides /proc/PID/schedstat
487 */
488static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
489			      struct pid *pid, struct task_struct *task)
490{
491	if (unlikely(!sched_info_on()))
492		seq_puts(m, "0 0 0\n");
493	else
494		seq_printf(m, "%llu %llu %lu\n",
495		   (unsigned long long)task->se.sum_exec_runtime,
496		   (unsigned long long)task->sched_info.run_delay,
497		   task->sched_info.pcount);
498
499	return 0;
500}
501#endif
502
503#ifdef CONFIG_LATENCYTOP
504static int lstats_show_proc(struct seq_file *m, void *v)
505{
506	int i;
507	struct inode *inode = m->private;
508	struct task_struct *task = get_proc_task(inode);
509
510	if (!task)
511		return -ESRCH;
512	seq_puts(m, "Latency Top version : v0.1\n");
513	for (i = 0; i < LT_SAVECOUNT; i++) {
514		struct latency_record *lr = &task->latency_record[i];
515		if (lr->backtrace[0]) {
516			int q;
517			seq_printf(m, "%i %li %li",
518				   lr->count, lr->time, lr->max);
519			for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
520				unsigned long bt = lr->backtrace[q];
521
522				if (!bt)
523					break;
524				seq_printf(m, " %ps", (void *)bt);
525			}
526			seq_putc(m, '\n');
527		}
528
529	}
530	put_task_struct(task);
531	return 0;
532}
533
534static int lstats_open(struct inode *inode, struct file *file)
535{
536	return single_open(file, lstats_show_proc, inode);
537}
538
539static ssize_t lstats_write(struct file *file, const char __user *buf,
540			    size_t count, loff_t *offs)
541{
542	struct task_struct *task = get_proc_task(file_inode(file));
543
544	if (!task)
545		return -ESRCH;
546	clear_tsk_latency_tracing(task);
547	put_task_struct(task);
548
549	return count;
550}
551
552static const struct file_operations proc_lstats_operations = {
553	.open		= lstats_open,
554	.read		= seq_read,
555	.write		= lstats_write,
556	.llseek		= seq_lseek,
557	.release	= single_release,
558};
559
560#endif
561
562static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
563			  struct pid *pid, struct task_struct *task)
564{
565	unsigned long totalpages = totalram_pages() + total_swap_pages;
566	unsigned long points = 0;
567	long badness;
568
569	badness = oom_badness(task, totalpages);
570	/*
571	 * Special case OOM_SCORE_ADJ_MIN for all others scale the
572	 * badness value into [0, 2000] range which we have been
573	 * exporting for a long time so userspace might depend on it.
574	 */
575	if (badness != LONG_MIN)
576		points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3;
577
578	seq_printf(m, "%lu\n", points);
579
580	return 0;
581}
582
583struct limit_names {
584	const char *name;
585	const char *unit;
586};
587
588static const struct limit_names lnames[RLIM_NLIMITS] = {
589	[RLIMIT_CPU] = {"Max cpu time", "seconds"},
590	[RLIMIT_FSIZE] = {"Max file size", "bytes"},
591	[RLIMIT_DATA] = {"Max data size", "bytes"},
592	[RLIMIT_STACK] = {"Max stack size", "bytes"},
593	[RLIMIT_CORE] = {"Max core file size", "bytes"},
594	[RLIMIT_RSS] = {"Max resident set", "bytes"},
595	[RLIMIT_NPROC] = {"Max processes", "processes"},
596	[RLIMIT_NOFILE] = {"Max open files", "files"},
597	[RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
598	[RLIMIT_AS] = {"Max address space", "bytes"},
599	[RLIMIT_LOCKS] = {"Max file locks", "locks"},
600	[RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
601	[RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
602	[RLIMIT_NICE] = {"Max nice priority", NULL},
603	[RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
604	[RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
605};
606
607/* Display limits for a process */
608static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
609			   struct pid *pid, struct task_struct *task)
610{
611	unsigned int i;
612	unsigned long flags;
613
614	struct rlimit rlim[RLIM_NLIMITS];
615
616	if (!lock_task_sighand(task, &flags))
617		return 0;
618	memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
619	unlock_task_sighand(task, &flags);
620
621	/*
622	 * print the file header
623	 */
624	seq_puts(m, "Limit                     "
625		"Soft Limit           "
626		"Hard Limit           "
627		"Units     \n");
628
629	for (i = 0; i < RLIM_NLIMITS; i++) {
630		if (rlim[i].rlim_cur == RLIM_INFINITY)
631			seq_printf(m, "%-25s %-20s ",
632				   lnames[i].name, "unlimited");
633		else
634			seq_printf(m, "%-25s %-20lu ",
635				   lnames[i].name, rlim[i].rlim_cur);
636
637		if (rlim[i].rlim_max == RLIM_INFINITY)
638			seq_printf(m, "%-20s ", "unlimited");
639		else
640			seq_printf(m, "%-20lu ", rlim[i].rlim_max);
641
642		if (lnames[i].unit)
643			seq_printf(m, "%-10s\n", lnames[i].unit);
644		else
645			seq_putc(m, '\n');
646	}
647
648	return 0;
649}
650
651#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
652static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
653			    struct pid *pid, struct task_struct *task)
654{
655	struct syscall_info info;
656	u64 *args = &info.data.args[0];
657	int res;
658
659	res = lock_trace(task);
660	if (res)
661		return res;
662
663	if (task_current_syscall(task, &info))
664		seq_puts(m, "running\n");
665	else if (info.data.nr < 0)
666		seq_printf(m, "%d 0x%llx 0x%llx\n",
667			   info.data.nr, info.sp, info.data.instruction_pointer);
668	else
669		seq_printf(m,
670		       "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n",
671		       info.data.nr,
672		       args[0], args[1], args[2], args[3], args[4], args[5],
673		       info.sp, info.data.instruction_pointer);
674	unlock_trace(task);
675
676	return 0;
677}
678#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
679
680/************************************************************************/
681/*                       Here the fs part begins                        */
682/************************************************************************/
683
684/* permission checks */
685static bool proc_fd_access_allowed(struct inode *inode)
686{
687	struct task_struct *task;
688	bool allowed = false;
689	/* Allow access to a task's file descriptors if it is us or we
690	 * may use ptrace attach to the process and find out that
691	 * information.
692	 */
693	task = get_proc_task(inode);
694	if (task) {
695		allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
696		put_task_struct(task);
697	}
698	return allowed;
699}
700
701int proc_setattr(struct mnt_idmap *idmap, struct dentry *dentry,
702		 struct iattr *attr)
703{
704	int error;
705	struct inode *inode = d_inode(dentry);
706
707	if (attr->ia_valid & ATTR_MODE)
708		return -EPERM;
709
710	error = setattr_prepare(&nop_mnt_idmap, dentry, attr);
711	if (error)
712		return error;
713
714	setattr_copy(&nop_mnt_idmap, inode, attr);
715	return 0;
716}
717
718/*
719 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
720 * or euid/egid (for hide_pid_min=2)?
721 */
722static bool has_pid_permissions(struct proc_fs_info *fs_info,
723				 struct task_struct *task,
724				 enum proc_hidepid hide_pid_min)
725{
726	/*
727	 * If 'hidpid' mount option is set force a ptrace check,
728	 * we indicate that we are using a filesystem syscall
729	 * by passing PTRACE_MODE_READ_FSCREDS
730	 */
731	if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE)
732		return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
733
734	if (fs_info->hide_pid < hide_pid_min)
735		return true;
736	if (in_group_p(fs_info->pid_gid))
737		return true;
738	return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
739}
740
741
742static int proc_pid_permission(struct mnt_idmap *idmap,
743			       struct inode *inode, int mask)
744{
745	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
746	struct task_struct *task;
747	bool has_perms;
748
749	task = get_proc_task(inode);
750	if (!task)
751		return -ESRCH;
752	has_perms = has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS);
753	put_task_struct(task);
754
755	if (!has_perms) {
756		if (fs_info->hide_pid == HIDEPID_INVISIBLE) {
757			/*
758			 * Let's make getdents(), stat(), and open()
759			 * consistent with each other.  If a process
760			 * may not stat() a file, it shouldn't be seen
761			 * in procfs at all.
762			 */
763			return -ENOENT;
764		}
765
766		return -EPERM;
767	}
768	return generic_permission(&nop_mnt_idmap, inode, mask);
769}
770
771
772
773static const struct inode_operations proc_def_inode_operations = {
774	.setattr	= proc_setattr,
775};
776
777static int proc_single_show(struct seq_file *m, void *v)
778{
779	struct inode *inode = m->private;
780	struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
781	struct pid *pid = proc_pid(inode);
782	struct task_struct *task;
783	int ret;
784
785	task = get_pid_task(pid, PIDTYPE_PID);
786	if (!task)
787		return -ESRCH;
788
789	ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
790
791	put_task_struct(task);
792	return ret;
793}
794
795static int proc_single_open(struct inode *inode, struct file *filp)
796{
797	return single_open(filp, proc_single_show, inode);
798}
799
800static const struct file_operations proc_single_file_operations = {
801	.open		= proc_single_open,
802	.read		= seq_read,
803	.llseek		= seq_lseek,
804	.release	= single_release,
805};
806
807
808struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
809{
810	struct task_struct *task = get_proc_task(inode);
811	struct mm_struct *mm = ERR_PTR(-ESRCH);
812
813	if (task) {
814		mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
815		put_task_struct(task);
816
817		if (!IS_ERR_OR_NULL(mm)) {
818			/* ensure this mm_struct can't be freed */
819			mmgrab(mm);
820			/* but do not pin its memory */
821			mmput(mm);
822		}
823	}
824
825	return mm;
826}
827
828static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
829{
830	struct mm_struct *mm = proc_mem_open(inode, mode);
831
832	if (IS_ERR(mm))
833		return PTR_ERR(mm);
834
835	file->private_data = mm;
836	return 0;
837}
838
839static int mem_open(struct inode *inode, struct file *file)
840{
841	int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
842
843	/* OK to pass negative loff_t, we can catch out-of-range */
844	file->f_mode |= FMODE_UNSIGNED_OFFSET;
845
846	return ret;
847}
848
849static ssize_t mem_rw(struct file *file, char __user *buf,
850			size_t count, loff_t *ppos, int write)
851{
852	struct mm_struct *mm = file->private_data;
853	unsigned long addr = *ppos;
854	ssize_t copied;
855	char *page;
856	unsigned int flags;
857
858	if (!mm)
859		return 0;
860
861	page = (char *)__get_free_page(GFP_KERNEL);
862	if (!page)
863		return -ENOMEM;
864
865	copied = 0;
866	if (!mmget_not_zero(mm))
867		goto free;
868
869	flags = FOLL_FORCE | (write ? FOLL_WRITE : 0);
870
871	while (count > 0) {
872		size_t this_len = min_t(size_t, count, PAGE_SIZE);
873
874		if (write && copy_from_user(page, buf, this_len)) {
875			copied = -EFAULT;
876			break;
877		}
878
879		this_len = access_remote_vm(mm, addr, page, this_len, flags);
880		if (!this_len) {
881			if (!copied)
882				copied = -EIO;
883			break;
884		}
885
886		if (!write && copy_to_user(buf, page, this_len)) {
887			copied = -EFAULT;
888			break;
889		}
890
891		buf += this_len;
892		addr += this_len;
893		copied += this_len;
894		count -= this_len;
895	}
896	*ppos = addr;
897
898	mmput(mm);
899free:
900	free_page((unsigned long) page);
901	return copied;
902}
903
904static ssize_t mem_read(struct file *file, char __user *buf,
905			size_t count, loff_t *ppos)
906{
907	return mem_rw(file, buf, count, ppos, 0);
908}
909
910static ssize_t mem_write(struct file *file, const char __user *buf,
911			 size_t count, loff_t *ppos)
912{
913	return mem_rw(file, (char __user*)buf, count, ppos, 1);
914}
915
916loff_t mem_lseek(struct file *file, loff_t offset, int orig)
917{
918	switch (orig) {
919	case 0:
920		file->f_pos = offset;
921		break;
922	case 1:
923		file->f_pos += offset;
924		break;
925	default:
926		return -EINVAL;
927	}
928	force_successful_syscall_return();
929	return file->f_pos;
930}
931
932static int mem_release(struct inode *inode, struct file *file)
933{
934	struct mm_struct *mm = file->private_data;
935	if (mm)
936		mmdrop(mm);
937	return 0;
938}
939
940static const struct file_operations proc_mem_operations = {
941	.llseek		= mem_lseek,
942	.read		= mem_read,
943	.write		= mem_write,
944	.open		= mem_open,
945	.release	= mem_release,
946};
947
948static int environ_open(struct inode *inode, struct file *file)
949{
950	return __mem_open(inode, file, PTRACE_MODE_READ);
951}
952
953static ssize_t environ_read(struct file *file, char __user *buf,
954			size_t count, loff_t *ppos)
955{
956	char *page;
957	unsigned long src = *ppos;
958	int ret = 0;
959	struct mm_struct *mm = file->private_data;
960	unsigned long env_start, env_end;
961
962	/* Ensure the process spawned far enough to have an environment. */
963	if (!mm || !mm->env_end)
964		return 0;
965
966	page = (char *)__get_free_page(GFP_KERNEL);
967	if (!page)
968		return -ENOMEM;
969
970	ret = 0;
971	if (!mmget_not_zero(mm))
972		goto free;
973
974	spin_lock(&mm->arg_lock);
975	env_start = mm->env_start;
976	env_end = mm->env_end;
977	spin_unlock(&mm->arg_lock);
978
979	while (count > 0) {
980		size_t this_len, max_len;
981		int retval;
982
983		if (src >= (env_end - env_start))
984			break;
985
986		this_len = env_end - (env_start + src);
987
988		max_len = min_t(size_t, PAGE_SIZE, count);
989		this_len = min(max_len, this_len);
990
991		retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON);
992
993		if (retval <= 0) {
994			ret = retval;
995			break;
996		}
997
998		if (copy_to_user(buf, page, retval)) {
999			ret = -EFAULT;
1000			break;
1001		}
1002
1003		ret += retval;
1004		src += retval;
1005		buf += retval;
1006		count -= retval;
1007	}
1008	*ppos = src;
1009	mmput(mm);
1010
1011free:
1012	free_page((unsigned long) page);
1013	return ret;
1014}
1015
1016static const struct file_operations proc_environ_operations = {
1017	.open		= environ_open,
1018	.read		= environ_read,
1019	.llseek		= generic_file_llseek,
1020	.release	= mem_release,
1021};
1022
1023static int auxv_open(struct inode *inode, struct file *file)
1024{
1025	return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
1026}
1027
1028static ssize_t auxv_read(struct file *file, char __user *buf,
1029			size_t count, loff_t *ppos)
1030{
1031	struct mm_struct *mm = file->private_data;
1032	unsigned int nwords = 0;
1033
1034	if (!mm)
1035		return 0;
1036	do {
1037		nwords += 2;
1038	} while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1039	return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
1040				       nwords * sizeof(mm->saved_auxv[0]));
1041}
1042
1043static const struct file_operations proc_auxv_operations = {
1044	.open		= auxv_open,
1045	.read		= auxv_read,
1046	.llseek		= generic_file_llseek,
1047	.release	= mem_release,
1048};
1049
1050static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1051			    loff_t *ppos)
1052{
1053	struct task_struct *task = get_proc_task(file_inode(file));
1054	char buffer[PROC_NUMBUF];
1055	int oom_adj = OOM_ADJUST_MIN;
1056	size_t len;
1057
1058	if (!task)
1059		return -ESRCH;
1060	if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1061		oom_adj = OOM_ADJUST_MAX;
1062	else
1063		oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1064			  OOM_SCORE_ADJ_MAX;
1065	put_task_struct(task);
1066	if (oom_adj > OOM_ADJUST_MAX)
1067		oom_adj = OOM_ADJUST_MAX;
1068	len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1069	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1070}
1071
1072static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1073{
1074	struct mm_struct *mm = NULL;
1075	struct task_struct *task;
1076	int err = 0;
1077
1078	task = get_proc_task(file_inode(file));
1079	if (!task)
1080		return -ESRCH;
1081
1082	mutex_lock(&oom_adj_mutex);
1083	if (legacy) {
1084		if (oom_adj < task->signal->oom_score_adj &&
1085				!capable(CAP_SYS_RESOURCE)) {
1086			err = -EACCES;
1087			goto err_unlock;
1088		}
1089		/*
1090		 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1091		 * /proc/pid/oom_score_adj instead.
1092		 */
1093		pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1094			  current->comm, task_pid_nr(current), task_pid_nr(task),
1095			  task_pid_nr(task));
1096	} else {
1097		if ((short)oom_adj < task->signal->oom_score_adj_min &&
1098				!capable(CAP_SYS_RESOURCE)) {
1099			err = -EACCES;
1100			goto err_unlock;
1101		}
1102	}
1103
1104	/*
1105	 * Make sure we will check other processes sharing the mm if this is
1106	 * not vfrok which wants its own oom_score_adj.
1107	 * pin the mm so it doesn't go away and get reused after task_unlock
1108	 */
1109	if (!task->vfork_done) {
1110		struct task_struct *p = find_lock_task_mm(task);
1111
1112		if (p) {
1113			if (test_bit(MMF_MULTIPROCESS, &p->mm->flags)) {
1114				mm = p->mm;
1115				mmgrab(mm);
1116			}
1117			task_unlock(p);
1118		}
1119	}
1120
1121	task->signal->oom_score_adj = oom_adj;
1122	if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1123		task->signal->oom_score_adj_min = (short)oom_adj;
1124	trace_oom_score_adj_update(task);
1125
1126	if (mm) {
1127		struct task_struct *p;
1128
1129		rcu_read_lock();
1130		for_each_process(p) {
1131			if (same_thread_group(task, p))
1132				continue;
1133
1134			/* do not touch kernel threads or the global init */
1135			if (p->flags & PF_KTHREAD || is_global_init(p))
1136				continue;
1137
1138			task_lock(p);
1139			if (!p->vfork_done && process_shares_mm(p, mm)) {
1140				p->signal->oom_score_adj = oom_adj;
1141				if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1142					p->signal->oom_score_adj_min = (short)oom_adj;
1143			}
1144			task_unlock(p);
1145		}
1146		rcu_read_unlock();
1147		mmdrop(mm);
1148	}
1149err_unlock:
1150	mutex_unlock(&oom_adj_mutex);
1151	put_task_struct(task);
1152	return err;
1153}
1154
1155/*
1156 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1157 * kernels.  The effective policy is defined by oom_score_adj, which has a
1158 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1159 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1160 * Processes that become oom disabled via oom_adj will still be oom disabled
1161 * with this implementation.
1162 *
1163 * oom_adj cannot be removed since existing userspace binaries use it.
1164 */
1165static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1166			     size_t count, loff_t *ppos)
1167{
1168	char buffer[PROC_NUMBUF];
1169	int oom_adj;
1170	int err;
1171
1172	memset(buffer, 0, sizeof(buffer));
1173	if (count > sizeof(buffer) - 1)
1174		count = sizeof(buffer) - 1;
1175	if (copy_from_user(buffer, buf, count)) {
1176		err = -EFAULT;
1177		goto out;
1178	}
1179
1180	err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1181	if (err)
1182		goto out;
1183	if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1184	     oom_adj != OOM_DISABLE) {
1185		err = -EINVAL;
1186		goto out;
1187	}
1188
1189	/*
1190	 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1191	 * value is always attainable.
1192	 */
1193	if (oom_adj == OOM_ADJUST_MAX)
1194		oom_adj = OOM_SCORE_ADJ_MAX;
1195	else
1196		oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1197
1198	err = __set_oom_adj(file, oom_adj, true);
1199out:
1200	return err < 0 ? err : count;
1201}
1202
1203static const struct file_operations proc_oom_adj_operations = {
1204	.read		= oom_adj_read,
1205	.write		= oom_adj_write,
1206	.llseek		= generic_file_llseek,
1207};
1208
1209static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1210					size_t count, loff_t *ppos)
1211{
1212	struct task_struct *task = get_proc_task(file_inode(file));
1213	char buffer[PROC_NUMBUF];
1214	short oom_score_adj = OOM_SCORE_ADJ_MIN;
1215	size_t len;
1216
1217	if (!task)
1218		return -ESRCH;
1219	oom_score_adj = task->signal->oom_score_adj;
1220	put_task_struct(task);
1221	len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1222	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1223}
1224
1225static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1226					size_t count, loff_t *ppos)
1227{
1228	char buffer[PROC_NUMBUF];
1229	int oom_score_adj;
1230	int err;
1231
1232	memset(buffer, 0, sizeof(buffer));
1233	if (count > sizeof(buffer) - 1)
1234		count = sizeof(buffer) - 1;
1235	if (copy_from_user(buffer, buf, count)) {
1236		err = -EFAULT;
1237		goto out;
1238	}
1239
1240	err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1241	if (err)
1242		goto out;
1243	if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1244			oom_score_adj > OOM_SCORE_ADJ_MAX) {
1245		err = -EINVAL;
1246		goto out;
1247	}
1248
1249	err = __set_oom_adj(file, oom_score_adj, false);
1250out:
1251	return err < 0 ? err : count;
1252}
1253
1254static const struct file_operations proc_oom_score_adj_operations = {
1255	.read		= oom_score_adj_read,
1256	.write		= oom_score_adj_write,
1257	.llseek		= default_llseek,
1258};
1259
1260#ifdef CONFIG_AUDIT
1261#define TMPBUFLEN 11
1262static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1263				  size_t count, loff_t *ppos)
1264{
1265	struct inode * inode = file_inode(file);
1266	struct task_struct *task = get_proc_task(inode);
1267	ssize_t length;
1268	char tmpbuf[TMPBUFLEN];
1269
1270	if (!task)
1271		return -ESRCH;
1272	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1273			   from_kuid(file->f_cred->user_ns,
1274				     audit_get_loginuid(task)));
1275	put_task_struct(task);
1276	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1277}
1278
1279static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1280				   size_t count, loff_t *ppos)
1281{
1282	struct inode * inode = file_inode(file);
1283	uid_t loginuid;
1284	kuid_t kloginuid;
1285	int rv;
1286
1287	/* Don't let kthreads write their own loginuid */
1288	if (current->flags & PF_KTHREAD)
1289		return -EPERM;
1290
1291	rcu_read_lock();
1292	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1293		rcu_read_unlock();
1294		return -EPERM;
1295	}
1296	rcu_read_unlock();
1297
1298	if (*ppos != 0) {
1299		/* No partial writes. */
1300		return -EINVAL;
1301	}
1302
1303	rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1304	if (rv < 0)
1305		return rv;
1306
1307	/* is userspace tring to explicitly UNSET the loginuid? */
1308	if (loginuid == AUDIT_UID_UNSET) {
1309		kloginuid = INVALID_UID;
1310	} else {
1311		kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1312		if (!uid_valid(kloginuid))
1313			return -EINVAL;
1314	}
1315
1316	rv = audit_set_loginuid(kloginuid);
1317	if (rv < 0)
1318		return rv;
1319	return count;
1320}
1321
1322static const struct file_operations proc_loginuid_operations = {
1323	.read		= proc_loginuid_read,
1324	.write		= proc_loginuid_write,
1325	.llseek		= generic_file_llseek,
1326};
1327
1328static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1329				  size_t count, loff_t *ppos)
1330{
1331	struct inode * inode = file_inode(file);
1332	struct task_struct *task = get_proc_task(inode);
1333	ssize_t length;
1334	char tmpbuf[TMPBUFLEN];
1335
1336	if (!task)
1337		return -ESRCH;
1338	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1339				audit_get_sessionid(task));
1340	put_task_struct(task);
1341	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1342}
1343
1344static const struct file_operations proc_sessionid_operations = {
1345	.read		= proc_sessionid_read,
1346	.llseek		= generic_file_llseek,
1347};
1348#endif
1349
1350#ifdef CONFIG_FAULT_INJECTION
1351static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1352				      size_t count, loff_t *ppos)
1353{
1354	struct task_struct *task = get_proc_task(file_inode(file));
1355	char buffer[PROC_NUMBUF];
1356	size_t len;
1357	int make_it_fail;
1358
1359	if (!task)
1360		return -ESRCH;
1361	make_it_fail = task->make_it_fail;
1362	put_task_struct(task);
1363
1364	len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1365
1366	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1367}
1368
1369static ssize_t proc_fault_inject_write(struct file * file,
1370			const char __user * buf, size_t count, loff_t *ppos)
1371{
1372	struct task_struct *task;
1373	char buffer[PROC_NUMBUF];
1374	int make_it_fail;
1375	int rv;
1376
1377	if (!capable(CAP_SYS_RESOURCE))
1378		return -EPERM;
1379	memset(buffer, 0, sizeof(buffer));
1380	if (count > sizeof(buffer) - 1)
1381		count = sizeof(buffer) - 1;
1382	if (copy_from_user(buffer, buf, count))
1383		return -EFAULT;
1384	rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1385	if (rv < 0)
1386		return rv;
1387	if (make_it_fail < 0 || make_it_fail > 1)
1388		return -EINVAL;
1389
1390	task = get_proc_task(file_inode(file));
1391	if (!task)
1392		return -ESRCH;
1393	task->make_it_fail = make_it_fail;
1394	put_task_struct(task);
1395
1396	return count;
1397}
1398
1399static const struct file_operations proc_fault_inject_operations = {
1400	.read		= proc_fault_inject_read,
1401	.write		= proc_fault_inject_write,
1402	.llseek		= generic_file_llseek,
1403};
1404
1405static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1406				   size_t count, loff_t *ppos)
1407{
1408	struct task_struct *task;
1409	int err;
1410	unsigned int n;
1411
1412	err = kstrtouint_from_user(buf, count, 0, &n);
1413	if (err)
1414		return err;
1415
1416	task = get_proc_task(file_inode(file));
1417	if (!task)
1418		return -ESRCH;
1419	task->fail_nth = n;
1420	put_task_struct(task);
1421
1422	return count;
1423}
1424
1425static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1426				  size_t count, loff_t *ppos)
1427{
1428	struct task_struct *task;
1429	char numbuf[PROC_NUMBUF];
1430	ssize_t len;
1431
1432	task = get_proc_task(file_inode(file));
1433	if (!task)
1434		return -ESRCH;
1435	len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth);
1436	put_task_struct(task);
1437	return simple_read_from_buffer(buf, count, ppos, numbuf, len);
1438}
1439
1440static const struct file_operations proc_fail_nth_operations = {
1441	.read		= proc_fail_nth_read,
1442	.write		= proc_fail_nth_write,
1443};
1444#endif
1445
1446
1447#ifdef CONFIG_SCHED_DEBUG
1448/*
1449 * Print out various scheduling related per-task fields:
1450 */
1451static int sched_show(struct seq_file *m, void *v)
1452{
1453	struct inode *inode = m->private;
1454	struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
1455	struct task_struct *p;
1456
1457	p = get_proc_task(inode);
1458	if (!p)
1459		return -ESRCH;
1460	proc_sched_show_task(p, ns, m);
1461
1462	put_task_struct(p);
1463
1464	return 0;
1465}
1466
1467static ssize_t
1468sched_write(struct file *file, const char __user *buf,
1469	    size_t count, loff_t *offset)
1470{
1471	struct inode *inode = file_inode(file);
1472	struct task_struct *p;
1473
1474	p = get_proc_task(inode);
1475	if (!p)
1476		return -ESRCH;
1477	proc_sched_set_task(p);
1478
1479	put_task_struct(p);
1480
1481	return count;
1482}
1483
1484static int sched_open(struct inode *inode, struct file *filp)
1485{
1486	return single_open(filp, sched_show, inode);
1487}
1488
1489static const struct file_operations proc_pid_sched_operations = {
1490	.open		= sched_open,
1491	.read		= seq_read,
1492	.write		= sched_write,
1493	.llseek		= seq_lseek,
1494	.release	= single_release,
1495};
1496
1497#endif
1498
1499#ifdef CONFIG_QOS_CTRL
1500long proc_qos_ctrl_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1501{
1502	return do_qos_ctrl_ioctl(QOS_IOCTL_ABI_AARCH64, file, cmd, arg);
1503}
1504
1505#ifdef CONFIG_COMPAT
1506long proc_qos_ctrl_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
1507{
1508	return do_qos_ctrl_ioctl(QOS_IOCTL_ABI_ARM32, file, cmd,
1509			(unsigned long)(compat_ptr((compat_uptr_t)arg)));
1510}
1511#endif
1512
1513int proc_qos_ctrl_open(struct inode *inode, struct file *filp)
1514{
1515	return 0;
1516}
1517
1518static const struct file_operations proc_qos_ctrl_operations = {
1519	.open   = proc_qos_ctrl_open,
1520	.unlocked_ioctl = proc_qos_ctrl_ioctl,
1521#ifdef CONFIG_COMPAT
1522	.compat_ioctl = proc_qos_ctrl_compat_ioctl,
1523#endif
1524};
1525#endif
1526
1527#ifdef CONFIG_SCHED_RTG
1528static const struct file_operations proc_rtg_operations = {
1529	.open		= proc_rtg_open,
1530	.unlocked_ioctl	= proc_rtg_ioctl,
1531#ifdef CONFIG_COMPAT
1532	.compat_ioctl	= proc_rtg_compat_ioctl,
1533#endif
1534};
1535#endif
1536
1537#ifdef CONFIG_SCHED_RTG_DEBUG
1538static int sched_group_id_show(struct seq_file *m, void *v)
1539{
1540	struct inode *inode = m->private;
1541	struct task_struct *p;
1542
1543	p = get_proc_task(inode);
1544	if (!p)
1545		return -ESRCH;
1546
1547	seq_printf(m, "%d\n", sched_get_group_id(p));
1548
1549	put_task_struct(p);
1550
1551	return 0;
1552}
1553
1554static ssize_t
1555sched_group_id_write(struct file *file, const char __user *buf,
1556	    size_t count, loff_t *offset)
1557{
1558	struct inode *inode = file_inode(file);
1559	struct task_struct *p;
1560	char buffer[PROC_NUMBUF];
1561	int group_id, err;
1562
1563	memset(buffer, 0, sizeof(buffer));
1564	if (count > sizeof(buffer) - 1)
1565		count = sizeof(buffer) - 1;
1566	if (copy_from_user(buffer, buf, count)) {
1567		err = -EFAULT;
1568		goto out;
1569	}
1570
1571	err = kstrtoint(strstrip(buffer), 0, &group_id);
1572	if (err)
1573		goto out;
1574
1575	p = get_proc_task(inode);
1576	if (!p)
1577		return -ESRCH;
1578
1579	err = sched_set_group_id(p, group_id);
1580
1581	put_task_struct(p);
1582
1583out:
1584	return err < 0 ? err : count;
1585}
1586
1587static int sched_group_id_open(struct inode *inode, struct file *filp)
1588{
1589	return single_open(filp, sched_group_id_show, inode);
1590}
1591
1592static const struct file_operations proc_pid_sched_group_id_operations = {
1593	.open		= sched_group_id_open,
1594	.read		= seq_read,
1595	.write		= sched_group_id_write,
1596	.llseek		= seq_lseek,
1597	.release	= single_release,
1598};
1599#endif	/* CONFIG_SCHED_RTG_DEBUG */
1600
1601#ifdef CONFIG_SCHED_AUTOGROUP
1602/*
1603 * Print out autogroup related information:
1604 */
1605static int sched_autogroup_show(struct seq_file *m, void *v)
1606{
1607	struct inode *inode = m->private;
1608	struct task_struct *p;
1609
1610	p = get_proc_task(inode);
1611	if (!p)
1612		return -ESRCH;
1613	proc_sched_autogroup_show_task(p, m);
1614
1615	put_task_struct(p);
1616
1617	return 0;
1618}
1619
1620static ssize_t
1621sched_autogroup_write(struct file *file, const char __user *buf,
1622	    size_t count, loff_t *offset)
1623{
1624	struct inode *inode = file_inode(file);
1625	struct task_struct *p;
1626	char buffer[PROC_NUMBUF];
1627	int nice;
1628	int err;
1629
1630	memset(buffer, 0, sizeof(buffer));
1631	if (count > sizeof(buffer) - 1)
1632		count = sizeof(buffer) - 1;
1633	if (copy_from_user(buffer, buf, count))
1634		return -EFAULT;
1635
1636	err = kstrtoint(strstrip(buffer), 0, &nice);
1637	if (err < 0)
1638		return err;
1639
1640	p = get_proc_task(inode);
1641	if (!p)
1642		return -ESRCH;
1643
1644	err = proc_sched_autogroup_set_nice(p, nice);
1645	if (err)
1646		count = err;
1647
1648	put_task_struct(p);
1649
1650	return count;
1651}
1652
1653static int sched_autogroup_open(struct inode *inode, struct file *filp)
1654{
1655	int ret;
1656
1657	ret = single_open(filp, sched_autogroup_show, NULL);
1658	if (!ret) {
1659		struct seq_file *m = filp->private_data;
1660
1661		m->private = inode;
1662	}
1663	return ret;
1664}
1665
1666static const struct file_operations proc_pid_sched_autogroup_operations = {
1667	.open		= sched_autogroup_open,
1668	.read		= seq_read,
1669	.write		= sched_autogroup_write,
1670	.llseek		= seq_lseek,
1671	.release	= single_release,
1672};
1673
1674#endif /* CONFIG_SCHED_AUTOGROUP */
1675
1676#ifdef CONFIG_SCHED_WALT
1677static int sched_init_task_load_show(struct seq_file *m, void *v)
1678{
1679	struct inode *inode = m->private;
1680	struct task_struct *p;
1681
1682	p = get_proc_task(inode);
1683	if (!p)
1684		return -ESRCH;
1685
1686	seq_printf(m, "%d\n", sched_get_init_task_load(p));
1687
1688	put_task_struct(p);
1689
1690	return 0;
1691}
1692
1693static ssize_t
1694sched_init_task_load_write(struct file *file, const char __user *buf,
1695	    size_t count, loff_t *offset)
1696{
1697	struct inode *inode = file_inode(file);
1698	struct task_struct *p;
1699	char buffer[PROC_NUMBUF];
1700	int init_task_load, err;
1701
1702	memset(buffer, 0, sizeof(buffer));
1703	if (count > sizeof(buffer) - 1)
1704		count = sizeof(buffer) - 1;
1705	if (copy_from_user(buffer, buf, count)) {
1706		err = -EFAULT;
1707		goto out;
1708	}
1709
1710	err = kstrtoint(strstrip(buffer), 0, &init_task_load);
1711	if (err)
1712		goto out;
1713
1714	p = get_proc_task(inode);
1715	if (!p)
1716		return -ESRCH;
1717
1718	err = sched_set_init_task_load(p, init_task_load);
1719
1720	put_task_struct(p);
1721
1722out:
1723	return err < 0 ? err : count;
1724}
1725
1726static int sched_init_task_load_open(struct inode *inode, struct file *filp)
1727{
1728	return single_open(filp, sched_init_task_load_show, inode);
1729}
1730
1731static const struct file_operations proc_pid_sched_init_task_load_operations = {
1732	.open		= sched_init_task_load_open,
1733	.read		= seq_read,
1734	.write		= sched_init_task_load_write,
1735	.llseek		= seq_lseek,
1736	.release	= single_release,
1737};
1738#endif	/* CONFIG_SCHED_WALT */
1739
1740#ifdef CONFIG_TIME_NS
1741static int timens_offsets_show(struct seq_file *m, void *v)
1742{
1743	struct task_struct *p;
1744
1745	p = get_proc_task(file_inode(m->file));
1746	if (!p)
1747		return -ESRCH;
1748	proc_timens_show_offsets(p, m);
1749
1750	put_task_struct(p);
1751
1752	return 0;
1753}
1754
1755static ssize_t timens_offsets_write(struct file *file, const char __user *buf,
1756				    size_t count, loff_t *ppos)
1757{
1758	struct inode *inode = file_inode(file);
1759	struct proc_timens_offset offsets[2];
1760	char *kbuf = NULL, *pos, *next_line;
1761	struct task_struct *p;
1762	int ret, noffsets;
1763
1764	/* Only allow < page size writes at the beginning of the file */
1765	if ((*ppos != 0) || (count >= PAGE_SIZE))
1766		return -EINVAL;
1767
1768	/* Slurp in the user data */
1769	kbuf = memdup_user_nul(buf, count);
1770	if (IS_ERR(kbuf))
1771		return PTR_ERR(kbuf);
1772
1773	/* Parse the user data */
1774	ret = -EINVAL;
1775	noffsets = 0;
1776	for (pos = kbuf; pos; pos = next_line) {
1777		struct proc_timens_offset *off = &offsets[noffsets];
1778		char clock[10];
1779		int err;
1780
1781		/* Find the end of line and ensure we don't look past it */
1782		next_line = strchr(pos, '\n');
1783		if (next_line) {
1784			*next_line = '\0';
1785			next_line++;
1786			if (*next_line == '\0')
1787				next_line = NULL;
1788		}
1789
1790		err = sscanf(pos, "%9s %lld %lu", clock,
1791				&off->val.tv_sec, &off->val.tv_nsec);
1792		if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC)
1793			goto out;
1794
1795		clock[sizeof(clock) - 1] = 0;
1796		if (strcmp(clock, "monotonic") == 0 ||
1797		    strcmp(clock, __stringify(CLOCK_MONOTONIC)) == 0)
1798			off->clockid = CLOCK_MONOTONIC;
1799		else if (strcmp(clock, "boottime") == 0 ||
1800			 strcmp(clock, __stringify(CLOCK_BOOTTIME)) == 0)
1801			off->clockid = CLOCK_BOOTTIME;
1802		else
1803			goto out;
1804
1805		noffsets++;
1806		if (noffsets == ARRAY_SIZE(offsets)) {
1807			if (next_line)
1808				count = next_line - kbuf;
1809			break;
1810		}
1811	}
1812
1813	ret = -ESRCH;
1814	p = get_proc_task(inode);
1815	if (!p)
1816		goto out;
1817	ret = proc_timens_set_offset(file, p, offsets, noffsets);
1818	put_task_struct(p);
1819	if (ret)
1820		goto out;
1821
1822	ret = count;
1823out:
1824	kfree(kbuf);
1825	return ret;
1826}
1827
1828static int timens_offsets_open(struct inode *inode, struct file *filp)
1829{
1830	return single_open(filp, timens_offsets_show, inode);
1831}
1832
1833static const struct file_operations proc_timens_offsets_operations = {
1834	.open		= timens_offsets_open,
1835	.read		= seq_read,
1836	.write		= timens_offsets_write,
1837	.llseek		= seq_lseek,
1838	.release	= single_release,
1839};
1840#endif /* CONFIG_TIME_NS */
1841
1842static ssize_t comm_write(struct file *file, const char __user *buf,
1843				size_t count, loff_t *offset)
1844{
1845	struct inode *inode = file_inode(file);
1846	struct task_struct *p;
1847	char buffer[TASK_COMM_LEN];
1848	const size_t maxlen = sizeof(buffer) - 1;
1849
1850	memset(buffer, 0, sizeof(buffer));
1851	if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1852		return -EFAULT;
1853
1854	p = get_proc_task(inode);
1855	if (!p)
1856		return -ESRCH;
1857
1858	if (same_thread_group(current, p)) {
1859		set_task_comm(p, buffer);
1860		proc_comm_connector(p);
1861	}
1862	else
1863		count = -EINVAL;
1864
1865	put_task_struct(p);
1866
1867	return count;
1868}
1869
1870static int comm_show(struct seq_file *m, void *v)
1871{
1872	struct inode *inode = m->private;
1873	struct task_struct *p;
1874
1875	p = get_proc_task(inode);
1876	if (!p)
1877		return -ESRCH;
1878
1879	proc_task_name(m, p, false);
1880	seq_putc(m, '\n');
1881
1882	put_task_struct(p);
1883
1884	return 0;
1885}
1886
1887static int comm_open(struct inode *inode, struct file *filp)
1888{
1889	return single_open(filp, comm_show, inode);
1890}
1891
1892static const struct file_operations proc_pid_set_comm_operations = {
1893	.open		= comm_open,
1894	.read		= seq_read,
1895	.write		= comm_write,
1896	.llseek		= seq_lseek,
1897	.release	= single_release,
1898};
1899
1900static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1901{
1902	struct task_struct *task;
1903	struct file *exe_file;
1904
1905	task = get_proc_task(d_inode(dentry));
1906	if (!task)
1907		return -ENOENT;
1908	exe_file = get_task_exe_file(task);
1909	put_task_struct(task);
1910	if (exe_file) {
1911		*exe_path = exe_file->f_path;
1912		path_get(&exe_file->f_path);
1913		fput(exe_file);
1914		return 0;
1915	} else
1916		return -ENOENT;
1917}
1918
1919static const char *proc_pid_get_link(struct dentry *dentry,
1920				     struct inode *inode,
1921				     struct delayed_call *done)
1922{
1923	struct path path;
1924	int error = -EACCES;
1925
1926	if (!dentry)
1927		return ERR_PTR(-ECHILD);
1928
1929	/* Are we allowed to snoop on the tasks file descriptors? */
1930	if (!proc_fd_access_allowed(inode))
1931		goto out;
1932
1933	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1934	if (error)
1935		goto out;
1936
1937	error = nd_jump_link(&path);
1938out:
1939	return ERR_PTR(error);
1940}
1941
1942static int do_proc_readlink(const struct path *path, char __user *buffer, int buflen)
1943{
1944	char *tmp = kmalloc(PATH_MAX, GFP_KERNEL);
1945	char *pathname;
1946	int len;
1947
1948	if (!tmp)
1949		return -ENOMEM;
1950
1951	pathname = d_path(path, tmp, PATH_MAX);
1952	len = PTR_ERR(pathname);
1953	if (IS_ERR(pathname))
1954		goto out;
1955	len = tmp + PATH_MAX - 1 - pathname;
1956
1957	if (len > buflen)
1958		len = buflen;
1959	if (copy_to_user(buffer, pathname, len))
1960		len = -EFAULT;
1961 out:
1962	kfree(tmp);
1963	return len;
1964}
1965
1966static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1967{
1968	int error = -EACCES;
1969	struct inode *inode = d_inode(dentry);
1970	struct path path;
1971
1972	/* Are we allowed to snoop on the tasks file descriptors? */
1973	if (!proc_fd_access_allowed(inode))
1974		goto out;
1975
1976	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1977	if (error)
1978		goto out;
1979
1980	error = do_proc_readlink(&path, buffer, buflen);
1981	path_put(&path);
1982out:
1983	return error;
1984}
1985
1986const struct inode_operations proc_pid_link_inode_operations = {
1987	.readlink	= proc_pid_readlink,
1988	.get_link	= proc_pid_get_link,
1989	.setattr	= proc_setattr,
1990};
1991
1992
1993/* building an inode */
1994
1995void task_dump_owner(struct task_struct *task, umode_t mode,
1996		     kuid_t *ruid, kgid_t *rgid)
1997{
1998	/* Depending on the state of dumpable compute who should own a
1999	 * proc file for a task.
2000	 */
2001	const struct cred *cred;
2002	kuid_t uid;
2003	kgid_t gid;
2004
2005	if (unlikely(task->flags & PF_KTHREAD)) {
2006		*ruid = GLOBAL_ROOT_UID;
2007		*rgid = GLOBAL_ROOT_GID;
2008		return;
2009	}
2010
2011	/* Default to the tasks effective ownership */
2012	rcu_read_lock();
2013	cred = __task_cred(task);
2014	uid = cred->euid;
2015	gid = cred->egid;
2016	rcu_read_unlock();
2017
2018	/*
2019	 * Before the /proc/pid/status file was created the only way to read
2020	 * the effective uid of a /process was to stat /proc/pid.  Reading
2021	 * /proc/pid/status is slow enough that procps and other packages
2022	 * kept stating /proc/pid.  To keep the rules in /proc simple I have
2023	 * made this apply to all per process world readable and executable
2024	 * directories.
2025	 */
2026	if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
2027		struct mm_struct *mm;
2028		task_lock(task);
2029		mm = task->mm;
2030		/* Make non-dumpable tasks owned by some root */
2031		if (mm) {
2032			if (get_dumpable(mm) != SUID_DUMP_USER) {
2033				struct user_namespace *user_ns = mm->user_ns;
2034
2035				uid = make_kuid(user_ns, 0);
2036				if (!uid_valid(uid))
2037					uid = GLOBAL_ROOT_UID;
2038
2039				gid = make_kgid(user_ns, 0);
2040				if (!gid_valid(gid))
2041					gid = GLOBAL_ROOT_GID;
2042			}
2043		} else {
2044			uid = GLOBAL_ROOT_UID;
2045			gid = GLOBAL_ROOT_GID;
2046		}
2047		task_unlock(task);
2048	}
2049	*ruid = uid;
2050	*rgid = gid;
2051}
2052
2053void proc_pid_evict_inode(struct proc_inode *ei)
2054{
2055	struct pid *pid = ei->pid;
2056
2057	if (S_ISDIR(ei->vfs_inode.i_mode)) {
2058		spin_lock(&pid->lock);
2059		hlist_del_init_rcu(&ei->sibling_inodes);
2060		spin_unlock(&pid->lock);
2061	}
2062
2063	put_pid(pid);
2064}
2065
2066struct inode *proc_pid_make_inode(struct super_block *sb,
2067				  struct task_struct *task, umode_t mode)
2068{
2069	struct inode * inode;
2070	struct proc_inode *ei;
2071	struct pid *pid;
2072
2073	/* We need a new inode */
2074
2075	inode = new_inode(sb);
2076	if (!inode)
2077		goto out;
2078
2079	/* Common stuff */
2080	ei = PROC_I(inode);
2081	inode->i_mode = mode;
2082	inode->i_ino = get_next_ino();
2083	inode->i_mtime = inode->i_atime = inode_set_ctime_current(inode);
2084	inode->i_op = &proc_def_inode_operations;
2085
2086	/*
2087	 * grab the reference to task.
2088	 */
2089	pid = get_task_pid(task, PIDTYPE_PID);
2090	if (!pid)
2091		goto out_unlock;
2092
2093	/* Let the pid remember us for quick removal */
2094	ei->pid = pid;
2095
2096	task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
2097	security_task_to_inode(task, inode);
2098
2099out:
2100	return inode;
2101
2102out_unlock:
2103	iput(inode);
2104	return NULL;
2105}
2106
2107/*
2108 * Generating an inode and adding it into @pid->inodes, so that task will
2109 * invalidate inode's dentry before being released.
2110 *
2111 * This helper is used for creating dir-type entries under '/proc' and
2112 * '/proc/<tgid>/task'. Other entries(eg. fd, stat) under '/proc/<tgid>'
2113 * can be released by invalidating '/proc/<tgid>' dentry.
2114 * In theory, dentries under '/proc/<tgid>/task' can also be released by
2115 * invalidating '/proc/<tgid>' dentry, we reserve it to handle single
2116 * thread exiting situation: Any one of threads should invalidate its
2117 * '/proc/<tgid>/task/<pid>' dentry before released.
2118 */
2119static struct inode *proc_pid_make_base_inode(struct super_block *sb,
2120				struct task_struct *task, umode_t mode)
2121{
2122	struct inode *inode;
2123	struct proc_inode *ei;
2124	struct pid *pid;
2125
2126	inode = proc_pid_make_inode(sb, task, mode);
2127	if (!inode)
2128		return NULL;
2129
2130	/* Let proc_flush_pid find this directory inode */
2131	ei = PROC_I(inode);
2132	pid = ei->pid;
2133	spin_lock(&pid->lock);
2134	hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes);
2135	spin_unlock(&pid->lock);
2136
2137	return inode;
2138}
2139
2140int pid_getattr(struct mnt_idmap *idmap, const struct path *path,
2141		struct kstat *stat, u32 request_mask, unsigned int query_flags)
2142{
2143	struct inode *inode = d_inode(path->dentry);
2144	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
2145	struct task_struct *task;
2146
2147	generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
2148
2149	stat->uid = GLOBAL_ROOT_UID;
2150	stat->gid = GLOBAL_ROOT_GID;
2151	rcu_read_lock();
2152	task = pid_task(proc_pid(inode), PIDTYPE_PID);
2153	if (task) {
2154		if (!has_pid_permissions(fs_info, task, HIDEPID_INVISIBLE)) {
2155			rcu_read_unlock();
2156			/*
2157			 * This doesn't prevent learning whether PID exists,
2158			 * it only makes getattr() consistent with readdir().
2159			 */
2160			return -ENOENT;
2161		}
2162		task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
2163	}
2164	rcu_read_unlock();
2165	return 0;
2166}
2167
2168/* dentry stuff */
2169
2170/*
2171 * Set <pid>/... inode ownership (can change due to setuid(), etc.)
2172 */
2173void pid_update_inode(struct task_struct *task, struct inode *inode)
2174{
2175	task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
2176
2177	inode->i_mode &= ~(S_ISUID | S_ISGID);
2178	security_task_to_inode(task, inode);
2179}
2180
2181/*
2182 * Rewrite the inode's ownerships here because the owning task may have
2183 * performed a setuid(), etc.
2184 *
2185 */
2186static int pid_revalidate(struct dentry *dentry, unsigned int flags)
2187{
2188	struct inode *inode;
2189	struct task_struct *task;
2190	int ret = 0;
2191
2192	rcu_read_lock();
2193	inode = d_inode_rcu(dentry);
2194	if (!inode)
2195		goto out;
2196	task = pid_task(proc_pid(inode), PIDTYPE_PID);
2197
2198	if (task) {
2199		pid_update_inode(task, inode);
2200		ret = 1;
2201	}
2202out:
2203	rcu_read_unlock();
2204	return ret;
2205}
2206
2207static inline bool proc_inode_is_dead(struct inode *inode)
2208{
2209	return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
2210}
2211
2212int pid_delete_dentry(const struct dentry *dentry)
2213{
2214	/* Is the task we represent dead?
2215	 * If so, then don't put the dentry on the lru list,
2216	 * kill it immediately.
2217	 */
2218	return proc_inode_is_dead(d_inode(dentry));
2219}
2220
2221const struct dentry_operations pid_dentry_operations =
2222{
2223	.d_revalidate	= pid_revalidate,
2224	.d_delete	= pid_delete_dentry,
2225};
2226
2227/* Lookups */
2228
2229/*
2230 * Fill a directory entry.
2231 *
2232 * If possible create the dcache entry and derive our inode number and
2233 * file type from dcache entry.
2234 *
2235 * Since all of the proc inode numbers are dynamically generated, the inode
2236 * numbers do not exist until the inode is cache.  This means creating
2237 * the dcache entry in readdir is necessary to keep the inode numbers
2238 * reported by readdir in sync with the inode numbers reported
2239 * by stat.
2240 */
2241bool proc_fill_cache(struct file *file, struct dir_context *ctx,
2242	const char *name, unsigned int len,
2243	instantiate_t instantiate, struct task_struct *task, const void *ptr)
2244{
2245	struct dentry *child, *dir = file->f_path.dentry;
2246	struct qstr qname = QSTR_INIT(name, len);
2247	struct inode *inode;
2248	unsigned type = DT_UNKNOWN;
2249	ino_t ino = 1;
2250
2251	child = d_hash_and_lookup(dir, &qname);
2252	if (!child) {
2253		DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
2254		child = d_alloc_parallel(dir, &qname, &wq);
2255		if (IS_ERR(child))
2256			goto end_instantiate;
2257		if (d_in_lookup(child)) {
2258			struct dentry *res;
2259			res = instantiate(child, task, ptr);
2260			d_lookup_done(child);
2261			if (unlikely(res)) {
2262				dput(child);
2263				child = res;
2264				if (IS_ERR(child))
2265					goto end_instantiate;
2266			}
2267		}
2268	}
2269	inode = d_inode(child);
2270	ino = inode->i_ino;
2271	type = inode->i_mode >> 12;
2272	dput(child);
2273end_instantiate:
2274	return dir_emit(ctx, name, len, ino, type);
2275}
2276
2277/*
2278 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2279 * which represent vma start and end addresses.
2280 */
2281static int dname_to_vma_addr(struct dentry *dentry,
2282			     unsigned long *start, unsigned long *end)
2283{
2284	const char *str = dentry->d_name.name;
2285	unsigned long long sval, eval;
2286	unsigned int len;
2287
2288	if (str[0] == '0' && str[1] != '-')
2289		return -EINVAL;
2290	len = _parse_integer(str, 16, &sval);
2291	if (len & KSTRTOX_OVERFLOW)
2292		return -EINVAL;
2293	if (sval != (unsigned long)sval)
2294		return -EINVAL;
2295	str += len;
2296
2297	if (*str != '-')
2298		return -EINVAL;
2299	str++;
2300
2301	if (str[0] == '0' && str[1])
2302		return -EINVAL;
2303	len = _parse_integer(str, 16, &eval);
2304	if (len & KSTRTOX_OVERFLOW)
2305		return -EINVAL;
2306	if (eval != (unsigned long)eval)
2307		return -EINVAL;
2308	str += len;
2309
2310	if (*str != '\0')
2311		return -EINVAL;
2312
2313	*start = sval;
2314	*end = eval;
2315
2316	return 0;
2317}
2318
2319static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
2320{
2321	unsigned long vm_start, vm_end;
2322	bool exact_vma_exists = false;
2323	struct mm_struct *mm = NULL;
2324	struct task_struct *task;
2325	struct inode *inode;
2326	int status = 0;
2327
2328	if (flags & LOOKUP_RCU)
2329		return -ECHILD;
2330
2331	inode = d_inode(dentry);
2332	task = get_proc_task(inode);
2333	if (!task)
2334		goto out_notask;
2335
2336	mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2337	if (IS_ERR_OR_NULL(mm))
2338		goto out;
2339
2340	if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2341		status = mmap_read_lock_killable(mm);
2342		if (!status) {
2343			exact_vma_exists = !!find_exact_vma(mm, vm_start,
2344							    vm_end);
2345			mmap_read_unlock(mm);
2346		}
2347	}
2348
2349	mmput(mm);
2350
2351	if (exact_vma_exists) {
2352		task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
2353
2354		security_task_to_inode(task, inode);
2355		status = 1;
2356	}
2357
2358out:
2359	put_task_struct(task);
2360
2361out_notask:
2362	return status;
2363}
2364
2365static const struct dentry_operations tid_map_files_dentry_operations = {
2366	.d_revalidate	= map_files_d_revalidate,
2367	.d_delete	= pid_delete_dentry,
2368};
2369
2370static int map_files_get_link(struct dentry *dentry, struct path *path)
2371{
2372	unsigned long vm_start, vm_end;
2373	struct vm_area_struct *vma;
2374	struct task_struct *task;
2375	struct mm_struct *mm;
2376	int rc;
2377
2378	rc = -ENOENT;
2379	task = get_proc_task(d_inode(dentry));
2380	if (!task)
2381		goto out;
2382
2383	mm = get_task_mm(task);
2384	put_task_struct(task);
2385	if (!mm)
2386		goto out;
2387
2388	rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2389	if (rc)
2390		goto out_mmput;
2391
2392	rc = mmap_read_lock_killable(mm);
2393	if (rc)
2394		goto out_mmput;
2395
2396	rc = -ENOENT;
2397	vma = find_exact_vma(mm, vm_start, vm_end);
2398	if (vma && vma->vm_file) {
2399		*path = vma->vm_file->f_path;
2400		path_get(path);
2401		rc = 0;
2402	}
2403	mmap_read_unlock(mm);
2404
2405out_mmput:
2406	mmput(mm);
2407out:
2408	return rc;
2409}
2410
2411struct map_files_info {
2412	unsigned long	start;
2413	unsigned long	end;
2414	fmode_t		mode;
2415};
2416
2417/*
2418 * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due
2419 * to concerns about how the symlinks may be used to bypass permissions on
2420 * ancestor directories in the path to the file in question.
2421 */
2422static const char *
2423proc_map_files_get_link(struct dentry *dentry,
2424			struct inode *inode,
2425		        struct delayed_call *done)
2426{
2427	if (!checkpoint_restore_ns_capable(&init_user_ns))
2428		return ERR_PTR(-EPERM);
2429
2430	return proc_pid_get_link(dentry, inode, done);
2431}
2432
2433/*
2434 * Identical to proc_pid_link_inode_operations except for get_link()
2435 */
2436static const struct inode_operations proc_map_files_link_inode_operations = {
2437	.readlink	= proc_pid_readlink,
2438	.get_link	= proc_map_files_get_link,
2439	.setattr	= proc_setattr,
2440};
2441
2442static struct dentry *
2443proc_map_files_instantiate(struct dentry *dentry,
2444			   struct task_struct *task, const void *ptr)
2445{
2446	fmode_t mode = (fmode_t)(unsigned long)ptr;
2447	struct proc_inode *ei;
2448	struct inode *inode;
2449
2450	inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK |
2451				    ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2452				    ((mode & FMODE_WRITE) ? S_IWUSR : 0));
2453	if (!inode)
2454		return ERR_PTR(-ENOENT);
2455
2456	ei = PROC_I(inode);
2457	ei->op.proc_get_link = map_files_get_link;
2458
2459	inode->i_op = &proc_map_files_link_inode_operations;
2460	inode->i_size = 64;
2461
2462	d_set_d_op(dentry, &tid_map_files_dentry_operations);
2463	return d_splice_alias(inode, dentry);
2464}
2465
2466static struct dentry *proc_map_files_lookup(struct inode *dir,
2467		struct dentry *dentry, unsigned int flags)
2468{
2469	unsigned long vm_start, vm_end;
2470	struct vm_area_struct *vma;
2471	struct task_struct *task;
2472	struct dentry *result;
2473	struct mm_struct *mm;
2474
2475	result = ERR_PTR(-ENOENT);
2476	task = get_proc_task(dir);
2477	if (!task)
2478		goto out;
2479
2480	result = ERR_PTR(-EACCES);
2481	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2482		goto out_put_task;
2483
2484	result = ERR_PTR(-ENOENT);
2485	if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2486		goto out_put_task;
2487
2488	mm = get_task_mm(task);
2489	if (!mm)
2490		goto out_put_task;
2491
2492	result = ERR_PTR(-EINTR);
2493	if (mmap_read_lock_killable(mm))
2494		goto out_put_mm;
2495
2496	result = ERR_PTR(-ENOENT);
2497	vma = find_exact_vma(mm, vm_start, vm_end);
2498	if (!vma)
2499		goto out_no_vma;
2500
2501	if (vma->vm_file)
2502		result = proc_map_files_instantiate(dentry, task,
2503				(void *)(unsigned long)vma->vm_file->f_mode);
2504
2505out_no_vma:
2506	mmap_read_unlock(mm);
2507out_put_mm:
2508	mmput(mm);
2509out_put_task:
2510	put_task_struct(task);
2511out:
2512	return result;
2513}
2514
2515static const struct inode_operations proc_map_files_inode_operations = {
2516	.lookup		= proc_map_files_lookup,
2517	.permission	= proc_fd_permission,
2518	.setattr	= proc_setattr,
2519};
2520
2521static int
2522proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2523{
2524	struct vm_area_struct *vma;
2525	struct task_struct *task;
2526	struct mm_struct *mm;
2527	unsigned long nr_files, pos, i;
2528	GENRADIX(struct map_files_info) fa;
2529	struct map_files_info *p;
2530	int ret;
2531	struct vma_iterator vmi;
2532
2533	genradix_init(&fa);
2534
2535	ret = -ENOENT;
2536	task = get_proc_task(file_inode(file));
2537	if (!task)
2538		goto out;
2539
2540	ret = -EACCES;
2541	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2542		goto out_put_task;
2543
2544	ret = 0;
2545	if (!dir_emit_dots(file, ctx))
2546		goto out_put_task;
2547
2548	mm = get_task_mm(task);
2549	if (!mm)
2550		goto out_put_task;
2551
2552	ret = mmap_read_lock_killable(mm);
2553	if (ret) {
2554		mmput(mm);
2555		goto out_put_task;
2556	}
2557
2558	nr_files = 0;
2559
2560	/*
2561	 * We need two passes here:
2562	 *
2563	 *  1) Collect vmas of mapped files with mmap_lock taken
2564	 *  2) Release mmap_lock and instantiate entries
2565	 *
2566	 * otherwise we get lockdep complained, since filldir()
2567	 * routine might require mmap_lock taken in might_fault().
2568	 */
2569
2570	pos = 2;
2571	vma_iter_init(&vmi, mm, 0);
2572	for_each_vma(vmi, vma) {
2573		if (!vma->vm_file)
2574			continue;
2575		if (++pos <= ctx->pos)
2576			continue;
2577
2578		p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL);
2579		if (!p) {
2580			ret = -ENOMEM;
2581			mmap_read_unlock(mm);
2582			mmput(mm);
2583			goto out_put_task;
2584		}
2585
2586		p->start = vma->vm_start;
2587		p->end = vma->vm_end;
2588		p->mode = vma->vm_file->f_mode;
2589	}
2590	mmap_read_unlock(mm);
2591	mmput(mm);
2592
2593	for (i = 0; i < nr_files; i++) {
2594		char buf[4 * sizeof(long) + 2];	/* max: %lx-%lx\0 */
2595		unsigned int len;
2596
2597		p = genradix_ptr(&fa, i);
2598		len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
2599		if (!proc_fill_cache(file, ctx,
2600				      buf, len,
2601				      proc_map_files_instantiate,
2602				      task,
2603				      (void *)(unsigned long)p->mode))
2604			break;
2605		ctx->pos++;
2606	}
2607
2608out_put_task:
2609	put_task_struct(task);
2610out:
2611	genradix_free(&fa);
2612	return ret;
2613}
2614
2615static const struct file_operations proc_map_files_operations = {
2616	.read		= generic_read_dir,
2617	.iterate_shared	= proc_map_files_readdir,
2618	.llseek		= generic_file_llseek,
2619};
2620
2621#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2622struct timers_private {
2623	struct pid *pid;
2624	struct task_struct *task;
2625	struct sighand_struct *sighand;
2626	struct pid_namespace *ns;
2627	unsigned long flags;
2628};
2629
2630static void *timers_start(struct seq_file *m, loff_t *pos)
2631{
2632	struct timers_private *tp = m->private;
2633
2634	tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2635	if (!tp->task)
2636		return ERR_PTR(-ESRCH);
2637
2638	tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2639	if (!tp->sighand)
2640		return ERR_PTR(-ESRCH);
2641
2642	return seq_list_start(&tp->task->signal->posix_timers, *pos);
2643}
2644
2645static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2646{
2647	struct timers_private *tp = m->private;
2648	return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2649}
2650
2651static void timers_stop(struct seq_file *m, void *v)
2652{
2653	struct timers_private *tp = m->private;
2654
2655	if (tp->sighand) {
2656		unlock_task_sighand(tp->task, &tp->flags);
2657		tp->sighand = NULL;
2658	}
2659
2660	if (tp->task) {
2661		put_task_struct(tp->task);
2662		tp->task = NULL;
2663	}
2664}
2665
2666static int show_timer(struct seq_file *m, void *v)
2667{
2668	struct k_itimer *timer;
2669	struct timers_private *tp = m->private;
2670	int notify;
2671	static const char * const nstr[] = {
2672		[SIGEV_SIGNAL] = "signal",
2673		[SIGEV_NONE] = "none",
2674		[SIGEV_THREAD] = "thread",
2675	};
2676
2677	timer = list_entry((struct list_head *)v, struct k_itimer, list);
2678	notify = timer->it_sigev_notify;
2679
2680	seq_printf(m, "ID: %d\n", timer->it_id);
2681	seq_printf(m, "signal: %d/%px\n",
2682		   timer->sigq->info.si_signo,
2683		   timer->sigq->info.si_value.sival_ptr);
2684	seq_printf(m, "notify: %s/%s.%d\n",
2685		   nstr[notify & ~SIGEV_THREAD_ID],
2686		   (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2687		   pid_nr_ns(timer->it_pid, tp->ns));
2688	seq_printf(m, "ClockID: %d\n", timer->it_clock);
2689
2690	return 0;
2691}
2692
2693static const struct seq_operations proc_timers_seq_ops = {
2694	.start	= timers_start,
2695	.next	= timers_next,
2696	.stop	= timers_stop,
2697	.show	= show_timer,
2698};
2699
2700static int proc_timers_open(struct inode *inode, struct file *file)
2701{
2702	struct timers_private *tp;
2703
2704	tp = __seq_open_private(file, &proc_timers_seq_ops,
2705			sizeof(struct timers_private));
2706	if (!tp)
2707		return -ENOMEM;
2708
2709	tp->pid = proc_pid(inode);
2710	tp->ns = proc_pid_ns(inode->i_sb);
2711	return 0;
2712}
2713
2714static const struct file_operations proc_timers_operations = {
2715	.open		= proc_timers_open,
2716	.read		= seq_read,
2717	.llseek		= seq_lseek,
2718	.release	= seq_release_private,
2719};
2720#endif
2721
2722static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2723					size_t count, loff_t *offset)
2724{
2725	struct inode *inode = file_inode(file);
2726	struct task_struct *p;
2727	u64 slack_ns;
2728	int err;
2729
2730	err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2731	if (err < 0)
2732		return err;
2733
2734	p = get_proc_task(inode);
2735	if (!p)
2736		return -ESRCH;
2737
2738	if (p != current) {
2739		rcu_read_lock();
2740		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2741			rcu_read_unlock();
2742			count = -EPERM;
2743			goto out;
2744		}
2745		rcu_read_unlock();
2746
2747		err = security_task_setscheduler(p);
2748		if (err) {
2749			count = err;
2750			goto out;
2751		}
2752	}
2753
2754	task_lock(p);
2755	if (slack_ns == 0)
2756		p->timer_slack_ns = p->default_timer_slack_ns;
2757	else
2758		p->timer_slack_ns = slack_ns;
2759	task_unlock(p);
2760
2761out:
2762	put_task_struct(p);
2763
2764	return count;
2765}
2766
2767static int timerslack_ns_show(struct seq_file *m, void *v)
2768{
2769	struct inode *inode = m->private;
2770	struct task_struct *p;
2771	int err = 0;
2772
2773	p = get_proc_task(inode);
2774	if (!p)
2775		return -ESRCH;
2776
2777	if (p != current) {
2778		rcu_read_lock();
2779		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2780			rcu_read_unlock();
2781			err = -EPERM;
2782			goto out;
2783		}
2784		rcu_read_unlock();
2785
2786		err = security_task_getscheduler(p);
2787		if (err)
2788			goto out;
2789	}
2790
2791	task_lock(p);
2792	seq_printf(m, "%llu\n", p->timer_slack_ns);
2793	task_unlock(p);
2794
2795out:
2796	put_task_struct(p);
2797
2798	return err;
2799}
2800
2801static int timerslack_ns_open(struct inode *inode, struct file *filp)
2802{
2803	return single_open(filp, timerslack_ns_show, inode);
2804}
2805
2806static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2807	.open		= timerslack_ns_open,
2808	.read		= seq_read,
2809	.write		= timerslack_ns_write,
2810	.llseek		= seq_lseek,
2811	.release	= single_release,
2812};
2813
2814static struct dentry *proc_pident_instantiate(struct dentry *dentry,
2815	struct task_struct *task, const void *ptr)
2816{
2817	const struct pid_entry *p = ptr;
2818	struct inode *inode;
2819	struct proc_inode *ei;
2820
2821	inode = proc_pid_make_inode(dentry->d_sb, task, p->mode);
2822	if (!inode)
2823		return ERR_PTR(-ENOENT);
2824
2825	ei = PROC_I(inode);
2826	if (S_ISDIR(inode->i_mode))
2827		set_nlink(inode, 2);	/* Use getattr to fix if necessary */
2828	if (p->iop)
2829		inode->i_op = p->iop;
2830	if (p->fop)
2831		inode->i_fop = p->fop;
2832	ei->op = p->op;
2833	pid_update_inode(task, inode);
2834	d_set_d_op(dentry, &pid_dentry_operations);
2835	return d_splice_alias(inode, dentry);
2836}
2837
2838static struct dentry *proc_pident_lookup(struct inode *dir,
2839					 struct dentry *dentry,
2840					 const struct pid_entry *p,
2841					 const struct pid_entry *end)
2842{
2843	struct task_struct *task = get_proc_task(dir);
2844	struct dentry *res = ERR_PTR(-ENOENT);
2845
2846	if (!task)
2847		goto out_no_task;
2848
2849	/*
2850	 * Yes, it does not scale. And it should not. Don't add
2851	 * new entries into /proc/<tgid>/ without very good reasons.
2852	 */
2853	for (; p < end; p++) {
2854		if (p->len != dentry->d_name.len)
2855			continue;
2856		if (!memcmp(dentry->d_name.name, p->name, p->len)) {
2857			res = proc_pident_instantiate(dentry, task, p);
2858			break;
2859		}
2860	}
2861	put_task_struct(task);
2862out_no_task:
2863	return res;
2864}
2865
2866static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2867		const struct pid_entry *ents, unsigned int nents)
2868{
2869	struct task_struct *task = get_proc_task(file_inode(file));
2870	const struct pid_entry *p;
2871
2872	if (!task)
2873		return -ENOENT;
2874
2875	if (!dir_emit_dots(file, ctx))
2876		goto out;
2877
2878	if (ctx->pos >= nents + 2)
2879		goto out;
2880
2881	for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
2882		if (!proc_fill_cache(file, ctx, p->name, p->len,
2883				proc_pident_instantiate, task, p))
2884			break;
2885		ctx->pos++;
2886	}
2887out:
2888	put_task_struct(task);
2889	return 0;
2890}
2891
2892#ifdef CONFIG_SECURITY
2893static int proc_pid_attr_open(struct inode *inode, struct file *file)
2894{
2895	file->private_data = NULL;
2896	__mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
2897	return 0;
2898}
2899
2900static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2901				  size_t count, loff_t *ppos)
2902{
2903	struct inode * inode = file_inode(file);
2904	char *p = NULL;
2905	ssize_t length;
2906	struct task_struct *task = get_proc_task(inode);
2907
2908	if (!task)
2909		return -ESRCH;
2910
2911	length = security_getprocattr(task, PROC_I(inode)->op.lsm,
2912				      file->f_path.dentry->d_name.name,
2913				      &p);
2914	put_task_struct(task);
2915	if (length > 0)
2916		length = simple_read_from_buffer(buf, count, ppos, p, length);
2917	kfree(p);
2918	return length;
2919}
2920
2921static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2922				   size_t count, loff_t *ppos)
2923{
2924	struct inode * inode = file_inode(file);
2925	struct task_struct *task;
2926	void *page;
2927	int rv;
2928
2929	/* A task may only write when it was the opener. */
2930	if (file->private_data != current->mm)
2931		return -EPERM;
2932
2933	rcu_read_lock();
2934	task = pid_task(proc_pid(inode), PIDTYPE_PID);
2935	if (!task) {
2936		rcu_read_unlock();
2937		return -ESRCH;
2938	}
2939	/* A task may only write its own attributes. */
2940	if (current != task) {
2941		rcu_read_unlock();
2942		return -EACCES;
2943	}
2944	/* Prevent changes to overridden credentials. */
2945	if (current_cred() != current_real_cred()) {
2946		rcu_read_unlock();
2947		return -EBUSY;
2948	}
2949	rcu_read_unlock();
2950
2951	if (count > PAGE_SIZE)
2952		count = PAGE_SIZE;
2953
2954	/* No partial writes. */
2955	if (*ppos != 0)
2956		return -EINVAL;
2957
2958	page = memdup_user(buf, count);
2959	if (IS_ERR(page)) {
2960		rv = PTR_ERR(page);
2961		goto out;
2962	}
2963
2964	/* Guard against adverse ptrace interaction */
2965	rv = mutex_lock_interruptible(&current->signal->cred_guard_mutex);
2966	if (rv < 0)
2967		goto out_free;
2968
2969	rv = security_setprocattr(PROC_I(inode)->op.lsm,
2970				  file->f_path.dentry->d_name.name, page,
2971				  count);
2972	mutex_unlock(&current->signal->cred_guard_mutex);
2973out_free:
2974	kfree(page);
2975out:
2976	return rv;
2977}
2978
2979static const struct file_operations proc_pid_attr_operations = {
2980	.open		= proc_pid_attr_open,
2981	.read		= proc_pid_attr_read,
2982	.write		= proc_pid_attr_write,
2983	.llseek		= generic_file_llseek,
2984	.release	= mem_release,
2985};
2986
2987#define LSM_DIR_OPS(LSM) \
2988static int proc_##LSM##_attr_dir_iterate(struct file *filp, \
2989			     struct dir_context *ctx) \
2990{ \
2991	return proc_pident_readdir(filp, ctx, \
2992				   LSM##_attr_dir_stuff, \
2993				   ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2994} \
2995\
2996static const struct file_operations proc_##LSM##_attr_dir_ops = { \
2997	.read		= generic_read_dir, \
2998	.iterate_shared	= proc_##LSM##_attr_dir_iterate, \
2999	.llseek		= default_llseek, \
3000}; \
3001\
3002static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \
3003				struct dentry *dentry, unsigned int flags) \
3004{ \
3005	return proc_pident_lookup(dir, dentry, \
3006				  LSM##_attr_dir_stuff, \
3007				  LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \
3008} \
3009\
3010static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \
3011	.lookup		= proc_##LSM##_attr_dir_lookup, \
3012	.getattr	= pid_getattr, \
3013	.setattr	= proc_setattr, \
3014}
3015
3016#ifdef CONFIG_SECURITY_SMACK
3017static const struct pid_entry smack_attr_dir_stuff[] = {
3018	ATTR("smack", "current",	0666),
3019};
3020LSM_DIR_OPS(smack);
3021#endif
3022
3023#ifdef CONFIG_SECURITY_APPARMOR
3024static const struct pid_entry apparmor_attr_dir_stuff[] = {
3025	ATTR("apparmor", "current",	0666),
3026	ATTR("apparmor", "prev",	0444),
3027	ATTR("apparmor", "exec",	0666),
3028};
3029LSM_DIR_OPS(apparmor);
3030#endif
3031
3032static const struct pid_entry attr_dir_stuff[] = {
3033	ATTR(NULL, "current",		0666),
3034	ATTR(NULL, "prev",		0444),
3035	ATTR(NULL, "exec",		0666),
3036	ATTR(NULL, "fscreate",		0666),
3037	ATTR(NULL, "keycreate",		0666),
3038	ATTR(NULL, "sockcreate",	0666),
3039#ifdef CONFIG_SECURITY_SMACK
3040	DIR("smack",			0555,
3041	    proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops),
3042#endif
3043#ifdef CONFIG_SECURITY_APPARMOR
3044	DIR("apparmor",			0555,
3045	    proc_apparmor_attr_dir_inode_ops, proc_apparmor_attr_dir_ops),
3046#endif
3047};
3048
3049static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
3050{
3051	return proc_pident_readdir(file, ctx,
3052				   attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
3053}
3054
3055static const struct file_operations proc_attr_dir_operations = {
3056	.read		= generic_read_dir,
3057	.iterate_shared	= proc_attr_dir_readdir,
3058	.llseek		= generic_file_llseek,
3059};
3060
3061static struct dentry *proc_attr_dir_lookup(struct inode *dir,
3062				struct dentry *dentry, unsigned int flags)
3063{
3064	return proc_pident_lookup(dir, dentry,
3065				  attr_dir_stuff,
3066				  attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff));
3067}
3068
3069static const struct inode_operations proc_attr_dir_inode_operations = {
3070	.lookup		= proc_attr_dir_lookup,
3071	.getattr	= pid_getattr,
3072	.setattr	= proc_setattr,
3073};
3074
3075#endif
3076
3077#ifdef CONFIG_ELF_CORE
3078static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
3079					 size_t count, loff_t *ppos)
3080{
3081	struct task_struct *task = get_proc_task(file_inode(file));
3082	struct mm_struct *mm;
3083	char buffer[PROC_NUMBUF];
3084	size_t len;
3085	int ret;
3086
3087	if (!task)
3088		return -ESRCH;
3089
3090	ret = 0;
3091	mm = get_task_mm(task);
3092	if (mm) {
3093		len = snprintf(buffer, sizeof(buffer), "%08lx\n",
3094			       ((mm->flags & MMF_DUMP_FILTER_MASK) >>
3095				MMF_DUMP_FILTER_SHIFT));
3096		mmput(mm);
3097		ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
3098	}
3099
3100	put_task_struct(task);
3101
3102	return ret;
3103}
3104
3105static ssize_t proc_coredump_filter_write(struct file *file,
3106					  const char __user *buf,
3107					  size_t count,
3108					  loff_t *ppos)
3109{
3110	struct task_struct *task;
3111	struct mm_struct *mm;
3112	unsigned int val;
3113	int ret;
3114	int i;
3115	unsigned long mask;
3116
3117	ret = kstrtouint_from_user(buf, count, 0, &val);
3118	if (ret < 0)
3119		return ret;
3120
3121	ret = -ESRCH;
3122	task = get_proc_task(file_inode(file));
3123	if (!task)
3124		goto out_no_task;
3125
3126	mm = get_task_mm(task);
3127	if (!mm)
3128		goto out_no_mm;
3129	ret = 0;
3130
3131	for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
3132		if (val & mask)
3133			set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
3134		else
3135			clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
3136	}
3137
3138	mmput(mm);
3139 out_no_mm:
3140	put_task_struct(task);
3141 out_no_task:
3142	if (ret < 0)
3143		return ret;
3144	return count;
3145}
3146
3147static const struct file_operations proc_coredump_filter_operations = {
3148	.read		= proc_coredump_filter_read,
3149	.write		= proc_coredump_filter_write,
3150	.llseek		= generic_file_llseek,
3151};
3152#endif
3153
3154#ifdef CONFIG_TASK_IO_ACCOUNTING
3155static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
3156{
3157	struct task_io_accounting acct = task->ioac;
3158	unsigned long flags;
3159	int result;
3160
3161	result = down_read_killable(&task->signal->exec_update_lock);
3162	if (result)
3163		return result;
3164
3165	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
3166		result = -EACCES;
3167		goto out_unlock;
3168	}
3169
3170	if (whole && lock_task_sighand(task, &flags)) {
3171		struct task_struct *t = task;
3172
3173		task_io_accounting_add(&acct, &task->signal->ioac);
3174		while_each_thread(task, t)
3175			task_io_accounting_add(&acct, &t->ioac);
3176
3177		unlock_task_sighand(task, &flags);
3178	}
3179	seq_printf(m,
3180		   "rchar: %llu\n"
3181		   "wchar: %llu\n"
3182		   "syscr: %llu\n"
3183		   "syscw: %llu\n"
3184		   "read_bytes: %llu\n"
3185		   "write_bytes: %llu\n"
3186		   "cancelled_write_bytes: %llu\n",
3187		   (unsigned long long)acct.rchar,
3188		   (unsigned long long)acct.wchar,
3189		   (unsigned long long)acct.syscr,
3190		   (unsigned long long)acct.syscw,
3191		   (unsigned long long)acct.read_bytes,
3192		   (unsigned long long)acct.write_bytes,
3193		   (unsigned long long)acct.cancelled_write_bytes);
3194	result = 0;
3195
3196out_unlock:
3197	up_read(&task->signal->exec_update_lock);
3198	return result;
3199}
3200
3201static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
3202				  struct pid *pid, struct task_struct *task)
3203{
3204	return do_io_accounting(task, m, 0);
3205}
3206
3207static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
3208				   struct pid *pid, struct task_struct *task)
3209{
3210	return do_io_accounting(task, m, 1);
3211}
3212#endif /* CONFIG_TASK_IO_ACCOUNTING */
3213
3214#ifdef CONFIG_USER_NS
3215static int proc_id_map_open(struct inode *inode, struct file *file,
3216	const struct seq_operations *seq_ops)
3217{
3218	struct user_namespace *ns = NULL;
3219	struct task_struct *task;
3220	struct seq_file *seq;
3221	int ret = -EINVAL;
3222
3223	task = get_proc_task(inode);
3224	if (task) {
3225		rcu_read_lock();
3226		ns = get_user_ns(task_cred_xxx(task, user_ns));
3227		rcu_read_unlock();
3228		put_task_struct(task);
3229	}
3230	if (!ns)
3231		goto err;
3232
3233	ret = seq_open(file, seq_ops);
3234	if (ret)
3235		goto err_put_ns;
3236
3237	seq = file->private_data;
3238	seq->private = ns;
3239
3240	return 0;
3241err_put_ns:
3242	put_user_ns(ns);
3243err:
3244	return ret;
3245}
3246
3247static int proc_id_map_release(struct inode *inode, struct file *file)
3248{
3249	struct seq_file *seq = file->private_data;
3250	struct user_namespace *ns = seq->private;
3251	put_user_ns(ns);
3252	return seq_release(inode, file);
3253}
3254
3255static int proc_uid_map_open(struct inode *inode, struct file *file)
3256{
3257	return proc_id_map_open(inode, file, &proc_uid_seq_operations);
3258}
3259
3260static int proc_gid_map_open(struct inode *inode, struct file *file)
3261{
3262	return proc_id_map_open(inode, file, &proc_gid_seq_operations);
3263}
3264
3265static int proc_projid_map_open(struct inode *inode, struct file *file)
3266{
3267	return proc_id_map_open(inode, file, &proc_projid_seq_operations);
3268}
3269
3270static const struct file_operations proc_uid_map_operations = {
3271	.open		= proc_uid_map_open,
3272	.write		= proc_uid_map_write,
3273	.read		= seq_read,
3274	.llseek		= seq_lseek,
3275	.release	= proc_id_map_release,
3276};
3277
3278static const struct file_operations proc_gid_map_operations = {
3279	.open		= proc_gid_map_open,
3280	.write		= proc_gid_map_write,
3281	.read		= seq_read,
3282	.llseek		= seq_lseek,
3283	.release	= proc_id_map_release,
3284};
3285
3286static const struct file_operations proc_projid_map_operations = {
3287	.open		= proc_projid_map_open,
3288	.write		= proc_projid_map_write,
3289	.read		= seq_read,
3290	.llseek		= seq_lseek,
3291	.release	= proc_id_map_release,
3292};
3293
3294static int proc_setgroups_open(struct inode *inode, struct file *file)
3295{
3296	struct user_namespace *ns = NULL;
3297	struct task_struct *task;
3298	int ret;
3299
3300	ret = -ESRCH;
3301	task = get_proc_task(inode);
3302	if (task) {
3303		rcu_read_lock();
3304		ns = get_user_ns(task_cred_xxx(task, user_ns));
3305		rcu_read_unlock();
3306		put_task_struct(task);
3307	}
3308	if (!ns)
3309		goto err;
3310
3311	if (file->f_mode & FMODE_WRITE) {
3312		ret = -EACCES;
3313		if (!ns_capable(ns, CAP_SYS_ADMIN))
3314			goto err_put_ns;
3315	}
3316
3317	ret = single_open(file, &proc_setgroups_show, ns);
3318	if (ret)
3319		goto err_put_ns;
3320
3321	return 0;
3322err_put_ns:
3323	put_user_ns(ns);
3324err:
3325	return ret;
3326}
3327
3328static int proc_setgroups_release(struct inode *inode, struct file *file)
3329{
3330	struct seq_file *seq = file->private_data;
3331	struct user_namespace *ns = seq->private;
3332	int ret = single_release(inode, file);
3333	put_user_ns(ns);
3334	return ret;
3335}
3336
3337static const struct file_operations proc_setgroups_operations = {
3338	.open		= proc_setgroups_open,
3339	.write		= proc_setgroups_write,
3340	.read		= seq_read,
3341	.llseek		= seq_lseek,
3342	.release	= proc_setgroups_release,
3343};
3344#endif /* CONFIG_USER_NS */
3345
3346static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3347				struct pid *pid, struct task_struct *task)
3348{
3349	int err = lock_trace(task);
3350	if (!err) {
3351		seq_printf(m, "%08x\n", task->personality);
3352		unlock_trace(task);
3353	}
3354	return err;
3355}
3356
3357#ifdef CONFIG_LIVEPATCH
3358static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
3359				struct pid *pid, struct task_struct *task)
3360{
3361	seq_printf(m, "%d\n", task->patch_state);
3362	return 0;
3363}
3364#endif /* CONFIG_LIVEPATCH */
3365
3366#ifdef CONFIG_KSM
3367static int proc_pid_ksm_merging_pages(struct seq_file *m, struct pid_namespace *ns,
3368				struct pid *pid, struct task_struct *task)
3369{
3370	struct mm_struct *mm;
3371
3372	mm = get_task_mm(task);
3373	if (mm) {
3374		seq_printf(m, "%lu\n", mm->ksm_merging_pages);
3375		mmput(mm);
3376	}
3377
3378	return 0;
3379}
3380static int proc_pid_ksm_stat(struct seq_file *m, struct pid_namespace *ns,
3381				struct pid *pid, struct task_struct *task)
3382{
3383	struct mm_struct *mm;
3384
3385	mm = get_task_mm(task);
3386	if (mm) {
3387		seq_printf(m, "ksm_rmap_items %lu\n", mm->ksm_rmap_items);
3388		seq_printf(m, "ksm_zero_pages %lu\n", mm->ksm_zero_pages);
3389		seq_printf(m, "ksm_merging_pages %lu\n", mm->ksm_merging_pages);
3390		seq_printf(m, "ksm_process_profit %ld\n", ksm_process_profit(mm));
3391		mmput(mm);
3392	}
3393
3394	return 0;
3395}
3396#endif /* CONFIG_KSM */
3397
3398#ifdef CONFIG_STACKLEAK_METRICS
3399static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns,
3400				struct pid *pid, struct task_struct *task)
3401{
3402	unsigned long prev_depth = THREAD_SIZE -
3403				(task->prev_lowest_stack & (THREAD_SIZE - 1));
3404	unsigned long depth = THREAD_SIZE -
3405				(task->lowest_stack & (THREAD_SIZE - 1));
3406
3407	seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n",
3408							prev_depth, depth);
3409	return 0;
3410}
3411#endif /* CONFIG_STACKLEAK_METRICS */
3412
3413#ifdef CONFIG_ACCESS_TOKENID
3414static int proc_token_operations(struct seq_file *m, struct pid_namespace *ns,
3415				 struct pid *pid, struct task_struct *task)
3416{
3417	seq_printf(m, "%#llx %#llx\n", task->token, task->ftoken);
3418	return 0;
3419}
3420#endif /* CONFIG_ACCESS_TOKENID */
3421
3422/*
3423 * Thread groups
3424 */
3425static const struct file_operations proc_task_operations;
3426static const struct inode_operations proc_task_inode_operations;
3427
3428static const struct pid_entry tgid_base_stuff[] = {
3429	DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3430	DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3431	DIR("map_files",  S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3432	DIR("fdinfo",     S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3433	DIR("ns",	  S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3434#ifdef CONFIG_NET
3435	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3436#endif
3437	REG("environ",    S_IRUSR, proc_environ_operations),
3438	REG("auxv",       S_IRUSR, proc_auxv_operations),
3439	ONE("status",     S_IRUGO, proc_pid_status),
3440	ONE("personality", S_IRUSR, proc_pid_personality),
3441	ONE("limits",	  S_IRUGO, proc_pid_limits),
3442#ifdef CONFIG_SCHED_WALT
3443	REG("sched_init_task_load", 00644, proc_pid_sched_init_task_load_operations),
3444#endif
3445#ifdef CONFIG_SCHED_DEBUG
3446	REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3447#endif
3448#ifdef CONFIG_SCHED_AUTOGROUP
3449	REG("autogroup",  S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3450#endif
3451#ifdef CONFIG_TIME_NS
3452	REG("timens_offsets",  S_IRUGO|S_IWUSR, proc_timens_offsets_operations),
3453#endif
3454	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3455#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3456	ONE("syscall",    S_IRUSR, proc_pid_syscall),
3457#endif
3458	REG("cmdline",    S_IRUGO, proc_pid_cmdline_ops),
3459	ONE("stat",       S_IRUGO, proc_tgid_stat),
3460	ONE("statm",      S_IRUGO, proc_pid_statm),
3461	REG("maps",       S_IRUGO, proc_pid_maps_operations),
3462#ifdef CONFIG_NUMA
3463	REG("numa_maps",  S_IRUGO, proc_pid_numa_maps_operations),
3464#endif
3465	REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
3466	LNK("cwd",        proc_cwd_link),
3467	LNK("root",       proc_root_link),
3468	LNK("exe",        proc_exe_link),
3469	REG("mounts",     S_IRUGO, proc_mounts_operations),
3470	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3471	REG("mountstats", S_IRUSR, proc_mountstats_operations),
3472#ifdef CONFIG_PROC_PAGE_MONITOR
3473	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3474	REG("smaps",      S_IRUGO, proc_pid_smaps_operations),
3475	REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3476	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3477#endif
3478#ifdef CONFIG_SECURITY
3479	DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3480#endif
3481#ifdef CONFIG_KALLSYMS
3482	ONE("wchan",      S_IRUGO, proc_pid_wchan),
3483#endif
3484#ifdef CONFIG_STACKTRACE
3485	ONE("stack",      S_IRUSR, proc_pid_stack),
3486#endif
3487#ifdef CONFIG_SCHED_INFO
3488	ONE("schedstat",  S_IRUGO, proc_pid_schedstat),
3489#endif
3490#ifdef CONFIG_LATENCYTOP
3491	REG("latency",  S_IRUGO, proc_lstats_operations),
3492#endif
3493#ifdef CONFIG_PROC_PID_CPUSET
3494	ONE("cpuset",     S_IRUGO, proc_cpuset_show),
3495#endif
3496#ifdef CONFIG_CGROUPS
3497	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3498#endif
3499#ifdef CONFIG_PROC_CPU_RESCTRL
3500	ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3501#endif
3502	ONE("oom_score",  S_IRUGO, proc_oom_score),
3503	REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3504	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3505#ifdef CONFIG_AUDIT
3506	REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
3507	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3508#endif
3509#ifdef CONFIG_FAULT_INJECTION
3510	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3511	REG("fail-nth", 0644, proc_fail_nth_operations),
3512#endif
3513#ifdef CONFIG_ELF_CORE
3514	REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3515#endif
3516#ifdef CONFIG_TASK_IO_ACCOUNTING
3517	ONE("io",	S_IRUSR, proc_tgid_io_accounting),
3518#endif
3519#ifdef CONFIG_USER_NS
3520	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3521	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3522	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3523	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3524#endif
3525#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3526	REG("timers",	  S_IRUGO, proc_timers_operations),
3527#endif
3528	REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
3529#ifdef CONFIG_LIVEPATCH
3530	ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3531#endif
3532#ifdef CONFIG_STACKLEAK_METRICS
3533	ONE("stack_depth", S_IRUGO, proc_stack_depth),
3534#endif
3535#ifdef CONFIG_PROC_PID_ARCH_STATUS
3536	ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3537#endif
3538#ifdef CONFIG_ACCESS_TOKENID
3539	ONE("tokenid", S_IRUSR, proc_token_operations),
3540#endif
3541#ifdef CONFIG_SCHED_RTG
3542	REG("sched_rtg_ctrl", S_IRUGO|S_IWUGO, proc_rtg_operations),
3543#endif
3544#ifdef CONFIG_SCHED_RTG_DEBUG
3545	REG("sched_group_id", S_IRUGO|S_IWUGO, proc_pid_sched_group_id_operations),
3546#endif
3547#ifdef CONFIG_SECCOMP_CACHE_DEBUG
3548	ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache),
3549#endif
3550#ifdef CONFIG_KSM
3551	ONE("ksm_merging_pages",  S_IRUSR, proc_pid_ksm_merging_pages),
3552	ONE("ksm_stat",  S_IRUSR, proc_pid_ksm_stat),
3553#endif
3554};
3555
3556static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
3557{
3558	return proc_pident_readdir(file, ctx,
3559				   tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3560}
3561
3562static const struct file_operations proc_tgid_base_operations = {
3563	.read		= generic_read_dir,
3564	.iterate_shared	= proc_tgid_base_readdir,
3565	.llseek		= generic_file_llseek,
3566};
3567
3568struct pid *tgid_pidfd_to_pid(const struct file *file)
3569{
3570	if (file->f_op != &proc_tgid_base_operations)
3571		return ERR_PTR(-EBADF);
3572
3573	return proc_pid(file_inode(file));
3574}
3575
3576static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3577{
3578	return proc_pident_lookup(dir, dentry,
3579				  tgid_base_stuff,
3580				  tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff));
3581}
3582
3583static const struct inode_operations proc_tgid_base_inode_operations = {
3584	.lookup		= proc_tgid_base_lookup,
3585	.getattr	= pid_getattr,
3586	.setattr	= proc_setattr,
3587	.permission	= proc_pid_permission,
3588};
3589
3590/**
3591 * proc_flush_pid -  Remove dcache entries for @pid from the /proc dcache.
3592 * @pid: pid that should be flushed.
3593 *
3594 * This function walks a list of inodes (that belong to any proc
3595 * filesystem) that are attached to the pid and flushes them from
3596 * the dentry cache.
3597 *
3598 * It is safe and reasonable to cache /proc entries for a task until
3599 * that task exits.  After that they just clog up the dcache with
3600 * useless entries, possibly causing useful dcache entries to be
3601 * flushed instead.  This routine is provided to flush those useless
3602 * dcache entries when a process is reaped.
3603 *
3604 * NOTE: This routine is just an optimization so it does not guarantee
3605 *       that no dcache entries will exist after a process is reaped
3606 *       it just makes it very unlikely that any will persist.
3607 */
3608
3609void proc_flush_pid(struct pid *pid)
3610{
3611	proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock);
3612}
3613
3614static struct dentry *proc_pid_instantiate(struct dentry * dentry,
3615				   struct task_struct *task, const void *ptr)
3616{
3617	struct inode *inode;
3618
3619	inode = proc_pid_make_base_inode(dentry->d_sb, task,
3620					 S_IFDIR | S_IRUGO | S_IXUGO);
3621	if (!inode)
3622		return ERR_PTR(-ENOENT);
3623
3624	inode->i_op = &proc_tgid_base_inode_operations;
3625	inode->i_fop = &proc_tgid_base_operations;
3626	inode->i_flags|=S_IMMUTABLE;
3627
3628	set_nlink(inode, nlink_tgid);
3629	pid_update_inode(task, inode);
3630
3631	d_set_d_op(dentry, &pid_dentry_operations);
3632	return d_splice_alias(inode, dentry);
3633}
3634
3635struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags)
3636{
3637	struct task_struct *task;
3638	unsigned tgid;
3639	struct proc_fs_info *fs_info;
3640	struct pid_namespace *ns;
3641	struct dentry *result = ERR_PTR(-ENOENT);
3642
3643	tgid = name_to_int(&dentry->d_name);
3644	if (tgid == ~0U)
3645		goto out;
3646
3647	fs_info = proc_sb_info(dentry->d_sb);
3648	ns = fs_info->pid_ns;
3649	rcu_read_lock();
3650	task = find_task_by_pid_ns(tgid, ns);
3651	if (task)
3652		get_task_struct(task);
3653	rcu_read_unlock();
3654	if (!task)
3655		goto out;
3656
3657	/* Limit procfs to only ptraceable tasks */
3658	if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) {
3659		if (!has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS))
3660			goto out_put_task;
3661	}
3662
3663	result = proc_pid_instantiate(dentry, task, NULL);
3664out_put_task:
3665	put_task_struct(task);
3666out:
3667	return result;
3668}
3669
3670/*
3671 * Find the first task with tgid >= tgid
3672 *
3673 */
3674struct tgid_iter {
3675	unsigned int tgid;
3676	struct task_struct *task;
3677};
3678static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3679{
3680	struct pid *pid;
3681
3682	if (iter.task)
3683		put_task_struct(iter.task);
3684	rcu_read_lock();
3685retry:
3686	iter.task = NULL;
3687	pid = find_ge_pid(iter.tgid, ns);
3688	if (pid) {
3689		iter.tgid = pid_nr_ns(pid, ns);
3690		iter.task = pid_task(pid, PIDTYPE_TGID);
3691		if (!iter.task) {
3692			iter.tgid += 1;
3693			goto retry;
3694		}
3695		get_task_struct(iter.task);
3696	}
3697	rcu_read_unlock();
3698	return iter;
3699}
3700
3701#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3702
3703/* for the /proc/ directory itself, after non-process stuff has been done */
3704int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3705{
3706	struct tgid_iter iter;
3707	struct proc_fs_info *fs_info = proc_sb_info(file_inode(file)->i_sb);
3708	struct pid_namespace *ns = proc_pid_ns(file_inode(file)->i_sb);
3709	loff_t pos = ctx->pos;
3710
3711	if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3712		return 0;
3713
3714	if (pos == TGID_OFFSET - 2) {
3715		struct inode *inode = d_inode(fs_info->proc_self);
3716		if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3717			return 0;
3718		ctx->pos = pos = pos + 1;
3719	}
3720	if (pos == TGID_OFFSET - 1) {
3721		struct inode *inode = d_inode(fs_info->proc_thread_self);
3722		if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3723			return 0;
3724		ctx->pos = pos = pos + 1;
3725	}
3726	iter.tgid = pos - TGID_OFFSET;
3727	iter.task = NULL;
3728	for (iter = next_tgid(ns, iter);
3729	     iter.task;
3730	     iter.tgid += 1, iter = next_tgid(ns, iter)) {
3731		char name[10 + 1];
3732		unsigned int len;
3733
3734		cond_resched();
3735		if (!has_pid_permissions(fs_info, iter.task, HIDEPID_INVISIBLE))
3736			continue;
3737
3738		len = snprintf(name, sizeof(name), "%u", iter.tgid);
3739		ctx->pos = iter.tgid + TGID_OFFSET;
3740		if (!proc_fill_cache(file, ctx, name, len,
3741				     proc_pid_instantiate, iter.task, NULL)) {
3742			put_task_struct(iter.task);
3743			return 0;
3744		}
3745	}
3746	ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3747	return 0;
3748}
3749
3750/*
3751 * proc_tid_comm_permission is a special permission function exclusively
3752 * used for the node /proc/<pid>/task/<tid>/comm.
3753 * It bypasses generic permission checks in the case where a task of the same
3754 * task group attempts to access the node.
3755 * The rationale behind this is that glibc and bionic access this node for
3756 * cross thread naming (pthread_set/getname_np(!self)). However, if
3757 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3758 * which locks out the cross thread naming implementation.
3759 * This function makes sure that the node is always accessible for members of
3760 * same thread group.
3761 */
3762static int proc_tid_comm_permission(struct mnt_idmap *idmap,
3763				    struct inode *inode, int mask)
3764{
3765	bool is_same_tgroup;
3766	struct task_struct *task;
3767
3768	task = get_proc_task(inode);
3769	if (!task)
3770		return -ESRCH;
3771	is_same_tgroup = same_thread_group(current, task);
3772	put_task_struct(task);
3773
3774	if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3775		/* This file (/proc/<pid>/task/<tid>/comm) can always be
3776		 * read or written by the members of the corresponding
3777		 * thread group.
3778		 */
3779		return 0;
3780	}
3781
3782	return generic_permission(&nop_mnt_idmap, inode, mask);
3783}
3784
3785static const struct inode_operations proc_tid_comm_inode_operations = {
3786		.setattr	= proc_setattr,
3787		.permission	= proc_tid_comm_permission,
3788};
3789
3790/*
3791 * Tasks
3792 */
3793static const struct pid_entry tid_base_stuff[] = {
3794	DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3795	DIR("fdinfo",    S_IRUGO|S_IXUGO, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3796	DIR("ns",	 S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3797#ifdef CONFIG_NET
3798	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3799#endif
3800	REG("environ",   S_IRUSR, proc_environ_operations),
3801	REG("auxv",      S_IRUSR, proc_auxv_operations),
3802	ONE("status",    S_IRUGO, proc_pid_status),
3803	ONE("personality", S_IRUSR, proc_pid_personality),
3804	ONE("limits",	 S_IRUGO, proc_pid_limits),
3805#ifdef CONFIG_SCHED_DEBUG
3806	REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3807#endif
3808	NOD("comm",      S_IFREG|S_IRUGO|S_IWUSR,
3809			 &proc_tid_comm_inode_operations,
3810			 &proc_pid_set_comm_operations, {}),
3811#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3812	ONE("syscall",   S_IRUSR, proc_pid_syscall),
3813#endif
3814	REG("cmdline",   S_IRUGO, proc_pid_cmdline_ops),
3815	ONE("stat",      S_IRUGO, proc_tid_stat),
3816	ONE("statm",     S_IRUGO, proc_pid_statm),
3817	REG("maps",      S_IRUGO, proc_pid_maps_operations),
3818#ifdef CONFIG_PROC_CHILDREN
3819	REG("children",  S_IRUGO, proc_tid_children_operations),
3820#endif
3821#ifdef CONFIG_NUMA
3822	REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3823#endif
3824	REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
3825	LNK("cwd",       proc_cwd_link),
3826	LNK("root",      proc_root_link),
3827	LNK("exe",       proc_exe_link),
3828	REG("mounts",    S_IRUGO, proc_mounts_operations),
3829	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3830#ifdef CONFIG_PROC_PAGE_MONITOR
3831	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3832	REG("smaps",     S_IRUGO, proc_pid_smaps_operations),
3833	REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3834	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3835#endif
3836#ifdef CONFIG_SECURITY
3837	DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3838#endif
3839#ifdef CONFIG_KALLSYMS
3840	ONE("wchan",     S_IRUGO, proc_pid_wchan),
3841#endif
3842#ifdef CONFIG_STACKTRACE
3843	ONE("stack",      S_IRUSR, proc_pid_stack),
3844#endif
3845#ifdef CONFIG_SCHED_INFO
3846	ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3847#endif
3848#ifdef CONFIG_LATENCYTOP
3849	REG("latency",  S_IRUGO, proc_lstats_operations),
3850#endif
3851#ifdef CONFIG_PROC_PID_CPUSET
3852	ONE("cpuset",    S_IRUGO, proc_cpuset_show),
3853#endif
3854#ifdef CONFIG_CGROUPS
3855	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3856#endif
3857#ifdef CONFIG_PROC_CPU_RESCTRL
3858	ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3859#endif
3860	ONE("oom_score", S_IRUGO, proc_oom_score),
3861	REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3862	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3863#ifdef CONFIG_AUDIT
3864	REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
3865	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3866#endif
3867#ifdef CONFIG_FAULT_INJECTION
3868	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3869	REG("fail-nth", 0644, proc_fail_nth_operations),
3870#endif
3871#ifdef CONFIG_TASK_IO_ACCOUNTING
3872	ONE("io",	S_IRUSR, proc_tid_io_accounting),
3873#endif
3874#ifdef CONFIG_USER_NS
3875	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3876	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3877	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3878	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3879#endif
3880#ifdef CONFIG_LIVEPATCH
3881	ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3882#endif
3883#ifdef CONFIG_PROC_PID_ARCH_STATUS
3884	ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3885#endif
3886#ifdef CONFIG_ACCESS_TOKENID
3887	ONE("tokenid", S_IRUSR, proc_token_operations),
3888#endif
3889#ifdef CONFIG_QOS_CTRL
3890	REG("sched_qos_ctrl", S_IRUGO|S_IWUGO, proc_qos_ctrl_operations),
3891#endif
3892#ifdef CONFIG_SCHED_RTG_DEBUG
3893	REG("sched_group_id", S_IRUGO|S_IWUGO, proc_pid_sched_group_id_operations),
3894#endif
3895#ifdef CONFIG_SECCOMP_CACHE_DEBUG
3896	ONE("seccomp_cache", S_IRUSR, proc_pid_seccomp_cache),
3897#endif
3898#ifdef CONFIG_KSM
3899	ONE("ksm_merging_pages",  S_IRUSR, proc_pid_ksm_merging_pages),
3900	ONE("ksm_stat",  S_IRUSR, proc_pid_ksm_stat),
3901#endif
3902};
3903
3904static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3905{
3906	return proc_pident_readdir(file, ctx,
3907				   tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3908}
3909
3910static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3911{
3912	return proc_pident_lookup(dir, dentry,
3913				  tid_base_stuff,
3914				  tid_base_stuff + ARRAY_SIZE(tid_base_stuff));
3915}
3916
3917static const struct file_operations proc_tid_base_operations = {
3918	.read		= generic_read_dir,
3919	.iterate_shared	= proc_tid_base_readdir,
3920	.llseek		= generic_file_llseek,
3921};
3922
3923static const struct inode_operations proc_tid_base_inode_operations = {
3924	.lookup		= proc_tid_base_lookup,
3925	.getattr	= pid_getattr,
3926	.setattr	= proc_setattr,
3927};
3928
3929static struct dentry *proc_task_instantiate(struct dentry *dentry,
3930	struct task_struct *task, const void *ptr)
3931{
3932	struct inode *inode;
3933	inode = proc_pid_make_base_inode(dentry->d_sb, task,
3934					 S_IFDIR | S_IRUGO | S_IXUGO);
3935	if (!inode)
3936		return ERR_PTR(-ENOENT);
3937
3938	inode->i_op = &proc_tid_base_inode_operations;
3939	inode->i_fop = &proc_tid_base_operations;
3940	inode->i_flags |= S_IMMUTABLE;
3941
3942	set_nlink(inode, nlink_tid);
3943	pid_update_inode(task, inode);
3944
3945	d_set_d_op(dentry, &pid_dentry_operations);
3946	return d_splice_alias(inode, dentry);
3947}
3948
3949static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3950{
3951	struct task_struct *task;
3952	struct task_struct *leader = get_proc_task(dir);
3953	unsigned tid;
3954	struct proc_fs_info *fs_info;
3955	struct pid_namespace *ns;
3956	struct dentry *result = ERR_PTR(-ENOENT);
3957
3958	if (!leader)
3959		goto out_no_task;
3960
3961	tid = name_to_int(&dentry->d_name);
3962	if (tid == ~0U)
3963		goto out;
3964
3965	fs_info = proc_sb_info(dentry->d_sb);
3966	ns = fs_info->pid_ns;
3967	rcu_read_lock();
3968	task = find_task_by_pid_ns(tid, ns);
3969	if (task)
3970		get_task_struct(task);
3971	rcu_read_unlock();
3972	if (!task)
3973		goto out;
3974	if (!same_thread_group(leader, task))
3975		goto out_drop_task;
3976
3977	result = proc_task_instantiate(dentry, task, NULL);
3978out_drop_task:
3979	put_task_struct(task);
3980out:
3981	put_task_struct(leader);
3982out_no_task:
3983	return result;
3984}
3985
3986/*
3987 * Find the first tid of a thread group to return to user space.
3988 *
3989 * Usually this is just the thread group leader, but if the users
3990 * buffer was too small or there was a seek into the middle of the
3991 * directory we have more work todo.
3992 *
3993 * In the case of a short read we start with find_task_by_pid.
3994 *
3995 * In the case of a seek we start with the leader and walk nr
3996 * threads past it.
3997 */
3998static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3999					struct pid_namespace *ns)
4000{
4001	struct task_struct *pos, *task;
4002	unsigned long nr = f_pos;
4003
4004	if (nr != f_pos)	/* 32bit overflow? */
4005		return NULL;
4006
4007	rcu_read_lock();
4008	task = pid_task(pid, PIDTYPE_PID);
4009	if (!task)
4010		goto fail;
4011
4012	/* Attempt to start with the tid of a thread */
4013	if (tid && nr) {
4014		pos = find_task_by_pid_ns(tid, ns);
4015		if (pos && same_thread_group(pos, task))
4016			goto found;
4017	}
4018
4019	/* If nr exceeds the number of threads there is nothing todo */
4020	if (nr >= get_nr_threads(task))
4021		goto fail;
4022
4023	/* If we haven't found our starting place yet start
4024	 * with the leader and walk nr threads forward.
4025	 */
4026	for_each_thread(task, pos) {
4027		if (!nr--)
4028			goto found;
4029	};
4030fail:
4031	pos = NULL;
4032	goto out;
4033found:
4034	get_task_struct(pos);
4035out:
4036	rcu_read_unlock();
4037	return pos;
4038}
4039
4040/*
4041 * Find the next thread in the thread list.
4042 * Return NULL if there is an error or no next thread.
4043 *
4044 * The reference to the input task_struct is released.
4045 */
4046static struct task_struct *next_tid(struct task_struct *start)
4047{
4048	struct task_struct *pos = NULL;
4049	rcu_read_lock();
4050	if (pid_alive(start)) {
4051		pos = next_thread(start);
4052		if (thread_group_leader(pos))
4053			pos = NULL;
4054		else
4055			get_task_struct(pos);
4056	}
4057	rcu_read_unlock();
4058	put_task_struct(start);
4059	return pos;
4060}
4061
4062/* for the /proc/TGID/task/ directories */
4063static int proc_task_readdir(struct file *file, struct dir_context *ctx)
4064{
4065	struct inode *inode = file_inode(file);
4066	struct task_struct *task;
4067	struct pid_namespace *ns;
4068	int tid;
4069
4070	if (proc_inode_is_dead(inode))
4071		return -ENOENT;
4072
4073	if (!dir_emit_dots(file, ctx))
4074		return 0;
4075
4076	/* f_version caches the tgid value that the last readdir call couldn't
4077	 * return. lseek aka telldir automagically resets f_version to 0.
4078	 */
4079	ns = proc_pid_ns(inode->i_sb);
4080	tid = (int)file->f_version;
4081	file->f_version = 0;
4082	for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
4083	     task;
4084	     task = next_tid(task), ctx->pos++) {
4085		char name[10 + 1];
4086		unsigned int len;
4087
4088		tid = task_pid_nr_ns(task, ns);
4089		if (!tid)
4090			continue;	/* The task has just exited. */
4091		len = snprintf(name, sizeof(name), "%u", tid);
4092		if (!proc_fill_cache(file, ctx, name, len,
4093				proc_task_instantiate, task, NULL)) {
4094			/* returning this tgid failed, save it as the first
4095			 * pid for the next readir call */
4096			file->f_version = (u64)tid;
4097			put_task_struct(task);
4098			break;
4099		}
4100	}
4101
4102	return 0;
4103}
4104
4105static int proc_task_getattr(struct mnt_idmap *idmap,
4106			     const struct path *path, struct kstat *stat,
4107			     u32 request_mask, unsigned int query_flags)
4108{
4109	struct inode *inode = d_inode(path->dentry);
4110	struct task_struct *p = get_proc_task(inode);
4111	generic_fillattr(&nop_mnt_idmap, request_mask, inode, stat);
4112
4113	if (p) {
4114		stat->nlink += get_nr_threads(p);
4115		put_task_struct(p);
4116	}
4117
4118	return 0;
4119}
4120
4121static const struct inode_operations proc_task_inode_operations = {
4122	.lookup		= proc_task_lookup,
4123	.getattr	= proc_task_getattr,
4124	.setattr	= proc_setattr,
4125	.permission	= proc_pid_permission,
4126};
4127
4128static const struct file_operations proc_task_operations = {
4129	.read		= generic_read_dir,
4130	.iterate_shared	= proc_task_readdir,
4131	.llseek		= generic_file_llseek,
4132};
4133
4134void __init set_proc_pid_nlink(void)
4135{
4136	nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
4137	nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
4138}
4139