xref: /openbmc/linux/fs/proc/base.c (revision b296a6d5)
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 #include <linux/task_io_accounting_ops.h>
58 #include <linux/init.h>
59 #include <linux/capability.h>
60 #include <linux/file.h>
61 #include <linux/fdtable.h>
62 #include <linux/generic-radix-tree.h>
63 #include <linux/string.h>
64 #include <linux/seq_file.h>
65 #include <linux/namei.h>
66 #include <linux/mnt_namespace.h>
67 #include <linux/mm.h>
68 #include <linux/swap.h>
69 #include <linux/rcupdate.h>
70 #include <linux/kallsyms.h>
71 #include <linux/stacktrace.h>
72 #include <linux/resource.h>
73 #include <linux/module.h>
74 #include <linux/mount.h>
75 #include <linux/security.h>
76 #include <linux/ptrace.h>
77 #include <linux/tracehook.h>
78 #include <linux/printk.h>
79 #include <linux/cache.h>
80 #include <linux/cgroup.h>
81 #include <linux/cpuset.h>
82 #include <linux/audit.h>
83 #include <linux/poll.h>
84 #include <linux/nsproxy.h>
85 #include <linux/oom.h>
86 #include <linux/elf.h>
87 #include <linux/pid_namespace.h>
88 #include <linux/user_namespace.h>
89 #include <linux/fs_struct.h>
90 #include <linux/slab.h>
91 #include <linux/sched/autogroup.h>
92 #include <linux/sched/mm.h>
93 #include <linux/sched/coredump.h>
94 #include <linux/sched/debug.h>
95 #include <linux/sched/stat.h>
96 #include <linux/posix-timers.h>
97 #include <linux/time_namespace.h>
98 #include <linux/resctrl.h>
99 #include <trace/events/oom.h>
100 #include "internal.h"
101 #include "fd.h"
102 
103 #include "../../lib/kstrtox.h"
104 
105 /* NOTE:
106  *	Implementing inode permission operations in /proc is almost
107  *	certainly an error.  Permission checks need to happen during
108  *	each system call not at open time.  The reason is that most of
109  *	what we wish to check for permissions in /proc varies at runtime.
110  *
111  *	The classic example of a problem is opening file descriptors
112  *	in /proc for a task before it execs a suid executable.
113  */
114 
115 static u8 nlink_tid __ro_after_init;
116 static u8 nlink_tgid __ro_after_init;
117 
118 struct pid_entry {
119 	const char *name;
120 	unsigned int len;
121 	umode_t mode;
122 	const struct inode_operations *iop;
123 	const struct file_operations *fop;
124 	union proc_op op;
125 };
126 
127 #define NOD(NAME, MODE, IOP, FOP, OP) {			\
128 	.name = (NAME),					\
129 	.len  = sizeof(NAME) - 1,			\
130 	.mode = MODE,					\
131 	.iop  = IOP,					\
132 	.fop  = FOP,					\
133 	.op   = OP,					\
134 }
135 
136 #define DIR(NAME, MODE, iops, fops)	\
137 	NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
138 #define LNK(NAME, get_link)					\
139 	NOD(NAME, (S_IFLNK|S_IRWXUGO),				\
140 		&proc_pid_link_inode_operations, NULL,		\
141 		{ .proc_get_link = get_link } )
142 #define REG(NAME, MODE, fops)				\
143 	NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
144 #define ONE(NAME, MODE, show)				\
145 	NOD(NAME, (S_IFREG|(MODE)),			\
146 		NULL, &proc_single_file_operations,	\
147 		{ .proc_show = show } )
148 #define ATTR(LSM, NAME, MODE)				\
149 	NOD(NAME, (S_IFREG|(MODE)),			\
150 		NULL, &proc_pid_attr_operations,	\
151 		{ .lsm = LSM })
152 
153 /*
154  * Count the number of hardlinks for the pid_entry table, excluding the .
155  * and .. links.
156  */
157 static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
158 	unsigned int n)
159 {
160 	unsigned int i;
161 	unsigned int count;
162 
163 	count = 2;
164 	for (i = 0; i < n; ++i) {
165 		if (S_ISDIR(entries[i].mode))
166 			++count;
167 	}
168 
169 	return count;
170 }
171 
172 static int get_task_root(struct task_struct *task, struct path *root)
173 {
174 	int result = -ENOENT;
175 
176 	task_lock(task);
177 	if (task->fs) {
178 		get_fs_root(task->fs, root);
179 		result = 0;
180 	}
181 	task_unlock(task);
182 	return result;
183 }
184 
185 static int proc_cwd_link(struct dentry *dentry, struct path *path)
186 {
187 	struct task_struct *task = get_proc_task(d_inode(dentry));
188 	int result = -ENOENT;
189 
190 	if (task) {
191 		task_lock(task);
192 		if (task->fs) {
193 			get_fs_pwd(task->fs, path);
194 			result = 0;
195 		}
196 		task_unlock(task);
197 		put_task_struct(task);
198 	}
199 	return result;
200 }
201 
202 static int proc_root_link(struct dentry *dentry, struct path *path)
203 {
204 	struct task_struct *task = get_proc_task(d_inode(dentry));
205 	int result = -ENOENT;
206 
207 	if (task) {
208 		result = get_task_root(task, path);
209 		put_task_struct(task);
210 	}
211 	return result;
212 }
213 
214 /*
215  * If the user used setproctitle(), we just get the string from
216  * user space at arg_start, and limit it to a maximum of one page.
217  */
218 static ssize_t get_mm_proctitle(struct mm_struct *mm, char __user *buf,
219 				size_t count, unsigned long pos,
220 				unsigned long arg_start)
221 {
222 	char *page;
223 	int ret, got;
224 
225 	if (pos >= PAGE_SIZE)
226 		return 0;
227 
228 	page = (char *)__get_free_page(GFP_KERNEL);
229 	if (!page)
230 		return -ENOMEM;
231 
232 	ret = 0;
233 	got = access_remote_vm(mm, arg_start, page, PAGE_SIZE, FOLL_ANON);
234 	if (got > 0) {
235 		int len = strnlen(page, got);
236 
237 		/* Include the NUL character if it was found */
238 		if (len < got)
239 			len++;
240 
241 		if (len > pos) {
242 			len -= pos;
243 			if (len > count)
244 				len = count;
245 			len -= copy_to_user(buf, page+pos, len);
246 			if (!len)
247 				len = -EFAULT;
248 			ret = len;
249 		}
250 	}
251 	free_page((unsigned long)page);
252 	return ret;
253 }
254 
255 static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf,
256 			      size_t count, loff_t *ppos)
257 {
258 	unsigned long arg_start, arg_end, env_start, env_end;
259 	unsigned long pos, len;
260 	char *page, c;
261 
262 	/* Check if process spawned far enough to have cmdline. */
263 	if (!mm->env_end)
264 		return 0;
265 
266 	spin_lock(&mm->arg_lock);
267 	arg_start = mm->arg_start;
268 	arg_end = mm->arg_end;
269 	env_start = mm->env_start;
270 	env_end = mm->env_end;
271 	spin_unlock(&mm->arg_lock);
272 
273 	if (arg_start >= arg_end)
274 		return 0;
275 
276 	/*
277 	 * We allow setproctitle() to overwrite the argument
278 	 * strings, and overflow past the original end. But
279 	 * only when it overflows into the environment area.
280 	 */
281 	if (env_start != arg_end || env_end < env_start)
282 		env_start = env_end = arg_end;
283 	len = env_end - arg_start;
284 
285 	/* We're not going to care if "*ppos" has high bits set */
286 	pos = *ppos;
287 	if (pos >= len)
288 		return 0;
289 	if (count > len - pos)
290 		count = len - pos;
291 	if (!count)
292 		return 0;
293 
294 	/*
295 	 * Magical special case: if the argv[] end byte is not
296 	 * zero, the user has overwritten it with setproctitle(3).
297 	 *
298 	 * Possible future enhancement: do this only once when
299 	 * pos is 0, and set a flag in the 'struct file'.
300 	 */
301 	if (access_remote_vm(mm, arg_end-1, &c, 1, FOLL_ANON) == 1 && c)
302 		return get_mm_proctitle(mm, buf, count, pos, arg_start);
303 
304 	/*
305 	 * For the non-setproctitle() case we limit things strictly
306 	 * to the [arg_start, arg_end[ range.
307 	 */
308 	pos += arg_start;
309 	if (pos < arg_start || pos >= arg_end)
310 		return 0;
311 	if (count > arg_end - pos)
312 		count = arg_end - pos;
313 
314 	page = (char *)__get_free_page(GFP_KERNEL);
315 	if (!page)
316 		return -ENOMEM;
317 
318 	len = 0;
319 	while (count) {
320 		int got;
321 		size_t size = min_t(size_t, PAGE_SIZE, count);
322 
323 		got = access_remote_vm(mm, pos, page, size, FOLL_ANON);
324 		if (got <= 0)
325 			break;
326 		got -= copy_to_user(buf, page, got);
327 		if (unlikely(!got)) {
328 			if (!len)
329 				len = -EFAULT;
330 			break;
331 		}
332 		pos += got;
333 		buf += got;
334 		len += got;
335 		count -= got;
336 	}
337 
338 	free_page((unsigned long)page);
339 	return len;
340 }
341 
342 static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf,
343 				size_t count, loff_t *pos)
344 {
345 	struct mm_struct *mm;
346 	ssize_t ret;
347 
348 	mm = get_task_mm(tsk);
349 	if (!mm)
350 		return 0;
351 
352 	ret = get_mm_cmdline(mm, buf, count, pos);
353 	mmput(mm);
354 	return ret;
355 }
356 
357 static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
358 				     size_t count, loff_t *pos)
359 {
360 	struct task_struct *tsk;
361 	ssize_t ret;
362 
363 	BUG_ON(*pos < 0);
364 
365 	tsk = get_proc_task(file_inode(file));
366 	if (!tsk)
367 		return -ESRCH;
368 	ret = get_task_cmdline(tsk, buf, count, pos);
369 	put_task_struct(tsk);
370 	if (ret > 0)
371 		*pos += ret;
372 	return ret;
373 }
374 
375 static const struct file_operations proc_pid_cmdline_ops = {
376 	.read	= proc_pid_cmdline_read,
377 	.llseek	= generic_file_llseek,
378 };
379 
380 #ifdef CONFIG_KALLSYMS
381 /*
382  * Provides a wchan file via kallsyms in a proper one-value-per-file format.
383  * Returns the resolved symbol.  If that fails, simply return the address.
384  */
385 static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
386 			  struct pid *pid, struct task_struct *task)
387 {
388 	unsigned long wchan;
389 	char symname[KSYM_NAME_LEN];
390 
391 	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
392 		goto print0;
393 
394 	wchan = get_wchan(task);
395 	if (wchan && !lookup_symbol_name(wchan, symname)) {
396 		seq_puts(m, symname);
397 		return 0;
398 	}
399 
400 print0:
401 	seq_putc(m, '0');
402 	return 0;
403 }
404 #endif /* CONFIG_KALLSYMS */
405 
406 static int lock_trace(struct task_struct *task)
407 {
408 	int err = mutex_lock_killable(&task->signal->exec_update_mutex);
409 	if (err)
410 		return err;
411 	if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
412 		mutex_unlock(&task->signal->exec_update_mutex);
413 		return -EPERM;
414 	}
415 	return 0;
416 }
417 
418 static void unlock_trace(struct task_struct *task)
419 {
420 	mutex_unlock(&task->signal->exec_update_mutex);
421 }
422 
423 #ifdef CONFIG_STACKTRACE
424 
425 #define MAX_STACK_TRACE_DEPTH	64
426 
427 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
428 			  struct pid *pid, struct task_struct *task)
429 {
430 	unsigned long *entries;
431 	int err;
432 
433 	/*
434 	 * The ability to racily run the kernel stack unwinder on a running task
435 	 * and then observe the unwinder output is scary; while it is useful for
436 	 * debugging kernel issues, it can also allow an attacker to leak kernel
437 	 * stack contents.
438 	 * Doing this in a manner that is at least safe from races would require
439 	 * some work to ensure that the remote task can not be scheduled; and
440 	 * even then, this would still expose the unwinder as local attack
441 	 * surface.
442 	 * Therefore, this interface is restricted to root.
443 	 */
444 	if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN))
445 		return -EACCES;
446 
447 	entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries),
448 				GFP_KERNEL);
449 	if (!entries)
450 		return -ENOMEM;
451 
452 	err = lock_trace(task);
453 	if (!err) {
454 		unsigned int i, nr_entries;
455 
456 		nr_entries = stack_trace_save_tsk(task, entries,
457 						  MAX_STACK_TRACE_DEPTH, 0);
458 
459 		for (i = 0; i < nr_entries; i++) {
460 			seq_printf(m, "[<0>] %pB\n", (void *)entries[i]);
461 		}
462 
463 		unlock_trace(task);
464 	}
465 	kfree(entries);
466 
467 	return err;
468 }
469 #endif
470 
471 #ifdef CONFIG_SCHED_INFO
472 /*
473  * Provides /proc/PID/schedstat
474  */
475 static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
476 			      struct pid *pid, struct task_struct *task)
477 {
478 	if (unlikely(!sched_info_on()))
479 		seq_puts(m, "0 0 0\n");
480 	else
481 		seq_printf(m, "%llu %llu %lu\n",
482 		   (unsigned long long)task->se.sum_exec_runtime,
483 		   (unsigned long long)task->sched_info.run_delay,
484 		   task->sched_info.pcount);
485 
486 	return 0;
487 }
488 #endif
489 
490 #ifdef CONFIG_LATENCYTOP
491 static int lstats_show_proc(struct seq_file *m, void *v)
492 {
493 	int i;
494 	struct inode *inode = m->private;
495 	struct task_struct *task = get_proc_task(inode);
496 
497 	if (!task)
498 		return -ESRCH;
499 	seq_puts(m, "Latency Top version : v0.1\n");
500 	for (i = 0; i < LT_SAVECOUNT; i++) {
501 		struct latency_record *lr = &task->latency_record[i];
502 		if (lr->backtrace[0]) {
503 			int q;
504 			seq_printf(m, "%i %li %li",
505 				   lr->count, lr->time, lr->max);
506 			for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
507 				unsigned long bt = lr->backtrace[q];
508 
509 				if (!bt)
510 					break;
511 				seq_printf(m, " %ps", (void *)bt);
512 			}
513 			seq_putc(m, '\n');
514 		}
515 
516 	}
517 	put_task_struct(task);
518 	return 0;
519 }
520 
521 static int lstats_open(struct inode *inode, struct file *file)
522 {
523 	return single_open(file, lstats_show_proc, inode);
524 }
525 
526 static ssize_t lstats_write(struct file *file, const char __user *buf,
527 			    size_t count, loff_t *offs)
528 {
529 	struct task_struct *task = get_proc_task(file_inode(file));
530 
531 	if (!task)
532 		return -ESRCH;
533 	clear_tsk_latency_tracing(task);
534 	put_task_struct(task);
535 
536 	return count;
537 }
538 
539 static const struct file_operations proc_lstats_operations = {
540 	.open		= lstats_open,
541 	.read		= seq_read,
542 	.write		= lstats_write,
543 	.llseek		= seq_lseek,
544 	.release	= single_release,
545 };
546 
547 #endif
548 
549 static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
550 			  struct pid *pid, struct task_struct *task)
551 {
552 	unsigned long totalpages = totalram_pages() + total_swap_pages;
553 	unsigned long points = 0;
554 	long badness;
555 
556 	badness = oom_badness(task, totalpages);
557 	/*
558 	 * Special case OOM_SCORE_ADJ_MIN for all others scale the
559 	 * badness value into [0, 2000] range which we have been
560 	 * exporting for a long time so userspace might depend on it.
561 	 */
562 	if (badness != LONG_MIN)
563 		points = (1000 + badness * 1000 / (long)totalpages) * 2 / 3;
564 
565 	seq_printf(m, "%lu\n", points);
566 
567 	return 0;
568 }
569 
570 struct limit_names {
571 	const char *name;
572 	const char *unit;
573 };
574 
575 static const struct limit_names lnames[RLIM_NLIMITS] = {
576 	[RLIMIT_CPU] = {"Max cpu time", "seconds"},
577 	[RLIMIT_FSIZE] = {"Max file size", "bytes"},
578 	[RLIMIT_DATA] = {"Max data size", "bytes"},
579 	[RLIMIT_STACK] = {"Max stack size", "bytes"},
580 	[RLIMIT_CORE] = {"Max core file size", "bytes"},
581 	[RLIMIT_RSS] = {"Max resident set", "bytes"},
582 	[RLIMIT_NPROC] = {"Max processes", "processes"},
583 	[RLIMIT_NOFILE] = {"Max open files", "files"},
584 	[RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
585 	[RLIMIT_AS] = {"Max address space", "bytes"},
586 	[RLIMIT_LOCKS] = {"Max file locks", "locks"},
587 	[RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
588 	[RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
589 	[RLIMIT_NICE] = {"Max nice priority", NULL},
590 	[RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
591 	[RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
592 };
593 
594 /* Display limits for a process */
595 static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
596 			   struct pid *pid, struct task_struct *task)
597 {
598 	unsigned int i;
599 	unsigned long flags;
600 
601 	struct rlimit rlim[RLIM_NLIMITS];
602 
603 	if (!lock_task_sighand(task, &flags))
604 		return 0;
605 	memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
606 	unlock_task_sighand(task, &flags);
607 
608 	/*
609 	 * print the file header
610 	 */
611 	seq_puts(m, "Limit                     "
612 		"Soft Limit           "
613 		"Hard Limit           "
614 		"Units     \n");
615 
616 	for (i = 0; i < RLIM_NLIMITS; i++) {
617 		if (rlim[i].rlim_cur == RLIM_INFINITY)
618 			seq_printf(m, "%-25s %-20s ",
619 				   lnames[i].name, "unlimited");
620 		else
621 			seq_printf(m, "%-25s %-20lu ",
622 				   lnames[i].name, rlim[i].rlim_cur);
623 
624 		if (rlim[i].rlim_max == RLIM_INFINITY)
625 			seq_printf(m, "%-20s ", "unlimited");
626 		else
627 			seq_printf(m, "%-20lu ", rlim[i].rlim_max);
628 
629 		if (lnames[i].unit)
630 			seq_printf(m, "%-10s\n", lnames[i].unit);
631 		else
632 			seq_putc(m, '\n');
633 	}
634 
635 	return 0;
636 }
637 
638 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
639 static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
640 			    struct pid *pid, struct task_struct *task)
641 {
642 	struct syscall_info info;
643 	u64 *args = &info.data.args[0];
644 	int res;
645 
646 	res = lock_trace(task);
647 	if (res)
648 		return res;
649 
650 	if (task_current_syscall(task, &info))
651 		seq_puts(m, "running\n");
652 	else if (info.data.nr < 0)
653 		seq_printf(m, "%d 0x%llx 0x%llx\n",
654 			   info.data.nr, info.sp, info.data.instruction_pointer);
655 	else
656 		seq_printf(m,
657 		       "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n",
658 		       info.data.nr,
659 		       args[0], args[1], args[2], args[3], args[4], args[5],
660 		       info.sp, info.data.instruction_pointer);
661 	unlock_trace(task);
662 
663 	return 0;
664 }
665 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
666 
667 /************************************************************************/
668 /*                       Here the fs part begins                        */
669 /************************************************************************/
670 
671 /* permission checks */
672 static int proc_fd_access_allowed(struct inode *inode)
673 {
674 	struct task_struct *task;
675 	int allowed = 0;
676 	/* Allow access to a task's file descriptors if it is us or we
677 	 * may use ptrace attach to the process and find out that
678 	 * information.
679 	 */
680 	task = get_proc_task(inode);
681 	if (task) {
682 		allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
683 		put_task_struct(task);
684 	}
685 	return allowed;
686 }
687 
688 int proc_setattr(struct dentry *dentry, struct iattr *attr)
689 {
690 	int error;
691 	struct inode *inode = d_inode(dentry);
692 
693 	if (attr->ia_valid & ATTR_MODE)
694 		return -EPERM;
695 
696 	error = setattr_prepare(dentry, attr);
697 	if (error)
698 		return error;
699 
700 	setattr_copy(inode, attr);
701 	mark_inode_dirty(inode);
702 	return 0;
703 }
704 
705 /*
706  * May current process learn task's sched/cmdline info (for hide_pid_min=1)
707  * or euid/egid (for hide_pid_min=2)?
708  */
709 static bool has_pid_permissions(struct proc_fs_info *fs_info,
710 				 struct task_struct *task,
711 				 enum proc_hidepid hide_pid_min)
712 {
713 	/*
714 	 * If 'hidpid' mount option is set force a ptrace check,
715 	 * we indicate that we are using a filesystem syscall
716 	 * by passing PTRACE_MODE_READ_FSCREDS
717 	 */
718 	if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE)
719 		return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
720 
721 	if (fs_info->hide_pid < hide_pid_min)
722 		return true;
723 	if (in_group_p(fs_info->pid_gid))
724 		return true;
725 	return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
726 }
727 
728 
729 static int proc_pid_permission(struct inode *inode, int mask)
730 {
731 	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
732 	struct task_struct *task;
733 	bool has_perms;
734 
735 	task = get_proc_task(inode);
736 	if (!task)
737 		return -ESRCH;
738 	has_perms = has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS);
739 	put_task_struct(task);
740 
741 	if (!has_perms) {
742 		if (fs_info->hide_pid == HIDEPID_INVISIBLE) {
743 			/*
744 			 * Let's make getdents(), stat(), and open()
745 			 * consistent with each other.  If a process
746 			 * may not stat() a file, it shouldn't be seen
747 			 * in procfs at all.
748 			 */
749 			return -ENOENT;
750 		}
751 
752 		return -EPERM;
753 	}
754 	return generic_permission(inode, mask);
755 }
756 
757 
758 
759 static const struct inode_operations proc_def_inode_operations = {
760 	.setattr	= proc_setattr,
761 };
762 
763 static int proc_single_show(struct seq_file *m, void *v)
764 {
765 	struct inode *inode = m->private;
766 	struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
767 	struct pid *pid = proc_pid(inode);
768 	struct task_struct *task;
769 	int ret;
770 
771 	task = get_pid_task(pid, PIDTYPE_PID);
772 	if (!task)
773 		return -ESRCH;
774 
775 	ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
776 
777 	put_task_struct(task);
778 	return ret;
779 }
780 
781 static int proc_single_open(struct inode *inode, struct file *filp)
782 {
783 	return single_open(filp, proc_single_show, inode);
784 }
785 
786 static const struct file_operations proc_single_file_operations = {
787 	.open		= proc_single_open,
788 	.read		= seq_read,
789 	.llseek		= seq_lseek,
790 	.release	= single_release,
791 };
792 
793 
794 struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
795 {
796 	struct task_struct *task = get_proc_task(inode);
797 	struct mm_struct *mm = ERR_PTR(-ESRCH);
798 
799 	if (task) {
800 		mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
801 		put_task_struct(task);
802 
803 		if (!IS_ERR_OR_NULL(mm)) {
804 			/* ensure this mm_struct can't be freed */
805 			mmgrab(mm);
806 			/* but do not pin its memory */
807 			mmput(mm);
808 		}
809 	}
810 
811 	return mm;
812 }
813 
814 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
815 {
816 	struct mm_struct *mm = proc_mem_open(inode, mode);
817 
818 	if (IS_ERR(mm))
819 		return PTR_ERR(mm);
820 
821 	file->private_data = mm;
822 	return 0;
823 }
824 
825 static int mem_open(struct inode *inode, struct file *file)
826 {
827 	int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
828 
829 	/* OK to pass negative loff_t, we can catch out-of-range */
830 	file->f_mode |= FMODE_UNSIGNED_OFFSET;
831 
832 	return ret;
833 }
834 
835 static ssize_t mem_rw(struct file *file, char __user *buf,
836 			size_t count, loff_t *ppos, int write)
837 {
838 	struct mm_struct *mm = file->private_data;
839 	unsigned long addr = *ppos;
840 	ssize_t copied;
841 	char *page;
842 	unsigned int flags;
843 
844 	if (!mm)
845 		return 0;
846 
847 	page = (char *)__get_free_page(GFP_KERNEL);
848 	if (!page)
849 		return -ENOMEM;
850 
851 	copied = 0;
852 	if (!mmget_not_zero(mm))
853 		goto free;
854 
855 	flags = FOLL_FORCE | (write ? FOLL_WRITE : 0);
856 
857 	while (count > 0) {
858 		int this_len = min_t(int, count, PAGE_SIZE);
859 
860 		if (write && copy_from_user(page, buf, this_len)) {
861 			copied = -EFAULT;
862 			break;
863 		}
864 
865 		this_len = access_remote_vm(mm, addr, page, this_len, flags);
866 		if (!this_len) {
867 			if (!copied)
868 				copied = -EIO;
869 			break;
870 		}
871 
872 		if (!write && copy_to_user(buf, page, this_len)) {
873 			copied = -EFAULT;
874 			break;
875 		}
876 
877 		buf += this_len;
878 		addr += this_len;
879 		copied += this_len;
880 		count -= this_len;
881 	}
882 	*ppos = addr;
883 
884 	mmput(mm);
885 free:
886 	free_page((unsigned long) page);
887 	return copied;
888 }
889 
890 static ssize_t mem_read(struct file *file, char __user *buf,
891 			size_t count, loff_t *ppos)
892 {
893 	return mem_rw(file, buf, count, ppos, 0);
894 }
895 
896 static ssize_t mem_write(struct file *file, const char __user *buf,
897 			 size_t count, loff_t *ppos)
898 {
899 	return mem_rw(file, (char __user*)buf, count, ppos, 1);
900 }
901 
902 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
903 {
904 	switch (orig) {
905 	case 0:
906 		file->f_pos = offset;
907 		break;
908 	case 1:
909 		file->f_pos += offset;
910 		break;
911 	default:
912 		return -EINVAL;
913 	}
914 	force_successful_syscall_return();
915 	return file->f_pos;
916 }
917 
918 static int mem_release(struct inode *inode, struct file *file)
919 {
920 	struct mm_struct *mm = file->private_data;
921 	if (mm)
922 		mmdrop(mm);
923 	return 0;
924 }
925 
926 static const struct file_operations proc_mem_operations = {
927 	.llseek		= mem_lseek,
928 	.read		= mem_read,
929 	.write		= mem_write,
930 	.open		= mem_open,
931 	.release	= mem_release,
932 };
933 
934 static int environ_open(struct inode *inode, struct file *file)
935 {
936 	return __mem_open(inode, file, PTRACE_MODE_READ);
937 }
938 
939 static ssize_t environ_read(struct file *file, char __user *buf,
940 			size_t count, loff_t *ppos)
941 {
942 	char *page;
943 	unsigned long src = *ppos;
944 	int ret = 0;
945 	struct mm_struct *mm = file->private_data;
946 	unsigned long env_start, env_end;
947 
948 	/* Ensure the process spawned far enough to have an environment. */
949 	if (!mm || !mm->env_end)
950 		return 0;
951 
952 	page = (char *)__get_free_page(GFP_KERNEL);
953 	if (!page)
954 		return -ENOMEM;
955 
956 	ret = 0;
957 	if (!mmget_not_zero(mm))
958 		goto free;
959 
960 	spin_lock(&mm->arg_lock);
961 	env_start = mm->env_start;
962 	env_end = mm->env_end;
963 	spin_unlock(&mm->arg_lock);
964 
965 	while (count > 0) {
966 		size_t this_len, max_len;
967 		int retval;
968 
969 		if (src >= (env_end - env_start))
970 			break;
971 
972 		this_len = env_end - (env_start + src);
973 
974 		max_len = min_t(size_t, PAGE_SIZE, count);
975 		this_len = min(max_len, this_len);
976 
977 		retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON);
978 
979 		if (retval <= 0) {
980 			ret = retval;
981 			break;
982 		}
983 
984 		if (copy_to_user(buf, page, retval)) {
985 			ret = -EFAULT;
986 			break;
987 		}
988 
989 		ret += retval;
990 		src += retval;
991 		buf += retval;
992 		count -= retval;
993 	}
994 	*ppos = src;
995 	mmput(mm);
996 
997 free:
998 	free_page((unsigned long) page);
999 	return ret;
1000 }
1001 
1002 static const struct file_operations proc_environ_operations = {
1003 	.open		= environ_open,
1004 	.read		= environ_read,
1005 	.llseek		= generic_file_llseek,
1006 	.release	= mem_release,
1007 };
1008 
1009 static int auxv_open(struct inode *inode, struct file *file)
1010 {
1011 	return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
1012 }
1013 
1014 static ssize_t auxv_read(struct file *file, char __user *buf,
1015 			size_t count, loff_t *ppos)
1016 {
1017 	struct mm_struct *mm = file->private_data;
1018 	unsigned int nwords = 0;
1019 
1020 	if (!mm)
1021 		return 0;
1022 	do {
1023 		nwords += 2;
1024 	} while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
1025 	return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
1026 				       nwords * sizeof(mm->saved_auxv[0]));
1027 }
1028 
1029 static const struct file_operations proc_auxv_operations = {
1030 	.open		= auxv_open,
1031 	.read		= auxv_read,
1032 	.llseek		= generic_file_llseek,
1033 	.release	= mem_release,
1034 };
1035 
1036 static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1037 			    loff_t *ppos)
1038 {
1039 	struct task_struct *task = get_proc_task(file_inode(file));
1040 	char buffer[PROC_NUMBUF];
1041 	int oom_adj = OOM_ADJUST_MIN;
1042 	size_t len;
1043 
1044 	if (!task)
1045 		return -ESRCH;
1046 	if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1047 		oom_adj = OOM_ADJUST_MAX;
1048 	else
1049 		oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1050 			  OOM_SCORE_ADJ_MAX;
1051 	put_task_struct(task);
1052 	len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1053 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1054 }
1055 
1056 static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1057 {
1058 	struct mm_struct *mm = NULL;
1059 	struct task_struct *task;
1060 	int err = 0;
1061 
1062 	task = get_proc_task(file_inode(file));
1063 	if (!task)
1064 		return -ESRCH;
1065 
1066 	mutex_lock(&oom_adj_mutex);
1067 	if (legacy) {
1068 		if (oom_adj < task->signal->oom_score_adj &&
1069 				!capable(CAP_SYS_RESOURCE)) {
1070 			err = -EACCES;
1071 			goto err_unlock;
1072 		}
1073 		/*
1074 		 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1075 		 * /proc/pid/oom_score_adj instead.
1076 		 */
1077 		pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1078 			  current->comm, task_pid_nr(current), task_pid_nr(task),
1079 			  task_pid_nr(task));
1080 	} else {
1081 		if ((short)oom_adj < task->signal->oom_score_adj_min &&
1082 				!capable(CAP_SYS_RESOURCE)) {
1083 			err = -EACCES;
1084 			goto err_unlock;
1085 		}
1086 	}
1087 
1088 	/*
1089 	 * Make sure we will check other processes sharing the mm if this is
1090 	 * not vfrok which wants its own oom_score_adj.
1091 	 * pin the mm so it doesn't go away and get reused after task_unlock
1092 	 */
1093 	if (!task->vfork_done) {
1094 		struct task_struct *p = find_lock_task_mm(task);
1095 
1096 		if (p) {
1097 			if (test_bit(MMF_MULTIPROCESS, &p->mm->flags)) {
1098 				mm = p->mm;
1099 				mmgrab(mm);
1100 			}
1101 			task_unlock(p);
1102 		}
1103 	}
1104 
1105 	task->signal->oom_score_adj = oom_adj;
1106 	if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1107 		task->signal->oom_score_adj_min = (short)oom_adj;
1108 	trace_oom_score_adj_update(task);
1109 
1110 	if (mm) {
1111 		struct task_struct *p;
1112 
1113 		rcu_read_lock();
1114 		for_each_process(p) {
1115 			if (same_thread_group(task, p))
1116 				continue;
1117 
1118 			/* do not touch kernel threads or the global init */
1119 			if (p->flags & PF_KTHREAD || is_global_init(p))
1120 				continue;
1121 
1122 			task_lock(p);
1123 			if (!p->vfork_done && process_shares_mm(p, mm)) {
1124 				p->signal->oom_score_adj = oom_adj;
1125 				if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1126 					p->signal->oom_score_adj_min = (short)oom_adj;
1127 			}
1128 			task_unlock(p);
1129 		}
1130 		rcu_read_unlock();
1131 		mmdrop(mm);
1132 	}
1133 err_unlock:
1134 	mutex_unlock(&oom_adj_mutex);
1135 	put_task_struct(task);
1136 	return err;
1137 }
1138 
1139 /*
1140  * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1141  * kernels.  The effective policy is defined by oom_score_adj, which has a
1142  * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1143  * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1144  * Processes that become oom disabled via oom_adj will still be oom disabled
1145  * with this implementation.
1146  *
1147  * oom_adj cannot be removed since existing userspace binaries use it.
1148  */
1149 static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1150 			     size_t count, loff_t *ppos)
1151 {
1152 	char buffer[PROC_NUMBUF];
1153 	int oom_adj;
1154 	int err;
1155 
1156 	memset(buffer, 0, sizeof(buffer));
1157 	if (count > sizeof(buffer) - 1)
1158 		count = sizeof(buffer) - 1;
1159 	if (copy_from_user(buffer, buf, count)) {
1160 		err = -EFAULT;
1161 		goto out;
1162 	}
1163 
1164 	err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1165 	if (err)
1166 		goto out;
1167 	if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1168 	     oom_adj != OOM_DISABLE) {
1169 		err = -EINVAL;
1170 		goto out;
1171 	}
1172 
1173 	/*
1174 	 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1175 	 * value is always attainable.
1176 	 */
1177 	if (oom_adj == OOM_ADJUST_MAX)
1178 		oom_adj = OOM_SCORE_ADJ_MAX;
1179 	else
1180 		oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1181 
1182 	err = __set_oom_adj(file, oom_adj, true);
1183 out:
1184 	return err < 0 ? err : count;
1185 }
1186 
1187 static const struct file_operations proc_oom_adj_operations = {
1188 	.read		= oom_adj_read,
1189 	.write		= oom_adj_write,
1190 	.llseek		= generic_file_llseek,
1191 };
1192 
1193 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1194 					size_t count, loff_t *ppos)
1195 {
1196 	struct task_struct *task = get_proc_task(file_inode(file));
1197 	char buffer[PROC_NUMBUF];
1198 	short oom_score_adj = OOM_SCORE_ADJ_MIN;
1199 	size_t len;
1200 
1201 	if (!task)
1202 		return -ESRCH;
1203 	oom_score_adj = task->signal->oom_score_adj;
1204 	put_task_struct(task);
1205 	len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
1206 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1207 }
1208 
1209 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1210 					size_t count, loff_t *ppos)
1211 {
1212 	char buffer[PROC_NUMBUF];
1213 	int oom_score_adj;
1214 	int err;
1215 
1216 	memset(buffer, 0, sizeof(buffer));
1217 	if (count > sizeof(buffer) - 1)
1218 		count = sizeof(buffer) - 1;
1219 	if (copy_from_user(buffer, buf, count)) {
1220 		err = -EFAULT;
1221 		goto out;
1222 	}
1223 
1224 	err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1225 	if (err)
1226 		goto out;
1227 	if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1228 			oom_score_adj > OOM_SCORE_ADJ_MAX) {
1229 		err = -EINVAL;
1230 		goto out;
1231 	}
1232 
1233 	err = __set_oom_adj(file, oom_score_adj, false);
1234 out:
1235 	return err < 0 ? err : count;
1236 }
1237 
1238 static const struct file_operations proc_oom_score_adj_operations = {
1239 	.read		= oom_score_adj_read,
1240 	.write		= oom_score_adj_write,
1241 	.llseek		= default_llseek,
1242 };
1243 
1244 #ifdef CONFIG_AUDIT
1245 #define TMPBUFLEN 11
1246 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1247 				  size_t count, loff_t *ppos)
1248 {
1249 	struct inode * inode = file_inode(file);
1250 	struct task_struct *task = get_proc_task(inode);
1251 	ssize_t length;
1252 	char tmpbuf[TMPBUFLEN];
1253 
1254 	if (!task)
1255 		return -ESRCH;
1256 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1257 			   from_kuid(file->f_cred->user_ns,
1258 				     audit_get_loginuid(task)));
1259 	put_task_struct(task);
1260 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1261 }
1262 
1263 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1264 				   size_t count, loff_t *ppos)
1265 {
1266 	struct inode * inode = file_inode(file);
1267 	uid_t loginuid;
1268 	kuid_t kloginuid;
1269 	int rv;
1270 
1271 	rcu_read_lock();
1272 	if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1273 		rcu_read_unlock();
1274 		return -EPERM;
1275 	}
1276 	rcu_read_unlock();
1277 
1278 	if (*ppos != 0) {
1279 		/* No partial writes. */
1280 		return -EINVAL;
1281 	}
1282 
1283 	rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1284 	if (rv < 0)
1285 		return rv;
1286 
1287 	/* is userspace tring to explicitly UNSET the loginuid? */
1288 	if (loginuid == AUDIT_UID_UNSET) {
1289 		kloginuid = INVALID_UID;
1290 	} else {
1291 		kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
1292 		if (!uid_valid(kloginuid))
1293 			return -EINVAL;
1294 	}
1295 
1296 	rv = audit_set_loginuid(kloginuid);
1297 	if (rv < 0)
1298 		return rv;
1299 	return count;
1300 }
1301 
1302 static const struct file_operations proc_loginuid_operations = {
1303 	.read		= proc_loginuid_read,
1304 	.write		= proc_loginuid_write,
1305 	.llseek		= generic_file_llseek,
1306 };
1307 
1308 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1309 				  size_t count, loff_t *ppos)
1310 {
1311 	struct inode * inode = file_inode(file);
1312 	struct task_struct *task = get_proc_task(inode);
1313 	ssize_t length;
1314 	char tmpbuf[TMPBUFLEN];
1315 
1316 	if (!task)
1317 		return -ESRCH;
1318 	length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1319 				audit_get_sessionid(task));
1320 	put_task_struct(task);
1321 	return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1322 }
1323 
1324 static const struct file_operations proc_sessionid_operations = {
1325 	.read		= proc_sessionid_read,
1326 	.llseek		= generic_file_llseek,
1327 };
1328 #endif
1329 
1330 #ifdef CONFIG_FAULT_INJECTION
1331 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1332 				      size_t count, loff_t *ppos)
1333 {
1334 	struct task_struct *task = get_proc_task(file_inode(file));
1335 	char buffer[PROC_NUMBUF];
1336 	size_t len;
1337 	int make_it_fail;
1338 
1339 	if (!task)
1340 		return -ESRCH;
1341 	make_it_fail = task->make_it_fail;
1342 	put_task_struct(task);
1343 
1344 	len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1345 
1346 	return simple_read_from_buffer(buf, count, ppos, buffer, len);
1347 }
1348 
1349 static ssize_t proc_fault_inject_write(struct file * file,
1350 			const char __user * buf, size_t count, loff_t *ppos)
1351 {
1352 	struct task_struct *task;
1353 	char buffer[PROC_NUMBUF];
1354 	int make_it_fail;
1355 	int rv;
1356 
1357 	if (!capable(CAP_SYS_RESOURCE))
1358 		return -EPERM;
1359 	memset(buffer, 0, sizeof(buffer));
1360 	if (count > sizeof(buffer) - 1)
1361 		count = sizeof(buffer) - 1;
1362 	if (copy_from_user(buffer, buf, count))
1363 		return -EFAULT;
1364 	rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1365 	if (rv < 0)
1366 		return rv;
1367 	if (make_it_fail < 0 || make_it_fail > 1)
1368 		return -EINVAL;
1369 
1370 	task = get_proc_task(file_inode(file));
1371 	if (!task)
1372 		return -ESRCH;
1373 	task->make_it_fail = make_it_fail;
1374 	put_task_struct(task);
1375 
1376 	return count;
1377 }
1378 
1379 static const struct file_operations proc_fault_inject_operations = {
1380 	.read		= proc_fault_inject_read,
1381 	.write		= proc_fault_inject_write,
1382 	.llseek		= generic_file_llseek,
1383 };
1384 
1385 static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1386 				   size_t count, loff_t *ppos)
1387 {
1388 	struct task_struct *task;
1389 	int err;
1390 	unsigned int n;
1391 
1392 	err = kstrtouint_from_user(buf, count, 0, &n);
1393 	if (err)
1394 		return err;
1395 
1396 	task = get_proc_task(file_inode(file));
1397 	if (!task)
1398 		return -ESRCH;
1399 	task->fail_nth = n;
1400 	put_task_struct(task);
1401 
1402 	return count;
1403 }
1404 
1405 static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1406 				  size_t count, loff_t *ppos)
1407 {
1408 	struct task_struct *task;
1409 	char numbuf[PROC_NUMBUF];
1410 	ssize_t len;
1411 
1412 	task = get_proc_task(file_inode(file));
1413 	if (!task)
1414 		return -ESRCH;
1415 	len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth);
1416 	put_task_struct(task);
1417 	return simple_read_from_buffer(buf, count, ppos, numbuf, len);
1418 }
1419 
1420 static const struct file_operations proc_fail_nth_operations = {
1421 	.read		= proc_fail_nth_read,
1422 	.write		= proc_fail_nth_write,
1423 };
1424 #endif
1425 
1426 
1427 #ifdef CONFIG_SCHED_DEBUG
1428 /*
1429  * Print out various scheduling related per-task fields:
1430  */
1431 static int sched_show(struct seq_file *m, void *v)
1432 {
1433 	struct inode *inode = m->private;
1434 	struct pid_namespace *ns = proc_pid_ns(inode->i_sb);
1435 	struct task_struct *p;
1436 
1437 	p = get_proc_task(inode);
1438 	if (!p)
1439 		return -ESRCH;
1440 	proc_sched_show_task(p, ns, m);
1441 
1442 	put_task_struct(p);
1443 
1444 	return 0;
1445 }
1446 
1447 static ssize_t
1448 sched_write(struct file *file, const char __user *buf,
1449 	    size_t count, loff_t *offset)
1450 {
1451 	struct inode *inode = file_inode(file);
1452 	struct task_struct *p;
1453 
1454 	p = get_proc_task(inode);
1455 	if (!p)
1456 		return -ESRCH;
1457 	proc_sched_set_task(p);
1458 
1459 	put_task_struct(p);
1460 
1461 	return count;
1462 }
1463 
1464 static int sched_open(struct inode *inode, struct file *filp)
1465 {
1466 	return single_open(filp, sched_show, inode);
1467 }
1468 
1469 static const struct file_operations proc_pid_sched_operations = {
1470 	.open		= sched_open,
1471 	.read		= seq_read,
1472 	.write		= sched_write,
1473 	.llseek		= seq_lseek,
1474 	.release	= single_release,
1475 };
1476 
1477 #endif
1478 
1479 #ifdef CONFIG_SCHED_AUTOGROUP
1480 /*
1481  * Print out autogroup related information:
1482  */
1483 static int sched_autogroup_show(struct seq_file *m, void *v)
1484 {
1485 	struct inode *inode = m->private;
1486 	struct task_struct *p;
1487 
1488 	p = get_proc_task(inode);
1489 	if (!p)
1490 		return -ESRCH;
1491 	proc_sched_autogroup_show_task(p, m);
1492 
1493 	put_task_struct(p);
1494 
1495 	return 0;
1496 }
1497 
1498 static ssize_t
1499 sched_autogroup_write(struct file *file, const char __user *buf,
1500 	    size_t count, loff_t *offset)
1501 {
1502 	struct inode *inode = file_inode(file);
1503 	struct task_struct *p;
1504 	char buffer[PROC_NUMBUF];
1505 	int nice;
1506 	int err;
1507 
1508 	memset(buffer, 0, sizeof(buffer));
1509 	if (count > sizeof(buffer) - 1)
1510 		count = sizeof(buffer) - 1;
1511 	if (copy_from_user(buffer, buf, count))
1512 		return -EFAULT;
1513 
1514 	err = kstrtoint(strstrip(buffer), 0, &nice);
1515 	if (err < 0)
1516 		return err;
1517 
1518 	p = get_proc_task(inode);
1519 	if (!p)
1520 		return -ESRCH;
1521 
1522 	err = proc_sched_autogroup_set_nice(p, nice);
1523 	if (err)
1524 		count = err;
1525 
1526 	put_task_struct(p);
1527 
1528 	return count;
1529 }
1530 
1531 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1532 {
1533 	int ret;
1534 
1535 	ret = single_open(filp, sched_autogroup_show, NULL);
1536 	if (!ret) {
1537 		struct seq_file *m = filp->private_data;
1538 
1539 		m->private = inode;
1540 	}
1541 	return ret;
1542 }
1543 
1544 static const struct file_operations proc_pid_sched_autogroup_operations = {
1545 	.open		= sched_autogroup_open,
1546 	.read		= seq_read,
1547 	.write		= sched_autogroup_write,
1548 	.llseek		= seq_lseek,
1549 	.release	= single_release,
1550 };
1551 
1552 #endif /* CONFIG_SCHED_AUTOGROUP */
1553 
1554 #ifdef CONFIG_TIME_NS
1555 static int timens_offsets_show(struct seq_file *m, void *v)
1556 {
1557 	struct task_struct *p;
1558 
1559 	p = get_proc_task(file_inode(m->file));
1560 	if (!p)
1561 		return -ESRCH;
1562 	proc_timens_show_offsets(p, m);
1563 
1564 	put_task_struct(p);
1565 
1566 	return 0;
1567 }
1568 
1569 static ssize_t timens_offsets_write(struct file *file, const char __user *buf,
1570 				    size_t count, loff_t *ppos)
1571 {
1572 	struct inode *inode = file_inode(file);
1573 	struct proc_timens_offset offsets[2];
1574 	char *kbuf = NULL, *pos, *next_line;
1575 	struct task_struct *p;
1576 	int ret, noffsets;
1577 
1578 	/* Only allow < page size writes at the beginning of the file */
1579 	if ((*ppos != 0) || (count >= PAGE_SIZE))
1580 		return -EINVAL;
1581 
1582 	/* Slurp in the user data */
1583 	kbuf = memdup_user_nul(buf, count);
1584 	if (IS_ERR(kbuf))
1585 		return PTR_ERR(kbuf);
1586 
1587 	/* Parse the user data */
1588 	ret = -EINVAL;
1589 	noffsets = 0;
1590 	for (pos = kbuf; pos; pos = next_line) {
1591 		struct proc_timens_offset *off = &offsets[noffsets];
1592 		char clock[10];
1593 		int err;
1594 
1595 		/* Find the end of line and ensure we don't look past it */
1596 		next_line = strchr(pos, '\n');
1597 		if (next_line) {
1598 			*next_line = '\0';
1599 			next_line++;
1600 			if (*next_line == '\0')
1601 				next_line = NULL;
1602 		}
1603 
1604 		err = sscanf(pos, "%9s %lld %lu", clock,
1605 				&off->val.tv_sec, &off->val.tv_nsec);
1606 		if (err != 3 || off->val.tv_nsec >= NSEC_PER_SEC)
1607 			goto out;
1608 
1609 		clock[sizeof(clock) - 1] = 0;
1610 		if (strcmp(clock, "monotonic") == 0 ||
1611 		    strcmp(clock, __stringify(CLOCK_MONOTONIC)) == 0)
1612 			off->clockid = CLOCK_MONOTONIC;
1613 		else if (strcmp(clock, "boottime") == 0 ||
1614 			 strcmp(clock, __stringify(CLOCK_BOOTTIME)) == 0)
1615 			off->clockid = CLOCK_BOOTTIME;
1616 		else
1617 			goto out;
1618 
1619 		noffsets++;
1620 		if (noffsets == ARRAY_SIZE(offsets)) {
1621 			if (next_line)
1622 				count = next_line - kbuf;
1623 			break;
1624 		}
1625 	}
1626 
1627 	ret = -ESRCH;
1628 	p = get_proc_task(inode);
1629 	if (!p)
1630 		goto out;
1631 	ret = proc_timens_set_offset(file, p, offsets, noffsets);
1632 	put_task_struct(p);
1633 	if (ret)
1634 		goto out;
1635 
1636 	ret = count;
1637 out:
1638 	kfree(kbuf);
1639 	return ret;
1640 }
1641 
1642 static int timens_offsets_open(struct inode *inode, struct file *filp)
1643 {
1644 	return single_open(filp, timens_offsets_show, inode);
1645 }
1646 
1647 static const struct file_operations proc_timens_offsets_operations = {
1648 	.open		= timens_offsets_open,
1649 	.read		= seq_read,
1650 	.write		= timens_offsets_write,
1651 	.llseek		= seq_lseek,
1652 	.release	= single_release,
1653 };
1654 #endif /* CONFIG_TIME_NS */
1655 
1656 static ssize_t comm_write(struct file *file, const char __user *buf,
1657 				size_t count, loff_t *offset)
1658 {
1659 	struct inode *inode = file_inode(file);
1660 	struct task_struct *p;
1661 	char buffer[TASK_COMM_LEN];
1662 	const size_t maxlen = sizeof(buffer) - 1;
1663 
1664 	memset(buffer, 0, sizeof(buffer));
1665 	if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
1666 		return -EFAULT;
1667 
1668 	p = get_proc_task(inode);
1669 	if (!p)
1670 		return -ESRCH;
1671 
1672 	if (same_thread_group(current, p))
1673 		set_task_comm(p, buffer);
1674 	else
1675 		count = -EINVAL;
1676 
1677 	put_task_struct(p);
1678 
1679 	return count;
1680 }
1681 
1682 static int comm_show(struct seq_file *m, void *v)
1683 {
1684 	struct inode *inode = m->private;
1685 	struct task_struct *p;
1686 
1687 	p = get_proc_task(inode);
1688 	if (!p)
1689 		return -ESRCH;
1690 
1691 	proc_task_name(m, p, false);
1692 	seq_putc(m, '\n');
1693 
1694 	put_task_struct(p);
1695 
1696 	return 0;
1697 }
1698 
1699 static int comm_open(struct inode *inode, struct file *filp)
1700 {
1701 	return single_open(filp, comm_show, inode);
1702 }
1703 
1704 static const struct file_operations proc_pid_set_comm_operations = {
1705 	.open		= comm_open,
1706 	.read		= seq_read,
1707 	.write		= comm_write,
1708 	.llseek		= seq_lseek,
1709 	.release	= single_release,
1710 };
1711 
1712 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1713 {
1714 	struct task_struct *task;
1715 	struct file *exe_file;
1716 
1717 	task = get_proc_task(d_inode(dentry));
1718 	if (!task)
1719 		return -ENOENT;
1720 	exe_file = get_task_exe_file(task);
1721 	put_task_struct(task);
1722 	if (exe_file) {
1723 		*exe_path = exe_file->f_path;
1724 		path_get(&exe_file->f_path);
1725 		fput(exe_file);
1726 		return 0;
1727 	} else
1728 		return -ENOENT;
1729 }
1730 
1731 static const char *proc_pid_get_link(struct dentry *dentry,
1732 				     struct inode *inode,
1733 				     struct delayed_call *done)
1734 {
1735 	struct path path;
1736 	int error = -EACCES;
1737 
1738 	if (!dentry)
1739 		return ERR_PTR(-ECHILD);
1740 
1741 	/* Are we allowed to snoop on the tasks file descriptors? */
1742 	if (!proc_fd_access_allowed(inode))
1743 		goto out;
1744 
1745 	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1746 	if (error)
1747 		goto out;
1748 
1749 	error = nd_jump_link(&path);
1750 out:
1751 	return ERR_PTR(error);
1752 }
1753 
1754 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1755 {
1756 	char *tmp = (char *)__get_free_page(GFP_KERNEL);
1757 	char *pathname;
1758 	int len;
1759 
1760 	if (!tmp)
1761 		return -ENOMEM;
1762 
1763 	pathname = d_path(path, tmp, PAGE_SIZE);
1764 	len = PTR_ERR(pathname);
1765 	if (IS_ERR(pathname))
1766 		goto out;
1767 	len = tmp + PAGE_SIZE - 1 - pathname;
1768 
1769 	if (len > buflen)
1770 		len = buflen;
1771 	if (copy_to_user(buffer, pathname, len))
1772 		len = -EFAULT;
1773  out:
1774 	free_page((unsigned long)tmp);
1775 	return len;
1776 }
1777 
1778 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1779 {
1780 	int error = -EACCES;
1781 	struct inode *inode = d_inode(dentry);
1782 	struct path path;
1783 
1784 	/* Are we allowed to snoop on the tasks file descriptors? */
1785 	if (!proc_fd_access_allowed(inode))
1786 		goto out;
1787 
1788 	error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1789 	if (error)
1790 		goto out;
1791 
1792 	error = do_proc_readlink(&path, buffer, buflen);
1793 	path_put(&path);
1794 out:
1795 	return error;
1796 }
1797 
1798 const struct inode_operations proc_pid_link_inode_operations = {
1799 	.readlink	= proc_pid_readlink,
1800 	.get_link	= proc_pid_get_link,
1801 	.setattr	= proc_setattr,
1802 };
1803 
1804 
1805 /* building an inode */
1806 
1807 void task_dump_owner(struct task_struct *task, umode_t mode,
1808 		     kuid_t *ruid, kgid_t *rgid)
1809 {
1810 	/* Depending on the state of dumpable compute who should own a
1811 	 * proc file for a task.
1812 	 */
1813 	const struct cred *cred;
1814 	kuid_t uid;
1815 	kgid_t gid;
1816 
1817 	if (unlikely(task->flags & PF_KTHREAD)) {
1818 		*ruid = GLOBAL_ROOT_UID;
1819 		*rgid = GLOBAL_ROOT_GID;
1820 		return;
1821 	}
1822 
1823 	/* Default to the tasks effective ownership */
1824 	rcu_read_lock();
1825 	cred = __task_cred(task);
1826 	uid = cred->euid;
1827 	gid = cred->egid;
1828 	rcu_read_unlock();
1829 
1830 	/*
1831 	 * Before the /proc/pid/status file was created the only way to read
1832 	 * the effective uid of a /process was to stat /proc/pid.  Reading
1833 	 * /proc/pid/status is slow enough that procps and other packages
1834 	 * kept stating /proc/pid.  To keep the rules in /proc simple I have
1835 	 * made this apply to all per process world readable and executable
1836 	 * directories.
1837 	 */
1838 	if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1839 		struct mm_struct *mm;
1840 		task_lock(task);
1841 		mm = task->mm;
1842 		/* Make non-dumpable tasks owned by some root */
1843 		if (mm) {
1844 			if (get_dumpable(mm) != SUID_DUMP_USER) {
1845 				struct user_namespace *user_ns = mm->user_ns;
1846 
1847 				uid = make_kuid(user_ns, 0);
1848 				if (!uid_valid(uid))
1849 					uid = GLOBAL_ROOT_UID;
1850 
1851 				gid = make_kgid(user_ns, 0);
1852 				if (!gid_valid(gid))
1853 					gid = GLOBAL_ROOT_GID;
1854 			}
1855 		} else {
1856 			uid = GLOBAL_ROOT_UID;
1857 			gid = GLOBAL_ROOT_GID;
1858 		}
1859 		task_unlock(task);
1860 	}
1861 	*ruid = uid;
1862 	*rgid = gid;
1863 }
1864 
1865 void proc_pid_evict_inode(struct proc_inode *ei)
1866 {
1867 	struct pid *pid = ei->pid;
1868 
1869 	if (S_ISDIR(ei->vfs_inode.i_mode)) {
1870 		spin_lock(&pid->lock);
1871 		hlist_del_init_rcu(&ei->sibling_inodes);
1872 		spin_unlock(&pid->lock);
1873 	}
1874 
1875 	put_pid(pid);
1876 }
1877 
1878 struct inode *proc_pid_make_inode(struct super_block * sb,
1879 				  struct task_struct *task, umode_t mode)
1880 {
1881 	struct inode * inode;
1882 	struct proc_inode *ei;
1883 	struct pid *pid;
1884 
1885 	/* We need a new inode */
1886 
1887 	inode = new_inode(sb);
1888 	if (!inode)
1889 		goto out;
1890 
1891 	/* Common stuff */
1892 	ei = PROC_I(inode);
1893 	inode->i_mode = mode;
1894 	inode->i_ino = get_next_ino();
1895 	inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
1896 	inode->i_op = &proc_def_inode_operations;
1897 
1898 	/*
1899 	 * grab the reference to task.
1900 	 */
1901 	pid = get_task_pid(task, PIDTYPE_PID);
1902 	if (!pid)
1903 		goto out_unlock;
1904 
1905 	/* Let the pid remember us for quick removal */
1906 	ei->pid = pid;
1907 	if (S_ISDIR(mode)) {
1908 		spin_lock(&pid->lock);
1909 		hlist_add_head_rcu(&ei->sibling_inodes, &pid->inodes);
1910 		spin_unlock(&pid->lock);
1911 	}
1912 
1913 	task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1914 	security_task_to_inode(task, inode);
1915 
1916 out:
1917 	return inode;
1918 
1919 out_unlock:
1920 	iput(inode);
1921 	return NULL;
1922 }
1923 
1924 int pid_getattr(const struct path *path, struct kstat *stat,
1925 		u32 request_mask, unsigned int query_flags)
1926 {
1927 	struct inode *inode = d_inode(path->dentry);
1928 	struct proc_fs_info *fs_info = proc_sb_info(inode->i_sb);
1929 	struct task_struct *task;
1930 
1931 	generic_fillattr(inode, stat);
1932 
1933 	stat->uid = GLOBAL_ROOT_UID;
1934 	stat->gid = GLOBAL_ROOT_GID;
1935 	rcu_read_lock();
1936 	task = pid_task(proc_pid(inode), PIDTYPE_PID);
1937 	if (task) {
1938 		if (!has_pid_permissions(fs_info, task, HIDEPID_INVISIBLE)) {
1939 			rcu_read_unlock();
1940 			/*
1941 			 * This doesn't prevent learning whether PID exists,
1942 			 * it only makes getattr() consistent with readdir().
1943 			 */
1944 			return -ENOENT;
1945 		}
1946 		task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
1947 	}
1948 	rcu_read_unlock();
1949 	return 0;
1950 }
1951 
1952 /* dentry stuff */
1953 
1954 /*
1955  * Set <pid>/... inode ownership (can change due to setuid(), etc.)
1956  */
1957 void pid_update_inode(struct task_struct *task, struct inode *inode)
1958 {
1959 	task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
1960 
1961 	inode->i_mode &= ~(S_ISUID | S_ISGID);
1962 	security_task_to_inode(task, inode);
1963 }
1964 
1965 /*
1966  * Rewrite the inode's ownerships here because the owning task may have
1967  * performed a setuid(), etc.
1968  *
1969  */
1970 static int pid_revalidate(struct dentry *dentry, unsigned int flags)
1971 {
1972 	struct inode *inode;
1973 	struct task_struct *task;
1974 
1975 	if (flags & LOOKUP_RCU)
1976 		return -ECHILD;
1977 
1978 	inode = d_inode(dentry);
1979 	task = get_proc_task(inode);
1980 
1981 	if (task) {
1982 		pid_update_inode(task, inode);
1983 		put_task_struct(task);
1984 		return 1;
1985 	}
1986 	return 0;
1987 }
1988 
1989 static inline bool proc_inode_is_dead(struct inode *inode)
1990 {
1991 	return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1992 }
1993 
1994 int pid_delete_dentry(const struct dentry *dentry)
1995 {
1996 	/* Is the task we represent dead?
1997 	 * If so, then don't put the dentry on the lru list,
1998 	 * kill it immediately.
1999 	 */
2000 	return proc_inode_is_dead(d_inode(dentry));
2001 }
2002 
2003 const struct dentry_operations pid_dentry_operations =
2004 {
2005 	.d_revalidate	= pid_revalidate,
2006 	.d_delete	= pid_delete_dentry,
2007 };
2008 
2009 /* Lookups */
2010 
2011 /*
2012  * Fill a directory entry.
2013  *
2014  * If possible create the dcache entry and derive our inode number and
2015  * file type from dcache entry.
2016  *
2017  * Since all of the proc inode numbers are dynamically generated, the inode
2018  * numbers do not exist until the inode is cache.  This means creating the
2019  * the dcache entry in readdir is necessary to keep the inode numbers
2020  * reported by readdir in sync with the inode numbers reported
2021  * by stat.
2022  */
2023 bool proc_fill_cache(struct file *file, struct dir_context *ctx,
2024 	const char *name, unsigned int len,
2025 	instantiate_t instantiate, struct task_struct *task, const void *ptr)
2026 {
2027 	struct dentry *child, *dir = file->f_path.dentry;
2028 	struct qstr qname = QSTR_INIT(name, len);
2029 	struct inode *inode;
2030 	unsigned type = DT_UNKNOWN;
2031 	ino_t ino = 1;
2032 
2033 	child = d_hash_and_lookup(dir, &qname);
2034 	if (!child) {
2035 		DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
2036 		child = d_alloc_parallel(dir, &qname, &wq);
2037 		if (IS_ERR(child))
2038 			goto end_instantiate;
2039 		if (d_in_lookup(child)) {
2040 			struct dentry *res;
2041 			res = instantiate(child, task, ptr);
2042 			d_lookup_done(child);
2043 			if (unlikely(res)) {
2044 				dput(child);
2045 				child = res;
2046 				if (IS_ERR(child))
2047 					goto end_instantiate;
2048 			}
2049 		}
2050 	}
2051 	inode = d_inode(child);
2052 	ino = inode->i_ino;
2053 	type = inode->i_mode >> 12;
2054 	dput(child);
2055 end_instantiate:
2056 	return dir_emit(ctx, name, len, ino, type);
2057 }
2058 
2059 /*
2060  * dname_to_vma_addr - maps a dentry name into two unsigned longs
2061  * which represent vma start and end addresses.
2062  */
2063 static int dname_to_vma_addr(struct dentry *dentry,
2064 			     unsigned long *start, unsigned long *end)
2065 {
2066 	const char *str = dentry->d_name.name;
2067 	unsigned long long sval, eval;
2068 	unsigned int len;
2069 
2070 	if (str[0] == '0' && str[1] != '-')
2071 		return -EINVAL;
2072 	len = _parse_integer(str, 16, &sval);
2073 	if (len & KSTRTOX_OVERFLOW)
2074 		return -EINVAL;
2075 	if (sval != (unsigned long)sval)
2076 		return -EINVAL;
2077 	str += len;
2078 
2079 	if (*str != '-')
2080 		return -EINVAL;
2081 	str++;
2082 
2083 	if (str[0] == '0' && str[1])
2084 		return -EINVAL;
2085 	len = _parse_integer(str, 16, &eval);
2086 	if (len & KSTRTOX_OVERFLOW)
2087 		return -EINVAL;
2088 	if (eval != (unsigned long)eval)
2089 		return -EINVAL;
2090 	str += len;
2091 
2092 	if (*str != '\0')
2093 		return -EINVAL;
2094 
2095 	*start = sval;
2096 	*end = eval;
2097 
2098 	return 0;
2099 }
2100 
2101 static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
2102 {
2103 	unsigned long vm_start, vm_end;
2104 	bool exact_vma_exists = false;
2105 	struct mm_struct *mm = NULL;
2106 	struct task_struct *task;
2107 	struct inode *inode;
2108 	int status = 0;
2109 
2110 	if (flags & LOOKUP_RCU)
2111 		return -ECHILD;
2112 
2113 	inode = d_inode(dentry);
2114 	task = get_proc_task(inode);
2115 	if (!task)
2116 		goto out_notask;
2117 
2118 	mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
2119 	if (IS_ERR_OR_NULL(mm))
2120 		goto out;
2121 
2122 	if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2123 		status = mmap_read_lock_killable(mm);
2124 		if (!status) {
2125 			exact_vma_exists = !!find_exact_vma(mm, vm_start,
2126 							    vm_end);
2127 			mmap_read_unlock(mm);
2128 		}
2129 	}
2130 
2131 	mmput(mm);
2132 
2133 	if (exact_vma_exists) {
2134 		task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
2135 
2136 		security_task_to_inode(task, inode);
2137 		status = 1;
2138 	}
2139 
2140 out:
2141 	put_task_struct(task);
2142 
2143 out_notask:
2144 	return status;
2145 }
2146 
2147 static const struct dentry_operations tid_map_files_dentry_operations = {
2148 	.d_revalidate	= map_files_d_revalidate,
2149 	.d_delete	= pid_delete_dentry,
2150 };
2151 
2152 static int map_files_get_link(struct dentry *dentry, struct path *path)
2153 {
2154 	unsigned long vm_start, vm_end;
2155 	struct vm_area_struct *vma;
2156 	struct task_struct *task;
2157 	struct mm_struct *mm;
2158 	int rc;
2159 
2160 	rc = -ENOENT;
2161 	task = get_proc_task(d_inode(dentry));
2162 	if (!task)
2163 		goto out;
2164 
2165 	mm = get_task_mm(task);
2166 	put_task_struct(task);
2167 	if (!mm)
2168 		goto out;
2169 
2170 	rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2171 	if (rc)
2172 		goto out_mmput;
2173 
2174 	rc = mmap_read_lock_killable(mm);
2175 	if (rc)
2176 		goto out_mmput;
2177 
2178 	rc = -ENOENT;
2179 	vma = find_exact_vma(mm, vm_start, vm_end);
2180 	if (vma && vma->vm_file) {
2181 		*path = vma->vm_file->f_path;
2182 		path_get(path);
2183 		rc = 0;
2184 	}
2185 	mmap_read_unlock(mm);
2186 
2187 out_mmput:
2188 	mmput(mm);
2189 out:
2190 	return rc;
2191 }
2192 
2193 struct map_files_info {
2194 	unsigned long	start;
2195 	unsigned long	end;
2196 	fmode_t		mode;
2197 };
2198 
2199 /*
2200  * Only allow CAP_SYS_ADMIN and CAP_CHECKPOINT_RESTORE to follow the links, due
2201  * to concerns about how the symlinks may be used to bypass permissions on
2202  * ancestor directories in the path to the file in question.
2203  */
2204 static const char *
2205 proc_map_files_get_link(struct dentry *dentry,
2206 			struct inode *inode,
2207 		        struct delayed_call *done)
2208 {
2209 	if (!checkpoint_restore_ns_capable(&init_user_ns))
2210 		return ERR_PTR(-EPERM);
2211 
2212 	return proc_pid_get_link(dentry, inode, done);
2213 }
2214 
2215 /*
2216  * Identical to proc_pid_link_inode_operations except for get_link()
2217  */
2218 static const struct inode_operations proc_map_files_link_inode_operations = {
2219 	.readlink	= proc_pid_readlink,
2220 	.get_link	= proc_map_files_get_link,
2221 	.setattr	= proc_setattr,
2222 };
2223 
2224 static struct dentry *
2225 proc_map_files_instantiate(struct dentry *dentry,
2226 			   struct task_struct *task, const void *ptr)
2227 {
2228 	fmode_t mode = (fmode_t)(unsigned long)ptr;
2229 	struct proc_inode *ei;
2230 	struct inode *inode;
2231 
2232 	inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK |
2233 				    ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2234 				    ((mode & FMODE_WRITE) ? S_IWUSR : 0));
2235 	if (!inode)
2236 		return ERR_PTR(-ENOENT);
2237 
2238 	ei = PROC_I(inode);
2239 	ei->op.proc_get_link = map_files_get_link;
2240 
2241 	inode->i_op = &proc_map_files_link_inode_operations;
2242 	inode->i_size = 64;
2243 
2244 	d_set_d_op(dentry, &tid_map_files_dentry_operations);
2245 	return d_splice_alias(inode, dentry);
2246 }
2247 
2248 static struct dentry *proc_map_files_lookup(struct inode *dir,
2249 		struct dentry *dentry, unsigned int flags)
2250 {
2251 	unsigned long vm_start, vm_end;
2252 	struct vm_area_struct *vma;
2253 	struct task_struct *task;
2254 	struct dentry *result;
2255 	struct mm_struct *mm;
2256 
2257 	result = ERR_PTR(-ENOENT);
2258 	task = get_proc_task(dir);
2259 	if (!task)
2260 		goto out;
2261 
2262 	result = ERR_PTR(-EACCES);
2263 	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2264 		goto out_put_task;
2265 
2266 	result = ERR_PTR(-ENOENT);
2267 	if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2268 		goto out_put_task;
2269 
2270 	mm = get_task_mm(task);
2271 	if (!mm)
2272 		goto out_put_task;
2273 
2274 	result = ERR_PTR(-EINTR);
2275 	if (mmap_read_lock_killable(mm))
2276 		goto out_put_mm;
2277 
2278 	result = ERR_PTR(-ENOENT);
2279 	vma = find_exact_vma(mm, vm_start, vm_end);
2280 	if (!vma)
2281 		goto out_no_vma;
2282 
2283 	if (vma->vm_file)
2284 		result = proc_map_files_instantiate(dentry, task,
2285 				(void *)(unsigned long)vma->vm_file->f_mode);
2286 
2287 out_no_vma:
2288 	mmap_read_unlock(mm);
2289 out_put_mm:
2290 	mmput(mm);
2291 out_put_task:
2292 	put_task_struct(task);
2293 out:
2294 	return result;
2295 }
2296 
2297 static const struct inode_operations proc_map_files_inode_operations = {
2298 	.lookup		= proc_map_files_lookup,
2299 	.permission	= proc_fd_permission,
2300 	.setattr	= proc_setattr,
2301 };
2302 
2303 static int
2304 proc_map_files_readdir(struct file *file, struct dir_context *ctx)
2305 {
2306 	struct vm_area_struct *vma;
2307 	struct task_struct *task;
2308 	struct mm_struct *mm;
2309 	unsigned long nr_files, pos, i;
2310 	GENRADIX(struct map_files_info) fa;
2311 	struct map_files_info *p;
2312 	int ret;
2313 
2314 	genradix_init(&fa);
2315 
2316 	ret = -ENOENT;
2317 	task = get_proc_task(file_inode(file));
2318 	if (!task)
2319 		goto out;
2320 
2321 	ret = -EACCES;
2322 	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
2323 		goto out_put_task;
2324 
2325 	ret = 0;
2326 	if (!dir_emit_dots(file, ctx))
2327 		goto out_put_task;
2328 
2329 	mm = get_task_mm(task);
2330 	if (!mm)
2331 		goto out_put_task;
2332 
2333 	ret = mmap_read_lock_killable(mm);
2334 	if (ret) {
2335 		mmput(mm);
2336 		goto out_put_task;
2337 	}
2338 
2339 	nr_files = 0;
2340 
2341 	/*
2342 	 * We need two passes here:
2343 	 *
2344 	 *  1) Collect vmas of mapped files with mmap_lock taken
2345 	 *  2) Release mmap_lock and instantiate entries
2346 	 *
2347 	 * otherwise we get lockdep complained, since filldir()
2348 	 * routine might require mmap_lock taken in might_fault().
2349 	 */
2350 
2351 	for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2352 		if (!vma->vm_file)
2353 			continue;
2354 		if (++pos <= ctx->pos)
2355 			continue;
2356 
2357 		p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL);
2358 		if (!p) {
2359 			ret = -ENOMEM;
2360 			mmap_read_unlock(mm);
2361 			mmput(mm);
2362 			goto out_put_task;
2363 		}
2364 
2365 		p->start = vma->vm_start;
2366 		p->end = vma->vm_end;
2367 		p->mode = vma->vm_file->f_mode;
2368 	}
2369 	mmap_read_unlock(mm);
2370 	mmput(mm);
2371 
2372 	for (i = 0; i < nr_files; i++) {
2373 		char buf[4 * sizeof(long) + 2];	/* max: %lx-%lx\0 */
2374 		unsigned int len;
2375 
2376 		p = genradix_ptr(&fa, i);
2377 		len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
2378 		if (!proc_fill_cache(file, ctx,
2379 				      buf, len,
2380 				      proc_map_files_instantiate,
2381 				      task,
2382 				      (void *)(unsigned long)p->mode))
2383 			break;
2384 		ctx->pos++;
2385 	}
2386 
2387 out_put_task:
2388 	put_task_struct(task);
2389 out:
2390 	genradix_free(&fa);
2391 	return ret;
2392 }
2393 
2394 static const struct file_operations proc_map_files_operations = {
2395 	.read		= generic_read_dir,
2396 	.iterate_shared	= proc_map_files_readdir,
2397 	.llseek		= generic_file_llseek,
2398 };
2399 
2400 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
2401 struct timers_private {
2402 	struct pid *pid;
2403 	struct task_struct *task;
2404 	struct sighand_struct *sighand;
2405 	struct pid_namespace *ns;
2406 	unsigned long flags;
2407 };
2408 
2409 static void *timers_start(struct seq_file *m, loff_t *pos)
2410 {
2411 	struct timers_private *tp = m->private;
2412 
2413 	tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2414 	if (!tp->task)
2415 		return ERR_PTR(-ESRCH);
2416 
2417 	tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2418 	if (!tp->sighand)
2419 		return ERR_PTR(-ESRCH);
2420 
2421 	return seq_list_start(&tp->task->signal->posix_timers, *pos);
2422 }
2423 
2424 static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2425 {
2426 	struct timers_private *tp = m->private;
2427 	return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2428 }
2429 
2430 static void timers_stop(struct seq_file *m, void *v)
2431 {
2432 	struct timers_private *tp = m->private;
2433 
2434 	if (tp->sighand) {
2435 		unlock_task_sighand(tp->task, &tp->flags);
2436 		tp->sighand = NULL;
2437 	}
2438 
2439 	if (tp->task) {
2440 		put_task_struct(tp->task);
2441 		tp->task = NULL;
2442 	}
2443 }
2444 
2445 static int show_timer(struct seq_file *m, void *v)
2446 {
2447 	struct k_itimer *timer;
2448 	struct timers_private *tp = m->private;
2449 	int notify;
2450 	static const char * const nstr[] = {
2451 		[SIGEV_SIGNAL] = "signal",
2452 		[SIGEV_NONE] = "none",
2453 		[SIGEV_THREAD] = "thread",
2454 	};
2455 
2456 	timer = list_entry((struct list_head *)v, struct k_itimer, list);
2457 	notify = timer->it_sigev_notify;
2458 
2459 	seq_printf(m, "ID: %d\n", timer->it_id);
2460 	seq_printf(m, "signal: %d/%px\n",
2461 		   timer->sigq->info.si_signo,
2462 		   timer->sigq->info.si_value.sival_ptr);
2463 	seq_printf(m, "notify: %s/%s.%d\n",
2464 		   nstr[notify & ~SIGEV_THREAD_ID],
2465 		   (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2466 		   pid_nr_ns(timer->it_pid, tp->ns));
2467 	seq_printf(m, "ClockID: %d\n", timer->it_clock);
2468 
2469 	return 0;
2470 }
2471 
2472 static const struct seq_operations proc_timers_seq_ops = {
2473 	.start	= timers_start,
2474 	.next	= timers_next,
2475 	.stop	= timers_stop,
2476 	.show	= show_timer,
2477 };
2478 
2479 static int proc_timers_open(struct inode *inode, struct file *file)
2480 {
2481 	struct timers_private *tp;
2482 
2483 	tp = __seq_open_private(file, &proc_timers_seq_ops,
2484 			sizeof(struct timers_private));
2485 	if (!tp)
2486 		return -ENOMEM;
2487 
2488 	tp->pid = proc_pid(inode);
2489 	tp->ns = proc_pid_ns(inode->i_sb);
2490 	return 0;
2491 }
2492 
2493 static const struct file_operations proc_timers_operations = {
2494 	.open		= proc_timers_open,
2495 	.read		= seq_read,
2496 	.llseek		= seq_lseek,
2497 	.release	= seq_release_private,
2498 };
2499 #endif
2500 
2501 static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2502 					size_t count, loff_t *offset)
2503 {
2504 	struct inode *inode = file_inode(file);
2505 	struct task_struct *p;
2506 	u64 slack_ns;
2507 	int err;
2508 
2509 	err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2510 	if (err < 0)
2511 		return err;
2512 
2513 	p = get_proc_task(inode);
2514 	if (!p)
2515 		return -ESRCH;
2516 
2517 	if (p != current) {
2518 		rcu_read_lock();
2519 		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2520 			rcu_read_unlock();
2521 			count = -EPERM;
2522 			goto out;
2523 		}
2524 		rcu_read_unlock();
2525 
2526 		err = security_task_setscheduler(p);
2527 		if (err) {
2528 			count = err;
2529 			goto out;
2530 		}
2531 	}
2532 
2533 	task_lock(p);
2534 	if (slack_ns == 0)
2535 		p->timer_slack_ns = p->default_timer_slack_ns;
2536 	else
2537 		p->timer_slack_ns = slack_ns;
2538 	task_unlock(p);
2539 
2540 out:
2541 	put_task_struct(p);
2542 
2543 	return count;
2544 }
2545 
2546 static int timerslack_ns_show(struct seq_file *m, void *v)
2547 {
2548 	struct inode *inode = m->private;
2549 	struct task_struct *p;
2550 	int err = 0;
2551 
2552 	p = get_proc_task(inode);
2553 	if (!p)
2554 		return -ESRCH;
2555 
2556 	if (p != current) {
2557 		rcu_read_lock();
2558 		if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2559 			rcu_read_unlock();
2560 			err = -EPERM;
2561 			goto out;
2562 		}
2563 		rcu_read_unlock();
2564 
2565 		err = security_task_getscheduler(p);
2566 		if (err)
2567 			goto out;
2568 	}
2569 
2570 	task_lock(p);
2571 	seq_printf(m, "%llu\n", p->timer_slack_ns);
2572 	task_unlock(p);
2573 
2574 out:
2575 	put_task_struct(p);
2576 
2577 	return err;
2578 }
2579 
2580 static int timerslack_ns_open(struct inode *inode, struct file *filp)
2581 {
2582 	return single_open(filp, timerslack_ns_show, inode);
2583 }
2584 
2585 static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2586 	.open		= timerslack_ns_open,
2587 	.read		= seq_read,
2588 	.write		= timerslack_ns_write,
2589 	.llseek		= seq_lseek,
2590 	.release	= single_release,
2591 };
2592 
2593 static struct dentry *proc_pident_instantiate(struct dentry *dentry,
2594 	struct task_struct *task, const void *ptr)
2595 {
2596 	const struct pid_entry *p = ptr;
2597 	struct inode *inode;
2598 	struct proc_inode *ei;
2599 
2600 	inode = proc_pid_make_inode(dentry->d_sb, task, p->mode);
2601 	if (!inode)
2602 		return ERR_PTR(-ENOENT);
2603 
2604 	ei = PROC_I(inode);
2605 	if (S_ISDIR(inode->i_mode))
2606 		set_nlink(inode, 2);	/* Use getattr to fix if necessary */
2607 	if (p->iop)
2608 		inode->i_op = p->iop;
2609 	if (p->fop)
2610 		inode->i_fop = p->fop;
2611 	ei->op = p->op;
2612 	pid_update_inode(task, inode);
2613 	d_set_d_op(dentry, &pid_dentry_operations);
2614 	return d_splice_alias(inode, dentry);
2615 }
2616 
2617 static struct dentry *proc_pident_lookup(struct inode *dir,
2618 					 struct dentry *dentry,
2619 					 const struct pid_entry *p,
2620 					 const struct pid_entry *end)
2621 {
2622 	struct task_struct *task = get_proc_task(dir);
2623 	struct dentry *res = ERR_PTR(-ENOENT);
2624 
2625 	if (!task)
2626 		goto out_no_task;
2627 
2628 	/*
2629 	 * Yes, it does not scale. And it should not. Don't add
2630 	 * new entries into /proc/<tgid>/ without very good reasons.
2631 	 */
2632 	for (; p < end; p++) {
2633 		if (p->len != dentry->d_name.len)
2634 			continue;
2635 		if (!memcmp(dentry->d_name.name, p->name, p->len)) {
2636 			res = proc_pident_instantiate(dentry, task, p);
2637 			break;
2638 		}
2639 	}
2640 	put_task_struct(task);
2641 out_no_task:
2642 	return res;
2643 }
2644 
2645 static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
2646 		const struct pid_entry *ents, unsigned int nents)
2647 {
2648 	struct task_struct *task = get_proc_task(file_inode(file));
2649 	const struct pid_entry *p;
2650 
2651 	if (!task)
2652 		return -ENOENT;
2653 
2654 	if (!dir_emit_dots(file, ctx))
2655 		goto out;
2656 
2657 	if (ctx->pos >= nents + 2)
2658 		goto out;
2659 
2660 	for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
2661 		if (!proc_fill_cache(file, ctx, p->name, p->len,
2662 				proc_pident_instantiate, task, p))
2663 			break;
2664 		ctx->pos++;
2665 	}
2666 out:
2667 	put_task_struct(task);
2668 	return 0;
2669 }
2670 
2671 #ifdef CONFIG_SECURITY
2672 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2673 				  size_t count, loff_t *ppos)
2674 {
2675 	struct inode * inode = file_inode(file);
2676 	char *p = NULL;
2677 	ssize_t length;
2678 	struct task_struct *task = get_proc_task(inode);
2679 
2680 	if (!task)
2681 		return -ESRCH;
2682 
2683 	length = security_getprocattr(task, PROC_I(inode)->op.lsm,
2684 				      (char*)file->f_path.dentry->d_name.name,
2685 				      &p);
2686 	put_task_struct(task);
2687 	if (length > 0)
2688 		length = simple_read_from_buffer(buf, count, ppos, p, length);
2689 	kfree(p);
2690 	return length;
2691 }
2692 
2693 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2694 				   size_t count, loff_t *ppos)
2695 {
2696 	struct inode * inode = file_inode(file);
2697 	struct task_struct *task;
2698 	void *page;
2699 	int rv;
2700 
2701 	rcu_read_lock();
2702 	task = pid_task(proc_pid(inode), PIDTYPE_PID);
2703 	if (!task) {
2704 		rcu_read_unlock();
2705 		return -ESRCH;
2706 	}
2707 	/* A task may only write its own attributes. */
2708 	if (current != task) {
2709 		rcu_read_unlock();
2710 		return -EACCES;
2711 	}
2712 	/* Prevent changes to overridden credentials. */
2713 	if (current_cred() != current_real_cred()) {
2714 		rcu_read_unlock();
2715 		return -EBUSY;
2716 	}
2717 	rcu_read_unlock();
2718 
2719 	if (count > PAGE_SIZE)
2720 		count = PAGE_SIZE;
2721 
2722 	/* No partial writes. */
2723 	if (*ppos != 0)
2724 		return -EINVAL;
2725 
2726 	page = memdup_user(buf, count);
2727 	if (IS_ERR(page)) {
2728 		rv = PTR_ERR(page);
2729 		goto out;
2730 	}
2731 
2732 	/* Guard against adverse ptrace interaction */
2733 	rv = mutex_lock_interruptible(&current->signal->cred_guard_mutex);
2734 	if (rv < 0)
2735 		goto out_free;
2736 
2737 	rv = security_setprocattr(PROC_I(inode)->op.lsm,
2738 				  file->f_path.dentry->d_name.name, page,
2739 				  count);
2740 	mutex_unlock(&current->signal->cred_guard_mutex);
2741 out_free:
2742 	kfree(page);
2743 out:
2744 	return rv;
2745 }
2746 
2747 static const struct file_operations proc_pid_attr_operations = {
2748 	.read		= proc_pid_attr_read,
2749 	.write		= proc_pid_attr_write,
2750 	.llseek		= generic_file_llseek,
2751 };
2752 
2753 #define LSM_DIR_OPS(LSM) \
2754 static int proc_##LSM##_attr_dir_iterate(struct file *filp, \
2755 			     struct dir_context *ctx) \
2756 { \
2757 	return proc_pident_readdir(filp, ctx, \
2758 				   LSM##_attr_dir_stuff, \
2759 				   ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2760 } \
2761 \
2762 static const struct file_operations proc_##LSM##_attr_dir_ops = { \
2763 	.read		= generic_read_dir, \
2764 	.iterate	= proc_##LSM##_attr_dir_iterate, \
2765 	.llseek		= default_llseek, \
2766 }; \
2767 \
2768 static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \
2769 				struct dentry *dentry, unsigned int flags) \
2770 { \
2771 	return proc_pident_lookup(dir, dentry, \
2772 				  LSM##_attr_dir_stuff, \
2773 				  LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2774 } \
2775 \
2776 static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \
2777 	.lookup		= proc_##LSM##_attr_dir_lookup, \
2778 	.getattr	= pid_getattr, \
2779 	.setattr	= proc_setattr, \
2780 }
2781 
2782 #ifdef CONFIG_SECURITY_SMACK
2783 static const struct pid_entry smack_attr_dir_stuff[] = {
2784 	ATTR("smack", "current",	0666),
2785 };
2786 LSM_DIR_OPS(smack);
2787 #endif
2788 
2789 #ifdef CONFIG_SECURITY_APPARMOR
2790 static const struct pid_entry apparmor_attr_dir_stuff[] = {
2791 	ATTR("apparmor", "current",	0666),
2792 	ATTR("apparmor", "prev",	0444),
2793 	ATTR("apparmor", "exec",	0666),
2794 };
2795 LSM_DIR_OPS(apparmor);
2796 #endif
2797 
2798 static const struct pid_entry attr_dir_stuff[] = {
2799 	ATTR(NULL, "current",		0666),
2800 	ATTR(NULL, "prev",		0444),
2801 	ATTR(NULL, "exec",		0666),
2802 	ATTR(NULL, "fscreate",		0666),
2803 	ATTR(NULL, "keycreate",		0666),
2804 	ATTR(NULL, "sockcreate",	0666),
2805 #ifdef CONFIG_SECURITY_SMACK
2806 	DIR("smack",			0555,
2807 	    proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops),
2808 #endif
2809 #ifdef CONFIG_SECURITY_APPARMOR
2810 	DIR("apparmor",			0555,
2811 	    proc_apparmor_attr_dir_inode_ops, proc_apparmor_attr_dir_ops),
2812 #endif
2813 };
2814 
2815 static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
2816 {
2817 	return proc_pident_readdir(file, ctx,
2818 				   attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2819 }
2820 
2821 static const struct file_operations proc_attr_dir_operations = {
2822 	.read		= generic_read_dir,
2823 	.iterate_shared	= proc_attr_dir_readdir,
2824 	.llseek		= generic_file_llseek,
2825 };
2826 
2827 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2828 				struct dentry *dentry, unsigned int flags)
2829 {
2830 	return proc_pident_lookup(dir, dentry,
2831 				  attr_dir_stuff,
2832 				  attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff));
2833 }
2834 
2835 static const struct inode_operations proc_attr_dir_inode_operations = {
2836 	.lookup		= proc_attr_dir_lookup,
2837 	.getattr	= pid_getattr,
2838 	.setattr	= proc_setattr,
2839 };
2840 
2841 #endif
2842 
2843 #ifdef CONFIG_ELF_CORE
2844 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2845 					 size_t count, loff_t *ppos)
2846 {
2847 	struct task_struct *task = get_proc_task(file_inode(file));
2848 	struct mm_struct *mm;
2849 	char buffer[PROC_NUMBUF];
2850 	size_t len;
2851 	int ret;
2852 
2853 	if (!task)
2854 		return -ESRCH;
2855 
2856 	ret = 0;
2857 	mm = get_task_mm(task);
2858 	if (mm) {
2859 		len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2860 			       ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2861 				MMF_DUMP_FILTER_SHIFT));
2862 		mmput(mm);
2863 		ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2864 	}
2865 
2866 	put_task_struct(task);
2867 
2868 	return ret;
2869 }
2870 
2871 static ssize_t proc_coredump_filter_write(struct file *file,
2872 					  const char __user *buf,
2873 					  size_t count,
2874 					  loff_t *ppos)
2875 {
2876 	struct task_struct *task;
2877 	struct mm_struct *mm;
2878 	unsigned int val;
2879 	int ret;
2880 	int i;
2881 	unsigned long mask;
2882 
2883 	ret = kstrtouint_from_user(buf, count, 0, &val);
2884 	if (ret < 0)
2885 		return ret;
2886 
2887 	ret = -ESRCH;
2888 	task = get_proc_task(file_inode(file));
2889 	if (!task)
2890 		goto out_no_task;
2891 
2892 	mm = get_task_mm(task);
2893 	if (!mm)
2894 		goto out_no_mm;
2895 	ret = 0;
2896 
2897 	for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2898 		if (val & mask)
2899 			set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2900 		else
2901 			clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2902 	}
2903 
2904 	mmput(mm);
2905  out_no_mm:
2906 	put_task_struct(task);
2907  out_no_task:
2908 	if (ret < 0)
2909 		return ret;
2910 	return count;
2911 }
2912 
2913 static const struct file_operations proc_coredump_filter_operations = {
2914 	.read		= proc_coredump_filter_read,
2915 	.write		= proc_coredump_filter_write,
2916 	.llseek		= generic_file_llseek,
2917 };
2918 #endif
2919 
2920 #ifdef CONFIG_TASK_IO_ACCOUNTING
2921 static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
2922 {
2923 	struct task_io_accounting acct = task->ioac;
2924 	unsigned long flags;
2925 	int result;
2926 
2927 	result = mutex_lock_killable(&task->signal->exec_update_mutex);
2928 	if (result)
2929 		return result;
2930 
2931 	if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
2932 		result = -EACCES;
2933 		goto out_unlock;
2934 	}
2935 
2936 	if (whole && lock_task_sighand(task, &flags)) {
2937 		struct task_struct *t = task;
2938 
2939 		task_io_accounting_add(&acct, &task->signal->ioac);
2940 		while_each_thread(task, t)
2941 			task_io_accounting_add(&acct, &t->ioac);
2942 
2943 		unlock_task_sighand(task, &flags);
2944 	}
2945 	seq_printf(m,
2946 		   "rchar: %llu\n"
2947 		   "wchar: %llu\n"
2948 		   "syscr: %llu\n"
2949 		   "syscw: %llu\n"
2950 		   "read_bytes: %llu\n"
2951 		   "write_bytes: %llu\n"
2952 		   "cancelled_write_bytes: %llu\n",
2953 		   (unsigned long long)acct.rchar,
2954 		   (unsigned long long)acct.wchar,
2955 		   (unsigned long long)acct.syscr,
2956 		   (unsigned long long)acct.syscw,
2957 		   (unsigned long long)acct.read_bytes,
2958 		   (unsigned long long)acct.write_bytes,
2959 		   (unsigned long long)acct.cancelled_write_bytes);
2960 	result = 0;
2961 
2962 out_unlock:
2963 	mutex_unlock(&task->signal->exec_update_mutex);
2964 	return result;
2965 }
2966 
2967 static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2968 				  struct pid *pid, struct task_struct *task)
2969 {
2970 	return do_io_accounting(task, m, 0);
2971 }
2972 
2973 static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2974 				   struct pid *pid, struct task_struct *task)
2975 {
2976 	return do_io_accounting(task, m, 1);
2977 }
2978 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2979 
2980 #ifdef CONFIG_USER_NS
2981 static int proc_id_map_open(struct inode *inode, struct file *file,
2982 	const struct seq_operations *seq_ops)
2983 {
2984 	struct user_namespace *ns = NULL;
2985 	struct task_struct *task;
2986 	struct seq_file *seq;
2987 	int ret = -EINVAL;
2988 
2989 	task = get_proc_task(inode);
2990 	if (task) {
2991 		rcu_read_lock();
2992 		ns = get_user_ns(task_cred_xxx(task, user_ns));
2993 		rcu_read_unlock();
2994 		put_task_struct(task);
2995 	}
2996 	if (!ns)
2997 		goto err;
2998 
2999 	ret = seq_open(file, seq_ops);
3000 	if (ret)
3001 		goto err_put_ns;
3002 
3003 	seq = file->private_data;
3004 	seq->private = ns;
3005 
3006 	return 0;
3007 err_put_ns:
3008 	put_user_ns(ns);
3009 err:
3010 	return ret;
3011 }
3012 
3013 static int proc_id_map_release(struct inode *inode, struct file *file)
3014 {
3015 	struct seq_file *seq = file->private_data;
3016 	struct user_namespace *ns = seq->private;
3017 	put_user_ns(ns);
3018 	return seq_release(inode, file);
3019 }
3020 
3021 static int proc_uid_map_open(struct inode *inode, struct file *file)
3022 {
3023 	return proc_id_map_open(inode, file, &proc_uid_seq_operations);
3024 }
3025 
3026 static int proc_gid_map_open(struct inode *inode, struct file *file)
3027 {
3028 	return proc_id_map_open(inode, file, &proc_gid_seq_operations);
3029 }
3030 
3031 static int proc_projid_map_open(struct inode *inode, struct file *file)
3032 {
3033 	return proc_id_map_open(inode, file, &proc_projid_seq_operations);
3034 }
3035 
3036 static const struct file_operations proc_uid_map_operations = {
3037 	.open		= proc_uid_map_open,
3038 	.write		= proc_uid_map_write,
3039 	.read		= seq_read,
3040 	.llseek		= seq_lseek,
3041 	.release	= proc_id_map_release,
3042 };
3043 
3044 static const struct file_operations proc_gid_map_operations = {
3045 	.open		= proc_gid_map_open,
3046 	.write		= proc_gid_map_write,
3047 	.read		= seq_read,
3048 	.llseek		= seq_lseek,
3049 	.release	= proc_id_map_release,
3050 };
3051 
3052 static const struct file_operations proc_projid_map_operations = {
3053 	.open		= proc_projid_map_open,
3054 	.write		= proc_projid_map_write,
3055 	.read		= seq_read,
3056 	.llseek		= seq_lseek,
3057 	.release	= proc_id_map_release,
3058 };
3059 
3060 static int proc_setgroups_open(struct inode *inode, struct file *file)
3061 {
3062 	struct user_namespace *ns = NULL;
3063 	struct task_struct *task;
3064 	int ret;
3065 
3066 	ret = -ESRCH;
3067 	task = get_proc_task(inode);
3068 	if (task) {
3069 		rcu_read_lock();
3070 		ns = get_user_ns(task_cred_xxx(task, user_ns));
3071 		rcu_read_unlock();
3072 		put_task_struct(task);
3073 	}
3074 	if (!ns)
3075 		goto err;
3076 
3077 	if (file->f_mode & FMODE_WRITE) {
3078 		ret = -EACCES;
3079 		if (!ns_capable(ns, CAP_SYS_ADMIN))
3080 			goto err_put_ns;
3081 	}
3082 
3083 	ret = single_open(file, &proc_setgroups_show, ns);
3084 	if (ret)
3085 		goto err_put_ns;
3086 
3087 	return 0;
3088 err_put_ns:
3089 	put_user_ns(ns);
3090 err:
3091 	return ret;
3092 }
3093 
3094 static int proc_setgroups_release(struct inode *inode, struct file *file)
3095 {
3096 	struct seq_file *seq = file->private_data;
3097 	struct user_namespace *ns = seq->private;
3098 	int ret = single_release(inode, file);
3099 	put_user_ns(ns);
3100 	return ret;
3101 }
3102 
3103 static const struct file_operations proc_setgroups_operations = {
3104 	.open		= proc_setgroups_open,
3105 	.write		= proc_setgroups_write,
3106 	.read		= seq_read,
3107 	.llseek		= seq_lseek,
3108 	.release	= proc_setgroups_release,
3109 };
3110 #endif /* CONFIG_USER_NS */
3111 
3112 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3113 				struct pid *pid, struct task_struct *task)
3114 {
3115 	int err = lock_trace(task);
3116 	if (!err) {
3117 		seq_printf(m, "%08x\n", task->personality);
3118 		unlock_trace(task);
3119 	}
3120 	return err;
3121 }
3122 
3123 #ifdef CONFIG_LIVEPATCH
3124 static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
3125 				struct pid *pid, struct task_struct *task)
3126 {
3127 	seq_printf(m, "%d\n", task->patch_state);
3128 	return 0;
3129 }
3130 #endif /* CONFIG_LIVEPATCH */
3131 
3132 #ifdef CONFIG_STACKLEAK_METRICS
3133 static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns,
3134 				struct pid *pid, struct task_struct *task)
3135 {
3136 	unsigned long prev_depth = THREAD_SIZE -
3137 				(task->prev_lowest_stack & (THREAD_SIZE - 1));
3138 	unsigned long depth = THREAD_SIZE -
3139 				(task->lowest_stack & (THREAD_SIZE - 1));
3140 
3141 	seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n",
3142 							prev_depth, depth);
3143 	return 0;
3144 }
3145 #endif /* CONFIG_STACKLEAK_METRICS */
3146 
3147 /*
3148  * Thread groups
3149  */
3150 static const struct file_operations proc_task_operations;
3151 static const struct inode_operations proc_task_inode_operations;
3152 
3153 static const struct pid_entry tgid_base_stuff[] = {
3154 	DIR("task",       S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3155 	DIR("fd",         S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3156 	DIR("map_files",  S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3157 	DIR("fdinfo",     S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3158 	DIR("ns",	  S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3159 #ifdef CONFIG_NET
3160 	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3161 #endif
3162 	REG("environ",    S_IRUSR, proc_environ_operations),
3163 	REG("auxv",       S_IRUSR, proc_auxv_operations),
3164 	ONE("status",     S_IRUGO, proc_pid_status),
3165 	ONE("personality", S_IRUSR, proc_pid_personality),
3166 	ONE("limits",	  S_IRUGO, proc_pid_limits),
3167 #ifdef CONFIG_SCHED_DEBUG
3168 	REG("sched",      S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3169 #endif
3170 #ifdef CONFIG_SCHED_AUTOGROUP
3171 	REG("autogroup",  S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3172 #endif
3173 #ifdef CONFIG_TIME_NS
3174 	REG("timens_offsets",  S_IRUGO|S_IWUSR, proc_timens_offsets_operations),
3175 #endif
3176 	REG("comm",      S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3177 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3178 	ONE("syscall",    S_IRUSR, proc_pid_syscall),
3179 #endif
3180 	REG("cmdline",    S_IRUGO, proc_pid_cmdline_ops),
3181 	ONE("stat",       S_IRUGO, proc_tgid_stat),
3182 	ONE("statm",      S_IRUGO, proc_pid_statm),
3183 	REG("maps",       S_IRUGO, proc_pid_maps_operations),
3184 #ifdef CONFIG_NUMA
3185 	REG("numa_maps",  S_IRUGO, proc_pid_numa_maps_operations),
3186 #endif
3187 	REG("mem",        S_IRUSR|S_IWUSR, proc_mem_operations),
3188 	LNK("cwd",        proc_cwd_link),
3189 	LNK("root",       proc_root_link),
3190 	LNK("exe",        proc_exe_link),
3191 	REG("mounts",     S_IRUGO, proc_mounts_operations),
3192 	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3193 	REG("mountstats", S_IRUSR, proc_mountstats_operations),
3194 #ifdef CONFIG_PROC_PAGE_MONITOR
3195 	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3196 	REG("smaps",      S_IRUGO, proc_pid_smaps_operations),
3197 	REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3198 	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3199 #endif
3200 #ifdef CONFIG_SECURITY
3201 	DIR("attr",       S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3202 #endif
3203 #ifdef CONFIG_KALLSYMS
3204 	ONE("wchan",      S_IRUGO, proc_pid_wchan),
3205 #endif
3206 #ifdef CONFIG_STACKTRACE
3207 	ONE("stack",      S_IRUSR, proc_pid_stack),
3208 #endif
3209 #ifdef CONFIG_SCHED_INFO
3210 	ONE("schedstat",  S_IRUGO, proc_pid_schedstat),
3211 #endif
3212 #ifdef CONFIG_LATENCYTOP
3213 	REG("latency",  S_IRUGO, proc_lstats_operations),
3214 #endif
3215 #ifdef CONFIG_PROC_PID_CPUSET
3216 	ONE("cpuset",     S_IRUGO, proc_cpuset_show),
3217 #endif
3218 #ifdef CONFIG_CGROUPS
3219 	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3220 #endif
3221 #ifdef CONFIG_PROC_CPU_RESCTRL
3222 	ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3223 #endif
3224 	ONE("oom_score",  S_IRUGO, proc_oom_score),
3225 	REG("oom_adj",    S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3226 	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3227 #ifdef CONFIG_AUDIT
3228 	REG("loginuid",   S_IWUSR|S_IRUGO, proc_loginuid_operations),
3229 	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3230 #endif
3231 #ifdef CONFIG_FAULT_INJECTION
3232 	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3233 	REG("fail-nth", 0644, proc_fail_nth_operations),
3234 #endif
3235 #ifdef CONFIG_ELF_CORE
3236 	REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3237 #endif
3238 #ifdef CONFIG_TASK_IO_ACCOUNTING
3239 	ONE("io",	S_IRUSR, proc_tgid_io_accounting),
3240 #endif
3241 #ifdef CONFIG_USER_NS
3242 	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3243 	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3244 	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3245 	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3246 #endif
3247 #if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
3248 	REG("timers",	  S_IRUGO, proc_timers_operations),
3249 #endif
3250 	REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
3251 #ifdef CONFIG_LIVEPATCH
3252 	ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3253 #endif
3254 #ifdef CONFIG_STACKLEAK_METRICS
3255 	ONE("stack_depth", S_IRUGO, proc_stack_depth),
3256 #endif
3257 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3258 	ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3259 #endif
3260 };
3261 
3262 static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
3263 {
3264 	return proc_pident_readdir(file, ctx,
3265 				   tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3266 }
3267 
3268 static const struct file_operations proc_tgid_base_operations = {
3269 	.read		= generic_read_dir,
3270 	.iterate_shared	= proc_tgid_base_readdir,
3271 	.llseek		= generic_file_llseek,
3272 };
3273 
3274 struct pid *tgid_pidfd_to_pid(const struct file *file)
3275 {
3276 	if (file->f_op != &proc_tgid_base_operations)
3277 		return ERR_PTR(-EBADF);
3278 
3279 	return proc_pid(file_inode(file));
3280 }
3281 
3282 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3283 {
3284 	return proc_pident_lookup(dir, dentry,
3285 				  tgid_base_stuff,
3286 				  tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff));
3287 }
3288 
3289 static const struct inode_operations proc_tgid_base_inode_operations = {
3290 	.lookup		= proc_tgid_base_lookup,
3291 	.getattr	= pid_getattr,
3292 	.setattr	= proc_setattr,
3293 	.permission	= proc_pid_permission,
3294 };
3295 
3296 /**
3297  * proc_flush_pid -  Remove dcache entries for @pid from the /proc dcache.
3298  * @pid: pid that should be flushed.
3299  *
3300  * This function walks a list of inodes (that belong to any proc
3301  * filesystem) that are attached to the pid and flushes them from
3302  * the dentry cache.
3303  *
3304  * It is safe and reasonable to cache /proc entries for a task until
3305  * that task exits.  After that they just clog up the dcache with
3306  * useless entries, possibly causing useful dcache entries to be
3307  * flushed instead.  This routine is provided to flush those useless
3308  * dcache entries when a process is reaped.
3309  *
3310  * NOTE: This routine is just an optimization so it does not guarantee
3311  *       that no dcache entries will exist after a process is reaped
3312  *       it just makes it very unlikely that any will persist.
3313  */
3314 
3315 void proc_flush_pid(struct pid *pid)
3316 {
3317 	proc_invalidate_siblings_dcache(&pid->inodes, &pid->lock);
3318 }
3319 
3320 static struct dentry *proc_pid_instantiate(struct dentry * dentry,
3321 				   struct task_struct *task, const void *ptr)
3322 {
3323 	struct inode *inode;
3324 
3325 	inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3326 	if (!inode)
3327 		return ERR_PTR(-ENOENT);
3328 
3329 	inode->i_op = &proc_tgid_base_inode_operations;
3330 	inode->i_fop = &proc_tgid_base_operations;
3331 	inode->i_flags|=S_IMMUTABLE;
3332 
3333 	set_nlink(inode, nlink_tgid);
3334 	pid_update_inode(task, inode);
3335 
3336 	d_set_d_op(dentry, &pid_dentry_operations);
3337 	return d_splice_alias(inode, dentry);
3338 }
3339 
3340 struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags)
3341 {
3342 	struct task_struct *task;
3343 	unsigned tgid;
3344 	struct proc_fs_info *fs_info;
3345 	struct pid_namespace *ns;
3346 	struct dentry *result = ERR_PTR(-ENOENT);
3347 
3348 	tgid = name_to_int(&dentry->d_name);
3349 	if (tgid == ~0U)
3350 		goto out;
3351 
3352 	fs_info = proc_sb_info(dentry->d_sb);
3353 	ns = fs_info->pid_ns;
3354 	rcu_read_lock();
3355 	task = find_task_by_pid_ns(tgid, ns);
3356 	if (task)
3357 		get_task_struct(task);
3358 	rcu_read_unlock();
3359 	if (!task)
3360 		goto out;
3361 
3362 	/* Limit procfs to only ptraceable tasks */
3363 	if (fs_info->hide_pid == HIDEPID_NOT_PTRACEABLE) {
3364 		if (!has_pid_permissions(fs_info, task, HIDEPID_NO_ACCESS))
3365 			goto out_put_task;
3366 	}
3367 
3368 	result = proc_pid_instantiate(dentry, task, NULL);
3369 out_put_task:
3370 	put_task_struct(task);
3371 out:
3372 	return result;
3373 }
3374 
3375 /*
3376  * Find the first task with tgid >= tgid
3377  *
3378  */
3379 struct tgid_iter {
3380 	unsigned int tgid;
3381 	struct task_struct *task;
3382 };
3383 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3384 {
3385 	struct pid *pid;
3386 
3387 	if (iter.task)
3388 		put_task_struct(iter.task);
3389 	rcu_read_lock();
3390 retry:
3391 	iter.task = NULL;
3392 	pid = find_ge_pid(iter.tgid, ns);
3393 	if (pid) {
3394 		iter.tgid = pid_nr_ns(pid, ns);
3395 		iter.task = pid_task(pid, PIDTYPE_TGID);
3396 		if (!iter.task) {
3397 			iter.tgid += 1;
3398 			goto retry;
3399 		}
3400 		get_task_struct(iter.task);
3401 	}
3402 	rcu_read_unlock();
3403 	return iter;
3404 }
3405 
3406 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
3407 
3408 /* for the /proc/ directory itself, after non-process stuff has been done */
3409 int proc_pid_readdir(struct file *file, struct dir_context *ctx)
3410 {
3411 	struct tgid_iter iter;
3412 	struct proc_fs_info *fs_info = proc_sb_info(file_inode(file)->i_sb);
3413 	struct pid_namespace *ns = proc_pid_ns(file_inode(file)->i_sb);
3414 	loff_t pos = ctx->pos;
3415 
3416 	if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
3417 		return 0;
3418 
3419 	if (pos == TGID_OFFSET - 2) {
3420 		struct inode *inode = d_inode(fs_info->proc_self);
3421 		if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
3422 			return 0;
3423 		ctx->pos = pos = pos + 1;
3424 	}
3425 	if (pos == TGID_OFFSET - 1) {
3426 		struct inode *inode = d_inode(fs_info->proc_thread_self);
3427 		if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3428 			return 0;
3429 		ctx->pos = pos = pos + 1;
3430 	}
3431 	iter.tgid = pos - TGID_OFFSET;
3432 	iter.task = NULL;
3433 	for (iter = next_tgid(ns, iter);
3434 	     iter.task;
3435 	     iter.tgid += 1, iter = next_tgid(ns, iter)) {
3436 		char name[10 + 1];
3437 		unsigned int len;
3438 
3439 		cond_resched();
3440 		if (!has_pid_permissions(fs_info, iter.task, HIDEPID_INVISIBLE))
3441 			continue;
3442 
3443 		len = snprintf(name, sizeof(name), "%u", iter.tgid);
3444 		ctx->pos = iter.tgid + TGID_OFFSET;
3445 		if (!proc_fill_cache(file, ctx, name, len,
3446 				     proc_pid_instantiate, iter.task, NULL)) {
3447 			put_task_struct(iter.task);
3448 			return 0;
3449 		}
3450 	}
3451 	ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
3452 	return 0;
3453 }
3454 
3455 /*
3456  * proc_tid_comm_permission is a special permission function exclusively
3457  * used for the node /proc/<pid>/task/<tid>/comm.
3458  * It bypasses generic permission checks in the case where a task of the same
3459  * task group attempts to access the node.
3460  * The rationale behind this is that glibc and bionic access this node for
3461  * cross thread naming (pthread_set/getname_np(!self)). However, if
3462  * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3463  * which locks out the cross thread naming implementation.
3464  * This function makes sure that the node is always accessible for members of
3465  * same thread group.
3466  */
3467 static int proc_tid_comm_permission(struct inode *inode, int mask)
3468 {
3469 	bool is_same_tgroup;
3470 	struct task_struct *task;
3471 
3472 	task = get_proc_task(inode);
3473 	if (!task)
3474 		return -ESRCH;
3475 	is_same_tgroup = same_thread_group(current, task);
3476 	put_task_struct(task);
3477 
3478 	if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3479 		/* This file (/proc/<pid>/task/<tid>/comm) can always be
3480 		 * read or written by the members of the corresponding
3481 		 * thread group.
3482 		 */
3483 		return 0;
3484 	}
3485 
3486 	return generic_permission(inode, mask);
3487 }
3488 
3489 static const struct inode_operations proc_tid_comm_inode_operations = {
3490 		.permission = proc_tid_comm_permission,
3491 };
3492 
3493 /*
3494  * Tasks
3495  */
3496 static const struct pid_entry tid_base_stuff[] = {
3497 	DIR("fd",        S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3498 	DIR("fdinfo",    S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3499 	DIR("ns",	 S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3500 #ifdef CONFIG_NET
3501 	DIR("net",        S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3502 #endif
3503 	REG("environ",   S_IRUSR, proc_environ_operations),
3504 	REG("auxv",      S_IRUSR, proc_auxv_operations),
3505 	ONE("status",    S_IRUGO, proc_pid_status),
3506 	ONE("personality", S_IRUSR, proc_pid_personality),
3507 	ONE("limits",	 S_IRUGO, proc_pid_limits),
3508 #ifdef CONFIG_SCHED_DEBUG
3509 	REG("sched",     S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3510 #endif
3511 	NOD("comm",      S_IFREG|S_IRUGO|S_IWUSR,
3512 			 &proc_tid_comm_inode_operations,
3513 			 &proc_pid_set_comm_operations, {}),
3514 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3515 	ONE("syscall",   S_IRUSR, proc_pid_syscall),
3516 #endif
3517 	REG("cmdline",   S_IRUGO, proc_pid_cmdline_ops),
3518 	ONE("stat",      S_IRUGO, proc_tid_stat),
3519 	ONE("statm",     S_IRUGO, proc_pid_statm),
3520 	REG("maps",      S_IRUGO, proc_pid_maps_operations),
3521 #ifdef CONFIG_PROC_CHILDREN
3522 	REG("children",  S_IRUGO, proc_tid_children_operations),
3523 #endif
3524 #ifdef CONFIG_NUMA
3525 	REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3526 #endif
3527 	REG("mem",       S_IRUSR|S_IWUSR, proc_mem_operations),
3528 	LNK("cwd",       proc_cwd_link),
3529 	LNK("root",      proc_root_link),
3530 	LNK("exe",       proc_exe_link),
3531 	REG("mounts",    S_IRUGO, proc_mounts_operations),
3532 	REG("mountinfo",  S_IRUGO, proc_mountinfo_operations),
3533 #ifdef CONFIG_PROC_PAGE_MONITOR
3534 	REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3535 	REG("smaps",     S_IRUGO, proc_pid_smaps_operations),
3536 	REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
3537 	REG("pagemap",    S_IRUSR, proc_pagemap_operations),
3538 #endif
3539 #ifdef CONFIG_SECURITY
3540 	DIR("attr",      S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3541 #endif
3542 #ifdef CONFIG_KALLSYMS
3543 	ONE("wchan",     S_IRUGO, proc_pid_wchan),
3544 #endif
3545 #ifdef CONFIG_STACKTRACE
3546 	ONE("stack",      S_IRUSR, proc_pid_stack),
3547 #endif
3548 #ifdef CONFIG_SCHED_INFO
3549 	ONE("schedstat", S_IRUGO, proc_pid_schedstat),
3550 #endif
3551 #ifdef CONFIG_LATENCYTOP
3552 	REG("latency",  S_IRUGO, proc_lstats_operations),
3553 #endif
3554 #ifdef CONFIG_PROC_PID_CPUSET
3555 	ONE("cpuset",    S_IRUGO, proc_cpuset_show),
3556 #endif
3557 #ifdef CONFIG_CGROUPS
3558 	ONE("cgroup",  S_IRUGO, proc_cgroup_show),
3559 #endif
3560 #ifdef CONFIG_PROC_CPU_RESCTRL
3561 	ONE("cpu_resctrl_groups", S_IRUGO, proc_resctrl_show),
3562 #endif
3563 	ONE("oom_score", S_IRUGO, proc_oom_score),
3564 	REG("oom_adj",   S_IRUGO|S_IWUSR, proc_oom_adj_operations),
3565 	REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3566 #ifdef CONFIG_AUDIT
3567 	REG("loginuid",  S_IWUSR|S_IRUGO, proc_loginuid_operations),
3568 	REG("sessionid",  S_IRUGO, proc_sessionid_operations),
3569 #endif
3570 #ifdef CONFIG_FAULT_INJECTION
3571 	REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3572 	REG("fail-nth", 0644, proc_fail_nth_operations),
3573 #endif
3574 #ifdef CONFIG_TASK_IO_ACCOUNTING
3575 	ONE("io",	S_IRUSR, proc_tid_io_accounting),
3576 #endif
3577 #ifdef CONFIG_USER_NS
3578 	REG("uid_map",    S_IRUGO|S_IWUSR, proc_uid_map_operations),
3579 	REG("gid_map",    S_IRUGO|S_IWUSR, proc_gid_map_operations),
3580 	REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
3581 	REG("setgroups",  S_IRUGO|S_IWUSR, proc_setgroups_operations),
3582 #endif
3583 #ifdef CONFIG_LIVEPATCH
3584 	ONE("patch_state",  S_IRUSR, proc_pid_patch_state),
3585 #endif
3586 #ifdef CONFIG_PROC_PID_ARCH_STATUS
3587 	ONE("arch_status", S_IRUGO, proc_pid_arch_status),
3588 #endif
3589 };
3590 
3591 static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
3592 {
3593 	return proc_pident_readdir(file, ctx,
3594 				   tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3595 }
3596 
3597 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3598 {
3599 	return proc_pident_lookup(dir, dentry,
3600 				  tid_base_stuff,
3601 				  tid_base_stuff + ARRAY_SIZE(tid_base_stuff));
3602 }
3603 
3604 static const struct file_operations proc_tid_base_operations = {
3605 	.read		= generic_read_dir,
3606 	.iterate_shared	= proc_tid_base_readdir,
3607 	.llseek		= generic_file_llseek,
3608 };
3609 
3610 static const struct inode_operations proc_tid_base_inode_operations = {
3611 	.lookup		= proc_tid_base_lookup,
3612 	.getattr	= pid_getattr,
3613 	.setattr	= proc_setattr,
3614 };
3615 
3616 static struct dentry *proc_task_instantiate(struct dentry *dentry,
3617 	struct task_struct *task, const void *ptr)
3618 {
3619 	struct inode *inode;
3620 	inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
3621 	if (!inode)
3622 		return ERR_PTR(-ENOENT);
3623 
3624 	inode->i_op = &proc_tid_base_inode_operations;
3625 	inode->i_fop = &proc_tid_base_operations;
3626 	inode->i_flags |= S_IMMUTABLE;
3627 
3628 	set_nlink(inode, nlink_tid);
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 
3635 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
3636 {
3637 	struct task_struct *task;
3638 	struct task_struct *leader = get_proc_task(dir);
3639 	unsigned tid;
3640 	struct proc_fs_info *fs_info;
3641 	struct pid_namespace *ns;
3642 	struct dentry *result = ERR_PTR(-ENOENT);
3643 
3644 	if (!leader)
3645 		goto out_no_task;
3646 
3647 	tid = name_to_int(&dentry->d_name);
3648 	if (tid == ~0U)
3649 		goto out;
3650 
3651 	fs_info = proc_sb_info(dentry->d_sb);
3652 	ns = fs_info->pid_ns;
3653 	rcu_read_lock();
3654 	task = find_task_by_pid_ns(tid, ns);
3655 	if (task)
3656 		get_task_struct(task);
3657 	rcu_read_unlock();
3658 	if (!task)
3659 		goto out;
3660 	if (!same_thread_group(leader, task))
3661 		goto out_drop_task;
3662 
3663 	result = proc_task_instantiate(dentry, task, NULL);
3664 out_drop_task:
3665 	put_task_struct(task);
3666 out:
3667 	put_task_struct(leader);
3668 out_no_task:
3669 	return result;
3670 }
3671 
3672 /*
3673  * Find the first tid of a thread group to return to user space.
3674  *
3675  * Usually this is just the thread group leader, but if the users
3676  * buffer was too small or there was a seek into the middle of the
3677  * directory we have more work todo.
3678  *
3679  * In the case of a short read we start with find_task_by_pid.
3680  *
3681  * In the case of a seek we start with the leader and walk nr
3682  * threads past it.
3683  */
3684 static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3685 					struct pid_namespace *ns)
3686 {
3687 	struct task_struct *pos, *task;
3688 	unsigned long nr = f_pos;
3689 
3690 	if (nr != f_pos)	/* 32bit overflow? */
3691 		return NULL;
3692 
3693 	rcu_read_lock();
3694 	task = pid_task(pid, PIDTYPE_PID);
3695 	if (!task)
3696 		goto fail;
3697 
3698 	/* Attempt to start with the tid of a thread */
3699 	if (tid && nr) {
3700 		pos = find_task_by_pid_ns(tid, ns);
3701 		if (pos && same_thread_group(pos, task))
3702 			goto found;
3703 	}
3704 
3705 	/* If nr exceeds the number of threads there is nothing todo */
3706 	if (nr >= get_nr_threads(task))
3707 		goto fail;
3708 
3709 	/* If we haven't found our starting place yet start
3710 	 * with the leader and walk nr threads forward.
3711 	 */
3712 	pos = task = task->group_leader;
3713 	do {
3714 		if (!nr--)
3715 			goto found;
3716 	} while_each_thread(task, pos);
3717 fail:
3718 	pos = NULL;
3719 	goto out;
3720 found:
3721 	get_task_struct(pos);
3722 out:
3723 	rcu_read_unlock();
3724 	return pos;
3725 }
3726 
3727 /*
3728  * Find the next thread in the thread list.
3729  * Return NULL if there is an error or no next thread.
3730  *
3731  * The reference to the input task_struct is released.
3732  */
3733 static struct task_struct *next_tid(struct task_struct *start)
3734 {
3735 	struct task_struct *pos = NULL;
3736 	rcu_read_lock();
3737 	if (pid_alive(start)) {
3738 		pos = next_thread(start);
3739 		if (thread_group_leader(pos))
3740 			pos = NULL;
3741 		else
3742 			get_task_struct(pos);
3743 	}
3744 	rcu_read_unlock();
3745 	put_task_struct(start);
3746 	return pos;
3747 }
3748 
3749 /* for the /proc/TGID/task/ directories */
3750 static int proc_task_readdir(struct file *file, struct dir_context *ctx)
3751 {
3752 	struct inode *inode = file_inode(file);
3753 	struct task_struct *task;
3754 	struct pid_namespace *ns;
3755 	int tid;
3756 
3757 	if (proc_inode_is_dead(inode))
3758 		return -ENOENT;
3759 
3760 	if (!dir_emit_dots(file, ctx))
3761 		return 0;
3762 
3763 	/* f_version caches the tgid value that the last readdir call couldn't
3764 	 * return. lseek aka telldir automagically resets f_version to 0.
3765 	 */
3766 	ns = proc_pid_ns(inode->i_sb);
3767 	tid = (int)file->f_version;
3768 	file->f_version = 0;
3769 	for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
3770 	     task;
3771 	     task = next_tid(task), ctx->pos++) {
3772 		char name[10 + 1];
3773 		unsigned int len;
3774 		tid = task_pid_nr_ns(task, ns);
3775 		len = snprintf(name, sizeof(name), "%u", tid);
3776 		if (!proc_fill_cache(file, ctx, name, len,
3777 				proc_task_instantiate, task, NULL)) {
3778 			/* returning this tgid failed, save it as the first
3779 			 * pid for the next readir call */
3780 			file->f_version = (u64)tid;
3781 			put_task_struct(task);
3782 			break;
3783 		}
3784 	}
3785 
3786 	return 0;
3787 }
3788 
3789 static int proc_task_getattr(const struct path *path, struct kstat *stat,
3790 			     u32 request_mask, unsigned int query_flags)
3791 {
3792 	struct inode *inode = d_inode(path->dentry);
3793 	struct task_struct *p = get_proc_task(inode);
3794 	generic_fillattr(inode, stat);
3795 
3796 	if (p) {
3797 		stat->nlink += get_nr_threads(p);
3798 		put_task_struct(p);
3799 	}
3800 
3801 	return 0;
3802 }
3803 
3804 static const struct inode_operations proc_task_inode_operations = {
3805 	.lookup		= proc_task_lookup,
3806 	.getattr	= proc_task_getattr,
3807 	.setattr	= proc_setattr,
3808 	.permission	= proc_pid_permission,
3809 };
3810 
3811 static const struct file_operations proc_task_operations = {
3812 	.read		= generic_read_dir,
3813 	.iterate_shared	= proc_task_readdir,
3814 	.llseek		= generic_file_llseek,
3815 };
3816 
3817 void __init set_proc_pid_nlink(void)
3818 {
3819 	nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3820 	nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3821 }
3822