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