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