blob: 255f6754c70d5800777434557898ca8b35721d1f [file] [log] [blame]
Greg Kroah-Hartmanb2441312017-11-01 15:07:57 +01001// SPDX-License-Identifier: GPL-2.0
Linus Torvalds1da177e2005-04-16 15:20:36 -07002/*
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.
Mauricio Line070ad42005-09-03 15:55:10 -070015 *
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.
Linus Torvalds1da177e2005-04-16 15:20:36 -070049 */
50
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080051#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070052
Linus Torvalds1da177e2005-04-16 15:20:36 -070053#include <linux/errno.h>
54#include <linux/time.h>
55#include <linux/proc_fs.h>
56#include <linux/stat.h>
Andrea Righi59954772008-07-27 17:29:15 +020057#include <linux/task_io_accounting_ops.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070058#include <linux/init.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080059#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070060#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040061#include <linux/fdtable.h>
Kent Overstreet94f8f3b2019-03-11 23:31:18 -070062#include <linux/generic-radix-tree.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070063#include <linux/string.h>
64#include <linux/seq_file.h>
65#include <linux/namei.h>
Kirill Korotaev6b3286e2006-12-08 02:37:56 -080066#include <linux/mnt_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070067#include <linux/mm.h>
David Rientjesa63d83f2010-08-09 17:19:46 -070068#include <linux/swap.h>
Dipankar Sarmab8359962005-09-09 13:04:14 -070069#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070070#include <linux/kallsyms.h>
Ken Chen2ec220e2008-11-10 11:26:08 +030071#include <linux/stacktrace.h>
Neil Hormand85f50d2007-10-18 23:40:37 -070072#include <linux/resource.h>
Kees Cook5096add2007-05-08 00:26:04 -070073#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070074#include <linux/mount.h>
75#include <linux/security.h>
76#include <linux/ptrace.h>
Roland McGrath0d094ef2008-07-25 19:45:49 -070077#include <linux/tracehook.h>
Andrew Morton87ebdc02013-02-27 17:03:16 -080078#include <linux/printk.h>
Alexey Dobriyanefb1a572018-02-06 15:37:24 -080079#include <linux/cache.h>
Paul Menagea4243162007-10-18 23:39:35 -070080#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070081#include <linux/cpuset.h>
82#include <linux/audit.h>
Al Viro5addc5d2005-11-07 17:15:49 -050083#include <linux/poll.h>
Serge E. Hallyn1651e142006-10-02 02:18:08 -070084#include <linux/nsproxy.h>
Alexey Dobriyan8ac773b2006-10-19 23:28:32 -070085#include <linux/oom.h>
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -070086#include <linux/elf.h>
Pavel Emelyanov60347f62007-10-18 23:40:03 -070087#include <linux/pid_namespace.h>
Eric W. Biederman22d917d2011-11-17 00:11:58 -080088#include <linux/user_namespace.h>
Al Viro5ad4e532009-03-29 19:50:06 -040089#include <linux/fs_struct.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090090#include <linux/slab.h>
Ingo Molnar4eb5aaa2017-02-08 18:51:29 +010091#include <linux/sched/autogroup.h>
Ingo Molnar6e84f312017-02-08 18:51:29 +010092#include <linux/sched/mm.h>
Ingo Molnarf7ccbae2017-02-08 18:51:30 +010093#include <linux/sched/coredump.h>
Ingo Molnarb17b0152017-02-08 18:51:35 +010094#include <linux/sched/debug.h>
Ingo Molnar3905f9a2017-02-05 12:07:04 +010095#include <linux/sched/stat.h>
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +040096#include <linux/posix-timers.h>
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -080097#include <trace/events/oom.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098#include "internal.h"
Cyrill Gorcunovfaf60af2012-08-23 14:43:24 +040099#include "fd.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
Alexey Dobriyanac7f1062018-02-06 15:36:59 -0800101#include "../../lib/kstrtox.h"
102
Eric W. Biederman0f2fe202006-06-26 00:25:46 -0700103/* NOTE:
104 * Implementing inode permission operations in /proc is almost
105 * certainly an error. Permission checks need to happen during
106 * each system call not at open time. The reason is that most of
107 * what we wish to check for permissions in /proc varies at runtime.
108 *
109 * The classic example of a problem is opening file descriptors
110 * in /proc for a task before it execs a suid executable.
111 */
112
Alexey Dobriyanefb1a572018-02-06 15:37:24 -0800113static u8 nlink_tid __ro_after_init;
114static u8 nlink_tgid __ro_after_init;
Alexey Dobriyan1270dd82016-12-12 16:45:32 -0800115
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116struct pid_entry {
Alexey Dobriyancedbcca2014-08-08 14:21:33 -0700117 const char *name;
Alexey Dobriyan623f5942016-12-12 16:45:08 -0800118 unsigned int len;
Al Virod161a132011-07-24 03:36:29 -0400119 umode_t mode;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800120 const struct inode_operations *iop;
Arjan van de Ven00977a52007-02-12 00:55:34 -0800121 const struct file_operations *fop;
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700122 union proc_op op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123};
124
Eric W. Biederman61a28782006-10-02 02:18:49 -0700125#define NOD(NAME, MODE, IOP, FOP, OP) { \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700126 .name = (NAME), \
Eric Dumazetc5141e62007-05-08 00:26:15 -0700127 .len = sizeof(NAME) - 1, \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700128 .mode = MODE, \
129 .iop = IOP, \
130 .fop = FOP, \
131 .op = OP, \
132}
133
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300134#define DIR(NAME, MODE, iops, fops) \
135 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
136#define LNK(NAME, get_link) \
Eric W. Biederman61a28782006-10-02 02:18:49 -0700137 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700138 &proc_pid_link_inode_operations, NULL, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300139 { .proc_get_link = get_link } )
140#define REG(NAME, MODE, fops) \
141 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300142#define ONE(NAME, MODE, show) \
Casey Schaufler6d9c9392018-09-21 17:16:59 -0700143 NOD(NAME, (S_IFREG|(MODE)), \
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800144 NULL, &proc_single_file_operations, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300145 { .proc_show = show } )
Casey Schaufler6d9c9392018-09-21 17:16:59 -0700146#define ATTR(LSM, NAME, MODE) \
147 NOD(NAME, (S_IFREG|(MODE)), \
148 NULL, &proc_pid_attr_operations, \
149 { .lsm = LSM })
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150
Vegard Nossumaed54172008-06-05 22:46:53 -0700151/*
152 * Count the number of hardlinks for the pid_entry table, excluding the .
153 * and .. links.
154 */
Alexey Dobriyan1270dd82016-12-12 16:45:32 -0800155static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
Vegard Nossumaed54172008-06-05 22:46:53 -0700156 unsigned int n)
157{
158 unsigned int i;
159 unsigned int count;
160
Alexey Dobriyan1270dd82016-12-12 16:45:32 -0800161 count = 2;
Vegard Nossumaed54172008-06-05 22:46:53 -0700162 for (i = 0; i < n; ++i) {
163 if (S_ISDIR(entries[i].mode))
164 ++count;
165 }
166
167 return count;
168}
169
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200170static int get_task_root(struct task_struct *task, struct path *root)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171{
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000172 int result = -ENOENT;
173
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700174 task_lock(task);
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200175 if (task->fs) {
176 get_fs_root(task->fs, root);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000177 result = 0;
178 }
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700179 task_unlock(task);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000180 return result;
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700181}
182
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800183static int proc_cwd_link(struct dentry *dentry, struct path *path)
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700184{
David Howells2b0143b2015-03-17 22:25:59 +0000185 struct task_struct *task = get_proc_task(d_inode(dentry));
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700186 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700187
188 if (task) {
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200189 task_lock(task);
190 if (task->fs) {
191 get_fs_pwd(task->fs, path);
192 result = 0;
193 }
194 task_unlock(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700195 put_task_struct(task);
196 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 return result;
198}
199
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800200static int proc_root_link(struct dentry *dentry, struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201{
David Howells2b0143b2015-03-17 22:25:59 +0000202 struct task_struct *task = get_proc_task(d_inode(dentry));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700204
205 if (task) {
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200206 result = get_task_root(task, path);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700207 put_task_struct(task);
208 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 return result;
210}
211
Linus Torvaldse4b4e442018-05-17 13:04:17 -0700212static ssize_t get_mm_cmdline(struct mm_struct *mm, char __user *buf,
Linus Torvalds5ab82712018-05-17 15:17:33 -0700213 size_t count, loff_t *ppos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214{
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700215 unsigned long arg_start, arg_end, env_start, env_end;
Linus Torvalds5ab82712018-05-17 15:17:33 -0700216 unsigned long pos, len;
217 char *page;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700218
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700219 /* Check if process spawned far enough to have cmdline. */
Linus Torvaldse4b4e442018-05-17 13:04:17 -0700220 if (!mm->env_end)
221 return 0;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700222
Yang Shi88aa7cc2018-06-07 17:05:28 -0700223 spin_lock(&mm->arg_lock);
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700224 arg_start = mm->arg_start;
225 arg_end = mm->arg_end;
226 env_start = mm->env_start;
227 env_end = mm->env_end;
Yang Shi88aa7cc2018-06-07 17:05:28 -0700228 spin_unlock(&mm->arg_lock);
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700229
Linus Torvalds5ab82712018-05-17 15:17:33 -0700230 if (arg_start >= arg_end)
231 return 0;
Alexey Dobriyan6a6cbe72018-06-07 17:09:59 -0700232
Alexey Dobriyan2ca66ff2014-08-08 14:21:41 -0700233 /*
Linus Torvalds5ab82712018-05-17 15:17:33 -0700234 * We have traditionally allowed the user to re-write
235 * the argument strings and overflow the end result
236 * into the environment section. But only do that if
237 * the environment area is contiguous to the arguments.
Alexey Dobriyan2ca66ff2014-08-08 14:21:41 -0700238 */
Linus Torvalds5ab82712018-05-17 15:17:33 -0700239 if (env_start != arg_end || env_start >= env_end)
240 env_start = env_end = arg_end;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700241
Linus Torvaldsf5b65342018-06-20 09:47:20 +0900242 /* .. and limit it to a maximum of one page of slop */
243 if (env_end >= arg_end + PAGE_SIZE)
244 env_end = arg_end + PAGE_SIZE - 1;
245
Linus Torvalds5ab82712018-05-17 15:17:33 -0700246 /* We're not going to care if "*ppos" has high bits set */
247 pos = arg_start + *ppos;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700248
Linus Torvalds5ab82712018-05-17 15:17:33 -0700249 /* .. but we do check the result is in the proper range */
250 if (pos < arg_start || pos >= env_end)
251 return 0;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700252
Linus Torvalds5ab82712018-05-17 15:17:33 -0700253 /* .. and we never go past env_end */
254 if (env_end - pos < count)
255 count = env_end - pos;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700256
Linus Torvalds5ab82712018-05-17 15:17:33 -0700257 page = (char *)__get_free_page(GFP_KERNEL);
258 if (!page)
259 return -ENOMEM;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700260
Linus Torvalds5ab82712018-05-17 15:17:33 -0700261 len = 0;
262 while (count) {
263 int got;
264 size_t size = min_t(size_t, PAGE_SIZE, count);
Linus Torvaldsf5b65342018-06-20 09:47:20 +0900265 long offset;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700266
Linus Torvaldsf5b65342018-06-20 09:47:20 +0900267 /*
268 * Are we already starting past the official end?
269 * We always include the last byte that is *supposed*
270 * to be NUL
271 */
272 offset = (pos >= arg_end) ? pos - arg_end + 1 : 0;
273
274 got = access_remote_vm(mm, pos - offset, page, size + offset, FOLL_ANON);
275 if (got <= offset)
Linus Torvalds5ab82712018-05-17 15:17:33 -0700276 break;
Linus Torvaldsf5b65342018-06-20 09:47:20 +0900277 got -= offset;
Linus Torvalds5ab82712018-05-17 15:17:33 -0700278
279 /* Don't walk past a NUL character once you hit arg_end */
280 if (pos + got >= arg_end) {
281 int n = 0;
282
283 /*
284 * If we started before 'arg_end' but ended up
285 * at or after it, we start the NUL character
286 * check at arg_end-1 (where we expect the normal
287 * EOF to be).
288 *
289 * NOTE! This is smaller than 'got', because
290 * pos + got >= arg_end
291 */
292 if (pos < arg_end)
293 n = arg_end - pos - 1;
294
295 /* Cut off at first NUL after 'n' */
Linus Torvaldsf5b65342018-06-20 09:47:20 +0900296 got = n + strnlen(page+n, offset+got-n);
297 if (got < offset)
Linus Torvalds5ab82712018-05-17 15:17:33 -0700298 break;
Linus Torvaldsf5b65342018-06-20 09:47:20 +0900299 got -= offset;
300
301 /* Include the NUL if it existed */
302 if (got < size)
303 got++;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700304 }
Alexey Dobriyana0a07b82017-02-24 15:00:20 -0800305
Linus Torvaldsf5b65342018-06-20 09:47:20 +0900306 got -= copy_to_user(buf, page+offset, got);
Linus Torvalds5ab82712018-05-17 15:17:33 -0700307 if (unlikely(!got)) {
308 if (!len)
309 len = -EFAULT;
310 break;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700311 }
Linus Torvalds5ab82712018-05-17 15:17:33 -0700312 pos += got;
313 buf += got;
314 len += got;
315 count -= got;
Alexey Dobriyan3cb4e162018-06-07 17:10:02 -0700316 }
317
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700318 free_page((unsigned long)page);
Linus Torvalds5ab82712018-05-17 15:17:33 -0700319 return len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320}
321
Linus Torvaldse4b4e442018-05-17 13:04:17 -0700322static ssize_t get_task_cmdline(struct task_struct *tsk, char __user *buf,
323 size_t count, loff_t *pos)
324{
325 struct mm_struct *mm;
326 ssize_t ret;
327
328 mm = get_task_mm(tsk);
329 if (!mm)
330 return 0;
331
332 ret = get_mm_cmdline(mm, buf, count, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 mmput(mm);
Linus Torvaldse4b4e442018-05-17 13:04:17 -0700334 return ret;
335}
336
337static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
338 size_t count, loff_t *pos)
339{
340 struct task_struct *tsk;
341 ssize_t ret;
342
343 BUG_ON(*pos < 0);
344
345 tsk = get_proc_task(file_inode(file));
346 if (!tsk)
347 return -ESRCH;
348 ret = get_task_cmdline(tsk, buf, count, pos);
349 put_task_struct(tsk);
350 if (ret > 0)
351 *pos += ret;
352 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353}
354
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700355static const struct file_operations proc_pid_cmdline_ops = {
356 .read = proc_pid_cmdline_read,
357 .llseek = generic_file_llseek,
358};
359
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360#ifdef CONFIG_KALLSYMS
361/*
362 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
363 * Returns the resolved symbol. If that fails, simply return the address.
364 */
Alexey Dobriyanedfcd602014-08-08 14:21:44 -0700365static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
366 struct pid *pid, struct task_struct *task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367{
Alexey Dobriyanffb45122007-05-08 00:28:41 -0700368 unsigned long wchan;
Tejun Heo9281ace2007-07-17 04:03:51 -0700369 char symname[KSYM_NAME_LEN];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370
Alexey Dobriyan24b2ec22018-04-10 16:31:30 -0700371 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
372 goto print0;
373
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374 wchan = get_wchan(task);
Alexey Dobriyan24b2ec22018-04-10 16:31:30 -0700375 if (wchan && !lookup_symbol_name(wchan, symname)) {
Alexey Dobriyan21dae0a2018-04-10 16:31:34 -0700376 seq_puts(m, symname);
Alexey Dobriyan24b2ec22018-04-10 16:31:30 -0700377 return 0;
378 }
Joe Perches25ce3192015-04-15 16:18:17 -0700379
Alexey Dobriyan24b2ec22018-04-10 16:31:30 -0700380print0:
381 seq_putc(m, '0');
Joe Perches25ce3192015-04-15 16:18:17 -0700382 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383}
384#endif /* CONFIG_KALLSYMS */
385
Al Viroa9712bc2011-03-23 15:52:50 -0400386static int lock_trace(struct task_struct *task)
387{
388 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
389 if (err)
390 return err;
Jann Horncaaee622016-01-20 15:00:04 -0800391 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
Al Viroa9712bc2011-03-23 15:52:50 -0400392 mutex_unlock(&task->signal->cred_guard_mutex);
393 return -EPERM;
394 }
395 return 0;
396}
397
398static void unlock_trace(struct task_struct *task)
399{
400 mutex_unlock(&task->signal->cred_guard_mutex);
401}
402
Ken Chen2ec220e2008-11-10 11:26:08 +0300403#ifdef CONFIG_STACKTRACE
404
405#define MAX_STACK_TRACE_DEPTH 64
406
407static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
408 struct pid *pid, struct task_struct *task)
409{
Ken Chen2ec220e2008-11-10 11:26:08 +0300410 unsigned long *entries;
Al Viroa9712bc2011-03-23 15:52:50 -0400411 int err;
Ken Chen2ec220e2008-11-10 11:26:08 +0300412
Jann Hornf8a00ce2018-10-05 15:51:58 -0700413 /*
414 * The ability to racily run the kernel stack unwinder on a running task
415 * and then observe the unwinder output is scary; while it is useful for
416 * debugging kernel issues, it can also allow an attacker to leak kernel
417 * stack contents.
418 * Doing this in a manner that is at least safe from races would require
419 * some work to ensure that the remote task can not be scheduled; and
420 * even then, this would still expose the unwinder as local attack
421 * surface.
422 * Therefore, this interface is restricted to root.
423 */
424 if (!file_ns_capable(m->file, &init_user_ns, CAP_SYS_ADMIN))
425 return -EACCES;
426
Kees Cook6da2ec52018-06-12 13:55:00 -0700427 entries = kmalloc_array(MAX_STACK_TRACE_DEPTH, sizeof(*entries),
428 GFP_KERNEL);
Ken Chen2ec220e2008-11-10 11:26:08 +0300429 if (!entries)
430 return -ENOMEM;
431
Al Viroa9712bc2011-03-23 15:52:50 -0400432 err = lock_trace(task);
433 if (!err) {
Thomas Gleixnere988e5e2019-04-25 11:44:58 +0200434 unsigned int i, nr_entries;
Alexey Dobriyan5d008fb2018-06-07 17:10:17 -0700435
Thomas Gleixnere988e5e2019-04-25 11:44:58 +0200436 nr_entries = stack_trace_save_tsk(task, entries,
437 MAX_STACK_TRACE_DEPTH, 0);
Al Viroa9712bc2011-03-23 15:52:50 -0400438
Thomas Gleixnere988e5e2019-04-25 11:44:58 +0200439 for (i = 0; i < nr_entries; i++) {
Linus Torvalds8f5abe82017-11-27 16:45:56 -0800440 seq_printf(m, "[<0>] %pB\n", (void *)entries[i]);
Al Viroa9712bc2011-03-23 15:52:50 -0400441 }
Thomas Gleixnere988e5e2019-04-25 11:44:58 +0200442
Al Viroa9712bc2011-03-23 15:52:50 -0400443 unlock_trace(task);
Ken Chen2ec220e2008-11-10 11:26:08 +0300444 }
445 kfree(entries);
446
Al Viroa9712bc2011-03-23 15:52:50 -0400447 return err;
Ken Chen2ec220e2008-11-10 11:26:08 +0300448}
449#endif
450
Naveen N. Rao5968cec2015-06-30 14:36:03 +0530451#ifdef CONFIG_SCHED_INFO
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452/*
453 * Provides /proc/PID/schedstat
454 */
Alexey Dobriyanf6e826c2014-08-08 14:21:46 -0700455static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
456 struct pid *pid, struct task_struct *task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457{
Naveen N. Rao5968cec2015-06-30 14:36:03 +0530458 if (unlikely(!sched_info_on()))
Alexey Dobriyan08b55772019-03-05 15:50:35 -0800459 seq_puts(m, "0 0 0\n");
Naveen N. Rao5968cec2015-06-30 14:36:03 +0530460 else
461 seq_printf(m, "%llu %llu %lu\n",
Joe Perches25ce3192015-04-15 16:18:17 -0700462 (unsigned long long)task->se.sum_exec_runtime,
463 (unsigned long long)task->sched_info.run_delay,
464 task->sched_info.pcount);
465
466 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467}
468#endif
469
Arjan van de Ven97455122008-01-25 21:08:34 +0100470#ifdef CONFIG_LATENCYTOP
471static int lstats_show_proc(struct seq_file *m, void *v)
472{
473 int i;
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800474 struct inode *inode = m->private;
475 struct task_struct *task = get_proc_task(inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100476
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800477 if (!task)
478 return -ESRCH;
479 seq_puts(m, "Latency Top version : v0.1\n");
Alexey Dobriyanf6d2f582018-08-21 21:54:34 -0700480 for (i = 0; i < LT_SAVECOUNT; i++) {
Joe Perches34e49d42011-01-12 17:00:30 -0800481 struct latency_record *lr = &task->latency_record[i];
482 if (lr->backtrace[0]) {
Arjan van de Ven97455122008-01-25 21:08:34 +0100483 int q;
Joe Perches34e49d42011-01-12 17:00:30 -0800484 seq_printf(m, "%i %li %li",
485 lr->count, lr->time, lr->max);
Arjan van de Ven97455122008-01-25 21:08:34 +0100486 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
Joe Perches34e49d42011-01-12 17:00:30 -0800487 unsigned long bt = lr->backtrace[q];
Thomas Gleixneraccddc42019-04-10 12:28:08 +0200488
Joe Perches34e49d42011-01-12 17:00:30 -0800489 if (!bt)
Arjan van de Ven97455122008-01-25 21:08:34 +0100490 break;
Joe Perches34e49d42011-01-12 17:00:30 -0800491 seq_printf(m, " %ps", (void *)bt);
Arjan van de Ven97455122008-01-25 21:08:34 +0100492 }
Alexey Dobriyan9d6de122011-01-12 17:00:32 -0800493 seq_putc(m, '\n');
Arjan van de Ven97455122008-01-25 21:08:34 +0100494 }
495
496 }
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800497 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100498 return 0;
499}
500
501static int lstats_open(struct inode *inode, struct file *file)
502{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800503 return single_open(file, lstats_show_proc, inode);
Hiroshi Shimamotod6643d12008-02-14 10:27:00 -0800504}
505
Arjan van de Ven97455122008-01-25 21:08:34 +0100506static ssize_t lstats_write(struct file *file, const char __user *buf,
507 size_t count, loff_t *offs)
508{
Al Viro496ad9a2013-01-23 17:07:38 -0500509 struct task_struct *task = get_proc_task(file_inode(file));
Arjan van de Ven97455122008-01-25 21:08:34 +0100510
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800511 if (!task)
512 return -ESRCH;
Lin Fenge02c9b02019-05-14 15:42:34 -0700513 clear_tsk_latency_tracing(task);
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800514 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100515
516 return count;
517}
518
519static const struct file_operations proc_lstats_operations = {
520 .open = lstats_open,
521 .read = seq_read,
522 .write = lstats_write,
523 .llseek = seq_lseek,
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800524 .release = single_release,
Arjan van de Ven97455122008-01-25 21:08:34 +0100525};
526
527#endif
528
Alexey Dobriyan6ba51e32014-08-08 14:21:48 -0700529static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
530 struct pid *pid, struct task_struct *task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531{
Arun KSca79b0c2018-12-28 00:34:29 -0800532 unsigned long totalpages = totalram_pages() + total_swap_pages;
Oleg Nesterovb95c35e2010-04-01 15:13:57 +0200533 unsigned long points = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534
Oleg Nesterovef419392016-08-02 14:03:19 -0700535 points = oom_badness(task, NULL, NULL, totalpages) *
536 1000 / totalpages;
Joe Perches25ce3192015-04-15 16:18:17 -0700537 seq_printf(m, "%lu\n", points);
538
539 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540}
541
Neil Hormand85f50d2007-10-18 23:40:37 -0700542struct limit_names {
Alexey Dobriyancedbcca2014-08-08 14:21:33 -0700543 const char *name;
544 const char *unit;
Neil Hormand85f50d2007-10-18 23:40:37 -0700545};
546
547static const struct limit_names lnames[RLIM_NLIMITS] = {
Kees Cookcff4edb2009-09-22 16:45:32 -0700548 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700549 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
550 [RLIMIT_DATA] = {"Max data size", "bytes"},
551 [RLIMIT_STACK] = {"Max stack size", "bytes"},
552 [RLIMIT_CORE] = {"Max core file size", "bytes"},
553 [RLIMIT_RSS] = {"Max resident set", "bytes"},
554 [RLIMIT_NPROC] = {"Max processes", "processes"},
555 [RLIMIT_NOFILE] = {"Max open files", "files"},
556 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
557 [RLIMIT_AS] = {"Max address space", "bytes"},
558 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
559 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
560 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
561 [RLIMIT_NICE] = {"Max nice priority", NULL},
562 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
Eugene Teo88081172008-02-23 15:23:52 -0800563 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700564};
565
566/* Display limits for a process */
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700567static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
568 struct pid *pid, struct task_struct *task)
Neil Hormand85f50d2007-10-18 23:40:37 -0700569{
570 unsigned int i;
Neil Hormand85f50d2007-10-18 23:40:37 -0700571 unsigned long flags;
Neil Hormand85f50d2007-10-18 23:40:37 -0700572
573 struct rlimit rlim[RLIM_NLIMITS];
574
Lai Jiangshana6bebbc2008-10-05 00:51:15 +0400575 if (!lock_task_sighand(task, &flags))
Neil Hormand85f50d2007-10-18 23:40:37 -0700576 return 0;
Neil Hormand85f50d2007-10-18 23:40:37 -0700577 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
578 unlock_task_sighand(task, &flags);
Neil Hormand85f50d2007-10-18 23:40:37 -0700579
580 /*
581 * print the file header
582 */
Alexey Dobriyanafe922c2019-01-03 15:26:09 -0800583 seq_puts(m, "Limit "
584 "Soft Limit "
585 "Hard Limit "
586 "Units \n");
Neil Hormand85f50d2007-10-18 23:40:37 -0700587
588 for (i = 0; i < RLIM_NLIMITS; i++) {
589 if (rlim[i].rlim_cur == RLIM_INFINITY)
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700590 seq_printf(m, "%-25s %-20s ",
Joe Perches25ce3192015-04-15 16:18:17 -0700591 lnames[i].name, "unlimited");
Neil Hormand85f50d2007-10-18 23:40:37 -0700592 else
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700593 seq_printf(m, "%-25s %-20lu ",
Joe Perches25ce3192015-04-15 16:18:17 -0700594 lnames[i].name, rlim[i].rlim_cur);
Neil Hormand85f50d2007-10-18 23:40:37 -0700595
596 if (rlim[i].rlim_max == RLIM_INFINITY)
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700597 seq_printf(m, "%-20s ", "unlimited");
Neil Hormand85f50d2007-10-18 23:40:37 -0700598 else
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700599 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
Neil Hormand85f50d2007-10-18 23:40:37 -0700600
601 if (lnames[i].unit)
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700602 seq_printf(m, "%-10s\n", lnames[i].unit);
Neil Hormand85f50d2007-10-18 23:40:37 -0700603 else
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700604 seq_putc(m, '\n');
Neil Hormand85f50d2007-10-18 23:40:37 -0700605 }
606
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700607 return 0;
Neil Hormand85f50d2007-10-18 23:40:37 -0700608}
609
Roland McGrathebcb6732008-07-25 19:46:00 -0700610#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -0700611static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
612 struct pid *pid, struct task_struct *task)
Roland McGrathebcb6732008-07-25 19:46:00 -0700613{
Steven Rostedt (Red Hat)631b7ab2016-11-07 16:26:35 -0500614 struct syscall_info info;
615 u64 *args = &info.data.args[0];
Joe Perches25ce3192015-04-15 16:18:17 -0700616 int res;
617
618 res = lock_trace(task);
Al Viroa9712bc2011-03-23 15:52:50 -0400619 if (res)
620 return res;
Roland McGrathebcb6732008-07-25 19:46:00 -0700621
Steven Rostedt (Red Hat)631b7ab2016-11-07 16:26:35 -0500622 if (task_current_syscall(task, &info))
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -0700623 seq_puts(m, "running\n");
Steven Rostedt (Red Hat)631b7ab2016-11-07 16:26:35 -0500624 else if (info.data.nr < 0)
625 seq_printf(m, "%d 0x%llx 0x%llx\n",
626 info.data.nr, info.sp, info.data.instruction_pointer);
Al Viroa9712bc2011-03-23 15:52:50 -0400627 else
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -0700628 seq_printf(m,
Steven Rostedt (Red Hat)631b7ab2016-11-07 16:26:35 -0500629 "%d 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx 0x%llx\n",
630 info.data.nr,
Roland McGrathebcb6732008-07-25 19:46:00 -0700631 args[0], args[1], args[2], args[3], args[4], args[5],
Steven Rostedt (Red Hat)631b7ab2016-11-07 16:26:35 -0500632 info.sp, info.data.instruction_pointer);
Al Viroa9712bc2011-03-23 15:52:50 -0400633 unlock_trace(task);
Joe Perches25ce3192015-04-15 16:18:17 -0700634
635 return 0;
Roland McGrathebcb6732008-07-25 19:46:00 -0700636}
637#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
638
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639/************************************************************************/
640/* Here the fs part begins */
641/************************************************************************/
642
643/* permission checks */
Eric W. Biederman778c1142006-06-26 00:25:58 -0700644static int proc_fd_access_allowed(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645{
Eric W. Biederman778c1142006-06-26 00:25:58 -0700646 struct task_struct *task;
647 int allowed = 0;
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700648 /* Allow access to a task's file descriptors if it is us or we
649 * may use ptrace attach to the process and find out that
650 * information.
Eric W. Biederman778c1142006-06-26 00:25:58 -0700651 */
652 task = get_proc_task(inode);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700653 if (task) {
Jann Horncaaee622016-01-20 15:00:04 -0800654 allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
Eric W. Biederman778c1142006-06-26 00:25:58 -0700655 put_task_struct(task);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700656 }
Eric W. Biederman778c1142006-06-26 00:25:58 -0700657 return allowed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658}
659
Eric W. Biederman6b4e3062010-03-07 16:41:34 -0800660int proc_setattr(struct dentry *dentry, struct iattr *attr)
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700661{
662 int error;
David Howells2b0143b2015-03-17 22:25:59 +0000663 struct inode *inode = d_inode(dentry);
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700664
665 if (attr->ia_valid & ATTR_MODE)
666 return -EPERM;
667
Jan Kara31051c82016-05-26 16:55:18 +0200668 error = setattr_prepare(dentry, attr);
Christoph Hellwig10257742010-06-04 11:30:02 +0200669 if (error)
670 return error;
671
Christoph Hellwig10257742010-06-04 11:30:02 +0200672 setattr_copy(inode, attr);
673 mark_inode_dirty(inode);
674 return 0;
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700675}
676
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800677/*
678 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
679 * or euid/egid (for hide_pid_min=2)?
680 */
681static bool has_pid_permissions(struct pid_namespace *pid,
682 struct task_struct *task,
683 int hide_pid_min)
684{
685 if (pid->hide_pid < hide_pid_min)
686 return true;
687 if (in_group_p(pid->pid_gid))
688 return true;
Jann Horncaaee622016-01-20 15:00:04 -0800689 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800690}
691
692
693static int proc_pid_permission(struct inode *inode, int mask)
694{
Christoph Hellwig76f668b2018-05-16 07:19:01 +0200695 struct pid_namespace *pid = proc_pid_ns(inode);
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800696 struct task_struct *task;
697 bool has_perms;
698
699 task = get_proc_task(inode);
Xiaotian Fenga2ef9902012-01-12 17:17:08 -0800700 if (!task)
701 return -ESRCH;
Lafcadio Wluiki796f5712017-02-24 15:00:23 -0800702 has_perms = has_pid_permissions(pid, task, HIDEPID_NO_ACCESS);
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800703 put_task_struct(task);
704
705 if (!has_perms) {
Lafcadio Wluiki796f5712017-02-24 15:00:23 -0800706 if (pid->hide_pid == HIDEPID_INVISIBLE) {
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800707 /*
708 * Let's make getdents(), stat(), and open()
709 * consistent with each other. If a process
710 * may not stat() a file, it shouldn't be seen
711 * in procfs at all.
712 */
713 return -ENOENT;
714 }
715
716 return -EPERM;
717 }
718 return generic_permission(inode, mask);
719}
720
721
722
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800723static const struct inode_operations proc_def_inode_operations = {
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700724 .setattr = proc_setattr,
725};
726
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800727static int proc_single_show(struct seq_file *m, void *v)
728{
729 struct inode *inode = m->private;
Christoph Hellwig76f668b2018-05-16 07:19:01 +0200730 struct pid_namespace *ns = proc_pid_ns(inode);
731 struct pid *pid = proc_pid(inode);
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800732 struct task_struct *task;
733 int ret;
734
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800735 task = get_pid_task(pid, PIDTYPE_PID);
736 if (!task)
737 return -ESRCH;
738
739 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
740
741 put_task_struct(task);
742 return ret;
743}
744
745static int proc_single_open(struct inode *inode, struct file *filp)
746{
Jovi Zhangc6a34052011-01-12 17:00:34 -0800747 return single_open(filp, proc_single_show, inode);
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800748}
749
750static const struct file_operations proc_single_file_operations = {
751 .open = proc_single_open,
752 .read = seq_read,
753 .llseek = seq_lseek,
754 .release = single_release,
755};
756
Oleg Nesterov5381e162014-10-09 15:25:24 -0700757
758struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
759{
760 struct task_struct *task = get_proc_task(inode);
761 struct mm_struct *mm = ERR_PTR(-ESRCH);
762
763 if (task) {
Jann Horncaaee622016-01-20 15:00:04 -0800764 mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
Oleg Nesterov5381e162014-10-09 15:25:24 -0700765 put_task_struct(task);
766
767 if (!IS_ERR_OR_NULL(mm)) {
768 /* ensure this mm_struct can't be freed */
Vegard Nossumf1f10072017-02-27 14:30:07 -0800769 mmgrab(mm);
Oleg Nesterov5381e162014-10-09 15:25:24 -0700770 /* but do not pin its memory */
771 mmput(mm);
772 }
773 }
774
775 return mm;
776}
777
Cong Wangb409e572012-05-31 16:26:17 -0700778static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779{
Oleg Nesterov5381e162014-10-09 15:25:24 -0700780 struct mm_struct *mm = proc_mem_open(inode, mode);
Linus Torvaldse2683372012-01-17 15:21:19 -0800781
782 if (IS_ERR(mm))
783 return PTR_ERR(mm);
784
Linus Torvaldse2683372012-01-17 15:21:19 -0800785 file->private_data = mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786 return 0;
787}
788
Cong Wangb409e572012-05-31 16:26:17 -0700789static int mem_open(struct inode *inode, struct file *file)
790{
Djalal Harounibc452b42012-07-30 14:42:28 -0700791 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
792
793 /* OK to pass negative loff_t, we can catch out-of-range */
794 file->f_mode |= FMODE_UNSIGNED_OFFSET;
795
796 return ret;
Cong Wangb409e572012-05-31 16:26:17 -0700797}
798
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100799static ssize_t mem_rw(struct file *file, char __user *buf,
800 size_t count, loff_t *ppos, int write)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801{
Linus Torvaldse2683372012-01-17 15:21:19 -0800802 struct mm_struct *mm = file->private_data;
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100803 unsigned long addr = *ppos;
804 ssize_t copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 char *page;
Linus Torvalds272ddc82016-10-24 19:00:44 -0700806 unsigned int flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807
Linus Torvaldse2683372012-01-17 15:21:19 -0800808 if (!mm)
809 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810
Michal Hocko0ee931c2017-09-13 16:28:29 -0700811 page = (char *)__get_free_page(GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 if (!page)
Linus Torvaldse2683372012-01-17 15:21:19 -0800813 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700815 copied = 0;
Vegard Nossum388f7932017-02-27 14:30:13 -0800816 if (!mmget_not_zero(mm))
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100817 goto free;
818
Linus Torvaldsf511c0b2017-05-30 12:38:59 -0700819 flags = FOLL_FORCE | (write ? FOLL_WRITE : 0);
Lorenzo Stoakes6347e8d2016-10-13 01:20:19 +0100820
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 while (count > 0) {
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100822 int this_len = min_t(int, count, PAGE_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100824 if (write && copy_from_user(page, buf, this_len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 copied = -EFAULT;
826 break;
827 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100828
Lorenzo Stoakes6347e8d2016-10-13 01:20:19 +0100829 this_len = access_remote_vm(mm, addr, page, this_len, flags);
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100830 if (!this_len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 if (!copied)
832 copied = -EIO;
833 break;
834 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100835
836 if (!write && copy_to_user(buf, page, this_len)) {
837 copied = -EFAULT;
838 break;
839 }
840
841 buf += this_len;
842 addr += this_len;
843 copied += this_len;
844 count -= this_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100846 *ppos = addr;
KOSAKI Motohiro30cd8902011-05-26 16:25:52 -0700847
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100848 mmput(mm);
849free:
KOSAKI Motohiro30cd8902011-05-26 16:25:52 -0700850 free_page((unsigned long) page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 return copied;
852}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100854static ssize_t mem_read(struct file *file, char __user *buf,
855 size_t count, loff_t *ppos)
856{
857 return mem_rw(file, buf, count, ppos, 0);
858}
859
860static ssize_t mem_write(struct file *file, const char __user *buf,
861 size_t count, loff_t *ppos)
862{
863 return mem_rw(file, (char __user*)buf, count, ppos, 1);
864}
865
Matt Mackall85863e42008-02-04 22:29:04 -0800866loff_t mem_lseek(struct file *file, loff_t offset, int orig)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867{
868 switch (orig) {
869 case 0:
870 file->f_pos = offset;
871 break;
872 case 1:
873 file->f_pos += offset;
874 break;
875 default:
876 return -EINVAL;
877 }
878 force_successful_syscall_return();
879 return file->f_pos;
880}
881
Linus Torvaldse2683372012-01-17 15:21:19 -0800882static int mem_release(struct inode *inode, struct file *file)
883{
884 struct mm_struct *mm = file->private_data;
Oleg Nesterov71879d32012-01-31 17:14:38 +0100885 if (mm)
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100886 mmdrop(mm);
Linus Torvaldse2683372012-01-17 15:21:19 -0800887 return 0;
888}
889
Arjan van de Ven00977a52007-02-12 00:55:34 -0800890static const struct file_operations proc_mem_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 .llseek = mem_lseek,
892 .read = mem_read,
893 .write = mem_write,
894 .open = mem_open,
Linus Torvaldse2683372012-01-17 15:21:19 -0800895 .release = mem_release,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896};
897
Cong Wangb409e572012-05-31 16:26:17 -0700898static int environ_open(struct inode *inode, struct file *file)
899{
900 return __mem_open(inode, file, PTRACE_MODE_READ);
901}
902
James Pearson315e28c2007-10-16 23:30:17 -0700903static ssize_t environ_read(struct file *file, char __user *buf,
904 size_t count, loff_t *ppos)
905{
James Pearson315e28c2007-10-16 23:30:17 -0700906 char *page;
907 unsigned long src = *ppos;
Cong Wangb409e572012-05-31 16:26:17 -0700908 int ret = 0;
909 struct mm_struct *mm = file->private_data;
Mateusz Guzika3b609e2016-01-20 15:01:05 -0800910 unsigned long env_start, env_end;
James Pearson315e28c2007-10-16 23:30:17 -0700911
Mathias Krause8148a732016-05-05 16:22:26 -0700912 /* Ensure the process spawned far enough to have an environment. */
913 if (!mm || !mm->env_end)
Cong Wangb409e572012-05-31 16:26:17 -0700914 return 0;
James Pearson315e28c2007-10-16 23:30:17 -0700915
Michal Hocko0ee931c2017-09-13 16:28:29 -0700916 page = (char *)__get_free_page(GFP_KERNEL);
James Pearson315e28c2007-10-16 23:30:17 -0700917 if (!page)
Cong Wangb409e572012-05-31 16:26:17 -0700918 return -ENOMEM;
James Pearson315e28c2007-10-16 23:30:17 -0700919
Al Virod6f64b82011-02-15 22:26:01 -0500920 ret = 0;
Vegard Nossum388f7932017-02-27 14:30:13 -0800921 if (!mmget_not_zero(mm))
Cong Wangb409e572012-05-31 16:26:17 -0700922 goto free;
Mateusz Guzika3b609e2016-01-20 15:01:05 -0800923
Yang Shi88aa7cc2018-06-07 17:05:28 -0700924 spin_lock(&mm->arg_lock);
Mateusz Guzika3b609e2016-01-20 15:01:05 -0800925 env_start = mm->env_start;
926 env_end = mm->env_end;
Yang Shi88aa7cc2018-06-07 17:05:28 -0700927 spin_unlock(&mm->arg_lock);
Mateusz Guzika3b609e2016-01-20 15:01:05 -0800928
James Pearson315e28c2007-10-16 23:30:17 -0700929 while (count > 0) {
Djalal Harounie8905ec2012-07-30 14:42:26 -0700930 size_t this_len, max_len;
931 int retval;
932
Mateusz Guzika3b609e2016-01-20 15:01:05 -0800933 if (src >= (env_end - env_start))
Djalal Harounie8905ec2012-07-30 14:42:26 -0700934 break;
James Pearson315e28c2007-10-16 23:30:17 -0700935
Mateusz Guzika3b609e2016-01-20 15:01:05 -0800936 this_len = env_end - (env_start + src);
James Pearson315e28c2007-10-16 23:30:17 -0700937
Djalal Harounie8905ec2012-07-30 14:42:26 -0700938 max_len = min_t(size_t, PAGE_SIZE, count);
939 this_len = min(max_len, this_len);
James Pearson315e28c2007-10-16 23:30:17 -0700940
Willy Tarreau7f7ccc22018-05-11 08:11:44 +0200941 retval = access_remote_vm(mm, (env_start + src), page, this_len, FOLL_ANON);
James Pearson315e28c2007-10-16 23:30:17 -0700942
943 if (retval <= 0) {
944 ret = retval;
945 break;
946 }
947
948 if (copy_to_user(buf, page, retval)) {
949 ret = -EFAULT;
950 break;
951 }
952
953 ret += retval;
954 src += retval;
955 buf += retval;
956 count -= retval;
957 }
958 *ppos = src;
James Pearson315e28c2007-10-16 23:30:17 -0700959 mmput(mm);
Cong Wangb409e572012-05-31 16:26:17 -0700960
961free:
James Pearson315e28c2007-10-16 23:30:17 -0700962 free_page((unsigned long) page);
James Pearson315e28c2007-10-16 23:30:17 -0700963 return ret;
964}
965
966static const struct file_operations proc_environ_operations = {
Cong Wangb409e572012-05-31 16:26:17 -0700967 .open = environ_open,
James Pearson315e28c2007-10-16 23:30:17 -0700968 .read = environ_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +0100969 .llseek = generic_file_llseek,
Cong Wangb409e572012-05-31 16:26:17 -0700970 .release = mem_release,
James Pearson315e28c2007-10-16 23:30:17 -0700971};
972
Al Viroc5317162016-10-05 18:43:43 -0400973static int auxv_open(struct inode *inode, struct file *file)
974{
975 return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
976}
977
978static ssize_t auxv_read(struct file *file, char __user *buf,
979 size_t count, loff_t *ppos)
980{
981 struct mm_struct *mm = file->private_data;
982 unsigned int nwords = 0;
Leon Yu06b28492016-10-27 17:46:50 -0700983
984 if (!mm)
985 return 0;
Al Viroc5317162016-10-05 18:43:43 -0400986 do {
987 nwords += 2;
988 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
989 return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
990 nwords * sizeof(mm->saved_auxv[0]));
991}
992
993static const struct file_operations proc_auxv_operations = {
994 .open = auxv_open,
995 .read = auxv_read,
996 .llseek = generic_file_llseek,
997 .release = mem_release,
998};
999
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001000static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1001 loff_t *ppos)
1002{
Al Viro496ad9a2013-01-23 17:07:38 -05001003 struct task_struct *task = get_proc_task(file_inode(file));
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001004 char buffer[PROC_NUMBUF];
1005 int oom_adj = OOM_ADJUST_MIN;
1006 size_t len;
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001007
1008 if (!task)
1009 return -ESRCH;
Michal Hockof913da52016-07-28 15:44:37 -07001010 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1011 oom_adj = OOM_ADJUST_MAX;
1012 else
1013 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1014 OOM_SCORE_ADJ_MAX;
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001015 put_task_struct(task);
1016 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1017 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1018}
1019
Michal Hocko1d5f0ac2016-07-28 15:44:40 -07001020static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1021{
1022 static DEFINE_MUTEX(oom_adj_mutex);
Michal Hocko44a70ade2016-07-28 15:44:43 -07001023 struct mm_struct *mm = NULL;
Michal Hocko1d5f0ac2016-07-28 15:44:40 -07001024 struct task_struct *task;
1025 int err = 0;
1026
1027 task = get_proc_task(file_inode(file));
1028 if (!task)
1029 return -ESRCH;
1030
1031 mutex_lock(&oom_adj_mutex);
1032 if (legacy) {
1033 if (oom_adj < task->signal->oom_score_adj &&
1034 !capable(CAP_SYS_RESOURCE)) {
1035 err = -EACCES;
1036 goto err_unlock;
1037 }
1038 /*
1039 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1040 * /proc/pid/oom_score_adj instead.
1041 */
1042 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1043 current->comm, task_pid_nr(current), task_pid_nr(task),
1044 task_pid_nr(task));
1045 } else {
1046 if ((short)oom_adj < task->signal->oom_score_adj_min &&
1047 !capable(CAP_SYS_RESOURCE)) {
1048 err = -EACCES;
1049 goto err_unlock;
1050 }
1051 }
1052
Michal Hocko44a70ade2016-07-28 15:44:43 -07001053 /*
1054 * Make sure we will check other processes sharing the mm if this is
1055 * not vfrok which wants its own oom_score_adj.
1056 * pin the mm so it doesn't go away and get reused after task_unlock
1057 */
1058 if (!task->vfork_done) {
1059 struct task_struct *p = find_lock_task_mm(task);
1060
1061 if (p) {
1062 if (atomic_read(&p->mm->mm_users) > 1) {
1063 mm = p->mm;
Vegard Nossumf1f10072017-02-27 14:30:07 -08001064 mmgrab(mm);
Michal Hocko44a70ade2016-07-28 15:44:43 -07001065 }
1066 task_unlock(p);
1067 }
1068 }
1069
Michal Hocko1d5f0ac2016-07-28 15:44:40 -07001070 task->signal->oom_score_adj = oom_adj;
1071 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1072 task->signal->oom_score_adj_min = (short)oom_adj;
1073 trace_oom_score_adj_update(task);
Michal Hocko44a70ade2016-07-28 15:44:43 -07001074
1075 if (mm) {
1076 struct task_struct *p;
1077
1078 rcu_read_lock();
1079 for_each_process(p) {
1080 if (same_thread_group(task, p))
1081 continue;
1082
1083 /* do not touch kernel threads or the global init */
1084 if (p->flags & PF_KTHREAD || is_global_init(p))
1085 continue;
1086
1087 task_lock(p);
1088 if (!p->vfork_done && process_shares_mm(p, mm)) {
Michal Hocko44a70ade2016-07-28 15:44:43 -07001089 p->signal->oom_score_adj = oom_adj;
1090 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1091 p->signal->oom_score_adj_min = (short)oom_adj;
1092 }
1093 task_unlock(p);
1094 }
1095 rcu_read_unlock();
1096 mmdrop(mm);
1097 }
Michal Hocko1d5f0ac2016-07-28 15:44:40 -07001098err_unlock:
1099 mutex_unlock(&oom_adj_mutex);
1100 put_task_struct(task);
1101 return err;
1102}
Michal Hockof913da52016-07-28 15:44:37 -07001103
David Rientjesb72bdfa2015-11-05 18:50:32 -08001104/*
1105 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1106 * kernels. The effective policy is defined by oom_score_adj, which has a
1107 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1108 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1109 * Processes that become oom disabled via oom_adj will still be oom disabled
1110 * with this implementation.
1111 *
1112 * oom_adj cannot be removed since existing userspace binaries use it.
1113 */
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001114static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1115 size_t count, loff_t *ppos)
1116{
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001117 char buffer[PROC_NUMBUF];
1118 int oom_adj;
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001119 int err;
1120
1121 memset(buffer, 0, sizeof(buffer));
1122 if (count > sizeof(buffer) - 1)
1123 count = sizeof(buffer) - 1;
1124 if (copy_from_user(buffer, buf, count)) {
1125 err = -EFAULT;
1126 goto out;
1127 }
1128
1129 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1130 if (err)
1131 goto out;
1132 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1133 oom_adj != OOM_DISABLE) {
1134 err = -EINVAL;
1135 goto out;
1136 }
1137
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001138 /*
1139 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1140 * value is always attainable.
1141 */
1142 if (oom_adj == OOM_ADJUST_MAX)
1143 oom_adj = OOM_SCORE_ADJ_MAX;
1144 else
1145 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1146
Michal Hocko1d5f0ac2016-07-28 15:44:40 -07001147 err = __set_oom_adj(file, oom_adj, true);
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001148out:
1149 return err < 0 ? err : count;
1150}
1151
1152static const struct file_operations proc_oom_adj_operations = {
1153 .read = oom_adj_read,
1154 .write = oom_adj_write,
1155 .llseek = generic_file_llseek,
1156};
1157
David Rientjesa63d83f2010-08-09 17:19:46 -07001158static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1159 size_t count, loff_t *ppos)
1160{
Al Viro496ad9a2013-01-23 17:07:38 -05001161 struct task_struct *task = get_proc_task(file_inode(file));
David Rientjesa63d83f2010-08-09 17:19:46 -07001162 char buffer[PROC_NUMBUF];
David Rientjesa9c58b902012-12-11 16:02:54 -08001163 short oom_score_adj = OOM_SCORE_ADJ_MIN;
David Rientjesa63d83f2010-08-09 17:19:46 -07001164 size_t len;
1165
1166 if (!task)
1167 return -ESRCH;
Michal Hockof913da52016-07-28 15:44:37 -07001168 oom_score_adj = task->signal->oom_score_adj;
David Rientjesa63d83f2010-08-09 17:19:46 -07001169 put_task_struct(task);
David Rientjesa9c58b902012-12-11 16:02:54 -08001170 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
David Rientjesa63d83f2010-08-09 17:19:46 -07001171 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1172}
1173
1174static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1175 size_t count, loff_t *ppos)
1176{
David Rientjesa63d83f2010-08-09 17:19:46 -07001177 char buffer[PROC_NUMBUF];
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001178 int oom_score_adj;
David Rientjesa63d83f2010-08-09 17:19:46 -07001179 int err;
1180
1181 memset(buffer, 0, sizeof(buffer));
1182 if (count > sizeof(buffer) - 1)
1183 count = sizeof(buffer) - 1;
David Rientjes723548b2010-10-26 14:21:25 -07001184 if (copy_from_user(buffer, buf, count)) {
1185 err = -EFAULT;
1186 goto out;
1187 }
David Rientjesa63d83f2010-08-09 17:19:46 -07001188
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001189 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
David Rientjesa63d83f2010-08-09 17:19:46 -07001190 if (err)
David Rientjes723548b2010-10-26 14:21:25 -07001191 goto out;
David Rientjesa63d83f2010-08-09 17:19:46 -07001192 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
David Rientjes723548b2010-10-26 14:21:25 -07001193 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1194 err = -EINVAL;
1195 goto out;
1196 }
David Rientjesa63d83f2010-08-09 17:19:46 -07001197
Michal Hocko1d5f0ac2016-07-28 15:44:40 -07001198 err = __set_oom_adj(file, oom_score_adj, false);
David Rientjes723548b2010-10-26 14:21:25 -07001199out:
1200 return err < 0 ? err : count;
David Rientjesa63d83f2010-08-09 17:19:46 -07001201}
1202
1203static const struct file_operations proc_oom_score_adj_operations = {
1204 .read = oom_score_adj_read,
1205 .write = oom_score_adj_write,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001206 .llseek = default_llseek,
David Rientjesa63d83f2010-08-09 17:19:46 -07001207};
1208
Richard Guy Briggs4b7d2482019-01-22 17:06:39 -05001209#ifdef CONFIG_AUDIT
Alexey Dobriyanb4eb4f72016-10-29 19:04:39 +03001210#define TMPBUFLEN 11
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1212 size_t count, loff_t *ppos)
1213{
Al Viro496ad9a2013-01-23 17:07:38 -05001214 struct inode * inode = file_inode(file);
Eric W. Biederman99f89552006-06-26 00:25:55 -07001215 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 ssize_t length;
1217 char tmpbuf[TMPBUFLEN];
1218
Eric W. Biederman99f89552006-06-26 00:25:55 -07001219 if (!task)
1220 return -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
Eric W. Biedermane1760bd2012-09-10 22:39:43 -07001222 from_kuid(file->f_cred->user_ns,
1223 audit_get_loginuid(task)));
Eric W. Biederman99f89552006-06-26 00:25:55 -07001224 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1226}
1227
1228static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1229 size_t count, loff_t *ppos)
1230{
Al Viro496ad9a2013-01-23 17:07:38 -05001231 struct inode * inode = file_inode(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 uid_t loginuid;
Eric W. Biedermane1760bd2012-09-10 22:39:43 -07001233 kuid_t kloginuid;
Alexey Dobriyan774636e2015-09-09 15:36:59 -07001234 int rv;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235
Paul E. McKenney7dc52152010-02-22 17:04:52 -08001236 rcu_read_lock();
1237 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1238 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 return -EPERM;
Paul E. McKenney7dc52152010-02-22 17:04:52 -08001240 }
1241 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243 if (*ppos != 0) {
1244 /* No partial writes. */
1245 return -EINVAL;
1246 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247
Alexey Dobriyan774636e2015-09-09 15:36:59 -07001248 rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1249 if (rv < 0)
1250 return rv;
Eric Paris81407c82013-05-24 09:49:14 -04001251
1252 /* is userspace tring to explicitly UNSET the loginuid? */
1253 if (loginuid == AUDIT_UID_UNSET) {
1254 kloginuid = INVALID_UID;
1255 } else {
1256 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
Alexey Dobriyan774636e2015-09-09 15:36:59 -07001257 if (!uid_valid(kloginuid))
1258 return -EINVAL;
Eric W. Biedermane1760bd2012-09-10 22:39:43 -07001259 }
1260
Alexey Dobriyan774636e2015-09-09 15:36:59 -07001261 rv = audit_set_loginuid(kloginuid);
1262 if (rv < 0)
1263 return rv;
1264 return count;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265}
1266
Arjan van de Ven00977a52007-02-12 00:55:34 -08001267static const struct file_operations proc_loginuid_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 .read = proc_loginuid_read,
1269 .write = proc_loginuid_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001270 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271};
Eric Paris1e0bd752008-03-13 08:15:31 -04001272
1273static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1274 size_t count, loff_t *ppos)
1275{
Al Viro496ad9a2013-01-23 17:07:38 -05001276 struct inode * inode = file_inode(file);
Eric Paris1e0bd752008-03-13 08:15:31 -04001277 struct task_struct *task = get_proc_task(inode);
1278 ssize_t length;
1279 char tmpbuf[TMPBUFLEN];
1280
1281 if (!task)
1282 return -ESRCH;
1283 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1284 audit_get_sessionid(task));
1285 put_task_struct(task);
1286 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1287}
1288
1289static const struct file_operations proc_sessionid_operations = {
1290 .read = proc_sessionid_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001291 .llseek = generic_file_llseek,
Eric Paris1e0bd752008-03-13 08:15:31 -04001292};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293#endif
1294
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001295#ifdef CONFIG_FAULT_INJECTION
1296static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1297 size_t count, loff_t *ppos)
1298{
Al Viro496ad9a2013-01-23 17:07:38 -05001299 struct task_struct *task = get_proc_task(file_inode(file));
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001300 char buffer[PROC_NUMBUF];
1301 size_t len;
1302 int make_it_fail;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001303
1304 if (!task)
1305 return -ESRCH;
1306 make_it_fail = task->make_it_fail;
1307 put_task_struct(task);
1308
1309 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001310
1311 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001312}
1313
1314static ssize_t proc_fault_inject_write(struct file * file,
1315 const char __user * buf, size_t count, loff_t *ppos)
1316{
1317 struct task_struct *task;
Alexey Dobriyan774636e2015-09-09 15:36:59 -07001318 char buffer[PROC_NUMBUF];
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001319 int make_it_fail;
Alexey Dobriyan774636e2015-09-09 15:36:59 -07001320 int rv;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001321
1322 if (!capable(CAP_SYS_RESOURCE))
1323 return -EPERM;
1324 memset(buffer, 0, sizeof(buffer));
1325 if (count > sizeof(buffer) - 1)
1326 count = sizeof(buffer) - 1;
1327 if (copy_from_user(buffer, buf, count))
1328 return -EFAULT;
Alexey Dobriyan774636e2015-09-09 15:36:59 -07001329 rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1330 if (rv < 0)
1331 return rv;
Dave Jones16caed32014-04-07 15:39:15 -07001332 if (make_it_fail < 0 || make_it_fail > 1)
1333 return -EINVAL;
1334
Al Viro496ad9a2013-01-23 17:07:38 -05001335 task = get_proc_task(file_inode(file));
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001336 if (!task)
1337 return -ESRCH;
1338 task->make_it_fail = make_it_fail;
1339 put_task_struct(task);
Vincent Licba8aaf2009-09-22 16:45:38 -07001340
1341 return count;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001342}
1343
Arjan van de Ven00977a52007-02-12 00:55:34 -08001344static const struct file_operations proc_fault_inject_operations = {
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001345 .read = proc_fault_inject_read,
1346 .write = proc_fault_inject_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001347 .llseek = generic_file_llseek,
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001348};
Dmitry Vyukove41d58182017-07-12 14:34:35 -07001349
1350static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1351 size_t count, loff_t *ppos)
1352{
1353 struct task_struct *task;
Akinobu Mita9049f2f2017-07-14 14:49:52 -07001354 int err;
1355 unsigned int n;
Dmitry Vyukove41d58182017-07-12 14:34:35 -07001356
Akinobu Mita9049f2f2017-07-14 14:49:52 -07001357 err = kstrtouint_from_user(buf, count, 0, &n);
Dmitry Vyukove41d58182017-07-12 14:34:35 -07001358 if (err)
1359 return err;
Akinobu Mita1203c8e2017-07-14 14:49:57 -07001360
1361 task = get_proc_task(file_inode(file));
1362 if (!task)
1363 return -ESRCH;
Alexey Dobriyan9f7118b2018-02-06 15:36:55 -08001364 task->fail_nth = n;
Akinobu Mita1203c8e2017-07-14 14:49:57 -07001365 put_task_struct(task);
1366
Dmitry Vyukove41d58182017-07-12 14:34:35 -07001367 return count;
1368}
1369
1370static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1371 size_t count, loff_t *ppos)
1372{
1373 struct task_struct *task;
Akinobu Mitabfc74092017-07-14 14:49:54 -07001374 char numbuf[PROC_NUMBUF];
1375 ssize_t len;
Dmitry Vyukove41d58182017-07-12 14:34:35 -07001376
1377 task = get_proc_task(file_inode(file));
1378 if (!task)
1379 return -ESRCH;
Alexey Dobriyan9f7118b2018-02-06 15:36:55 -08001380 len = snprintf(numbuf, sizeof(numbuf), "%u\n", task->fail_nth);
Akinobu Mita1203c8e2017-07-14 14:49:57 -07001381 put_task_struct(task);
Alexey Dobriyana44937f2018-08-21 21:54:27 -07001382 return simple_read_from_buffer(buf, count, ppos, numbuf, len);
Dmitry Vyukove41d58182017-07-12 14:34:35 -07001383}
1384
1385static const struct file_operations proc_fail_nth_operations = {
1386 .read = proc_fail_nth_read,
1387 .write = proc_fail_nth_write,
1388};
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001389#endif
1390
Arjan van de Ven97455122008-01-25 21:08:34 +01001391
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001392#ifdef CONFIG_SCHED_DEBUG
1393/*
1394 * Print out various scheduling related per-task fields:
1395 */
1396static int sched_show(struct seq_file *m, void *v)
1397{
1398 struct inode *inode = m->private;
Christoph Hellwig76f668b2018-05-16 07:19:01 +02001399 struct pid_namespace *ns = proc_pid_ns(inode);
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001400 struct task_struct *p;
1401
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001402 p = get_proc_task(inode);
1403 if (!p)
1404 return -ESRCH;
Aleksa Sarai74dc3382017-08-06 14:41:41 +10001405 proc_sched_show_task(p, ns, m);
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001406
1407 put_task_struct(p);
1408
1409 return 0;
1410}
1411
1412static ssize_t
1413sched_write(struct file *file, const char __user *buf,
1414 size_t count, loff_t *offset)
1415{
Al Viro496ad9a2013-01-23 17:07:38 -05001416 struct inode *inode = file_inode(file);
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001417 struct task_struct *p;
1418
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001419 p = get_proc_task(inode);
1420 if (!p)
1421 return -ESRCH;
1422 proc_sched_set_task(p);
1423
1424 put_task_struct(p);
1425
1426 return count;
1427}
1428
1429static int sched_open(struct inode *inode, struct file *filp)
1430{
Jovi Zhangc6a34052011-01-12 17:00:34 -08001431 return single_open(filp, sched_show, inode);
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001432}
1433
1434static const struct file_operations proc_pid_sched_operations = {
1435 .open = sched_open,
1436 .read = seq_read,
1437 .write = sched_write,
1438 .llseek = seq_lseek,
Alexey Dobriyan5ea473a2007-07-31 00:38:50 -07001439 .release = single_release,
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001440};
1441
1442#endif
1443
Mike Galbraith5091faa2010-11-30 14:18:03 +01001444#ifdef CONFIG_SCHED_AUTOGROUP
1445/*
1446 * Print out autogroup related information:
1447 */
1448static int sched_autogroup_show(struct seq_file *m, void *v)
1449{
1450 struct inode *inode = m->private;
1451 struct task_struct *p;
1452
1453 p = get_proc_task(inode);
1454 if (!p)
1455 return -ESRCH;
1456 proc_sched_autogroup_show_task(p, m);
1457
1458 put_task_struct(p);
1459
1460 return 0;
1461}
1462
1463static ssize_t
1464sched_autogroup_write(struct file *file, const char __user *buf,
1465 size_t count, loff_t *offset)
1466{
Al Viro496ad9a2013-01-23 17:07:38 -05001467 struct inode *inode = file_inode(file);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001468 struct task_struct *p;
1469 char buffer[PROC_NUMBUF];
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001470 int nice;
Mike Galbraith5091faa2010-11-30 14:18:03 +01001471 int err;
1472
1473 memset(buffer, 0, sizeof(buffer));
1474 if (count > sizeof(buffer) - 1)
1475 count = sizeof(buffer) - 1;
1476 if (copy_from_user(buffer, buf, count))
1477 return -EFAULT;
1478
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001479 err = kstrtoint(strstrip(buffer), 0, &nice);
1480 if (err < 0)
1481 return err;
Mike Galbraith5091faa2010-11-30 14:18:03 +01001482
1483 p = get_proc_task(inode);
1484 if (!p)
1485 return -ESRCH;
1486
Hiroshi Shimamoto2e5b5b32012-02-23 17:41:27 +09001487 err = proc_sched_autogroup_set_nice(p, nice);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001488 if (err)
1489 count = err;
1490
1491 put_task_struct(p);
1492
1493 return count;
1494}
1495
1496static int sched_autogroup_open(struct inode *inode, struct file *filp)
1497{
1498 int ret;
1499
1500 ret = single_open(filp, sched_autogroup_show, NULL);
1501 if (!ret) {
1502 struct seq_file *m = filp->private_data;
1503
1504 m->private = inode;
1505 }
1506 return ret;
1507}
1508
1509static const struct file_operations proc_pid_sched_autogroup_operations = {
1510 .open = sched_autogroup_open,
1511 .read = seq_read,
1512 .write = sched_autogroup_write,
1513 .llseek = seq_lseek,
1514 .release = single_release,
1515};
1516
1517#endif /* CONFIG_SCHED_AUTOGROUP */
1518
john stultz4614a696b2009-12-14 18:00:05 -08001519static ssize_t comm_write(struct file *file, const char __user *buf,
1520 size_t count, loff_t *offset)
1521{
Al Viro496ad9a2013-01-23 17:07:38 -05001522 struct inode *inode = file_inode(file);
john stultz4614a696b2009-12-14 18:00:05 -08001523 struct task_struct *p;
1524 char buffer[TASK_COMM_LEN];
David Rientjes830e0fc2013-04-30 15:28:18 -07001525 const size_t maxlen = sizeof(buffer) - 1;
john stultz4614a696b2009-12-14 18:00:05 -08001526
1527 memset(buffer, 0, sizeof(buffer));
David Rientjes830e0fc2013-04-30 15:28:18 -07001528 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
john stultz4614a696b2009-12-14 18:00:05 -08001529 return -EFAULT;
1530
1531 p = get_proc_task(inode);
1532 if (!p)
1533 return -ESRCH;
1534
1535 if (same_thread_group(current, p))
1536 set_task_comm(p, buffer);
1537 else
1538 count = -EINVAL;
1539
1540 put_task_struct(p);
1541
1542 return count;
1543}
1544
1545static int comm_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
Tejun Heo88b72b32018-05-18 08:47:13 -07001554 proc_task_name(m, p, false);
1555 seq_putc(m, '\n');
john stultz4614a696b2009-12-14 18:00:05 -08001556
1557 put_task_struct(p);
1558
1559 return 0;
1560}
1561
1562static int comm_open(struct inode *inode, struct file *filp)
1563{
Jovi Zhangc6a34052011-01-12 17:00:34 -08001564 return single_open(filp, comm_show, inode);
john stultz4614a696b2009-12-14 18:00:05 -08001565}
1566
1567static const struct file_operations proc_pid_set_comm_operations = {
1568 .open = comm_open,
1569 .read = seq_read,
1570 .write = comm_write,
1571 .llseek = seq_lseek,
1572 .release = single_release,
1573};
1574
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -08001575static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
Matt Helsley925d1c42008-04-29 01:01:36 -07001576{
1577 struct task_struct *task;
Matt Helsley925d1c42008-04-29 01:01:36 -07001578 struct file *exe_file;
1579
David Howells2b0143b2015-03-17 22:25:59 +00001580 task = get_proc_task(d_inode(dentry));
Matt Helsley925d1c42008-04-29 01:01:36 -07001581 if (!task)
1582 return -ENOENT;
Mateusz Guzikcd81a9172016-08-23 16:20:38 +02001583 exe_file = get_task_exe_file(task);
Matt Helsley925d1c42008-04-29 01:01:36 -07001584 put_task_struct(task);
Matt Helsley925d1c42008-04-29 01:01:36 -07001585 if (exe_file) {
1586 *exe_path = exe_file->f_path;
1587 path_get(&exe_file->f_path);
1588 fput(exe_file);
1589 return 0;
1590 } else
1591 return -ENOENT;
1592}
1593
Al Viro6b255392015-11-17 10:20:54 -05001594static const char *proc_pid_get_link(struct dentry *dentry,
Al Virofceef392015-12-29 15:58:39 -05001595 struct inode *inode,
1596 struct delayed_call *done)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597{
Christoph Hellwig408ef012012-06-18 10:47:03 -04001598 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599 int error = -EACCES;
1600
Al Viro6b255392015-11-17 10:20:54 -05001601 if (!dentry)
1602 return ERR_PTR(-ECHILD);
1603
Eric W. Biederman778c1142006-06-26 00:25:58 -07001604 /* Are we allowed to snoop on the tasks file descriptors? */
1605 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607
Christoph Hellwig408ef012012-06-18 10:47:03 -04001608 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1609 if (error)
1610 goto out;
1611
Al Viro6e77137b2015-05-02 13:37:52 -04001612 nd_jump_link(&path);
Christoph Hellwig408ef012012-06-18 10:47:03 -04001613 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614out:
Al Viro008b1502005-08-20 00:17:39 +01001615 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616}
1617
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001618static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619{
Michal Hocko0ee931c2017-09-13 16:28:29 -07001620 char *tmp = (char *)__get_free_page(GFP_KERNEL);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001621 char *pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622 int len;
1623
1624 if (!tmp)
1625 return -ENOMEM;
Akinobu Mita0c28f282007-05-08 00:31:41 -07001626
Eric W. Biederman7b2a69b2010-12-05 15:51:21 -08001627 pathname = d_path(path, tmp, PAGE_SIZE);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001628 len = PTR_ERR(pathname);
1629 if (IS_ERR(pathname))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 goto out;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001631 len = tmp + PAGE_SIZE - 1 - pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632
1633 if (len > buflen)
1634 len = buflen;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001635 if (copy_to_user(buffer, pathname, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 len = -EFAULT;
1637 out:
1638 free_page((unsigned long)tmp);
1639 return len;
1640}
1641
1642static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1643{
1644 int error = -EACCES;
David Howells2b0143b2015-03-17 22:25:59 +00001645 struct inode *inode = d_inode(dentry);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001646 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647
Eric W. Biederman778c1142006-06-26 00:25:58 -07001648 /* Are we allowed to snoop on the tasks file descriptors? */
1649 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -08001652 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 if (error)
1654 goto out;
1655
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001656 error = do_proc_readlink(&path, buffer, buflen);
1657 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 return error;
1660}
1661
Cyrill Gorcunovfaf60af2012-08-23 14:43:24 +04001662const struct inode_operations proc_pid_link_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 .readlink = proc_pid_readlink,
Al Viro6b255392015-11-17 10:20:54 -05001664 .get_link = proc_pid_get_link,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001665 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666};
1667
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001668
1669/* building an inode */
1670
Al Viroc6eb50d2017-09-30 14:45:42 -04001671void task_dump_owner(struct task_struct *task, umode_t mode,
Eric W. Biederman68eb94f2017-01-03 10:23:11 +13001672 kuid_t *ruid, kgid_t *rgid)
1673{
1674 /* Depending on the state of dumpable compute who should own a
1675 * proc file for a task.
1676 */
1677 const struct cred *cred;
1678 kuid_t uid;
1679 kgid_t gid;
1680
Alexey Dobriyan2e0ad552018-04-20 14:56:03 -07001681 if (unlikely(task->flags & PF_KTHREAD)) {
1682 *ruid = GLOBAL_ROOT_UID;
1683 *rgid = GLOBAL_ROOT_GID;
1684 return;
1685 }
1686
Eric W. Biederman68eb94f2017-01-03 10:23:11 +13001687 /* Default to the tasks effective ownership */
1688 rcu_read_lock();
1689 cred = __task_cred(task);
1690 uid = cred->euid;
1691 gid = cred->egid;
1692 rcu_read_unlock();
1693
1694 /*
1695 * Before the /proc/pid/status file was created the only way to read
1696 * the effective uid of a /process was to stat /proc/pid. Reading
1697 * /proc/pid/status is slow enough that procps and other packages
1698 * kept stating /proc/pid. To keep the rules in /proc simple I have
1699 * made this apply to all per process world readable and executable
1700 * directories.
1701 */
1702 if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1703 struct mm_struct *mm;
1704 task_lock(task);
1705 mm = task->mm;
1706 /* Make non-dumpable tasks owned by some root */
1707 if (mm) {
1708 if (get_dumpable(mm) != SUID_DUMP_USER) {
1709 struct user_namespace *user_ns = mm->user_ns;
1710
1711 uid = make_kuid(user_ns, 0);
1712 if (!uid_valid(uid))
1713 uid = GLOBAL_ROOT_UID;
1714
1715 gid = make_kgid(user_ns, 0);
1716 if (!gid_valid(gid))
1717 gid = GLOBAL_ROOT_GID;
1718 }
1719 } else {
1720 uid = GLOBAL_ROOT_UID;
1721 gid = GLOBAL_ROOT_GID;
1722 }
1723 task_unlock(task);
1724 }
1725 *ruid = uid;
1726 *rgid = gid;
1727}
1728
Andreas Gruenbacherdb978da2016-11-10 22:18:28 +01001729struct inode *proc_pid_make_inode(struct super_block * sb,
1730 struct task_struct *task, umode_t mode)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001731{
1732 struct inode * inode;
1733 struct proc_inode *ei;
1734
1735 /* We need a new inode */
1736
1737 inode = new_inode(sb);
1738 if (!inode)
1739 goto out;
1740
1741 /* Common stuff */
1742 ei = PROC_I(inode);
Andreas Gruenbacherdb978da2016-11-10 22:18:28 +01001743 inode->i_mode = mode;
Christoph Hellwig85fe4022010-10-23 11:19:54 -04001744 inode->i_ino = get_next_ino();
Deepa Dinamani078cd822016-09-14 07:48:04 -07001745 inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001746 inode->i_op = &proc_def_inode_operations;
1747
1748 /*
1749 * grab the reference to task.
1750 */
Oleg Nesterov1a657f782006-10-02 02:18:59 -07001751 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001752 if (!ei->pid)
1753 goto out_unlock;
1754
Eric W. Biederman68eb94f2017-01-03 10:23:11 +13001755 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001756 security_task_to_inode(task, inode);
1757
1758out:
1759 return inode;
1760
1761out_unlock:
1762 iput(inode);
1763 return NULL;
1764}
1765
David Howellsa528d352017-01-31 16:46:22 +00001766int pid_getattr(const struct path *path, struct kstat *stat,
1767 u32 request_mask, unsigned int query_flags)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001768{
David Howellsa528d352017-01-31 16:46:22 +00001769 struct inode *inode = d_inode(path->dentry);
Christoph Hellwig76f668b2018-05-16 07:19:01 +02001770 struct pid_namespace *pid = proc_pid_ns(inode);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001771 struct task_struct *task;
David Howellsc69e8d92008-11-14 10:39:19 +11001772
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001773 generic_fillattr(inode, stat);
1774
Eric W. Biedermandcb0f222012-02-09 08:48:21 -08001775 stat->uid = GLOBAL_ROOT_UID;
1776 stat->gid = GLOBAL_ROOT_GID;
Alexey Dobriyan94116922018-06-07 17:10:07 -07001777 rcu_read_lock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001778 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1779 if (task) {
Lafcadio Wluiki796f5712017-02-24 15:00:23 -08001780 if (!has_pid_permissions(pid, task, HIDEPID_INVISIBLE)) {
Vasiliy Kulikov04996802012-01-10 15:11:31 -08001781 rcu_read_unlock();
1782 /*
1783 * This doesn't prevent learning whether PID exists,
1784 * it only makes getattr() consistent with readdir().
1785 */
1786 return -ENOENT;
1787 }
Eric W. Biederman68eb94f2017-01-03 10:23:11 +13001788 task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001789 }
1790 rcu_read_unlock();
1791 return 0;
1792}
1793
1794/* dentry stuff */
1795
1796/*
Al Viro1bbc5512018-05-02 21:26:16 -04001797 * Set <pid>/... inode ownership (can change due to setuid(), etc.)
1798 */
1799void pid_update_inode(struct task_struct *task, struct inode *inode)
1800{
1801 task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
1802
1803 inode->i_mode &= ~(S_ISUID | S_ISGID);
1804 security_task_to_inode(task, inode);
1805}
1806
1807/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001808 * Rewrite the inode's ownerships here because the owning task may have
1809 * performed a setuid(), etc.
1810 *
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001811 */
Al Viro1bbc5512018-05-02 21:26:16 -04001812static int pid_revalidate(struct dentry *dentry, unsigned int flags)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001813{
Nick Piggin34286d62011-01-07 17:49:57 +11001814 struct inode *inode;
1815 struct task_struct *task;
David Howellsc69e8d92008-11-14 10:39:19 +11001816
Al Viro0b728e12012-06-10 16:03:43 -04001817 if (flags & LOOKUP_RCU)
Nick Piggin34286d62011-01-07 17:49:57 +11001818 return -ECHILD;
1819
David Howells2b0143b2015-03-17 22:25:59 +00001820 inode = d_inode(dentry);
Nick Piggin34286d62011-01-07 17:49:57 +11001821 task = get_proc_task(inode);
1822
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001823 if (task) {
Al Viro1bbc5512018-05-02 21:26:16 -04001824 pid_update_inode(task, inode);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001825 put_task_struct(task);
1826 return 1;
1827 }
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001828 return 0;
1829}
1830
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08001831static inline bool proc_inode_is_dead(struct inode *inode)
1832{
1833 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1834}
1835
David Howells1dd704b2013-04-12 01:08:50 +01001836int pid_delete_dentry(const struct dentry *dentry)
1837{
1838 /* Is the task we represent dead?
1839 * If so, then don't put the dentry on the lru list,
1840 * kill it immediately.
1841 */
David Howells2b0143b2015-03-17 22:25:59 +00001842 return proc_inode_is_dead(d_inode(dentry));
David Howells1dd704b2013-04-12 01:08:50 +01001843}
1844
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001845const struct dentry_operations pid_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001846{
1847 .d_revalidate = pid_revalidate,
1848 .d_delete = pid_delete_dentry,
1849};
1850
1851/* Lookups */
1852
Eric W. Biederman1c0d04c2006-10-02 02:18:57 -07001853/*
1854 * Fill a directory entry.
1855 *
1856 * If possible create the dcache entry and derive our inode number and
1857 * file type from dcache entry.
1858 *
1859 * Since all of the proc inode numbers are dynamically generated, the inode
1860 * numbers do not exist until the inode is cache. This means creating the
1861 * the dcache entry in readdir is necessary to keep the inode numbers
1862 * reported by readdir in sync with the inode numbers reported
1863 * by stat.
1864 */
Al Virof0c3b502013-05-16 12:07:31 -04001865bool proc_fill_cache(struct file *file, struct dir_context *ctx,
Alexey Dobriyana4ef3892018-06-07 17:10:10 -07001866 const char *name, unsigned int len,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001867 instantiate_t instantiate, struct task_struct *task, const void *ptr)
Eric W. Biederman61a28782006-10-02 02:18:49 -07001868{
Al Virof0c3b502013-05-16 12:07:31 -04001869 struct dentry *child, *dir = file->f_path.dentry;
Al Viro1df98b82013-06-15 11:33:10 +04001870 struct qstr qname = QSTR_INIT(name, len);
Eric W. Biederman61a28782006-10-02 02:18:49 -07001871 struct inode *inode;
Al Viro0168b9e2018-05-03 09:21:05 -04001872 unsigned type = DT_UNKNOWN;
1873 ino_t ino = 1;
Eric W. Biederman61a28782006-10-02 02:18:49 -07001874
Al Viro1df98b82013-06-15 11:33:10 +04001875 child = d_hash_and_lookup(dir, &qname);
Eric W. Biederman61a28782006-10-02 02:18:49 -07001876 if (!child) {
Al Viro37817642016-04-20 16:31:31 -04001877 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
1878 child = d_alloc_parallel(dir, &qname, &wq);
1879 if (IS_ERR(child))
Al Viro1df98b82013-06-15 11:33:10 +04001880 goto end_instantiate;
Al Viro37817642016-04-20 16:31:31 -04001881 if (d_in_lookup(child)) {
Al Viro0168b9e2018-05-03 09:21:05 -04001882 struct dentry *res;
1883 res = instantiate(child, task, ptr);
Al Viro37817642016-04-20 16:31:31 -04001884 d_lookup_done(child);
Al Viro0168b9e2018-05-03 09:21:05 -04001885 if (unlikely(res)) {
1886 dput(child);
1887 child = res;
Al Virod85b3992018-06-08 01:17:11 -04001888 if (IS_ERR(child))
1889 goto end_instantiate;
Al Viro37817642016-04-20 16:31:31 -04001890 }
Eric W. Biederman61a28782006-10-02 02:18:49 -07001891 }
1892 }
David Howells2b0143b2015-03-17 22:25:59 +00001893 inode = d_inode(child);
Al Viro147ce692013-06-15 10:26:35 +04001894 ino = inode->i_ino;
1895 type = inode->i_mode >> 12;
Eric W. Biederman61a28782006-10-02 02:18:49 -07001896 dput(child);
Al Virod85b3992018-06-08 01:17:11 -04001897end_instantiate:
Al Virof0c3b502013-05-16 12:07:31 -04001898 return dir_emit(ctx, name, len, ino, type);
Eric W. Biederman61a28782006-10-02 02:18:49 -07001899}
1900
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001901/*
1902 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1903 * which represent vma start and end addresses.
1904 */
1905static int dname_to_vma_addr(struct dentry *dentry,
1906 unsigned long *start, unsigned long *end)
1907{
Alexey Dobriyanac7f1062018-02-06 15:36:59 -08001908 const char *str = dentry->d_name.name;
1909 unsigned long long sval, eval;
1910 unsigned int len;
1911
Alexey Dobriyan35318db2018-04-10 16:41:14 -07001912 if (str[0] == '0' && str[1] != '-')
1913 return -EINVAL;
Alexey Dobriyanac7f1062018-02-06 15:36:59 -08001914 len = _parse_integer(str, 16, &sval);
1915 if (len & KSTRTOX_OVERFLOW)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001916 return -EINVAL;
Alexey Dobriyanac7f1062018-02-06 15:36:59 -08001917 if (sval != (unsigned long)sval)
1918 return -EINVAL;
1919 str += len;
1920
1921 if (*str != '-')
1922 return -EINVAL;
1923 str++;
1924
Alexey Dobriyan35318db2018-04-10 16:41:14 -07001925 if (str[0] == '0' && str[1])
1926 return -EINVAL;
Alexey Dobriyanac7f1062018-02-06 15:36:59 -08001927 len = _parse_integer(str, 16, &eval);
1928 if (len & KSTRTOX_OVERFLOW)
1929 return -EINVAL;
1930 if (eval != (unsigned long)eval)
1931 return -EINVAL;
1932 str += len;
1933
1934 if (*str != '\0')
1935 return -EINVAL;
1936
1937 *start = sval;
1938 *end = eval;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001939
1940 return 0;
1941}
1942
Al Viro0b728e12012-06-10 16:03:43 -04001943static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001944{
1945 unsigned long vm_start, vm_end;
1946 bool exact_vma_exists = false;
1947 struct mm_struct *mm = NULL;
1948 struct task_struct *task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001949 struct inode *inode;
1950 int status = 0;
1951
Al Viro0b728e12012-06-10 16:03:43 -04001952 if (flags & LOOKUP_RCU)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001953 return -ECHILD;
1954
David Howells2b0143b2015-03-17 22:25:59 +00001955 inode = d_inode(dentry);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001956 task = get_proc_task(inode);
1957 if (!task)
1958 goto out_notask;
1959
Jann Horncaaee622016-01-20 15:00:04 -08001960 mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
Cong Wang2344bec2012-05-31 16:26:18 -07001961 if (IS_ERR_OR_NULL(mm))
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001962 goto out;
1963
1964 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1965 down_read(&mm->mmap_sem);
1966 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1967 up_read(&mm->mmap_sem);
1968 }
1969
1970 mmput(mm);
1971
1972 if (exact_vma_exists) {
Eric W. Biederman68eb94f2017-01-03 10:23:11 +13001973 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1974
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001975 security_task_to_inode(task, inode);
1976 status = 1;
1977 }
1978
1979out:
1980 put_task_struct(task);
1981
1982out_notask:
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001983 return status;
1984}
1985
1986static const struct dentry_operations tid_map_files_dentry_operations = {
1987 .d_revalidate = map_files_d_revalidate,
1988 .d_delete = pid_delete_dentry,
1989};
1990
Al Viro6b255392015-11-17 10:20:54 -05001991static int map_files_get_link(struct dentry *dentry, struct path *path)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001992{
1993 unsigned long vm_start, vm_end;
1994 struct vm_area_struct *vma;
1995 struct task_struct *task;
1996 struct mm_struct *mm;
1997 int rc;
1998
1999 rc = -ENOENT;
David Howells2b0143b2015-03-17 22:25:59 +00002000 task = get_proc_task(d_inode(dentry));
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002001 if (!task)
2002 goto out;
2003
2004 mm = get_task_mm(task);
2005 put_task_struct(task);
2006 if (!mm)
2007 goto out;
2008
2009 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2010 if (rc)
2011 goto out_mmput;
2012
Artem Fetishev70335ab2014-03-10 15:49:45 -07002013 rc = -ENOENT;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002014 down_read(&mm->mmap_sem);
2015 vma = find_exact_vma(mm, vm_start, vm_end);
2016 if (vma && vma->vm_file) {
2017 *path = vma->vm_file->f_path;
2018 path_get(path);
2019 rc = 0;
2020 }
2021 up_read(&mm->mmap_sem);
2022
2023out_mmput:
2024 mmput(mm);
2025out:
2026 return rc;
2027}
2028
2029struct map_files_info {
Alexey Dobriyan20d28cd2018-02-06 15:37:06 -08002030 unsigned long start;
2031 unsigned long end;
Al Viro7b540d02012-08-27 14:55:26 -04002032 fmode_t mode;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002033};
2034
Calvin Owensbdb4d102015-09-09 15:35:54 -07002035/*
2036 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
2037 * symlinks may be used to bypass permissions on ancestor directories in the
2038 * path to the file in question.
2039 */
2040static const char *
Al Viro6b255392015-11-17 10:20:54 -05002041proc_map_files_get_link(struct dentry *dentry,
Al Virofceef392015-12-29 15:58:39 -05002042 struct inode *inode,
2043 struct delayed_call *done)
Calvin Owensbdb4d102015-09-09 15:35:54 -07002044{
2045 if (!capable(CAP_SYS_ADMIN))
2046 return ERR_PTR(-EPERM);
2047
Al Virofceef392015-12-29 15:58:39 -05002048 return proc_pid_get_link(dentry, inode, done);
Calvin Owensbdb4d102015-09-09 15:35:54 -07002049}
2050
2051/*
Al Viro6b255392015-11-17 10:20:54 -05002052 * Identical to proc_pid_link_inode_operations except for get_link()
Calvin Owensbdb4d102015-09-09 15:35:54 -07002053 */
2054static const struct inode_operations proc_map_files_link_inode_operations = {
2055 .readlink = proc_pid_readlink,
Al Viro6b255392015-11-17 10:20:54 -05002056 .get_link = proc_map_files_get_link,
Calvin Owensbdb4d102015-09-09 15:35:54 -07002057 .setattr = proc_setattr,
2058};
2059
Al Viro0168b9e2018-05-03 09:21:05 -04002060static struct dentry *
2061proc_map_files_instantiate(struct dentry *dentry,
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002062 struct task_struct *task, const void *ptr)
2063{
Al Viro7b540d02012-08-27 14:55:26 -04002064 fmode_t mode = (fmode_t)(unsigned long)ptr;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002065 struct proc_inode *ei;
2066 struct inode *inode;
2067
Al Viro0168b9e2018-05-03 09:21:05 -04002068 inode = proc_pid_make_inode(dentry->d_sb, task, S_IFLNK |
Andreas Gruenbacherdb978da2016-11-10 22:18:28 +01002069 ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2070 ((mode & FMODE_WRITE) ? S_IWUSR : 0));
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002071 if (!inode)
Al Viro0168b9e2018-05-03 09:21:05 -04002072 return ERR_PTR(-ENOENT);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002073
2074 ei = PROC_I(inode);
Al Viro6b255392015-11-17 10:20:54 -05002075 ei->op.proc_get_link = map_files_get_link;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002076
Calvin Owensbdb4d102015-09-09 15:35:54 -07002077 inode->i_op = &proc_map_files_link_inode_operations;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002078 inode->i_size = 64;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002079
2080 d_set_d_op(dentry, &tid_map_files_dentry_operations);
Al Viro0168b9e2018-05-03 09:21:05 -04002081 return d_splice_alias(inode, dentry);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002082}
2083
2084static struct dentry *proc_map_files_lookup(struct inode *dir,
Al Viro00cd8dd2012-06-10 17:13:09 -04002085 struct dentry *dentry, unsigned int flags)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002086{
2087 unsigned long vm_start, vm_end;
2088 struct vm_area_struct *vma;
2089 struct task_struct *task;
Al Viro0168b9e2018-05-03 09:21:05 -04002090 struct dentry *result;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002091 struct mm_struct *mm;
2092
Al Viro0168b9e2018-05-03 09:21:05 -04002093 result = ERR_PTR(-ENOENT);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002094 task = get_proc_task(dir);
2095 if (!task)
2096 goto out;
2097
Al Viro0168b9e2018-05-03 09:21:05 -04002098 result = ERR_PTR(-EACCES);
Jann Horncaaee622016-01-20 15:00:04 -08002099 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002100 goto out_put_task;
2101
Al Viro0168b9e2018-05-03 09:21:05 -04002102 result = ERR_PTR(-ENOENT);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002103 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07002104 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002105
2106 mm = get_task_mm(task);
2107 if (!mm)
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07002108 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002109
2110 down_read(&mm->mmap_sem);
2111 vma = find_exact_vma(mm, vm_start, vm_end);
2112 if (!vma)
2113 goto out_no_vma;
2114
Stanislav Kinsbursky05f56482012-11-26 16:29:42 -08002115 if (vma->vm_file)
Al Viro0168b9e2018-05-03 09:21:05 -04002116 result = proc_map_files_instantiate(dentry, task,
Stanislav Kinsbursky05f56482012-11-26 16:29:42 -08002117 (void *)(unsigned long)vma->vm_file->f_mode);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002118
2119out_no_vma:
2120 up_read(&mm->mmap_sem);
2121 mmput(mm);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002122out_put_task:
2123 put_task_struct(task);
2124out:
Al Viro0168b9e2018-05-03 09:21:05 -04002125 return result;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002126}
2127
2128static const struct inode_operations proc_map_files_inode_operations = {
2129 .lookup = proc_map_files_lookup,
2130 .permission = proc_fd_permission,
2131 .setattr = proc_setattr,
2132};
2133
2134static int
Al Virof0c3b502013-05-16 12:07:31 -04002135proc_map_files_readdir(struct file *file, struct dir_context *ctx)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002136{
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002137 struct vm_area_struct *vma;
2138 struct task_struct *task;
2139 struct mm_struct *mm;
Al Virof0c3b502013-05-16 12:07:31 -04002140 unsigned long nr_files, pos, i;
Kent Overstreet94f8f3b2019-03-11 23:31:18 -07002141 GENRADIX(struct map_files_info) fa;
Al Virof0c3b502013-05-16 12:07:31 -04002142 struct map_files_info *p;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002143 int ret;
2144
Kent Overstreet94f8f3b2019-03-11 23:31:18 -07002145 genradix_init(&fa);
2146
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002147 ret = -ENOENT;
Al Virof0c3b502013-05-16 12:07:31 -04002148 task = get_proc_task(file_inode(file));
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002149 if (!task)
2150 goto out;
2151
2152 ret = -EACCES;
Jann Horncaaee622016-01-20 15:00:04 -08002153 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002154 goto out_put_task;
2155
2156 ret = 0;
Al Virof0c3b502013-05-16 12:07:31 -04002157 if (!dir_emit_dots(file, ctx))
2158 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002159
Al Virof0c3b502013-05-16 12:07:31 -04002160 mm = get_task_mm(task);
2161 if (!mm)
2162 goto out_put_task;
2163 down_read(&mm->mmap_sem);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002164
Al Virof0c3b502013-05-16 12:07:31 -04002165 nr_files = 0;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002166
Al Virof0c3b502013-05-16 12:07:31 -04002167 /*
2168 * We need two passes here:
2169 *
2170 * 1) Collect vmas of mapped files with mmap_sem taken
2171 * 2) Release mmap_sem and instantiate entries
2172 *
2173 * otherwise we get lockdep complained, since filldir()
2174 * routine might require mmap_sem taken in might_fault().
2175 */
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002176
Al Virof0c3b502013-05-16 12:07:31 -04002177 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
Kent Overstreet94f8f3b2019-03-11 23:31:18 -07002178 if (!vma->vm_file)
2179 continue;
2180 if (++pos <= ctx->pos)
2181 continue;
Al Virof0c3b502013-05-16 12:07:31 -04002182
Kent Overstreet94f8f3b2019-03-11 23:31:18 -07002183 p = genradix_ptr_alloc(&fa, nr_files++, GFP_KERNEL);
2184 if (!p) {
Al Virof0c3b502013-05-16 12:07:31 -04002185 ret = -ENOMEM;
Al Virof0c3b502013-05-16 12:07:31 -04002186 up_read(&mm->mmap_sem);
2187 mmput(mm);
2188 goto out_put_task;
2189 }
Al Virof0c3b502013-05-16 12:07:31 -04002190
Kent Overstreet94f8f3b2019-03-11 23:31:18 -07002191 p->start = vma->vm_start;
2192 p->end = vma->vm_end;
2193 p->mode = vma->vm_file->f_mode;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002194 }
Al Virof0c3b502013-05-16 12:07:31 -04002195 up_read(&mm->mmap_sem);
Alexey Dobriyanfe079a52018-04-10 16:32:05 -07002196 mmput(mm);
Al Virof0c3b502013-05-16 12:07:31 -04002197
2198 for (i = 0; i < nr_files; i++) {
Alexey Dobriyan20d28cd2018-02-06 15:37:06 -08002199 char buf[4 * sizeof(long) + 2]; /* max: %lx-%lx\0 */
2200 unsigned int len;
2201
Kent Overstreet94f8f3b2019-03-11 23:31:18 -07002202 p = genradix_ptr(&fa, i);
Alexey Dobriyan20d28cd2018-02-06 15:37:06 -08002203 len = snprintf(buf, sizeof(buf), "%lx-%lx", p->start, p->end);
Al Virof0c3b502013-05-16 12:07:31 -04002204 if (!proc_fill_cache(file, ctx,
Alexey Dobriyan20d28cd2018-02-06 15:37:06 -08002205 buf, len,
Al Virof0c3b502013-05-16 12:07:31 -04002206 proc_map_files_instantiate,
2207 task,
2208 (void *)(unsigned long)p->mode))
2209 break;
2210 ctx->pos++;
2211 }
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002212
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002213out_put_task:
2214 put_task_struct(task);
2215out:
Kent Overstreet94f8f3b2019-03-11 23:31:18 -07002216 genradix_free(&fa);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002217 return ret;
2218}
2219
2220static const struct file_operations proc_map_files_operations = {
2221 .read = generic_read_dir,
Al Virof50752e2016-04-20 17:13:54 -04002222 .iterate_shared = proc_map_files_readdir,
2223 .llseek = generic_file_llseek,
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002224};
2225
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05002226#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002227struct timers_private {
2228 struct pid *pid;
2229 struct task_struct *task;
2230 struct sighand_struct *sighand;
Pavel Emelyanov57b80152013-03-11 13:13:08 +04002231 struct pid_namespace *ns;
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002232 unsigned long flags;
2233};
2234
2235static void *timers_start(struct seq_file *m, loff_t *pos)
2236{
2237 struct timers_private *tp = m->private;
2238
2239 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2240 if (!tp->task)
2241 return ERR_PTR(-ESRCH);
2242
2243 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2244 if (!tp->sighand)
2245 return ERR_PTR(-ESRCH);
2246
2247 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2248}
2249
2250static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2251{
2252 struct timers_private *tp = m->private;
2253 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2254}
2255
2256static void timers_stop(struct seq_file *m, void *v)
2257{
2258 struct timers_private *tp = m->private;
2259
2260 if (tp->sighand) {
2261 unlock_task_sighand(tp->task, &tp->flags);
2262 tp->sighand = NULL;
2263 }
2264
2265 if (tp->task) {
2266 put_task_struct(tp->task);
2267 tp->task = NULL;
2268 }
2269}
2270
2271static int show_timer(struct seq_file *m, void *v)
2272{
2273 struct k_itimer *timer;
Pavel Emelyanov57b80152013-03-11 13:13:08 +04002274 struct timers_private *tp = m->private;
2275 int notify;
Alexey Dobriyancedbcca2014-08-08 14:21:33 -07002276 static const char * const nstr[] = {
Pavel Emelyanov57b80152013-03-11 13:13:08 +04002277 [SIGEV_SIGNAL] = "signal",
2278 [SIGEV_NONE] = "none",
2279 [SIGEV_THREAD] = "thread",
2280 };
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002281
2282 timer = list_entry((struct list_head *)v, struct k_itimer, list);
Pavel Emelyanov57b80152013-03-11 13:13:08 +04002283 notify = timer->it_sigev_notify;
2284
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002285 seq_printf(m, "ID: %d\n", timer->it_id);
Linus Torvaldsba3edf1f2017-12-06 18:23:27 -08002286 seq_printf(m, "signal: %d/%px\n",
Joe Perches25ce3192015-04-15 16:18:17 -07002287 timer->sigq->info.si_signo,
2288 timer->sigq->info.si_value.sival_ptr);
Pavel Emelyanov57b80152013-03-11 13:13:08 +04002289 seq_printf(m, "notify: %s/%s.%d\n",
Joe Perches25ce3192015-04-15 16:18:17 -07002290 nstr[notify & ~SIGEV_THREAD_ID],
2291 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2292 pid_nr_ns(timer->it_pid, tp->ns));
Pavel Tikhomirov15ef0292013-05-17 02:12:03 +04002293 seq_printf(m, "ClockID: %d\n", timer->it_clock);
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002294
2295 return 0;
2296}
2297
2298static const struct seq_operations proc_timers_seq_ops = {
2299 .start = timers_start,
2300 .next = timers_next,
2301 .stop = timers_stop,
2302 .show = show_timer,
2303};
2304
2305static int proc_timers_open(struct inode *inode, struct file *file)
2306{
2307 struct timers_private *tp;
2308
2309 tp = __seq_open_private(file, &proc_timers_seq_ops,
2310 sizeof(struct timers_private));
2311 if (!tp)
2312 return -ENOMEM;
2313
2314 tp->pid = proc_pid(inode);
Christoph Hellwig76f668b2018-05-16 07:19:01 +02002315 tp->ns = proc_pid_ns(inode);
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002316 return 0;
2317}
2318
2319static const struct file_operations proc_timers_operations = {
2320 .open = proc_timers_open,
2321 .read = seq_read,
2322 .llseek = seq_lseek,
2323 .release = seq_release_private,
2324};
Eric Engestromb5946be2016-03-17 14:20:57 -07002325#endif
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002326
John Stultz5de23d42016-03-17 14:20:54 -07002327static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2328 size_t count, loff_t *offset)
2329{
2330 struct inode *inode = file_inode(file);
2331 struct task_struct *p;
2332 u64 slack_ns;
2333 int err;
2334
2335 err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2336 if (err < 0)
2337 return err;
2338
2339 p = get_proc_task(inode);
2340 if (!p)
2341 return -ESRCH;
2342
John Stultz4b2bd5f2016-10-07 17:02:33 -07002343 if (p != current) {
Benjamin Gordon8da0b4f2019-01-03 15:25:56 -08002344 rcu_read_lock();
2345 if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2346 rcu_read_unlock();
John Stultz4b2bd5f2016-10-07 17:02:33 -07002347 count = -EPERM;
2348 goto out;
2349 }
Benjamin Gordon8da0b4f2019-01-03 15:25:56 -08002350 rcu_read_unlock();
John Stultz5de23d42016-03-17 14:20:54 -07002351
John Stultz4b2bd5f2016-10-07 17:02:33 -07002352 err = security_task_setscheduler(p);
2353 if (err) {
2354 count = err;
2355 goto out;
2356 }
John Stultz904763e2016-10-07 17:02:29 -07002357 }
2358
John Stultz7abbaf92016-10-07 17:02:26 -07002359 task_lock(p);
2360 if (slack_ns == 0)
2361 p->timer_slack_ns = p->default_timer_slack_ns;
2362 else
2363 p->timer_slack_ns = slack_ns;
2364 task_unlock(p);
2365
2366out:
John Stultz5de23d42016-03-17 14:20:54 -07002367 put_task_struct(p);
2368
2369 return count;
2370}
2371
2372static int timerslack_ns_show(struct seq_file *m, void *v)
2373{
2374 struct inode *inode = m->private;
2375 struct task_struct *p;
John Stultz7abbaf92016-10-07 17:02:26 -07002376 int err = 0;
John Stultz5de23d42016-03-17 14:20:54 -07002377
2378 p = get_proc_task(inode);
2379 if (!p)
2380 return -ESRCH;
2381
John Stultz4b2bd5f2016-10-07 17:02:33 -07002382 if (p != current) {
Benjamin Gordon8da0b4f2019-01-03 15:25:56 -08002383 rcu_read_lock();
2384 if (!ns_capable(__task_cred(p)->user_ns, CAP_SYS_NICE)) {
2385 rcu_read_unlock();
John Stultz4b2bd5f2016-10-07 17:02:33 -07002386 err = -EPERM;
2387 goto out;
2388 }
Benjamin Gordon8da0b4f2019-01-03 15:25:56 -08002389 rcu_read_unlock();
2390
John Stultz4b2bd5f2016-10-07 17:02:33 -07002391 err = security_task_getscheduler(p);
2392 if (err)
2393 goto out;
2394 }
John Stultz904763e2016-10-07 17:02:29 -07002395
John Stultz7abbaf92016-10-07 17:02:26 -07002396 task_lock(p);
2397 seq_printf(m, "%llu\n", p->timer_slack_ns);
2398 task_unlock(p);
2399
2400out:
John Stultz5de23d42016-03-17 14:20:54 -07002401 put_task_struct(p);
2402
2403 return err;
2404}
2405
2406static int timerslack_ns_open(struct inode *inode, struct file *filp)
2407{
2408 return single_open(filp, timerslack_ns_show, inode);
2409}
2410
2411static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2412 .open = timerslack_ns_open,
2413 .read = seq_read,
2414 .write = timerslack_ns_write,
2415 .llseek = seq_lseek,
2416 .release = single_release,
2417};
2418
Al Viro0168b9e2018-05-03 09:21:05 -04002419static struct dentry *proc_pident_instantiate(struct dentry *dentry,
2420 struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002421{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002422 const struct pid_entry *p = ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002423 struct inode *inode;
2424 struct proc_inode *ei;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002425
Al Viro0168b9e2018-05-03 09:21:05 -04002426 inode = proc_pid_make_inode(dentry->d_sb, task, p->mode);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002427 if (!inode)
Al Viro0168b9e2018-05-03 09:21:05 -04002428 return ERR_PTR(-ENOENT);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002429
2430 ei = PROC_I(inode);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002431 if (S_ISDIR(inode->i_mode))
Miklos Szeredibfe86842011-10-28 14:13:29 +02002432 set_nlink(inode, 2); /* Use getattr to fix if necessary */
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002433 if (p->iop)
2434 inode->i_op = p->iop;
2435 if (p->fop)
2436 inode->i_fop = p->fop;
2437 ei->op = p->op;
Al Viro1bbc5512018-05-02 21:26:16 -04002438 pid_update_inode(task, inode);
Nick Pigginfb045ad2011-01-07 17:49:55 +11002439 d_set_d_op(dentry, &pid_dentry_operations);
Al Viro0168b9e2018-05-03 09:21:05 -04002440 return d_splice_alias(inode, dentry);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002441}
2442
Linus Torvalds1da177e2005-04-16 15:20:36 -07002443static struct dentry *proc_pident_lookup(struct inode *dir,
2444 struct dentry *dentry,
Alexey Dobriyand5a572a2019-03-11 23:28:51 -07002445 const struct pid_entry *p,
2446 const struct pid_entry *end)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002447{
Eric W. Biederman99f89552006-06-26 00:25:55 -07002448 struct task_struct *task = get_proc_task(dir);
Al Viro0168b9e2018-05-03 09:21:05 -04002449 struct dentry *res = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450
Eric W. Biederman99f89552006-06-26 00:25:55 -07002451 if (!task)
2452 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002453
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002454 /*
2455 * Yes, it does not scale. And it should not. Don't add
2456 * new entries into /proc/<tgid>/ without very good reasons.
2457 */
Alexey Dobriyand5a572a2019-03-11 23:28:51 -07002458 for (; p < end; p++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002459 if (p->len != dentry->d_name.len)
2460 continue;
Alexey Dobriyan26b95132018-06-14 15:27:17 -07002461 if (!memcmp(dentry->d_name.name, p->name, p->len)) {
2462 res = proc_pident_instantiate(dentry, task, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002463 break;
Alexey Dobriyan26b95132018-06-14 15:27:17 -07002464 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002465 }
Eric W. Biederman99f89552006-06-26 00:25:55 -07002466 put_task_struct(task);
2467out_no_task:
Al Viro0168b9e2018-05-03 09:21:05 -04002468 return res;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002469}
2470
Al Virof0c3b502013-05-16 12:07:31 -04002471static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002472 const struct pid_entry *ents, unsigned int nents)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002473{
Al Virof0c3b502013-05-16 12:07:31 -04002474 struct task_struct *task = get_proc_task(file_inode(file));
2475 const struct pid_entry *p;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002476
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002477 if (!task)
Al Virof0c3b502013-05-16 12:07:31 -04002478 return -ENOENT;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002479
Al Virof0c3b502013-05-16 12:07:31 -04002480 if (!dir_emit_dots(file, ctx))
2481 goto out;
2482
2483 if (ctx->pos >= nents + 2)
2484 goto out;
2485
Alexey Dobriyanbac5f5d2016-12-12 16:45:28 -08002486 for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
Al Virof0c3b502013-05-16 12:07:31 -04002487 if (!proc_fill_cache(file, ctx, p->name, p->len,
2488 proc_pident_instantiate, task, p))
2489 break;
2490 ctx->pos++;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002491 }
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002492out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07002493 put_task_struct(task);
Al Virof0c3b502013-05-16 12:07:31 -04002494 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002495}
2496
Linus Torvalds1da177e2005-04-16 15:20:36 -07002497#ifdef CONFIG_SECURITY
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002498static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2499 size_t count, loff_t *ppos)
2500{
Al Viro496ad9a2013-01-23 17:07:38 -05002501 struct inode * inode = file_inode(file);
Al Viro04ff9702007-03-12 16:17:58 +00002502 char *p = NULL;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002503 ssize_t length;
2504 struct task_struct *task = get_proc_task(inode);
2505
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002506 if (!task)
Al Viro04ff9702007-03-12 16:17:58 +00002507 return -ESRCH;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002508
Casey Schaufler6d9c9392018-09-21 17:16:59 -07002509 length = security_getprocattr(task, PROC_I(inode)->op.lsm,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002510 (char*)file->f_path.dentry->d_name.name,
Al Viro04ff9702007-03-12 16:17:58 +00002511 &p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002512 put_task_struct(task);
Al Viro04ff9702007-03-12 16:17:58 +00002513 if (length > 0)
2514 length = simple_read_from_buffer(buf, count, ppos, p, length);
2515 kfree(p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002516 return length;
2517}
2518
2519static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2520 size_t count, loff_t *ppos)
2521{
Al Viro496ad9a2013-01-23 17:07:38 -05002522 struct inode * inode = file_inode(file);
Alexey Dobriyan41089b62018-08-21 21:54:30 -07002523 struct task_struct *task;
Al Virobb646cd2015-12-24 00:16:30 -05002524 void *page;
Alexey Dobriyan41089b62018-08-21 21:54:30 -07002525 int rv;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002526
Alexey Dobriyan41089b62018-08-21 21:54:30 -07002527 rcu_read_lock();
2528 task = pid_task(proc_pid(inode), PIDTYPE_PID);
2529 if (!task) {
2530 rcu_read_unlock();
2531 return -ESRCH;
2532 }
Stephen Smalleyb21507e2017-01-09 10:07:31 -05002533 /* A task may only write its own attributes. */
Alexey Dobriyan41089b62018-08-21 21:54:30 -07002534 if (current != task) {
2535 rcu_read_unlock();
2536 return -EACCES;
2537 }
Paul Moore35a196b2019-04-19 14:55:12 -04002538 /* Prevent changes to overridden credentials. */
2539 if (current_cred() != current_real_cred()) {
2540 rcu_read_unlock();
2541 return -EBUSY;
2542 }
Alexey Dobriyan41089b62018-08-21 21:54:30 -07002543 rcu_read_unlock();
Stephen Smalleyb21507e2017-01-09 10:07:31 -05002544
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002545 if (count > PAGE_SIZE)
2546 count = PAGE_SIZE;
2547
2548 /* No partial writes. */
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002549 if (*ppos != 0)
Alexey Dobriyan41089b62018-08-21 21:54:30 -07002550 return -EINVAL;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002551
Al Virobb646cd2015-12-24 00:16:30 -05002552 page = memdup_user(buf, count);
2553 if (IS_ERR(page)) {
Alexey Dobriyan41089b62018-08-21 21:54:30 -07002554 rv = PTR_ERR(page);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002555 goto out;
Al Virobb646cd2015-12-24 00:16:30 -05002556 }
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002557
David Howells107db7c2009-05-08 13:55:27 +01002558 /* Guard against adverse ptrace interaction */
Alexey Dobriyan41089b62018-08-21 21:54:30 -07002559 rv = mutex_lock_interruptible(&current->signal->cred_guard_mutex);
2560 if (rv < 0)
David Howells107db7c2009-05-08 13:55:27 +01002561 goto out_free;
2562
Casey Schaufler6d9c9392018-09-21 17:16:59 -07002563 rv = security_setprocattr(PROC_I(inode)->op.lsm,
2564 file->f_path.dentry->d_name.name, page,
2565 count);
Stephen Smalleyb21507e2017-01-09 10:07:31 -05002566 mutex_unlock(&current->signal->cred_guard_mutex);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002567out_free:
Al Virobb646cd2015-12-24 00:16:30 -05002568 kfree(page);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002569out:
Alexey Dobriyan41089b62018-08-21 21:54:30 -07002570 return rv;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002571}
2572
Arjan van de Ven00977a52007-02-12 00:55:34 -08002573static const struct file_operations proc_pid_attr_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002574 .read = proc_pid_attr_read,
2575 .write = proc_pid_attr_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01002576 .llseek = generic_file_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002577};
2578
Casey Schaufler6d9c9392018-09-21 17:16:59 -07002579#define LSM_DIR_OPS(LSM) \
2580static int proc_##LSM##_attr_dir_iterate(struct file *filp, \
2581 struct dir_context *ctx) \
2582{ \
2583 return proc_pident_readdir(filp, ctx, \
2584 LSM##_attr_dir_stuff, \
2585 ARRAY_SIZE(LSM##_attr_dir_stuff)); \
2586} \
2587\
2588static const struct file_operations proc_##LSM##_attr_dir_ops = { \
2589 .read = generic_read_dir, \
2590 .iterate = proc_##LSM##_attr_dir_iterate, \
2591 .llseek = default_llseek, \
2592}; \
2593\
2594static struct dentry *proc_##LSM##_attr_dir_lookup(struct inode *dir, \
2595 struct dentry *dentry, unsigned int flags) \
2596{ \
2597 return proc_pident_lookup(dir, dentry, \
2598 LSM##_attr_dir_stuff, \
Alexey Dobriyand5a572a2019-03-11 23:28:51 -07002599 LSM##_attr_dir_stuff + ARRAY_SIZE(LSM##_attr_dir_stuff)); \
Casey Schaufler6d9c9392018-09-21 17:16:59 -07002600} \
2601\
2602static const struct inode_operations proc_##LSM##_attr_dir_inode_ops = { \
2603 .lookup = proc_##LSM##_attr_dir_lookup, \
2604 .getattr = pid_getattr, \
2605 .setattr = proc_setattr, \
2606}
2607
2608#ifdef CONFIG_SECURITY_SMACK
2609static const struct pid_entry smack_attr_dir_stuff[] = {
2610 ATTR("smack", "current", 0666),
2611};
2612LSM_DIR_OPS(smack);
2613#endif
2614
Eric Dumazetc5141e62007-05-08 00:26:15 -07002615static const struct pid_entry attr_dir_stuff[] = {
Casey Schaufler6d9c9392018-09-21 17:16:59 -07002616 ATTR(NULL, "current", 0666),
2617 ATTR(NULL, "prev", 0444),
2618 ATTR(NULL, "exec", 0666),
2619 ATTR(NULL, "fscreate", 0666),
2620 ATTR(NULL, "keycreate", 0666),
2621 ATTR(NULL, "sockcreate", 0666),
2622#ifdef CONFIG_SECURITY_SMACK
2623 DIR("smack", 0555,
2624 proc_smack_attr_dir_inode_ops, proc_smack_attr_dir_ops),
2625#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002626};
2627
Al Virof0c3b502013-05-16 12:07:31 -04002628static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002629{
Al Virof0c3b502013-05-16 12:07:31 -04002630 return proc_pident_readdir(file, ctx,
2631 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002632}
2633
Arjan van de Ven00977a52007-02-12 00:55:34 -08002634static const struct file_operations proc_attr_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002635 .read = generic_read_dir,
Al Virof50752e2016-04-20 17:13:54 -04002636 .iterate_shared = proc_attr_dir_readdir,
2637 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002638};
2639
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002640static struct dentry *proc_attr_dir_lookup(struct inode *dir,
Al Viro00cd8dd2012-06-10 17:13:09 -04002641 struct dentry *dentry, unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002642{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002643 return proc_pident_lookup(dir, dentry,
Alexey Dobriyand5a572a2019-03-11 23:28:51 -07002644 attr_dir_stuff,
2645 attr_dir_stuff + ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002646}
2647
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002648static const struct inode_operations proc_attr_dir_inode_operations = {
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002649 .lookup = proc_attr_dir_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07002650 .getattr = pid_getattr,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07002651 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002652};
2653
Linus Torvalds1da177e2005-04-16 15:20:36 -07002654#endif
2655
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08002656#ifdef CONFIG_ELF_CORE
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002657static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2658 size_t count, loff_t *ppos)
2659{
Al Viro496ad9a2013-01-23 17:07:38 -05002660 struct task_struct *task = get_proc_task(file_inode(file));
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002661 struct mm_struct *mm;
2662 char buffer[PROC_NUMBUF];
2663 size_t len;
2664 int ret;
2665
2666 if (!task)
2667 return -ESRCH;
2668
2669 ret = 0;
2670 mm = get_task_mm(task);
2671 if (mm) {
2672 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2673 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2674 MMF_DUMP_FILTER_SHIFT));
2675 mmput(mm);
2676 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2677 }
2678
2679 put_task_struct(task);
2680
2681 return ret;
2682}
2683
2684static ssize_t proc_coredump_filter_write(struct file *file,
2685 const char __user *buf,
2686 size_t count,
2687 loff_t *ppos)
2688{
2689 struct task_struct *task;
2690 struct mm_struct *mm;
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002691 unsigned int val;
2692 int ret;
2693 int i;
2694 unsigned long mask;
2695
Alexey Dobriyan774636e2015-09-09 15:36:59 -07002696 ret = kstrtouint_from_user(buf, count, 0, &val);
2697 if (ret < 0)
2698 return ret;
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002699
2700 ret = -ESRCH;
Al Viro496ad9a2013-01-23 17:07:38 -05002701 task = get_proc_task(file_inode(file));
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002702 if (!task)
2703 goto out_no_task;
2704
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002705 mm = get_task_mm(task);
2706 if (!mm)
2707 goto out_no_mm;
Colin Ian King41a0c2492015-12-18 14:22:01 -08002708 ret = 0;
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002709
2710 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2711 if (val & mask)
2712 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2713 else
2714 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2715 }
2716
2717 mmput(mm);
2718 out_no_mm:
2719 put_task_struct(task);
2720 out_no_task:
Alexey Dobriyan774636e2015-09-09 15:36:59 -07002721 if (ret < 0)
2722 return ret;
2723 return count;
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002724}
2725
2726static const struct file_operations proc_coredump_filter_operations = {
2727 .read = proc_coredump_filter_read,
2728 .write = proc_coredump_filter_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01002729 .llseek = generic_file_llseek,
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002730};
2731#endif
2732
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002733#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002734static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002735{
Andrea Righi940389b2008-07-28 00:48:12 +02002736 struct task_io_accounting acct = task->ioac;
Andrea Righi59954772008-07-27 17:29:15 +02002737 unsigned long flags;
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002738 int result;
Andrea Righi297c5d92008-07-25 01:48:49 -07002739
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002740 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2741 if (result)
2742 return result;
2743
Jann Horncaaee622016-01-20 15:00:04 -08002744 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002745 result = -EACCES;
2746 goto out_unlock;
2747 }
Vasiliy Kulikov1d1221f2011-06-24 16:08:38 +04002748
Andrea Righi59954772008-07-27 17:29:15 +02002749 if (whole && lock_task_sighand(task, &flags)) {
2750 struct task_struct *t = task;
Andrea Righib2d002d2008-07-26 15:22:27 -07002751
Andrea Righi59954772008-07-27 17:29:15 +02002752 task_io_accounting_add(&acct, &task->signal->ioac);
2753 while_each_thread(task, t)
2754 task_io_accounting_add(&acct, &t->ioac);
Andrea Righi297c5d92008-07-25 01:48:49 -07002755
Andrea Righi59954772008-07-27 17:29:15 +02002756 unlock_task_sighand(task, &flags);
Andrea Righi297c5d92008-07-25 01:48:49 -07002757 }
Joe Perches25ce3192015-04-15 16:18:17 -07002758 seq_printf(m,
2759 "rchar: %llu\n"
2760 "wchar: %llu\n"
2761 "syscr: %llu\n"
2762 "syscw: %llu\n"
2763 "read_bytes: %llu\n"
2764 "write_bytes: %llu\n"
2765 "cancelled_write_bytes: %llu\n",
2766 (unsigned long long)acct.rchar,
2767 (unsigned long long)acct.wchar,
2768 (unsigned long long)acct.syscr,
2769 (unsigned long long)acct.syscw,
2770 (unsigned long long)acct.read_bytes,
2771 (unsigned long long)acct.write_bytes,
2772 (unsigned long long)acct.cancelled_write_bytes);
2773 result = 0;
2774
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002775out_unlock:
2776 mutex_unlock(&task->signal->cred_guard_mutex);
2777 return result;
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002778}
Andrea Righi297c5d92008-07-25 01:48:49 -07002779
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002780static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2781 struct pid *pid, struct task_struct *task)
Andrea Righi297c5d92008-07-25 01:48:49 -07002782{
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002783 return do_io_accounting(task, m, 0);
Andrea Righi297c5d92008-07-25 01:48:49 -07002784}
2785
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002786static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2787 struct pid *pid, struct task_struct *task)
Andrea Righi297c5d92008-07-25 01:48:49 -07002788{
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002789 return do_io_accounting(task, m, 1);
Andrea Righi297c5d92008-07-25 01:48:49 -07002790}
2791#endif /* CONFIG_TASK_IO_ACCOUNTING */
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002792
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002793#ifdef CONFIG_USER_NS
2794static int proc_id_map_open(struct inode *inode, struct file *file,
Fabian Frederickccf94f12014-08-08 14:21:22 -07002795 const struct seq_operations *seq_ops)
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002796{
2797 struct user_namespace *ns = NULL;
2798 struct task_struct *task;
2799 struct seq_file *seq;
2800 int ret = -EINVAL;
2801
2802 task = get_proc_task(inode);
2803 if (task) {
2804 rcu_read_lock();
2805 ns = get_user_ns(task_cred_xxx(task, user_ns));
2806 rcu_read_unlock();
2807 put_task_struct(task);
2808 }
2809 if (!ns)
2810 goto err;
2811
2812 ret = seq_open(file, seq_ops);
2813 if (ret)
2814 goto err_put_ns;
2815
2816 seq = file->private_data;
2817 seq->private = ns;
2818
2819 return 0;
2820err_put_ns:
2821 put_user_ns(ns);
2822err:
2823 return ret;
2824}
2825
2826static int proc_id_map_release(struct inode *inode, struct file *file)
2827{
2828 struct seq_file *seq = file->private_data;
2829 struct user_namespace *ns = seq->private;
2830 put_user_ns(ns);
2831 return seq_release(inode, file);
2832}
2833
2834static int proc_uid_map_open(struct inode *inode, struct file *file)
2835{
2836 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2837}
2838
2839static int proc_gid_map_open(struct inode *inode, struct file *file)
2840{
2841 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2842}
2843
Eric W. Biedermanf76d2072012-08-30 01:24:05 -07002844static int proc_projid_map_open(struct inode *inode, struct file *file)
2845{
2846 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2847}
2848
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002849static const struct file_operations proc_uid_map_operations = {
2850 .open = proc_uid_map_open,
2851 .write = proc_uid_map_write,
2852 .read = seq_read,
2853 .llseek = seq_lseek,
2854 .release = proc_id_map_release,
2855};
2856
2857static const struct file_operations proc_gid_map_operations = {
2858 .open = proc_gid_map_open,
2859 .write = proc_gid_map_write,
2860 .read = seq_read,
2861 .llseek = seq_lseek,
2862 .release = proc_id_map_release,
2863};
Eric W. Biedermanf76d2072012-08-30 01:24:05 -07002864
2865static const struct file_operations proc_projid_map_operations = {
2866 .open = proc_projid_map_open,
2867 .write = proc_projid_map_write,
2868 .read = seq_read,
2869 .llseek = seq_lseek,
2870 .release = proc_id_map_release,
2871};
Eric W. Biederman9cc46512014-12-02 12:27:26 -06002872
2873static int proc_setgroups_open(struct inode *inode, struct file *file)
2874{
2875 struct user_namespace *ns = NULL;
2876 struct task_struct *task;
2877 int ret;
2878
2879 ret = -ESRCH;
2880 task = get_proc_task(inode);
2881 if (task) {
2882 rcu_read_lock();
2883 ns = get_user_ns(task_cred_xxx(task, user_ns));
2884 rcu_read_unlock();
2885 put_task_struct(task);
2886 }
2887 if (!ns)
2888 goto err;
2889
2890 if (file->f_mode & FMODE_WRITE) {
2891 ret = -EACCES;
2892 if (!ns_capable(ns, CAP_SYS_ADMIN))
2893 goto err_put_ns;
2894 }
2895
2896 ret = single_open(file, &proc_setgroups_show, ns);
2897 if (ret)
2898 goto err_put_ns;
2899
2900 return 0;
2901err_put_ns:
2902 put_user_ns(ns);
2903err:
2904 return ret;
2905}
2906
2907static int proc_setgroups_release(struct inode *inode, struct file *file)
2908{
2909 struct seq_file *seq = file->private_data;
2910 struct user_namespace *ns = seq->private;
2911 int ret = single_release(inode, file);
2912 put_user_ns(ns);
2913 return ret;
2914}
2915
2916static const struct file_operations proc_setgroups_operations = {
2917 .open = proc_setgroups_open,
2918 .write = proc_setgroups_write,
2919 .read = seq_read,
2920 .llseek = seq_lseek,
2921 .release = proc_setgroups_release,
2922};
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002923#endif /* CONFIG_USER_NS */
2924
Kees Cook47830722008-10-06 03:11:58 +04002925static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2926 struct pid *pid, struct task_struct *task)
2927{
Al Viroa9712bc2011-03-23 15:52:50 -04002928 int err = lock_trace(task);
2929 if (!err) {
2930 seq_printf(m, "%08x\n", task->personality);
2931 unlock_trace(task);
2932 }
2933 return err;
Kees Cook47830722008-10-06 03:11:58 +04002934}
2935
Josh Poimboeuf7c23b332017-02-13 19:42:41 -06002936#ifdef CONFIG_LIVEPATCH
2937static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
2938 struct pid *pid, struct task_struct *task)
2939{
2940 seq_printf(m, "%d\n", task->patch_state);
2941 return 0;
2942}
2943#endif /* CONFIG_LIVEPATCH */
2944
Alexander Popovc8d12622018-08-17 01:17:01 +03002945#ifdef CONFIG_STACKLEAK_METRICS
2946static int proc_stack_depth(struct seq_file *m, struct pid_namespace *ns,
2947 struct pid *pid, struct task_struct *task)
2948{
2949 unsigned long prev_depth = THREAD_SIZE -
2950 (task->prev_lowest_stack & (THREAD_SIZE - 1));
2951 unsigned long depth = THREAD_SIZE -
2952 (task->lowest_stack & (THREAD_SIZE - 1));
2953
2954 seq_printf(m, "previous stack depth: %lu\nstack depth: %lu\n",
2955 prev_depth, depth);
2956 return 0;
2957}
2958#endif /* CONFIG_STACKLEAK_METRICS */
2959
Eric W. Biederman801199c2006-10-02 02:18:48 -07002960/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002961 * Thread groups
2962 */
Arjan van de Ven00977a52007-02-12 00:55:34 -08002963static const struct file_operations proc_task_operations;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002964static const struct inode_operations proc_task_inode_operations;
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002965
Eric Dumazetc5141e62007-05-08 00:26:15 -07002966static const struct pid_entry tgid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002967 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2968 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002969 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002970 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08002971 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002972#ifdef CONFIG_NET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002973 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002974#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002975 REG("environ", S_IRUSR, proc_environ_operations),
Al Viroc5317162016-10-05 18:43:43 -04002976 REG("auxv", S_IRUSR, proc_auxv_operations),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002977 ONE("status", S_IRUGO, proc_pid_status),
Djalal Harouni35a35042014-04-07 15:38:36 -07002978 ONE("personality", S_IRUSR, proc_pid_personality),
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -07002979 ONE("limits", S_IRUGO, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002980#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002981 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002982#endif
Mike Galbraith5091faa2010-11-30 14:18:03 +01002983#ifdef CONFIG_SCHED_AUTOGROUP
2984 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2985#endif
john stultz4614a696b2009-12-14 18:00:05 -08002986 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
Roland McGrathebcb6732008-07-25 19:46:00 -07002987#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -07002988 ONE("syscall", S_IRUSR, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07002989#endif
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -07002990 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002991 ONE("stat", S_IRUGO, proc_tgid_stat),
2992 ONE("statm", S_IRUGO, proc_pid_statm),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07002993 REG("maps", S_IRUGO, proc_pid_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002994#ifdef CONFIG_NUMA
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07002995 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002996#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002997 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2998 LNK("cwd", proc_cwd_link),
2999 LNK("root", proc_root_link),
3000 LNK("exe", proc_exe_link),
3001 REG("mounts", S_IRUGO, proc_mounts_operations),
3002 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3003 REG("mountstats", S_IRUSR, proc_mountstats_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08003004#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003005 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07003006 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
Daniel Colascione493b0e92017-09-06 16:25:08 -07003007 REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
Djalal Harouni32ed74a2014-04-07 15:38:38 -07003008 REG("pagemap", S_IRUSR, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003009#endif
3010#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003011 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003012#endif
3013#ifdef CONFIG_KALLSYMS
Alexey Dobriyanedfcd602014-08-08 14:21:44 -07003014 ONE("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003015#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03003016#ifdef CONFIG_STACKTRACE
Djalal Harouni35a35042014-04-07 15:38:36 -07003017 ONE("stack", S_IRUSR, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003018#endif
Naveen N. Rao5968cec2015-06-30 14:36:03 +05303019#ifdef CONFIG_SCHED_INFO
Alexey Dobriyanf6e826c2014-08-08 14:21:46 -07003020 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003021#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01003022#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003023 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01003024#endif
Paul Menage8793d852007-10-18 23:39:39 -07003025#ifdef CONFIG_PROC_PID_CPUSET
Zefan Li52de4772014-09-18 16:03:36 +08003026 ONE("cpuset", S_IRUGO, proc_cpuset_show),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003027#endif
Paul Menagea4243162007-10-18 23:39:35 -07003028#ifdef CONFIG_CGROUPS
Zefan Li006f4ac2014-09-18 16:03:15 +08003029 ONE("cgroup", S_IRUGO, proc_cgroup_show),
Paul Menagea4243162007-10-18 23:39:35 -07003030#endif
Alexey Dobriyan6ba51e32014-08-08 14:21:48 -07003031 ONE("oom_score", S_IRUGO, proc_oom_score),
David Rientjesfa0cbbf2012-11-12 17:53:04 -08003032 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
David Rientjesa63d83f2010-08-09 17:19:46 -07003033 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
Richard Guy Briggs4b7d2482019-01-22 17:06:39 -05003034#ifdef CONFIG_AUDIT
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003035 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3036 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003037#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003038#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003039 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mita168c42b2017-07-14 14:50:00 -07003040 REG("fail-nth", 0644, proc_fail_nth_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003041#endif
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08003042#ifdef CONFIG_ELF_CORE
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003043 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07003044#endif
Andrew Mortonaba76fd2006-12-10 02:19:48 -08003045#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07003046 ONE("io", S_IRUSR, proc_tgid_io_accounting),
Andrew Mortonaba76fd2006-12-10 02:19:48 -08003047#endif
Eric W. Biederman22d917d2011-11-17 00:11:58 -08003048#ifdef CONFIG_USER_NS
3049 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3050 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
Eric W. Biedermanf76d2072012-08-30 01:24:05 -07003051 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
Eric W. Biederman9cc46512014-12-02 12:27:26 -06003052 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
Eric W. Biederman22d917d2011-11-17 00:11:58 -08003053#endif
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05003054#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04003055 REG("timers", S_IRUGO, proc_timers_operations),
3056#endif
John Stultz5de23d42016-03-17 14:20:54 -07003057 REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
Josh Poimboeuf7c23b332017-02-13 19:42:41 -06003058#ifdef CONFIG_LIVEPATCH
3059 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3060#endif
Alexander Popovc8d12622018-08-17 01:17:01 +03003061#ifdef CONFIG_STACKLEAK_METRICS
3062 ONE("stack_depth", S_IRUGO, proc_stack_depth),
3063#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003064};
3065
Al Virof0c3b502013-05-16 12:07:31 -04003066static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003067{
Al Virof0c3b502013-05-16 12:07:31 -04003068 return proc_pident_readdir(file, ctx,
3069 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003070}
3071
Arjan van de Ven00977a52007-02-12 00:55:34 -08003072static const struct file_operations proc_tgid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003073 .read = generic_read_dir,
Al Virof50752e2016-04-20 17:13:54 -04003074 .iterate_shared = proc_tgid_base_readdir,
3075 .llseek = generic_file_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003076};
3077
Christian Brauner3eb39f42018-11-19 00:51:56 +01003078struct pid *tgid_pidfd_to_pid(const struct file *file)
3079{
Christian Brauner30d158b2019-06-27 11:35:14 +02003080 if (file->f_op != &proc_tgid_base_operations)
Christian Brauner3eb39f42018-11-19 00:51:56 +01003081 return ERR_PTR(-EBADF);
3082
3083 return proc_pid(file_inode(file));
3084}
3085
Al Viro00cd8dd2012-06-10 17:13:09 -04003086static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3087{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003088 return proc_pident_lookup(dir, dentry,
Alexey Dobriyand5a572a2019-03-11 23:28:51 -07003089 tgid_base_stuff,
3090 tgid_base_stuff + ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003091}
3092
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003093static const struct inode_operations proc_tgid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003094 .lookup = proc_tgid_base_lookup,
3095 .getattr = pid_getattr,
3096 .setattr = proc_setattr,
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003097 .permission = proc_pid_permission,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003098};
3099
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003100static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003101{
Eric W. Biederman48e64842006-06-26 00:25:48 -07003102 struct dentry *dentry, *leader, *dir;
Alexey Dobriyane3912ac2018-02-06 15:36:51 -08003103 char buf[10 + 1];
Eric W. Biederman48e64842006-06-26 00:25:48 -07003104 struct qstr name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003105
Eric W. Biederman48e64842006-06-26 00:25:48 -07003106 name.name = buf;
Alexey Dobriyane3912ac2018-02-06 15:36:51 -08003107 name.len = snprintf(buf, sizeof(buf), "%u", pid);
Al Viro4f522a22013-02-11 23:20:37 -05003108 /* no ->d_hash() rejects on procfs */
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003109 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003110 if (dentry) {
Eric W. Biedermanbbd51922014-02-13 10:24:23 -08003111 d_invalidate(dentry);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003112 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003113 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003114
Oleg Nesterovc35a7f12014-12-10 15:54:56 -08003115 if (pid == tgid)
3116 return;
3117
Eric W. Biederman48e64842006-06-26 00:25:48 -07003118 name.name = buf;
Alexey Dobriyane3912ac2018-02-06 15:36:51 -08003119 name.len = snprintf(buf, sizeof(buf), "%u", tgid);
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003120 leader = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003121 if (!leader)
3122 goto out;
3123
3124 name.name = "task";
3125 name.len = strlen(name.name);
3126 dir = d_hash_and_lookup(leader, &name);
3127 if (!dir)
3128 goto out_put_leader;
3129
3130 name.name = buf;
Alexey Dobriyane3912ac2018-02-06 15:36:51 -08003131 name.len = snprintf(buf, sizeof(buf), "%u", pid);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003132 dentry = d_hash_and_lookup(dir, &name);
3133 if (dentry) {
Eric W. Biedermanbbd51922014-02-13 10:24:23 -08003134 d_invalidate(dentry);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003135 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003136 }
Eric W. Biederman48e64842006-06-26 00:25:48 -07003137
3138 dput(dir);
3139out_put_leader:
3140 dput(leader);
3141out:
3142 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003143}
3144
Randy Dunlap0895e912007-10-21 21:00:10 -07003145/**
3146 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3147 * @task: task that should be flushed.
3148 *
3149 * When flushing dentries from proc, one needs to flush them from global
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003150 * proc (proc_mnt) and from all the namespaces' procs this task was seen
Randy Dunlap0895e912007-10-21 21:00:10 -07003151 * in. This call is supposed to do all of this job.
3152 *
3153 * Looks in the dcache for
3154 * /proc/@pid
3155 * /proc/@tgid/task/@pid
3156 * if either directory is present flushes it and all of it'ts children
3157 * from the dcache.
3158 *
3159 * It is safe and reasonable to cache /proc entries for a task until
3160 * that task exits. After that they just clog up the dcache with
3161 * useless entries, possibly causing useful dcache entries to be
3162 * flushed instead. This routine is proved to flush those useless
3163 * dcache entries at process exit time.
3164 *
3165 * NOTE: This routine is just an optimization so it does not guarantee
3166 * that no dcache entries will exist at process exit time it
3167 * just makes it very unlikely that any will persist.
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003168 */
3169
3170void proc_flush_task(struct task_struct *task)
3171{
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08003172 int i;
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07003173 struct pid *pid, *tgid;
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003174 struct upid *upid;
3175
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003176 pid = task_pid(task);
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07003177 tgid = task_tgid(task);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003178
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08003179 for (i = 0; i <= pid->level; i++) {
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003180 upid = &pid->numbers[i];
3181 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07003182 tgid->numbers[i].nr);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003183 }
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003184}
3185
Al Viro0168b9e2018-05-03 09:21:05 -04003186static struct dentry *proc_pid_instantiate(struct dentry * dentry,
Al Viroc52a47a2013-06-15 11:15:20 +04003187 struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003188{
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003189 struct inode *inode;
3190
Al Viro0168b9e2018-05-03 09:21:05 -04003191 inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003192 if (!inode)
Al Viro0168b9e2018-05-03 09:21:05 -04003193 return ERR_PTR(-ENOENT);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003194
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003195 inode->i_op = &proc_tgid_base_inode_operations;
3196 inode->i_fop = &proc_tgid_base_operations;
3197 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07003198
Alexey Dobriyan1270dd82016-12-12 16:45:32 -08003199 set_nlink(inode, nlink_tgid);
Al Viro1bbc5512018-05-02 21:26:16 -04003200 pid_update_inode(task, inode);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003201
Nick Pigginfb045ad2011-01-07 17:49:55 +11003202 d_set_d_op(dentry, &pid_dentry_operations);
Al Viro0168b9e2018-05-03 09:21:05 -04003203 return d_splice_alias(inode, dentry);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003204}
3205
Zhikang Zhang867aacc2019-03-05 15:50:29 -08003206struct dentry *proc_pid_lookup(struct dentry *dentry, unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003207{
3208 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003209 unsigned tgid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003210 struct pid_namespace *ns;
Al Viro0168b9e2018-05-03 09:21:05 -04003211 struct dentry *result = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003212
Alexey Dobriyandbcdb502014-08-08 14:21:25 -07003213 tgid = name_to_int(&dentry->d_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003214 if (tgid == ~0U)
3215 goto out;
3216
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003217 ns = dentry->d_sb->s_fs_info;
Eric W. Biedermande758732006-06-26 00:25:51 -07003218 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003219 task = find_task_by_pid_ns(tgid, ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003220 if (task)
3221 get_task_struct(task);
Eric W. Biedermande758732006-06-26 00:25:51 -07003222 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003223 if (!task)
3224 goto out;
3225
Al Viro0168b9e2018-05-03 09:21:05 -04003226 result = proc_pid_instantiate(dentry, task, NULL);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003227 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003228out:
Al Viro0168b9e2018-05-03 09:21:05 -04003229 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003230}
3231
Linus Torvalds1da177e2005-04-16 15:20:36 -07003232/*
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003233 * Find the first task with tgid >= tgid
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003234 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07003235 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003236struct tgid_iter {
3237 unsigned int tgid;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003238 struct task_struct *task;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003239};
3240static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3241{
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003242 struct pid *pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003243
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003244 if (iter.task)
3245 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003246 rcu_read_lock();
3247retry:
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003248 iter.task = NULL;
3249 pid = find_ge_pid(iter.tgid, ns);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003250 if (pid) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003251 iter.tgid = pid_nr_ns(pid, ns);
3252 iter.task = pid_task(pid, PIDTYPE_PID);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003253 /* What we to know is if the pid we have find is the
3254 * pid of a thread_group_leader. Testing for task
3255 * being a thread_group_leader is the obvious thing
3256 * todo but there is a window when it fails, due to
3257 * the pid transfer logic in de_thread.
3258 *
3259 * So we perform the straight forward test of seeing
3260 * if the pid we have found is the pid of a thread
3261 * group leader, and don't worry if the task we have
3262 * found doesn't happen to be a thread group leader.
3263 * As we don't care in the case of readdir.
3264 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003265 if (!iter.task || !has_group_leader_pid(iter.task)) {
3266 iter.tgid += 1;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003267 goto retry;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003268 }
3269 get_task_struct(iter.task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003270 }
Eric W. Biederman454cc102006-06-26 00:25:51 -07003271 rcu_read_unlock();
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003272 return iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003273}
3274
Eric W. Biederman00978752014-07-31 03:10:50 -07003275#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003276
3277/* for the /proc/ directory itself, after non-process stuff has been done */
Al Virof0c3b502013-05-16 12:07:31 -04003278int proc_pid_readdir(struct file *file, struct dir_context *ctx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003279{
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003280 struct tgid_iter iter;
Christoph Hellwig76f668b2018-05-16 07:19:01 +02003281 struct pid_namespace *ns = proc_pid_ns(file_inode(file));
Al Virof0c3b502013-05-16 12:07:31 -04003282 loff_t pos = ctx->pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003283
Al Viro021ada72013-03-29 19:27:05 -04003284 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
Al Virof0c3b502013-05-16 12:07:31 -04003285 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003286
Eric W. Biederman00978752014-07-31 03:10:50 -07003287 if (pos == TGID_OFFSET - 2) {
David Howells2b0143b2015-03-17 22:25:59 +00003288 struct inode *inode = d_inode(ns->proc_self);
Al Virodb963162013-06-15 10:45:10 +04003289 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
Al Virof0c3b502013-05-16 12:07:31 -04003290 return 0;
Eric W. Biederman00978752014-07-31 03:10:50 -07003291 ctx->pos = pos = pos + 1;
Al Viro021ada72013-03-29 19:27:05 -04003292 }
Eric W. Biederman00978752014-07-31 03:10:50 -07003293 if (pos == TGID_OFFSET - 1) {
David Howells2b0143b2015-03-17 22:25:59 +00003294 struct inode *inode = d_inode(ns->proc_thread_self);
Eric W. Biederman00978752014-07-31 03:10:50 -07003295 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3296 return 0;
3297 ctx->pos = pos = pos + 1;
3298 }
3299 iter.tgid = pos - TGID_OFFSET;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003300 iter.task = NULL;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003301 for (iter = next_tgid(ns, iter);
3302 iter.task;
3303 iter.tgid += 1, iter = next_tgid(ns, iter)) {
Alexey Dobriyane3912ac2018-02-06 15:36:51 -08003304 char name[10 + 1];
Alexey Dobriyana4ef3892018-06-07 17:10:10 -07003305 unsigned int len;
Eric Dumazet3ba4bce2017-01-24 15:18:07 -08003306
3307 cond_resched();
Lafcadio Wluiki796f5712017-02-24 15:00:23 -08003308 if (!has_pid_permissions(ns, iter.task, HIDEPID_INVISIBLE))
Al Virof0c3b502013-05-16 12:07:31 -04003309 continue;
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003310
Alexey Dobriyane3912ac2018-02-06 15:36:51 -08003311 len = snprintf(name, sizeof(name), "%u", iter.tgid);
Al Virof0c3b502013-05-16 12:07:31 -04003312 ctx->pos = iter.tgid + TGID_OFFSET;
3313 if (!proc_fill_cache(file, ctx, name, len,
3314 proc_pid_instantiate, iter.task, NULL)) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003315 put_task_struct(iter.task);
Al Virof0c3b502013-05-16 12:07:31 -04003316 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003317 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003318 }
Al Virof0c3b502013-05-16 12:07:31 -04003319 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003320 return 0;
3321}
3322
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003323/*
Janis Danisevskis1b3044e2016-05-20 17:00:08 -07003324 * proc_tid_comm_permission is a special permission function exclusively
3325 * used for the node /proc/<pid>/task/<tid>/comm.
3326 * It bypasses generic permission checks in the case where a task of the same
3327 * task group attempts to access the node.
3328 * The rationale behind this is that glibc and bionic access this node for
3329 * cross thread naming (pthread_set/getname_np(!self)). However, if
3330 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3331 * which locks out the cross thread naming implementation.
3332 * This function makes sure that the node is always accessible for members of
3333 * same thread group.
3334 */
3335static int proc_tid_comm_permission(struct inode *inode, int mask)
3336{
3337 bool is_same_tgroup;
3338 struct task_struct *task;
3339
3340 task = get_proc_task(inode);
3341 if (!task)
3342 return -ESRCH;
3343 is_same_tgroup = same_thread_group(current, task);
3344 put_task_struct(task);
3345
3346 if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3347 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3348 * read or written by the members of the corresponding
3349 * thread group.
3350 */
3351 return 0;
3352 }
3353
3354 return generic_permission(inode, mask);
3355}
3356
3357static const struct inode_operations proc_tid_comm_inode_operations = {
3358 .permission = proc_tid_comm_permission,
3359};
3360
3361/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003362 * Tasks
3363 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07003364static const struct pid_entry tid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003365 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
Jerome Marchand38355412010-04-27 13:13:06 -07003366 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08003367 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
Eric W. Biederman6ba8ed72014-07-31 16:27:08 -07003368#ifdef CONFIG_NET
3369 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3370#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003371 REG("environ", S_IRUSR, proc_environ_operations),
Al Viroc5317162016-10-05 18:43:43 -04003372 REG("auxv", S_IRUSR, proc_auxv_operations),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003373 ONE("status", S_IRUGO, proc_pid_status),
Djalal Harouni35a35042014-04-07 15:38:36 -07003374 ONE("personality", S_IRUSR, proc_pid_personality),
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -07003375 ONE("limits", S_IRUGO, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02003376#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003377 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02003378#endif
Janis Danisevskis1b3044e2016-05-20 17:00:08 -07003379 NOD("comm", S_IFREG|S_IRUGO|S_IWUSR,
3380 &proc_tid_comm_inode_operations,
3381 &proc_pid_set_comm_operations, {}),
Roland McGrathebcb6732008-07-25 19:46:00 -07003382#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -07003383 ONE("syscall", S_IRUSR, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07003384#endif
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -07003385 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003386 ONE("stat", S_IRUGO, proc_tid_stat),
3387 ONE("statm", S_IRUGO, proc_pid_statm),
Vlastimil Babka871305b2018-08-21 21:52:48 -07003388 REG("maps", S_IRUGO, proc_pid_maps_operations),
Iago López Galeiras2e13ba52015-06-25 15:00:57 -07003389#ifdef CONFIG_PROC_CHILDREN
Cyrill Gorcunov818411612012-05-31 16:26:43 -07003390 REG("children", S_IRUGO, proc_tid_children_operations),
3391#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003392#ifdef CONFIG_NUMA
Vlastimil Babka871305b2018-08-21 21:52:48 -07003393 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003394#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003395 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3396 LNK("cwd", proc_cwd_link),
3397 LNK("root", proc_root_link),
3398 LNK("exe", proc_exe_link),
3399 REG("mounts", S_IRUGO, proc_mounts_operations),
3400 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08003401#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003402 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
Vlastimil Babka871305b2018-08-21 21:52:48 -07003403 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
Daniel Colascione493b0e92017-09-06 16:25:08 -07003404 REG("smaps_rollup", S_IRUGO, proc_pid_smaps_rollup_operations),
Djalal Harouni32ed74a2014-04-07 15:38:38 -07003405 REG("pagemap", S_IRUSR, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003406#endif
3407#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003408 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003409#endif
3410#ifdef CONFIG_KALLSYMS
Alexey Dobriyanedfcd602014-08-08 14:21:44 -07003411 ONE("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003412#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03003413#ifdef CONFIG_STACKTRACE
Djalal Harouni35a35042014-04-07 15:38:36 -07003414 ONE("stack", S_IRUSR, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003415#endif
Naveen N. Rao5968cec2015-06-30 14:36:03 +05303416#ifdef CONFIG_SCHED_INFO
Alexey Dobriyanf6e826c2014-08-08 14:21:46 -07003417 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003418#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01003419#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003420 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01003421#endif
Paul Menage8793d852007-10-18 23:39:39 -07003422#ifdef CONFIG_PROC_PID_CPUSET
Zefan Li52de4772014-09-18 16:03:36 +08003423 ONE("cpuset", S_IRUGO, proc_cpuset_show),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003424#endif
Paul Menagea4243162007-10-18 23:39:35 -07003425#ifdef CONFIG_CGROUPS
Zefan Li006f4ac2014-09-18 16:03:15 +08003426 ONE("cgroup", S_IRUGO, proc_cgroup_show),
Paul Menagea4243162007-10-18 23:39:35 -07003427#endif
Alexey Dobriyan6ba51e32014-08-08 14:21:48 -07003428 ONE("oom_score", S_IRUGO, proc_oom_score),
David Rientjesfa0cbbf2012-11-12 17:53:04 -08003429 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
David Rientjesa63d83f2010-08-09 17:19:46 -07003430 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
Richard Guy Briggs4b7d2482019-01-22 17:06:39 -05003431#ifdef CONFIG_AUDIT
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003432 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
Al Viro26ec3c62011-02-15 21:24:05 -05003433 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003434#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003435#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003436 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mita1203c8e2017-07-14 14:49:57 -07003437 REG("fail-nth", 0644, proc_fail_nth_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003438#endif
Andrea Righi297c5d92008-07-25 01:48:49 -07003439#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07003440 ONE("io", S_IRUSR, proc_tid_io_accounting),
Andrea Righi297c5d92008-07-25 01:48:49 -07003441#endif
Eric W. Biederman22d917d2011-11-17 00:11:58 -08003442#ifdef CONFIG_USER_NS
3443 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3444 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
Eric W. Biedermanf76d2072012-08-30 01:24:05 -07003445 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
Eric W. Biederman9cc46512014-12-02 12:27:26 -06003446 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
Eric W. Biederman22d917d2011-11-17 00:11:58 -08003447#endif
Josh Poimboeuf7c23b332017-02-13 19:42:41 -06003448#ifdef CONFIG_LIVEPATCH
3449 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3450#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003451};
3452
Al Virof0c3b502013-05-16 12:07:31 -04003453static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003454{
Al Virof0c3b502013-05-16 12:07:31 -04003455 return proc_pident_readdir(file, ctx,
3456 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003457}
3458
Al Viro00cd8dd2012-06-10 17:13:09 -04003459static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3460{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003461 return proc_pident_lookup(dir, dentry,
Alexey Dobriyand5a572a2019-03-11 23:28:51 -07003462 tid_base_stuff,
3463 tid_base_stuff + ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003464}
3465
Arjan van de Ven00977a52007-02-12 00:55:34 -08003466static const struct file_operations proc_tid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003467 .read = generic_read_dir,
Al Virof50752e2016-04-20 17:13:54 -04003468 .iterate_shared = proc_tid_base_readdir,
3469 .llseek = generic_file_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003470};
3471
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003472static const struct inode_operations proc_tid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003473 .lookup = proc_tid_base_lookup,
3474 .getattr = pid_getattr,
3475 .setattr = proc_setattr,
3476};
3477
Al Viro0168b9e2018-05-03 09:21:05 -04003478static struct dentry *proc_task_instantiate(struct dentry *dentry,
3479 struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003480{
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003481 struct inode *inode;
Al Viro0168b9e2018-05-03 09:21:05 -04003482 inode = proc_pid_make_inode(dentry->d_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003483 if (!inode)
Al Viro0168b9e2018-05-03 09:21:05 -04003484 return ERR_PTR(-ENOENT);
Al Viro1bbc5512018-05-02 21:26:16 -04003485
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003486 inode->i_op = &proc_tid_base_inode_operations;
3487 inode->i_fop = &proc_tid_base_operations;
Al Viro1bbc5512018-05-02 21:26:16 -04003488 inode->i_flags |= S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07003489
Alexey Dobriyan1270dd82016-12-12 16:45:32 -08003490 set_nlink(inode, nlink_tid);
Al Viro1bbc5512018-05-02 21:26:16 -04003491 pid_update_inode(task, inode);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003492
Nick Pigginfb045ad2011-01-07 17:49:55 +11003493 d_set_d_op(dentry, &pid_dentry_operations);
Al Viro0168b9e2018-05-03 09:21:05 -04003494 return d_splice_alias(inode, dentry);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003495}
3496
Al Viro00cd8dd2012-06-10 17:13:09 -04003497static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003498{
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003499 struct task_struct *task;
3500 struct task_struct *leader = get_proc_task(dir);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003501 unsigned tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003502 struct pid_namespace *ns;
Al Viro0168b9e2018-05-03 09:21:05 -04003503 struct dentry *result = ERR_PTR(-ENOENT);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003504
3505 if (!leader)
3506 goto out_no_task;
3507
Alexey Dobriyandbcdb502014-08-08 14:21:25 -07003508 tid = name_to_int(&dentry->d_name);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003509 if (tid == ~0U)
3510 goto out;
3511
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003512 ns = dentry->d_sb->s_fs_info;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003513 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003514 task = find_task_by_pid_ns(tid, ns);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003515 if (task)
3516 get_task_struct(task);
3517 rcu_read_unlock();
3518 if (!task)
3519 goto out;
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07003520 if (!same_thread_group(leader, task))
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003521 goto out_drop_task;
3522
Al Viro0168b9e2018-05-03 09:21:05 -04003523 result = proc_task_instantiate(dentry, task, NULL);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003524out_drop_task:
3525 put_task_struct(task);
3526out:
3527 put_task_struct(leader);
3528out_no_task:
Al Viro0168b9e2018-05-03 09:21:05 -04003529 return result;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003530}
3531
3532/*
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003533 * Find the first tid of a thread group to return to user space.
3534 *
3535 * Usually this is just the thread group leader, but if the users
3536 * buffer was too small or there was a seek into the middle of the
3537 * directory we have more work todo.
3538 *
3539 * In the case of a short read we start with find_task_by_pid.
3540 *
3541 * In the case of a seek we start with the leader and walk nr
3542 * threads past it.
3543 */
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003544static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3545 struct pid_namespace *ns)
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003546{
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003547 struct task_struct *pos, *task;
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003548 unsigned long nr = f_pos;
3549
3550 if (nr != f_pos) /* 32bit overflow? */
3551 return NULL;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003552
Eric W. Biedermancc288732006-06-26 00:26:01 -07003553 rcu_read_lock();
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003554 task = pid_task(pid, PIDTYPE_PID);
3555 if (!task)
3556 goto fail;
3557
3558 /* Attempt to start with the tid of a thread */
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003559 if (tid && nr) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003560 pos = find_task_by_pid_ns(tid, ns);
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003561 if (pos && same_thread_group(pos, task))
Oleg Nesterova872ff02006-06-26 00:26:01 -07003562 goto found;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003563 }
3564
3565 /* If nr exceeds the number of threads there is nothing todo */
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003566 if (nr >= get_nr_threads(task))
Oleg Nesterovc986c142014-01-23 15:55:38 -08003567 goto fail;
Oleg Nesterova872ff02006-06-26 00:26:01 -07003568
3569 /* If we haven't found our starting place yet start
3570 * with the leader and walk nr threads forward.
3571 */
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003572 pos = task = task->group_leader;
Oleg Nesterovc986c142014-01-23 15:55:38 -08003573 do {
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003574 if (!nr--)
Oleg Nesterovc986c142014-01-23 15:55:38 -08003575 goto found;
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003576 } while_each_thread(task, pos);
Oleg Nesterovc986c142014-01-23 15:55:38 -08003577fail:
3578 pos = NULL;
3579 goto out;
Oleg Nesterova872ff02006-06-26 00:26:01 -07003580found:
3581 get_task_struct(pos);
3582out:
Eric W. Biedermancc288732006-06-26 00:26:01 -07003583 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003584 return pos;
3585}
3586
3587/*
3588 * Find the next thread in the thread list.
3589 * Return NULL if there is an error or no next thread.
3590 *
3591 * The reference to the input task_struct is released.
3592 */
3593static struct task_struct *next_tid(struct task_struct *start)
3594{
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003595 struct task_struct *pos = NULL;
Eric W. Biedermancc288732006-06-26 00:26:01 -07003596 rcu_read_lock();
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003597 if (pid_alive(start)) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003598 pos = next_thread(start);
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003599 if (thread_group_leader(pos))
3600 pos = NULL;
3601 else
3602 get_task_struct(pos);
3603 }
Eric W. Biedermancc288732006-06-26 00:26:01 -07003604 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003605 put_task_struct(start);
3606 return pos;
3607}
3608
Linus Torvalds1da177e2005-04-16 15:20:36 -07003609/* for the /proc/TGID/task/ directories */
Al Virof0c3b502013-05-16 12:07:31 -04003610static int proc_task_readdir(struct file *file, struct dir_context *ctx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003611{
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003612 struct inode *inode = file_inode(file);
3613 struct task_struct *task;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003614 struct pid_namespace *ns;
Al Virof0c3b502013-05-16 12:07:31 -04003615 int tid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003616
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003617 if (proc_inode_is_dead(inode))
Al Virof0c3b502013-05-16 12:07:31 -04003618 return -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003619
Al Virof0c3b502013-05-16 12:07:31 -04003620 if (!dir_emit_dots(file, ctx))
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003621 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003622
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003623 /* f_version caches the tgid value that the last readdir call couldn't
3624 * return. lseek aka telldir automagically resets f_version to 0.
3625 */
Christoph Hellwig76f668b2018-05-16 07:19:01 +02003626 ns = proc_pid_ns(inode);
Al Virof0c3b502013-05-16 12:07:31 -04003627 tid = (int)file->f_version;
3628 file->f_version = 0;
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003629 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003630 task;
Al Virof0c3b502013-05-16 12:07:31 -04003631 task = next_tid(task), ctx->pos++) {
Alexey Dobriyane3912ac2018-02-06 15:36:51 -08003632 char name[10 + 1];
Alexey Dobriyana4ef3892018-06-07 17:10:10 -07003633 unsigned int len;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003634 tid = task_pid_nr_ns(task, ns);
Alexey Dobriyane3912ac2018-02-06 15:36:51 -08003635 len = snprintf(name, sizeof(name), "%u", tid);
Al Virof0c3b502013-05-16 12:07:31 -04003636 if (!proc_fill_cache(file, ctx, name, len,
3637 proc_task_instantiate, task, NULL)) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003638 /* returning this tgid failed, save it as the first
3639 * pid for the next readir call */
Al Virof0c3b502013-05-16 12:07:31 -04003640 file->f_version = (u64)tid;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003641 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003642 break;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003643 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003644 }
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003645
Al Virof0c3b502013-05-16 12:07:31 -04003646 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003647}
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003648
David Howellsa528d352017-01-31 16:46:22 +00003649static int proc_task_getattr(const struct path *path, struct kstat *stat,
3650 u32 request_mask, unsigned int query_flags)
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003651{
David Howellsa528d352017-01-31 16:46:22 +00003652 struct inode *inode = d_inode(path->dentry);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003653 struct task_struct *p = get_proc_task(inode);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003654 generic_fillattr(inode, stat);
3655
Eric W. Biederman99f89552006-06-26 00:25:55 -07003656 if (p) {
Eric W. Biederman99f89552006-06-26 00:25:55 -07003657 stat->nlink += get_nr_threads(p);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003658 put_task_struct(p);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003659 }
3660
3661 return 0;
3662}
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003663
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003664static const struct inode_operations proc_task_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003665 .lookup = proc_task_lookup,
3666 .getattr = proc_task_getattr,
3667 .setattr = proc_setattr,
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003668 .permission = proc_pid_permission,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003669};
3670
Arjan van de Ven00977a52007-02-12 00:55:34 -08003671static const struct file_operations proc_task_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003672 .read = generic_read_dir,
Al Virof50752e2016-04-20 17:13:54 -04003673 .iterate_shared = proc_task_readdir,
3674 .llseek = generic_file_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003675};
Alexey Dobriyan1270dd82016-12-12 16:45:32 -08003676
3677void __init set_proc_pid_nlink(void)
3678{
3679 nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3680 nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3681}