blob: c1fdaecb8d23c99009d2aec5ec635f86b99ebd5b [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/proc/base.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * proc base directory handling functions
7 *
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
Mauricio Line070ad42005-09-03 15:55:10 -070014 *
15 *
16 * Changelog:
17 * 17-Jan-2005
18 * Allan Bezerra
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
23 *
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
25 *
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
32 *
33 * Changelog:
34 * 21-Feb-2005
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
37 *
38 * ChangeLog:
39 * 10-Mar-2005
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
42 *
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
45 *
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
Linus Torvalds1da177e2005-04-16 15:20:36 -070048 */
49
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080050#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070051
Linus Torvalds1da177e2005-04-16 15:20:36 -070052#include <linux/errno.h>
53#include <linux/time.h>
54#include <linux/proc_fs.h>
55#include <linux/stat.h>
Andrea Righi59954772008-07-27 17:29:15 +020056#include <linux/task_io_accounting_ops.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070057#include <linux/init.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080058#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070059#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040060#include <linux/fdtable.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#include <linux/string.h>
62#include <linux/seq_file.h>
63#include <linux/namei.h>
Kirill Korotaev6b3286e2006-12-08 02:37:56 -080064#include <linux/mnt_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070065#include <linux/mm.h>
David Rientjesa63d83f2010-08-09 17:19:46 -070066#include <linux/swap.h>
Dipankar Sarmab8359962005-09-09 13:04:14 -070067#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <linux/kallsyms.h>
Ken Chen2ec220e2008-11-10 11:26:08 +030069#include <linux/stacktrace.h>
Neil Hormand85f50d2007-10-18 23:40:37 -070070#include <linux/resource.h>
Kees Cook5096add2007-05-08 00:26:04 -070071#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070072#include <linux/mount.h>
73#include <linux/security.h>
74#include <linux/ptrace.h>
Roland McGrath0d094ef2008-07-25 19:45:49 -070075#include <linux/tracehook.h>
Andrew Morton87ebdc02013-02-27 17:03:16 -080076#include <linux/printk.h>
Paul Menagea4243162007-10-18 23:39:35 -070077#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070078#include <linux/cpuset.h>
79#include <linux/audit.h>
Al Viro5addc5d2005-11-07 17:15:49 -050080#include <linux/poll.h>
Serge E. Hallyn1651e142006-10-02 02:18:08 -070081#include <linux/nsproxy.h>
Alexey Dobriyan8ac773b2006-10-19 23:28:32 -070082#include <linux/oom.h>
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -070083#include <linux/elf.h>
Pavel Emelyanov60347f62007-10-18 23:40:03 -070084#include <linux/pid_namespace.h>
Eric W. Biederman22d917d2011-11-17 00:11:58 -080085#include <linux/user_namespace.h>
Al Viro5ad4e532009-03-29 19:50:06 -040086#include <linux/fs_struct.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090087#include <linux/slab.h>
Ingo Molnar4eb5aaa2017-02-08 18:51:29 +010088#include <linux/sched/autogroup.h>
Ingo Molnar6e84f312017-02-08 18:51:29 +010089#include <linux/sched/mm.h>
Ingo Molnarf7ccbae2017-02-08 18:51:30 +010090#include <linux/sched/coredump.h>
Ingo Molnarb17b0152017-02-08 18:51:35 +010091#include <linux/sched/debug.h>
Ingo Molnar3905f9a2017-02-05 12:07:04 +010092#include <linux/sched/stat.h>
Pavel Emelyanov640708a2012-01-10 15:11:23 -080093#include <linux/flex_array.h>
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +040094#include <linux/posix-timers.h>
Chris Metcalff133ecc2011-05-26 12:40:09 -040095#ifdef CONFIG_HARDWALL
96#include <asm/hardwall.h>
97#endif
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -080098#include <trace/events/oom.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099#include "internal.h"
Cyrill Gorcunovfaf60af2012-08-23 14:43:24 +0400100#include "fd.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101
Eric W. Biederman0f2fe202006-06-26 00:25:46 -0700102/* NOTE:
103 * Implementing inode permission operations in /proc is almost
104 * certainly an error. Permission checks need to happen during
105 * each system call not at open time. The reason is that most of
106 * what we wish to check for permissions in /proc varies at runtime.
107 *
108 * The classic example of a problem is opening file descriptors
109 * in /proc for a task before it execs a suid executable.
110 */
111
Alexey Dobriyan1270dd82016-12-12 16:45:32 -0800112static u8 nlink_tid;
113static u8 nlink_tgid;
114
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115struct pid_entry {
Alexey Dobriyancedbcca2014-08-08 14:21:33 -0700116 const char *name;
Alexey Dobriyan623f5942016-12-12 16:45:08 -0800117 unsigned int len;
Al Virod161a132011-07-24 03:36:29 -0400118 umode_t mode;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800119 const struct inode_operations *iop;
Arjan van de Ven00977a52007-02-12 00:55:34 -0800120 const struct file_operations *fop;
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700121 union proc_op op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122};
123
Eric W. Biederman61a28782006-10-02 02:18:49 -0700124#define NOD(NAME, MODE, IOP, FOP, OP) { \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700125 .name = (NAME), \
Eric Dumazetc5141e62007-05-08 00:26:15 -0700126 .len = sizeof(NAME) - 1, \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700127 .mode = MODE, \
128 .iop = IOP, \
129 .fop = FOP, \
130 .op = OP, \
131}
132
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300133#define DIR(NAME, MODE, iops, fops) \
134 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
135#define LNK(NAME, get_link) \
Eric W. Biederman61a28782006-10-02 02:18:49 -0700136 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700137 &proc_pid_link_inode_operations, NULL, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300138 { .proc_get_link = get_link } )
139#define REG(NAME, MODE, fops) \
140 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300141#define ONE(NAME, MODE, show) \
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800142 NOD(NAME, (S_IFREG|(MODE)), \
143 NULL, &proc_single_file_operations, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300144 { .proc_show = show } )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145
Vegard Nossumaed54172008-06-05 22:46:53 -0700146/*
147 * Count the number of hardlinks for the pid_entry table, excluding the .
148 * and .. links.
149 */
Alexey Dobriyan1270dd82016-12-12 16:45:32 -0800150static unsigned int __init pid_entry_nlink(const struct pid_entry *entries,
Vegard Nossumaed54172008-06-05 22:46:53 -0700151 unsigned int n)
152{
153 unsigned int i;
154 unsigned int count;
155
Alexey Dobriyan1270dd82016-12-12 16:45:32 -0800156 count = 2;
Vegard Nossumaed54172008-06-05 22:46:53 -0700157 for (i = 0; i < n; ++i) {
158 if (S_ISDIR(entries[i].mode))
159 ++count;
160 }
161
162 return count;
163}
164
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200165static int get_task_root(struct task_struct *task, struct path *root)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166{
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000167 int result = -ENOENT;
168
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700169 task_lock(task);
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200170 if (task->fs) {
171 get_fs_root(task->fs, root);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000172 result = 0;
173 }
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700174 task_unlock(task);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000175 return result;
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700176}
177
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800178static int proc_cwd_link(struct dentry *dentry, struct path *path)
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700179{
David Howells2b0143b2015-03-17 22:25:59 +0000180 struct task_struct *task = get_proc_task(d_inode(dentry));
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700181 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700182
183 if (task) {
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200184 task_lock(task);
185 if (task->fs) {
186 get_fs_pwd(task->fs, path);
187 result = 0;
188 }
189 task_unlock(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700190 put_task_struct(task);
191 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 return result;
193}
194
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800195static int proc_root_link(struct dentry *dentry, struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196{
David Howells2b0143b2015-03-17 22:25:59 +0000197 struct task_struct *task = get_proc_task(d_inode(dentry));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700199
200 if (task) {
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200201 result = get_task_root(task, path);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700202 put_task_struct(task);
203 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 return result;
205}
206
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700207static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
208 size_t _count, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209{
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700210 struct task_struct *tsk;
211 struct mm_struct *mm;
212 char *page;
213 unsigned long count = _count;
214 unsigned long arg_start, arg_end, env_start, env_end;
215 unsigned long len1, len2, len;
216 unsigned long p;
217 char c;
218 ssize_t rv;
219
220 BUG_ON(*pos < 0);
221
222 tsk = get_proc_task(file_inode(file));
223 if (!tsk)
224 return -ESRCH;
225 mm = get_task_mm(tsk);
226 put_task_struct(tsk);
227 if (!mm)
228 return 0;
229 /* Check if process spawned far enough to have cmdline. */
230 if (!mm->env_end) {
231 rv = 0;
232 goto out_mmput;
233 }
234
235 page = (char *)__get_free_page(GFP_TEMPORARY);
236 if (!page) {
237 rv = -ENOMEM;
238 goto out_mmput;
239 }
240
241 down_read(&mm->mmap_sem);
242 arg_start = mm->arg_start;
243 arg_end = mm->arg_end;
244 env_start = mm->env_start;
245 env_end = mm->env_end;
246 up_read(&mm->mmap_sem);
247
248 BUG_ON(arg_start > arg_end);
249 BUG_ON(env_start > env_end);
250
251 len1 = arg_end - arg_start;
252 len2 = env_end - env_start;
253
Alexey Dobriyan3581d452015-07-17 16:24:09 -0700254 /* Empty ARGV. */
255 if (len1 == 0) {
256 rv = 0;
257 goto out_free_page;
258 }
Alexey Dobriyan2ca66ff2014-08-08 14:21:41 -0700259 /*
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700260 * Inherently racy -- command line shares address space
261 * with code and data.
Alexey Dobriyan2ca66ff2014-08-08 14:21:41 -0700262 */
Linus Torvalds272ddc82016-10-24 19:00:44 -0700263 rv = access_remote_vm(mm, arg_end - 1, &c, 1, 0);
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700264 if (rv <= 0)
265 goto out_free_page;
266
267 rv = 0;
268
269 if (c == '\0') {
270 /* Command line (set of strings) occupies whole ARGV. */
271 if (len1 <= *pos)
272 goto out_free_page;
273
274 p = arg_start + *pos;
275 len = len1 - *pos;
276 while (count > 0 && len > 0) {
277 unsigned int _count;
278 int nr_read;
279
280 _count = min3(count, len, PAGE_SIZE);
Linus Torvalds272ddc82016-10-24 19:00:44 -0700281 nr_read = access_remote_vm(mm, p, page, _count, 0);
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700282 if (nr_read < 0)
283 rv = nr_read;
284 if (nr_read <= 0)
285 goto out_free_page;
286
287 if (copy_to_user(buf, page, nr_read)) {
288 rv = -EFAULT;
289 goto out_free_page;
290 }
291
292 p += nr_read;
293 len -= nr_read;
294 buf += nr_read;
295 count -= nr_read;
296 rv += nr_read;
297 }
298 } else {
299 /*
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700300 * Command line (1 string) occupies ARGV and
301 * extends into ENVP.
302 */
Alexey Dobriyana0a07b82017-02-24 15:00:20 -0800303 struct {
304 unsigned long p;
305 unsigned long len;
306 } cmdline[2] = {
307 { .p = arg_start, .len = len1 },
308 { .p = env_start, .len = len2 },
309 };
310 loff_t pos1 = *pos;
311 unsigned int i;
312
313 i = 0;
314 while (i < 2 && pos1 >= cmdline[i].len) {
315 pos1 -= cmdline[i].len;
316 i++;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700317 }
Alexey Dobriyana0a07b82017-02-24 15:00:20 -0800318 while (i < 2) {
319 p = cmdline[i].p + pos1;
320 len = cmdline[i].len - pos1;
321 while (count > 0 && len > 0) {
322 unsigned int _count, l;
323 int nr_read;
324 bool final;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700325
Alexey Dobriyana0a07b82017-02-24 15:00:20 -0800326 _count = min3(count, len, PAGE_SIZE);
327 nr_read = access_remote_vm(mm, p, page, _count, 0);
328 if (nr_read < 0)
329 rv = nr_read;
330 if (nr_read <= 0)
331 goto out_free_page;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700332
Alexey Dobriyana0a07b82017-02-24 15:00:20 -0800333 /*
334 * Command line can be shorter than whole ARGV
335 * even if last "marker" byte says it is not.
336 */
337 final = false;
338 l = strnlen(page, nr_read);
339 if (l < nr_read) {
340 nr_read = l;
341 final = true;
342 }
343
344 if (copy_to_user(buf, page, nr_read)) {
345 rv = -EFAULT;
346 goto out_free_page;
347 }
348
349 p += nr_read;
350 len -= nr_read;
351 buf += nr_read;
352 count -= nr_read;
353 rv += nr_read;
354
355 if (final)
356 goto out_free_page;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700357 }
358
Alexey Dobriyana0a07b82017-02-24 15:00:20 -0800359 /* Only first chunk can be read partially. */
360 pos1 = 0;
361 i++;
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700362 }
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700363 }
364
365out_free_page:
366 free_page((unsigned long)page);
367out_mmput:
368 mmput(mm);
369 if (rv > 0)
370 *pos += rv;
371 return rv;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372}
373
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -0700374static const struct file_operations proc_pid_cmdline_ops = {
375 .read = proc_pid_cmdline_read,
376 .llseek = generic_file_llseek,
377};
378
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379#ifdef CONFIG_KALLSYMS
380/*
381 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
382 * Returns the resolved symbol. If that fails, simply return the address.
383 */
Alexey Dobriyanedfcd602014-08-08 14:21:44 -0700384static int proc_pid_wchan(struct seq_file *m, struct pid_namespace *ns,
385 struct pid *pid, struct task_struct *task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386{
Alexey Dobriyanffb45122007-05-08 00:28:41 -0700387 unsigned long wchan;
Tejun Heo9281ace2007-07-17 04:03:51 -0700388 char symname[KSYM_NAME_LEN];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389
390 wchan = get_wchan(task);
391
Jann Horncaaee622016-01-20 15:00:04 -0800392 if (wchan && ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)
393 && !lookup_symbol_name(wchan, symname))
Joe Perches25ce3192015-04-15 16:18:17 -0700394 seq_printf(m, "%s", symname);
Ingo Molnarb2f73922015-09-30 15:59:17 +0200395 else
Robin Humble1e92a612016-05-09 16:28:46 -0700396 seq_putc(m, '0');
Joe Perches25ce3192015-04-15 16:18:17 -0700397
398 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399}
400#endif /* CONFIG_KALLSYMS */
401
Al Viroa9712bc2011-03-23 15:52:50 -0400402static int lock_trace(struct task_struct *task)
403{
404 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
405 if (err)
406 return err;
Jann Horncaaee622016-01-20 15:00:04 -0800407 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH_FSCREDS)) {
Al Viroa9712bc2011-03-23 15:52:50 -0400408 mutex_unlock(&task->signal->cred_guard_mutex);
409 return -EPERM;
410 }
411 return 0;
412}
413
414static void unlock_trace(struct task_struct *task)
415{
416 mutex_unlock(&task->signal->cred_guard_mutex);
417}
418
Ken Chen2ec220e2008-11-10 11:26:08 +0300419#ifdef CONFIG_STACKTRACE
420
421#define MAX_STACK_TRACE_DEPTH 64
422
423static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
424 struct pid *pid, struct task_struct *task)
425{
426 struct stack_trace trace;
427 unsigned long *entries;
Al Viroa9712bc2011-03-23 15:52:50 -0400428 int err;
Ken Chen2ec220e2008-11-10 11:26:08 +0300429 int i;
430
431 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
432 if (!entries)
433 return -ENOMEM;
434
435 trace.nr_entries = 0;
436 trace.max_entries = MAX_STACK_TRACE_DEPTH;
437 trace.entries = entries;
438 trace.skip = 0;
Ken Chen2ec220e2008-11-10 11:26:08 +0300439
Al Viroa9712bc2011-03-23 15:52:50 -0400440 err = lock_trace(task);
441 if (!err) {
442 save_stack_trace_tsk(task, &trace);
443
444 for (i = 0; i < trace.nr_entries; i++) {
Josh Poimboeuf8b927d72016-08-18 10:59:07 -0500445 seq_printf(m, "[<%pK>] %pB\n",
Al Viroa9712bc2011-03-23 15:52:50 -0400446 (void *)entries[i], (void *)entries[i]);
447 }
448 unlock_trace(task);
Ken Chen2ec220e2008-11-10 11:26:08 +0300449 }
450 kfree(entries);
451
Al Viroa9712bc2011-03-23 15:52:50 -0400452 return err;
Ken Chen2ec220e2008-11-10 11:26:08 +0300453}
454#endif
455
Naveen N. Rao5968cec2015-06-30 14:36:03 +0530456#ifdef CONFIG_SCHED_INFO
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457/*
458 * Provides /proc/PID/schedstat
459 */
Alexey Dobriyanf6e826c2014-08-08 14:21:46 -0700460static int proc_pid_schedstat(struct seq_file *m, struct pid_namespace *ns,
461 struct pid *pid, struct task_struct *task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462{
Naveen N. Rao5968cec2015-06-30 14:36:03 +0530463 if (unlikely(!sched_info_on()))
464 seq_printf(m, "0 0 0\n");
465 else
466 seq_printf(m, "%llu %llu %lu\n",
Joe Perches25ce3192015-04-15 16:18:17 -0700467 (unsigned long long)task->se.sum_exec_runtime,
468 (unsigned long long)task->sched_info.run_delay,
469 task->sched_info.pcount);
470
471 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472}
473#endif
474
Arjan van de Ven97455122008-01-25 21:08:34 +0100475#ifdef CONFIG_LATENCYTOP
476static int lstats_show_proc(struct seq_file *m, void *v)
477{
478 int i;
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800479 struct inode *inode = m->private;
480 struct task_struct *task = get_proc_task(inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100481
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800482 if (!task)
483 return -ESRCH;
484 seq_puts(m, "Latency Top version : v0.1\n");
Arjan van de Ven97455122008-01-25 21:08:34 +0100485 for (i = 0; i < 32; i++) {
Joe Perches34e49d42011-01-12 17:00:30 -0800486 struct latency_record *lr = &task->latency_record[i];
487 if (lr->backtrace[0]) {
Arjan van de Ven97455122008-01-25 21:08:34 +0100488 int q;
Joe Perches34e49d42011-01-12 17:00:30 -0800489 seq_printf(m, "%i %li %li",
490 lr->count, lr->time, lr->max);
Arjan van de Ven97455122008-01-25 21:08:34 +0100491 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
Joe Perches34e49d42011-01-12 17:00:30 -0800492 unsigned long bt = lr->backtrace[q];
493 if (!bt)
Arjan van de Ven97455122008-01-25 21:08:34 +0100494 break;
Joe Perches34e49d42011-01-12 17:00:30 -0800495 if (bt == ULONG_MAX)
Arjan van de Ven97455122008-01-25 21:08:34 +0100496 break;
Joe Perches34e49d42011-01-12 17:00:30 -0800497 seq_printf(m, " %ps", (void *)bt);
Arjan van de Ven97455122008-01-25 21:08:34 +0100498 }
Alexey Dobriyan9d6de122011-01-12 17:00:32 -0800499 seq_putc(m, '\n');
Arjan van de Ven97455122008-01-25 21:08:34 +0100500 }
501
502 }
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800503 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100504 return 0;
505}
506
507static int lstats_open(struct inode *inode, struct file *file)
508{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800509 return single_open(file, lstats_show_proc, inode);
Hiroshi Shimamotod6643d12008-02-14 10:27:00 -0800510}
511
Arjan van de Ven97455122008-01-25 21:08:34 +0100512static ssize_t lstats_write(struct file *file, const char __user *buf,
513 size_t count, loff_t *offs)
514{
Al Viro496ad9a2013-01-23 17:07:38 -0500515 struct task_struct *task = get_proc_task(file_inode(file));
Arjan van de Ven97455122008-01-25 21:08:34 +0100516
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800517 if (!task)
518 return -ESRCH;
Arjan van de Ven97455122008-01-25 21:08:34 +0100519 clear_all_latency_tracing(task);
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800520 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100521
522 return count;
523}
524
525static const struct file_operations proc_lstats_operations = {
526 .open = lstats_open,
527 .read = seq_read,
528 .write = lstats_write,
529 .llseek = seq_lseek,
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800530 .release = single_release,
Arjan van de Ven97455122008-01-25 21:08:34 +0100531};
532
533#endif
534
Alexey Dobriyan6ba51e32014-08-08 14:21:48 -0700535static int proc_oom_score(struct seq_file *m, struct pid_namespace *ns,
536 struct pid *pid, struct task_struct *task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537{
David Rientjesa7f638f2012-05-29 15:06:47 -0700538 unsigned long totalpages = totalram_pages + total_swap_pages;
Oleg Nesterovb95c35e2010-04-01 15:13:57 +0200539 unsigned long points = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540
Oleg Nesterovef419392016-08-02 14:03:19 -0700541 points = oom_badness(task, NULL, NULL, totalpages) *
542 1000 / totalpages;
Joe Perches25ce3192015-04-15 16:18:17 -0700543 seq_printf(m, "%lu\n", points);
544
545 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546}
547
Neil Hormand85f50d2007-10-18 23:40:37 -0700548struct limit_names {
Alexey Dobriyancedbcca2014-08-08 14:21:33 -0700549 const char *name;
550 const char *unit;
Neil Hormand85f50d2007-10-18 23:40:37 -0700551};
552
553static const struct limit_names lnames[RLIM_NLIMITS] = {
Kees Cookcff4edb2009-09-22 16:45:32 -0700554 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700555 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
556 [RLIMIT_DATA] = {"Max data size", "bytes"},
557 [RLIMIT_STACK] = {"Max stack size", "bytes"},
558 [RLIMIT_CORE] = {"Max core file size", "bytes"},
559 [RLIMIT_RSS] = {"Max resident set", "bytes"},
560 [RLIMIT_NPROC] = {"Max processes", "processes"},
561 [RLIMIT_NOFILE] = {"Max open files", "files"},
562 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
563 [RLIMIT_AS] = {"Max address space", "bytes"},
564 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
565 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
566 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
567 [RLIMIT_NICE] = {"Max nice priority", NULL},
568 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
Eugene Teo88081172008-02-23 15:23:52 -0800569 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700570};
571
572/* Display limits for a process */
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700573static int proc_pid_limits(struct seq_file *m, struct pid_namespace *ns,
574 struct pid *pid, struct task_struct *task)
Neil Hormand85f50d2007-10-18 23:40:37 -0700575{
576 unsigned int i;
Neil Hormand85f50d2007-10-18 23:40:37 -0700577 unsigned long flags;
Neil Hormand85f50d2007-10-18 23:40:37 -0700578
579 struct rlimit rlim[RLIM_NLIMITS];
580
Lai Jiangshana6bebbc2008-10-05 00:51:15 +0400581 if (!lock_task_sighand(task, &flags))
Neil Hormand85f50d2007-10-18 23:40:37 -0700582 return 0;
Neil Hormand85f50d2007-10-18 23:40:37 -0700583 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
584 unlock_task_sighand(task, &flags);
Neil Hormand85f50d2007-10-18 23:40:37 -0700585
586 /*
587 * print the file header
588 */
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700589 seq_printf(m, "%-25s %-20s %-20s %-10s\n",
Joe Perches25ce3192015-04-15 16:18:17 -0700590 "Limit", "Soft Limit", "Hard Limit", "Units");
Neil Hormand85f50d2007-10-18 23:40:37 -0700591
592 for (i = 0; i < RLIM_NLIMITS; i++) {
593 if (rlim[i].rlim_cur == RLIM_INFINITY)
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700594 seq_printf(m, "%-25s %-20s ",
Joe Perches25ce3192015-04-15 16:18:17 -0700595 lnames[i].name, "unlimited");
Neil Hormand85f50d2007-10-18 23:40:37 -0700596 else
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700597 seq_printf(m, "%-25s %-20lu ",
Joe Perches25ce3192015-04-15 16:18:17 -0700598 lnames[i].name, rlim[i].rlim_cur);
Neil Hormand85f50d2007-10-18 23:40:37 -0700599
600 if (rlim[i].rlim_max == RLIM_INFINITY)
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700601 seq_printf(m, "%-20s ", "unlimited");
Neil Hormand85f50d2007-10-18 23:40:37 -0700602 else
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700603 seq_printf(m, "%-20lu ", rlim[i].rlim_max);
Neil Hormand85f50d2007-10-18 23:40:37 -0700604
605 if (lnames[i].unit)
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700606 seq_printf(m, "%-10s\n", lnames[i].unit);
Neil Hormand85f50d2007-10-18 23:40:37 -0700607 else
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700608 seq_putc(m, '\n');
Neil Hormand85f50d2007-10-18 23:40:37 -0700609 }
610
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -0700611 return 0;
Neil Hormand85f50d2007-10-18 23:40:37 -0700612}
613
Roland McGrathebcb6732008-07-25 19:46:00 -0700614#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -0700615static int proc_pid_syscall(struct seq_file *m, struct pid_namespace *ns,
616 struct pid *pid, struct task_struct *task)
Roland McGrathebcb6732008-07-25 19:46:00 -0700617{
618 long nr;
619 unsigned long args[6], sp, pc;
Joe Perches25ce3192015-04-15 16:18:17 -0700620 int res;
621
622 res = lock_trace(task);
Al Viroa9712bc2011-03-23 15:52:50 -0400623 if (res)
624 return res;
Roland McGrathebcb6732008-07-25 19:46:00 -0700625
626 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -0700627 seq_puts(m, "running\n");
Al Viroa9712bc2011-03-23 15:52:50 -0400628 else if (nr < 0)
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -0700629 seq_printf(m, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
Al Viroa9712bc2011-03-23 15:52:50 -0400630 else
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -0700631 seq_printf(m,
Roland McGrathebcb6732008-07-25 19:46:00 -0700632 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
633 nr,
634 args[0], args[1], args[2], args[3], args[4], args[5],
635 sp, pc);
Al Viroa9712bc2011-03-23 15:52:50 -0400636 unlock_trace(task);
Joe Perches25ce3192015-04-15 16:18:17 -0700637
638 return 0;
Roland McGrathebcb6732008-07-25 19:46:00 -0700639}
640#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
641
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642/************************************************************************/
643/* Here the fs part begins */
644/************************************************************************/
645
646/* permission checks */
Eric W. Biederman778c1142006-06-26 00:25:58 -0700647static int proc_fd_access_allowed(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648{
Eric W. Biederman778c1142006-06-26 00:25:58 -0700649 struct task_struct *task;
650 int allowed = 0;
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700651 /* Allow access to a task's file descriptors if it is us or we
652 * may use ptrace attach to the process and find out that
653 * information.
Eric W. Biederman778c1142006-06-26 00:25:58 -0700654 */
655 task = get_proc_task(inode);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700656 if (task) {
Jann Horncaaee622016-01-20 15:00:04 -0800657 allowed = ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
Eric W. Biederman778c1142006-06-26 00:25:58 -0700658 put_task_struct(task);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700659 }
Eric W. Biederman778c1142006-06-26 00:25:58 -0700660 return allowed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661}
662
Eric W. Biederman6b4e3062010-03-07 16:41:34 -0800663int proc_setattr(struct dentry *dentry, struct iattr *attr)
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700664{
665 int error;
David Howells2b0143b2015-03-17 22:25:59 +0000666 struct inode *inode = d_inode(dentry);
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700667
668 if (attr->ia_valid & ATTR_MODE)
669 return -EPERM;
670
Jan Kara31051c82016-05-26 16:55:18 +0200671 error = setattr_prepare(dentry, attr);
Christoph Hellwig10257742010-06-04 11:30:02 +0200672 if (error)
673 return error;
674
Christoph Hellwig10257742010-06-04 11:30:02 +0200675 setattr_copy(inode, attr);
676 mark_inode_dirty(inode);
677 return 0;
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700678}
679
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800680/*
681 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
682 * or euid/egid (for hide_pid_min=2)?
683 */
684static bool has_pid_permissions(struct pid_namespace *pid,
685 struct task_struct *task,
686 int hide_pid_min)
687{
688 if (pid->hide_pid < hide_pid_min)
689 return true;
690 if (in_group_p(pid->pid_gid))
691 return true;
Jann Horncaaee622016-01-20 15:00:04 -0800692 return ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS);
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800693}
694
695
696static int proc_pid_permission(struct inode *inode, int mask)
697{
698 struct pid_namespace *pid = inode->i_sb->s_fs_info;
699 struct task_struct *task;
700 bool has_perms;
701
702 task = get_proc_task(inode);
Xiaotian Fenga2ef9902012-01-12 17:17:08 -0800703 if (!task)
704 return -ESRCH;
Lafcadio Wluiki796f5712017-02-24 15:00:23 -0800705 has_perms = has_pid_permissions(pid, task, HIDEPID_NO_ACCESS);
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800706 put_task_struct(task);
707
708 if (!has_perms) {
Lafcadio Wluiki796f5712017-02-24 15:00:23 -0800709 if (pid->hide_pid == HIDEPID_INVISIBLE) {
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800710 /*
711 * Let's make getdents(), stat(), and open()
712 * consistent with each other. If a process
713 * may not stat() a file, it shouldn't be seen
714 * in procfs at all.
715 */
716 return -ENOENT;
717 }
718
719 return -EPERM;
720 }
721 return generic_permission(inode, mask);
722}
723
724
725
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800726static const struct inode_operations proc_def_inode_operations = {
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700727 .setattr = proc_setattr,
728};
729
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800730static int proc_single_show(struct seq_file *m, void *v)
731{
732 struct inode *inode = m->private;
733 struct pid_namespace *ns;
734 struct pid *pid;
735 struct task_struct *task;
736 int ret;
737
738 ns = inode->i_sb->s_fs_info;
739 pid = proc_pid(inode);
740 task = get_pid_task(pid, PIDTYPE_PID);
741 if (!task)
742 return -ESRCH;
743
744 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
745
746 put_task_struct(task);
747 return ret;
748}
749
750static int proc_single_open(struct inode *inode, struct file *filp)
751{
Jovi Zhangc6a34052011-01-12 17:00:34 -0800752 return single_open(filp, proc_single_show, inode);
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800753}
754
755static const struct file_operations proc_single_file_operations = {
756 .open = proc_single_open,
757 .read = seq_read,
758 .llseek = seq_lseek,
759 .release = single_release,
760};
761
Oleg Nesterov5381e162014-10-09 15:25:24 -0700762
763struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
764{
765 struct task_struct *task = get_proc_task(inode);
766 struct mm_struct *mm = ERR_PTR(-ESRCH);
767
768 if (task) {
Jann Horncaaee622016-01-20 15:00:04 -0800769 mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
Oleg Nesterov5381e162014-10-09 15:25:24 -0700770 put_task_struct(task);
771
772 if (!IS_ERR_OR_NULL(mm)) {
773 /* ensure this mm_struct can't be freed */
Vegard Nossumf1f10072017-02-27 14:30:07 -0800774 mmgrab(mm);
Oleg Nesterov5381e162014-10-09 15:25:24 -0700775 /* but do not pin its memory */
776 mmput(mm);
777 }
778 }
779
780 return mm;
781}
782
Cong Wangb409e572012-05-31 16:26:17 -0700783static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784{
Oleg Nesterov5381e162014-10-09 15:25:24 -0700785 struct mm_struct *mm = proc_mem_open(inode, mode);
Linus Torvaldse2683372012-01-17 15:21:19 -0800786
787 if (IS_ERR(mm))
788 return PTR_ERR(mm);
789
Linus Torvaldse2683372012-01-17 15:21:19 -0800790 file->private_data = mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 return 0;
792}
793
Cong Wangb409e572012-05-31 16:26:17 -0700794static int mem_open(struct inode *inode, struct file *file)
795{
Djalal Harounibc452b42012-07-30 14:42:28 -0700796 int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
797
798 /* OK to pass negative loff_t, we can catch out-of-range */
799 file->f_mode |= FMODE_UNSIGNED_OFFSET;
800
801 return ret;
Cong Wangb409e572012-05-31 16:26:17 -0700802}
803
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100804static ssize_t mem_rw(struct file *file, char __user *buf,
805 size_t count, loff_t *ppos, int write)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806{
Linus Torvaldse2683372012-01-17 15:21:19 -0800807 struct mm_struct *mm = file->private_data;
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100808 unsigned long addr = *ppos;
809 ssize_t copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 char *page;
Linus Torvalds272ddc82016-10-24 19:00:44 -0700811 unsigned int flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812
Linus Torvaldse2683372012-01-17 15:21:19 -0800813 if (!mm)
814 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815
Mel Gormane12ba742007-10-16 01:25:52 -0700816 page = (char *)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 if (!page)
Linus Torvaldse2683372012-01-17 15:21:19 -0800818 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700820 copied = 0;
Vegard Nossum388f7932017-02-27 14:30:13 -0800821 if (!mmget_not_zero(mm))
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100822 goto free;
823
Linus Torvaldsf511c0b2017-05-30 12:38:59 -0700824 flags = FOLL_FORCE | (write ? FOLL_WRITE : 0);
Lorenzo Stoakes6347e8d2016-10-13 01:20:19 +0100825
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 while (count > 0) {
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100827 int this_len = min_t(int, count, PAGE_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100829 if (write && copy_from_user(page, buf, this_len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 copied = -EFAULT;
831 break;
832 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100833
Lorenzo Stoakes6347e8d2016-10-13 01:20:19 +0100834 this_len = access_remote_vm(mm, addr, page, this_len, flags);
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100835 if (!this_len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 if (!copied)
837 copied = -EIO;
838 break;
839 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100840
841 if (!write && copy_to_user(buf, page, this_len)) {
842 copied = -EFAULT;
843 break;
844 }
845
846 buf += this_len;
847 addr += this_len;
848 copied += this_len;
849 count -= this_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100851 *ppos = addr;
KOSAKI Motohiro30cd8902011-05-26 16:25:52 -0700852
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100853 mmput(mm);
854free:
KOSAKI Motohiro30cd8902011-05-26 16:25:52 -0700855 free_page((unsigned long) page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 return copied;
857}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100859static ssize_t mem_read(struct file *file, char __user *buf,
860 size_t count, loff_t *ppos)
861{
862 return mem_rw(file, buf, count, ppos, 0);
863}
864
865static ssize_t mem_write(struct file *file, const char __user *buf,
866 size_t count, loff_t *ppos)
867{
868 return mem_rw(file, (char __user*)buf, count, ppos, 1);
869}
870
Matt Mackall85863e42008-02-04 22:29:04 -0800871loff_t mem_lseek(struct file *file, loff_t offset, int orig)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872{
873 switch (orig) {
874 case 0:
875 file->f_pos = offset;
876 break;
877 case 1:
878 file->f_pos += offset;
879 break;
880 default:
881 return -EINVAL;
882 }
883 force_successful_syscall_return();
884 return file->f_pos;
885}
886
Linus Torvaldse2683372012-01-17 15:21:19 -0800887static int mem_release(struct inode *inode, struct file *file)
888{
889 struct mm_struct *mm = file->private_data;
Oleg Nesterov71879d32012-01-31 17:14:38 +0100890 if (mm)
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100891 mmdrop(mm);
Linus Torvaldse2683372012-01-17 15:21:19 -0800892 return 0;
893}
894
Arjan van de Ven00977a52007-02-12 00:55:34 -0800895static const struct file_operations proc_mem_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 .llseek = mem_lseek,
897 .read = mem_read,
898 .write = mem_write,
899 .open = mem_open,
Linus Torvaldse2683372012-01-17 15:21:19 -0800900 .release = mem_release,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901};
902
Cong Wangb409e572012-05-31 16:26:17 -0700903static int environ_open(struct inode *inode, struct file *file)
904{
905 return __mem_open(inode, file, PTRACE_MODE_READ);
906}
907
James Pearson315e28c2007-10-16 23:30:17 -0700908static ssize_t environ_read(struct file *file, char __user *buf,
909 size_t count, loff_t *ppos)
910{
James Pearson315e28c2007-10-16 23:30:17 -0700911 char *page;
912 unsigned long src = *ppos;
Cong Wangb409e572012-05-31 16:26:17 -0700913 int ret = 0;
914 struct mm_struct *mm = file->private_data;
Mateusz Guzika3b609e2016-01-20 15:01:05 -0800915 unsigned long env_start, env_end;
James Pearson315e28c2007-10-16 23:30:17 -0700916
Mathias Krause8148a732016-05-05 16:22:26 -0700917 /* Ensure the process spawned far enough to have an environment. */
918 if (!mm || !mm->env_end)
Cong Wangb409e572012-05-31 16:26:17 -0700919 return 0;
James Pearson315e28c2007-10-16 23:30:17 -0700920
James Pearson315e28c2007-10-16 23:30:17 -0700921 page = (char *)__get_free_page(GFP_TEMPORARY);
922 if (!page)
Cong Wangb409e572012-05-31 16:26:17 -0700923 return -ENOMEM;
James Pearson315e28c2007-10-16 23:30:17 -0700924
Al Virod6f64b82011-02-15 22:26:01 -0500925 ret = 0;
Vegard Nossum388f7932017-02-27 14:30:13 -0800926 if (!mmget_not_zero(mm))
Cong Wangb409e572012-05-31 16:26:17 -0700927 goto free;
Mateusz Guzika3b609e2016-01-20 15:01:05 -0800928
929 down_read(&mm->mmap_sem);
930 env_start = mm->env_start;
931 env_end = mm->env_end;
932 up_read(&mm->mmap_sem);
933
James Pearson315e28c2007-10-16 23:30:17 -0700934 while (count > 0) {
Djalal Harounie8905ec2012-07-30 14:42:26 -0700935 size_t this_len, max_len;
936 int retval;
937
Mateusz Guzika3b609e2016-01-20 15:01:05 -0800938 if (src >= (env_end - env_start))
Djalal Harounie8905ec2012-07-30 14:42:26 -0700939 break;
James Pearson315e28c2007-10-16 23:30:17 -0700940
Mateusz Guzika3b609e2016-01-20 15:01:05 -0800941 this_len = env_end - (env_start + src);
James Pearson315e28c2007-10-16 23:30:17 -0700942
Djalal Harounie8905ec2012-07-30 14:42:26 -0700943 max_len = min_t(size_t, PAGE_SIZE, count);
944 this_len = min(max_len, this_len);
James Pearson315e28c2007-10-16 23:30:17 -0700945
Linus Torvalds272ddc82016-10-24 19:00:44 -0700946 retval = access_remote_vm(mm, (env_start + src), page, this_len, 0);
James Pearson315e28c2007-10-16 23:30:17 -0700947
948 if (retval <= 0) {
949 ret = retval;
950 break;
951 }
952
953 if (copy_to_user(buf, page, retval)) {
954 ret = -EFAULT;
955 break;
956 }
957
958 ret += retval;
959 src += retval;
960 buf += retval;
961 count -= retval;
962 }
963 *ppos = src;
James Pearson315e28c2007-10-16 23:30:17 -0700964 mmput(mm);
Cong Wangb409e572012-05-31 16:26:17 -0700965
966free:
James Pearson315e28c2007-10-16 23:30:17 -0700967 free_page((unsigned long) page);
James Pearson315e28c2007-10-16 23:30:17 -0700968 return ret;
969}
970
971static const struct file_operations proc_environ_operations = {
Cong Wangb409e572012-05-31 16:26:17 -0700972 .open = environ_open,
James Pearson315e28c2007-10-16 23:30:17 -0700973 .read = environ_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +0100974 .llseek = generic_file_llseek,
Cong Wangb409e572012-05-31 16:26:17 -0700975 .release = mem_release,
James Pearson315e28c2007-10-16 23:30:17 -0700976};
977
Al Viroc5317162016-10-05 18:43:43 -0400978static int auxv_open(struct inode *inode, struct file *file)
979{
980 return __mem_open(inode, file, PTRACE_MODE_READ_FSCREDS);
981}
982
983static ssize_t auxv_read(struct file *file, char __user *buf,
984 size_t count, loff_t *ppos)
985{
986 struct mm_struct *mm = file->private_data;
987 unsigned int nwords = 0;
Leon Yu06b28492016-10-27 17:46:50 -0700988
989 if (!mm)
990 return 0;
Al Viroc5317162016-10-05 18:43:43 -0400991 do {
992 nwords += 2;
993 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
994 return simple_read_from_buffer(buf, count, ppos, mm->saved_auxv,
995 nwords * sizeof(mm->saved_auxv[0]));
996}
997
998static const struct file_operations proc_auxv_operations = {
999 .open = auxv_open,
1000 .read = auxv_read,
1001 .llseek = generic_file_llseek,
1002 .release = mem_release,
1003};
1004
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001005static ssize_t oom_adj_read(struct file *file, char __user *buf, size_t count,
1006 loff_t *ppos)
1007{
Al Viro496ad9a2013-01-23 17:07:38 -05001008 struct task_struct *task = get_proc_task(file_inode(file));
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001009 char buffer[PROC_NUMBUF];
1010 int oom_adj = OOM_ADJUST_MIN;
1011 size_t len;
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001012
1013 if (!task)
1014 return -ESRCH;
Michal Hockof913da52016-07-28 15:44:37 -07001015 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MAX)
1016 oom_adj = OOM_ADJUST_MAX;
1017 else
1018 oom_adj = (task->signal->oom_score_adj * -OOM_DISABLE) /
1019 OOM_SCORE_ADJ_MAX;
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001020 put_task_struct(task);
1021 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_adj);
1022 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1023}
1024
Michal Hocko1d5f0ac2016-07-28 15:44:40 -07001025static int __set_oom_adj(struct file *file, int oom_adj, bool legacy)
1026{
1027 static DEFINE_MUTEX(oom_adj_mutex);
Michal Hocko44a70ade2016-07-28 15:44:43 -07001028 struct mm_struct *mm = NULL;
Michal Hocko1d5f0ac2016-07-28 15:44:40 -07001029 struct task_struct *task;
1030 int err = 0;
1031
1032 task = get_proc_task(file_inode(file));
1033 if (!task)
1034 return -ESRCH;
1035
1036 mutex_lock(&oom_adj_mutex);
1037 if (legacy) {
1038 if (oom_adj < task->signal->oom_score_adj &&
1039 !capable(CAP_SYS_RESOURCE)) {
1040 err = -EACCES;
1041 goto err_unlock;
1042 }
1043 /*
1044 * /proc/pid/oom_adj is provided for legacy purposes, ask users to use
1045 * /proc/pid/oom_score_adj instead.
1046 */
1047 pr_warn_once("%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
1048 current->comm, task_pid_nr(current), task_pid_nr(task),
1049 task_pid_nr(task));
1050 } else {
1051 if ((short)oom_adj < task->signal->oom_score_adj_min &&
1052 !capable(CAP_SYS_RESOURCE)) {
1053 err = -EACCES;
1054 goto err_unlock;
1055 }
1056 }
1057
Michal Hocko44a70ade2016-07-28 15:44:43 -07001058 /*
1059 * Make sure we will check other processes sharing the mm if this is
1060 * not vfrok which wants its own oom_score_adj.
1061 * pin the mm so it doesn't go away and get reused after task_unlock
1062 */
1063 if (!task->vfork_done) {
1064 struct task_struct *p = find_lock_task_mm(task);
1065
1066 if (p) {
1067 if (atomic_read(&p->mm->mm_users) > 1) {
1068 mm = p->mm;
Vegard Nossumf1f10072017-02-27 14:30:07 -08001069 mmgrab(mm);
Michal Hocko44a70ade2016-07-28 15:44:43 -07001070 }
1071 task_unlock(p);
1072 }
1073 }
1074
Michal Hocko1d5f0ac2016-07-28 15:44:40 -07001075 task->signal->oom_score_adj = oom_adj;
1076 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1077 task->signal->oom_score_adj_min = (short)oom_adj;
1078 trace_oom_score_adj_update(task);
Michal Hocko44a70ade2016-07-28 15:44:43 -07001079
1080 if (mm) {
1081 struct task_struct *p;
1082
1083 rcu_read_lock();
1084 for_each_process(p) {
1085 if (same_thread_group(task, p))
1086 continue;
1087
1088 /* do not touch kernel threads or the global init */
1089 if (p->flags & PF_KTHREAD || is_global_init(p))
1090 continue;
1091
1092 task_lock(p);
1093 if (!p->vfork_done && process_shares_mm(p, mm)) {
1094 pr_info("updating oom_score_adj for %d (%s) from %d to %d because it shares mm with %d (%s). Report if this is unexpected.\n",
1095 task_pid_nr(p), p->comm,
1096 p->signal->oom_score_adj, oom_adj,
1097 task_pid_nr(task), task->comm);
1098 p->signal->oom_score_adj = oom_adj;
1099 if (!legacy && has_capability_noaudit(current, CAP_SYS_RESOURCE))
1100 p->signal->oom_score_adj_min = (short)oom_adj;
1101 }
1102 task_unlock(p);
1103 }
1104 rcu_read_unlock();
1105 mmdrop(mm);
1106 }
Michal Hocko1d5f0ac2016-07-28 15:44:40 -07001107err_unlock:
1108 mutex_unlock(&oom_adj_mutex);
1109 put_task_struct(task);
1110 return err;
1111}
Michal Hockof913da52016-07-28 15:44:37 -07001112
David Rientjesb72bdfa2015-11-05 18:50:32 -08001113/*
1114 * /proc/pid/oom_adj exists solely for backwards compatibility with previous
1115 * kernels. The effective policy is defined by oom_score_adj, which has a
1116 * different scale: oom_adj grew exponentially and oom_score_adj grows linearly.
1117 * Values written to oom_adj are simply mapped linearly to oom_score_adj.
1118 * Processes that become oom disabled via oom_adj will still be oom disabled
1119 * with this implementation.
1120 *
1121 * oom_adj cannot be removed since existing userspace binaries use it.
1122 */
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001123static ssize_t oom_adj_write(struct file *file, const char __user *buf,
1124 size_t count, loff_t *ppos)
1125{
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001126 char buffer[PROC_NUMBUF];
1127 int oom_adj;
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001128 int err;
1129
1130 memset(buffer, 0, sizeof(buffer));
1131 if (count > sizeof(buffer) - 1)
1132 count = sizeof(buffer) - 1;
1133 if (copy_from_user(buffer, buf, count)) {
1134 err = -EFAULT;
1135 goto out;
1136 }
1137
1138 err = kstrtoint(strstrip(buffer), 0, &oom_adj);
1139 if (err)
1140 goto out;
1141 if ((oom_adj < OOM_ADJUST_MIN || oom_adj > OOM_ADJUST_MAX) &&
1142 oom_adj != OOM_DISABLE) {
1143 err = -EINVAL;
1144 goto out;
1145 }
1146
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001147 /*
1148 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1149 * value is always attainable.
1150 */
1151 if (oom_adj == OOM_ADJUST_MAX)
1152 oom_adj = OOM_SCORE_ADJ_MAX;
1153 else
1154 oom_adj = (oom_adj * OOM_SCORE_ADJ_MAX) / -OOM_DISABLE;
1155
Michal Hocko1d5f0ac2016-07-28 15:44:40 -07001156 err = __set_oom_adj(file, oom_adj, true);
David Rientjesfa0cbbf2012-11-12 17:53:04 -08001157out:
1158 return err < 0 ? err : count;
1159}
1160
1161static const struct file_operations proc_oom_adj_operations = {
1162 .read = oom_adj_read,
1163 .write = oom_adj_write,
1164 .llseek = generic_file_llseek,
1165};
1166
David Rientjesa63d83f2010-08-09 17:19:46 -07001167static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1168 size_t count, loff_t *ppos)
1169{
Al Viro496ad9a2013-01-23 17:07:38 -05001170 struct task_struct *task = get_proc_task(file_inode(file));
David Rientjesa63d83f2010-08-09 17:19:46 -07001171 char buffer[PROC_NUMBUF];
David Rientjesa9c58b902012-12-11 16:02:54 -08001172 short oom_score_adj = OOM_SCORE_ADJ_MIN;
David Rientjesa63d83f2010-08-09 17:19:46 -07001173 size_t len;
1174
1175 if (!task)
1176 return -ESRCH;
Michal Hockof913da52016-07-28 15:44:37 -07001177 oom_score_adj = task->signal->oom_score_adj;
David Rientjesa63d83f2010-08-09 17:19:46 -07001178 put_task_struct(task);
David Rientjesa9c58b902012-12-11 16:02:54 -08001179 len = snprintf(buffer, sizeof(buffer), "%hd\n", oom_score_adj);
David Rientjesa63d83f2010-08-09 17:19:46 -07001180 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1181}
1182
1183static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1184 size_t count, loff_t *ppos)
1185{
David Rientjesa63d83f2010-08-09 17:19:46 -07001186 char buffer[PROC_NUMBUF];
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001187 int oom_score_adj;
David Rientjesa63d83f2010-08-09 17:19:46 -07001188 int err;
1189
1190 memset(buffer, 0, sizeof(buffer));
1191 if (count > sizeof(buffer) - 1)
1192 count = sizeof(buffer) - 1;
David Rientjes723548b2010-10-26 14:21:25 -07001193 if (copy_from_user(buffer, buf, count)) {
1194 err = -EFAULT;
1195 goto out;
1196 }
David Rientjesa63d83f2010-08-09 17:19:46 -07001197
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001198 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
David Rientjesa63d83f2010-08-09 17:19:46 -07001199 if (err)
David Rientjes723548b2010-10-26 14:21:25 -07001200 goto out;
David Rientjesa63d83f2010-08-09 17:19:46 -07001201 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
David Rientjes723548b2010-10-26 14:21:25 -07001202 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1203 err = -EINVAL;
1204 goto out;
1205 }
David Rientjesa63d83f2010-08-09 17:19:46 -07001206
Michal Hocko1d5f0ac2016-07-28 15:44:40 -07001207 err = __set_oom_adj(file, oom_score_adj, false);
David Rientjes723548b2010-10-26 14:21:25 -07001208out:
1209 return err < 0 ? err : count;
David Rientjesa63d83f2010-08-09 17:19:46 -07001210}
1211
1212static const struct file_operations proc_oom_score_adj_operations = {
1213 .read = oom_score_adj_read,
1214 .write = oom_score_adj_write,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001215 .llseek = default_llseek,
David Rientjesa63d83f2010-08-09 17:19:46 -07001216};
1217
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyanb4eb4f72016-10-29 19:04:39 +03001219#define TMPBUFLEN 11
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1221 size_t count, loff_t *ppos)
1222{
Al Viro496ad9a2013-01-23 17:07:38 -05001223 struct inode * inode = file_inode(file);
Eric W. Biederman99f89552006-06-26 00:25:55 -07001224 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225 ssize_t length;
1226 char tmpbuf[TMPBUFLEN];
1227
Eric W. Biederman99f89552006-06-26 00:25:55 -07001228 if (!task)
1229 return -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
Eric W. Biedermane1760bd2012-09-10 22:39:43 -07001231 from_kuid(file->f_cred->user_ns,
1232 audit_get_loginuid(task)));
Eric W. Biederman99f89552006-06-26 00:25:55 -07001233 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1235}
1236
1237static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1238 size_t count, loff_t *ppos)
1239{
Al Viro496ad9a2013-01-23 17:07:38 -05001240 struct inode * inode = file_inode(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241 uid_t loginuid;
Eric W. Biedermane1760bd2012-09-10 22:39:43 -07001242 kuid_t kloginuid;
Alexey Dobriyan774636e2015-09-09 15:36:59 -07001243 int rv;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244
Paul E. McKenney7dc52152010-02-22 17:04:52 -08001245 rcu_read_lock();
1246 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1247 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 return -EPERM;
Paul E. McKenney7dc52152010-02-22 17:04:52 -08001249 }
1250 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252 if (*ppos != 0) {
1253 /* No partial writes. */
1254 return -EINVAL;
1255 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256
Alexey Dobriyan774636e2015-09-09 15:36:59 -07001257 rv = kstrtou32_from_user(buf, count, 10, &loginuid);
1258 if (rv < 0)
1259 return rv;
Eric Paris81407c82013-05-24 09:49:14 -04001260
1261 /* is userspace tring to explicitly UNSET the loginuid? */
1262 if (loginuid == AUDIT_UID_UNSET) {
1263 kloginuid = INVALID_UID;
1264 } else {
1265 kloginuid = make_kuid(file->f_cred->user_ns, loginuid);
Alexey Dobriyan774636e2015-09-09 15:36:59 -07001266 if (!uid_valid(kloginuid))
1267 return -EINVAL;
Eric W. Biedermane1760bd2012-09-10 22:39:43 -07001268 }
1269
Alexey Dobriyan774636e2015-09-09 15:36:59 -07001270 rv = audit_set_loginuid(kloginuid);
1271 if (rv < 0)
1272 return rv;
1273 return count;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274}
1275
Arjan van de Ven00977a52007-02-12 00:55:34 -08001276static const struct file_operations proc_loginuid_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277 .read = proc_loginuid_read,
1278 .write = proc_loginuid_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001279 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280};
Eric Paris1e0bd752008-03-13 08:15:31 -04001281
1282static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1283 size_t count, loff_t *ppos)
1284{
Al Viro496ad9a2013-01-23 17:07:38 -05001285 struct inode * inode = file_inode(file);
Eric Paris1e0bd752008-03-13 08:15:31 -04001286 struct task_struct *task = get_proc_task(inode);
1287 ssize_t length;
1288 char tmpbuf[TMPBUFLEN];
1289
1290 if (!task)
1291 return -ESRCH;
1292 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1293 audit_get_sessionid(task));
1294 put_task_struct(task);
1295 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1296}
1297
1298static const struct file_operations proc_sessionid_operations = {
1299 .read = proc_sessionid_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001300 .llseek = generic_file_llseek,
Eric Paris1e0bd752008-03-13 08:15:31 -04001301};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302#endif
1303
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001304#ifdef CONFIG_FAULT_INJECTION
1305static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1306 size_t count, loff_t *ppos)
1307{
Al Viro496ad9a2013-01-23 17:07:38 -05001308 struct task_struct *task = get_proc_task(file_inode(file));
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001309 char buffer[PROC_NUMBUF];
1310 size_t len;
1311 int make_it_fail;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001312
1313 if (!task)
1314 return -ESRCH;
1315 make_it_fail = task->make_it_fail;
1316 put_task_struct(task);
1317
1318 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001319
1320 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001321}
1322
1323static ssize_t proc_fault_inject_write(struct file * file,
1324 const char __user * buf, size_t count, loff_t *ppos)
1325{
1326 struct task_struct *task;
Alexey Dobriyan774636e2015-09-09 15:36:59 -07001327 char buffer[PROC_NUMBUF];
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001328 int make_it_fail;
Alexey Dobriyan774636e2015-09-09 15:36:59 -07001329 int rv;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001330
1331 if (!capable(CAP_SYS_RESOURCE))
1332 return -EPERM;
1333 memset(buffer, 0, sizeof(buffer));
1334 if (count > sizeof(buffer) - 1)
1335 count = sizeof(buffer) - 1;
1336 if (copy_from_user(buffer, buf, count))
1337 return -EFAULT;
Alexey Dobriyan774636e2015-09-09 15:36:59 -07001338 rv = kstrtoint(strstrip(buffer), 0, &make_it_fail);
1339 if (rv < 0)
1340 return rv;
Dave Jones16caed32014-04-07 15:39:15 -07001341 if (make_it_fail < 0 || make_it_fail > 1)
1342 return -EINVAL;
1343
Al Viro496ad9a2013-01-23 17:07:38 -05001344 task = get_proc_task(file_inode(file));
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001345 if (!task)
1346 return -ESRCH;
1347 task->make_it_fail = make_it_fail;
1348 put_task_struct(task);
Vincent Licba8aaf2009-09-22 16:45:38 -07001349
1350 return count;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001351}
1352
Arjan van de Ven00977a52007-02-12 00:55:34 -08001353static const struct file_operations proc_fault_inject_operations = {
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001354 .read = proc_fault_inject_read,
1355 .write = proc_fault_inject_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001356 .llseek = generic_file_llseek,
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001357};
Dmitry Vyukove41d58182017-07-12 14:34:35 -07001358
1359static ssize_t proc_fail_nth_write(struct file *file, const char __user *buf,
1360 size_t count, loff_t *ppos)
1361{
1362 struct task_struct *task;
Akinobu Mita9049f2f2017-07-14 14:49:52 -07001363 int err;
1364 unsigned int n;
Dmitry Vyukove41d58182017-07-12 14:34:35 -07001365
1366 task = get_proc_task(file_inode(file));
1367 if (!task)
1368 return -ESRCH;
1369 put_task_struct(task);
1370 if (task != current)
1371 return -EPERM;
Akinobu Mita9049f2f2017-07-14 14:49:52 -07001372 err = kstrtouint_from_user(buf, count, 0, &n);
Dmitry Vyukove41d58182017-07-12 14:34:35 -07001373 if (err)
1374 return err;
Akinobu Mita9049f2f2017-07-14 14:49:52 -07001375 current->fail_nth = n;
Dmitry Vyukove41d58182017-07-12 14:34:35 -07001376 return count;
1377}
1378
1379static ssize_t proc_fail_nth_read(struct file *file, char __user *buf,
1380 size_t count, loff_t *ppos)
1381{
1382 struct task_struct *task;
1383 int err;
1384
1385 task = get_proc_task(file_inode(file));
1386 if (!task)
1387 return -ESRCH;
1388 put_task_struct(task);
1389 if (task != current)
1390 return -EPERM;
1391 if (count < 1)
1392 return -EINVAL;
1393 err = put_user((char)(current->fail_nth ? 'N' : 'Y'), buf);
1394 if (err)
1395 return err;
1396 current->fail_nth = 0;
1397 return 1;
1398}
1399
1400static const struct file_operations proc_fail_nth_operations = {
1401 .read = proc_fail_nth_read,
1402 .write = proc_fail_nth_write,
1403};
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001404#endif
1405
Arjan van de Ven97455122008-01-25 21:08:34 +01001406
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001407#ifdef CONFIG_SCHED_DEBUG
1408/*
1409 * Print out various scheduling related per-task fields:
1410 */
1411static int sched_show(struct seq_file *m, void *v)
1412{
1413 struct inode *inode = m->private;
1414 struct task_struct *p;
1415
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001416 p = get_proc_task(inode);
1417 if (!p)
1418 return -ESRCH;
1419 proc_sched_show_task(p, m);
1420
1421 put_task_struct(p);
1422
1423 return 0;
1424}
1425
1426static ssize_t
1427sched_write(struct file *file, const char __user *buf,
1428 size_t count, loff_t *offset)
1429{
Al Viro496ad9a2013-01-23 17:07:38 -05001430 struct inode *inode = file_inode(file);
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001431 struct task_struct *p;
1432
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001433 p = get_proc_task(inode);
1434 if (!p)
1435 return -ESRCH;
1436 proc_sched_set_task(p);
1437
1438 put_task_struct(p);
1439
1440 return count;
1441}
1442
1443static int sched_open(struct inode *inode, struct file *filp)
1444{
Jovi Zhangc6a34052011-01-12 17:00:34 -08001445 return single_open(filp, sched_show, inode);
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001446}
1447
1448static const struct file_operations proc_pid_sched_operations = {
1449 .open = sched_open,
1450 .read = seq_read,
1451 .write = sched_write,
1452 .llseek = seq_lseek,
Alexey Dobriyan5ea473a2007-07-31 00:38:50 -07001453 .release = single_release,
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001454};
1455
1456#endif
1457
Mike Galbraith5091faa2010-11-30 14:18:03 +01001458#ifdef CONFIG_SCHED_AUTOGROUP
1459/*
1460 * Print out autogroup related information:
1461 */
1462static int sched_autogroup_show(struct seq_file *m, void *v)
1463{
1464 struct inode *inode = m->private;
1465 struct task_struct *p;
1466
1467 p = get_proc_task(inode);
1468 if (!p)
1469 return -ESRCH;
1470 proc_sched_autogroup_show_task(p, m);
1471
1472 put_task_struct(p);
1473
1474 return 0;
1475}
1476
1477static ssize_t
1478sched_autogroup_write(struct file *file, const char __user *buf,
1479 size_t count, loff_t *offset)
1480{
Al Viro496ad9a2013-01-23 17:07:38 -05001481 struct inode *inode = file_inode(file);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001482 struct task_struct *p;
1483 char buffer[PROC_NUMBUF];
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001484 int nice;
Mike Galbraith5091faa2010-11-30 14:18:03 +01001485 int err;
1486
1487 memset(buffer, 0, sizeof(buffer));
1488 if (count > sizeof(buffer) - 1)
1489 count = sizeof(buffer) - 1;
1490 if (copy_from_user(buffer, buf, count))
1491 return -EFAULT;
1492
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001493 err = kstrtoint(strstrip(buffer), 0, &nice);
1494 if (err < 0)
1495 return err;
Mike Galbraith5091faa2010-11-30 14:18:03 +01001496
1497 p = get_proc_task(inode);
1498 if (!p)
1499 return -ESRCH;
1500
Hiroshi Shimamoto2e5b5b32012-02-23 17:41:27 +09001501 err = proc_sched_autogroup_set_nice(p, nice);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001502 if (err)
1503 count = err;
1504
1505 put_task_struct(p);
1506
1507 return count;
1508}
1509
1510static int sched_autogroup_open(struct inode *inode, struct file *filp)
1511{
1512 int ret;
1513
1514 ret = single_open(filp, sched_autogroup_show, NULL);
1515 if (!ret) {
1516 struct seq_file *m = filp->private_data;
1517
1518 m->private = inode;
1519 }
1520 return ret;
1521}
1522
1523static const struct file_operations proc_pid_sched_autogroup_operations = {
1524 .open = sched_autogroup_open,
1525 .read = seq_read,
1526 .write = sched_autogroup_write,
1527 .llseek = seq_lseek,
1528 .release = single_release,
1529};
1530
1531#endif /* CONFIG_SCHED_AUTOGROUP */
1532
john stultz4614a696b2009-12-14 18:00:05 -08001533static ssize_t comm_write(struct file *file, const char __user *buf,
1534 size_t count, loff_t *offset)
1535{
Al Viro496ad9a2013-01-23 17:07:38 -05001536 struct inode *inode = file_inode(file);
john stultz4614a696b2009-12-14 18:00:05 -08001537 struct task_struct *p;
1538 char buffer[TASK_COMM_LEN];
David Rientjes830e0fc2013-04-30 15:28:18 -07001539 const size_t maxlen = sizeof(buffer) - 1;
john stultz4614a696b2009-12-14 18:00:05 -08001540
1541 memset(buffer, 0, sizeof(buffer));
David Rientjes830e0fc2013-04-30 15:28:18 -07001542 if (copy_from_user(buffer, buf, count > maxlen ? maxlen : count))
john stultz4614a696b2009-12-14 18:00:05 -08001543 return -EFAULT;
1544
1545 p = get_proc_task(inode);
1546 if (!p)
1547 return -ESRCH;
1548
1549 if (same_thread_group(current, p))
1550 set_task_comm(p, buffer);
1551 else
1552 count = -EINVAL;
1553
1554 put_task_struct(p);
1555
1556 return count;
1557}
1558
1559static int comm_show(struct seq_file *m, void *v)
1560{
1561 struct inode *inode = m->private;
1562 struct task_struct *p;
1563
1564 p = get_proc_task(inode);
1565 if (!p)
1566 return -ESRCH;
1567
1568 task_lock(p);
1569 seq_printf(m, "%s\n", p->comm);
1570 task_unlock(p);
1571
1572 put_task_struct(p);
1573
1574 return 0;
1575}
1576
1577static int comm_open(struct inode *inode, struct file *filp)
1578{
Jovi Zhangc6a34052011-01-12 17:00:34 -08001579 return single_open(filp, comm_show, inode);
john stultz4614a696b2009-12-14 18:00:05 -08001580}
1581
1582static const struct file_operations proc_pid_set_comm_operations = {
1583 .open = comm_open,
1584 .read = seq_read,
1585 .write = comm_write,
1586 .llseek = seq_lseek,
1587 .release = single_release,
1588};
1589
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -08001590static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
Matt Helsley925d1c42008-04-29 01:01:36 -07001591{
1592 struct task_struct *task;
Matt Helsley925d1c42008-04-29 01:01:36 -07001593 struct file *exe_file;
1594
David Howells2b0143b2015-03-17 22:25:59 +00001595 task = get_proc_task(d_inode(dentry));
Matt Helsley925d1c42008-04-29 01:01:36 -07001596 if (!task)
1597 return -ENOENT;
Mateusz Guzikcd81a9172016-08-23 16:20:38 +02001598 exe_file = get_task_exe_file(task);
Matt Helsley925d1c42008-04-29 01:01:36 -07001599 put_task_struct(task);
Matt Helsley925d1c42008-04-29 01:01:36 -07001600 if (exe_file) {
1601 *exe_path = exe_file->f_path;
1602 path_get(&exe_file->f_path);
1603 fput(exe_file);
1604 return 0;
1605 } else
1606 return -ENOENT;
1607}
1608
Al Viro6b255392015-11-17 10:20:54 -05001609static const char *proc_pid_get_link(struct dentry *dentry,
Al Virofceef392015-12-29 15:58:39 -05001610 struct inode *inode,
1611 struct delayed_call *done)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612{
Christoph Hellwig408ef012012-06-18 10:47:03 -04001613 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614 int error = -EACCES;
1615
Al Viro6b255392015-11-17 10:20:54 -05001616 if (!dentry)
1617 return ERR_PTR(-ECHILD);
1618
Eric W. Biederman778c1142006-06-26 00:25:58 -07001619 /* Are we allowed to snoop on the tasks file descriptors? */
1620 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622
Christoph Hellwig408ef012012-06-18 10:47:03 -04001623 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1624 if (error)
1625 goto out;
1626
Al Viro6e77137b2015-05-02 13:37:52 -04001627 nd_jump_link(&path);
Christoph Hellwig408ef012012-06-18 10:47:03 -04001628 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629out:
Al Viro008b1502005-08-20 00:17:39 +01001630 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631}
1632
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001633static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634{
Mel Gormane12ba742007-10-16 01:25:52 -07001635 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001636 char *pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637 int len;
1638
1639 if (!tmp)
1640 return -ENOMEM;
Akinobu Mita0c28f282007-05-08 00:31:41 -07001641
Eric W. Biederman7b2a69b2010-12-05 15:51:21 -08001642 pathname = d_path(path, tmp, PAGE_SIZE);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001643 len = PTR_ERR(pathname);
1644 if (IS_ERR(pathname))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 goto out;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001646 len = tmp + PAGE_SIZE - 1 - pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647
1648 if (len > buflen)
1649 len = buflen;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001650 if (copy_to_user(buffer, pathname, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 len = -EFAULT;
1652 out:
1653 free_page((unsigned long)tmp);
1654 return len;
1655}
1656
1657static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1658{
1659 int error = -EACCES;
David Howells2b0143b2015-03-17 22:25:59 +00001660 struct inode *inode = d_inode(dentry);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001661 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662
Eric W. Biederman778c1142006-06-26 00:25:58 -07001663 /* Are we allowed to snoop on the tasks file descriptors? */
1664 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -08001667 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668 if (error)
1669 goto out;
1670
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001671 error = do_proc_readlink(&path, buffer, buflen);
1672 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 return error;
1675}
1676
Cyrill Gorcunovfaf60af2012-08-23 14:43:24 +04001677const struct inode_operations proc_pid_link_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 .readlink = proc_pid_readlink,
Al Viro6b255392015-11-17 10:20:54 -05001679 .get_link = proc_pid_get_link,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001680 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681};
1682
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001683
1684/* building an inode */
1685
Eric W. Biederman68eb94f2017-01-03 10:23:11 +13001686void task_dump_owner(struct task_struct *task, mode_t mode,
1687 kuid_t *ruid, kgid_t *rgid)
1688{
1689 /* Depending on the state of dumpable compute who should own a
1690 * proc file for a task.
1691 */
1692 const struct cred *cred;
1693 kuid_t uid;
1694 kgid_t gid;
1695
1696 /* Default to the tasks effective ownership */
1697 rcu_read_lock();
1698 cred = __task_cred(task);
1699 uid = cred->euid;
1700 gid = cred->egid;
1701 rcu_read_unlock();
1702
1703 /*
1704 * Before the /proc/pid/status file was created the only way to read
1705 * the effective uid of a /process was to stat /proc/pid. Reading
1706 * /proc/pid/status is slow enough that procps and other packages
1707 * kept stating /proc/pid. To keep the rules in /proc simple I have
1708 * made this apply to all per process world readable and executable
1709 * directories.
1710 */
1711 if (mode != (S_IFDIR|S_IRUGO|S_IXUGO)) {
1712 struct mm_struct *mm;
1713 task_lock(task);
1714 mm = task->mm;
1715 /* Make non-dumpable tasks owned by some root */
1716 if (mm) {
1717 if (get_dumpable(mm) != SUID_DUMP_USER) {
1718 struct user_namespace *user_ns = mm->user_ns;
1719
1720 uid = make_kuid(user_ns, 0);
1721 if (!uid_valid(uid))
1722 uid = GLOBAL_ROOT_UID;
1723
1724 gid = make_kgid(user_ns, 0);
1725 if (!gid_valid(gid))
1726 gid = GLOBAL_ROOT_GID;
1727 }
1728 } else {
1729 uid = GLOBAL_ROOT_UID;
1730 gid = GLOBAL_ROOT_GID;
1731 }
1732 task_unlock(task);
1733 }
1734 *ruid = uid;
1735 *rgid = gid;
1736}
1737
Andreas Gruenbacherdb978da2016-11-10 22:18:28 +01001738struct inode *proc_pid_make_inode(struct super_block * sb,
1739 struct task_struct *task, umode_t mode)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001740{
1741 struct inode * inode;
1742 struct proc_inode *ei;
1743
1744 /* We need a new inode */
1745
1746 inode = new_inode(sb);
1747 if (!inode)
1748 goto out;
1749
1750 /* Common stuff */
1751 ei = PROC_I(inode);
Andreas Gruenbacherdb978da2016-11-10 22:18:28 +01001752 inode->i_mode = mode;
Christoph Hellwig85fe4022010-10-23 11:19:54 -04001753 inode->i_ino = get_next_ino();
Deepa Dinamani078cd822016-09-14 07:48:04 -07001754 inode->i_mtime = inode->i_atime = inode->i_ctime = current_time(inode);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001755 inode->i_op = &proc_def_inode_operations;
1756
1757 /*
1758 * grab the reference to task.
1759 */
Oleg Nesterov1a657f782006-10-02 02:18:59 -07001760 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001761 if (!ei->pid)
1762 goto out_unlock;
1763
Eric W. Biederman68eb94f2017-01-03 10:23:11 +13001764 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001765 security_task_to_inode(task, inode);
1766
1767out:
1768 return inode;
1769
1770out_unlock:
1771 iput(inode);
1772 return NULL;
1773}
1774
David Howellsa528d352017-01-31 16:46:22 +00001775int pid_getattr(const struct path *path, struct kstat *stat,
1776 u32 request_mask, unsigned int query_flags)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001777{
David Howellsa528d352017-01-31 16:46:22 +00001778 struct inode *inode = d_inode(path->dentry);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001779 struct task_struct *task;
David Howellsa528d352017-01-31 16:46:22 +00001780 struct pid_namespace *pid = path->dentry->d_sb->s_fs_info;
David Howellsc69e8d92008-11-14 10:39:19 +11001781
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001782 generic_fillattr(inode, stat);
1783
1784 rcu_read_lock();
Eric W. Biedermandcb0f222012-02-09 08:48:21 -08001785 stat->uid = GLOBAL_ROOT_UID;
1786 stat->gid = GLOBAL_ROOT_GID;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001787 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1788 if (task) {
Lafcadio Wluiki796f5712017-02-24 15:00:23 -08001789 if (!has_pid_permissions(pid, task, HIDEPID_INVISIBLE)) {
Vasiliy Kulikov04996802012-01-10 15:11:31 -08001790 rcu_read_unlock();
1791 /*
1792 * This doesn't prevent learning whether PID exists,
1793 * it only makes getattr() consistent with readdir().
1794 */
1795 return -ENOENT;
1796 }
Eric W. Biederman68eb94f2017-01-03 10:23:11 +13001797 task_dump_owner(task, inode->i_mode, &stat->uid, &stat->gid);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001798 }
1799 rcu_read_unlock();
1800 return 0;
1801}
1802
1803/* dentry stuff */
1804
1805/*
1806 * Exceptional case: normally we are not allowed to unhash a busy
1807 * directory. In this case, however, we can do it - no aliasing problems
1808 * due to the way we treat inodes.
1809 *
1810 * Rewrite the inode's ownerships here because the owning task may have
1811 * performed a setuid(), etc.
1812 *
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001813 */
Al Viro0b728e12012-06-10 16:03:43 -04001814int pid_revalidate(struct dentry *dentry, unsigned int flags)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001815{
Nick Piggin34286d62011-01-07 17:49:57 +11001816 struct inode *inode;
1817 struct task_struct *task;
David Howellsc69e8d92008-11-14 10:39:19 +11001818
Al Viro0b728e12012-06-10 16:03:43 -04001819 if (flags & LOOKUP_RCU)
Nick Piggin34286d62011-01-07 17:49:57 +11001820 return -ECHILD;
1821
David Howells2b0143b2015-03-17 22:25:59 +00001822 inode = d_inode(dentry);
Nick Piggin34286d62011-01-07 17:49:57 +11001823 task = get_proc_task(inode);
1824
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001825 if (task) {
Eric W. Biederman68eb94f2017-01-03 10:23:11 +13001826 task_dump_owner(task, inode->i_mode, &inode->i_uid, &inode->i_gid);
1827
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001828 inode->i_mode &= ~(S_ISUID | S_ISGID);
1829 security_task_to_inode(task, inode);
1830 put_task_struct(task);
1831 return 1;
1832 }
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001833 return 0;
1834}
1835
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08001836static inline bool proc_inode_is_dead(struct inode *inode)
1837{
1838 return !proc_pid(inode)->tasks[PIDTYPE_PID].first;
1839}
1840
David Howells1dd704b2013-04-12 01:08:50 +01001841int pid_delete_dentry(const struct dentry *dentry)
1842{
1843 /* Is the task we represent dead?
1844 * If so, then don't put the dentry on the lru list,
1845 * kill it immediately.
1846 */
David Howells2b0143b2015-03-17 22:25:59 +00001847 return proc_inode_is_dead(d_inode(dentry));
David Howells1dd704b2013-04-12 01:08:50 +01001848}
1849
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001850const struct dentry_operations pid_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001851{
1852 .d_revalidate = pid_revalidate,
1853 .d_delete = pid_delete_dentry,
1854};
1855
1856/* Lookups */
1857
Eric W. Biederman1c0d04c2006-10-02 02:18:57 -07001858/*
1859 * Fill a directory entry.
1860 *
1861 * If possible create the dcache entry and derive our inode number and
1862 * file type from dcache entry.
1863 *
1864 * Since all of the proc inode numbers are dynamically generated, the inode
1865 * numbers do not exist until the inode is cache. This means creating the
1866 * the dcache entry in readdir is necessary to keep the inode numbers
1867 * reported by readdir in sync with the inode numbers reported
1868 * by stat.
1869 */
Al Virof0c3b502013-05-16 12:07:31 -04001870bool proc_fill_cache(struct file *file, struct dir_context *ctx,
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001871 const char *name, int len,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001872 instantiate_t instantiate, struct task_struct *task, const void *ptr)
Eric W. Biederman61a28782006-10-02 02:18:49 -07001873{
Al Virof0c3b502013-05-16 12:07:31 -04001874 struct dentry *child, *dir = file->f_path.dentry;
Al Viro1df98b82013-06-15 11:33:10 +04001875 struct qstr qname = QSTR_INIT(name, len);
Eric W. Biederman61a28782006-10-02 02:18:49 -07001876 struct inode *inode;
Al Viro1df98b82013-06-15 11:33:10 +04001877 unsigned type;
1878 ino_t ino;
Eric W. Biederman61a28782006-10-02 02:18:49 -07001879
Al Viro1df98b82013-06-15 11:33:10 +04001880 child = d_hash_and_lookup(dir, &qname);
Eric W. Biederman61a28782006-10-02 02:18:49 -07001881 if (!child) {
Al Viro37817642016-04-20 16:31:31 -04001882 DECLARE_WAIT_QUEUE_HEAD_ONSTACK(wq);
1883 child = d_alloc_parallel(dir, &qname, &wq);
1884 if (IS_ERR(child))
Al Viro1df98b82013-06-15 11:33:10 +04001885 goto end_instantiate;
Al Viro37817642016-04-20 16:31:31 -04001886 if (d_in_lookup(child)) {
1887 int err = instantiate(d_inode(dir), child, task, ptr);
1888 d_lookup_done(child);
1889 if (err < 0) {
1890 dput(child);
1891 goto end_instantiate;
1892 }
Eric W. Biederman61a28782006-10-02 02:18:49 -07001893 }
1894 }
David Howells2b0143b2015-03-17 22:25:59 +00001895 inode = d_inode(child);
Al Viro147ce692013-06-15 10:26:35 +04001896 ino = inode->i_ino;
1897 type = inode->i_mode >> 12;
Eric W. Biederman61a28782006-10-02 02:18:49 -07001898 dput(child);
Al Virof0c3b502013-05-16 12:07:31 -04001899 return dir_emit(ctx, name, len, ino, type);
Al Viro1df98b82013-06-15 11:33:10 +04001900
1901end_instantiate:
1902 return dir_emit(ctx, name, len, 1, DT_UNKNOWN);
Eric W. Biederman61a28782006-10-02 02:18:49 -07001903}
1904
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001905/*
1906 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1907 * which represent vma start and end addresses.
1908 */
1909static int dname_to_vma_addr(struct dentry *dentry,
1910 unsigned long *start, unsigned long *end)
1911{
1912 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
1913 return -EINVAL;
1914
1915 return 0;
1916}
1917
Al Viro0b728e12012-06-10 16:03:43 -04001918static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001919{
1920 unsigned long vm_start, vm_end;
1921 bool exact_vma_exists = false;
1922 struct mm_struct *mm = NULL;
1923 struct task_struct *task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001924 struct inode *inode;
1925 int status = 0;
1926
Al Viro0b728e12012-06-10 16:03:43 -04001927 if (flags & LOOKUP_RCU)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001928 return -ECHILD;
1929
David Howells2b0143b2015-03-17 22:25:59 +00001930 inode = d_inode(dentry);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001931 task = get_proc_task(inode);
1932 if (!task)
1933 goto out_notask;
1934
Jann Horncaaee622016-01-20 15:00:04 -08001935 mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
Cong Wang2344bec2012-05-31 16:26:18 -07001936 if (IS_ERR_OR_NULL(mm))
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001937 goto out;
1938
1939 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
1940 down_read(&mm->mmap_sem);
1941 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
1942 up_read(&mm->mmap_sem);
1943 }
1944
1945 mmput(mm);
1946
1947 if (exact_vma_exists) {
Eric W. Biederman68eb94f2017-01-03 10:23:11 +13001948 task_dump_owner(task, 0, &inode->i_uid, &inode->i_gid);
1949
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001950 security_task_to_inode(task, inode);
1951 status = 1;
1952 }
1953
1954out:
1955 put_task_struct(task);
1956
1957out_notask:
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001958 return status;
1959}
1960
1961static const struct dentry_operations tid_map_files_dentry_operations = {
1962 .d_revalidate = map_files_d_revalidate,
1963 .d_delete = pid_delete_dentry,
1964};
1965
Al Viro6b255392015-11-17 10:20:54 -05001966static int map_files_get_link(struct dentry *dentry, struct path *path)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001967{
1968 unsigned long vm_start, vm_end;
1969 struct vm_area_struct *vma;
1970 struct task_struct *task;
1971 struct mm_struct *mm;
1972 int rc;
1973
1974 rc = -ENOENT;
David Howells2b0143b2015-03-17 22:25:59 +00001975 task = get_proc_task(d_inode(dentry));
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001976 if (!task)
1977 goto out;
1978
1979 mm = get_task_mm(task);
1980 put_task_struct(task);
1981 if (!mm)
1982 goto out;
1983
1984 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
1985 if (rc)
1986 goto out_mmput;
1987
Artem Fetishev70335ab2014-03-10 15:49:45 -07001988 rc = -ENOENT;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001989 down_read(&mm->mmap_sem);
1990 vma = find_exact_vma(mm, vm_start, vm_end);
1991 if (vma && vma->vm_file) {
1992 *path = vma->vm_file->f_path;
1993 path_get(path);
1994 rc = 0;
1995 }
1996 up_read(&mm->mmap_sem);
1997
1998out_mmput:
1999 mmput(mm);
2000out:
2001 return rc;
2002}
2003
2004struct map_files_info {
Al Viro7b540d02012-08-27 14:55:26 -04002005 fmode_t mode;
Alexey Dobriyan9a87fe02016-12-12 16:45:11 -08002006 unsigned int len;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002007 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2008};
2009
Calvin Owensbdb4d102015-09-09 15:35:54 -07002010/*
2011 * Only allow CAP_SYS_ADMIN to follow the links, due to concerns about how the
2012 * symlinks may be used to bypass permissions on ancestor directories in the
2013 * path to the file in question.
2014 */
2015static const char *
Al Viro6b255392015-11-17 10:20:54 -05002016proc_map_files_get_link(struct dentry *dentry,
Al Virofceef392015-12-29 15:58:39 -05002017 struct inode *inode,
2018 struct delayed_call *done)
Calvin Owensbdb4d102015-09-09 15:35:54 -07002019{
2020 if (!capable(CAP_SYS_ADMIN))
2021 return ERR_PTR(-EPERM);
2022
Al Virofceef392015-12-29 15:58:39 -05002023 return proc_pid_get_link(dentry, inode, done);
Calvin Owensbdb4d102015-09-09 15:35:54 -07002024}
2025
2026/*
Al Viro6b255392015-11-17 10:20:54 -05002027 * Identical to proc_pid_link_inode_operations except for get_link()
Calvin Owensbdb4d102015-09-09 15:35:54 -07002028 */
2029static const struct inode_operations proc_map_files_link_inode_operations = {
2030 .readlink = proc_pid_readlink,
Al Viro6b255392015-11-17 10:20:54 -05002031 .get_link = proc_map_files_get_link,
Calvin Owensbdb4d102015-09-09 15:35:54 -07002032 .setattr = proc_setattr,
2033};
2034
Al Viroc52a47a2013-06-15 11:15:20 +04002035static int
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002036proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2037 struct task_struct *task, const void *ptr)
2038{
Al Viro7b540d02012-08-27 14:55:26 -04002039 fmode_t mode = (fmode_t)(unsigned long)ptr;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002040 struct proc_inode *ei;
2041 struct inode *inode;
2042
Andreas Gruenbacherdb978da2016-11-10 22:18:28 +01002043 inode = proc_pid_make_inode(dir->i_sb, task, S_IFLNK |
2044 ((mode & FMODE_READ ) ? S_IRUSR : 0) |
2045 ((mode & FMODE_WRITE) ? S_IWUSR : 0));
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002046 if (!inode)
Al Viroc52a47a2013-06-15 11:15:20 +04002047 return -ENOENT;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002048
2049 ei = PROC_I(inode);
Al Viro6b255392015-11-17 10:20:54 -05002050 ei->op.proc_get_link = map_files_get_link;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002051
Calvin Owensbdb4d102015-09-09 15:35:54 -07002052 inode->i_op = &proc_map_files_link_inode_operations;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002053 inode->i_size = 64;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002054
2055 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2056 d_add(dentry, inode);
2057
Al Viroc52a47a2013-06-15 11:15:20 +04002058 return 0;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002059}
2060
2061static struct dentry *proc_map_files_lookup(struct inode *dir,
Al Viro00cd8dd2012-06-10 17:13:09 -04002062 struct dentry *dentry, unsigned int flags)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002063{
2064 unsigned long vm_start, vm_end;
2065 struct vm_area_struct *vma;
2066 struct task_struct *task;
Al Viroc52a47a2013-06-15 11:15:20 +04002067 int result;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002068 struct mm_struct *mm;
2069
Al Viroc52a47a2013-06-15 11:15:20 +04002070 result = -ENOENT;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002071 task = get_proc_task(dir);
2072 if (!task)
2073 goto out;
2074
Al Viroc52a47a2013-06-15 11:15:20 +04002075 result = -EACCES;
Jann Horncaaee622016-01-20 15:00:04 -08002076 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002077 goto out_put_task;
2078
Al Viroc52a47a2013-06-15 11:15:20 +04002079 result = -ENOENT;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002080 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07002081 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002082
2083 mm = get_task_mm(task);
2084 if (!mm)
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07002085 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002086
2087 down_read(&mm->mmap_sem);
2088 vma = find_exact_vma(mm, vm_start, vm_end);
2089 if (!vma)
2090 goto out_no_vma;
2091
Stanislav Kinsbursky05f56482012-11-26 16:29:42 -08002092 if (vma->vm_file)
2093 result = proc_map_files_instantiate(dir, dentry, task,
2094 (void *)(unsigned long)vma->vm_file->f_mode);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002095
2096out_no_vma:
2097 up_read(&mm->mmap_sem);
2098 mmput(mm);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002099out_put_task:
2100 put_task_struct(task);
2101out:
Al Viroc52a47a2013-06-15 11:15:20 +04002102 return ERR_PTR(result);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002103}
2104
2105static const struct inode_operations proc_map_files_inode_operations = {
2106 .lookup = proc_map_files_lookup,
2107 .permission = proc_fd_permission,
2108 .setattr = proc_setattr,
2109};
2110
2111static int
Al Virof0c3b502013-05-16 12:07:31 -04002112proc_map_files_readdir(struct file *file, struct dir_context *ctx)
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002113{
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002114 struct vm_area_struct *vma;
2115 struct task_struct *task;
2116 struct mm_struct *mm;
Al Virof0c3b502013-05-16 12:07:31 -04002117 unsigned long nr_files, pos, i;
2118 struct flex_array *fa = NULL;
2119 struct map_files_info info;
2120 struct map_files_info *p;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002121 int ret;
2122
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002123 ret = -ENOENT;
Al Virof0c3b502013-05-16 12:07:31 -04002124 task = get_proc_task(file_inode(file));
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002125 if (!task)
2126 goto out;
2127
2128 ret = -EACCES;
Jann Horncaaee622016-01-20 15:00:04 -08002129 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS))
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002130 goto out_put_task;
2131
2132 ret = 0;
Al Virof0c3b502013-05-16 12:07:31 -04002133 if (!dir_emit_dots(file, ctx))
2134 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002135
Al Virof0c3b502013-05-16 12:07:31 -04002136 mm = get_task_mm(task);
2137 if (!mm)
2138 goto out_put_task;
2139 down_read(&mm->mmap_sem);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002140
Al Virof0c3b502013-05-16 12:07:31 -04002141 nr_files = 0;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002142
Al Virof0c3b502013-05-16 12:07:31 -04002143 /*
2144 * We need two passes here:
2145 *
2146 * 1) Collect vmas of mapped files with mmap_sem taken
2147 * 2) Release mmap_sem and instantiate entries
2148 *
2149 * otherwise we get lockdep complained, since filldir()
2150 * routine might require mmap_sem taken in might_fault().
2151 */
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002152
Al Virof0c3b502013-05-16 12:07:31 -04002153 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2154 if (vma->vm_file && ++pos > ctx->pos)
2155 nr_files++;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002156 }
Al Virof0c3b502013-05-16 12:07:31 -04002157
2158 if (nr_files) {
2159 fa = flex_array_alloc(sizeof(info), nr_files,
2160 GFP_KERNEL);
2161 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2162 GFP_KERNEL)) {
2163 ret = -ENOMEM;
2164 if (fa)
2165 flex_array_free(fa);
2166 up_read(&mm->mmap_sem);
2167 mmput(mm);
2168 goto out_put_task;
2169 }
2170 for (i = 0, vma = mm->mmap, pos = 2; vma;
2171 vma = vma->vm_next) {
2172 if (!vma->vm_file)
2173 continue;
2174 if (++pos <= ctx->pos)
2175 continue;
2176
2177 info.mode = vma->vm_file->f_mode;
2178 info.len = snprintf(info.name,
2179 sizeof(info.name), "%lx-%lx",
2180 vma->vm_start, vma->vm_end);
2181 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2182 BUG();
2183 }
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002184 }
Al Virof0c3b502013-05-16 12:07:31 -04002185 up_read(&mm->mmap_sem);
2186
2187 for (i = 0; i < nr_files; i++) {
2188 p = flex_array_get(fa, i);
2189 if (!proc_fill_cache(file, ctx,
2190 p->name, p->len,
2191 proc_map_files_instantiate,
2192 task,
2193 (void *)(unsigned long)p->mode))
2194 break;
2195 ctx->pos++;
2196 }
2197 if (fa)
2198 flex_array_free(fa);
2199 mmput(mm);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002200
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002201out_put_task:
2202 put_task_struct(task);
2203out:
2204 return ret;
2205}
2206
2207static const struct file_operations proc_map_files_operations = {
2208 .read = generic_read_dir,
Al Virof50752e2016-04-20 17:13:54 -04002209 .iterate_shared = proc_map_files_readdir,
2210 .llseek = generic_file_llseek,
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002211};
2212
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05002213#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002214struct timers_private {
2215 struct pid *pid;
2216 struct task_struct *task;
2217 struct sighand_struct *sighand;
Pavel Emelyanov57b80152013-03-11 13:13:08 +04002218 struct pid_namespace *ns;
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002219 unsigned long flags;
2220};
2221
2222static void *timers_start(struct seq_file *m, loff_t *pos)
2223{
2224 struct timers_private *tp = m->private;
2225
2226 tp->task = get_pid_task(tp->pid, PIDTYPE_PID);
2227 if (!tp->task)
2228 return ERR_PTR(-ESRCH);
2229
2230 tp->sighand = lock_task_sighand(tp->task, &tp->flags);
2231 if (!tp->sighand)
2232 return ERR_PTR(-ESRCH);
2233
2234 return seq_list_start(&tp->task->signal->posix_timers, *pos);
2235}
2236
2237static void *timers_next(struct seq_file *m, void *v, loff_t *pos)
2238{
2239 struct timers_private *tp = m->private;
2240 return seq_list_next(v, &tp->task->signal->posix_timers, pos);
2241}
2242
2243static void timers_stop(struct seq_file *m, void *v)
2244{
2245 struct timers_private *tp = m->private;
2246
2247 if (tp->sighand) {
2248 unlock_task_sighand(tp->task, &tp->flags);
2249 tp->sighand = NULL;
2250 }
2251
2252 if (tp->task) {
2253 put_task_struct(tp->task);
2254 tp->task = NULL;
2255 }
2256}
2257
2258static int show_timer(struct seq_file *m, void *v)
2259{
2260 struct k_itimer *timer;
Pavel Emelyanov57b80152013-03-11 13:13:08 +04002261 struct timers_private *tp = m->private;
2262 int notify;
Alexey Dobriyancedbcca2014-08-08 14:21:33 -07002263 static const char * const nstr[] = {
Pavel Emelyanov57b80152013-03-11 13:13:08 +04002264 [SIGEV_SIGNAL] = "signal",
2265 [SIGEV_NONE] = "none",
2266 [SIGEV_THREAD] = "thread",
2267 };
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002268
2269 timer = list_entry((struct list_head *)v, struct k_itimer, list);
Pavel Emelyanov57b80152013-03-11 13:13:08 +04002270 notify = timer->it_sigev_notify;
2271
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002272 seq_printf(m, "ID: %d\n", timer->it_id);
Joe Perches25ce3192015-04-15 16:18:17 -07002273 seq_printf(m, "signal: %d/%p\n",
2274 timer->sigq->info.si_signo,
2275 timer->sigq->info.si_value.sival_ptr);
Pavel Emelyanov57b80152013-03-11 13:13:08 +04002276 seq_printf(m, "notify: %s/%s.%d\n",
Joe Perches25ce3192015-04-15 16:18:17 -07002277 nstr[notify & ~SIGEV_THREAD_ID],
2278 (notify & SIGEV_THREAD_ID) ? "tid" : "pid",
2279 pid_nr_ns(timer->it_pid, tp->ns));
Pavel Tikhomirov15ef0292013-05-17 02:12:03 +04002280 seq_printf(m, "ClockID: %d\n", timer->it_clock);
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002281
2282 return 0;
2283}
2284
2285static const struct seq_operations proc_timers_seq_ops = {
2286 .start = timers_start,
2287 .next = timers_next,
2288 .stop = timers_stop,
2289 .show = show_timer,
2290};
2291
2292static int proc_timers_open(struct inode *inode, struct file *file)
2293{
2294 struct timers_private *tp;
2295
2296 tp = __seq_open_private(file, &proc_timers_seq_ops,
2297 sizeof(struct timers_private));
2298 if (!tp)
2299 return -ENOMEM;
2300
2301 tp->pid = proc_pid(inode);
Pavel Emelyanov57b80152013-03-11 13:13:08 +04002302 tp->ns = inode->i_sb->s_fs_info;
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002303 return 0;
2304}
2305
2306static const struct file_operations proc_timers_operations = {
2307 .open = proc_timers_open,
2308 .read = seq_read,
2309 .llseek = seq_lseek,
2310 .release = seq_release_private,
2311};
Eric Engestromb5946be2016-03-17 14:20:57 -07002312#endif
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002313
John Stultz5de23d42016-03-17 14:20:54 -07002314static ssize_t timerslack_ns_write(struct file *file, const char __user *buf,
2315 size_t count, loff_t *offset)
2316{
2317 struct inode *inode = file_inode(file);
2318 struct task_struct *p;
2319 u64 slack_ns;
2320 int err;
2321
2322 err = kstrtoull_from_user(buf, count, 10, &slack_ns);
2323 if (err < 0)
2324 return err;
2325
2326 p = get_proc_task(inode);
2327 if (!p)
2328 return -ESRCH;
2329
John Stultz4b2bd5f2016-10-07 17:02:33 -07002330 if (p != current) {
2331 if (!capable(CAP_SYS_NICE)) {
2332 count = -EPERM;
2333 goto out;
2334 }
John Stultz5de23d42016-03-17 14:20:54 -07002335
John Stultz4b2bd5f2016-10-07 17:02:33 -07002336 err = security_task_setscheduler(p);
2337 if (err) {
2338 count = err;
2339 goto out;
2340 }
John Stultz904763e2016-10-07 17:02:29 -07002341 }
2342
John Stultz7abbaf92016-10-07 17:02:26 -07002343 task_lock(p);
2344 if (slack_ns == 0)
2345 p->timer_slack_ns = p->default_timer_slack_ns;
2346 else
2347 p->timer_slack_ns = slack_ns;
2348 task_unlock(p);
2349
2350out:
John Stultz5de23d42016-03-17 14:20:54 -07002351 put_task_struct(p);
2352
2353 return count;
2354}
2355
2356static int timerslack_ns_show(struct seq_file *m, void *v)
2357{
2358 struct inode *inode = m->private;
2359 struct task_struct *p;
John Stultz7abbaf92016-10-07 17:02:26 -07002360 int err = 0;
John Stultz5de23d42016-03-17 14:20:54 -07002361
2362 p = get_proc_task(inode);
2363 if (!p)
2364 return -ESRCH;
2365
John Stultz4b2bd5f2016-10-07 17:02:33 -07002366 if (p != current) {
John Stultz5de23d42016-03-17 14:20:54 -07002367
John Stultz4b2bd5f2016-10-07 17:02:33 -07002368 if (!capable(CAP_SYS_NICE)) {
2369 err = -EPERM;
2370 goto out;
2371 }
2372 err = security_task_getscheduler(p);
2373 if (err)
2374 goto out;
2375 }
John Stultz904763e2016-10-07 17:02:29 -07002376
John Stultz7abbaf92016-10-07 17:02:26 -07002377 task_lock(p);
2378 seq_printf(m, "%llu\n", p->timer_slack_ns);
2379 task_unlock(p);
2380
2381out:
John Stultz5de23d42016-03-17 14:20:54 -07002382 put_task_struct(p);
2383
2384 return err;
2385}
2386
2387static int timerslack_ns_open(struct inode *inode, struct file *filp)
2388{
2389 return single_open(filp, timerslack_ns_show, inode);
2390}
2391
2392static const struct file_operations proc_pid_set_timerslack_ns_operations = {
2393 .open = timerslack_ns_open,
2394 .read = seq_read,
2395 .write = timerslack_ns_write,
2396 .llseek = seq_lseek,
2397 .release = single_release,
2398};
2399
Al Viroc52a47a2013-06-15 11:15:20 +04002400static int proc_pident_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002401 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002402{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002403 const struct pid_entry *p = ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002404 struct inode *inode;
2405 struct proc_inode *ei;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002406
Andreas Gruenbacherdb978da2016-11-10 22:18:28 +01002407 inode = proc_pid_make_inode(dir->i_sb, task, p->mode);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002408 if (!inode)
2409 goto out;
2410
2411 ei = PROC_I(inode);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002412 if (S_ISDIR(inode->i_mode))
Miklos Szeredibfe86842011-10-28 14:13:29 +02002413 set_nlink(inode, 2); /* Use getattr to fix if necessary */
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002414 if (p->iop)
2415 inode->i_op = p->iop;
2416 if (p->fop)
2417 inode->i_fop = p->fop;
2418 ei->op = p->op;
Nick Pigginfb045ad2011-01-07 17:49:55 +11002419 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002420 d_add(dentry, inode);
2421 /* Close the race of the process dying before we return the dentry */
Al Viro0b728e12012-06-10 16:03:43 -04002422 if (pid_revalidate(dentry, 0))
Al Viroc52a47a2013-06-15 11:15:20 +04002423 return 0;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002424out:
Al Viroc52a47a2013-06-15 11:15:20 +04002425 return -ENOENT;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002426}
2427
Linus Torvalds1da177e2005-04-16 15:20:36 -07002428static struct dentry *proc_pident_lookup(struct inode *dir,
2429 struct dentry *dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002430 const struct pid_entry *ents,
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002431 unsigned int nents)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002432{
Al Viroc52a47a2013-06-15 11:15:20 +04002433 int error;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002434 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002435 const struct pid_entry *p, *last;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002436
Al Viroc52a47a2013-06-15 11:15:20 +04002437 error = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002438
Eric W. Biederman99f89552006-06-26 00:25:55 -07002439 if (!task)
2440 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002441
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002442 /*
2443 * Yes, it does not scale. And it should not. Don't add
2444 * new entries into /proc/<tgid>/ without very good reasons.
2445 */
Alexey Dobriyanbac5f5d2016-12-12 16:45:28 -08002446 last = &ents[nents];
2447 for (p = ents; p < last; p++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002448 if (p->len != dentry->d_name.len)
2449 continue;
2450 if (!memcmp(dentry->d_name.name, p->name, p->len))
2451 break;
2452 }
Alexey Dobriyanbac5f5d2016-12-12 16:45:28 -08002453 if (p >= last)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002454 goto out;
2455
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002456 error = proc_pident_instantiate(dir, dentry, task, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002457out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07002458 put_task_struct(task);
2459out_no_task:
Al Viroc52a47a2013-06-15 11:15:20 +04002460 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002461}
2462
Al Virof0c3b502013-05-16 12:07:31 -04002463static int proc_pident_readdir(struct file *file, struct dir_context *ctx,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002464 const struct pid_entry *ents, unsigned int nents)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002465{
Al Virof0c3b502013-05-16 12:07:31 -04002466 struct task_struct *task = get_proc_task(file_inode(file));
2467 const struct pid_entry *p;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002468
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002469 if (!task)
Al Virof0c3b502013-05-16 12:07:31 -04002470 return -ENOENT;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002471
Al Virof0c3b502013-05-16 12:07:31 -04002472 if (!dir_emit_dots(file, ctx))
2473 goto out;
2474
2475 if (ctx->pos >= nents + 2)
2476 goto out;
2477
Alexey Dobriyanbac5f5d2016-12-12 16:45:28 -08002478 for (p = ents + (ctx->pos - 2); p < ents + nents; p++) {
Al Virof0c3b502013-05-16 12:07:31 -04002479 if (!proc_fill_cache(file, ctx, p->name, p->len,
2480 proc_pident_instantiate, task, p))
2481 break;
2482 ctx->pos++;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002483 }
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002484out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07002485 put_task_struct(task);
Al Virof0c3b502013-05-16 12:07:31 -04002486 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002487}
2488
Linus Torvalds1da177e2005-04-16 15:20:36 -07002489#ifdef CONFIG_SECURITY
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002490static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2491 size_t count, loff_t *ppos)
2492{
Al Viro496ad9a2013-01-23 17:07:38 -05002493 struct inode * inode = file_inode(file);
Al Viro04ff9702007-03-12 16:17:58 +00002494 char *p = NULL;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002495 ssize_t length;
2496 struct task_struct *task = get_proc_task(inode);
2497
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002498 if (!task)
Al Viro04ff9702007-03-12 16:17:58 +00002499 return -ESRCH;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002500
2501 length = security_getprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002502 (char*)file->f_path.dentry->d_name.name,
Al Viro04ff9702007-03-12 16:17:58 +00002503 &p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002504 put_task_struct(task);
Al Viro04ff9702007-03-12 16:17:58 +00002505 if (length > 0)
2506 length = simple_read_from_buffer(buf, count, ppos, p, length);
2507 kfree(p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002508 return length;
2509}
2510
2511static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2512 size_t count, loff_t *ppos)
2513{
Al Viro496ad9a2013-01-23 17:07:38 -05002514 struct inode * inode = file_inode(file);
Al Virobb646cd2015-12-24 00:16:30 -05002515 void *page;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002516 ssize_t length;
2517 struct task_struct *task = get_proc_task(inode);
2518
2519 length = -ESRCH;
2520 if (!task)
2521 goto out_no_task;
Stephen Smalleyb21507e2017-01-09 10:07:31 -05002522
2523 /* A task may only write its own attributes. */
2524 length = -EACCES;
2525 if (current != task)
2526 goto out;
2527
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002528 if (count > PAGE_SIZE)
2529 count = PAGE_SIZE;
2530
2531 /* No partial writes. */
2532 length = -EINVAL;
2533 if (*ppos != 0)
2534 goto out;
2535
Al Virobb646cd2015-12-24 00:16:30 -05002536 page = memdup_user(buf, count);
2537 if (IS_ERR(page)) {
2538 length = PTR_ERR(page);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002539 goto out;
Al Virobb646cd2015-12-24 00:16:30 -05002540 }
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002541
David Howells107db7c2009-05-08 13:55:27 +01002542 /* Guard against adverse ptrace interaction */
Stephen Smalleyb21507e2017-01-09 10:07:31 -05002543 length = mutex_lock_interruptible(&current->signal->cred_guard_mutex);
David Howells107db7c2009-05-08 13:55:27 +01002544 if (length < 0)
2545 goto out_free;
2546
Stephen Smalleyb21507e2017-01-09 10:07:31 -05002547 length = security_setprocattr(file->f_path.dentry->d_name.name,
Al Virobb646cd2015-12-24 00:16:30 -05002548 page, count);
Stephen Smalleyb21507e2017-01-09 10:07:31 -05002549 mutex_unlock(&current->signal->cred_guard_mutex);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002550out_free:
Al Virobb646cd2015-12-24 00:16:30 -05002551 kfree(page);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002552out:
2553 put_task_struct(task);
2554out_no_task:
2555 return length;
2556}
2557
Arjan van de Ven00977a52007-02-12 00:55:34 -08002558static const struct file_operations proc_pid_attr_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002559 .read = proc_pid_attr_read,
2560 .write = proc_pid_attr_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01002561 .llseek = generic_file_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002562};
2563
Eric Dumazetc5141e62007-05-08 00:26:15 -07002564static const struct pid_entry attr_dir_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002565 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2566 REG("prev", S_IRUGO, proc_pid_attr_operations),
2567 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2568 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2569 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2570 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002571};
2572
Al Virof0c3b502013-05-16 12:07:31 -04002573static int proc_attr_dir_readdir(struct file *file, struct dir_context *ctx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002574{
Al Virof0c3b502013-05-16 12:07:31 -04002575 return proc_pident_readdir(file, ctx,
2576 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002577}
2578
Arjan van de Ven00977a52007-02-12 00:55:34 -08002579static const struct file_operations proc_attr_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002580 .read = generic_read_dir,
Al Virof50752e2016-04-20 17:13:54 -04002581 .iterate_shared = proc_attr_dir_readdir,
2582 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002583};
2584
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002585static struct dentry *proc_attr_dir_lookup(struct inode *dir,
Al Viro00cd8dd2012-06-10 17:13:09 -04002586 struct dentry *dentry, unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002587{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002588 return proc_pident_lookup(dir, dentry,
2589 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002590}
2591
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002592static const struct inode_operations proc_attr_dir_inode_operations = {
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002593 .lookup = proc_attr_dir_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07002594 .getattr = pid_getattr,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07002595 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002596};
2597
Linus Torvalds1da177e2005-04-16 15:20:36 -07002598#endif
2599
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08002600#ifdef CONFIG_ELF_CORE
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002601static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2602 size_t count, loff_t *ppos)
2603{
Al Viro496ad9a2013-01-23 17:07:38 -05002604 struct task_struct *task = get_proc_task(file_inode(file));
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002605 struct mm_struct *mm;
2606 char buffer[PROC_NUMBUF];
2607 size_t len;
2608 int ret;
2609
2610 if (!task)
2611 return -ESRCH;
2612
2613 ret = 0;
2614 mm = get_task_mm(task);
2615 if (mm) {
2616 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2617 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2618 MMF_DUMP_FILTER_SHIFT));
2619 mmput(mm);
2620 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2621 }
2622
2623 put_task_struct(task);
2624
2625 return ret;
2626}
2627
2628static ssize_t proc_coredump_filter_write(struct file *file,
2629 const char __user *buf,
2630 size_t count,
2631 loff_t *ppos)
2632{
2633 struct task_struct *task;
2634 struct mm_struct *mm;
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002635 unsigned int val;
2636 int ret;
2637 int i;
2638 unsigned long mask;
2639
Alexey Dobriyan774636e2015-09-09 15:36:59 -07002640 ret = kstrtouint_from_user(buf, count, 0, &val);
2641 if (ret < 0)
2642 return ret;
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002643
2644 ret = -ESRCH;
Al Viro496ad9a2013-01-23 17:07:38 -05002645 task = get_proc_task(file_inode(file));
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002646 if (!task)
2647 goto out_no_task;
2648
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002649 mm = get_task_mm(task);
2650 if (!mm)
2651 goto out_no_mm;
Colin Ian King41a0c2492015-12-18 14:22:01 -08002652 ret = 0;
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002653
2654 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2655 if (val & mask)
2656 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2657 else
2658 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2659 }
2660
2661 mmput(mm);
2662 out_no_mm:
2663 put_task_struct(task);
2664 out_no_task:
Alexey Dobriyan774636e2015-09-09 15:36:59 -07002665 if (ret < 0)
2666 return ret;
2667 return count;
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002668}
2669
2670static const struct file_operations proc_coredump_filter_operations = {
2671 .read = proc_coredump_filter_read,
2672 .write = proc_coredump_filter_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01002673 .llseek = generic_file_llseek,
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002674};
2675#endif
2676
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002677#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002678static int do_io_accounting(struct task_struct *task, struct seq_file *m, int whole)
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002679{
Andrea Righi940389b2008-07-28 00:48:12 +02002680 struct task_io_accounting acct = task->ioac;
Andrea Righi59954772008-07-27 17:29:15 +02002681 unsigned long flags;
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002682 int result;
Andrea Righi297c5d92008-07-25 01:48:49 -07002683
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002684 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2685 if (result)
2686 return result;
2687
Jann Horncaaee622016-01-20 15:00:04 -08002688 if (!ptrace_may_access(task, PTRACE_MODE_READ_FSCREDS)) {
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002689 result = -EACCES;
2690 goto out_unlock;
2691 }
Vasiliy Kulikov1d1221f2011-06-24 16:08:38 +04002692
Andrea Righi59954772008-07-27 17:29:15 +02002693 if (whole && lock_task_sighand(task, &flags)) {
2694 struct task_struct *t = task;
Andrea Righib2d002d2008-07-26 15:22:27 -07002695
Andrea Righi59954772008-07-27 17:29:15 +02002696 task_io_accounting_add(&acct, &task->signal->ioac);
2697 while_each_thread(task, t)
2698 task_io_accounting_add(&acct, &t->ioac);
Andrea Righi297c5d92008-07-25 01:48:49 -07002699
Andrea Righi59954772008-07-27 17:29:15 +02002700 unlock_task_sighand(task, &flags);
Andrea Righi297c5d92008-07-25 01:48:49 -07002701 }
Joe Perches25ce3192015-04-15 16:18:17 -07002702 seq_printf(m,
2703 "rchar: %llu\n"
2704 "wchar: %llu\n"
2705 "syscr: %llu\n"
2706 "syscw: %llu\n"
2707 "read_bytes: %llu\n"
2708 "write_bytes: %llu\n"
2709 "cancelled_write_bytes: %llu\n",
2710 (unsigned long long)acct.rchar,
2711 (unsigned long long)acct.wchar,
2712 (unsigned long long)acct.syscr,
2713 (unsigned long long)acct.syscw,
2714 (unsigned long long)acct.read_bytes,
2715 (unsigned long long)acct.write_bytes,
2716 (unsigned long long)acct.cancelled_write_bytes);
2717 result = 0;
2718
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002719out_unlock:
2720 mutex_unlock(&task->signal->cred_guard_mutex);
2721 return result;
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002722}
Andrea Righi297c5d92008-07-25 01:48:49 -07002723
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002724static int proc_tid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2725 struct pid *pid, struct task_struct *task)
Andrea Righi297c5d92008-07-25 01:48:49 -07002726{
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002727 return do_io_accounting(task, m, 0);
Andrea Righi297c5d92008-07-25 01:48:49 -07002728}
2729
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002730static int proc_tgid_io_accounting(struct seq_file *m, struct pid_namespace *ns,
2731 struct pid *pid, struct task_struct *task)
Andrea Righi297c5d92008-07-25 01:48:49 -07002732{
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002733 return do_io_accounting(task, m, 1);
Andrea Righi297c5d92008-07-25 01:48:49 -07002734}
2735#endif /* CONFIG_TASK_IO_ACCOUNTING */
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002736
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002737#ifdef CONFIG_USER_NS
2738static int proc_id_map_open(struct inode *inode, struct file *file,
Fabian Frederickccf94f12014-08-08 14:21:22 -07002739 const struct seq_operations *seq_ops)
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002740{
2741 struct user_namespace *ns = NULL;
2742 struct task_struct *task;
2743 struct seq_file *seq;
2744 int ret = -EINVAL;
2745
2746 task = get_proc_task(inode);
2747 if (task) {
2748 rcu_read_lock();
2749 ns = get_user_ns(task_cred_xxx(task, user_ns));
2750 rcu_read_unlock();
2751 put_task_struct(task);
2752 }
2753 if (!ns)
2754 goto err;
2755
2756 ret = seq_open(file, seq_ops);
2757 if (ret)
2758 goto err_put_ns;
2759
2760 seq = file->private_data;
2761 seq->private = ns;
2762
2763 return 0;
2764err_put_ns:
2765 put_user_ns(ns);
2766err:
2767 return ret;
2768}
2769
2770static int proc_id_map_release(struct inode *inode, struct file *file)
2771{
2772 struct seq_file *seq = file->private_data;
2773 struct user_namespace *ns = seq->private;
2774 put_user_ns(ns);
2775 return seq_release(inode, file);
2776}
2777
2778static int proc_uid_map_open(struct inode *inode, struct file *file)
2779{
2780 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2781}
2782
2783static int proc_gid_map_open(struct inode *inode, struct file *file)
2784{
2785 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2786}
2787
Eric W. Biedermanf76d2072012-08-30 01:24:05 -07002788static int proc_projid_map_open(struct inode *inode, struct file *file)
2789{
2790 return proc_id_map_open(inode, file, &proc_projid_seq_operations);
2791}
2792
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002793static const struct file_operations proc_uid_map_operations = {
2794 .open = proc_uid_map_open,
2795 .write = proc_uid_map_write,
2796 .read = seq_read,
2797 .llseek = seq_lseek,
2798 .release = proc_id_map_release,
2799};
2800
2801static const struct file_operations proc_gid_map_operations = {
2802 .open = proc_gid_map_open,
2803 .write = proc_gid_map_write,
2804 .read = seq_read,
2805 .llseek = seq_lseek,
2806 .release = proc_id_map_release,
2807};
Eric W. Biedermanf76d2072012-08-30 01:24:05 -07002808
2809static const struct file_operations proc_projid_map_operations = {
2810 .open = proc_projid_map_open,
2811 .write = proc_projid_map_write,
2812 .read = seq_read,
2813 .llseek = seq_lseek,
2814 .release = proc_id_map_release,
2815};
Eric W. Biederman9cc46512014-12-02 12:27:26 -06002816
2817static int proc_setgroups_open(struct inode *inode, struct file *file)
2818{
2819 struct user_namespace *ns = NULL;
2820 struct task_struct *task;
2821 int ret;
2822
2823 ret = -ESRCH;
2824 task = get_proc_task(inode);
2825 if (task) {
2826 rcu_read_lock();
2827 ns = get_user_ns(task_cred_xxx(task, user_ns));
2828 rcu_read_unlock();
2829 put_task_struct(task);
2830 }
2831 if (!ns)
2832 goto err;
2833
2834 if (file->f_mode & FMODE_WRITE) {
2835 ret = -EACCES;
2836 if (!ns_capable(ns, CAP_SYS_ADMIN))
2837 goto err_put_ns;
2838 }
2839
2840 ret = single_open(file, &proc_setgroups_show, ns);
2841 if (ret)
2842 goto err_put_ns;
2843
2844 return 0;
2845err_put_ns:
2846 put_user_ns(ns);
2847err:
2848 return ret;
2849}
2850
2851static int proc_setgroups_release(struct inode *inode, struct file *file)
2852{
2853 struct seq_file *seq = file->private_data;
2854 struct user_namespace *ns = seq->private;
2855 int ret = single_release(inode, file);
2856 put_user_ns(ns);
2857 return ret;
2858}
2859
2860static const struct file_operations proc_setgroups_operations = {
2861 .open = proc_setgroups_open,
2862 .write = proc_setgroups_write,
2863 .read = seq_read,
2864 .llseek = seq_lseek,
2865 .release = proc_setgroups_release,
2866};
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002867#endif /* CONFIG_USER_NS */
2868
Kees Cook47830722008-10-06 03:11:58 +04002869static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2870 struct pid *pid, struct task_struct *task)
2871{
Al Viroa9712bc2011-03-23 15:52:50 -04002872 int err = lock_trace(task);
2873 if (!err) {
2874 seq_printf(m, "%08x\n", task->personality);
2875 unlock_trace(task);
2876 }
2877 return err;
Kees Cook47830722008-10-06 03:11:58 +04002878}
2879
Josh Poimboeuf7c23b332017-02-13 19:42:41 -06002880#ifdef CONFIG_LIVEPATCH
2881static int proc_pid_patch_state(struct seq_file *m, struct pid_namespace *ns,
2882 struct pid *pid, struct task_struct *task)
2883{
2884 seq_printf(m, "%d\n", task->patch_state);
2885 return 0;
2886}
2887#endif /* CONFIG_LIVEPATCH */
2888
Eric W. Biederman801199c2006-10-02 02:18:48 -07002889/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002890 * Thread groups
2891 */
Arjan van de Ven00977a52007-02-12 00:55:34 -08002892static const struct file_operations proc_task_operations;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002893static const struct inode_operations proc_task_inode_operations;
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002894
Eric Dumazetc5141e62007-05-08 00:26:15 -07002895static const struct pid_entry tgid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002896 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2897 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002898 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002899 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08002900 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002901#ifdef CONFIG_NET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002902 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002903#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002904 REG("environ", S_IRUSR, proc_environ_operations),
Al Viroc5317162016-10-05 18:43:43 -04002905 REG("auxv", S_IRUSR, proc_auxv_operations),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002906 ONE("status", S_IRUGO, proc_pid_status),
Djalal Harouni35a35042014-04-07 15:38:36 -07002907 ONE("personality", S_IRUSR, proc_pid_personality),
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -07002908 ONE("limits", S_IRUGO, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002909#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002910 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002911#endif
Mike Galbraith5091faa2010-11-30 14:18:03 +01002912#ifdef CONFIG_SCHED_AUTOGROUP
2913 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2914#endif
john stultz4614a696b2009-12-14 18:00:05 -08002915 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
Roland McGrathebcb6732008-07-25 19:46:00 -07002916#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -07002917 ONE("syscall", S_IRUSR, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07002918#endif
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -07002919 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002920 ONE("stat", S_IRUGO, proc_tgid_stat),
2921 ONE("statm", S_IRUGO, proc_pid_statm),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07002922 REG("maps", S_IRUGO, proc_pid_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002923#ifdef CONFIG_NUMA
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07002924 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002925#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002926 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2927 LNK("cwd", proc_cwd_link),
2928 LNK("root", proc_root_link),
2929 LNK("exe", proc_exe_link),
2930 REG("mounts", S_IRUGO, proc_mounts_operations),
2931 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2932 REG("mountstats", S_IRUSR, proc_mountstats_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08002933#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002934 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07002935 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
Djalal Harouni32ed74a2014-04-07 15:38:38 -07002936 REG("pagemap", S_IRUSR, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002937#endif
2938#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002939 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002940#endif
2941#ifdef CONFIG_KALLSYMS
Alexey Dobriyanedfcd602014-08-08 14:21:44 -07002942 ONE("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002943#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03002944#ifdef CONFIG_STACKTRACE
Djalal Harouni35a35042014-04-07 15:38:36 -07002945 ONE("stack", S_IRUSR, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002946#endif
Naveen N. Rao5968cec2015-06-30 14:36:03 +05302947#ifdef CONFIG_SCHED_INFO
Alexey Dobriyanf6e826c2014-08-08 14:21:46 -07002948 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002949#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01002950#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002951 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01002952#endif
Paul Menage8793d852007-10-18 23:39:39 -07002953#ifdef CONFIG_PROC_PID_CPUSET
Zefan Li52de4772014-09-18 16:03:36 +08002954 ONE("cpuset", S_IRUGO, proc_cpuset_show),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002955#endif
Paul Menagea4243162007-10-18 23:39:35 -07002956#ifdef CONFIG_CGROUPS
Zefan Li006f4ac2014-09-18 16:03:15 +08002957 ONE("cgroup", S_IRUGO, proc_cgroup_show),
Paul Menagea4243162007-10-18 23:39:35 -07002958#endif
Alexey Dobriyan6ba51e32014-08-08 14:21:48 -07002959 ONE("oom_score", S_IRUGO, proc_oom_score),
David Rientjesfa0cbbf2012-11-12 17:53:04 -08002960 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
David Rientjesa63d83f2010-08-09 17:19:46 -07002961 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002962#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002963 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2964 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002965#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002966#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002967 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002968#endif
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08002969#ifdef CONFIG_ELF_CORE
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002970 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002971#endif
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002972#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07002973 ONE("io", S_IRUSR, proc_tgid_io_accounting),
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002974#endif
Chris Metcalff133ecc2011-05-26 12:40:09 -04002975#ifdef CONFIG_HARDWALL
Alexey Dobriyand962c142014-08-08 14:21:52 -07002976 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
Chris Metcalff133ecc2011-05-26 12:40:09 -04002977#endif
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002978#ifdef CONFIG_USER_NS
2979 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
2980 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
Eric W. Biedermanf76d2072012-08-30 01:24:05 -07002981 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
Eric W. Biederman9cc46512014-12-02 12:27:26 -06002982 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002983#endif
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05002984#if defined(CONFIG_CHECKPOINT_RESTORE) && defined(CONFIG_POSIX_TIMERS)
Pavel Emelyanov48f6a7a2013-03-11 13:12:45 +04002985 REG("timers", S_IRUGO, proc_timers_operations),
2986#endif
John Stultz5de23d42016-03-17 14:20:54 -07002987 REG("timerslack_ns", S_IRUGO|S_IWUGO, proc_pid_set_timerslack_ns_operations),
Josh Poimboeuf7c23b332017-02-13 19:42:41 -06002988#ifdef CONFIG_LIVEPATCH
2989 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
2990#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002991};
2992
Al Virof0c3b502013-05-16 12:07:31 -04002993static int proc_tgid_base_readdir(struct file *file, struct dir_context *ctx)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002994{
Al Virof0c3b502013-05-16 12:07:31 -04002995 return proc_pident_readdir(file, ctx,
2996 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002997}
2998
Arjan van de Ven00977a52007-02-12 00:55:34 -08002999static const struct file_operations proc_tgid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003000 .read = generic_read_dir,
Al Virof50752e2016-04-20 17:13:54 -04003001 .iterate_shared = proc_tgid_base_readdir,
3002 .llseek = generic_file_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003003};
3004
Al Viro00cd8dd2012-06-10 17:13:09 -04003005static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3006{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003007 return proc_pident_lookup(dir, dentry,
3008 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003009}
3010
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003011static const struct inode_operations proc_tgid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003012 .lookup = proc_tgid_base_lookup,
3013 .getattr = pid_getattr,
3014 .setattr = proc_setattr,
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003015 .permission = proc_pid_permission,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003016};
3017
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003018static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003019{
Eric W. Biederman48e64842006-06-26 00:25:48 -07003020 struct dentry *dentry, *leader, *dir;
Eric W. Biederman8578cea2006-06-26 00:25:54 -07003021 char buf[PROC_NUMBUF];
Eric W. Biederman48e64842006-06-26 00:25:48 -07003022 struct qstr name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003023
Eric W. Biederman48e64842006-06-26 00:25:48 -07003024 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003025 name.len = snprintf(buf, sizeof(buf), "%d", pid);
Al Viro4f522a22013-02-11 23:20:37 -05003026 /* no ->d_hash() rejects on procfs */
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003027 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003028 if (dentry) {
Eric W. Biedermanbbd51922014-02-13 10:24:23 -08003029 d_invalidate(dentry);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003030 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003031 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003032
Oleg Nesterovc35a7f12014-12-10 15:54:56 -08003033 if (pid == tgid)
3034 return;
3035
Eric W. Biederman48e64842006-06-26 00:25:48 -07003036 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003037 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3038 leader = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003039 if (!leader)
3040 goto out;
3041
3042 name.name = "task";
3043 name.len = strlen(name.name);
3044 dir = d_hash_and_lookup(leader, &name);
3045 if (!dir)
3046 goto out_put_leader;
3047
3048 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003049 name.len = snprintf(buf, sizeof(buf), "%d", pid);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003050 dentry = d_hash_and_lookup(dir, &name);
3051 if (dentry) {
Eric W. Biedermanbbd51922014-02-13 10:24:23 -08003052 d_invalidate(dentry);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003053 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003054 }
Eric W. Biederman48e64842006-06-26 00:25:48 -07003055
3056 dput(dir);
3057out_put_leader:
3058 dput(leader);
3059out:
3060 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003061}
3062
Randy Dunlap0895e912007-10-21 21:00:10 -07003063/**
3064 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3065 * @task: task that should be flushed.
3066 *
3067 * When flushing dentries from proc, one needs to flush them from global
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003068 * proc (proc_mnt) and from all the namespaces' procs this task was seen
Randy Dunlap0895e912007-10-21 21:00:10 -07003069 * in. This call is supposed to do all of this job.
3070 *
3071 * Looks in the dcache for
3072 * /proc/@pid
3073 * /proc/@tgid/task/@pid
3074 * if either directory is present flushes it and all of it'ts children
3075 * from the dcache.
3076 *
3077 * It is safe and reasonable to cache /proc entries for a task until
3078 * that task exits. After that they just clog up the dcache with
3079 * useless entries, possibly causing useful dcache entries to be
3080 * flushed instead. This routine is proved to flush those useless
3081 * dcache entries at process exit time.
3082 *
3083 * NOTE: This routine is just an optimization so it does not guarantee
3084 * that no dcache entries will exist at process exit time it
3085 * just makes it very unlikely that any will persist.
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003086 */
3087
3088void proc_flush_task(struct task_struct *task)
3089{
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08003090 int i;
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07003091 struct pid *pid, *tgid;
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003092 struct upid *upid;
3093
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003094 pid = task_pid(task);
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07003095 tgid = task_tgid(task);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003096
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08003097 for (i = 0; i <= pid->level; i++) {
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003098 upid = &pid->numbers[i];
3099 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07003100 tgid->numbers[i].nr);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003101 }
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003102}
3103
Al Viroc52a47a2013-06-15 11:15:20 +04003104static int proc_pid_instantiate(struct inode *dir,
3105 struct dentry * dentry,
3106 struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003107{
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003108 struct inode *inode;
3109
Andreas Gruenbacherdb978da2016-11-10 22:18:28 +01003110 inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003111 if (!inode)
3112 goto out;
3113
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003114 inode->i_op = &proc_tgid_base_inode_operations;
3115 inode->i_fop = &proc_tgid_base_operations;
3116 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07003117
Alexey Dobriyan1270dd82016-12-12 16:45:32 -08003118 set_nlink(inode, nlink_tgid);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003119
Nick Pigginfb045ad2011-01-07 17:49:55 +11003120 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003121
3122 d_add(dentry, inode);
3123 /* Close the race of the process dying before we return the dentry */
Al Viro0b728e12012-06-10 16:03:43 -04003124 if (pid_revalidate(dentry, 0))
Al Viroc52a47a2013-06-15 11:15:20 +04003125 return 0;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003126out:
Al Viroc52a47a2013-06-15 11:15:20 +04003127 return -ENOENT;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003128}
3129
Al Viro00cd8dd2012-06-10 17:13:09 -04003130struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003131{
Alexey Dobriyan335eb532014-08-08 14:21:27 -07003132 int result = -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003133 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003134 unsigned tgid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003135 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003136
Alexey Dobriyandbcdb502014-08-08 14:21:25 -07003137 tgid = name_to_int(&dentry->d_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003138 if (tgid == ~0U)
3139 goto out;
3140
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003141 ns = dentry->d_sb->s_fs_info;
Eric W. Biedermande758732006-06-26 00:25:51 -07003142 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003143 task = find_task_by_pid_ns(tgid, ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003144 if (task)
3145 get_task_struct(task);
Eric W. Biedermande758732006-06-26 00:25:51 -07003146 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003147 if (!task)
3148 goto out;
3149
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003150 result = proc_pid_instantiate(dir, dentry, task, NULL);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003151 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003152out:
Al Viroc52a47a2013-06-15 11:15:20 +04003153 return ERR_PTR(result);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003154}
3155
Linus Torvalds1da177e2005-04-16 15:20:36 -07003156/*
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003157 * Find the first task with tgid >= tgid
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003158 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07003159 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003160struct tgid_iter {
3161 unsigned int tgid;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003162 struct task_struct *task;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003163};
3164static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3165{
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003166 struct pid *pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003167
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003168 if (iter.task)
3169 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003170 rcu_read_lock();
3171retry:
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003172 iter.task = NULL;
3173 pid = find_ge_pid(iter.tgid, ns);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003174 if (pid) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003175 iter.tgid = pid_nr_ns(pid, ns);
3176 iter.task = pid_task(pid, PIDTYPE_PID);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003177 /* What we to know is if the pid we have find is the
3178 * pid of a thread_group_leader. Testing for task
3179 * being a thread_group_leader is the obvious thing
3180 * todo but there is a window when it fails, due to
3181 * the pid transfer logic in de_thread.
3182 *
3183 * So we perform the straight forward test of seeing
3184 * if the pid we have found is the pid of a thread
3185 * group leader, and don't worry if the task we have
3186 * found doesn't happen to be a thread group leader.
3187 * As we don't care in the case of readdir.
3188 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003189 if (!iter.task || !has_group_leader_pid(iter.task)) {
3190 iter.tgid += 1;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003191 goto retry;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003192 }
3193 get_task_struct(iter.task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003194 }
Eric W. Biederman454cc102006-06-26 00:25:51 -07003195 rcu_read_unlock();
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003196 return iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003197}
3198
Eric W. Biederman00978752014-07-31 03:10:50 -07003199#define TGID_OFFSET (FIRST_PROCESS_ENTRY + 2)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003200
3201/* for the /proc/ directory itself, after non-process stuff has been done */
Al Virof0c3b502013-05-16 12:07:31 -04003202int proc_pid_readdir(struct file *file, struct dir_context *ctx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003203{
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003204 struct tgid_iter iter;
Al Viro3aa33772014-10-31 00:42:35 -04003205 struct pid_namespace *ns = file_inode(file)->i_sb->s_fs_info;
Al Virof0c3b502013-05-16 12:07:31 -04003206 loff_t pos = ctx->pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003207
Al Viro021ada72013-03-29 19:27:05 -04003208 if (pos >= PID_MAX_LIMIT + TGID_OFFSET)
Al Virof0c3b502013-05-16 12:07:31 -04003209 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003210
Eric W. Biederman00978752014-07-31 03:10:50 -07003211 if (pos == TGID_OFFSET - 2) {
David Howells2b0143b2015-03-17 22:25:59 +00003212 struct inode *inode = d_inode(ns->proc_self);
Al Virodb963162013-06-15 10:45:10 +04003213 if (!dir_emit(ctx, "self", 4, inode->i_ino, DT_LNK))
Al Virof0c3b502013-05-16 12:07:31 -04003214 return 0;
Eric W. Biederman00978752014-07-31 03:10:50 -07003215 ctx->pos = pos = pos + 1;
Al Viro021ada72013-03-29 19:27:05 -04003216 }
Eric W. Biederman00978752014-07-31 03:10:50 -07003217 if (pos == TGID_OFFSET - 1) {
David Howells2b0143b2015-03-17 22:25:59 +00003218 struct inode *inode = d_inode(ns->proc_thread_self);
Eric W. Biederman00978752014-07-31 03:10:50 -07003219 if (!dir_emit(ctx, "thread-self", 11, inode->i_ino, DT_LNK))
3220 return 0;
3221 ctx->pos = pos = pos + 1;
3222 }
3223 iter.tgid = pos - TGID_OFFSET;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003224 iter.task = NULL;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003225 for (iter = next_tgid(ns, iter);
3226 iter.task;
3227 iter.tgid += 1, iter = next_tgid(ns, iter)) {
Al Virof0c3b502013-05-16 12:07:31 -04003228 char name[PROC_NUMBUF];
3229 int len;
Eric Dumazet3ba4bce2017-01-24 15:18:07 -08003230
3231 cond_resched();
Lafcadio Wluiki796f5712017-02-24 15:00:23 -08003232 if (!has_pid_permissions(ns, iter.task, HIDEPID_INVISIBLE))
Al Virof0c3b502013-05-16 12:07:31 -04003233 continue;
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003234
Al Virof0c3b502013-05-16 12:07:31 -04003235 len = snprintf(name, sizeof(name), "%d", iter.tgid);
3236 ctx->pos = iter.tgid + TGID_OFFSET;
3237 if (!proc_fill_cache(file, ctx, name, len,
3238 proc_pid_instantiate, iter.task, NULL)) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003239 put_task_struct(iter.task);
Al Virof0c3b502013-05-16 12:07:31 -04003240 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003241 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003242 }
Al Virof0c3b502013-05-16 12:07:31 -04003243 ctx->pos = PID_MAX_LIMIT + TGID_OFFSET;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003244 return 0;
3245}
3246
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003247/*
Janis Danisevskis1b3044e2016-05-20 17:00:08 -07003248 * proc_tid_comm_permission is a special permission function exclusively
3249 * used for the node /proc/<pid>/task/<tid>/comm.
3250 * It bypasses generic permission checks in the case where a task of the same
3251 * task group attempts to access the node.
3252 * The rationale behind this is that glibc and bionic access this node for
3253 * cross thread naming (pthread_set/getname_np(!self)). However, if
3254 * PR_SET_DUMPABLE gets set to 0 this node among others becomes uid=0 gid=0,
3255 * which locks out the cross thread naming implementation.
3256 * This function makes sure that the node is always accessible for members of
3257 * same thread group.
3258 */
3259static int proc_tid_comm_permission(struct inode *inode, int mask)
3260{
3261 bool is_same_tgroup;
3262 struct task_struct *task;
3263
3264 task = get_proc_task(inode);
3265 if (!task)
3266 return -ESRCH;
3267 is_same_tgroup = same_thread_group(current, task);
3268 put_task_struct(task);
3269
3270 if (likely(is_same_tgroup && !(mask & MAY_EXEC))) {
3271 /* This file (/proc/<pid>/task/<tid>/comm) can always be
3272 * read or written by the members of the corresponding
3273 * thread group.
3274 */
3275 return 0;
3276 }
3277
3278 return generic_permission(inode, mask);
3279}
3280
3281static const struct inode_operations proc_tid_comm_inode_operations = {
3282 .permission = proc_tid_comm_permission,
3283};
3284
3285/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003286 * Tasks
3287 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07003288static const struct pid_entry tid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003289 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
Jerome Marchand38355412010-04-27 13:13:06 -07003290 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08003291 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
Eric W. Biederman6ba8ed72014-07-31 16:27:08 -07003292#ifdef CONFIG_NET
3293 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3294#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003295 REG("environ", S_IRUSR, proc_environ_operations),
Al Viroc5317162016-10-05 18:43:43 -04003296 REG("auxv", S_IRUSR, proc_auxv_operations),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003297 ONE("status", S_IRUGO, proc_pid_status),
Djalal Harouni35a35042014-04-07 15:38:36 -07003298 ONE("personality", S_IRUSR, proc_pid_personality),
Alexey Dobriyan1c963eb2014-08-08 14:21:37 -07003299 ONE("limits", S_IRUGO, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02003300#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003301 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02003302#endif
Janis Danisevskis1b3044e2016-05-20 17:00:08 -07003303 NOD("comm", S_IFREG|S_IRUGO|S_IWUSR,
3304 &proc_tid_comm_inode_operations,
3305 &proc_pid_set_comm_operations, {}),
Roland McGrathebcb6732008-07-25 19:46:00 -07003306#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan09d93bd2014-08-08 14:21:39 -07003307 ONE("syscall", S_IRUSR, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07003308#endif
Alexey Dobriyanc2c0bb42015-06-25 15:00:54 -07003309 REG("cmdline", S_IRUGO, proc_pid_cmdline_ops),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003310 ONE("stat", S_IRUGO, proc_tid_stat),
3311 ONE("statm", S_IRUGO, proc_pid_statm),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07003312 REG("maps", S_IRUGO, proc_tid_maps_operations),
Iago López Galeiras2e13ba52015-06-25 15:00:57 -07003313#ifdef CONFIG_PROC_CHILDREN
Cyrill Gorcunov818411612012-05-31 16:26:43 -07003314 REG("children", S_IRUGO, proc_tid_children_operations),
3315#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003316#ifdef CONFIG_NUMA
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07003317 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003318#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003319 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3320 LNK("cwd", proc_cwd_link),
3321 LNK("root", proc_root_link),
3322 LNK("exe", proc_exe_link),
3323 REG("mounts", S_IRUGO, proc_mounts_operations),
3324 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08003325#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003326 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07003327 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
Djalal Harouni32ed74a2014-04-07 15:38:38 -07003328 REG("pagemap", S_IRUSR, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003329#endif
3330#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003331 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003332#endif
3333#ifdef CONFIG_KALLSYMS
Alexey Dobriyanedfcd602014-08-08 14:21:44 -07003334 ONE("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003335#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03003336#ifdef CONFIG_STACKTRACE
Djalal Harouni35a35042014-04-07 15:38:36 -07003337 ONE("stack", S_IRUSR, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003338#endif
Naveen N. Rao5968cec2015-06-30 14:36:03 +05303339#ifdef CONFIG_SCHED_INFO
Alexey Dobriyanf6e826c2014-08-08 14:21:46 -07003340 ONE("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003341#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01003342#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003343 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01003344#endif
Paul Menage8793d852007-10-18 23:39:39 -07003345#ifdef CONFIG_PROC_PID_CPUSET
Zefan Li52de4772014-09-18 16:03:36 +08003346 ONE("cpuset", S_IRUGO, proc_cpuset_show),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003347#endif
Paul Menagea4243162007-10-18 23:39:35 -07003348#ifdef CONFIG_CGROUPS
Zefan Li006f4ac2014-09-18 16:03:15 +08003349 ONE("cgroup", S_IRUGO, proc_cgroup_show),
Paul Menagea4243162007-10-18 23:39:35 -07003350#endif
Alexey Dobriyan6ba51e32014-08-08 14:21:48 -07003351 ONE("oom_score", S_IRUGO, proc_oom_score),
David Rientjesfa0cbbf2012-11-12 17:53:04 -08003352 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adj_operations),
David Rientjesa63d83f2010-08-09 17:19:46 -07003353 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003354#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003355 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
Al Viro26ec3c62011-02-15 21:24:05 -05003356 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003357#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003358#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003359 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Dmitry Vyukove41d58182017-07-12 14:34:35 -07003360 /*
3361 * Operations on the file check that the task is current,
3362 * so we create it with 0666 to support testing under unprivileged user.
3363 */
3364 REG("fail-nth", 0666, proc_fail_nth_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003365#endif
Andrea Righi297c5d92008-07-25 01:48:49 -07003366#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan19aadc92014-08-08 14:21:50 -07003367 ONE("io", S_IRUSR, proc_tid_io_accounting),
Andrea Righi297c5d92008-07-25 01:48:49 -07003368#endif
Chris Metcalff133ecc2011-05-26 12:40:09 -04003369#ifdef CONFIG_HARDWALL
Alexey Dobriyand962c142014-08-08 14:21:52 -07003370 ONE("hardwall", S_IRUGO, proc_pid_hardwall),
Chris Metcalff133ecc2011-05-26 12:40:09 -04003371#endif
Eric W. Biederman22d917d2011-11-17 00:11:58 -08003372#ifdef CONFIG_USER_NS
3373 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3374 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
Eric W. Biedermanf76d2072012-08-30 01:24:05 -07003375 REG("projid_map", S_IRUGO|S_IWUSR, proc_projid_map_operations),
Eric W. Biederman9cc46512014-12-02 12:27:26 -06003376 REG("setgroups", S_IRUGO|S_IWUSR, proc_setgroups_operations),
Eric W. Biederman22d917d2011-11-17 00:11:58 -08003377#endif
Josh Poimboeuf7c23b332017-02-13 19:42:41 -06003378#ifdef CONFIG_LIVEPATCH
3379 ONE("patch_state", S_IRUSR, proc_pid_patch_state),
3380#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003381};
3382
Al Virof0c3b502013-05-16 12:07:31 -04003383static int proc_tid_base_readdir(struct file *file, struct dir_context *ctx)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003384{
Al Virof0c3b502013-05-16 12:07:31 -04003385 return proc_pident_readdir(file, ctx,
3386 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003387}
3388
Al Viro00cd8dd2012-06-10 17:13:09 -04003389static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
3390{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003391 return proc_pident_lookup(dir, dentry,
3392 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003393}
3394
Arjan van de Ven00977a52007-02-12 00:55:34 -08003395static const struct file_operations proc_tid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003396 .read = generic_read_dir,
Al Virof50752e2016-04-20 17:13:54 -04003397 .iterate_shared = proc_tid_base_readdir,
3398 .llseek = generic_file_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003399};
3400
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003401static const struct inode_operations proc_tid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003402 .lookup = proc_tid_base_lookup,
3403 .getattr = pid_getattr,
3404 .setattr = proc_setattr,
3405};
3406
Al Viroc52a47a2013-06-15 11:15:20 +04003407static int proc_task_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07003408 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003409{
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003410 struct inode *inode;
Andreas Gruenbacherdb978da2016-11-10 22:18:28 +01003411 inode = proc_pid_make_inode(dir->i_sb, task, S_IFDIR | S_IRUGO | S_IXUGO);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003412
3413 if (!inode)
3414 goto out;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003415 inode->i_op = &proc_tid_base_inode_operations;
3416 inode->i_fop = &proc_tid_base_operations;
3417 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07003418
Alexey Dobriyan1270dd82016-12-12 16:45:32 -08003419 set_nlink(inode, nlink_tid);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003420
Nick Pigginfb045ad2011-01-07 17:49:55 +11003421 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003422
3423 d_add(dentry, inode);
3424 /* Close the race of the process dying before we return the dentry */
Al Viro0b728e12012-06-10 16:03:43 -04003425 if (pid_revalidate(dentry, 0))
Al Viroc52a47a2013-06-15 11:15:20 +04003426 return 0;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003427out:
Al Viroc52a47a2013-06-15 11:15:20 +04003428 return -ENOENT;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003429}
3430
Al Viro00cd8dd2012-06-10 17:13:09 -04003431static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, unsigned int flags)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003432{
Al Viroc52a47a2013-06-15 11:15:20 +04003433 int result = -ENOENT;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003434 struct task_struct *task;
3435 struct task_struct *leader = get_proc_task(dir);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003436 unsigned tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003437 struct pid_namespace *ns;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003438
3439 if (!leader)
3440 goto out_no_task;
3441
Alexey Dobriyandbcdb502014-08-08 14:21:25 -07003442 tid = name_to_int(&dentry->d_name);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003443 if (tid == ~0U)
3444 goto out;
3445
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003446 ns = dentry->d_sb->s_fs_info;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003447 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003448 task = find_task_by_pid_ns(tid, ns);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003449 if (task)
3450 get_task_struct(task);
3451 rcu_read_unlock();
3452 if (!task)
3453 goto out;
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07003454 if (!same_thread_group(leader, task))
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003455 goto out_drop_task;
3456
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003457 result = proc_task_instantiate(dir, dentry, task, NULL);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003458out_drop_task:
3459 put_task_struct(task);
3460out:
3461 put_task_struct(leader);
3462out_no_task:
Al Viroc52a47a2013-06-15 11:15:20 +04003463 return ERR_PTR(result);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003464}
3465
3466/*
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003467 * Find the first tid of a thread group to return to user space.
3468 *
3469 * Usually this is just the thread group leader, but if the users
3470 * buffer was too small or there was a seek into the middle of the
3471 * directory we have more work todo.
3472 *
3473 * In the case of a short read we start with find_task_by_pid.
3474 *
3475 * In the case of a seek we start with the leader and walk nr
3476 * threads past it.
3477 */
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003478static struct task_struct *first_tid(struct pid *pid, int tid, loff_t f_pos,
3479 struct pid_namespace *ns)
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003480{
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003481 struct task_struct *pos, *task;
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003482 unsigned long nr = f_pos;
3483
3484 if (nr != f_pos) /* 32bit overflow? */
3485 return NULL;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003486
Eric W. Biedermancc288732006-06-26 00:26:01 -07003487 rcu_read_lock();
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003488 task = pid_task(pid, PIDTYPE_PID);
3489 if (!task)
3490 goto fail;
3491
3492 /* Attempt to start with the tid of a thread */
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003493 if (tid && nr) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003494 pos = find_task_by_pid_ns(tid, ns);
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003495 if (pos && same_thread_group(pos, task))
Oleg Nesterova872ff02006-06-26 00:26:01 -07003496 goto found;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003497 }
3498
3499 /* If nr exceeds the number of threads there is nothing todo */
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003500 if (nr >= get_nr_threads(task))
Oleg Nesterovc986c142014-01-23 15:55:38 -08003501 goto fail;
Oleg Nesterova872ff02006-06-26 00:26:01 -07003502
3503 /* If we haven't found our starting place yet start
3504 * with the leader and walk nr threads forward.
3505 */
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003506 pos = task = task->group_leader;
Oleg Nesterovc986c142014-01-23 15:55:38 -08003507 do {
Oleg Nesterov9f6e9632014-01-23 15:55:40 -08003508 if (!nr--)
Oleg Nesterovc986c142014-01-23 15:55:38 -08003509 goto found;
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003510 } while_each_thread(task, pos);
Oleg Nesterovc986c142014-01-23 15:55:38 -08003511fail:
3512 pos = NULL;
3513 goto out;
Oleg Nesterova872ff02006-06-26 00:26:01 -07003514found:
3515 get_task_struct(pos);
3516out:
Eric W. Biedermancc288732006-06-26 00:26:01 -07003517 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003518 return pos;
3519}
3520
3521/*
3522 * Find the next thread in the thread list.
3523 * Return NULL if there is an error or no next thread.
3524 *
3525 * The reference to the input task_struct is released.
3526 */
3527static struct task_struct *next_tid(struct task_struct *start)
3528{
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003529 struct task_struct *pos = NULL;
Eric W. Biedermancc288732006-06-26 00:26:01 -07003530 rcu_read_lock();
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003531 if (pid_alive(start)) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003532 pos = next_thread(start);
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003533 if (thread_group_leader(pos))
3534 pos = NULL;
3535 else
3536 get_task_struct(pos);
3537 }
Eric W. Biedermancc288732006-06-26 00:26:01 -07003538 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003539 put_task_struct(start);
3540 return pos;
3541}
3542
Linus Torvalds1da177e2005-04-16 15:20:36 -07003543/* for the /proc/TGID/task/ directories */
Al Virof0c3b502013-05-16 12:07:31 -04003544static int proc_task_readdir(struct file *file, struct dir_context *ctx)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003545{
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003546 struct inode *inode = file_inode(file);
3547 struct task_struct *task;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003548 struct pid_namespace *ns;
Al Virof0c3b502013-05-16 12:07:31 -04003549 int tid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003550
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003551 if (proc_inode_is_dead(inode))
Al Virof0c3b502013-05-16 12:07:31 -04003552 return -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003553
Al Virof0c3b502013-05-16 12:07:31 -04003554 if (!dir_emit_dots(file, ctx))
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003555 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003556
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003557 /* f_version caches the tgid value that the last readdir call couldn't
3558 * return. lseek aka telldir automagically resets f_version to 0.
3559 */
Al Viro3aa33772014-10-31 00:42:35 -04003560 ns = inode->i_sb->s_fs_info;
Al Virof0c3b502013-05-16 12:07:31 -04003561 tid = (int)file->f_version;
3562 file->f_version = 0;
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003563 for (task = first_tid(proc_pid(inode), tid, ctx->pos - 2, ns);
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003564 task;
Al Virof0c3b502013-05-16 12:07:31 -04003565 task = next_tid(task), ctx->pos++) {
3566 char name[PROC_NUMBUF];
3567 int len;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003568 tid = task_pid_nr_ns(task, ns);
Al Virof0c3b502013-05-16 12:07:31 -04003569 len = snprintf(name, sizeof(name), "%d", tid);
3570 if (!proc_fill_cache(file, ctx, name, len,
3571 proc_task_instantiate, task, NULL)) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003572 /* returning this tgid failed, save it as the first
3573 * pid for the next readir call */
Al Virof0c3b502013-05-16 12:07:31 -04003574 file->f_version = (u64)tid;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003575 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003576 break;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003577 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003578 }
Oleg Nesterovd855a4b2014-01-23 15:55:39 -08003579
Al Virof0c3b502013-05-16 12:07:31 -04003580 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003581}
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003582
David Howellsa528d352017-01-31 16:46:22 +00003583static int proc_task_getattr(const struct path *path, struct kstat *stat,
3584 u32 request_mask, unsigned int query_flags)
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003585{
David Howellsa528d352017-01-31 16:46:22 +00003586 struct inode *inode = d_inode(path->dentry);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003587 struct task_struct *p = get_proc_task(inode);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003588 generic_fillattr(inode, stat);
3589
Eric W. Biederman99f89552006-06-26 00:25:55 -07003590 if (p) {
Eric W. Biederman99f89552006-06-26 00:25:55 -07003591 stat->nlink += get_nr_threads(p);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003592 put_task_struct(p);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003593 }
3594
3595 return 0;
3596}
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003597
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003598static const struct inode_operations proc_task_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003599 .lookup = proc_task_lookup,
3600 .getattr = proc_task_getattr,
3601 .setattr = proc_setattr,
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003602 .permission = proc_pid_permission,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003603};
3604
Arjan van de Ven00977a52007-02-12 00:55:34 -08003605static const struct file_operations proc_task_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003606 .read = generic_read_dir,
Al Virof50752e2016-04-20 17:13:54 -04003607 .iterate_shared = proc_task_readdir,
3608 .llseek = generic_file_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003609};
Alexey Dobriyan1270dd82016-12-12 16:45:32 -08003610
3611void __init set_proc_pid_nlink(void)
3612{
3613 nlink_tid = pid_entry_nlink(tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3614 nlink_tgid = pid_entry_nlink(tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3615}