blob: 851ba3dcdc290ab6b750793c2840bdffac755de7 [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
50#include <asm/uaccess.h>
51
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>
Paul Menagea4243162007-10-18 23:39:35 -070076#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070077#include <linux/cpuset.h>
78#include <linux/audit.h>
Al Viro5addc5d2005-11-07 17:15:49 -050079#include <linux/poll.h>
Serge E. Hallyn1651e142006-10-02 02:18:08 -070080#include <linux/nsproxy.h>
Alexey Dobriyan8ac773b2006-10-19 23:28:32 -070081#include <linux/oom.h>
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -070082#include <linux/elf.h>
Pavel Emelyanov60347f62007-10-18 23:40:03 -070083#include <linux/pid_namespace.h>
Al Viro5ad4e532009-03-29 19:50:06 -040084#include <linux/fs_struct.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090085#include <linux/slab.h>
Chris Metcalff133ecc2011-05-26 12:40:09 -040086#ifdef CONFIG_HARDWALL
87#include <asm/hardwall.h>
88#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070089#include "internal.h"
90
Eric W. Biederman0f2fe202006-06-26 00:25:46 -070091/* NOTE:
92 * Implementing inode permission operations in /proc is almost
93 * certainly an error. Permission checks need to happen during
94 * each system call not at open time. The reason is that most of
95 * what we wish to check for permissions in /proc varies at runtime.
96 *
97 * The classic example of a problem is opening file descriptors
98 * in /proc for a task before it execs a suid executable.
99 */
100
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101struct pid_entry {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102 char *name;
Eric Dumazetc5141e62007-05-08 00:26:15 -0700103 int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104 mode_t mode;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800105 const struct inode_operations *iop;
Arjan van de Ven00977a52007-02-12 00:55:34 -0800106 const struct file_operations *fop;
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700107 union proc_op op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108};
109
Eric W. Biederman61a28782006-10-02 02:18:49 -0700110#define NOD(NAME, MODE, IOP, FOP, OP) { \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700111 .name = (NAME), \
Eric Dumazetc5141e62007-05-08 00:26:15 -0700112 .len = sizeof(NAME) - 1, \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700113 .mode = MODE, \
114 .iop = IOP, \
115 .fop = FOP, \
116 .op = OP, \
117}
118
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300119#define DIR(NAME, MODE, iops, fops) \
120 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
121#define LNK(NAME, get_link) \
Eric W. Biederman61a28782006-10-02 02:18:49 -0700122 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700123 &proc_pid_link_inode_operations, NULL, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300124 { .proc_get_link = get_link } )
125#define REG(NAME, MODE, fops) \
126 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
127#define INF(NAME, MODE, read) \
Eric W. Biederman61a28782006-10-02 02:18:49 -0700128 NOD(NAME, (S_IFREG|(MODE)), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700129 NULL, &proc_info_file_operations, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300130 { .proc_read = read } )
131#define ONE(NAME, MODE, show) \
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800132 NOD(NAME, (S_IFREG|(MODE)), \
133 NULL, &proc_single_file_operations, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300134 { .proc_show = show } )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135
Vegard Nossumaed54172008-06-05 22:46:53 -0700136/*
137 * Count the number of hardlinks for the pid_entry table, excluding the .
138 * and .. links.
139 */
140static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
141 unsigned int n)
142{
143 unsigned int i;
144 unsigned int count;
145
146 count = 0;
147 for (i = 0; i < n; ++i) {
148 if (S_ISDIR(entries[i].mode))
149 ++count;
150 }
151
152 return count;
153}
154
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200155static int get_task_root(struct task_struct *task, struct path *root)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156{
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000157 int result = -ENOENT;
158
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700159 task_lock(task);
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200160 if (task->fs) {
161 get_fs_root(task->fs, root);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000162 result = 0;
163 }
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700164 task_unlock(task);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000165 return result;
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700166}
167
Jan Blunck3dcd25f2008-02-14 19:38:35 -0800168static int proc_cwd_link(struct inode *inode, struct path *path)
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700169{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700170 struct task_struct *task = get_proc_task(inode);
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700171 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700172
173 if (task) {
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200174 task_lock(task);
175 if (task->fs) {
176 get_fs_pwd(task->fs, path);
177 result = 0;
178 }
179 task_unlock(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700180 put_task_struct(task);
181 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 return result;
183}
184
Jan Blunck3dcd25f2008-02-14 19:38:35 -0800185static int proc_root_link(struct inode *inode, struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700187 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700189
190 if (task) {
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200191 result = get_task_root(task, path);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700192 put_task_struct(task);
193 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 return result;
195}
196
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400197static struct mm_struct *__check_mem_permission(struct task_struct *task)
Roland McGrath638fa202008-04-29 01:01:38 -0700198{
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400199 struct mm_struct *mm;
200
201 mm = get_task_mm(task);
202 if (!mm)
203 return ERR_PTR(-EINVAL);
204
Roland McGrath638fa202008-04-29 01:01:38 -0700205 /*
206 * A task can always look at itself, in case it chooses
207 * to use system calls instead of load instructions.
208 */
209 if (task == current)
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400210 return mm;
Roland McGrath638fa202008-04-29 01:01:38 -0700211
212 /*
213 * If current is actively ptrace'ing, and would also be
214 * permitted to freshly attach with ptrace now, permit it.
215 */
Roland McGrath0d094ef2008-07-25 19:45:49 -0700216 if (task_is_stopped_or_traced(task)) {
217 int match;
218 rcu_read_lock();
Tejun Heo06d98472011-06-17 16:50:40 +0200219 match = (ptrace_parent(task) == current);
Roland McGrath0d094ef2008-07-25 19:45:49 -0700220 rcu_read_unlock();
221 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400222 return mm;
Roland McGrath0d094ef2008-07-25 19:45:49 -0700223 }
Roland McGrath638fa202008-04-29 01:01:38 -0700224
225 /*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300226 * No one else is allowed.
Roland McGrath638fa202008-04-29 01:01:38 -0700227 */
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400228 mmput(mm);
229 return ERR_PTR(-EPERM);
Roland McGrath638fa202008-04-29 01:01:38 -0700230}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231
Stephen Wilson18f661b2011-03-13 15:49:22 -0400232/*
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400233 * If current may access user memory in @task return a reference to the
234 * corresponding mm, otherwise ERR_PTR.
Stephen Wilson18f661b2011-03-13 15:49:22 -0400235 */
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400236static struct mm_struct *check_mem_permission(struct task_struct *task)
Stephen Wilson18f661b2011-03-13 15:49:22 -0400237{
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400238 struct mm_struct *mm;
Stephen Wilson18f661b2011-03-13 15:49:22 -0400239 int err;
240
241 /*
242 * Avoid racing if task exec's as we might get a new mm but validate
243 * against old credentials.
244 */
245 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
246 if (err)
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400247 return ERR_PTR(err);
Stephen Wilson18f661b2011-03-13 15:49:22 -0400248
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400249 mm = __check_mem_permission(task);
Stephen Wilson18f661b2011-03-13 15:49:22 -0400250 mutex_unlock(&task->signal->cred_guard_mutex);
251
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400252 return mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253}
254
Al Viro831830b2008-01-02 14:09:57 +0000255struct mm_struct *mm_for_maps(struct task_struct *task)
256{
Oleg Nesterov704b8362009-07-10 03:27:40 +0200257 struct mm_struct *mm;
Al Viroec6fd8a2011-02-15 22:22:54 -0500258 int err;
Oleg Nesterov00f89d22009-07-10 03:27:38 +0200259
Al Viroec6fd8a2011-02-15 22:22:54 -0500260 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
261 if (err)
262 return ERR_PTR(err);
Oleg Nesterov704b8362009-07-10 03:27:40 +0200263
264 mm = get_task_mm(task);
265 if (mm && mm != current->mm &&
266 !ptrace_may_access(task, PTRACE_MODE_READ)) {
267 mmput(mm);
Al Viroec6fd8a2011-02-15 22:22:54 -0500268 mm = ERR_PTR(-EACCES);
Oleg Nesterov13f0fea2009-06-23 21:25:32 +0200269 }
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -0700270 mutex_unlock(&task->signal->cred_guard_mutex);
Oleg Nesterov704b8362009-07-10 03:27:40 +0200271
Al Viro831830b2008-01-02 14:09:57 +0000272 return mm;
Al Viro831830b2008-01-02 14:09:57 +0000273}
274
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275static int proc_pid_cmdline(struct task_struct *task, char * buffer)
276{
277 int res = 0;
278 unsigned int len;
279 struct mm_struct *mm = get_task_mm(task);
280 if (!mm)
281 goto out;
282 if (!mm->arg_end)
283 goto out_mm; /* Shh! No looking before we're done */
284
285 len = mm->arg_end - mm->arg_start;
286
287 if (len > PAGE_SIZE)
288 len = PAGE_SIZE;
289
290 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
291
292 // If the nul at the end of args has been overwritten, then
293 // assume application is using setproctitle(3).
294 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
295 len = strnlen(buffer, res);
296 if (len < res) {
297 res = len;
298 } else {
299 len = mm->env_end - mm->env_start;
300 if (len > PAGE_SIZE - res)
301 len = PAGE_SIZE - res;
302 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
303 res = strnlen(buffer, res);
304 }
305 }
306out_mm:
307 mmput(mm);
308out:
309 return res;
310}
311
312static int proc_pid_auxv(struct task_struct *task, char *buffer)
313{
Al Viro2fadaef2011-02-15 22:52:11 -0500314 struct mm_struct *mm = mm_for_maps(task);
315 int res = PTR_ERR(mm);
316 if (mm && !IS_ERR(mm)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 unsigned int nwords = 0;
Hannes Ederdfe6b7d2008-12-30 18:49:13 +0300318 do {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 nwords += 2;
Hannes Ederdfe6b7d2008-12-30 18:49:13 +0300320 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 res = nwords * sizeof(mm->saved_auxv[0]);
322 if (res > PAGE_SIZE)
323 res = PAGE_SIZE;
324 memcpy(buffer, mm->saved_auxv, res);
325 mmput(mm);
326 }
327 return res;
328}
329
330
331#ifdef CONFIG_KALLSYMS
332/*
333 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
334 * Returns the resolved symbol. If that fails, simply return the address.
335 */
336static int proc_pid_wchan(struct task_struct *task, char *buffer)
337{
Alexey Dobriyanffb45122007-05-08 00:28:41 -0700338 unsigned long wchan;
Tejun Heo9281ace2007-07-17 04:03:51 -0700339 char symname[KSYM_NAME_LEN];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340
341 wchan = get_wchan(task);
342
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700343 if (lookup_symbol_name(wchan, symname) < 0)
Jake Edgef83ce3e2009-05-04 12:51:14 -0600344 if (!ptrace_may_access(task, PTRACE_MODE_READ))
345 return 0;
346 else
347 return sprintf(buffer, "%lu", wchan);
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700348 else
349 return sprintf(buffer, "%s", symname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350}
351#endif /* CONFIG_KALLSYMS */
352
Al Viroa9712bc2011-03-23 15:52:50 -0400353static int lock_trace(struct task_struct *task)
354{
355 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
356 if (err)
357 return err;
358 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
359 mutex_unlock(&task->signal->cred_guard_mutex);
360 return -EPERM;
361 }
362 return 0;
363}
364
365static void unlock_trace(struct task_struct *task)
366{
367 mutex_unlock(&task->signal->cred_guard_mutex);
368}
369
Ken Chen2ec220e2008-11-10 11:26:08 +0300370#ifdef CONFIG_STACKTRACE
371
372#define MAX_STACK_TRACE_DEPTH 64
373
374static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
375 struct pid *pid, struct task_struct *task)
376{
377 struct stack_trace trace;
378 unsigned long *entries;
Al Viroa9712bc2011-03-23 15:52:50 -0400379 int err;
Ken Chen2ec220e2008-11-10 11:26:08 +0300380 int i;
381
382 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
383 if (!entries)
384 return -ENOMEM;
385
386 trace.nr_entries = 0;
387 trace.max_entries = MAX_STACK_TRACE_DEPTH;
388 trace.entries = entries;
389 trace.skip = 0;
Ken Chen2ec220e2008-11-10 11:26:08 +0300390
Al Viroa9712bc2011-03-23 15:52:50 -0400391 err = lock_trace(task);
392 if (!err) {
393 save_stack_trace_tsk(task, &trace);
394
395 for (i = 0; i < trace.nr_entries; i++) {
Linus Torvaldsb81a6182011-03-23 20:51:42 -0700396 seq_printf(m, "[<%pK>] %pS\n",
Al Viroa9712bc2011-03-23 15:52:50 -0400397 (void *)entries[i], (void *)entries[i]);
398 }
399 unlock_trace(task);
Ken Chen2ec220e2008-11-10 11:26:08 +0300400 }
401 kfree(entries);
402
Al Viroa9712bc2011-03-23 15:52:50 -0400403 return err;
Ken Chen2ec220e2008-11-10 11:26:08 +0300404}
405#endif
406
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407#ifdef CONFIG_SCHEDSTATS
408/*
409 * Provides /proc/PID/schedstat
410 */
411static int proc_pid_schedstat(struct task_struct *task, char *buffer)
412{
Balbir Singh172ba842007-07-09 18:52:00 +0200413 return sprintf(buffer, "%llu %llu %lu\n",
Ingo Molnar826e08b2008-12-22 07:37:41 +0100414 (unsigned long long)task->se.sum_exec_runtime,
415 (unsigned long long)task->sched_info.run_delay,
Ingo Molnar2d723762007-10-15 17:00:12 +0200416 task->sched_info.pcount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417}
418#endif
419
Arjan van de Ven97455122008-01-25 21:08:34 +0100420#ifdef CONFIG_LATENCYTOP
421static int lstats_show_proc(struct seq_file *m, void *v)
422{
423 int i;
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800424 struct inode *inode = m->private;
425 struct task_struct *task = get_proc_task(inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100426
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800427 if (!task)
428 return -ESRCH;
429 seq_puts(m, "Latency Top version : v0.1\n");
Arjan van de Ven97455122008-01-25 21:08:34 +0100430 for (i = 0; i < 32; i++) {
Joe Perches34e49d42011-01-12 17:00:30 -0800431 struct latency_record *lr = &task->latency_record[i];
432 if (lr->backtrace[0]) {
Arjan van de Ven97455122008-01-25 21:08:34 +0100433 int q;
Joe Perches34e49d42011-01-12 17:00:30 -0800434 seq_printf(m, "%i %li %li",
435 lr->count, lr->time, lr->max);
Arjan van de Ven97455122008-01-25 21:08:34 +0100436 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
Joe Perches34e49d42011-01-12 17:00:30 -0800437 unsigned long bt = lr->backtrace[q];
438 if (!bt)
Arjan van de Ven97455122008-01-25 21:08:34 +0100439 break;
Joe Perches34e49d42011-01-12 17:00:30 -0800440 if (bt == ULONG_MAX)
Arjan van de Ven97455122008-01-25 21:08:34 +0100441 break;
Joe Perches34e49d42011-01-12 17:00:30 -0800442 seq_printf(m, " %ps", (void *)bt);
Arjan van de Ven97455122008-01-25 21:08:34 +0100443 }
Alexey Dobriyan9d6de122011-01-12 17:00:32 -0800444 seq_putc(m, '\n');
Arjan van de Ven97455122008-01-25 21:08:34 +0100445 }
446
447 }
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800448 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100449 return 0;
450}
451
452static int lstats_open(struct inode *inode, struct file *file)
453{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800454 return single_open(file, lstats_show_proc, inode);
Hiroshi Shimamotod6643d12008-02-14 10:27:00 -0800455}
456
Arjan van de Ven97455122008-01-25 21:08:34 +0100457static ssize_t lstats_write(struct file *file, const char __user *buf,
458 size_t count, loff_t *offs)
459{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800460 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100461
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800462 if (!task)
463 return -ESRCH;
Arjan van de Ven97455122008-01-25 21:08:34 +0100464 clear_all_latency_tracing(task);
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800465 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100466
467 return count;
468}
469
470static const struct file_operations proc_lstats_operations = {
471 .open = lstats_open,
472 .read = seq_read,
473 .write = lstats_write,
474 .llseek = seq_lseek,
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800475 .release = single_release,
Arjan van de Ven97455122008-01-25 21:08:34 +0100476};
477
478#endif
479
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480static int proc_oom_score(struct task_struct *task, char *buffer)
481{
Oleg Nesterovb95c35e2010-04-01 15:13:57 +0200482 unsigned long points = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700484 read_lock(&tasklist_lock);
Oleg Nesterovb95c35e2010-04-01 15:13:57 +0200485 if (pid_alive(task))
David Rientjesa63d83f2010-08-09 17:19:46 -0700486 points = oom_badness(task, NULL, NULL,
487 totalram_pages + total_swap_pages);
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700488 read_unlock(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 return sprintf(buffer, "%lu\n", points);
490}
491
Neil Hormand85f50d2007-10-18 23:40:37 -0700492struct limit_names {
493 char *name;
494 char *unit;
495};
496
497static const struct limit_names lnames[RLIM_NLIMITS] = {
Kees Cookcff4edb2009-09-22 16:45:32 -0700498 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700499 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
500 [RLIMIT_DATA] = {"Max data size", "bytes"},
501 [RLIMIT_STACK] = {"Max stack size", "bytes"},
502 [RLIMIT_CORE] = {"Max core file size", "bytes"},
503 [RLIMIT_RSS] = {"Max resident set", "bytes"},
504 [RLIMIT_NPROC] = {"Max processes", "processes"},
505 [RLIMIT_NOFILE] = {"Max open files", "files"},
506 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
507 [RLIMIT_AS] = {"Max address space", "bytes"},
508 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
509 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
510 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
511 [RLIMIT_NICE] = {"Max nice priority", NULL},
512 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
Eugene Teo88081172008-02-23 15:23:52 -0800513 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700514};
515
516/* Display limits for a process */
517static int proc_pid_limits(struct task_struct *task, char *buffer)
518{
519 unsigned int i;
520 int count = 0;
521 unsigned long flags;
522 char *bufptr = buffer;
523
524 struct rlimit rlim[RLIM_NLIMITS];
525
Lai Jiangshana6bebbc2008-10-05 00:51:15 +0400526 if (!lock_task_sighand(task, &flags))
Neil Hormand85f50d2007-10-18 23:40:37 -0700527 return 0;
Neil Hormand85f50d2007-10-18 23:40:37 -0700528 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
529 unlock_task_sighand(task, &flags);
Neil Hormand85f50d2007-10-18 23:40:37 -0700530
531 /*
532 * print the file header
533 */
534 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
535 "Limit", "Soft Limit", "Hard Limit", "Units");
536
537 for (i = 0; i < RLIM_NLIMITS; i++) {
538 if (rlim[i].rlim_cur == RLIM_INFINITY)
539 count += sprintf(&bufptr[count], "%-25s %-20s ",
540 lnames[i].name, "unlimited");
541 else
542 count += sprintf(&bufptr[count], "%-25s %-20lu ",
543 lnames[i].name, rlim[i].rlim_cur);
544
545 if (rlim[i].rlim_max == RLIM_INFINITY)
546 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
547 else
548 count += sprintf(&bufptr[count], "%-20lu ",
549 rlim[i].rlim_max);
550
551 if (lnames[i].unit)
552 count += sprintf(&bufptr[count], "%-10s\n",
553 lnames[i].unit);
554 else
555 count += sprintf(&bufptr[count], "\n");
556 }
557
558 return count;
559}
560
Roland McGrathebcb6732008-07-25 19:46:00 -0700561#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
562static int proc_pid_syscall(struct task_struct *task, char *buffer)
563{
564 long nr;
565 unsigned long args[6], sp, pc;
Al Viroa9712bc2011-03-23 15:52:50 -0400566 int res = lock_trace(task);
567 if (res)
568 return res;
Roland McGrathebcb6732008-07-25 19:46:00 -0700569
570 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
Al Viroa9712bc2011-03-23 15:52:50 -0400571 res = sprintf(buffer, "running\n");
572 else if (nr < 0)
573 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
574 else
575 res = sprintf(buffer,
Roland McGrathebcb6732008-07-25 19:46:00 -0700576 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
577 nr,
578 args[0], args[1], args[2], args[3], args[4], args[5],
579 sp, pc);
Al Viroa9712bc2011-03-23 15:52:50 -0400580 unlock_trace(task);
581 return res;
Roland McGrathebcb6732008-07-25 19:46:00 -0700582}
583#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
584
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585/************************************************************************/
586/* Here the fs part begins */
587/************************************************************************/
588
589/* permission checks */
Eric W. Biederman778c1142006-06-26 00:25:58 -0700590static int proc_fd_access_allowed(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591{
Eric W. Biederman778c1142006-06-26 00:25:58 -0700592 struct task_struct *task;
593 int allowed = 0;
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700594 /* Allow access to a task's file descriptors if it is us or we
595 * may use ptrace attach to the process and find out that
596 * information.
Eric W. Biederman778c1142006-06-26 00:25:58 -0700597 */
598 task = get_proc_task(inode);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700599 if (task) {
Stephen Smalley006ebb42008-05-19 08:32:49 -0400600 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
Eric W. Biederman778c1142006-06-26 00:25:58 -0700601 put_task_struct(task);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700602 }
Eric W. Biederman778c1142006-06-26 00:25:58 -0700603 return allowed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604}
605
Eric W. Biederman6b4e3062010-03-07 16:41:34 -0800606int proc_setattr(struct dentry *dentry, struct iattr *attr)
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700607{
608 int error;
609 struct inode *inode = dentry->d_inode;
610
611 if (attr->ia_valid & ATTR_MODE)
612 return -EPERM;
613
614 error = inode_change_ok(inode, attr);
Christoph Hellwig10257742010-06-04 11:30:02 +0200615 if (error)
616 return error;
617
618 if ((attr->ia_valid & ATTR_SIZE) &&
619 attr->ia_size != i_size_read(inode)) {
620 error = vmtruncate(inode, attr->ia_size);
621 if (error)
622 return error;
623 }
624
625 setattr_copy(inode, attr);
626 mark_inode_dirty(inode);
627 return 0;
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700628}
629
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800630static const struct inode_operations proc_def_inode_operations = {
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700631 .setattr = proc_setattr,
632};
633
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100634static int mounts_open_common(struct inode *inode, struct file *file,
635 const struct seq_operations *op)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700637 struct task_struct *task = get_proc_task(inode);
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700638 struct nsproxy *nsp;
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800639 struct mnt_namespace *ns = NULL;
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100640 struct path root;
Al Viro5addc5d2005-11-07 17:15:49 -0500641 struct proc_mounts *p;
642 int ret = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643
Eric W. Biederman99f89552006-06-26 00:25:55 -0700644 if (task) {
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700645 rcu_read_lock();
646 nsp = task_nsproxy(task);
647 if (nsp) {
648 ns = nsp->mnt_ns;
Alexey Dobriyan863c4702007-01-26 00:56:53 -0800649 if (ns)
650 get_mnt_ns(ns);
651 }
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700652 rcu_read_unlock();
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200653 if (ns && get_task_root(task, &root) == 0)
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000654 ret = 0;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700655 put_task_struct(task);
656 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100658 if (!ns)
659 goto err;
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000660 if (ret)
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100661 goto err_put_ns;
662
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100663 ret = -ENOMEM;
664 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
665 if (!p)
666 goto err_put_path;
667
668 file->private_data = &p->m;
669 ret = seq_open(file, op);
670 if (ret)
671 goto err_free;
672
673 p->m.private = p;
674 p->ns = ns;
675 p->root = root;
Kay Sieversf1514632011-07-12 20:48:39 +0200676 p->m.poll_event = ns->event;
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100677
678 return 0;
679
680 err_free:
681 kfree(p);
682 err_put_path:
683 path_put(&root);
684 err_put_ns:
685 put_mnt_ns(ns);
686 err:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 return ret;
688}
689
690static int mounts_release(struct inode *inode, struct file *file)
691{
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100692 struct proc_mounts *p = file->private_data;
693 path_put(&p->root);
694 put_mnt_ns(p->ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 return seq_release(inode, file);
696}
697
Al Viro5addc5d2005-11-07 17:15:49 -0500698static unsigned mounts_poll(struct file *file, poll_table *wait)
699{
700 struct proc_mounts *p = file->private_data;
KOSAKI Motohiro31b07092009-04-09 13:57:59 +0900701 unsigned res = POLLIN | POLLRDNORM;
Al Viro5addc5d2005-11-07 17:15:49 -0500702
Al Viro9f5596a2010-02-05 00:40:25 -0500703 poll_wait(file, &p->ns->poll, wait);
704 if (mnt_had_events(p))
KOSAKI Motohiro31b07092009-04-09 13:57:59 +0900705 res |= POLLERR | POLLPRI;
Al Viro5addc5d2005-11-07 17:15:49 -0500706
707 return res;
708}
709
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100710static int mounts_open(struct inode *inode, struct file *file)
711{
712 return mounts_open_common(inode, file, &mounts_op);
713}
714
Arjan van de Ven00977a52007-02-12 00:55:34 -0800715static const struct file_operations proc_mounts_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 .open = mounts_open,
717 .read = seq_read,
718 .llseek = seq_lseek,
719 .release = mounts_release,
Al Viro5addc5d2005-11-07 17:15:49 -0500720 .poll = mounts_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721};
722
Ram Pai2d4d4862008-03-27 13:06:25 +0100723static int mountinfo_open(struct inode *inode, struct file *file)
724{
725 return mounts_open_common(inode, file, &mountinfo_op);
726}
727
728static const struct file_operations proc_mountinfo_operations = {
729 .open = mountinfo_open,
730 .read = seq_read,
731 .llseek = seq_lseek,
732 .release = mounts_release,
733 .poll = mounts_poll,
734};
735
Chuck Leverb4629fe2006-03-20 13:44:12 -0500736static int mountstats_open(struct inode *inode, struct file *file)
737{
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100738 return mounts_open_common(inode, file, &mountstats_op);
Chuck Leverb4629fe2006-03-20 13:44:12 -0500739}
740
Arjan van de Ven00977a52007-02-12 00:55:34 -0800741static const struct file_operations proc_mountstats_operations = {
Chuck Leverb4629fe2006-03-20 13:44:12 -0500742 .open = mountstats_open,
743 .read = seq_read,
744 .llseek = seq_lseek,
745 .release = mounts_release,
746};
747
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
749
750static ssize_t proc_info_read(struct file * file, char __user * buf,
751 size_t count, loff_t *ppos)
752{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800753 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 unsigned long page;
755 ssize_t length;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700756 struct task_struct *task = get_proc_task(inode);
757
758 length = -ESRCH;
759 if (!task)
760 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761
762 if (count > PROC_BLOCK_SIZE)
763 count = PROC_BLOCK_SIZE;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700764
765 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700766 if (!(page = __get_free_page(GFP_TEMPORARY)))
Eric W. Biederman99f89552006-06-26 00:25:55 -0700767 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768
769 length = PROC_I(inode)->op.proc_read(task, (char*)page);
770
771 if (length >= 0)
772 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
773 free_page(page);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700774out:
775 put_task_struct(task);
776out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 return length;
778}
779
Arjan van de Ven00977a52007-02-12 00:55:34 -0800780static const struct file_operations proc_info_file_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781 .read = proc_info_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +0100782 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783};
784
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800785static int proc_single_show(struct seq_file *m, void *v)
786{
787 struct inode *inode = m->private;
788 struct pid_namespace *ns;
789 struct pid *pid;
790 struct task_struct *task;
791 int ret;
792
793 ns = inode->i_sb->s_fs_info;
794 pid = proc_pid(inode);
795 task = get_pid_task(pid, PIDTYPE_PID);
796 if (!task)
797 return -ESRCH;
798
799 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
800
801 put_task_struct(task);
802 return ret;
803}
804
805static int proc_single_open(struct inode *inode, struct file *filp)
806{
Jovi Zhangc6a34052011-01-12 17:00:34 -0800807 return single_open(filp, proc_single_show, inode);
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800808}
809
810static const struct file_operations proc_single_file_operations = {
811 .open = proc_single_open,
812 .read = seq_read,
813 .llseek = seq_lseek,
814 .release = single_release,
815};
816
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817static int mem_open(struct inode* inode, struct file* file)
818{
819 file->private_data = (void*)((long)current->self_exec_id);
KAMEZAWA Hiroyuki4a3956c2010-10-01 14:20:22 -0700820 /* OK to pass negative loff_t, we can catch out-of-range */
821 file->f_mode |= FMODE_UNSIGNED_OFFSET;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 return 0;
823}
824
825static ssize_t mem_read(struct file * file, char __user * buf,
826 size_t count, loff_t *ppos)
827{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800828 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 char *page;
830 unsigned long src = *ppos;
831 int ret = -ESRCH;
832 struct mm_struct *mm;
833
Eric W. Biederman99f89552006-06-26 00:25:55 -0700834 if (!task)
835 goto out_no_task;
836
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837 ret = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700838 page = (char *)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 if (!page)
840 goto out;
841
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400842 mm = check_mem_permission(task);
843 ret = PTR_ERR(mm);
844 if (IS_ERR(mm))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 goto out_free;
846
847 ret = -EIO;
848
849 if (file->private_data != (void*)((long)current->self_exec_id))
850 goto out_put;
851
852 ret = 0;
853
854 while (count > 0) {
855 int this_len, retval;
856
857 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400858 retval = access_remote_vm(mm, src, page, this_len, 0);
859 if (!retval) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 if (!ret)
861 ret = -EIO;
862 break;
863 }
864
865 if (copy_to_user(buf, page, retval)) {
866 ret = -EFAULT;
867 break;
868 }
869
870 ret += retval;
871 src += retval;
872 buf += retval;
873 count -= retval;
874 }
875 *ppos = src;
876
877out_put:
878 mmput(mm);
879out_free:
880 free_page((unsigned long) page);
881out:
Eric W. Biederman99f89552006-06-26 00:25:55 -0700882 put_task_struct(task);
883out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 return ret;
885}
886
Glauber de Oliveira Costa63967fa2007-02-20 13:58:12 -0800887static ssize_t mem_write(struct file * file, const char __user *buf,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 size_t count, loff_t *ppos)
889{
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700890 int copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 char *page;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800892 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 unsigned long dst = *ppos;
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400894 struct mm_struct *mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895
Eric W. Biederman99f89552006-06-26 00:25:55 -0700896 copied = -ESRCH;
897 if (!task)
898 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899
Eric W. Biederman99f89552006-06-26 00:25:55 -0700900 copied = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700901 page = (char *)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 if (!page)
KOSAKI Motohiro30cd8902011-05-26 16:25:52 -0700903 goto out_task;
904
905 mm = check_mem_permission(task);
906 copied = PTR_ERR(mm);
907 if (IS_ERR(mm))
908 goto out_free;
909
910 copied = -EIO;
911 if (file->private_data != (void *)((long)current->self_exec_id))
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400912 goto out_mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700914 copied = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 while (count > 0) {
916 int this_len, retval;
917
918 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
919 if (copy_from_user(page, buf, this_len)) {
920 copied = -EFAULT;
921 break;
922 }
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400923 retval = access_remote_vm(mm, dst, page, this_len, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 if (!retval) {
925 if (!copied)
926 copied = -EIO;
927 break;
928 }
929 copied += retval;
930 buf += retval;
931 dst += retval;
932 count -= retval;
933 }
934 *ppos = dst;
KOSAKI Motohiro30cd8902011-05-26 16:25:52 -0700935
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400936out_mm:
937 mmput(mm);
KOSAKI Motohiro30cd8902011-05-26 16:25:52 -0700938out_free:
939 free_page((unsigned long) page);
Stephen Wilson8b0db9d2011-03-13 15:49:23 -0400940out_task:
Eric W. Biederman99f89552006-06-26 00:25:55 -0700941 put_task_struct(task);
942out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 return copied;
944}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945
Matt Mackall85863e42008-02-04 22:29:04 -0800946loff_t mem_lseek(struct file *file, loff_t offset, int orig)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947{
948 switch (orig) {
949 case 0:
950 file->f_pos = offset;
951 break;
952 case 1:
953 file->f_pos += offset;
954 break;
955 default:
956 return -EINVAL;
957 }
958 force_successful_syscall_return();
959 return file->f_pos;
960}
961
Arjan van de Ven00977a52007-02-12 00:55:34 -0800962static const struct file_operations proc_mem_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 .llseek = mem_lseek,
964 .read = mem_read,
965 .write = mem_write,
966 .open = mem_open,
967};
968
James Pearson315e28c2007-10-16 23:30:17 -0700969static ssize_t environ_read(struct file *file, char __user *buf,
970 size_t count, loff_t *ppos)
971{
972 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
973 char *page;
974 unsigned long src = *ppos;
975 int ret = -ESRCH;
976 struct mm_struct *mm;
977
978 if (!task)
979 goto out_no_task;
980
James Pearson315e28c2007-10-16 23:30:17 -0700981 ret = -ENOMEM;
982 page = (char *)__get_free_page(GFP_TEMPORARY);
983 if (!page)
984 goto out;
985
James Pearson315e28c2007-10-16 23:30:17 -0700986
Al Virod6f64b82011-02-15 22:26:01 -0500987 mm = mm_for_maps(task);
988 ret = PTR_ERR(mm);
989 if (!mm || IS_ERR(mm))
James Pearson315e28c2007-10-16 23:30:17 -0700990 goto out_free;
991
Al Virod6f64b82011-02-15 22:26:01 -0500992 ret = 0;
James Pearson315e28c2007-10-16 23:30:17 -0700993 while (count > 0) {
994 int this_len, retval, max_len;
995
996 this_len = mm->env_end - (mm->env_start + src);
997
998 if (this_len <= 0)
999 break;
1000
1001 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
1002 this_len = (this_len > max_len) ? max_len : this_len;
1003
1004 retval = access_process_vm(task, (mm->env_start + src),
1005 page, this_len, 0);
1006
1007 if (retval <= 0) {
1008 ret = retval;
1009 break;
1010 }
1011
1012 if (copy_to_user(buf, page, retval)) {
1013 ret = -EFAULT;
1014 break;
1015 }
1016
1017 ret += retval;
1018 src += retval;
1019 buf += retval;
1020 count -= retval;
1021 }
1022 *ppos = src;
1023
1024 mmput(mm);
1025out_free:
1026 free_page((unsigned long) page);
1027out:
1028 put_task_struct(task);
1029out_no_task:
1030 return ret;
1031}
1032
1033static const struct file_operations proc_environ_operations = {
1034 .read = environ_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001035 .llseek = generic_file_llseek,
James Pearson315e28c2007-10-16 23:30:17 -07001036};
1037
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038static ssize_t oom_adjust_read(struct file *file, char __user *buf,
1039 size_t count, loff_t *ppos)
1040{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001041 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001042 char buffer[PROC_NUMBUF];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 size_t len;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001044 int oom_adjust = OOM_DISABLE;
1045 unsigned long flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046
Eric W. Biederman99f89552006-06-26 00:25:55 -07001047 if (!task)
1048 return -ESRCH;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001049
1050 if (lock_task_sighand(task, &flags)) {
1051 oom_adjust = task->signal->oom_adj;
1052 unlock_task_sighand(task, &flags);
1053 }
1054
Eric W. Biederman99f89552006-06-26 00:25:55 -07001055 put_task_struct(task);
1056
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001057 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001058
1059 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060}
1061
1062static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1063 size_t count, loff_t *ppos)
1064{
Eric W. Biederman99f89552006-06-26 00:25:55 -07001065 struct task_struct *task;
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -07001066 char buffer[PROC_NUMBUF];
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001067 int oom_adjust;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001068 unsigned long flags;
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -07001069 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001071 memset(buffer, 0, sizeof(buffer));
1072 if (count > sizeof(buffer) - 1)
1073 count = sizeof(buffer) - 1;
David Rientjes723548b2010-10-26 14:21:25 -07001074 if (copy_from_user(buffer, buf, count)) {
1075 err = -EFAULT;
1076 goto out;
1077 }
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -07001078
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001079 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -07001080 if (err)
David Rientjes723548b2010-10-26 14:21:25 -07001081 goto out;
Alexey Dobriyan8ac773b2006-10-19 23:28:32 -07001082 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
David Rientjes723548b2010-10-26 14:21:25 -07001083 oom_adjust != OOM_DISABLE) {
1084 err = -EINVAL;
1085 goto out;
1086 }
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -07001087
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001088 task = get_proc_task(file->f_path.dentry->d_inode);
David Rientjes723548b2010-10-26 14:21:25 -07001089 if (!task) {
1090 err = -ESRCH;
1091 goto out;
1092 }
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001093
Ying Han3d5992d2010-10-26 14:21:23 -07001094 task_lock(task);
1095 if (!task->mm) {
David Rientjes723548b2010-10-26 14:21:25 -07001096 err = -EINVAL;
1097 goto err_task_lock;
Ying Han3d5992d2010-10-26 14:21:23 -07001098 }
1099
David Rientjesd19d5472010-10-26 14:21:26 -07001100 if (!lock_task_sighand(task, &flags)) {
1101 err = -ESRCH;
1102 goto err_task_lock;
1103 }
1104
1105 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1106 err = -EACCES;
1107 goto err_sighand;
1108 }
1109
David Rientjes51b1bd22010-08-09 17:19:47 -07001110 /*
1111 * Warn that /proc/pid/oom_adj is deprecated, see
1112 * Documentation/feature-removal-schedule.txt.
1113 */
Linus Torvaldsc2142702011-08-06 11:43:08 -07001114 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
David Rientjesbe8f6842011-07-25 17:12:18 -07001115 current->comm, task_pid_nr(current), task_pid_nr(task),
1116 task_pid_nr(task));
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001117 task->signal->oom_adj = oom_adjust;
David Rientjesa63d83f2010-08-09 17:19:46 -07001118 /*
1119 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1120 * value is always attainable.
1121 */
1122 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1123 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1124 else
1125 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1126 -OOM_DISABLE;
David Rientjes723548b2010-10-26 14:21:25 -07001127err_sighand:
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001128 unlock_task_sighand(task, &flags);
David Rientjesd19d5472010-10-26 14:21:26 -07001129err_task_lock:
1130 task_unlock(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -07001131 put_task_struct(task);
David Rientjes723548b2010-10-26 14:21:25 -07001132out:
1133 return err < 0 ? err : count;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134}
1135
Arjan van de Ven00977a52007-02-12 00:55:34 -08001136static const struct file_operations proc_oom_adjust_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 .read = oom_adjust_read,
1138 .write = oom_adjust_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001139 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140};
1141
David Rientjesa63d83f2010-08-09 17:19:46 -07001142static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1143 size_t count, loff_t *ppos)
1144{
1145 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1146 char buffer[PROC_NUMBUF];
1147 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1148 unsigned long flags;
1149 size_t len;
1150
1151 if (!task)
1152 return -ESRCH;
1153 if (lock_task_sighand(task, &flags)) {
1154 oom_score_adj = task->signal->oom_score_adj;
1155 unlock_task_sighand(task, &flags);
1156 }
1157 put_task_struct(task);
1158 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1159 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1160}
1161
1162static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1163 size_t count, loff_t *ppos)
1164{
1165 struct task_struct *task;
1166 char buffer[PROC_NUMBUF];
1167 unsigned long flags;
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001168 int oom_score_adj;
David Rientjesa63d83f2010-08-09 17:19:46 -07001169 int err;
1170
1171 memset(buffer, 0, sizeof(buffer));
1172 if (count > sizeof(buffer) - 1)
1173 count = sizeof(buffer) - 1;
David Rientjes723548b2010-10-26 14:21:25 -07001174 if (copy_from_user(buffer, buf, count)) {
1175 err = -EFAULT;
1176 goto out;
1177 }
David Rientjesa63d83f2010-08-09 17:19:46 -07001178
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001179 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
David Rientjesa63d83f2010-08-09 17:19:46 -07001180 if (err)
David Rientjes723548b2010-10-26 14:21:25 -07001181 goto out;
David Rientjesa63d83f2010-08-09 17:19:46 -07001182 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
David Rientjes723548b2010-10-26 14:21:25 -07001183 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1184 err = -EINVAL;
1185 goto out;
1186 }
David Rientjesa63d83f2010-08-09 17:19:46 -07001187
1188 task = get_proc_task(file->f_path.dentry->d_inode);
David Rientjes723548b2010-10-26 14:21:25 -07001189 if (!task) {
1190 err = -ESRCH;
1191 goto out;
1192 }
David Rientjesa63d83f2010-08-09 17:19:46 -07001193
Ying Han3d5992d2010-10-26 14:21:23 -07001194 task_lock(task);
1195 if (!task->mm) {
David Rientjes723548b2010-10-26 14:21:25 -07001196 err = -EINVAL;
1197 goto err_task_lock;
Ying Han3d5992d2010-10-26 14:21:23 -07001198 }
David Rientjesd19d5472010-10-26 14:21:26 -07001199
1200 if (!lock_task_sighand(task, &flags)) {
1201 err = -ESRCH;
1202 goto err_task_lock;
1203 }
1204
Mandeep Singh Bainesdabb16f2011-01-13 15:46:05 -08001205 if (oom_score_adj < task->signal->oom_score_adj_min &&
David Rientjesd19d5472010-10-26 14:21:26 -07001206 !capable(CAP_SYS_RESOURCE)) {
1207 err = -EACCES;
1208 goto err_sighand;
1209 }
1210
David Rientjesa63d83f2010-08-09 17:19:46 -07001211 task->signal->oom_score_adj = oom_score_adj;
Mandeep Singh Bainesdabb16f2011-01-13 15:46:05 -08001212 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1213 task->signal->oom_score_adj_min = oom_score_adj;
David Rientjesa63d83f2010-08-09 17:19:46 -07001214 /*
1215 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1216 * always attainable.
1217 */
1218 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1219 task->signal->oom_adj = OOM_DISABLE;
1220 else
1221 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1222 OOM_SCORE_ADJ_MAX;
David Rientjes723548b2010-10-26 14:21:25 -07001223err_sighand:
David Rientjesa63d83f2010-08-09 17:19:46 -07001224 unlock_task_sighand(task, &flags);
David Rientjesd19d5472010-10-26 14:21:26 -07001225err_task_lock:
1226 task_unlock(task);
David Rientjesa63d83f2010-08-09 17:19:46 -07001227 put_task_struct(task);
David Rientjes723548b2010-10-26 14:21:25 -07001228out:
1229 return err < 0 ? err : count;
David Rientjesa63d83f2010-08-09 17:19:46 -07001230}
1231
1232static const struct file_operations proc_oom_score_adj_operations = {
1233 .read = oom_score_adj_read,
1234 .write = oom_score_adj_write,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001235 .llseek = default_llseek,
David Rientjesa63d83f2010-08-09 17:19:46 -07001236};
1237
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238#ifdef CONFIG_AUDITSYSCALL
1239#define TMPBUFLEN 21
1240static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1241 size_t count, loff_t *ppos)
1242{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001243 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001244 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 ssize_t length;
1246 char tmpbuf[TMPBUFLEN];
1247
Eric W. Biederman99f89552006-06-26 00:25:55 -07001248 if (!task)
1249 return -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
Al Viro0c11b942008-01-10 04:20:52 -05001251 audit_get_loginuid(task));
Eric W. Biederman99f89552006-06-26 00:25:55 -07001252 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1254}
1255
1256static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1257 size_t count, loff_t *ppos)
1258{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001259 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 char *page, *tmp;
1261 ssize_t length;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262 uid_t loginuid;
1263
1264 if (!capable(CAP_AUDIT_CONTROL))
1265 return -EPERM;
1266
Paul E. McKenney7dc52152010-02-22 17:04:52 -08001267 rcu_read_lock();
1268 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1269 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270 return -EPERM;
Paul E. McKenney7dc52152010-02-22 17:04:52 -08001271 }
1272 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273
Al Viroe0182902006-05-18 08:28:02 -04001274 if (count >= PAGE_SIZE)
1275 count = PAGE_SIZE - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276
1277 if (*ppos != 0) {
1278 /* No partial writes. */
1279 return -EINVAL;
1280 }
Mel Gormane12ba742007-10-16 01:25:52 -07001281 page = (char*)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282 if (!page)
1283 return -ENOMEM;
1284 length = -EFAULT;
1285 if (copy_from_user(page, buf, count))
1286 goto out_free_page;
1287
Al Viroe0182902006-05-18 08:28:02 -04001288 page[count] = '\0';
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 loginuid = simple_strtoul(page, &tmp, 10);
1290 if (tmp == page) {
1291 length = -EINVAL;
1292 goto out_free_page;
1293
1294 }
Eric W. Biederman99f89552006-06-26 00:25:55 -07001295 length = audit_set_loginuid(current, loginuid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 if (likely(length == 0))
1297 length = count;
1298
1299out_free_page:
1300 free_page((unsigned long) page);
1301 return length;
1302}
1303
Arjan van de Ven00977a52007-02-12 00:55:34 -08001304static const struct file_operations proc_loginuid_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305 .read = proc_loginuid_read,
1306 .write = proc_loginuid_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001307 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308};
Eric Paris1e0bd752008-03-13 08:15:31 -04001309
1310static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1311 size_t count, loff_t *ppos)
1312{
1313 struct inode * inode = file->f_path.dentry->d_inode;
1314 struct task_struct *task = get_proc_task(inode);
1315 ssize_t length;
1316 char tmpbuf[TMPBUFLEN];
1317
1318 if (!task)
1319 return -ESRCH;
1320 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1321 audit_get_sessionid(task));
1322 put_task_struct(task);
1323 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1324}
1325
1326static const struct file_operations proc_sessionid_operations = {
1327 .read = proc_sessionid_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001328 .llseek = generic_file_llseek,
Eric Paris1e0bd752008-03-13 08:15:31 -04001329};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330#endif
1331
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001332#ifdef CONFIG_FAULT_INJECTION
1333static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1334 size_t count, loff_t *ppos)
1335{
1336 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1337 char buffer[PROC_NUMBUF];
1338 size_t len;
1339 int make_it_fail;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001340
1341 if (!task)
1342 return -ESRCH;
1343 make_it_fail = task->make_it_fail;
1344 put_task_struct(task);
1345
1346 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001347
1348 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001349}
1350
1351static ssize_t proc_fault_inject_write(struct file * file,
1352 const char __user * buf, size_t count, loff_t *ppos)
1353{
1354 struct task_struct *task;
1355 char buffer[PROC_NUMBUF], *end;
1356 int make_it_fail;
1357
1358 if (!capable(CAP_SYS_RESOURCE))
1359 return -EPERM;
1360 memset(buffer, 0, sizeof(buffer));
1361 if (count > sizeof(buffer) - 1)
1362 count = sizeof(buffer) - 1;
1363 if (copy_from_user(buffer, buf, count))
1364 return -EFAULT;
Vincent Licba8aaf2009-09-22 16:45:38 -07001365 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1366 if (*end)
1367 return -EINVAL;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001368 task = get_proc_task(file->f_dentry->d_inode);
1369 if (!task)
1370 return -ESRCH;
1371 task->make_it_fail = make_it_fail;
1372 put_task_struct(task);
Vincent Licba8aaf2009-09-22 16:45:38 -07001373
1374 return count;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001375}
1376
Arjan van de Ven00977a52007-02-12 00:55:34 -08001377static const struct file_operations proc_fault_inject_operations = {
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001378 .read = proc_fault_inject_read,
1379 .write = proc_fault_inject_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001380 .llseek = generic_file_llseek,
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001381};
1382#endif
1383
Arjan van de Ven97455122008-01-25 21:08:34 +01001384
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001385#ifdef CONFIG_SCHED_DEBUG
1386/*
1387 * Print out various scheduling related per-task fields:
1388 */
1389static int sched_show(struct seq_file *m, void *v)
1390{
1391 struct inode *inode = m->private;
1392 struct task_struct *p;
1393
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001394 p = get_proc_task(inode);
1395 if (!p)
1396 return -ESRCH;
1397 proc_sched_show_task(p, m);
1398
1399 put_task_struct(p);
1400
1401 return 0;
1402}
1403
1404static ssize_t
1405sched_write(struct file *file, const char __user *buf,
1406 size_t count, loff_t *offset)
1407{
1408 struct inode *inode = file->f_path.dentry->d_inode;
1409 struct task_struct *p;
1410
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001411 p = get_proc_task(inode);
1412 if (!p)
1413 return -ESRCH;
1414 proc_sched_set_task(p);
1415
1416 put_task_struct(p);
1417
1418 return count;
1419}
1420
1421static int sched_open(struct inode *inode, struct file *filp)
1422{
Jovi Zhangc6a34052011-01-12 17:00:34 -08001423 return single_open(filp, sched_show, inode);
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001424}
1425
1426static const struct file_operations proc_pid_sched_operations = {
1427 .open = sched_open,
1428 .read = seq_read,
1429 .write = sched_write,
1430 .llseek = seq_lseek,
Alexey Dobriyan5ea473a2007-07-31 00:38:50 -07001431 .release = single_release,
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001432};
1433
1434#endif
1435
Mike Galbraith5091faa2010-11-30 14:18:03 +01001436#ifdef CONFIG_SCHED_AUTOGROUP
1437/*
1438 * Print out autogroup related information:
1439 */
1440static int sched_autogroup_show(struct seq_file *m, void *v)
1441{
1442 struct inode *inode = m->private;
1443 struct task_struct *p;
1444
1445 p = get_proc_task(inode);
1446 if (!p)
1447 return -ESRCH;
1448 proc_sched_autogroup_show_task(p, m);
1449
1450 put_task_struct(p);
1451
1452 return 0;
1453}
1454
1455static ssize_t
1456sched_autogroup_write(struct file *file, const char __user *buf,
1457 size_t count, loff_t *offset)
1458{
1459 struct inode *inode = file->f_path.dentry->d_inode;
1460 struct task_struct *p;
1461 char buffer[PROC_NUMBUF];
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001462 int nice;
Mike Galbraith5091faa2010-11-30 14:18:03 +01001463 int err;
1464
1465 memset(buffer, 0, sizeof(buffer));
1466 if (count > sizeof(buffer) - 1)
1467 count = sizeof(buffer) - 1;
1468 if (copy_from_user(buffer, buf, count))
1469 return -EFAULT;
1470
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001471 err = kstrtoint(strstrip(buffer), 0, &nice);
1472 if (err < 0)
1473 return err;
Mike Galbraith5091faa2010-11-30 14:18:03 +01001474
1475 p = get_proc_task(inode);
1476 if (!p)
1477 return -ESRCH;
1478
1479 err = nice;
1480 err = proc_sched_autogroup_set_nice(p, &err);
1481 if (err)
1482 count = err;
1483
1484 put_task_struct(p);
1485
1486 return count;
1487}
1488
1489static int sched_autogroup_open(struct inode *inode, struct file *filp)
1490{
1491 int ret;
1492
1493 ret = single_open(filp, sched_autogroup_show, NULL);
1494 if (!ret) {
1495 struct seq_file *m = filp->private_data;
1496
1497 m->private = inode;
1498 }
1499 return ret;
1500}
1501
1502static const struct file_operations proc_pid_sched_autogroup_operations = {
1503 .open = sched_autogroup_open,
1504 .read = seq_read,
1505 .write = sched_autogroup_write,
1506 .llseek = seq_lseek,
1507 .release = single_release,
1508};
1509
1510#endif /* CONFIG_SCHED_AUTOGROUP */
1511
john stultz4614a696b2009-12-14 18:00:05 -08001512static ssize_t comm_write(struct file *file, const char __user *buf,
1513 size_t count, loff_t *offset)
1514{
1515 struct inode *inode = file->f_path.dentry->d_inode;
1516 struct task_struct *p;
1517 char buffer[TASK_COMM_LEN];
1518
1519 memset(buffer, 0, sizeof(buffer));
1520 if (count > sizeof(buffer) - 1)
1521 count = sizeof(buffer) - 1;
1522 if (copy_from_user(buffer, buf, count))
1523 return -EFAULT;
1524
1525 p = get_proc_task(inode);
1526 if (!p)
1527 return -ESRCH;
1528
1529 if (same_thread_group(current, p))
1530 set_task_comm(p, buffer);
1531 else
1532 count = -EINVAL;
1533
1534 put_task_struct(p);
1535
1536 return count;
1537}
1538
1539static int comm_show(struct seq_file *m, void *v)
1540{
1541 struct inode *inode = m->private;
1542 struct task_struct *p;
1543
1544 p = get_proc_task(inode);
1545 if (!p)
1546 return -ESRCH;
1547
1548 task_lock(p);
1549 seq_printf(m, "%s\n", p->comm);
1550 task_unlock(p);
1551
1552 put_task_struct(p);
1553
1554 return 0;
1555}
1556
1557static int comm_open(struct inode *inode, struct file *filp)
1558{
Jovi Zhangc6a34052011-01-12 17:00:34 -08001559 return single_open(filp, comm_show, inode);
john stultz4614a696b2009-12-14 18:00:05 -08001560}
1561
1562static const struct file_operations proc_pid_set_comm_operations = {
1563 .open = comm_open,
1564 .read = seq_read,
1565 .write = comm_write,
1566 .llseek = seq_lseek,
1567 .release = single_release,
1568};
1569
Matt Helsley925d1c42008-04-29 01:01:36 -07001570static int proc_exe_link(struct inode *inode, struct path *exe_path)
1571{
1572 struct task_struct *task;
1573 struct mm_struct *mm;
1574 struct file *exe_file;
1575
1576 task = get_proc_task(inode);
1577 if (!task)
1578 return -ENOENT;
1579 mm = get_task_mm(task);
1580 put_task_struct(task);
1581 if (!mm)
1582 return -ENOENT;
1583 exe_file = get_mm_exe_file(mm);
1584 mmput(mm);
1585 if (exe_file) {
1586 *exe_path = exe_file->f_path;
1587 path_get(&exe_file->f_path);
1588 fput(exe_file);
1589 return 0;
1590 } else
1591 return -ENOENT;
1592}
1593
Al Viro008b1502005-08-20 00:17:39 +01001594static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595{
1596 struct inode *inode = dentry->d_inode;
1597 int error = -EACCES;
1598
1599 /* We don't need a base pointer in the /proc filesystem */
Jan Blunck1d957f92008-02-14 19:34:35 -08001600 path_put(&nd->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601
Eric W. Biederman778c1142006-06-26 00:25:58 -07001602 /* Are we allowed to snoop on the tasks file descriptors? */
1603 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001606 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607out:
Al Viro008b1502005-08-20 00:17:39 +01001608 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609}
1610
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001611static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612{
Mel Gormane12ba742007-10-16 01:25:52 -07001613 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001614 char *pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615 int len;
1616
1617 if (!tmp)
1618 return -ENOMEM;
Akinobu Mita0c28f282007-05-08 00:31:41 -07001619
Eric W. Biederman7b2a69b2010-12-05 15:51:21 -08001620 pathname = d_path(path, tmp, PAGE_SIZE);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001621 len = PTR_ERR(pathname);
1622 if (IS_ERR(pathname))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623 goto out;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001624 len = tmp + PAGE_SIZE - 1 - pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625
1626 if (len > buflen)
1627 len = buflen;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001628 if (copy_to_user(buffer, pathname, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629 len = -EFAULT;
1630 out:
1631 free_page((unsigned long)tmp);
1632 return len;
1633}
1634
1635static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1636{
1637 int error = -EACCES;
1638 struct inode *inode = dentry->d_inode;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001639 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640
Eric W. Biederman778c1142006-06-26 00:25:58 -07001641 /* Are we allowed to snoop on the tasks file descriptors? */
1642 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001645 error = PROC_I(inode)->op.proc_get_link(inode, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 if (error)
1647 goto out;
1648
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001649 error = do_proc_readlink(&path, buffer, buflen);
1650 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652 return error;
1653}
1654
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08001655static const struct inode_operations proc_pid_link_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656 .readlink = proc_pid_readlink,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001657 .follow_link = proc_pid_follow_link,
1658 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659};
1660
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001661
1662/* building an inode */
1663
1664static int task_dumpable(struct task_struct *task)
1665{
1666 int dumpable = 0;
1667 struct mm_struct *mm;
1668
1669 task_lock(task);
1670 mm = task->mm;
1671 if (mm)
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001672 dumpable = get_dumpable(mm);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001673 task_unlock(task);
1674 if(dumpable == 1)
1675 return 1;
1676 return 0;
1677}
1678
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001679struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001680{
1681 struct inode * inode;
1682 struct proc_inode *ei;
David Howellsc69e8d92008-11-14 10:39:19 +11001683 const struct cred *cred;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001684
1685 /* We need a new inode */
1686
1687 inode = new_inode(sb);
1688 if (!inode)
1689 goto out;
1690
1691 /* Common stuff */
1692 ei = PROC_I(inode);
Christoph Hellwig85fe4022010-10-23 11:19:54 -04001693 inode->i_ino = get_next_ino();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001694 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001695 inode->i_op = &proc_def_inode_operations;
1696
1697 /*
1698 * grab the reference to task.
1699 */
Oleg Nesterov1a657f72006-10-02 02:18:59 -07001700 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001701 if (!ei->pid)
1702 goto out_unlock;
1703
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001704 if (task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001705 rcu_read_lock();
1706 cred = __task_cred(task);
1707 inode->i_uid = cred->euid;
1708 inode->i_gid = cred->egid;
1709 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001710 }
1711 security_task_to_inode(task, inode);
1712
1713out:
1714 return inode;
1715
1716out_unlock:
1717 iput(inode);
1718 return NULL;
1719}
1720
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001721int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001722{
1723 struct inode *inode = dentry->d_inode;
1724 struct task_struct *task;
David Howellsc69e8d92008-11-14 10:39:19 +11001725 const struct cred *cred;
1726
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001727 generic_fillattr(inode, stat);
1728
1729 rcu_read_lock();
1730 stat->uid = 0;
1731 stat->gid = 0;
1732 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1733 if (task) {
1734 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1735 task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001736 cred = __task_cred(task);
1737 stat->uid = cred->euid;
1738 stat->gid = cred->egid;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001739 }
1740 }
1741 rcu_read_unlock();
1742 return 0;
1743}
1744
1745/* dentry stuff */
1746
1747/*
1748 * Exceptional case: normally we are not allowed to unhash a busy
1749 * directory. In this case, however, we can do it - no aliasing problems
1750 * due to the way we treat inodes.
1751 *
1752 * Rewrite the inode's ownerships here because the owning task may have
1753 * performed a setuid(), etc.
1754 *
1755 * Before the /proc/pid/status file was created the only way to read
1756 * the effective uid of a /process was to stat /proc/pid. Reading
1757 * /proc/pid/status is slow enough that procps and other packages
1758 * kept stating /proc/pid. To keep the rules in /proc simple I have
1759 * made this apply to all per process world readable and executable
1760 * directories.
1761 */
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001762int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001763{
Nick Piggin34286d62011-01-07 17:49:57 +11001764 struct inode *inode;
1765 struct task_struct *task;
David Howellsc69e8d92008-11-14 10:39:19 +11001766 const struct cred *cred;
1767
Nick Piggin34286d62011-01-07 17:49:57 +11001768 if (nd && nd->flags & LOOKUP_RCU)
1769 return -ECHILD;
1770
1771 inode = dentry->d_inode;
1772 task = get_proc_task(inode);
1773
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001774 if (task) {
1775 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1776 task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001777 rcu_read_lock();
1778 cred = __task_cred(task);
1779 inode->i_uid = cred->euid;
1780 inode->i_gid = cred->egid;
1781 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001782 } else {
1783 inode->i_uid = 0;
1784 inode->i_gid = 0;
1785 }
1786 inode->i_mode &= ~(S_ISUID | S_ISGID);
1787 security_task_to_inode(task, inode);
1788 put_task_struct(task);
1789 return 1;
1790 }
1791 d_drop(dentry);
1792 return 0;
1793}
1794
Nick Pigginfe15ce42011-01-07 17:49:23 +11001795static int pid_delete_dentry(const struct dentry * dentry)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001796{
1797 /* Is the task we represent dead?
1798 * If so, then don't put the dentry on the lru list,
1799 * kill it immediately.
1800 */
1801 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1802}
1803
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001804const struct dentry_operations pid_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001805{
1806 .d_revalidate = pid_revalidate,
1807 .d_delete = pid_delete_dentry,
1808};
1809
1810/* Lookups */
1811
Eric W. Biederman1c0d04c2006-10-02 02:18:57 -07001812/*
1813 * Fill a directory entry.
1814 *
1815 * If possible create the dcache entry and derive our inode number and
1816 * file type from dcache entry.
1817 *
1818 * Since all of the proc inode numbers are dynamically generated, the inode
1819 * numbers do not exist until the inode is cache. This means creating the
1820 * the dcache entry in readdir is necessary to keep the inode numbers
1821 * reported by readdir in sync with the inode numbers reported
1822 * by stat.
1823 */
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001824int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1825 const char *name, int len,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001826 instantiate_t instantiate, struct task_struct *task, const void *ptr)
Eric W. Biederman61a28782006-10-02 02:18:49 -07001827{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001828 struct dentry *child, *dir = filp->f_path.dentry;
Eric W. Biederman61a28782006-10-02 02:18:49 -07001829 struct inode *inode;
1830 struct qstr qname;
1831 ino_t ino = 0;
1832 unsigned type = DT_UNKNOWN;
1833
1834 qname.name = name;
1835 qname.len = len;
1836 qname.hash = full_name_hash(name, len);
1837
1838 child = d_lookup(dir, &qname);
1839 if (!child) {
1840 struct dentry *new;
1841 new = d_alloc(dir, &qname);
1842 if (new) {
1843 child = instantiate(dir->d_inode, new, task, ptr);
1844 if (child)
1845 dput(new);
1846 else
1847 child = new;
1848 }
1849 }
1850 if (!child || IS_ERR(child) || !child->d_inode)
1851 goto end_instantiate;
1852 inode = child->d_inode;
1853 if (inode) {
1854 ino = inode->i_ino;
1855 type = inode->i_mode >> 12;
1856 }
1857 dput(child);
1858end_instantiate:
1859 if (!ino)
1860 ino = find_inode_number(dir, &qname);
1861 if (!ino)
1862 ino = 1;
1863 return filldir(dirent, name, len, filp->f_pos, ino, type);
1864}
1865
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001866static unsigned name_to_int(struct dentry *dentry)
1867{
1868 const char *name = dentry->d_name.name;
1869 int len = dentry->d_name.len;
1870 unsigned n = 0;
1871
1872 if (len > 1 && *name == '0')
1873 goto out;
1874 while (len-- > 0) {
1875 unsigned c = *name++ - '0';
1876 if (c > 9)
1877 goto out;
1878 if (n >= (~0U-9)/10)
1879 goto out;
1880 n *= 10;
1881 n += c;
1882 }
1883 return n;
1884out:
1885 return ~0U;
1886}
1887
Miklos Szeredi27932742007-05-08 00:26:17 -07001888#define PROC_FDINFO_MAX 64
1889
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001890static int proc_fd_info(struct inode *inode, struct path *path, char *info)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001891{
Linus Torvalds5e442a42011-11-09 18:16:00 -05001892 struct task_struct *task = get_proc_task(inode);
1893 struct files_struct *files = NULL;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001894 struct file *file;
1895 int fd = proc_fd(inode);
1896
Linus Torvalds5e442a42011-11-09 18:16:00 -05001897 if (task) {
1898 files = get_files_struct(task);
1899 put_task_struct(task);
1900 }
1901 if (files) {
1902 /*
1903 * We are not taking a ref to the file structure, so we must
1904 * hold ->file_lock.
1905 */
1906 spin_lock(&files->file_lock);
1907 file = fcheck_files(files, fd);
1908 if (file) {
1909 unsigned int f_flags;
1910 struct fdtable *fdt;
Linus Torvalds1117f722011-08-06 11:51:33 -07001911
Linus Torvalds5e442a42011-11-09 18:16:00 -05001912 fdt = files_fdtable(files);
1913 f_flags = file->f_flags & ~O_CLOEXEC;
1914 if (FD_ISSET(fd, fdt->close_on_exec))
1915 f_flags |= O_CLOEXEC;
Linus Torvalds1117f722011-08-06 11:51:33 -07001916
Linus Torvalds5e442a42011-11-09 18:16:00 -05001917 if (path) {
1918 *path = file->f_path;
1919 path_get(&file->f_path);
1920 }
1921 if (info)
1922 snprintf(info, PROC_FDINFO_MAX,
1923 "pos:\t%lli\n"
1924 "flags:\t0%o\n",
1925 (long long) file->f_pos,
1926 f_flags);
1927 spin_unlock(&files->file_lock);
1928 put_files_struct(files);
1929 return 0;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001930 }
Linus Torvalds5e442a42011-11-09 18:16:00 -05001931 spin_unlock(&files->file_lock);
1932 put_files_struct(files);
1933 }
1934 return -ENOENT;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001935}
1936
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001937static int proc_fd_link(struct inode *inode, struct path *path)
Miklos Szeredi27932742007-05-08 00:26:17 -07001938{
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001939 return proc_fd_info(inode, path, NULL);
Miklos Szeredi27932742007-05-08 00:26:17 -07001940}
1941
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001942static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1943{
Nick Piggin34286d62011-01-07 17:49:57 +11001944 struct inode *inode;
1945 struct task_struct *task;
1946 int fd;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001947 struct files_struct *files;
David Howellsc69e8d92008-11-14 10:39:19 +11001948 const struct cred *cred;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001949
Nick Piggin34286d62011-01-07 17:49:57 +11001950 if (nd && nd->flags & LOOKUP_RCU)
1951 return -ECHILD;
1952
1953 inode = dentry->d_inode;
1954 task = get_proc_task(inode);
1955 fd = proc_fd(inode);
1956
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001957 if (task) {
1958 files = get_files_struct(task);
1959 if (files) {
1960 rcu_read_lock();
1961 if (fcheck_files(files, fd)) {
1962 rcu_read_unlock();
1963 put_files_struct(files);
1964 if (task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001965 rcu_read_lock();
1966 cred = __task_cred(task);
1967 inode->i_uid = cred->euid;
1968 inode->i_gid = cred->egid;
1969 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001970 } else {
1971 inode->i_uid = 0;
1972 inode->i_gid = 0;
1973 }
1974 inode->i_mode &= ~(S_ISUID | S_ISGID);
1975 security_task_to_inode(task, inode);
1976 put_task_struct(task);
1977 return 1;
1978 }
1979 rcu_read_unlock();
1980 put_files_struct(files);
1981 }
1982 put_task_struct(task);
1983 }
1984 d_drop(dentry);
1985 return 0;
1986}
1987
Al Virod72f71e2009-02-20 05:58:47 +00001988static const struct dentry_operations tid_fd_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001989{
1990 .d_revalidate = tid_fd_revalidate,
1991 .d_delete = pid_delete_dentry,
1992};
1993
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001994static struct dentry *proc_fd_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001995 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001996{
Eric Dumazetc5141e62007-05-08 00:26:15 -07001997 unsigned fd = *(const unsigned *)ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001998 struct file *file;
1999 struct files_struct *files;
2000 struct inode *inode;
2001 struct proc_inode *ei;
2002 struct dentry *error = ERR_PTR(-ENOENT);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002003
Eric W. Biederman61a28782006-10-02 02:18:49 -07002004 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002005 if (!inode)
2006 goto out;
2007 ei = PROC_I(inode);
2008 ei->fd = fd;
2009 files = get_files_struct(task);
2010 if (!files)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002011 goto out_iput;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002012 inode->i_mode = S_IFLNK;
2013
2014 /*
2015 * We are not taking a ref to the file structure, so we must
2016 * hold ->file_lock.
2017 */
2018 spin_lock(&files->file_lock);
2019 file = fcheck_files(files, fd);
2020 if (!file)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002021 goto out_unlock;
Al Viroaeb5d722008-09-02 15:28:45 -04002022 if (file->f_mode & FMODE_READ)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002023 inode->i_mode |= S_IRUSR | S_IXUSR;
Al Viroaeb5d722008-09-02 15:28:45 -04002024 if (file->f_mode & FMODE_WRITE)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002025 inode->i_mode |= S_IWUSR | S_IXUSR;
2026 spin_unlock(&files->file_lock);
2027 put_files_struct(files);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002028
Linus Torvalds5e442a42011-11-09 18:16:00 -05002029 inode->i_op = &proc_pid_link_inode_operations;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002030 inode->i_size = 64;
2031 ei->op.proc_get_link = proc_fd_link;
Nick Pigginfb045ad2011-01-07 17:49:55 +11002032 d_set_d_op(dentry, &tid_fd_dentry_operations);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002033 d_add(dentry, inode);
2034 /* Close the race of the process dying before we return the dentry */
2035 if (tid_fd_revalidate(dentry, NULL))
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002036 error = NULL;
2037
2038 out:
2039 return error;
2040out_unlock:
2041 spin_unlock(&files->file_lock);
2042 put_files_struct(files);
2043out_iput:
2044 iput(inode);
2045 goto out;
2046}
2047
Miklos Szeredi27932742007-05-08 00:26:17 -07002048static struct dentry *proc_lookupfd_common(struct inode *dir,
2049 struct dentry *dentry,
2050 instantiate_t instantiate)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002051{
2052 struct task_struct *task = get_proc_task(dir);
2053 unsigned fd = name_to_int(dentry);
2054 struct dentry *result = ERR_PTR(-ENOENT);
2055
2056 if (!task)
2057 goto out_no_task;
2058 if (fd == ~0U)
2059 goto out;
2060
Miklos Szeredi27932742007-05-08 00:26:17 -07002061 result = instantiate(dir, dentry, task, &fd);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002062out:
2063 put_task_struct(task);
2064out_no_task:
2065 return result;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002066}
2067
Miklos Szeredi27932742007-05-08 00:26:17 -07002068static int proc_readfd_common(struct file * filp, void * dirent,
2069 filldir_t filldir, instantiate_t instantiate)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002070{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002071 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman56347082006-06-26 00:25:40 -07002072 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002073 struct task_struct *p = get_proc_task(inode);
Pavel Emelyanov457c2512007-10-18 23:40:43 -07002074 unsigned int fd, ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076 struct files_struct * files;
2077
2078 retval = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002079 if (!p)
2080 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002081 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082
2083 fd = filp->f_pos;
2084 switch (fd) {
2085 case 0:
2086 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
Linus Torvalds5e442a42011-11-09 18:16:00 -05002087 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088 filp->f_pos++;
2089 case 1:
Eric W. Biederman56347082006-06-26 00:25:40 -07002090 ino = parent_ino(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
Linus Torvalds5e442a42011-11-09 18:16:00 -05002092 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093 filp->f_pos++;
2094 default:
2095 files = get_files_struct(p);
2096 if (!files)
Linus Torvalds5e442a42011-11-09 18:16:00 -05002097 goto out;
Dipankar Sarmab8359962005-09-09 13:04:14 -07002098 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099 for (fd = filp->f_pos-2;
Al Viro9b4f5262008-04-22 01:32:44 -04002100 fd < files_fdtable(files)->max_fds;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101 fd++, filp->f_pos++) {
Miklos Szeredi27932742007-05-08 00:26:17 -07002102 char name[PROC_NUMBUF];
2103 int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104
2105 if (!fcheck_files(files, fd))
2106 continue;
Dipankar Sarmab8359962005-09-09 13:04:14 -07002107 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108
Miklos Szeredi27932742007-05-08 00:26:17 -07002109 len = snprintf(name, sizeof(name), "%d", fd);
2110 if (proc_fill_cache(filp, dirent, filldir,
2111 name, len, instantiate,
2112 p, &fd) < 0) {
Dipankar Sarmab8359962005-09-09 13:04:14 -07002113 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114 break;
2115 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07002116 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07002118 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119 put_files_struct(files);
2120 }
2121out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07002122 put_task_struct(p);
2123out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002124 return retval;
2125}
2126
Miklos Szeredi27932742007-05-08 00:26:17 -07002127static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2128 struct nameidata *nd)
2129{
2130 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2131}
2132
2133static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2134{
2135 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2136}
2137
2138static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2139 size_t len, loff_t *ppos)
2140{
2141 char tmp[PROC_FDINFO_MAX];
Jan Blunck3dcd25f2008-02-14 19:38:35 -08002142 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
Miklos Szeredi27932742007-05-08 00:26:17 -07002143 if (!err)
2144 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2145 return err;
2146}
2147
2148static const struct file_operations proc_fdinfo_file_operations = {
Arnd Bergmann87df8422010-03-17 23:06:02 +01002149 .open = nonseekable_open,
Miklos Szeredi27932742007-05-08 00:26:17 -07002150 .read = proc_fdinfo_read,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002151 .llseek = no_llseek,
Miklos Szeredi27932742007-05-08 00:26:17 -07002152};
2153
Arjan van de Ven00977a52007-02-12 00:55:34 -08002154static const struct file_operations proc_fd_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155 .read = generic_read_dir,
2156 .readdir = proc_readfd,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002157 .llseek = default_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002158};
2159
Linus Torvalds1da177e2005-04-16 15:20:36 -07002160/*
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002161 * /proc/pid/fd needs a special permission handler so that a process can still
2162 * access /proc/self/fd after it has executed a setuid().
2163 */
Al Viro10556cb2011-06-20 19:28:19 -04002164static int proc_fd_permission(struct inode *inode, int mask)
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002165{
Al Viro2830ba72011-06-20 19:16:29 -04002166 int rv = generic_permission(inode, mask);
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002167 if (rv == 0)
2168 return 0;
2169 if (task_pid(current) == proc_pid(inode))
2170 rv = 0;
2171 return rv;
2172}
2173
2174/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002175 * proc directories can do almost nothing..
2176 */
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002177static const struct inode_operations proc_fd_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002178 .lookup = proc_lookupfd,
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002179 .permission = proc_fd_permission,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07002180 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181};
2182
Miklos Szeredi27932742007-05-08 00:26:17 -07002183static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2184 struct dentry *dentry, struct task_struct *task, const void *ptr)
2185{
2186 unsigned fd = *(unsigned *)ptr;
2187 struct inode *inode;
2188 struct proc_inode *ei;
2189 struct dentry *error = ERR_PTR(-ENOENT);
2190
2191 inode = proc_pid_make_inode(dir->i_sb, task);
2192 if (!inode)
2193 goto out;
2194 ei = PROC_I(inode);
2195 ei->fd = fd;
2196 inode->i_mode = S_IFREG | S_IRUSR;
2197 inode->i_fop = &proc_fdinfo_file_operations;
Nick Pigginfb045ad2011-01-07 17:49:55 +11002198 d_set_d_op(dentry, &tid_fd_dentry_operations);
Miklos Szeredi27932742007-05-08 00:26:17 -07002199 d_add(dentry, inode);
2200 /* Close the race of the process dying before we return the dentry */
2201 if (tid_fd_revalidate(dentry, NULL))
2202 error = NULL;
2203
2204 out:
2205 return error;
2206}
2207
2208static struct dentry *proc_lookupfdinfo(struct inode *dir,
2209 struct dentry *dentry,
2210 struct nameidata *nd)
2211{
2212 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2213}
2214
2215static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2216{
2217 return proc_readfd_common(filp, dirent, filldir,
2218 proc_fdinfo_instantiate);
2219}
2220
2221static const struct file_operations proc_fdinfo_operations = {
2222 .read = generic_read_dir,
2223 .readdir = proc_readfdinfo,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002224 .llseek = default_llseek,
Miklos Szeredi27932742007-05-08 00:26:17 -07002225};
2226
2227/*
2228 * proc directories can do almost nothing..
2229 */
2230static const struct inode_operations proc_fdinfo_inode_operations = {
2231 .lookup = proc_lookupfdinfo,
2232 .setattr = proc_setattr,
2233};
2234
2235
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002236static struct dentry *proc_pident_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002237 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002238{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002239 const struct pid_entry *p = ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002240 struct inode *inode;
2241 struct proc_inode *ei;
KOSAKI Motohirobd6daba2009-05-28 14:34:21 -07002242 struct dentry *error = ERR_PTR(-ENOENT);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002243
Eric W. Biederman61a28782006-10-02 02:18:49 -07002244 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002245 if (!inode)
2246 goto out;
2247
2248 ei = PROC_I(inode);
2249 inode->i_mode = p->mode;
2250 if (S_ISDIR(inode->i_mode))
Miklos Szeredibfe86842011-10-28 14:13:29 +02002251 set_nlink(inode, 2); /* Use getattr to fix if necessary */
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002252 if (p->iop)
2253 inode->i_op = p->iop;
2254 if (p->fop)
2255 inode->i_fop = p->fop;
2256 ei->op = p->op;
Nick Pigginfb045ad2011-01-07 17:49:55 +11002257 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002258 d_add(dentry, inode);
2259 /* Close the race of the process dying before we return the dentry */
2260 if (pid_revalidate(dentry, NULL))
2261 error = NULL;
2262out:
2263 return error;
2264}
2265
Linus Torvalds1da177e2005-04-16 15:20:36 -07002266static struct dentry *proc_pident_lookup(struct inode *dir,
2267 struct dentry *dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002268 const struct pid_entry *ents,
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002269 unsigned int nents)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002270{
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002271 struct dentry *error;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002272 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002273 const struct pid_entry *p, *last;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002275 error = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002276
Eric W. Biederman99f89552006-06-26 00:25:55 -07002277 if (!task)
2278 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002279
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002280 /*
2281 * Yes, it does not scale. And it should not. Don't add
2282 * new entries into /proc/<tgid>/ without very good reasons.
2283 */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002284 last = &ents[nents - 1];
2285 for (p = ents; p <= last; p++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286 if (p->len != dentry->d_name.len)
2287 continue;
2288 if (!memcmp(dentry->d_name.name, p->name, p->len))
2289 break;
2290 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002291 if (p > last)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002292 goto out;
2293
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002294 error = proc_pident_instantiate(dir, dentry, task, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002295out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07002296 put_task_struct(task);
2297out_no_task:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002298 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002299}
2300
Eric Dumazetc5141e62007-05-08 00:26:15 -07002301static int proc_pident_fill_cache(struct file *filp, void *dirent,
2302 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002303{
2304 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2305 proc_pident_instantiate, task, p);
2306}
2307
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002308static int proc_pident_readdir(struct file *filp,
2309 void *dirent, filldir_t filldir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002310 const struct pid_entry *ents, unsigned int nents)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002311{
2312 int i;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002313 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002314 struct inode *inode = dentry->d_inode;
2315 struct task_struct *task = get_proc_task(inode);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002316 const struct pid_entry *p, *last;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002317 ino_t ino;
2318 int ret;
2319
2320 ret = -ENOENT;
2321 if (!task)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002322 goto out_no_task;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002323
2324 ret = 0;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002325 i = filp->f_pos;
2326 switch (i) {
2327 case 0:
2328 ino = inode->i_ino;
2329 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2330 goto out;
2331 i++;
2332 filp->f_pos++;
2333 /* fall through */
2334 case 1:
2335 ino = parent_ino(dentry);
2336 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2337 goto out;
2338 i++;
2339 filp->f_pos++;
2340 /* fall through */
2341 default:
2342 i -= 2;
2343 if (i >= nents) {
2344 ret = 1;
2345 goto out;
2346 }
2347 p = ents + i;
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002348 last = &ents[nents - 1];
2349 while (p <= last) {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002350 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002351 goto out;
2352 filp->f_pos++;
2353 p++;
2354 }
2355 }
2356
2357 ret = 1;
2358out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07002359 put_task_struct(task);
2360out_no_task:
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002361 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002362}
2363
Linus Torvalds1da177e2005-04-16 15:20:36 -07002364#ifdef CONFIG_SECURITY
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002365static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2366 size_t count, loff_t *ppos)
2367{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002368 struct inode * inode = file->f_path.dentry->d_inode;
Al Viro04ff9702007-03-12 16:17:58 +00002369 char *p = NULL;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002370 ssize_t length;
2371 struct task_struct *task = get_proc_task(inode);
2372
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002373 if (!task)
Al Viro04ff9702007-03-12 16:17:58 +00002374 return -ESRCH;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002375
2376 length = security_getprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002377 (char*)file->f_path.dentry->d_name.name,
Al Viro04ff9702007-03-12 16:17:58 +00002378 &p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002379 put_task_struct(task);
Al Viro04ff9702007-03-12 16:17:58 +00002380 if (length > 0)
2381 length = simple_read_from_buffer(buf, count, ppos, p, length);
2382 kfree(p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002383 return length;
2384}
2385
2386static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2387 size_t count, loff_t *ppos)
2388{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002389 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002390 char *page;
2391 ssize_t length;
2392 struct task_struct *task = get_proc_task(inode);
2393
2394 length = -ESRCH;
2395 if (!task)
2396 goto out_no_task;
2397 if (count > PAGE_SIZE)
2398 count = PAGE_SIZE;
2399
2400 /* No partial writes. */
2401 length = -EINVAL;
2402 if (*ppos != 0)
2403 goto out;
2404
2405 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -07002406 page = (char*)__get_free_page(GFP_TEMPORARY);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002407 if (!page)
2408 goto out;
2409
2410 length = -EFAULT;
2411 if (copy_from_user(page, buf, count))
2412 goto out_free;
2413
David Howells107db7c2009-05-08 13:55:27 +01002414 /* Guard against adverse ptrace interaction */
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07002415 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
David Howells107db7c2009-05-08 13:55:27 +01002416 if (length < 0)
2417 goto out_free;
2418
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002419 length = security_setprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002420 (char*)file->f_path.dentry->d_name.name,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002421 (void*)page, count);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07002422 mutex_unlock(&task->signal->cred_guard_mutex);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002423out_free:
2424 free_page((unsigned long) page);
2425out:
2426 put_task_struct(task);
2427out_no_task:
2428 return length;
2429}
2430
Arjan van de Ven00977a52007-02-12 00:55:34 -08002431static const struct file_operations proc_pid_attr_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002432 .read = proc_pid_attr_read,
2433 .write = proc_pid_attr_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01002434 .llseek = generic_file_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002435};
2436
Eric Dumazetc5141e62007-05-08 00:26:15 -07002437static const struct pid_entry attr_dir_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002438 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2439 REG("prev", S_IRUGO, proc_pid_attr_operations),
2440 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2441 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2442 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2443 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002444};
2445
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002446static int proc_attr_dir_readdir(struct file * filp,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002447 void * dirent, filldir_t filldir)
2448{
2449 return proc_pident_readdir(filp,dirent,filldir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002450 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002451}
2452
Arjan van de Ven00977a52007-02-12 00:55:34 -08002453static const struct file_operations proc_attr_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002454 .read = generic_read_dir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002455 .readdir = proc_attr_dir_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002456 .llseek = default_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002457};
2458
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002459static struct dentry *proc_attr_dir_lookup(struct inode *dir,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002460 struct dentry *dentry, struct nameidata *nd)
2461{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002462 return proc_pident_lookup(dir, dentry,
2463 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002464}
2465
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002466static const struct inode_operations proc_attr_dir_inode_operations = {
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002467 .lookup = proc_attr_dir_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07002468 .getattr = pid_getattr,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07002469 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002470};
2471
Linus Torvalds1da177e2005-04-16 15:20:36 -07002472#endif
2473
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08002474#ifdef CONFIG_ELF_CORE
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002475static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2476 size_t count, loff_t *ppos)
2477{
2478 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2479 struct mm_struct *mm;
2480 char buffer[PROC_NUMBUF];
2481 size_t len;
2482 int ret;
2483
2484 if (!task)
2485 return -ESRCH;
2486
2487 ret = 0;
2488 mm = get_task_mm(task);
2489 if (mm) {
2490 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2491 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2492 MMF_DUMP_FILTER_SHIFT));
2493 mmput(mm);
2494 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2495 }
2496
2497 put_task_struct(task);
2498
2499 return ret;
2500}
2501
2502static ssize_t proc_coredump_filter_write(struct file *file,
2503 const char __user *buf,
2504 size_t count,
2505 loff_t *ppos)
2506{
2507 struct task_struct *task;
2508 struct mm_struct *mm;
2509 char buffer[PROC_NUMBUF], *end;
2510 unsigned int val;
2511 int ret;
2512 int i;
2513 unsigned long mask;
2514
2515 ret = -EFAULT;
2516 memset(buffer, 0, sizeof(buffer));
2517 if (count > sizeof(buffer) - 1)
2518 count = sizeof(buffer) - 1;
2519 if (copy_from_user(buffer, buf, count))
2520 goto out_no_task;
2521
2522 ret = -EINVAL;
2523 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2524 if (*end == '\n')
2525 end++;
2526 if (end - buffer == 0)
2527 goto out_no_task;
2528
2529 ret = -ESRCH;
2530 task = get_proc_task(file->f_dentry->d_inode);
2531 if (!task)
2532 goto out_no_task;
2533
2534 ret = end - buffer;
2535 mm = get_task_mm(task);
2536 if (!mm)
2537 goto out_no_mm;
2538
2539 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2540 if (val & mask)
2541 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2542 else
2543 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2544 }
2545
2546 mmput(mm);
2547 out_no_mm:
2548 put_task_struct(task);
2549 out_no_task:
2550 return ret;
2551}
2552
2553static const struct file_operations proc_coredump_filter_operations = {
2554 .read = proc_coredump_filter_read,
2555 .write = proc_coredump_filter_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01002556 .llseek = generic_file_llseek,
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002557};
2558#endif
2559
Linus Torvalds1da177e2005-04-16 15:20:36 -07002560/*
2561 * /proc/self:
2562 */
2563static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2564 int buflen)
2565{
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002566 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002567 pid_t tgid = task_tgid_nr_ns(current, ns);
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002568 char tmp[PROC_NUMBUF];
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002569 if (!tgid)
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002570 return -ENOENT;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002571 sprintf(tmp, "%d", tgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002572 return vfs_readlink(dentry,buffer,buflen,tmp);
2573}
2574
Al Viro008b1502005-08-20 00:17:39 +01002575static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002576{
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002577 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002578 pid_t tgid = task_tgid_nr_ns(current, ns);
Al Viro7fee4862010-01-14 01:03:28 -05002579 char *name = ERR_PTR(-ENOENT);
2580 if (tgid) {
2581 name = __getname();
2582 if (!name)
2583 name = ERR_PTR(-ENOMEM);
2584 else
2585 sprintf(name, "%d", tgid);
2586 }
2587 nd_set_link(nd, name);
2588 return NULL;
2589}
2590
2591static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2592 void *cookie)
2593{
2594 char *s = nd_get_link(nd);
2595 if (!IS_ERR(s))
2596 __putname(s);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002597}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002598
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002599static const struct inode_operations proc_self_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002600 .readlink = proc_self_readlink,
2601 .follow_link = proc_self_follow_link,
Al Viro7fee4862010-01-14 01:03:28 -05002602 .put_link = proc_self_put_link,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002603};
2604
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002605/*
Eric W. Biederman801199c2006-10-02 02:18:48 -07002606 * proc base
2607 *
2608 * These are the directory entries in the root directory of /proc
2609 * that properly belong to the /proc filesystem, as they describe
2610 * describe something that is process related.
2611 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07002612static const struct pid_entry proc_base_stuff[] = {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002613 NOD("self", S_IFLNK|S_IRWXUGO,
Eric W. Biederman801199c2006-10-02 02:18:48 -07002614 &proc_self_inode_operations, NULL, {}),
Eric W. Biederman801199c2006-10-02 02:18:48 -07002615};
2616
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002617static struct dentry *proc_base_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002618 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman801199c2006-10-02 02:18:48 -07002619{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002620 const struct pid_entry *p = ptr;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002621 struct inode *inode;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002622 struct proc_inode *ei;
Dan Carpenter73d36462010-05-26 14:43:25 -07002623 struct dentry *error;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002624
2625 /* Allocate the inode */
2626 error = ERR_PTR(-ENOMEM);
2627 inode = new_inode(dir->i_sb);
2628 if (!inode)
2629 goto out;
2630
2631 /* Initialize the inode */
2632 ei = PROC_I(inode);
Christoph Hellwig85fe4022010-10-23 11:19:54 -04002633 inode->i_ino = get_next_ino();
Eric W. Biederman801199c2006-10-02 02:18:48 -07002634 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002635
2636 /*
2637 * grab the reference to the task.
2638 */
Oleg Nesterov1a657f72006-10-02 02:18:59 -07002639 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002640 if (!ei->pid)
2641 goto out_iput;
2642
Eric W. Biederman801199c2006-10-02 02:18:48 -07002643 inode->i_mode = p->mode;
2644 if (S_ISDIR(inode->i_mode))
Miklos Szeredibfe86842011-10-28 14:13:29 +02002645 set_nlink(inode, 2);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002646 if (S_ISLNK(inode->i_mode))
2647 inode->i_size = 64;
2648 if (p->iop)
2649 inode->i_op = p->iop;
2650 if (p->fop)
2651 inode->i_fop = p->fop;
2652 ei->op = p->op;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002653 d_add(dentry, inode);
2654 error = NULL;
2655out:
Eric W. Biederman801199c2006-10-02 02:18:48 -07002656 return error;
2657out_iput:
2658 iput(inode);
2659 goto out;
2660}
2661
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002662static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2663{
2664 struct dentry *error;
2665 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002666 const struct pid_entry *p, *last;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002667
2668 error = ERR_PTR(-ENOENT);
2669
2670 if (!task)
2671 goto out_no_task;
2672
2673 /* Lookup the directory entry */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002674 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2675 for (p = proc_base_stuff; p <= last; p++) {
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002676 if (p->len != dentry->d_name.len)
2677 continue;
2678 if (!memcmp(dentry->d_name.name, p->name, p->len))
2679 break;
2680 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002681 if (p > last)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002682 goto out;
2683
2684 error = proc_base_instantiate(dir, dentry, task, p);
2685
2686out:
2687 put_task_struct(task);
2688out_no_task:
2689 return error;
2690}
2691
Eric Dumazetc5141e62007-05-08 00:26:15 -07002692static int proc_base_fill_cache(struct file *filp, void *dirent,
2693 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002694{
2695 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2696 proc_base_instantiate, task, p);
2697}
2698
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002699#ifdef CONFIG_TASK_IO_ACCOUNTING
Andrea Righi297c5d92008-07-25 01:48:49 -07002700static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002701{
Andrea Righi940389b2008-07-28 00:48:12 +02002702 struct task_io_accounting acct = task->ioac;
Andrea Righi59954772008-07-27 17:29:15 +02002703 unsigned long flags;
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002704 int result;
Andrea Righi297c5d92008-07-25 01:48:49 -07002705
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002706 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2707 if (result)
2708 return result;
2709
2710 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2711 result = -EACCES;
2712 goto out_unlock;
2713 }
Vasiliy Kulikov1d1221f2011-06-24 16:08:38 +04002714
Andrea Righi59954772008-07-27 17:29:15 +02002715 if (whole && lock_task_sighand(task, &flags)) {
2716 struct task_struct *t = task;
Andrea Righib2d002d2008-07-26 15:22:27 -07002717
Andrea Righi59954772008-07-27 17:29:15 +02002718 task_io_accounting_add(&acct, &task->signal->ioac);
2719 while_each_thread(task, t)
2720 task_io_accounting_add(&acct, &t->ioac);
Andrea Righi297c5d92008-07-25 01:48:49 -07002721
Andrea Righi59954772008-07-27 17:29:15 +02002722 unlock_task_sighand(task, &flags);
Andrea Righi297c5d92008-07-25 01:48:49 -07002723 }
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002724 result = sprintf(buffer,
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002725 "rchar: %llu\n"
2726 "wchar: %llu\n"
2727 "syscr: %llu\n"
2728 "syscw: %llu\n"
2729 "read_bytes: %llu\n"
2730 "write_bytes: %llu\n"
2731 "cancelled_write_bytes: %llu\n",
Alexander Beregalov7c443192008-08-05 13:01:34 -07002732 (unsigned long long)acct.rchar,
2733 (unsigned long long)acct.wchar,
2734 (unsigned long long)acct.syscr,
2735 (unsigned long long)acct.syscw,
2736 (unsigned long long)acct.read_bytes,
2737 (unsigned long long)acct.write_bytes,
2738 (unsigned long long)acct.cancelled_write_bytes);
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002739out_unlock:
2740 mutex_unlock(&task->signal->cred_guard_mutex);
2741 return result;
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002742}
Andrea Righi297c5d92008-07-25 01:48:49 -07002743
2744static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2745{
2746 return do_io_accounting(task, buffer, 0);
2747}
2748
2749static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2750{
2751 return do_io_accounting(task, buffer, 1);
2752}
2753#endif /* CONFIG_TASK_IO_ACCOUNTING */
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002754
Kees Cook47830722008-10-06 03:11:58 +04002755static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2756 struct pid *pid, struct task_struct *task)
2757{
Al Viroa9712bc2011-03-23 15:52:50 -04002758 int err = lock_trace(task);
2759 if (!err) {
2760 seq_printf(m, "%08x\n", task->personality);
2761 unlock_trace(task);
2762 }
2763 return err;
Kees Cook47830722008-10-06 03:11:58 +04002764}
2765
Eric W. Biederman801199c2006-10-02 02:18:48 -07002766/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002767 * Thread groups
2768 */
Arjan van de Ven00977a52007-02-12 00:55:34 -08002769static const struct file_operations proc_task_operations;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002770static const struct inode_operations proc_task_inode_operations;
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002771
Eric Dumazetc5141e62007-05-08 00:26:15 -07002772static const struct pid_entry tgid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002773 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2774 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2775 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08002776 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002777#ifdef CONFIG_NET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002778 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002779#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002780 REG("environ", S_IRUSR, proc_environ_operations),
2781 INF("auxv", S_IRUSR, proc_pid_auxv),
2782 ONE("status", S_IRUGO, proc_pid_status),
Al Viroa9712bc2011-03-23 15:52:50 -04002783 ONE("personality", S_IRUGO, proc_pid_personality),
Jiri Olsa3036e7b2010-09-30 15:15:33 -07002784 INF("limits", S_IRUGO, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002785#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002786 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002787#endif
Mike Galbraith5091faa2010-11-30 14:18:03 +01002788#ifdef CONFIG_SCHED_AUTOGROUP
2789 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2790#endif
john stultz4614a696b2009-12-14 18:00:05 -08002791 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
Roland McGrathebcb6732008-07-25 19:46:00 -07002792#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Al Viroa9712bc2011-03-23 15:52:50 -04002793 INF("syscall", S_IRUGO, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07002794#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002795 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2796 ONE("stat", S_IRUGO, proc_tgid_stat),
2797 ONE("statm", S_IRUGO, proc_pid_statm),
2798 REG("maps", S_IRUGO, proc_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002799#ifdef CONFIG_NUMA
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002800 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002801#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002802 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2803 LNK("cwd", proc_cwd_link),
2804 LNK("root", proc_root_link),
2805 LNK("exe", proc_exe_link),
2806 REG("mounts", S_IRUGO, proc_mounts_operations),
2807 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2808 REG("mountstats", S_IRUSR, proc_mountstats_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08002809#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002810 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2811 REG("smaps", S_IRUGO, proc_smaps_operations),
Al Viroca6b0bf2011-02-15 22:04:37 -05002812 REG("pagemap", S_IRUGO, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002813#endif
2814#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002815 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002816#endif
2817#ifdef CONFIG_KALLSYMS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002818 INF("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002819#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03002820#ifdef CONFIG_STACKTRACE
Al Viroa9712bc2011-03-23 15:52:50 -04002821 ONE("stack", S_IRUGO, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002822#endif
2823#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002824 INF("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002825#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01002826#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002827 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01002828#endif
Paul Menage8793d852007-10-18 23:39:39 -07002829#ifdef CONFIG_PROC_PID_CPUSET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002830 REG("cpuset", S_IRUGO, proc_cpuset_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002831#endif
Paul Menagea4243162007-10-18 23:39:35 -07002832#ifdef CONFIG_CGROUPS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002833 REG("cgroup", S_IRUGO, proc_cgroup_operations),
Paul Menagea4243162007-10-18 23:39:35 -07002834#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002835 INF("oom_score", S_IRUGO, proc_oom_score),
2836 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
David Rientjesa63d83f2010-08-09 17:19:46 -07002837 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002838#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002839 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2840 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002841#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002842#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002843 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002844#endif
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08002845#ifdef CONFIG_ELF_CORE
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002846 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002847#endif
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002848#ifdef CONFIG_TASK_IO_ACCOUNTING
Vasiliy Kulikov1d1221f2011-06-24 16:08:38 +04002849 INF("io", S_IRUSR, proc_tgid_io_accounting),
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002850#endif
Chris Metcalff133ecc2011-05-26 12:40:09 -04002851#ifdef CONFIG_HARDWALL
2852 INF("hardwall", S_IRUGO, proc_pid_hardwall),
2853#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002854};
2855
2856static int proc_tgid_base_readdir(struct file * filp,
2857 void * dirent, filldir_t filldir)
2858{
2859 return proc_pident_readdir(filp,dirent,filldir,
2860 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2861}
2862
Arjan van de Ven00977a52007-02-12 00:55:34 -08002863static const struct file_operations proc_tgid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002864 .read = generic_read_dir,
2865 .readdir = proc_tgid_base_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002866 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002867};
2868
2869static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002870 return proc_pident_lookup(dir, dentry,
2871 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002872}
2873
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002874static const struct inode_operations proc_tgid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002875 .lookup = proc_tgid_base_lookup,
2876 .getattr = pid_getattr,
2877 .setattr = proc_setattr,
2878};
2879
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002880static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002881{
Eric W. Biederman48e64842006-06-26 00:25:48 -07002882 struct dentry *dentry, *leader, *dir;
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002883 char buf[PROC_NUMBUF];
Eric W. Biederman48e64842006-06-26 00:25:48 -07002884 struct qstr name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002885
Eric W. Biederman48e64842006-06-26 00:25:48 -07002886 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002887 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2888 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002889 if (dentry) {
Sukadev Bhattiprolu29f12ca2009-11-11 14:26:32 -08002890 shrink_dcache_parent(dentry);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002891 d_drop(dentry);
2892 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002893 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002894
Eric W. Biederman48e64842006-06-26 00:25:48 -07002895 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002896 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2897 leader = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002898 if (!leader)
2899 goto out;
2900
2901 name.name = "task";
2902 name.len = strlen(name.name);
2903 dir = d_hash_and_lookup(leader, &name);
2904 if (!dir)
2905 goto out_put_leader;
2906
2907 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002908 name.len = snprintf(buf, sizeof(buf), "%d", pid);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002909 dentry = d_hash_and_lookup(dir, &name);
2910 if (dentry) {
2911 shrink_dcache_parent(dentry);
2912 d_drop(dentry);
2913 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002914 }
Eric W. Biederman48e64842006-06-26 00:25:48 -07002915
2916 dput(dir);
2917out_put_leader:
2918 dput(leader);
2919out:
2920 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002921}
2922
Randy Dunlap0895e912007-10-21 21:00:10 -07002923/**
2924 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2925 * @task: task that should be flushed.
2926 *
2927 * When flushing dentries from proc, one needs to flush them from global
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002928 * proc (proc_mnt) and from all the namespaces' procs this task was seen
Randy Dunlap0895e912007-10-21 21:00:10 -07002929 * in. This call is supposed to do all of this job.
2930 *
2931 * Looks in the dcache for
2932 * /proc/@pid
2933 * /proc/@tgid/task/@pid
2934 * if either directory is present flushes it and all of it'ts children
2935 * from the dcache.
2936 *
2937 * It is safe and reasonable to cache /proc entries for a task until
2938 * that task exits. After that they just clog up the dcache with
2939 * useless entries, possibly causing useful dcache entries to be
2940 * flushed instead. This routine is proved to flush those useless
2941 * dcache entries at process exit time.
2942 *
2943 * NOTE: This routine is just an optimization so it does not guarantee
2944 * that no dcache entries will exist at process exit time it
2945 * just makes it very unlikely that any will persist.
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002946 */
2947
2948void proc_flush_task(struct task_struct *task)
2949{
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002950 int i;
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07002951 struct pid *pid, *tgid;
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002952 struct upid *upid;
2953
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002954 pid = task_pid(task);
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07002955 tgid = task_tgid(task);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002956
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002957 for (i = 0; i <= pid->level; i++) {
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002958 upid = &pid->numbers[i];
2959 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07002960 tgid->numbers[i].nr);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002961 }
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -07002962
2963 upid = &pid->numbers[pid->level];
2964 if (upid->nr == 1)
2965 pid_ns_release_proc(upid->ns);
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002966}
2967
Adrian Bunk9711ef92006-12-06 20:38:31 -08002968static struct dentry *proc_pid_instantiate(struct inode *dir,
2969 struct dentry * dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002970 struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002971{
2972 struct dentry *error = ERR_PTR(-ENOENT);
2973 struct inode *inode;
2974
Eric W. Biederman61a28782006-10-02 02:18:49 -07002975 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002976 if (!inode)
2977 goto out;
2978
2979 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2980 inode->i_op = &proc_tgid_base_inode_operations;
2981 inode->i_fop = &proc_tgid_base_operations;
2982 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07002983
Miklos Szeredibfe86842011-10-28 14:13:29 +02002984 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
2985 ARRAY_SIZE(tgid_base_stuff)));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002986
Nick Pigginfb045ad2011-01-07 17:49:55 +11002987 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002988
2989 d_add(dentry, inode);
2990 /* Close the race of the process dying before we return the dentry */
2991 if (pid_revalidate(dentry, NULL))
2992 error = NULL;
2993out:
2994 return error;
2995}
2996
Linus Torvalds1da177e2005-04-16 15:20:36 -07002997struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2998{
Dan Carpenter73d36462010-05-26 14:43:25 -07002999 struct dentry *result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003000 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003001 unsigned tgid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003002 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003003
Eric W. Biederman801199c2006-10-02 02:18:48 -07003004 result = proc_base_lookup(dir, dentry);
3005 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3006 goto out;
3007
Linus Torvalds1da177e2005-04-16 15:20:36 -07003008 tgid = name_to_int(dentry);
3009 if (tgid == ~0U)
3010 goto out;
3011
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003012 ns = dentry->d_sb->s_fs_info;
Eric W. Biedermande758732006-06-26 00:25:51 -07003013 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003014 task = find_task_by_pid_ns(tgid, ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003015 if (task)
3016 get_task_struct(task);
Eric W. Biedermande758732006-06-26 00:25:51 -07003017 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003018 if (!task)
3019 goto out;
3020
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003021 result = proc_pid_instantiate(dir, dentry, task, NULL);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003022 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003023out:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07003024 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003025}
3026
Linus Torvalds1da177e2005-04-16 15:20:36 -07003027/*
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003028 * Find the first task with tgid >= tgid
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003029 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07003030 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003031struct tgid_iter {
3032 unsigned int tgid;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003033 struct task_struct *task;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003034};
3035static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3036{
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003037 struct pid *pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003038
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003039 if (iter.task)
3040 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003041 rcu_read_lock();
3042retry:
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003043 iter.task = NULL;
3044 pid = find_ge_pid(iter.tgid, ns);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003045 if (pid) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003046 iter.tgid = pid_nr_ns(pid, ns);
3047 iter.task = pid_task(pid, PIDTYPE_PID);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003048 /* What we to know is if the pid we have find is the
3049 * pid of a thread_group_leader. Testing for task
3050 * being a thread_group_leader is the obvious thing
3051 * todo but there is a window when it fails, due to
3052 * the pid transfer logic in de_thread.
3053 *
3054 * So we perform the straight forward test of seeing
3055 * if the pid we have found is the pid of a thread
3056 * group leader, and don't worry if the task we have
3057 * found doesn't happen to be a thread group leader.
3058 * As we don't care in the case of readdir.
3059 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003060 if (!iter.task || !has_group_leader_pid(iter.task)) {
3061 iter.tgid += 1;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003062 goto retry;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003063 }
3064 get_task_struct(iter.task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003065 }
Eric W. Biederman454cc102006-06-26 00:25:51 -07003066 rcu_read_unlock();
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003067 return iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003068}
3069
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003070#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003071
Eric W. Biederman61a28782006-10-02 02:18:49 -07003072static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003073 struct tgid_iter iter)
Eric W. Biederman61a28782006-10-02 02:18:49 -07003074{
3075 char name[PROC_NUMBUF];
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003076 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003077 return proc_fill_cache(filp, dirent, filldir, name, len,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003078 proc_pid_instantiate, iter.task, NULL);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003079}
3080
Linus Torvalds1da177e2005-04-16 15:20:36 -07003081/* for the /proc/ directory itself, after non-process stuff has been done */
3082int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3083{
Linus Torvaldsd8bdc592011-04-18 10:36:54 -07003084 unsigned int nr;
3085 struct task_struct *reaper;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003086 struct tgid_iter iter;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003087 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003088
Linus Torvaldsd8bdc592011-04-18 10:36:54 -07003089 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3090 goto out_no_task;
3091 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3092
3093 reaper = get_proc_task(filp->f_path.dentry->d_inode);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003094 if (!reaper)
3095 goto out_no_task;
3096
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003097 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
Eric Dumazetc5141e62007-05-08 00:26:15 -07003098 const struct pid_entry *p = &proc_base_stuff[nr];
Eric W. Biederman61a28782006-10-02 02:18:49 -07003099 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
Eric W. Biederman801199c2006-10-02 02:18:48 -07003100 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003101 }
3102
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003103 ns = filp->f_dentry->d_sb->s_fs_info;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003104 iter.task = NULL;
3105 iter.tgid = filp->f_pos - TGID_OFFSET;
3106 for (iter = next_tgid(ns, iter);
3107 iter.task;
3108 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3109 filp->f_pos = iter.tgid + TGID_OFFSET;
3110 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3111 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003112 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003113 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003114 }
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003115 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3116out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07003117 put_task_struct(reaper);
3118out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07003119 return 0;
3120}
3121
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003122/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003123 * Tasks
3124 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07003125static const struct pid_entry tid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003126 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
Jerome Marchand38355412010-04-27 13:13:06 -07003127 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08003128 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003129 REG("environ", S_IRUSR, proc_environ_operations),
3130 INF("auxv", S_IRUSR, proc_pid_auxv),
3131 ONE("status", S_IRUGO, proc_pid_status),
Al Viroa9712bc2011-03-23 15:52:50 -04003132 ONE("personality", S_IRUGO, proc_pid_personality),
Jiri Olsa3036e7b2010-09-30 15:15:33 -07003133 INF("limits", S_IRUGO, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02003134#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003135 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02003136#endif
john stultz4614a696b2009-12-14 18:00:05 -08003137 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
Roland McGrathebcb6732008-07-25 19:46:00 -07003138#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Al Viroa9712bc2011-03-23 15:52:50 -04003139 INF("syscall", S_IRUGO, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07003140#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003141 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3142 ONE("stat", S_IRUGO, proc_tid_stat),
3143 ONE("statm", S_IRUGO, proc_pid_statm),
3144 REG("maps", S_IRUGO, proc_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003145#ifdef CONFIG_NUMA
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003146 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003147#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003148 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3149 LNK("cwd", proc_cwd_link),
3150 LNK("root", proc_root_link),
3151 LNK("exe", proc_exe_link),
3152 REG("mounts", S_IRUGO, proc_mounts_operations),
3153 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08003154#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003155 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3156 REG("smaps", S_IRUGO, proc_smaps_operations),
Al Viroca6b0bf2011-02-15 22:04:37 -05003157 REG("pagemap", S_IRUGO, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003158#endif
3159#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003160 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003161#endif
3162#ifdef CONFIG_KALLSYMS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003163 INF("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003164#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03003165#ifdef CONFIG_STACKTRACE
Al Viroa9712bc2011-03-23 15:52:50 -04003166 ONE("stack", S_IRUGO, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003167#endif
3168#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003169 INF("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003170#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01003171#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003172 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01003173#endif
Paul Menage8793d852007-10-18 23:39:39 -07003174#ifdef CONFIG_PROC_PID_CPUSET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003175 REG("cpuset", S_IRUGO, proc_cpuset_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003176#endif
Paul Menagea4243162007-10-18 23:39:35 -07003177#ifdef CONFIG_CGROUPS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003178 REG("cgroup", S_IRUGO, proc_cgroup_operations),
Paul Menagea4243162007-10-18 23:39:35 -07003179#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003180 INF("oom_score", S_IRUGO, proc_oom_score),
3181 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
David Rientjesa63d83f2010-08-09 17:19:46 -07003182 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003183#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003184 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
Al Viro26ec3c62011-02-15 21:24:05 -05003185 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003186#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003187#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003188 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003189#endif
Andrea Righi297c5d92008-07-25 01:48:49 -07003190#ifdef CONFIG_TASK_IO_ACCOUNTING
Vasiliy Kulikov1d1221f2011-06-24 16:08:38 +04003191 INF("io", S_IRUSR, proc_tid_io_accounting),
Andrea Righi297c5d92008-07-25 01:48:49 -07003192#endif
Chris Metcalff133ecc2011-05-26 12:40:09 -04003193#ifdef CONFIG_HARDWALL
3194 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3195#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003196};
3197
3198static int proc_tid_base_readdir(struct file * filp,
3199 void * dirent, filldir_t filldir)
3200{
3201 return proc_pident_readdir(filp,dirent,filldir,
3202 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3203}
3204
3205static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003206 return proc_pident_lookup(dir, dentry,
3207 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003208}
3209
Arjan van de Ven00977a52007-02-12 00:55:34 -08003210static const struct file_operations proc_tid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003211 .read = generic_read_dir,
3212 .readdir = proc_tid_base_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02003213 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003214};
3215
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003216static const struct inode_operations proc_tid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003217 .lookup = proc_tid_base_lookup,
3218 .getattr = pid_getattr,
3219 .setattr = proc_setattr,
3220};
3221
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003222static struct dentry *proc_task_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07003223 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003224{
3225 struct dentry *error = ERR_PTR(-ENOENT);
3226 struct inode *inode;
Eric W. Biederman61a28782006-10-02 02:18:49 -07003227 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003228
3229 if (!inode)
3230 goto out;
3231 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3232 inode->i_op = &proc_tid_base_inode_operations;
3233 inode->i_fop = &proc_tid_base_operations;
3234 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07003235
Miklos Szeredibfe86842011-10-28 14:13:29 +02003236 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3237 ARRAY_SIZE(tid_base_stuff)));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003238
Nick Pigginfb045ad2011-01-07 17:49:55 +11003239 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003240
3241 d_add(dentry, inode);
3242 /* Close the race of the process dying before we return the dentry */
3243 if (pid_revalidate(dentry, NULL))
3244 error = NULL;
3245out:
3246 return error;
3247}
3248
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003249static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3250{
3251 struct dentry *result = ERR_PTR(-ENOENT);
3252 struct task_struct *task;
3253 struct task_struct *leader = get_proc_task(dir);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003254 unsigned tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003255 struct pid_namespace *ns;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003256
3257 if (!leader)
3258 goto out_no_task;
3259
3260 tid = name_to_int(dentry);
3261 if (tid == ~0U)
3262 goto out;
3263
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003264 ns = dentry->d_sb->s_fs_info;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003265 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003266 task = find_task_by_pid_ns(tid, ns);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003267 if (task)
3268 get_task_struct(task);
3269 rcu_read_unlock();
3270 if (!task)
3271 goto out;
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07003272 if (!same_thread_group(leader, task))
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003273 goto out_drop_task;
3274
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003275 result = proc_task_instantiate(dir, dentry, task, NULL);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003276out_drop_task:
3277 put_task_struct(task);
3278out:
3279 put_task_struct(leader);
3280out_no_task:
3281 return result;
3282}
3283
3284/*
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003285 * Find the first tid of a thread group to return to user space.
3286 *
3287 * Usually this is just the thread group leader, but if the users
3288 * buffer was too small or there was a seek into the middle of the
3289 * directory we have more work todo.
3290 *
3291 * In the case of a short read we start with find_task_by_pid.
3292 *
3293 * In the case of a seek we start with the leader and walk nr
3294 * threads past it.
3295 */
Eric W. Biedermancc288732006-06-26 00:26:01 -07003296static struct task_struct *first_tid(struct task_struct *leader,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003297 int tid, int nr, struct pid_namespace *ns)
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003298{
Oleg Nesterova872ff02006-06-26 00:26:01 -07003299 struct task_struct *pos;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003300
Eric W. Biedermancc288732006-06-26 00:26:01 -07003301 rcu_read_lock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003302 /* Attempt to start with the pid of a thread */
3303 if (tid && (nr > 0)) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003304 pos = find_task_by_pid_ns(tid, ns);
Oleg Nesterova872ff02006-06-26 00:26:01 -07003305 if (pos && (pos->group_leader == leader))
3306 goto found;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003307 }
3308
3309 /* If nr exceeds the number of threads there is nothing todo */
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003310 pos = NULL;
Oleg Nesterova872ff02006-06-26 00:26:01 -07003311 if (nr && nr >= get_nr_threads(leader))
3312 goto out;
3313
3314 /* If we haven't found our starting place yet start
3315 * with the leader and walk nr threads forward.
3316 */
3317 for (pos = leader; nr > 0; --nr) {
3318 pos = next_thread(pos);
3319 if (pos == leader) {
3320 pos = NULL;
3321 goto out;
3322 }
3323 }
3324found:
3325 get_task_struct(pos);
3326out:
Eric W. Biedermancc288732006-06-26 00:26:01 -07003327 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003328 return pos;
3329}
3330
3331/*
3332 * Find the next thread in the thread list.
3333 * Return NULL if there is an error or no next thread.
3334 *
3335 * The reference to the input task_struct is released.
3336 */
3337static struct task_struct *next_tid(struct task_struct *start)
3338{
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003339 struct task_struct *pos = NULL;
Eric W. Biedermancc288732006-06-26 00:26:01 -07003340 rcu_read_lock();
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003341 if (pid_alive(start)) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003342 pos = next_thread(start);
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003343 if (thread_group_leader(pos))
3344 pos = NULL;
3345 else
3346 get_task_struct(pos);
3347 }
Eric W. Biedermancc288732006-06-26 00:26:01 -07003348 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003349 put_task_struct(start);
3350 return pos;
3351}
3352
Eric W. Biederman61a28782006-10-02 02:18:49 -07003353static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3354 struct task_struct *task, int tid)
3355{
3356 char name[PROC_NUMBUF];
3357 int len = snprintf(name, sizeof(name), "%d", tid);
3358 return proc_fill_cache(filp, dirent, filldir, name, len,
3359 proc_task_instantiate, task, NULL);
3360}
3361
Linus Torvalds1da177e2005-04-16 15:20:36 -07003362/* for the /proc/TGID/task/ directories */
3363static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3364{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08003365 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003366 struct inode *inode = dentry->d_inode;
Guillaume Chazarain7d895242007-01-31 23:48:14 -08003367 struct task_struct *leader = NULL;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003368 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003369 int retval = -ENOENT;
3370 ino_t ino;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003371 int tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003372 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003373
Guillaume Chazarain7d895242007-01-31 23:48:14 -08003374 task = get_proc_task(inode);
3375 if (!task)
3376 goto out_no_task;
3377 rcu_read_lock();
3378 if (pid_alive(task)) {
3379 leader = task->group_leader;
3380 get_task_struct(leader);
3381 }
3382 rcu_read_unlock();
3383 put_task_struct(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003384 if (!leader)
3385 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003386 retval = 0;
3387
Linus Torvaldsee568b22009-03-17 10:02:35 -07003388 switch ((unsigned long)filp->f_pos) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003389 case 0:
3390 ino = inode->i_ino;
Zhang Leee6f7792009-03-16 14:44:31 +08003391 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003392 goto out;
Zhang Leee6f7792009-03-16 14:44:31 +08003393 filp->f_pos++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003394 /* fall through */
3395 case 1:
3396 ino = parent_ino(dentry);
Zhang Leee6f7792009-03-16 14:44:31 +08003397 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003398 goto out;
Zhang Leee6f7792009-03-16 14:44:31 +08003399 filp->f_pos++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003400 /* fall through */
3401 }
3402
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003403 /* f_version caches the tgid value that the last readdir call couldn't
3404 * return. lseek aka telldir automagically resets f_version to 0.
3405 */
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003406 ns = filp->f_dentry->d_sb->s_fs_info;
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07003407 tid = (int)filp->f_version;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003408 filp->f_version = 0;
Zhang Leee6f7792009-03-16 14:44:31 +08003409 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003410 task;
Zhang Leee6f7792009-03-16 14:44:31 +08003411 task = next_tid(task), filp->f_pos++) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003412 tid = task_pid_nr_ns(task, ns);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003413 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003414 /* returning this tgid failed, save it as the first
3415 * pid for the next readir call */
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07003416 filp->f_version = (u64)tid;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003417 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003418 break;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003419 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003420 }
3421out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07003422 put_task_struct(leader);
3423out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07003424 return retval;
3425}
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003426
3427static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3428{
3429 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07003430 struct task_struct *p = get_proc_task(inode);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003431 generic_fillattr(inode, stat);
3432
Eric W. Biederman99f89552006-06-26 00:25:55 -07003433 if (p) {
Eric W. Biederman99f89552006-06-26 00:25:55 -07003434 stat->nlink += get_nr_threads(p);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003435 put_task_struct(p);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003436 }
3437
3438 return 0;
3439}
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003440
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003441static const struct inode_operations proc_task_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003442 .lookup = proc_task_lookup,
3443 .getattr = proc_task_getattr,
3444 .setattr = proc_setattr,
3445};
3446
Arjan van de Ven00977a52007-02-12 00:55:34 -08003447static const struct file_operations proc_task_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003448 .read = generic_read_dir,
3449 .readdir = proc_task_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02003450 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003451};