blob: 437195f204e14e908bdda87ace0e8b0ce81efb21 [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>
Eric W. Biederman22d917d2011-11-17 00:11:58 -080084#include <linux/user_namespace.h>
Al Viro5ad4e532009-03-29 19:50:06 -040085#include <linux/fs_struct.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090086#include <linux/slab.h>
Pavel Emelyanov640708a2012-01-10 15:11:23 -080087#include <linux/flex_array.h>
Chris Metcalff133ecc2011-05-26 12:40:09 -040088#ifdef CONFIG_HARDWALL
89#include <asm/hardwall.h>
90#endif
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -080091#include <trace/events/oom.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070092#include "internal.h"
93
Eric W. Biederman0f2fe202006-06-26 00:25:46 -070094/* NOTE:
95 * Implementing inode permission operations in /proc is almost
96 * certainly an error. Permission checks need to happen during
97 * each system call not at open time. The reason is that most of
98 * what we wish to check for permissions in /proc varies at runtime.
99 *
100 * The classic example of a problem is opening file descriptors
101 * in /proc for a task before it execs a suid executable.
102 */
103
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104struct pid_entry {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 char *name;
Eric Dumazetc5141e62007-05-08 00:26:15 -0700106 int len;
Al Virod161a132011-07-24 03:36:29 -0400107 umode_t mode;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800108 const struct inode_operations *iop;
Arjan van de Ven00977a52007-02-12 00:55:34 -0800109 const struct file_operations *fop;
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700110 union proc_op op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111};
112
Eric W. Biederman61a28782006-10-02 02:18:49 -0700113#define NOD(NAME, MODE, IOP, FOP, OP) { \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700114 .name = (NAME), \
Eric Dumazetc5141e62007-05-08 00:26:15 -0700115 .len = sizeof(NAME) - 1, \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700116 .mode = MODE, \
117 .iop = IOP, \
118 .fop = FOP, \
119 .op = OP, \
120}
121
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300122#define DIR(NAME, MODE, iops, fops) \
123 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
124#define LNK(NAME, get_link) \
Eric W. Biederman61a28782006-10-02 02:18:49 -0700125 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700126 &proc_pid_link_inode_operations, NULL, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300127 { .proc_get_link = get_link } )
128#define REG(NAME, MODE, fops) \
129 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
130#define INF(NAME, MODE, read) \
Eric W. Biederman61a28782006-10-02 02:18:49 -0700131 NOD(NAME, (S_IFREG|(MODE)), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700132 NULL, &proc_info_file_operations, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300133 { .proc_read = read } )
134#define ONE(NAME, MODE, show) \
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800135 NOD(NAME, (S_IFREG|(MODE)), \
136 NULL, &proc_single_file_operations, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300137 { .proc_show = show } )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138
Pavel Emelyanov640708a2012-01-10 15:11:23 -0800139static int proc_fd_permission(struct inode *inode, int mask);
140
Vegard Nossumaed54172008-06-05 22:46:53 -0700141/*
142 * Count the number of hardlinks for the pid_entry table, excluding the .
143 * and .. links.
144 */
145static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
146 unsigned int n)
147{
148 unsigned int i;
149 unsigned int count;
150
151 count = 0;
152 for (i = 0; i < n; ++i) {
153 if (S_ISDIR(entries[i].mode))
154 ++count;
155 }
156
157 return count;
158}
159
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200160static int get_task_root(struct task_struct *task, struct path *root)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161{
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000162 int result = -ENOENT;
163
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700164 task_lock(task);
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200165 if (task->fs) {
166 get_fs_root(task->fs, root);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000167 result = 0;
168 }
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700169 task_unlock(task);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000170 return result;
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700171}
172
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800173static int proc_cwd_link(struct dentry *dentry, struct path *path)
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700174{
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800175 struct task_struct *task = get_proc_task(dentry->d_inode);
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700176 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700177
178 if (task) {
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200179 task_lock(task);
180 if (task->fs) {
181 get_fs_pwd(task->fs, path);
182 result = 0;
183 }
184 task_unlock(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700185 put_task_struct(task);
186 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 return result;
188}
189
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800190static int proc_root_link(struct dentry *dentry, struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191{
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800192 struct task_struct *task = get_proc_task(dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700194
195 if (task) {
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200196 result = get_task_root(task, path);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700197 put_task_struct(task);
198 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 return result;
200}
201
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202static int proc_pid_cmdline(struct task_struct *task, char * buffer)
203{
204 int res = 0;
205 unsigned int len;
206 struct mm_struct *mm = get_task_mm(task);
207 if (!mm)
208 goto out;
209 if (!mm->arg_end)
210 goto out_mm; /* Shh! No looking before we're done */
211
212 len = mm->arg_end - mm->arg_start;
213
214 if (len > PAGE_SIZE)
215 len = PAGE_SIZE;
216
217 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
218
219 // If the nul at the end of args has been overwritten, then
220 // assume application is using setproctitle(3).
221 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
222 len = strnlen(buffer, res);
223 if (len < res) {
224 res = len;
225 } else {
226 len = mm->env_end - mm->env_start;
227 if (len > PAGE_SIZE - res)
228 len = PAGE_SIZE - res;
229 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
230 res = strnlen(buffer, res);
231 }
232 }
233out_mm:
234 mmput(mm);
235out:
236 return res;
237}
238
239static int proc_pid_auxv(struct task_struct *task, char *buffer)
240{
Cong Wange7dcd992012-05-31 16:26:17 -0700241 struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ);
Al Viro2fadaef2011-02-15 22:52:11 -0500242 int res = PTR_ERR(mm);
243 if (mm && !IS_ERR(mm)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 unsigned int nwords = 0;
Hannes Ederdfe6b7d2008-12-30 18:49:13 +0300245 do {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 nwords += 2;
Hannes Ederdfe6b7d2008-12-30 18:49:13 +0300247 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 res = nwords * sizeof(mm->saved_auxv[0]);
249 if (res > PAGE_SIZE)
250 res = PAGE_SIZE;
251 memcpy(buffer, mm->saved_auxv, res);
252 mmput(mm);
253 }
254 return res;
255}
256
257
258#ifdef CONFIG_KALLSYMS
259/*
260 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
261 * Returns the resolved symbol. If that fails, simply return the address.
262 */
263static int proc_pid_wchan(struct task_struct *task, char *buffer)
264{
Alexey Dobriyanffb45122007-05-08 00:28:41 -0700265 unsigned long wchan;
Tejun Heo9281ace2007-07-17 04:03:51 -0700266 char symname[KSYM_NAME_LEN];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267
268 wchan = get_wchan(task);
269
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700270 if (lookup_symbol_name(wchan, symname) < 0)
Jake Edgef83ce3e2009-05-04 12:51:14 -0600271 if (!ptrace_may_access(task, PTRACE_MODE_READ))
272 return 0;
273 else
274 return sprintf(buffer, "%lu", wchan);
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700275 else
276 return sprintf(buffer, "%s", symname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277}
278#endif /* CONFIG_KALLSYMS */
279
Al Viroa9712bc2011-03-23 15:52:50 -0400280static int lock_trace(struct task_struct *task)
281{
282 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
283 if (err)
284 return err;
285 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
286 mutex_unlock(&task->signal->cred_guard_mutex);
287 return -EPERM;
288 }
289 return 0;
290}
291
292static void unlock_trace(struct task_struct *task)
293{
294 mutex_unlock(&task->signal->cred_guard_mutex);
295}
296
Ken Chen2ec220e2008-11-10 11:26:08 +0300297#ifdef CONFIG_STACKTRACE
298
299#define MAX_STACK_TRACE_DEPTH 64
300
301static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
302 struct pid *pid, struct task_struct *task)
303{
304 struct stack_trace trace;
305 unsigned long *entries;
Al Viroa9712bc2011-03-23 15:52:50 -0400306 int err;
Ken Chen2ec220e2008-11-10 11:26:08 +0300307 int i;
308
309 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
310 if (!entries)
311 return -ENOMEM;
312
313 trace.nr_entries = 0;
314 trace.max_entries = MAX_STACK_TRACE_DEPTH;
315 trace.entries = entries;
316 trace.skip = 0;
Ken Chen2ec220e2008-11-10 11:26:08 +0300317
Al Viroa9712bc2011-03-23 15:52:50 -0400318 err = lock_trace(task);
319 if (!err) {
320 save_stack_trace_tsk(task, &trace);
321
322 for (i = 0; i < trace.nr_entries; i++) {
Linus Torvaldsb81a6182011-03-23 20:51:42 -0700323 seq_printf(m, "[<%pK>] %pS\n",
Al Viroa9712bc2011-03-23 15:52:50 -0400324 (void *)entries[i], (void *)entries[i]);
325 }
326 unlock_trace(task);
Ken Chen2ec220e2008-11-10 11:26:08 +0300327 }
328 kfree(entries);
329
Al Viroa9712bc2011-03-23 15:52:50 -0400330 return err;
Ken Chen2ec220e2008-11-10 11:26:08 +0300331}
332#endif
333
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334#ifdef CONFIG_SCHEDSTATS
335/*
336 * Provides /proc/PID/schedstat
337 */
338static int proc_pid_schedstat(struct task_struct *task, char *buffer)
339{
Balbir Singh172ba842007-07-09 18:52:00 +0200340 return sprintf(buffer, "%llu %llu %lu\n",
Ingo Molnar826e08b2008-12-22 07:37:41 +0100341 (unsigned long long)task->se.sum_exec_runtime,
342 (unsigned long long)task->sched_info.run_delay,
Ingo Molnar2d723762007-10-15 17:00:12 +0200343 task->sched_info.pcount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344}
345#endif
346
Arjan van de Ven97455122008-01-25 21:08:34 +0100347#ifdef CONFIG_LATENCYTOP
348static int lstats_show_proc(struct seq_file *m, void *v)
349{
350 int i;
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800351 struct inode *inode = m->private;
352 struct task_struct *task = get_proc_task(inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100353
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800354 if (!task)
355 return -ESRCH;
356 seq_puts(m, "Latency Top version : v0.1\n");
Arjan van de Ven97455122008-01-25 21:08:34 +0100357 for (i = 0; i < 32; i++) {
Joe Perches34e49d42011-01-12 17:00:30 -0800358 struct latency_record *lr = &task->latency_record[i];
359 if (lr->backtrace[0]) {
Arjan van de Ven97455122008-01-25 21:08:34 +0100360 int q;
Joe Perches34e49d42011-01-12 17:00:30 -0800361 seq_printf(m, "%i %li %li",
362 lr->count, lr->time, lr->max);
Arjan van de Ven97455122008-01-25 21:08:34 +0100363 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
Joe Perches34e49d42011-01-12 17:00:30 -0800364 unsigned long bt = lr->backtrace[q];
365 if (!bt)
Arjan van de Ven97455122008-01-25 21:08:34 +0100366 break;
Joe Perches34e49d42011-01-12 17:00:30 -0800367 if (bt == ULONG_MAX)
Arjan van de Ven97455122008-01-25 21:08:34 +0100368 break;
Joe Perches34e49d42011-01-12 17:00:30 -0800369 seq_printf(m, " %ps", (void *)bt);
Arjan van de Ven97455122008-01-25 21:08:34 +0100370 }
Alexey Dobriyan9d6de122011-01-12 17:00:32 -0800371 seq_putc(m, '\n');
Arjan van de Ven97455122008-01-25 21:08:34 +0100372 }
373
374 }
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800375 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100376 return 0;
377}
378
379static int lstats_open(struct inode *inode, struct file *file)
380{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800381 return single_open(file, lstats_show_proc, inode);
Hiroshi Shimamotod6643d12008-02-14 10:27:00 -0800382}
383
Arjan van de Ven97455122008-01-25 21:08:34 +0100384static ssize_t lstats_write(struct file *file, const char __user *buf,
385 size_t count, loff_t *offs)
386{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800387 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100388
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800389 if (!task)
390 return -ESRCH;
Arjan van de Ven97455122008-01-25 21:08:34 +0100391 clear_all_latency_tracing(task);
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800392 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100393
394 return count;
395}
396
397static const struct file_operations proc_lstats_operations = {
398 .open = lstats_open,
399 .read = seq_read,
400 .write = lstats_write,
401 .llseek = seq_lseek,
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800402 .release = single_release,
Arjan van de Ven97455122008-01-25 21:08:34 +0100403};
404
405#endif
406
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407static int proc_oom_score(struct task_struct *task, char *buffer)
408{
David Rientjesa7f638f2012-05-29 15:06:47 -0700409 unsigned long totalpages = totalram_pages + total_swap_pages;
Oleg Nesterovb95c35e2010-04-01 15:13:57 +0200410 unsigned long points = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700412 read_lock(&tasklist_lock);
Oleg Nesterovb95c35e2010-04-01 15:13:57 +0200413 if (pid_alive(task))
David Rientjesa7f638f2012-05-29 15:06:47 -0700414 points = oom_badness(task, NULL, NULL, totalpages) *
415 1000 / totalpages;
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700416 read_unlock(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 return sprintf(buffer, "%lu\n", points);
418}
419
Neil Hormand85f50d2007-10-18 23:40:37 -0700420struct limit_names {
421 char *name;
422 char *unit;
423};
424
425static const struct limit_names lnames[RLIM_NLIMITS] = {
Kees Cookcff4edb2009-09-22 16:45:32 -0700426 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700427 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
428 [RLIMIT_DATA] = {"Max data size", "bytes"},
429 [RLIMIT_STACK] = {"Max stack size", "bytes"},
430 [RLIMIT_CORE] = {"Max core file size", "bytes"},
431 [RLIMIT_RSS] = {"Max resident set", "bytes"},
432 [RLIMIT_NPROC] = {"Max processes", "processes"},
433 [RLIMIT_NOFILE] = {"Max open files", "files"},
434 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
435 [RLIMIT_AS] = {"Max address space", "bytes"},
436 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
437 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
438 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
439 [RLIMIT_NICE] = {"Max nice priority", NULL},
440 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
Eugene Teo88081172008-02-23 15:23:52 -0800441 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700442};
443
444/* Display limits for a process */
445static int proc_pid_limits(struct task_struct *task, char *buffer)
446{
447 unsigned int i;
448 int count = 0;
449 unsigned long flags;
450 char *bufptr = buffer;
451
452 struct rlimit rlim[RLIM_NLIMITS];
453
Lai Jiangshana6bebbc2008-10-05 00:51:15 +0400454 if (!lock_task_sighand(task, &flags))
Neil Hormand85f50d2007-10-18 23:40:37 -0700455 return 0;
Neil Hormand85f50d2007-10-18 23:40:37 -0700456 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
457 unlock_task_sighand(task, &flags);
Neil Hormand85f50d2007-10-18 23:40:37 -0700458
459 /*
460 * print the file header
461 */
462 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
463 "Limit", "Soft Limit", "Hard Limit", "Units");
464
465 for (i = 0; i < RLIM_NLIMITS; i++) {
466 if (rlim[i].rlim_cur == RLIM_INFINITY)
467 count += sprintf(&bufptr[count], "%-25s %-20s ",
468 lnames[i].name, "unlimited");
469 else
470 count += sprintf(&bufptr[count], "%-25s %-20lu ",
471 lnames[i].name, rlim[i].rlim_cur);
472
473 if (rlim[i].rlim_max == RLIM_INFINITY)
474 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
475 else
476 count += sprintf(&bufptr[count], "%-20lu ",
477 rlim[i].rlim_max);
478
479 if (lnames[i].unit)
480 count += sprintf(&bufptr[count], "%-10s\n",
481 lnames[i].unit);
482 else
483 count += sprintf(&bufptr[count], "\n");
484 }
485
486 return count;
487}
488
Roland McGrathebcb6732008-07-25 19:46:00 -0700489#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
490static int proc_pid_syscall(struct task_struct *task, char *buffer)
491{
492 long nr;
493 unsigned long args[6], sp, pc;
Al Viroa9712bc2011-03-23 15:52:50 -0400494 int res = lock_trace(task);
495 if (res)
496 return res;
Roland McGrathebcb6732008-07-25 19:46:00 -0700497
498 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
Al Viroa9712bc2011-03-23 15:52:50 -0400499 res = sprintf(buffer, "running\n");
500 else if (nr < 0)
501 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
502 else
503 res = sprintf(buffer,
Roland McGrathebcb6732008-07-25 19:46:00 -0700504 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
505 nr,
506 args[0], args[1], args[2], args[3], args[4], args[5],
507 sp, pc);
Al Viroa9712bc2011-03-23 15:52:50 -0400508 unlock_trace(task);
509 return res;
Roland McGrathebcb6732008-07-25 19:46:00 -0700510}
511#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
512
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513/************************************************************************/
514/* Here the fs part begins */
515/************************************************************************/
516
517/* permission checks */
Eric W. Biederman778c1142006-06-26 00:25:58 -0700518static int proc_fd_access_allowed(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519{
Eric W. Biederman778c1142006-06-26 00:25:58 -0700520 struct task_struct *task;
521 int allowed = 0;
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700522 /* Allow access to a task's file descriptors if it is us or we
523 * may use ptrace attach to the process and find out that
524 * information.
Eric W. Biederman778c1142006-06-26 00:25:58 -0700525 */
526 task = get_proc_task(inode);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700527 if (task) {
Stephen Smalley006ebb42008-05-19 08:32:49 -0400528 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
Eric W. Biederman778c1142006-06-26 00:25:58 -0700529 put_task_struct(task);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700530 }
Eric W. Biederman778c1142006-06-26 00:25:58 -0700531 return allowed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532}
533
Eric W. Biederman6b4e3062010-03-07 16:41:34 -0800534int proc_setattr(struct dentry *dentry, struct iattr *attr)
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700535{
536 int error;
537 struct inode *inode = dentry->d_inode;
538
539 if (attr->ia_valid & ATTR_MODE)
540 return -EPERM;
541
542 error = inode_change_ok(inode, attr);
Christoph Hellwig10257742010-06-04 11:30:02 +0200543 if (error)
544 return error;
545
546 if ((attr->ia_valid & ATTR_SIZE) &&
547 attr->ia_size != i_size_read(inode)) {
548 error = vmtruncate(inode, attr->ia_size);
549 if (error)
550 return error;
551 }
552
553 setattr_copy(inode, attr);
554 mark_inode_dirty(inode);
555 return 0;
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700556}
557
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800558/*
559 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
560 * or euid/egid (for hide_pid_min=2)?
561 */
562static bool has_pid_permissions(struct pid_namespace *pid,
563 struct task_struct *task,
564 int hide_pid_min)
565{
566 if (pid->hide_pid < hide_pid_min)
567 return true;
568 if (in_group_p(pid->pid_gid))
569 return true;
570 return ptrace_may_access(task, PTRACE_MODE_READ);
571}
572
573
574static int proc_pid_permission(struct inode *inode, int mask)
575{
576 struct pid_namespace *pid = inode->i_sb->s_fs_info;
577 struct task_struct *task;
578 bool has_perms;
579
580 task = get_proc_task(inode);
Xiaotian Fenga2ef9902012-01-12 17:17:08 -0800581 if (!task)
582 return -ESRCH;
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800583 has_perms = has_pid_permissions(pid, task, 1);
584 put_task_struct(task);
585
586 if (!has_perms) {
587 if (pid->hide_pid == 2) {
588 /*
589 * Let's make getdents(), stat(), and open()
590 * consistent with each other. If a process
591 * may not stat() a file, it shouldn't be seen
592 * in procfs at all.
593 */
594 return -ENOENT;
595 }
596
597 return -EPERM;
598 }
599 return generic_permission(inode, mask);
600}
601
602
603
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800604static const struct inode_operations proc_def_inode_operations = {
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700605 .setattr = proc_setattr,
606};
607
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
609
610static ssize_t proc_info_read(struct file * file, char __user * buf,
611 size_t count, loff_t *ppos)
612{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800613 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 unsigned long page;
615 ssize_t length;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700616 struct task_struct *task = get_proc_task(inode);
617
618 length = -ESRCH;
619 if (!task)
620 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621
622 if (count > PROC_BLOCK_SIZE)
623 count = PROC_BLOCK_SIZE;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700624
625 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700626 if (!(page = __get_free_page(GFP_TEMPORARY)))
Eric W. Biederman99f89552006-06-26 00:25:55 -0700627 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628
629 length = PROC_I(inode)->op.proc_read(task, (char*)page);
630
631 if (length >= 0)
632 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
633 free_page(page);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700634out:
635 put_task_struct(task);
636out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 return length;
638}
639
Arjan van de Ven00977a52007-02-12 00:55:34 -0800640static const struct file_operations proc_info_file_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 .read = proc_info_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +0100642 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643};
644
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800645static int proc_single_show(struct seq_file *m, void *v)
646{
647 struct inode *inode = m->private;
648 struct pid_namespace *ns;
649 struct pid *pid;
650 struct task_struct *task;
651 int ret;
652
653 ns = inode->i_sb->s_fs_info;
654 pid = proc_pid(inode);
655 task = get_pid_task(pid, PIDTYPE_PID);
656 if (!task)
657 return -ESRCH;
658
659 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
660
661 put_task_struct(task);
662 return ret;
663}
664
665static int proc_single_open(struct inode *inode, struct file *filp)
666{
Jovi Zhangc6a34052011-01-12 17:00:34 -0800667 return single_open(filp, proc_single_show, inode);
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800668}
669
670static const struct file_operations proc_single_file_operations = {
671 .open = proc_single_open,
672 .read = seq_read,
673 .llseek = seq_lseek,
674 .release = single_release,
675};
676
Cong Wangb409e572012-05-31 16:26:17 -0700677static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678{
Linus Torvaldse2683372012-01-17 15:21:19 -0800679 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
680 struct mm_struct *mm;
681
682 if (!task)
683 return -ESRCH;
684
Cong Wangb409e572012-05-31 16:26:17 -0700685 mm = mm_access(task, mode);
Linus Torvaldse2683372012-01-17 15:21:19 -0800686 put_task_struct(task);
687
688 if (IS_ERR(mm))
689 return PTR_ERR(mm);
690
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100691 if (mm) {
692 /* ensure this mm_struct can't be freed */
693 atomic_inc(&mm->mm_count);
694 /* but do not pin its memory */
695 mmput(mm);
696 }
697
KAMEZAWA Hiroyuki4a3956c2010-10-01 14:20:22 -0700698 /* OK to pass negative loff_t, we can catch out-of-range */
699 file->f_mode |= FMODE_UNSIGNED_OFFSET;
Linus Torvaldse2683372012-01-17 15:21:19 -0800700 file->private_data = mm;
701
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 return 0;
703}
704
Cong Wangb409e572012-05-31 16:26:17 -0700705static int mem_open(struct inode *inode, struct file *file)
706{
707 return __mem_open(inode, file, PTRACE_MODE_ATTACH);
708}
709
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100710static ssize_t mem_rw(struct file *file, char __user *buf,
711 size_t count, loff_t *ppos, int write)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712{
Linus Torvaldse2683372012-01-17 15:21:19 -0800713 struct mm_struct *mm = file->private_data;
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100714 unsigned long addr = *ppos;
715 ssize_t copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 char *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717
Linus Torvaldse2683372012-01-17 15:21:19 -0800718 if (!mm)
719 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720
Mel Gormane12ba742007-10-16 01:25:52 -0700721 page = (char *)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722 if (!page)
Linus Torvaldse2683372012-01-17 15:21:19 -0800723 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700725 copied = 0;
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100726 if (!atomic_inc_not_zero(&mm->mm_users))
727 goto free;
728
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 while (count > 0) {
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100730 int this_len = min_t(int, count, PAGE_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100732 if (write && copy_from_user(page, buf, this_len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 copied = -EFAULT;
734 break;
735 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100736
737 this_len = access_remote_vm(mm, addr, page, this_len, write);
738 if (!this_len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739 if (!copied)
740 copied = -EIO;
741 break;
742 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100743
744 if (!write && copy_to_user(buf, page, this_len)) {
745 copied = -EFAULT;
746 break;
747 }
748
749 buf += this_len;
750 addr += this_len;
751 copied += this_len;
752 count -= this_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100754 *ppos = addr;
KOSAKI Motohiro30cd8902011-05-26 16:25:52 -0700755
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100756 mmput(mm);
757free:
KOSAKI Motohiro30cd8902011-05-26 16:25:52 -0700758 free_page((unsigned long) page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 return copied;
760}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100762static ssize_t mem_read(struct file *file, char __user *buf,
763 size_t count, loff_t *ppos)
764{
765 return mem_rw(file, buf, count, ppos, 0);
766}
767
768static ssize_t mem_write(struct file *file, const char __user *buf,
769 size_t count, loff_t *ppos)
770{
771 return mem_rw(file, (char __user*)buf, count, ppos, 1);
772}
773
Matt Mackall85863e42008-02-04 22:29:04 -0800774loff_t mem_lseek(struct file *file, loff_t offset, int orig)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775{
776 switch (orig) {
777 case 0:
778 file->f_pos = offset;
779 break;
780 case 1:
781 file->f_pos += offset;
782 break;
783 default:
784 return -EINVAL;
785 }
786 force_successful_syscall_return();
787 return file->f_pos;
788}
789
Linus Torvaldse2683372012-01-17 15:21:19 -0800790static int mem_release(struct inode *inode, struct file *file)
791{
792 struct mm_struct *mm = file->private_data;
Oleg Nesterov71879d32012-01-31 17:14:38 +0100793 if (mm)
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100794 mmdrop(mm);
Linus Torvaldse2683372012-01-17 15:21:19 -0800795 return 0;
796}
797
Arjan van de Ven00977a52007-02-12 00:55:34 -0800798static const struct file_operations proc_mem_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 .llseek = mem_lseek,
800 .read = mem_read,
801 .write = mem_write,
802 .open = mem_open,
Linus Torvaldse2683372012-01-17 15:21:19 -0800803 .release = mem_release,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804};
805
Cong Wangb409e572012-05-31 16:26:17 -0700806static int environ_open(struct inode *inode, struct file *file)
807{
808 return __mem_open(inode, file, PTRACE_MODE_READ);
809}
810
James Pearson315e28c2007-10-16 23:30:17 -0700811static ssize_t environ_read(struct file *file, char __user *buf,
812 size_t count, loff_t *ppos)
813{
James Pearson315e28c2007-10-16 23:30:17 -0700814 char *page;
815 unsigned long src = *ppos;
Cong Wangb409e572012-05-31 16:26:17 -0700816 int ret = 0;
817 struct mm_struct *mm = file->private_data;
James Pearson315e28c2007-10-16 23:30:17 -0700818
Cong Wangb409e572012-05-31 16:26:17 -0700819 if (!mm)
820 return 0;
James Pearson315e28c2007-10-16 23:30:17 -0700821
James Pearson315e28c2007-10-16 23:30:17 -0700822 page = (char *)__get_free_page(GFP_TEMPORARY);
823 if (!page)
Cong Wangb409e572012-05-31 16:26:17 -0700824 return -ENOMEM;
James Pearson315e28c2007-10-16 23:30:17 -0700825
Al Virod6f64b82011-02-15 22:26:01 -0500826 ret = 0;
Cong Wangb409e572012-05-31 16:26:17 -0700827 if (!atomic_inc_not_zero(&mm->mm_users))
828 goto free;
James Pearson315e28c2007-10-16 23:30:17 -0700829 while (count > 0) {
830 int this_len, retval, max_len;
831
832 this_len = mm->env_end - (mm->env_start + src);
833
834 if (this_len <= 0)
835 break;
836
837 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
838 this_len = (this_len > max_len) ? max_len : this_len;
839
Cong Wangb409e572012-05-31 16:26:17 -0700840 retval = access_remote_vm(mm, (mm->env_start + src),
James Pearson315e28c2007-10-16 23:30:17 -0700841 page, this_len, 0);
842
843 if (retval <= 0) {
844 ret = retval;
845 break;
846 }
847
848 if (copy_to_user(buf, page, retval)) {
849 ret = -EFAULT;
850 break;
851 }
852
853 ret += retval;
854 src += retval;
855 buf += retval;
856 count -= retval;
857 }
858 *ppos = src;
James Pearson315e28c2007-10-16 23:30:17 -0700859 mmput(mm);
Cong Wangb409e572012-05-31 16:26:17 -0700860
861free:
James Pearson315e28c2007-10-16 23:30:17 -0700862 free_page((unsigned long) page);
James Pearson315e28c2007-10-16 23:30:17 -0700863 return ret;
864}
865
866static const struct file_operations proc_environ_operations = {
Cong Wangb409e572012-05-31 16:26:17 -0700867 .open = environ_open,
James Pearson315e28c2007-10-16 23:30:17 -0700868 .read = environ_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +0100869 .llseek = generic_file_llseek,
Cong Wangb409e572012-05-31 16:26:17 -0700870 .release = mem_release,
James Pearson315e28c2007-10-16 23:30:17 -0700871};
872
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873static ssize_t oom_adjust_read(struct file *file, char __user *buf,
874 size_t count, loff_t *ppos)
875{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800876 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Eric W. Biederman8578cea2006-06-26 00:25:54 -0700877 char buffer[PROC_NUMBUF];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 size_t len;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -0700879 int oom_adjust = OOM_DISABLE;
880 unsigned long flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881
Eric W. Biederman99f89552006-06-26 00:25:55 -0700882 if (!task)
883 return -ESRCH;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -0700884
885 if (lock_task_sighand(task, &flags)) {
886 oom_adjust = task->signal->oom_adj;
887 unlock_task_sighand(task, &flags);
888 }
889
Eric W. Biederman99f89552006-06-26 00:25:55 -0700890 put_task_struct(task);
891
Eric W. Biederman8578cea2006-06-26 00:25:54 -0700892 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
Akinobu Mita0c28f282007-05-08 00:31:41 -0700893
894 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895}
896
897static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
898 size_t count, loff_t *ppos)
899{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700900 struct task_struct *task;
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -0700901 char buffer[PROC_NUMBUF];
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -0700902 int oom_adjust;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -0700903 unsigned long flags;
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -0700904 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905
Eric W. Biederman8578cea2006-06-26 00:25:54 -0700906 memset(buffer, 0, sizeof(buffer));
907 if (count > sizeof(buffer) - 1)
908 count = sizeof(buffer) - 1;
David Rientjes723548b2010-10-26 14:21:25 -0700909 if (copy_from_user(buffer, buf, count)) {
910 err = -EFAULT;
911 goto out;
912 }
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -0700913
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -0700914 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -0700915 if (err)
David Rientjes723548b2010-10-26 14:21:25 -0700916 goto out;
Alexey Dobriyan8ac773b2006-10-19 23:28:32 -0700917 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
David Rientjes723548b2010-10-26 14:21:25 -0700918 oom_adjust != OOM_DISABLE) {
919 err = -EINVAL;
920 goto out;
921 }
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -0700922
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800923 task = get_proc_task(file->f_path.dentry->d_inode);
David Rientjes723548b2010-10-26 14:21:25 -0700924 if (!task) {
925 err = -ESRCH;
926 goto out;
927 }
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -0700928
Ying Han3d5992d2010-10-26 14:21:23 -0700929 task_lock(task);
930 if (!task->mm) {
David Rientjes723548b2010-10-26 14:21:25 -0700931 err = -EINVAL;
932 goto err_task_lock;
Ying Han3d5992d2010-10-26 14:21:23 -0700933 }
934
David Rientjesd19d5472010-10-26 14:21:26 -0700935 if (!lock_task_sighand(task, &flags)) {
936 err = -ESRCH;
937 goto err_task_lock;
938 }
939
940 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
941 err = -EACCES;
942 goto err_sighand;
943 }
944
David Rientjes51b1bd22010-08-09 17:19:47 -0700945 /*
946 * Warn that /proc/pid/oom_adj is deprecated, see
947 * Documentation/feature-removal-schedule.txt.
948 */
Linus Torvaldsc2142702011-08-06 11:43:08 -0700949 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 -0700950 current->comm, task_pid_nr(current), task_pid_nr(task),
951 task_pid_nr(task));
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -0700952 task->signal->oom_adj = oom_adjust;
David Rientjesa63d83f2010-08-09 17:19:46 -0700953 /*
954 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
955 * value is always attainable.
956 */
957 if (task->signal->oom_adj == OOM_ADJUST_MAX)
958 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
959 else
960 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
961 -OOM_DISABLE;
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -0800962 trace_oom_score_adj_update(task);
David Rientjes723548b2010-10-26 14:21:25 -0700963err_sighand:
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -0700964 unlock_task_sighand(task, &flags);
David Rientjesd19d5472010-10-26 14:21:26 -0700965err_task_lock:
966 task_unlock(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700967 put_task_struct(task);
David Rientjes723548b2010-10-26 14:21:25 -0700968out:
969 return err < 0 ? err : count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970}
971
Arjan van de Ven00977a52007-02-12 00:55:34 -0800972static const struct file_operations proc_oom_adjust_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 .read = oom_adjust_read,
974 .write = oom_adjust_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +0100975 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976};
977
David Rientjesa63d83f2010-08-09 17:19:46 -0700978static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
979 size_t count, loff_t *ppos)
980{
981 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
982 char buffer[PROC_NUMBUF];
983 int oom_score_adj = OOM_SCORE_ADJ_MIN;
984 unsigned long flags;
985 size_t len;
986
987 if (!task)
988 return -ESRCH;
989 if (lock_task_sighand(task, &flags)) {
990 oom_score_adj = task->signal->oom_score_adj;
991 unlock_task_sighand(task, &flags);
992 }
993 put_task_struct(task);
994 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
995 return simple_read_from_buffer(buf, count, ppos, buffer, len);
996}
997
998static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
999 size_t count, loff_t *ppos)
1000{
1001 struct task_struct *task;
1002 char buffer[PROC_NUMBUF];
1003 unsigned long flags;
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001004 int oom_score_adj;
David Rientjesa63d83f2010-08-09 17:19:46 -07001005 int err;
1006
1007 memset(buffer, 0, sizeof(buffer));
1008 if (count > sizeof(buffer) - 1)
1009 count = sizeof(buffer) - 1;
David Rientjes723548b2010-10-26 14:21:25 -07001010 if (copy_from_user(buffer, buf, count)) {
1011 err = -EFAULT;
1012 goto out;
1013 }
David Rientjesa63d83f2010-08-09 17:19:46 -07001014
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001015 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
David Rientjesa63d83f2010-08-09 17:19:46 -07001016 if (err)
David Rientjes723548b2010-10-26 14:21:25 -07001017 goto out;
David Rientjesa63d83f2010-08-09 17:19:46 -07001018 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
David Rientjes723548b2010-10-26 14:21:25 -07001019 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1020 err = -EINVAL;
1021 goto out;
1022 }
David Rientjesa63d83f2010-08-09 17:19:46 -07001023
1024 task = get_proc_task(file->f_path.dentry->d_inode);
David Rientjes723548b2010-10-26 14:21:25 -07001025 if (!task) {
1026 err = -ESRCH;
1027 goto out;
1028 }
David Rientjesa63d83f2010-08-09 17:19:46 -07001029
Ying Han3d5992d2010-10-26 14:21:23 -07001030 task_lock(task);
1031 if (!task->mm) {
David Rientjes723548b2010-10-26 14:21:25 -07001032 err = -EINVAL;
1033 goto err_task_lock;
Ying Han3d5992d2010-10-26 14:21:23 -07001034 }
David Rientjesd19d5472010-10-26 14:21:26 -07001035
1036 if (!lock_task_sighand(task, &flags)) {
1037 err = -ESRCH;
1038 goto err_task_lock;
1039 }
1040
Mandeep Singh Bainesdabb16f2011-01-13 15:46:05 -08001041 if (oom_score_adj < task->signal->oom_score_adj_min &&
David Rientjesd19d5472010-10-26 14:21:26 -07001042 !capable(CAP_SYS_RESOURCE)) {
1043 err = -EACCES;
1044 goto err_sighand;
1045 }
1046
David Rientjesa63d83f2010-08-09 17:19:46 -07001047 task->signal->oom_score_adj = oom_score_adj;
Mandeep Singh Bainesdabb16f2011-01-13 15:46:05 -08001048 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1049 task->signal->oom_score_adj_min = oom_score_adj;
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08001050 trace_oom_score_adj_update(task);
David Rientjesa63d83f2010-08-09 17:19:46 -07001051 /*
1052 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1053 * always attainable.
1054 */
1055 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1056 task->signal->oom_adj = OOM_DISABLE;
1057 else
1058 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1059 OOM_SCORE_ADJ_MAX;
David Rientjes723548b2010-10-26 14:21:25 -07001060err_sighand:
David Rientjesa63d83f2010-08-09 17:19:46 -07001061 unlock_task_sighand(task, &flags);
David Rientjesd19d5472010-10-26 14:21:26 -07001062err_task_lock:
1063 task_unlock(task);
David Rientjesa63d83f2010-08-09 17:19:46 -07001064 put_task_struct(task);
David Rientjes723548b2010-10-26 14:21:25 -07001065out:
1066 return err < 0 ? err : count;
David Rientjesa63d83f2010-08-09 17:19:46 -07001067}
1068
1069static const struct file_operations proc_oom_score_adj_operations = {
1070 .read = oom_score_adj_read,
1071 .write = oom_score_adj_write,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001072 .llseek = default_llseek,
David Rientjesa63d83f2010-08-09 17:19:46 -07001073};
1074
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075#ifdef CONFIG_AUDITSYSCALL
1076#define TMPBUFLEN 21
1077static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1078 size_t count, loff_t *ppos)
1079{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001080 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001081 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 ssize_t length;
1083 char tmpbuf[TMPBUFLEN];
1084
Eric W. Biederman99f89552006-06-26 00:25:55 -07001085 if (!task)
1086 return -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
Al Viro0c11b942008-01-10 04:20:52 -05001088 audit_get_loginuid(task));
Eric W. Biederman99f89552006-06-26 00:25:55 -07001089 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1091}
1092
1093static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1094 size_t count, loff_t *ppos)
1095{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001096 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097 char *page, *tmp;
1098 ssize_t length;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 uid_t loginuid;
1100
Paul E. McKenney7dc52152010-02-22 17:04:52 -08001101 rcu_read_lock();
1102 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1103 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 return -EPERM;
Paul E. McKenney7dc52152010-02-22 17:04:52 -08001105 }
1106 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107
Al Viroe0182902006-05-18 08:28:02 -04001108 if (count >= PAGE_SIZE)
1109 count = PAGE_SIZE - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110
1111 if (*ppos != 0) {
1112 /* No partial writes. */
1113 return -EINVAL;
1114 }
Mel Gormane12ba742007-10-16 01:25:52 -07001115 page = (char*)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 if (!page)
1117 return -ENOMEM;
1118 length = -EFAULT;
1119 if (copy_from_user(page, buf, count))
1120 goto out_free_page;
1121
Al Viroe0182902006-05-18 08:28:02 -04001122 page[count] = '\0';
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 loginuid = simple_strtoul(page, &tmp, 10);
1124 if (tmp == page) {
1125 length = -EINVAL;
1126 goto out_free_page;
1127
1128 }
Eric Paris0a300be2012-01-03 14:23:08 -05001129 length = audit_set_loginuid(loginuid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 if (likely(length == 0))
1131 length = count;
1132
1133out_free_page:
1134 free_page((unsigned long) page);
1135 return length;
1136}
1137
Arjan van de Ven00977a52007-02-12 00:55:34 -08001138static const struct file_operations proc_loginuid_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 .read = proc_loginuid_read,
1140 .write = proc_loginuid_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001141 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142};
Eric Paris1e0bd752008-03-13 08:15:31 -04001143
1144static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1145 size_t count, loff_t *ppos)
1146{
1147 struct inode * inode = file->f_path.dentry->d_inode;
1148 struct task_struct *task = get_proc_task(inode);
1149 ssize_t length;
1150 char tmpbuf[TMPBUFLEN];
1151
1152 if (!task)
1153 return -ESRCH;
1154 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1155 audit_get_sessionid(task));
1156 put_task_struct(task);
1157 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1158}
1159
1160static const struct file_operations proc_sessionid_operations = {
1161 .read = proc_sessionid_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001162 .llseek = generic_file_llseek,
Eric Paris1e0bd752008-03-13 08:15:31 -04001163};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164#endif
1165
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001166#ifdef CONFIG_FAULT_INJECTION
1167static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1168 size_t count, loff_t *ppos)
1169{
1170 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1171 char buffer[PROC_NUMBUF];
1172 size_t len;
1173 int make_it_fail;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001174
1175 if (!task)
1176 return -ESRCH;
1177 make_it_fail = task->make_it_fail;
1178 put_task_struct(task);
1179
1180 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001181
1182 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001183}
1184
1185static ssize_t proc_fault_inject_write(struct file * file,
1186 const char __user * buf, size_t count, loff_t *ppos)
1187{
1188 struct task_struct *task;
1189 char buffer[PROC_NUMBUF], *end;
1190 int make_it_fail;
1191
1192 if (!capable(CAP_SYS_RESOURCE))
1193 return -EPERM;
1194 memset(buffer, 0, sizeof(buffer));
1195 if (count > sizeof(buffer) - 1)
1196 count = sizeof(buffer) - 1;
1197 if (copy_from_user(buffer, buf, count))
1198 return -EFAULT;
Vincent Licba8aaf2009-09-22 16:45:38 -07001199 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1200 if (*end)
1201 return -EINVAL;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001202 task = get_proc_task(file->f_dentry->d_inode);
1203 if (!task)
1204 return -ESRCH;
1205 task->make_it_fail = make_it_fail;
1206 put_task_struct(task);
Vincent Licba8aaf2009-09-22 16:45:38 -07001207
1208 return count;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001209}
1210
Arjan van de Ven00977a52007-02-12 00:55:34 -08001211static const struct file_operations proc_fault_inject_operations = {
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001212 .read = proc_fault_inject_read,
1213 .write = proc_fault_inject_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001214 .llseek = generic_file_llseek,
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001215};
1216#endif
1217
Arjan van de Ven97455122008-01-25 21:08:34 +01001218
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001219#ifdef CONFIG_SCHED_DEBUG
1220/*
1221 * Print out various scheduling related per-task fields:
1222 */
1223static int sched_show(struct seq_file *m, void *v)
1224{
1225 struct inode *inode = m->private;
1226 struct task_struct *p;
1227
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001228 p = get_proc_task(inode);
1229 if (!p)
1230 return -ESRCH;
1231 proc_sched_show_task(p, m);
1232
1233 put_task_struct(p);
1234
1235 return 0;
1236}
1237
1238static ssize_t
1239sched_write(struct file *file, const char __user *buf,
1240 size_t count, loff_t *offset)
1241{
1242 struct inode *inode = file->f_path.dentry->d_inode;
1243 struct task_struct *p;
1244
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001245 p = get_proc_task(inode);
1246 if (!p)
1247 return -ESRCH;
1248 proc_sched_set_task(p);
1249
1250 put_task_struct(p);
1251
1252 return count;
1253}
1254
1255static int sched_open(struct inode *inode, struct file *filp)
1256{
Jovi Zhangc6a34052011-01-12 17:00:34 -08001257 return single_open(filp, sched_show, inode);
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001258}
1259
1260static const struct file_operations proc_pid_sched_operations = {
1261 .open = sched_open,
1262 .read = seq_read,
1263 .write = sched_write,
1264 .llseek = seq_lseek,
Alexey Dobriyan5ea473a2007-07-31 00:38:50 -07001265 .release = single_release,
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001266};
1267
1268#endif
1269
Mike Galbraith5091faa2010-11-30 14:18:03 +01001270#ifdef CONFIG_SCHED_AUTOGROUP
1271/*
1272 * Print out autogroup related information:
1273 */
1274static int sched_autogroup_show(struct seq_file *m, void *v)
1275{
1276 struct inode *inode = m->private;
1277 struct task_struct *p;
1278
1279 p = get_proc_task(inode);
1280 if (!p)
1281 return -ESRCH;
1282 proc_sched_autogroup_show_task(p, m);
1283
1284 put_task_struct(p);
1285
1286 return 0;
1287}
1288
1289static ssize_t
1290sched_autogroup_write(struct file *file, const char __user *buf,
1291 size_t count, loff_t *offset)
1292{
1293 struct inode *inode = file->f_path.dentry->d_inode;
1294 struct task_struct *p;
1295 char buffer[PROC_NUMBUF];
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001296 int nice;
Mike Galbraith5091faa2010-11-30 14:18:03 +01001297 int err;
1298
1299 memset(buffer, 0, sizeof(buffer));
1300 if (count > sizeof(buffer) - 1)
1301 count = sizeof(buffer) - 1;
1302 if (copy_from_user(buffer, buf, count))
1303 return -EFAULT;
1304
Alexey Dobriyan0a8cb8e2011-05-26 16:25:50 -07001305 err = kstrtoint(strstrip(buffer), 0, &nice);
1306 if (err < 0)
1307 return err;
Mike Galbraith5091faa2010-11-30 14:18:03 +01001308
1309 p = get_proc_task(inode);
1310 if (!p)
1311 return -ESRCH;
1312
Hiroshi Shimamoto2e5b5b32012-02-23 17:41:27 +09001313 err = proc_sched_autogroup_set_nice(p, nice);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001314 if (err)
1315 count = err;
1316
1317 put_task_struct(p);
1318
1319 return count;
1320}
1321
1322static int sched_autogroup_open(struct inode *inode, struct file *filp)
1323{
1324 int ret;
1325
1326 ret = single_open(filp, sched_autogroup_show, NULL);
1327 if (!ret) {
1328 struct seq_file *m = filp->private_data;
1329
1330 m->private = inode;
1331 }
1332 return ret;
1333}
1334
1335static const struct file_operations proc_pid_sched_autogroup_operations = {
1336 .open = sched_autogroup_open,
1337 .read = seq_read,
1338 .write = sched_autogroup_write,
1339 .llseek = seq_lseek,
1340 .release = single_release,
1341};
1342
1343#endif /* CONFIG_SCHED_AUTOGROUP */
1344
john stultz4614a696b2009-12-14 18:00:05 -08001345static ssize_t comm_write(struct file *file, const char __user *buf,
1346 size_t count, loff_t *offset)
1347{
1348 struct inode *inode = file->f_path.dentry->d_inode;
1349 struct task_struct *p;
1350 char buffer[TASK_COMM_LEN];
1351
1352 memset(buffer, 0, sizeof(buffer));
1353 if (count > sizeof(buffer) - 1)
1354 count = sizeof(buffer) - 1;
1355 if (copy_from_user(buffer, buf, count))
1356 return -EFAULT;
1357
1358 p = get_proc_task(inode);
1359 if (!p)
1360 return -ESRCH;
1361
1362 if (same_thread_group(current, p))
1363 set_task_comm(p, buffer);
1364 else
1365 count = -EINVAL;
1366
1367 put_task_struct(p);
1368
1369 return count;
1370}
1371
1372static int comm_show(struct seq_file *m, void *v)
1373{
1374 struct inode *inode = m->private;
1375 struct task_struct *p;
1376
1377 p = get_proc_task(inode);
1378 if (!p)
1379 return -ESRCH;
1380
1381 task_lock(p);
1382 seq_printf(m, "%s\n", p->comm);
1383 task_unlock(p);
1384
1385 put_task_struct(p);
1386
1387 return 0;
1388}
1389
1390static int comm_open(struct inode *inode, struct file *filp)
1391{
Jovi Zhangc6a34052011-01-12 17:00:34 -08001392 return single_open(filp, comm_show, inode);
john stultz4614a696b2009-12-14 18:00:05 -08001393}
1394
1395static const struct file_operations proc_pid_set_comm_operations = {
1396 .open = comm_open,
1397 .read = seq_read,
1398 .write = comm_write,
1399 .llseek = seq_lseek,
1400 .release = single_release,
1401};
1402
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -08001403static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
Matt Helsley925d1c42008-04-29 01:01:36 -07001404{
1405 struct task_struct *task;
1406 struct mm_struct *mm;
1407 struct file *exe_file;
1408
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -08001409 task = get_proc_task(dentry->d_inode);
Matt Helsley925d1c42008-04-29 01:01:36 -07001410 if (!task)
1411 return -ENOENT;
1412 mm = get_task_mm(task);
1413 put_task_struct(task);
1414 if (!mm)
1415 return -ENOENT;
1416 exe_file = get_mm_exe_file(mm);
1417 mmput(mm);
1418 if (exe_file) {
1419 *exe_path = exe_file->f_path;
1420 path_get(&exe_file->f_path);
1421 fput(exe_file);
1422 return 0;
1423 } else
1424 return -ENOENT;
1425}
1426
Al Viro008b1502005-08-20 00:17:39 +01001427static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428{
1429 struct inode *inode = dentry->d_inode;
1430 int error = -EACCES;
1431
1432 /* We don't need a base pointer in the /proc filesystem */
Jan Blunck1d957f92008-02-14 19:34:35 -08001433 path_put(&nd->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434
Eric W. Biederman778c1142006-06-26 00:25:58 -07001435 /* Are we allowed to snoop on the tasks file descriptors? */
1436 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -08001439 error = PROC_I(inode)->op.proc_get_link(dentry, &nd->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440out:
Al Viro008b1502005-08-20 00:17:39 +01001441 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442}
1443
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001444static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445{
Mel Gormane12ba742007-10-16 01:25:52 -07001446 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001447 char *pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 int len;
1449
1450 if (!tmp)
1451 return -ENOMEM;
Akinobu Mita0c28f282007-05-08 00:31:41 -07001452
Eric W. Biederman7b2a69b2010-12-05 15:51:21 -08001453 pathname = d_path(path, tmp, PAGE_SIZE);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001454 len = PTR_ERR(pathname);
1455 if (IS_ERR(pathname))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456 goto out;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001457 len = tmp + PAGE_SIZE - 1 - pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458
1459 if (len > buflen)
1460 len = buflen;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001461 if (copy_to_user(buffer, pathname, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 len = -EFAULT;
1463 out:
1464 free_page((unsigned long)tmp);
1465 return len;
1466}
1467
1468static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1469{
1470 int error = -EACCES;
1471 struct inode *inode = dentry->d_inode;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001472 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473
Eric W. Biederman778c1142006-06-26 00:25:58 -07001474 /* Are we allowed to snoop on the tasks file descriptors? */
1475 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -08001478 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 if (error)
1480 goto out;
1481
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001482 error = do_proc_readlink(&path, buffer, buflen);
1483 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 return error;
1486}
1487
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08001488static const struct inode_operations proc_pid_link_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489 .readlink = proc_pid_readlink,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001490 .follow_link = proc_pid_follow_link,
1491 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492};
1493
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001494
1495/* building an inode */
1496
1497static int task_dumpable(struct task_struct *task)
1498{
1499 int dumpable = 0;
1500 struct mm_struct *mm;
1501
1502 task_lock(task);
1503 mm = task->mm;
1504 if (mm)
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001505 dumpable = get_dumpable(mm);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001506 task_unlock(task);
1507 if(dumpable == 1)
1508 return 1;
1509 return 0;
1510}
1511
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001512struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001513{
1514 struct inode * inode;
1515 struct proc_inode *ei;
David Howellsc69e8d92008-11-14 10:39:19 +11001516 const struct cred *cred;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001517
1518 /* We need a new inode */
1519
1520 inode = new_inode(sb);
1521 if (!inode)
1522 goto out;
1523
1524 /* Common stuff */
1525 ei = PROC_I(inode);
Christoph Hellwig85fe4022010-10-23 11:19:54 -04001526 inode->i_ino = get_next_ino();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001527 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001528 inode->i_op = &proc_def_inode_operations;
1529
1530 /*
1531 * grab the reference to task.
1532 */
Oleg Nesterov1a657f782006-10-02 02:18:59 -07001533 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001534 if (!ei->pid)
1535 goto out_unlock;
1536
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001537 if (task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001538 rcu_read_lock();
1539 cred = __task_cred(task);
1540 inode->i_uid = cred->euid;
1541 inode->i_gid = cred->egid;
1542 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001543 }
1544 security_task_to_inode(task, inode);
1545
1546out:
1547 return inode;
1548
1549out_unlock:
1550 iput(inode);
1551 return NULL;
1552}
1553
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001554int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001555{
1556 struct inode *inode = dentry->d_inode;
1557 struct task_struct *task;
David Howellsc69e8d92008-11-14 10:39:19 +11001558 const struct cred *cred;
Vasiliy Kulikov04996802012-01-10 15:11:31 -08001559 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
David Howellsc69e8d92008-11-14 10:39:19 +11001560
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001561 generic_fillattr(inode, stat);
1562
1563 rcu_read_lock();
Eric W. Biedermandcb0f222012-02-09 08:48:21 -08001564 stat->uid = GLOBAL_ROOT_UID;
1565 stat->gid = GLOBAL_ROOT_GID;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001566 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1567 if (task) {
Vasiliy Kulikov04996802012-01-10 15:11:31 -08001568 if (!has_pid_permissions(pid, task, 2)) {
1569 rcu_read_unlock();
1570 /*
1571 * This doesn't prevent learning whether PID exists,
1572 * it only makes getattr() consistent with readdir().
1573 */
1574 return -ENOENT;
1575 }
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001576 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1577 task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001578 cred = __task_cred(task);
1579 stat->uid = cred->euid;
1580 stat->gid = cred->egid;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001581 }
1582 }
1583 rcu_read_unlock();
1584 return 0;
1585}
1586
1587/* dentry stuff */
1588
1589/*
1590 * Exceptional case: normally we are not allowed to unhash a busy
1591 * directory. In this case, however, we can do it - no aliasing problems
1592 * due to the way we treat inodes.
1593 *
1594 * Rewrite the inode's ownerships here because the owning task may have
1595 * performed a setuid(), etc.
1596 *
1597 * Before the /proc/pid/status file was created the only way to read
1598 * the effective uid of a /process was to stat /proc/pid. Reading
1599 * /proc/pid/status is slow enough that procps and other packages
1600 * kept stating /proc/pid. To keep the rules in /proc simple I have
1601 * made this apply to all per process world readable and executable
1602 * directories.
1603 */
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001604int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001605{
Nick Piggin34286d62011-01-07 17:49:57 +11001606 struct inode *inode;
1607 struct task_struct *task;
David Howellsc69e8d92008-11-14 10:39:19 +11001608 const struct cred *cred;
1609
Nick Piggin34286d62011-01-07 17:49:57 +11001610 if (nd && nd->flags & LOOKUP_RCU)
1611 return -ECHILD;
1612
1613 inode = dentry->d_inode;
1614 task = get_proc_task(inode);
1615
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001616 if (task) {
1617 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1618 task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001619 rcu_read_lock();
1620 cred = __task_cred(task);
1621 inode->i_uid = cred->euid;
1622 inode->i_gid = cred->egid;
1623 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001624 } else {
Eric W. Biedermandcb0f222012-02-09 08:48:21 -08001625 inode->i_uid = GLOBAL_ROOT_UID;
1626 inode->i_gid = GLOBAL_ROOT_GID;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001627 }
1628 inode->i_mode &= ~(S_ISUID | S_ISGID);
1629 security_task_to_inode(task, inode);
1630 put_task_struct(task);
1631 return 1;
1632 }
1633 d_drop(dentry);
1634 return 0;
1635}
1636
Nick Pigginfe15ce42011-01-07 17:49:23 +11001637static int pid_delete_dentry(const struct dentry * dentry)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001638{
1639 /* Is the task we represent dead?
1640 * If so, then don't put the dentry on the lru list,
1641 * kill it immediately.
1642 */
1643 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1644}
1645
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001646const struct dentry_operations pid_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001647{
1648 .d_revalidate = pid_revalidate,
1649 .d_delete = pid_delete_dentry,
1650};
1651
1652/* Lookups */
1653
Eric W. Biederman1c0d04c2006-10-02 02:18:57 -07001654/*
1655 * Fill a directory entry.
1656 *
1657 * If possible create the dcache entry and derive our inode number and
1658 * file type from dcache entry.
1659 *
1660 * Since all of the proc inode numbers are dynamically generated, the inode
1661 * numbers do not exist until the inode is cache. This means creating the
1662 * the dcache entry in readdir is necessary to keep the inode numbers
1663 * reported by readdir in sync with the inode numbers reported
1664 * by stat.
1665 */
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08001666int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1667 const char *name, int len,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001668 instantiate_t instantiate, struct task_struct *task, const void *ptr)
Eric W. Biederman61a28782006-10-02 02:18:49 -07001669{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001670 struct dentry *child, *dir = filp->f_path.dentry;
Eric W. Biederman61a28782006-10-02 02:18:49 -07001671 struct inode *inode;
1672 struct qstr qname;
1673 ino_t ino = 0;
1674 unsigned type = DT_UNKNOWN;
1675
1676 qname.name = name;
1677 qname.len = len;
1678 qname.hash = full_name_hash(name, len);
1679
1680 child = d_lookup(dir, &qname);
1681 if (!child) {
1682 struct dentry *new;
1683 new = d_alloc(dir, &qname);
1684 if (new) {
1685 child = instantiate(dir->d_inode, new, task, ptr);
1686 if (child)
1687 dput(new);
1688 else
1689 child = new;
1690 }
1691 }
1692 if (!child || IS_ERR(child) || !child->d_inode)
1693 goto end_instantiate;
1694 inode = child->d_inode;
1695 if (inode) {
1696 ino = inode->i_ino;
1697 type = inode->i_mode >> 12;
1698 }
1699 dput(child);
1700end_instantiate:
1701 if (!ino)
1702 ino = find_inode_number(dir, &qname);
1703 if (!ino)
1704 ino = 1;
1705 return filldir(dirent, name, len, filp->f_pos, ino, type);
1706}
1707
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001708static unsigned name_to_int(struct dentry *dentry)
1709{
1710 const char *name = dentry->d_name.name;
1711 int len = dentry->d_name.len;
1712 unsigned n = 0;
1713
1714 if (len > 1 && *name == '0')
1715 goto out;
1716 while (len-- > 0) {
1717 unsigned c = *name++ - '0';
1718 if (c > 9)
1719 goto out;
1720 if (n >= (~0U-9)/10)
1721 goto out;
1722 n *= 10;
1723 n += c;
1724 }
1725 return n;
1726out:
1727 return ~0U;
1728}
1729
Miklos Szeredi27932742007-05-08 00:26:17 -07001730#define PROC_FDINFO_MAX 64
1731
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001732static int proc_fd_info(struct inode *inode, struct path *path, char *info)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001733{
Linus Torvalds5e442a42011-11-09 18:16:00 -05001734 struct task_struct *task = get_proc_task(inode);
1735 struct files_struct *files = NULL;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001736 struct file *file;
1737 int fd = proc_fd(inode);
1738
Linus Torvalds5e442a42011-11-09 18:16:00 -05001739 if (task) {
1740 files = get_files_struct(task);
1741 put_task_struct(task);
1742 }
1743 if (files) {
1744 /*
1745 * We are not taking a ref to the file structure, so we must
1746 * hold ->file_lock.
1747 */
1748 spin_lock(&files->file_lock);
1749 file = fcheck_files(files, fd);
1750 if (file) {
1751 unsigned int f_flags;
1752 struct fdtable *fdt;
Linus Torvalds1117f722011-08-06 11:51:33 -07001753
Linus Torvalds5e442a42011-11-09 18:16:00 -05001754 fdt = files_fdtable(files);
1755 f_flags = file->f_flags & ~O_CLOEXEC;
David Howells1dce27c2012-02-16 17:49:42 +00001756 if (close_on_exec(fd, fdt))
Linus Torvalds5e442a42011-11-09 18:16:00 -05001757 f_flags |= O_CLOEXEC;
Linus Torvalds1117f722011-08-06 11:51:33 -07001758
Linus Torvalds5e442a42011-11-09 18:16:00 -05001759 if (path) {
1760 *path = file->f_path;
1761 path_get(&file->f_path);
1762 }
1763 if (info)
1764 snprintf(info, PROC_FDINFO_MAX,
1765 "pos:\t%lli\n"
1766 "flags:\t0%o\n",
1767 (long long) file->f_pos,
1768 f_flags);
1769 spin_unlock(&files->file_lock);
1770 put_files_struct(files);
1771 return 0;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001772 }
Linus Torvalds5e442a42011-11-09 18:16:00 -05001773 spin_unlock(&files->file_lock);
1774 put_files_struct(files);
1775 }
1776 return -ENOENT;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001777}
1778
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -08001779static int proc_fd_link(struct dentry *dentry, struct path *path)
Miklos Szeredi27932742007-05-08 00:26:17 -07001780{
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -08001781 return proc_fd_info(dentry->d_inode, path, NULL);
Miklos Szeredi27932742007-05-08 00:26:17 -07001782}
1783
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001784static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1785{
Nick Piggin34286d62011-01-07 17:49:57 +11001786 struct inode *inode;
1787 struct task_struct *task;
1788 int fd;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001789 struct files_struct *files;
David Howellsc69e8d92008-11-14 10:39:19 +11001790 const struct cred *cred;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001791
Nick Piggin34286d62011-01-07 17:49:57 +11001792 if (nd && nd->flags & LOOKUP_RCU)
1793 return -ECHILD;
1794
1795 inode = dentry->d_inode;
1796 task = get_proc_task(inode);
1797 fd = proc_fd(inode);
1798
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001799 if (task) {
1800 files = get_files_struct(task);
1801 if (files) {
Linus Torvalds30a08bf2012-05-18 11:32:15 -07001802 struct file *file;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001803 rcu_read_lock();
Linus Torvalds30a08bf2012-05-18 11:32:15 -07001804 file = fcheck_files(files, fd);
1805 if (file) {
Linus Torvalds06401132012-06-04 11:00:45 -07001806 unsigned f_mode = file->f_mode;
Linus Torvalds30a08bf2012-05-18 11:32:15 -07001807
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001808 rcu_read_unlock();
1809 put_files_struct(files);
Linus Torvalds30a08bf2012-05-18 11:32:15 -07001810
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001811 if (task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001812 rcu_read_lock();
1813 cred = __task_cred(task);
1814 inode->i_uid = cred->euid;
1815 inode->i_gid = cred->egid;
1816 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001817 } else {
Eric W. Biedermandcb0f222012-02-09 08:48:21 -08001818 inode->i_uid = GLOBAL_ROOT_UID;
1819 inode->i_gid = GLOBAL_ROOT_GID;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001820 }
Linus Torvalds30a08bf2012-05-18 11:32:15 -07001821
Linus Torvalds06401132012-06-04 11:00:45 -07001822 if (S_ISLNK(inode->i_mode)) {
1823 unsigned i_mode = S_IFLNK;
1824 if (f_mode & FMODE_READ)
1825 i_mode |= S_IRUSR | S_IXUSR;
1826 if (f_mode & FMODE_WRITE)
1827 i_mode |= S_IWUSR | S_IXUSR;
1828 inode->i_mode = i_mode;
1829 }
Linus Torvalds30a08bf2012-05-18 11:32:15 -07001830
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001831 security_task_to_inode(task, inode);
1832 put_task_struct(task);
1833 return 1;
1834 }
1835 rcu_read_unlock();
1836 put_files_struct(files);
1837 }
1838 put_task_struct(task);
1839 }
1840 d_drop(dentry);
1841 return 0;
1842}
1843
Al Virod72f71e2009-02-20 05:58:47 +00001844static const struct dentry_operations tid_fd_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001845{
1846 .d_revalidate = tid_fd_revalidate,
1847 .d_delete = pid_delete_dentry,
1848};
1849
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001850static struct dentry *proc_fd_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001851 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001852{
Alexey Dobriyanaf5e6172012-05-31 16:26:18 -07001853 unsigned fd = (unsigned long)ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001854 struct inode *inode;
1855 struct proc_inode *ei;
1856 struct dentry *error = ERR_PTR(-ENOENT);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001857
Eric W. Biederman61a28782006-10-02 02:18:49 -07001858 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001859 if (!inode)
1860 goto out;
1861 ei = PROC_I(inode);
1862 ei->fd = fd;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001863
Linus Torvalds06401132012-06-04 11:00:45 -07001864 inode->i_mode = S_IFLNK;
Linus Torvalds5e442a42011-11-09 18:16:00 -05001865 inode->i_op = &proc_pid_link_inode_operations;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001866 inode->i_size = 64;
1867 ei->op.proc_get_link = proc_fd_link;
Nick Pigginfb045ad2011-01-07 17:49:55 +11001868 d_set_d_op(dentry, &tid_fd_dentry_operations);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001869 d_add(dentry, inode);
1870 /* Close the race of the process dying before we return the dentry */
1871 if (tid_fd_revalidate(dentry, NULL))
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001872 error = NULL;
1873
1874 out:
1875 return error;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001876}
1877
Miklos Szeredi27932742007-05-08 00:26:17 -07001878static struct dentry *proc_lookupfd_common(struct inode *dir,
1879 struct dentry *dentry,
1880 instantiate_t instantiate)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001881{
1882 struct task_struct *task = get_proc_task(dir);
1883 unsigned fd = name_to_int(dentry);
1884 struct dentry *result = ERR_PTR(-ENOENT);
1885
1886 if (!task)
1887 goto out_no_task;
1888 if (fd == ~0U)
1889 goto out;
1890
Alexey Dobriyanaf5e6172012-05-31 16:26:18 -07001891 result = instantiate(dir, dentry, task, (void *)(unsigned long)fd);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001892out:
1893 put_task_struct(task);
1894out_no_task:
1895 return result;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001896}
1897
Miklos Szeredi27932742007-05-08 00:26:17 -07001898static int proc_readfd_common(struct file * filp, void * dirent,
1899 filldir_t filldir, instantiate_t instantiate)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001901 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman56347082006-06-26 00:25:40 -07001902 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001903 struct task_struct *p = get_proc_task(inode);
Pavel Emelyanov457c2512007-10-18 23:40:43 -07001904 unsigned int fd, ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906 struct files_struct * files;
1907
1908 retval = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001909 if (!p)
1910 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912
1913 fd = filp->f_pos;
1914 switch (fd) {
1915 case 0:
1916 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
Linus Torvalds5e442a42011-11-09 18:16:00 -05001917 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001918 filp->f_pos++;
1919 case 1:
Eric W. Biederman56347082006-06-26 00:25:40 -07001920 ino = parent_ino(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
Linus Torvalds5e442a42011-11-09 18:16:00 -05001922 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 filp->f_pos++;
1924 default:
1925 files = get_files_struct(p);
1926 if (!files)
Linus Torvalds5e442a42011-11-09 18:16:00 -05001927 goto out;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001928 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929 for (fd = filp->f_pos-2;
Al Viro9b4f5262008-04-22 01:32:44 -04001930 fd < files_fdtable(files)->max_fds;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931 fd++, filp->f_pos++) {
Miklos Szeredi27932742007-05-08 00:26:17 -07001932 char name[PROC_NUMBUF];
1933 int len;
Alexey Dobriyanf05ed3f2012-05-31 16:26:18 -07001934 int rv;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935
1936 if (!fcheck_files(files, fd))
1937 continue;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001938 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939
Miklos Szeredi27932742007-05-08 00:26:17 -07001940 len = snprintf(name, sizeof(name), "%d", fd);
Alexey Dobriyanf05ed3f2012-05-31 16:26:18 -07001941 rv = proc_fill_cache(filp, dirent, filldir,
1942 name, len, instantiate, p,
Alexey Dobriyanaf5e6172012-05-31 16:26:18 -07001943 (void *)(unsigned long)fd);
Alexey Dobriyanf05ed3f2012-05-31 16:26:18 -07001944 if (rv < 0)
1945 goto out_fd_loop;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001946 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07001948 rcu_read_unlock();
Alexey Dobriyanf05ed3f2012-05-31 16:26:18 -07001949out_fd_loop:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950 put_files_struct(files);
1951 }
1952out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07001953 put_task_struct(p);
1954out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955 return retval;
1956}
1957
Miklos Szeredi27932742007-05-08 00:26:17 -07001958static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1959 struct nameidata *nd)
1960{
1961 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1962}
1963
1964static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1965{
1966 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1967}
1968
1969static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1970 size_t len, loff_t *ppos)
1971{
1972 char tmp[PROC_FDINFO_MAX];
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001973 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
Miklos Szeredi27932742007-05-08 00:26:17 -07001974 if (!err)
1975 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1976 return err;
1977}
1978
1979static const struct file_operations proc_fdinfo_file_operations = {
Arnd Bergmann87df8422010-03-17 23:06:02 +01001980 .open = nonseekable_open,
Miklos Szeredi27932742007-05-08 00:26:17 -07001981 .read = proc_fdinfo_read,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001982 .llseek = no_llseek,
Miklos Szeredi27932742007-05-08 00:26:17 -07001983};
1984
Arjan van de Ven00977a52007-02-12 00:55:34 -08001985static const struct file_operations proc_fd_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986 .read = generic_read_dir,
1987 .readdir = proc_readfd,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001988 .llseek = default_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989};
1990
Pavel Emelyanov640708a2012-01-10 15:11:23 -08001991#ifdef CONFIG_CHECKPOINT_RESTORE
1992
1993/*
1994 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1995 * which represent vma start and end addresses.
1996 */
1997static int dname_to_vma_addr(struct dentry *dentry,
1998 unsigned long *start, unsigned long *end)
1999{
2000 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
2001 return -EINVAL;
2002
2003 return 0;
2004}
2005
2006static int map_files_d_revalidate(struct dentry *dentry, struct nameidata *nd)
2007{
2008 unsigned long vm_start, vm_end;
2009 bool exact_vma_exists = false;
2010 struct mm_struct *mm = NULL;
2011 struct task_struct *task;
2012 const struct cred *cred;
2013 struct inode *inode;
2014 int status = 0;
2015
2016 if (nd && nd->flags & LOOKUP_RCU)
2017 return -ECHILD;
2018
2019 if (!capable(CAP_SYS_ADMIN)) {
2020 status = -EACCES;
2021 goto out_notask;
2022 }
2023
2024 inode = dentry->d_inode;
2025 task = get_proc_task(inode);
2026 if (!task)
2027 goto out_notask;
2028
Cong Wang2344bec2012-05-31 16:26:18 -07002029 mm = mm_access(task, PTRACE_MODE_READ);
2030 if (IS_ERR_OR_NULL(mm))
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002031 goto out;
2032
2033 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2034 down_read(&mm->mmap_sem);
2035 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
2036 up_read(&mm->mmap_sem);
2037 }
2038
2039 mmput(mm);
2040
2041 if (exact_vma_exists) {
2042 if (task_dumpable(task)) {
2043 rcu_read_lock();
2044 cred = __task_cred(task);
2045 inode->i_uid = cred->euid;
2046 inode->i_gid = cred->egid;
2047 rcu_read_unlock();
2048 } else {
Eric W. Biedermandcb0f222012-02-09 08:48:21 -08002049 inode->i_uid = GLOBAL_ROOT_UID;
2050 inode->i_gid = GLOBAL_ROOT_GID;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002051 }
2052 security_task_to_inode(task, inode);
2053 status = 1;
2054 }
2055
2056out:
2057 put_task_struct(task);
2058
2059out_notask:
2060 if (status <= 0)
2061 d_drop(dentry);
2062
2063 return status;
2064}
2065
2066static const struct dentry_operations tid_map_files_dentry_operations = {
2067 .d_revalidate = map_files_d_revalidate,
2068 .d_delete = pid_delete_dentry,
2069};
2070
2071static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
2072{
2073 unsigned long vm_start, vm_end;
2074 struct vm_area_struct *vma;
2075 struct task_struct *task;
2076 struct mm_struct *mm;
2077 int rc;
2078
2079 rc = -ENOENT;
2080 task = get_proc_task(dentry->d_inode);
2081 if (!task)
2082 goto out;
2083
2084 mm = get_task_mm(task);
2085 put_task_struct(task);
2086 if (!mm)
2087 goto out;
2088
2089 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2090 if (rc)
2091 goto out_mmput;
2092
2093 down_read(&mm->mmap_sem);
2094 vma = find_exact_vma(mm, vm_start, vm_end);
2095 if (vma && vma->vm_file) {
2096 *path = vma->vm_file->f_path;
2097 path_get(path);
2098 rc = 0;
2099 }
2100 up_read(&mm->mmap_sem);
2101
2102out_mmput:
2103 mmput(mm);
2104out:
2105 return rc;
2106}
2107
2108struct map_files_info {
2109 struct file *file;
2110 unsigned long len;
2111 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2112};
2113
2114static struct dentry *
2115proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2116 struct task_struct *task, const void *ptr)
2117{
2118 const struct file *file = ptr;
2119 struct proc_inode *ei;
2120 struct inode *inode;
2121
2122 if (!file)
2123 return ERR_PTR(-ENOENT);
2124
2125 inode = proc_pid_make_inode(dir->i_sb, task);
2126 if (!inode)
2127 return ERR_PTR(-ENOENT);
2128
2129 ei = PROC_I(inode);
2130 ei->op.proc_get_link = proc_map_files_get_link;
2131
2132 inode->i_op = &proc_pid_link_inode_operations;
2133 inode->i_size = 64;
2134 inode->i_mode = S_IFLNK;
2135
2136 if (file->f_mode & FMODE_READ)
2137 inode->i_mode |= S_IRUSR;
2138 if (file->f_mode & FMODE_WRITE)
2139 inode->i_mode |= S_IWUSR;
2140
2141 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2142 d_add(dentry, inode);
2143
2144 return NULL;
2145}
2146
2147static struct dentry *proc_map_files_lookup(struct inode *dir,
2148 struct dentry *dentry, struct nameidata *nd)
2149{
2150 unsigned long vm_start, vm_end;
2151 struct vm_area_struct *vma;
2152 struct task_struct *task;
2153 struct dentry *result;
2154 struct mm_struct *mm;
2155
2156 result = ERR_PTR(-EACCES);
2157 if (!capable(CAP_SYS_ADMIN))
2158 goto out;
2159
2160 result = ERR_PTR(-ENOENT);
2161 task = get_proc_task(dir);
2162 if (!task)
2163 goto out;
2164
2165 result = ERR_PTR(-EACCES);
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07002166 if (!ptrace_may_access(task, PTRACE_MODE_READ))
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002167 goto out_put_task;
2168
2169 result = ERR_PTR(-ENOENT);
2170 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07002171 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002172
2173 mm = get_task_mm(task);
2174 if (!mm)
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07002175 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002176
2177 down_read(&mm->mmap_sem);
2178 vma = find_exact_vma(mm, vm_start, vm_end);
2179 if (!vma)
2180 goto out_no_vma;
2181
2182 result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
2183
2184out_no_vma:
2185 up_read(&mm->mmap_sem);
2186 mmput(mm);
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002187out_put_task:
2188 put_task_struct(task);
2189out:
2190 return result;
2191}
2192
2193static const struct inode_operations proc_map_files_inode_operations = {
2194 .lookup = proc_map_files_lookup,
2195 .permission = proc_fd_permission,
2196 .setattr = proc_setattr,
2197};
2198
2199static int
2200proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2201{
2202 struct dentry *dentry = filp->f_path.dentry;
2203 struct inode *inode = dentry->d_inode;
2204 struct vm_area_struct *vma;
2205 struct task_struct *task;
2206 struct mm_struct *mm;
2207 ino_t ino;
2208 int ret;
2209
2210 ret = -EACCES;
2211 if (!capable(CAP_SYS_ADMIN))
2212 goto out;
2213
2214 ret = -ENOENT;
2215 task = get_proc_task(inode);
2216 if (!task)
2217 goto out;
2218
2219 ret = -EACCES;
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07002220 if (!ptrace_may_access(task, PTRACE_MODE_READ))
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002221 goto out_put_task;
2222
2223 ret = 0;
2224 switch (filp->f_pos) {
2225 case 0:
2226 ino = inode->i_ino;
2227 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07002228 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002229 filp->f_pos++;
2230 case 1:
2231 ino = parent_ino(dentry);
2232 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07002233 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002234 filp->f_pos++;
2235 default:
2236 {
2237 unsigned long nr_files, pos, i;
2238 struct flex_array *fa = NULL;
2239 struct map_files_info info;
2240 struct map_files_info *p;
2241
2242 mm = get_task_mm(task);
2243 if (!mm)
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07002244 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002245 down_read(&mm->mmap_sem);
2246
2247 nr_files = 0;
2248
2249 /*
2250 * We need two passes here:
2251 *
2252 * 1) Collect vmas of mapped files with mmap_sem taken
2253 * 2) Release mmap_sem and instantiate entries
2254 *
2255 * otherwise we get lockdep complained, since filldir()
2256 * routine might require mmap_sem taken in might_fault().
2257 */
2258
2259 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2260 if (vma->vm_file && ++pos > filp->f_pos)
2261 nr_files++;
2262 }
2263
2264 if (nr_files) {
2265 fa = flex_array_alloc(sizeof(info), nr_files,
2266 GFP_KERNEL);
2267 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2268 GFP_KERNEL)) {
2269 ret = -ENOMEM;
2270 if (fa)
2271 flex_array_free(fa);
2272 up_read(&mm->mmap_sem);
2273 mmput(mm);
Cyrill Gorcunoveb94cd92012-05-17 17:03:25 -07002274 goto out_put_task;
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002275 }
2276 for (i = 0, vma = mm->mmap, pos = 2; vma;
2277 vma = vma->vm_next) {
2278 if (!vma->vm_file)
2279 continue;
2280 if (++pos <= filp->f_pos)
2281 continue;
2282
2283 get_file(vma->vm_file);
2284 info.file = vma->vm_file;
2285 info.len = snprintf(info.name,
2286 sizeof(info.name), "%lx-%lx",
2287 vma->vm_start, vma->vm_end);
2288 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2289 BUG();
2290 }
2291 }
2292 up_read(&mm->mmap_sem);
2293
2294 for (i = 0; i < nr_files; i++) {
2295 p = flex_array_get(fa, i);
2296 ret = proc_fill_cache(filp, dirent, filldir,
2297 p->name, p->len,
2298 proc_map_files_instantiate,
2299 task, p->file);
2300 if (ret)
2301 break;
2302 filp->f_pos++;
2303 fput(p->file);
2304 }
2305 for (; i < nr_files; i++) {
2306 /*
2307 * In case of error don't forget
2308 * to put rest of file refs.
2309 */
2310 p = flex_array_get(fa, i);
2311 fput(p->file);
2312 }
2313 if (fa)
2314 flex_array_free(fa);
2315 mmput(mm);
2316 }
2317 }
2318
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002319out_put_task:
2320 put_task_struct(task);
2321out:
2322 return ret;
2323}
2324
2325static const struct file_operations proc_map_files_operations = {
2326 .read = generic_read_dir,
2327 .readdir = proc_map_files_readdir,
2328 .llseek = default_llseek,
2329};
2330
2331#endif /* CONFIG_CHECKPOINT_RESTORE */
2332
Linus Torvalds1da177e2005-04-16 15:20:36 -07002333/*
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002334 * /proc/pid/fd needs a special permission handler so that a process can still
2335 * access /proc/self/fd after it has executed a setuid().
2336 */
Al Viro10556cb2011-06-20 19:28:19 -04002337static int proc_fd_permission(struct inode *inode, int mask)
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002338{
Al Viro2830ba72011-06-20 19:16:29 -04002339 int rv = generic_permission(inode, mask);
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002340 if (rv == 0)
2341 return 0;
2342 if (task_pid(current) == proc_pid(inode))
2343 rv = 0;
2344 return rv;
2345}
2346
2347/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348 * proc directories can do almost nothing..
2349 */
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002350static const struct inode_operations proc_fd_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002351 .lookup = proc_lookupfd,
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002352 .permission = proc_fd_permission,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07002353 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002354};
2355
Miklos Szeredi27932742007-05-08 00:26:17 -07002356static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2357 struct dentry *dentry, struct task_struct *task, const void *ptr)
2358{
Alexey Dobriyanaf5e6172012-05-31 16:26:18 -07002359 unsigned fd = (unsigned long)ptr;
Miklos Szeredi27932742007-05-08 00:26:17 -07002360 struct inode *inode;
2361 struct proc_inode *ei;
2362 struct dentry *error = ERR_PTR(-ENOENT);
2363
2364 inode = proc_pid_make_inode(dir->i_sb, task);
2365 if (!inode)
2366 goto out;
2367 ei = PROC_I(inode);
2368 ei->fd = fd;
2369 inode->i_mode = S_IFREG | S_IRUSR;
2370 inode->i_fop = &proc_fdinfo_file_operations;
Nick Pigginfb045ad2011-01-07 17:49:55 +11002371 d_set_d_op(dentry, &tid_fd_dentry_operations);
Miklos Szeredi27932742007-05-08 00:26:17 -07002372 d_add(dentry, inode);
2373 /* Close the race of the process dying before we return the dentry */
2374 if (tid_fd_revalidate(dentry, NULL))
2375 error = NULL;
2376
2377 out:
2378 return error;
2379}
2380
2381static struct dentry *proc_lookupfdinfo(struct inode *dir,
2382 struct dentry *dentry,
2383 struct nameidata *nd)
2384{
2385 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2386}
2387
2388static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2389{
2390 return proc_readfd_common(filp, dirent, filldir,
2391 proc_fdinfo_instantiate);
2392}
2393
2394static const struct file_operations proc_fdinfo_operations = {
2395 .read = generic_read_dir,
2396 .readdir = proc_readfdinfo,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002397 .llseek = default_llseek,
Miklos Szeredi27932742007-05-08 00:26:17 -07002398};
2399
2400/*
2401 * proc directories can do almost nothing..
2402 */
2403static const struct inode_operations proc_fdinfo_inode_operations = {
2404 .lookup = proc_lookupfdinfo,
2405 .setattr = proc_setattr,
2406};
2407
2408
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002409static struct dentry *proc_pident_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002410 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002411{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002412 const struct pid_entry *p = ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002413 struct inode *inode;
2414 struct proc_inode *ei;
KOSAKI Motohirobd6daba2009-05-28 14:34:21 -07002415 struct dentry *error = ERR_PTR(-ENOENT);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002416
Eric W. Biederman61a28782006-10-02 02:18:49 -07002417 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002418 if (!inode)
2419 goto out;
2420
2421 ei = PROC_I(inode);
2422 inode->i_mode = p->mode;
2423 if (S_ISDIR(inode->i_mode))
Miklos Szeredibfe86842011-10-28 14:13:29 +02002424 set_nlink(inode, 2); /* Use getattr to fix if necessary */
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002425 if (p->iop)
2426 inode->i_op = p->iop;
2427 if (p->fop)
2428 inode->i_fop = p->fop;
2429 ei->op = p->op;
Nick Pigginfb045ad2011-01-07 17:49:55 +11002430 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002431 d_add(dentry, inode);
2432 /* Close the race of the process dying before we return the dentry */
2433 if (pid_revalidate(dentry, NULL))
2434 error = NULL;
2435out:
2436 return error;
2437}
2438
Linus Torvalds1da177e2005-04-16 15:20:36 -07002439static struct dentry *proc_pident_lookup(struct inode *dir,
2440 struct dentry *dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002441 const struct pid_entry *ents,
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002442 unsigned int nents)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002443{
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002444 struct dentry *error;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002445 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002446 const struct pid_entry *p, *last;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002447
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002448 error = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002449
Eric W. Biederman99f89552006-06-26 00:25:55 -07002450 if (!task)
2451 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002453 /*
2454 * Yes, it does not scale. And it should not. Don't add
2455 * new entries into /proc/<tgid>/ without very good reasons.
2456 */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002457 last = &ents[nents - 1];
2458 for (p = ents; p <= last; p++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002459 if (p->len != dentry->d_name.len)
2460 continue;
2461 if (!memcmp(dentry->d_name.name, p->name, p->len))
2462 break;
2463 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002464 if (p > last)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002465 goto out;
2466
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002467 error = proc_pident_instantiate(dir, dentry, task, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002468out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07002469 put_task_struct(task);
2470out_no_task:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002471 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002472}
2473
Eric Dumazetc5141e62007-05-08 00:26:15 -07002474static int proc_pident_fill_cache(struct file *filp, void *dirent,
2475 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002476{
2477 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2478 proc_pident_instantiate, task, p);
2479}
2480
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002481static int proc_pident_readdir(struct file *filp,
2482 void *dirent, filldir_t filldir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002483 const struct pid_entry *ents, unsigned int nents)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002484{
2485 int i;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002486 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002487 struct inode *inode = dentry->d_inode;
2488 struct task_struct *task = get_proc_task(inode);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002489 const struct pid_entry *p, *last;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002490 ino_t ino;
2491 int ret;
2492
2493 ret = -ENOENT;
2494 if (!task)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002495 goto out_no_task;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002496
2497 ret = 0;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002498 i = filp->f_pos;
2499 switch (i) {
2500 case 0:
2501 ino = inode->i_ino;
2502 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2503 goto out;
2504 i++;
2505 filp->f_pos++;
2506 /* fall through */
2507 case 1:
2508 ino = parent_ino(dentry);
2509 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2510 goto out;
2511 i++;
2512 filp->f_pos++;
2513 /* fall through */
2514 default:
2515 i -= 2;
2516 if (i >= nents) {
2517 ret = 1;
2518 goto out;
2519 }
2520 p = ents + i;
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002521 last = &ents[nents - 1];
2522 while (p <= last) {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002523 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002524 goto out;
2525 filp->f_pos++;
2526 p++;
2527 }
2528 }
2529
2530 ret = 1;
2531out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07002532 put_task_struct(task);
2533out_no_task:
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002534 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002535}
2536
Linus Torvalds1da177e2005-04-16 15:20:36 -07002537#ifdef CONFIG_SECURITY
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002538static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2539 size_t count, loff_t *ppos)
2540{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002541 struct inode * inode = file->f_path.dentry->d_inode;
Al Viro04ff9702007-03-12 16:17:58 +00002542 char *p = NULL;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002543 ssize_t length;
2544 struct task_struct *task = get_proc_task(inode);
2545
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002546 if (!task)
Al Viro04ff9702007-03-12 16:17:58 +00002547 return -ESRCH;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002548
2549 length = security_getprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002550 (char*)file->f_path.dentry->d_name.name,
Al Viro04ff9702007-03-12 16:17:58 +00002551 &p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002552 put_task_struct(task);
Al Viro04ff9702007-03-12 16:17:58 +00002553 if (length > 0)
2554 length = simple_read_from_buffer(buf, count, ppos, p, length);
2555 kfree(p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002556 return length;
2557}
2558
2559static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2560 size_t count, loff_t *ppos)
2561{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002562 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002563 char *page;
2564 ssize_t length;
2565 struct task_struct *task = get_proc_task(inode);
2566
2567 length = -ESRCH;
2568 if (!task)
2569 goto out_no_task;
2570 if (count > PAGE_SIZE)
2571 count = PAGE_SIZE;
2572
2573 /* No partial writes. */
2574 length = -EINVAL;
2575 if (*ppos != 0)
2576 goto out;
2577
2578 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -07002579 page = (char*)__get_free_page(GFP_TEMPORARY);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002580 if (!page)
2581 goto out;
2582
2583 length = -EFAULT;
2584 if (copy_from_user(page, buf, count))
2585 goto out_free;
2586
David Howells107db7c2009-05-08 13:55:27 +01002587 /* Guard against adverse ptrace interaction */
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07002588 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
David Howells107db7c2009-05-08 13:55:27 +01002589 if (length < 0)
2590 goto out_free;
2591
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002592 length = security_setprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002593 (char*)file->f_path.dentry->d_name.name,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002594 (void*)page, count);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07002595 mutex_unlock(&task->signal->cred_guard_mutex);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002596out_free:
2597 free_page((unsigned long) page);
2598out:
2599 put_task_struct(task);
2600out_no_task:
2601 return length;
2602}
2603
Arjan van de Ven00977a52007-02-12 00:55:34 -08002604static const struct file_operations proc_pid_attr_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002605 .read = proc_pid_attr_read,
2606 .write = proc_pid_attr_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01002607 .llseek = generic_file_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002608};
2609
Eric Dumazetc5141e62007-05-08 00:26:15 -07002610static const struct pid_entry attr_dir_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002611 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2612 REG("prev", S_IRUGO, proc_pid_attr_operations),
2613 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2614 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2615 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2616 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002617};
2618
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002619static int proc_attr_dir_readdir(struct file * filp,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002620 void * dirent, filldir_t filldir)
2621{
2622 return proc_pident_readdir(filp,dirent,filldir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002623 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002624}
2625
Arjan van de Ven00977a52007-02-12 00:55:34 -08002626static const struct file_operations proc_attr_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002627 .read = generic_read_dir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002628 .readdir = proc_attr_dir_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002629 .llseek = default_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002630};
2631
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002632static struct dentry *proc_attr_dir_lookup(struct inode *dir,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002633 struct dentry *dentry, struct nameidata *nd)
2634{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002635 return proc_pident_lookup(dir, dentry,
2636 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002637}
2638
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002639static const struct inode_operations proc_attr_dir_inode_operations = {
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002640 .lookup = proc_attr_dir_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07002641 .getattr = pid_getattr,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07002642 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002643};
2644
Linus Torvalds1da177e2005-04-16 15:20:36 -07002645#endif
2646
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08002647#ifdef CONFIG_ELF_CORE
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002648static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2649 size_t count, loff_t *ppos)
2650{
2651 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2652 struct mm_struct *mm;
2653 char buffer[PROC_NUMBUF];
2654 size_t len;
2655 int ret;
2656
2657 if (!task)
2658 return -ESRCH;
2659
2660 ret = 0;
2661 mm = get_task_mm(task);
2662 if (mm) {
2663 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2664 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2665 MMF_DUMP_FILTER_SHIFT));
2666 mmput(mm);
2667 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2668 }
2669
2670 put_task_struct(task);
2671
2672 return ret;
2673}
2674
2675static ssize_t proc_coredump_filter_write(struct file *file,
2676 const char __user *buf,
2677 size_t count,
2678 loff_t *ppos)
2679{
2680 struct task_struct *task;
2681 struct mm_struct *mm;
2682 char buffer[PROC_NUMBUF], *end;
2683 unsigned int val;
2684 int ret;
2685 int i;
2686 unsigned long mask;
2687
2688 ret = -EFAULT;
2689 memset(buffer, 0, sizeof(buffer));
2690 if (count > sizeof(buffer) - 1)
2691 count = sizeof(buffer) - 1;
2692 if (copy_from_user(buffer, buf, count))
2693 goto out_no_task;
2694
2695 ret = -EINVAL;
2696 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2697 if (*end == '\n')
2698 end++;
2699 if (end - buffer == 0)
2700 goto out_no_task;
2701
2702 ret = -ESRCH;
2703 task = get_proc_task(file->f_dentry->d_inode);
2704 if (!task)
2705 goto out_no_task;
2706
2707 ret = end - buffer;
2708 mm = get_task_mm(task);
2709 if (!mm)
2710 goto out_no_mm;
2711
2712 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2713 if (val & mask)
2714 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2715 else
2716 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2717 }
2718
2719 mmput(mm);
2720 out_no_mm:
2721 put_task_struct(task);
2722 out_no_task:
2723 return ret;
2724}
2725
2726static const struct file_operations proc_coredump_filter_operations = {
2727 .read = proc_coredump_filter_read,
2728 .write = proc_coredump_filter_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01002729 .llseek = generic_file_llseek,
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002730};
2731#endif
2732
Linus Torvalds1da177e2005-04-16 15:20:36 -07002733/*
2734 * /proc/self:
2735 */
2736static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2737 int buflen)
2738{
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002739 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002740 pid_t tgid = task_tgid_nr_ns(current, ns);
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002741 char tmp[PROC_NUMBUF];
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002742 if (!tgid)
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002743 return -ENOENT;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002744 sprintf(tmp, "%d", tgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002745 return vfs_readlink(dentry,buffer,buflen,tmp);
2746}
2747
Al Viro008b1502005-08-20 00:17:39 +01002748static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002749{
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002750 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002751 pid_t tgid = task_tgid_nr_ns(current, ns);
Al Viro7fee4862010-01-14 01:03:28 -05002752 char *name = ERR_PTR(-ENOENT);
2753 if (tgid) {
2754 name = __getname();
2755 if (!name)
2756 name = ERR_PTR(-ENOMEM);
2757 else
2758 sprintf(name, "%d", tgid);
2759 }
2760 nd_set_link(nd, name);
2761 return NULL;
2762}
2763
2764static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2765 void *cookie)
2766{
2767 char *s = nd_get_link(nd);
2768 if (!IS_ERR(s))
2769 __putname(s);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002770}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002771
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002772static const struct inode_operations proc_self_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002773 .readlink = proc_self_readlink,
2774 .follow_link = proc_self_follow_link,
Al Viro7fee4862010-01-14 01:03:28 -05002775 .put_link = proc_self_put_link,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002776};
2777
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002778/*
Eric W. Biederman801199c2006-10-02 02:18:48 -07002779 * proc base
2780 *
2781 * These are the directory entries in the root directory of /proc
2782 * that properly belong to the /proc filesystem, as they describe
2783 * describe something that is process related.
2784 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07002785static const struct pid_entry proc_base_stuff[] = {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002786 NOD("self", S_IFLNK|S_IRWXUGO,
Eric W. Biederman801199c2006-10-02 02:18:48 -07002787 &proc_self_inode_operations, NULL, {}),
Eric W. Biederman801199c2006-10-02 02:18:48 -07002788};
2789
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002790static struct dentry *proc_base_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002791 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman801199c2006-10-02 02:18:48 -07002792{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002793 const struct pid_entry *p = ptr;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002794 struct inode *inode;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002795 struct proc_inode *ei;
Dan Carpenter73d36462010-05-26 14:43:25 -07002796 struct dentry *error;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002797
2798 /* Allocate the inode */
2799 error = ERR_PTR(-ENOMEM);
2800 inode = new_inode(dir->i_sb);
2801 if (!inode)
2802 goto out;
2803
2804 /* Initialize the inode */
2805 ei = PROC_I(inode);
Christoph Hellwig85fe4022010-10-23 11:19:54 -04002806 inode->i_ino = get_next_ino();
Eric W. Biederman801199c2006-10-02 02:18:48 -07002807 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002808
2809 /*
2810 * grab the reference to the task.
2811 */
Oleg Nesterov1a657f782006-10-02 02:18:59 -07002812 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002813 if (!ei->pid)
2814 goto out_iput;
2815
Eric W. Biederman801199c2006-10-02 02:18:48 -07002816 inode->i_mode = p->mode;
2817 if (S_ISDIR(inode->i_mode))
Miklos Szeredibfe86842011-10-28 14:13:29 +02002818 set_nlink(inode, 2);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002819 if (S_ISLNK(inode->i_mode))
2820 inode->i_size = 64;
2821 if (p->iop)
2822 inode->i_op = p->iop;
2823 if (p->fop)
2824 inode->i_fop = p->fop;
2825 ei->op = p->op;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002826 d_add(dentry, inode);
2827 error = NULL;
2828out:
Eric W. Biederman801199c2006-10-02 02:18:48 -07002829 return error;
2830out_iput:
2831 iput(inode);
2832 goto out;
2833}
2834
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002835static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2836{
2837 struct dentry *error;
2838 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002839 const struct pid_entry *p, *last;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002840
2841 error = ERR_PTR(-ENOENT);
2842
2843 if (!task)
2844 goto out_no_task;
2845
2846 /* Lookup the directory entry */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002847 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2848 for (p = proc_base_stuff; p <= last; p++) {
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002849 if (p->len != dentry->d_name.len)
2850 continue;
2851 if (!memcmp(dentry->d_name.name, p->name, p->len))
2852 break;
2853 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002854 if (p > last)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002855 goto out;
2856
2857 error = proc_base_instantiate(dir, dentry, task, p);
2858
2859out:
2860 put_task_struct(task);
2861out_no_task:
2862 return error;
2863}
2864
Eric Dumazetc5141e62007-05-08 00:26:15 -07002865static int proc_base_fill_cache(struct file *filp, void *dirent,
2866 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002867{
2868 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2869 proc_base_instantiate, task, p);
2870}
2871
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002872#ifdef CONFIG_TASK_IO_ACCOUNTING
Andrea Righi297c5d92008-07-25 01:48:49 -07002873static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002874{
Andrea Righi940389b2008-07-28 00:48:12 +02002875 struct task_io_accounting acct = task->ioac;
Andrea Righi59954772008-07-27 17:29:15 +02002876 unsigned long flags;
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002877 int result;
Andrea Righi297c5d92008-07-25 01:48:49 -07002878
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002879 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2880 if (result)
2881 return result;
2882
2883 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2884 result = -EACCES;
2885 goto out_unlock;
2886 }
Vasiliy Kulikov1d1221f2011-06-24 16:08:38 +04002887
Andrea Righi59954772008-07-27 17:29:15 +02002888 if (whole && lock_task_sighand(task, &flags)) {
2889 struct task_struct *t = task;
Andrea Righib2d002d2008-07-26 15:22:27 -07002890
Andrea Righi59954772008-07-27 17:29:15 +02002891 task_io_accounting_add(&acct, &task->signal->ioac);
2892 while_each_thread(task, t)
2893 task_io_accounting_add(&acct, &t->ioac);
Andrea Righi297c5d92008-07-25 01:48:49 -07002894
Andrea Righi59954772008-07-27 17:29:15 +02002895 unlock_task_sighand(task, &flags);
Andrea Righi297c5d92008-07-25 01:48:49 -07002896 }
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002897 result = sprintf(buffer,
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002898 "rchar: %llu\n"
2899 "wchar: %llu\n"
2900 "syscr: %llu\n"
2901 "syscw: %llu\n"
2902 "read_bytes: %llu\n"
2903 "write_bytes: %llu\n"
2904 "cancelled_write_bytes: %llu\n",
Alexander Beregalov7c443192008-08-05 13:01:34 -07002905 (unsigned long long)acct.rchar,
2906 (unsigned long long)acct.wchar,
2907 (unsigned long long)acct.syscr,
2908 (unsigned long long)acct.syscw,
2909 (unsigned long long)acct.read_bytes,
2910 (unsigned long long)acct.write_bytes,
2911 (unsigned long long)acct.cancelled_write_bytes);
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002912out_unlock:
2913 mutex_unlock(&task->signal->cred_guard_mutex);
2914 return result;
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002915}
Andrea Righi297c5d92008-07-25 01:48:49 -07002916
2917static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2918{
2919 return do_io_accounting(task, buffer, 0);
2920}
2921
2922static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2923{
2924 return do_io_accounting(task, buffer, 1);
2925}
2926#endif /* CONFIG_TASK_IO_ACCOUNTING */
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002927
Eric W. Biederman22d917d2011-11-17 00:11:58 -08002928#ifdef CONFIG_USER_NS
2929static int proc_id_map_open(struct inode *inode, struct file *file,
2930 struct seq_operations *seq_ops)
2931{
2932 struct user_namespace *ns = NULL;
2933 struct task_struct *task;
2934 struct seq_file *seq;
2935 int ret = -EINVAL;
2936
2937 task = get_proc_task(inode);
2938 if (task) {
2939 rcu_read_lock();
2940 ns = get_user_ns(task_cred_xxx(task, user_ns));
2941 rcu_read_unlock();
2942 put_task_struct(task);
2943 }
2944 if (!ns)
2945 goto err;
2946
2947 ret = seq_open(file, seq_ops);
2948 if (ret)
2949 goto err_put_ns;
2950
2951 seq = file->private_data;
2952 seq->private = ns;
2953
2954 return 0;
2955err_put_ns:
2956 put_user_ns(ns);
2957err:
2958 return ret;
2959}
2960
2961static int proc_id_map_release(struct inode *inode, struct file *file)
2962{
2963 struct seq_file *seq = file->private_data;
2964 struct user_namespace *ns = seq->private;
2965 put_user_ns(ns);
2966 return seq_release(inode, file);
2967}
2968
2969static int proc_uid_map_open(struct inode *inode, struct file *file)
2970{
2971 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2972}
2973
2974static int proc_gid_map_open(struct inode *inode, struct file *file)
2975{
2976 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2977}
2978
2979static const struct file_operations proc_uid_map_operations = {
2980 .open = proc_uid_map_open,
2981 .write = proc_uid_map_write,
2982 .read = seq_read,
2983 .llseek = seq_lseek,
2984 .release = proc_id_map_release,
2985};
2986
2987static const struct file_operations proc_gid_map_operations = {
2988 .open = proc_gid_map_open,
2989 .write = proc_gid_map_write,
2990 .read = seq_read,
2991 .llseek = seq_lseek,
2992 .release = proc_id_map_release,
2993};
2994#endif /* CONFIG_USER_NS */
2995
Kees Cook47830722008-10-06 03:11:58 +04002996static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2997 struct pid *pid, struct task_struct *task)
2998{
Al Viroa9712bc2011-03-23 15:52:50 -04002999 int err = lock_trace(task);
3000 if (!err) {
3001 seq_printf(m, "%08x\n", task->personality);
3002 unlock_trace(task);
3003 }
3004 return err;
Kees Cook47830722008-10-06 03:11:58 +04003005}
3006
Eric W. Biederman801199c2006-10-02 02:18:48 -07003007/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003008 * Thread groups
3009 */
Arjan van de Ven00977a52007-02-12 00:55:34 -08003010static const struct file_operations proc_task_operations;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003011static const struct inode_operations proc_task_inode_operations;
Eric W. Biederman20cdc892006-10-02 02:17:07 -07003012
Eric Dumazetc5141e62007-05-08 00:26:15 -07003013static const struct pid_entry tgid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003014 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3015 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
Pavel Emelyanov640708a2012-01-10 15:11:23 -08003016#ifdef CONFIG_CHECKPOINT_RESTORE
3017 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3018#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003019 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08003020 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07003021#ifdef CONFIG_NET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003022 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07003023#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003024 REG("environ", S_IRUSR, proc_environ_operations),
3025 INF("auxv", S_IRUSR, proc_pid_auxv),
3026 ONE("status", S_IRUGO, proc_pid_status),
Al Viroa9712bc2011-03-23 15:52:50 -04003027 ONE("personality", S_IRUGO, proc_pid_personality),
Jiri Olsa3036e7b2010-09-30 15:15:33 -07003028 INF("limits", S_IRUGO, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02003029#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003030 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02003031#endif
Mike Galbraith5091faa2010-11-30 14:18:03 +01003032#ifdef CONFIG_SCHED_AUTOGROUP
3033 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3034#endif
john stultz4614a696b2009-12-14 18:00:05 -08003035 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
Roland McGrathebcb6732008-07-25 19:46:00 -07003036#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Al Viroa9712bc2011-03-23 15:52:50 -04003037 INF("syscall", S_IRUGO, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07003038#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003039 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3040 ONE("stat", S_IRUGO, proc_tgid_stat),
3041 ONE("statm", S_IRUGO, proc_pid_statm),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07003042 REG("maps", S_IRUGO, proc_pid_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003043#ifdef CONFIG_NUMA
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07003044 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003045#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003046 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3047 LNK("cwd", proc_cwd_link),
3048 LNK("root", proc_root_link),
3049 LNK("exe", proc_exe_link),
3050 REG("mounts", S_IRUGO, proc_mounts_operations),
3051 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3052 REG("mountstats", S_IRUSR, proc_mountstats_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08003053#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003054 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07003055 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
Al Viroca6b0bf2011-02-15 22:04:37 -05003056 REG("pagemap", S_IRUGO, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003057#endif
3058#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003059 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003060#endif
3061#ifdef CONFIG_KALLSYMS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003062 INF("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003063#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03003064#ifdef CONFIG_STACKTRACE
Al Viroa9712bc2011-03-23 15:52:50 -04003065 ONE("stack", S_IRUGO, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003066#endif
3067#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003068 INF("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003069#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01003070#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003071 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01003072#endif
Paul Menage8793d852007-10-18 23:39:39 -07003073#ifdef CONFIG_PROC_PID_CPUSET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003074 REG("cpuset", S_IRUGO, proc_cpuset_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003075#endif
Paul Menagea4243162007-10-18 23:39:35 -07003076#ifdef CONFIG_CGROUPS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003077 REG("cgroup", S_IRUGO, proc_cgroup_operations),
Paul Menagea4243162007-10-18 23:39:35 -07003078#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003079 INF("oom_score", S_IRUGO, proc_oom_score),
3080 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
David Rientjesa63d83f2010-08-09 17:19:46 -07003081 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003082#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003083 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3084 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003085#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003086#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003087 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003088#endif
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08003089#ifdef CONFIG_ELF_CORE
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003090 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07003091#endif
Andrew Mortonaba76fd2006-12-10 02:19:48 -08003092#ifdef CONFIG_TASK_IO_ACCOUNTING
Vasiliy Kulikov1d1221f2011-06-24 16:08:38 +04003093 INF("io", S_IRUSR, proc_tgid_io_accounting),
Andrew Mortonaba76fd2006-12-10 02:19:48 -08003094#endif
Chris Metcalff133ecc2011-05-26 12:40:09 -04003095#ifdef CONFIG_HARDWALL
3096 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3097#endif
Eric W. Biederman22d917d2011-11-17 00:11:58 -08003098#ifdef CONFIG_USER_NS
3099 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3100 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3101#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003102};
3103
3104static int proc_tgid_base_readdir(struct file * filp,
3105 void * dirent, filldir_t filldir)
3106{
3107 return proc_pident_readdir(filp,dirent,filldir,
3108 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
3109}
3110
Arjan van de Ven00977a52007-02-12 00:55:34 -08003111static const struct file_operations proc_tgid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003112 .read = generic_read_dir,
3113 .readdir = proc_tgid_base_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02003114 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003115};
3116
3117static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003118 return proc_pident_lookup(dir, dentry,
3119 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003120}
3121
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003122static const struct inode_operations proc_tgid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003123 .lookup = proc_tgid_base_lookup,
3124 .getattr = pid_getattr,
3125 .setattr = proc_setattr,
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003126 .permission = proc_pid_permission,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003127};
3128
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003129static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003130{
Eric W. Biederman48e64842006-06-26 00:25:48 -07003131 struct dentry *dentry, *leader, *dir;
Eric W. Biederman8578cea2006-06-26 00:25:54 -07003132 char buf[PROC_NUMBUF];
Eric W. Biederman48e64842006-06-26 00:25:48 -07003133 struct qstr name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003134
Eric W. Biederman48e64842006-06-26 00:25:48 -07003135 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003136 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3137 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003138 if (dentry) {
Sukadev Bhattiprolu29f12ca2009-11-11 14:26:32 -08003139 shrink_dcache_parent(dentry);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003140 d_drop(dentry);
3141 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003142 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003143
Eric W. Biederman48e64842006-06-26 00:25:48 -07003144 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003145 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3146 leader = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003147 if (!leader)
3148 goto out;
3149
3150 name.name = "task";
3151 name.len = strlen(name.name);
3152 dir = d_hash_and_lookup(leader, &name);
3153 if (!dir)
3154 goto out_put_leader;
3155
3156 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003157 name.len = snprintf(buf, sizeof(buf), "%d", pid);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003158 dentry = d_hash_and_lookup(dir, &name);
3159 if (dentry) {
3160 shrink_dcache_parent(dentry);
3161 d_drop(dentry);
3162 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003163 }
Eric W. Biederman48e64842006-06-26 00:25:48 -07003164
3165 dput(dir);
3166out_put_leader:
3167 dput(leader);
3168out:
3169 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003170}
3171
Randy Dunlap0895e912007-10-21 21:00:10 -07003172/**
3173 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3174 * @task: task that should be flushed.
3175 *
3176 * When flushing dentries from proc, one needs to flush them from global
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003177 * proc (proc_mnt) and from all the namespaces' procs this task was seen
Randy Dunlap0895e912007-10-21 21:00:10 -07003178 * in. This call is supposed to do all of this job.
3179 *
3180 * Looks in the dcache for
3181 * /proc/@pid
3182 * /proc/@tgid/task/@pid
3183 * if either directory is present flushes it and all of it'ts children
3184 * from the dcache.
3185 *
3186 * It is safe and reasonable to cache /proc entries for a task until
3187 * that task exits. After that they just clog up the dcache with
3188 * useless entries, possibly causing useful dcache entries to be
3189 * flushed instead. This routine is proved to flush those useless
3190 * dcache entries at process exit time.
3191 *
3192 * NOTE: This routine is just an optimization so it does not guarantee
3193 * that no dcache entries will exist at process exit time it
3194 * just makes it very unlikely that any will persist.
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003195 */
3196
3197void proc_flush_task(struct task_struct *task)
3198{
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08003199 int i;
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07003200 struct pid *pid, *tgid;
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003201 struct upid *upid;
3202
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003203 pid = task_pid(task);
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07003204 tgid = task_tgid(task);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003205
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08003206 for (i = 0; i <= pid->level; i++) {
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003207 upid = &pid->numbers[i];
3208 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07003209 tgid->numbers[i].nr);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003210 }
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -07003211
3212 upid = &pid->numbers[pid->level];
3213 if (upid->nr == 1)
3214 pid_ns_release_proc(upid->ns);
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003215}
3216
Adrian Bunk9711ef92006-12-06 20:38:31 -08003217static struct dentry *proc_pid_instantiate(struct inode *dir,
3218 struct dentry * dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07003219 struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003220{
3221 struct dentry *error = ERR_PTR(-ENOENT);
3222 struct inode *inode;
3223
Eric W. Biederman61a28782006-10-02 02:18:49 -07003224 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003225 if (!inode)
3226 goto out;
3227
3228 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3229 inode->i_op = &proc_tgid_base_inode_operations;
3230 inode->i_fop = &proc_tgid_base_operations;
3231 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07003232
Miklos Szeredibfe86842011-10-28 14:13:29 +02003233 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3234 ARRAY_SIZE(tgid_base_stuff)));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003235
Nick Pigginfb045ad2011-01-07 17:49:55 +11003236 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003237
3238 d_add(dentry, inode);
3239 /* Close the race of the process dying before we return the dentry */
3240 if (pid_revalidate(dentry, NULL))
3241 error = NULL;
3242out:
3243 return error;
3244}
3245
Linus Torvalds1da177e2005-04-16 15:20:36 -07003246struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3247{
Dan Carpenter73d36462010-05-26 14:43:25 -07003248 struct dentry *result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003249 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003250 unsigned tgid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003251 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003252
Eric W. Biederman801199c2006-10-02 02:18:48 -07003253 result = proc_base_lookup(dir, dentry);
3254 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3255 goto out;
3256
Linus Torvalds1da177e2005-04-16 15:20:36 -07003257 tgid = name_to_int(dentry);
3258 if (tgid == ~0U)
3259 goto out;
3260
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003261 ns = dentry->d_sb->s_fs_info;
Eric W. Biedermande758732006-06-26 00:25:51 -07003262 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003263 task = find_task_by_pid_ns(tgid, ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003264 if (task)
3265 get_task_struct(task);
Eric W. Biedermande758732006-06-26 00:25:51 -07003266 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003267 if (!task)
3268 goto out;
3269
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003270 result = proc_pid_instantiate(dir, dentry, task, NULL);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003271 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003272out:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07003273 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003274}
3275
Linus Torvalds1da177e2005-04-16 15:20:36 -07003276/*
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003277 * Find the first task with tgid >= tgid
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003278 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07003279 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003280struct tgid_iter {
3281 unsigned int tgid;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003282 struct task_struct *task;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003283};
3284static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3285{
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003286 struct pid *pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003287
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003288 if (iter.task)
3289 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003290 rcu_read_lock();
3291retry:
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003292 iter.task = NULL;
3293 pid = find_ge_pid(iter.tgid, ns);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003294 if (pid) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003295 iter.tgid = pid_nr_ns(pid, ns);
3296 iter.task = pid_task(pid, PIDTYPE_PID);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003297 /* What we to know is if the pid we have find is the
3298 * pid of a thread_group_leader. Testing for task
3299 * being a thread_group_leader is the obvious thing
3300 * todo but there is a window when it fails, due to
3301 * the pid transfer logic in de_thread.
3302 *
3303 * So we perform the straight forward test of seeing
3304 * if the pid we have found is the pid of a thread
3305 * group leader, and don't worry if the task we have
3306 * found doesn't happen to be a thread group leader.
3307 * As we don't care in the case of readdir.
3308 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003309 if (!iter.task || !has_group_leader_pid(iter.task)) {
3310 iter.tgid += 1;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003311 goto retry;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003312 }
3313 get_task_struct(iter.task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003314 }
Eric W. Biederman454cc102006-06-26 00:25:51 -07003315 rcu_read_unlock();
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003316 return iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003317}
3318
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003319#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003320
Eric W. Biederman61a28782006-10-02 02:18:49 -07003321static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003322 struct tgid_iter iter)
Eric W. Biederman61a28782006-10-02 02:18:49 -07003323{
3324 char name[PROC_NUMBUF];
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003325 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003326 return proc_fill_cache(filp, dirent, filldir, name, len,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003327 proc_pid_instantiate, iter.task, NULL);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003328}
3329
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003330static int fake_filldir(void *buf, const char *name, int namelen,
3331 loff_t offset, u64 ino, unsigned d_type)
3332{
3333 return 0;
3334}
3335
Linus Torvalds1da177e2005-04-16 15:20:36 -07003336/* for the /proc/ directory itself, after non-process stuff has been done */
3337int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3338{
Linus Torvaldsd8bdc592011-04-18 10:36:54 -07003339 unsigned int nr;
3340 struct task_struct *reaper;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003341 struct tgid_iter iter;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003342 struct pid_namespace *ns;
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003343 filldir_t __filldir;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003344
Linus Torvaldsd8bdc592011-04-18 10:36:54 -07003345 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3346 goto out_no_task;
3347 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3348
3349 reaper = get_proc_task(filp->f_path.dentry->d_inode);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003350 if (!reaper)
3351 goto out_no_task;
3352
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003353 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
Eric Dumazetc5141e62007-05-08 00:26:15 -07003354 const struct pid_entry *p = &proc_base_stuff[nr];
Eric W. Biederman61a28782006-10-02 02:18:49 -07003355 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
Eric W. Biederman801199c2006-10-02 02:18:48 -07003356 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003357 }
3358
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003359 ns = filp->f_dentry->d_sb->s_fs_info;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003360 iter.task = NULL;
3361 iter.tgid = filp->f_pos - TGID_OFFSET;
3362 for (iter = next_tgid(ns, iter);
3363 iter.task;
3364 iter.tgid += 1, iter = next_tgid(ns, iter)) {
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003365 if (has_pid_permissions(ns, iter.task, 2))
3366 __filldir = filldir;
3367 else
3368 __filldir = fake_filldir;
3369
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003370 filp->f_pos = iter.tgid + TGID_OFFSET;
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003371 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003372 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003373 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003374 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003375 }
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003376 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3377out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07003378 put_task_struct(reaper);
3379out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07003380 return 0;
3381}
3382
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003383/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003384 * Tasks
3385 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07003386static const struct pid_entry tid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003387 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
Jerome Marchand38355412010-04-27 13:13:06 -07003388 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08003389 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003390 REG("environ", S_IRUSR, proc_environ_operations),
3391 INF("auxv", S_IRUSR, proc_pid_auxv),
3392 ONE("status", S_IRUGO, proc_pid_status),
Al Viroa9712bc2011-03-23 15:52:50 -04003393 ONE("personality", S_IRUGO, proc_pid_personality),
Jiri Olsa3036e7b2010-09-30 15:15:33 -07003394 INF("limits", S_IRUGO, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02003395#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003396 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02003397#endif
john stultz4614a696b2009-12-14 18:00:05 -08003398 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
Roland McGrathebcb6732008-07-25 19:46:00 -07003399#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Al Viroa9712bc2011-03-23 15:52:50 -04003400 INF("syscall", S_IRUGO, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07003401#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003402 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3403 ONE("stat", S_IRUGO, proc_tid_stat),
3404 ONE("statm", S_IRUGO, proc_pid_statm),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07003405 REG("maps", S_IRUGO, proc_tid_maps_operations),
Cyrill Gorcunov818411612012-05-31 16:26:43 -07003406#ifdef CONFIG_CHECKPOINT_RESTORE
3407 REG("children", S_IRUGO, proc_tid_children_operations),
3408#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003409#ifdef CONFIG_NUMA
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07003410 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003411#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003412 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3413 LNK("cwd", proc_cwd_link),
3414 LNK("root", proc_root_link),
3415 LNK("exe", proc_exe_link),
3416 REG("mounts", S_IRUGO, proc_mounts_operations),
3417 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08003418#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003419 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07003420 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
Al Viroca6b0bf2011-02-15 22:04:37 -05003421 REG("pagemap", S_IRUGO, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003422#endif
3423#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003424 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003425#endif
3426#ifdef CONFIG_KALLSYMS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003427 INF("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003428#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03003429#ifdef CONFIG_STACKTRACE
Al Viroa9712bc2011-03-23 15:52:50 -04003430 ONE("stack", S_IRUGO, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003431#endif
3432#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003433 INF("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003434#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01003435#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003436 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01003437#endif
Paul Menage8793d852007-10-18 23:39:39 -07003438#ifdef CONFIG_PROC_PID_CPUSET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003439 REG("cpuset", S_IRUGO, proc_cpuset_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003440#endif
Paul Menagea4243162007-10-18 23:39:35 -07003441#ifdef CONFIG_CGROUPS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003442 REG("cgroup", S_IRUGO, proc_cgroup_operations),
Paul Menagea4243162007-10-18 23:39:35 -07003443#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003444 INF("oom_score", S_IRUGO, proc_oom_score),
3445 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
David Rientjesa63d83f2010-08-09 17:19:46 -07003446 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003447#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003448 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
Al Viro26ec3c62011-02-15 21:24:05 -05003449 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003450#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003451#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003452 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003453#endif
Andrea Righi297c5d92008-07-25 01:48:49 -07003454#ifdef CONFIG_TASK_IO_ACCOUNTING
Vasiliy Kulikov1d1221f2011-06-24 16:08:38 +04003455 INF("io", S_IRUSR, proc_tid_io_accounting),
Andrea Righi297c5d92008-07-25 01:48:49 -07003456#endif
Chris Metcalff133ecc2011-05-26 12:40:09 -04003457#ifdef CONFIG_HARDWALL
3458 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3459#endif
Eric W. Biederman22d917d2011-11-17 00:11:58 -08003460#ifdef CONFIG_USER_NS
3461 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3462 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3463#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003464};
3465
3466static int proc_tid_base_readdir(struct file * filp,
3467 void * dirent, filldir_t filldir)
3468{
3469 return proc_pident_readdir(filp,dirent,filldir,
3470 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3471}
3472
3473static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003474 return proc_pident_lookup(dir, dentry,
3475 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003476}
3477
Arjan van de Ven00977a52007-02-12 00:55:34 -08003478static const struct file_operations proc_tid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003479 .read = generic_read_dir,
3480 .readdir = proc_tid_base_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02003481 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003482};
3483
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003484static const struct inode_operations proc_tid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003485 .lookup = proc_tid_base_lookup,
3486 .getattr = pid_getattr,
3487 .setattr = proc_setattr,
3488};
3489
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003490static struct dentry *proc_task_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07003491 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003492{
3493 struct dentry *error = ERR_PTR(-ENOENT);
3494 struct inode *inode;
Eric W. Biederman61a28782006-10-02 02:18:49 -07003495 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003496
3497 if (!inode)
3498 goto out;
3499 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3500 inode->i_op = &proc_tid_base_inode_operations;
3501 inode->i_fop = &proc_tid_base_operations;
3502 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07003503
Miklos Szeredibfe86842011-10-28 14:13:29 +02003504 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3505 ARRAY_SIZE(tid_base_stuff)));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003506
Nick Pigginfb045ad2011-01-07 17:49:55 +11003507 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003508
3509 d_add(dentry, inode);
3510 /* Close the race of the process dying before we return the dentry */
3511 if (pid_revalidate(dentry, NULL))
3512 error = NULL;
3513out:
3514 return error;
3515}
3516
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003517static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3518{
3519 struct dentry *result = ERR_PTR(-ENOENT);
3520 struct task_struct *task;
3521 struct task_struct *leader = get_proc_task(dir);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003522 unsigned tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003523 struct pid_namespace *ns;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003524
3525 if (!leader)
3526 goto out_no_task;
3527
3528 tid = name_to_int(dentry);
3529 if (tid == ~0U)
3530 goto out;
3531
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003532 ns = dentry->d_sb->s_fs_info;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003533 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003534 task = find_task_by_pid_ns(tid, ns);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003535 if (task)
3536 get_task_struct(task);
3537 rcu_read_unlock();
3538 if (!task)
3539 goto out;
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07003540 if (!same_thread_group(leader, task))
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003541 goto out_drop_task;
3542
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003543 result = proc_task_instantiate(dir, dentry, task, NULL);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003544out_drop_task:
3545 put_task_struct(task);
3546out:
3547 put_task_struct(leader);
3548out_no_task:
3549 return result;
3550}
3551
3552/*
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003553 * Find the first tid of a thread group to return to user space.
3554 *
3555 * Usually this is just the thread group leader, but if the users
3556 * buffer was too small or there was a seek into the middle of the
3557 * directory we have more work todo.
3558 *
3559 * In the case of a short read we start with find_task_by_pid.
3560 *
3561 * In the case of a seek we start with the leader and walk nr
3562 * threads past it.
3563 */
Eric W. Biedermancc288732006-06-26 00:26:01 -07003564static struct task_struct *first_tid(struct task_struct *leader,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003565 int tid, int nr, struct pid_namespace *ns)
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003566{
Oleg Nesterova872ff02006-06-26 00:26:01 -07003567 struct task_struct *pos;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003568
Eric W. Biedermancc288732006-06-26 00:26:01 -07003569 rcu_read_lock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003570 /* Attempt to start with the pid of a thread */
3571 if (tid && (nr > 0)) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003572 pos = find_task_by_pid_ns(tid, ns);
Oleg Nesterova872ff02006-06-26 00:26:01 -07003573 if (pos && (pos->group_leader == leader))
3574 goto found;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003575 }
3576
3577 /* If nr exceeds the number of threads there is nothing todo */
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003578 pos = NULL;
Oleg Nesterova872ff02006-06-26 00:26:01 -07003579 if (nr && nr >= get_nr_threads(leader))
3580 goto out;
3581
3582 /* If we haven't found our starting place yet start
3583 * with the leader and walk nr threads forward.
3584 */
3585 for (pos = leader; nr > 0; --nr) {
3586 pos = next_thread(pos);
3587 if (pos == leader) {
3588 pos = NULL;
3589 goto out;
3590 }
3591 }
3592found:
3593 get_task_struct(pos);
3594out:
Eric W. Biedermancc288732006-06-26 00:26:01 -07003595 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003596 return pos;
3597}
3598
3599/*
3600 * Find the next thread in the thread list.
3601 * Return NULL if there is an error or no next thread.
3602 *
3603 * The reference to the input task_struct is released.
3604 */
3605static struct task_struct *next_tid(struct task_struct *start)
3606{
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003607 struct task_struct *pos = NULL;
Eric W. Biedermancc288732006-06-26 00:26:01 -07003608 rcu_read_lock();
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003609 if (pid_alive(start)) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003610 pos = next_thread(start);
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003611 if (thread_group_leader(pos))
3612 pos = NULL;
3613 else
3614 get_task_struct(pos);
3615 }
Eric W. Biedermancc288732006-06-26 00:26:01 -07003616 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003617 put_task_struct(start);
3618 return pos;
3619}
3620
Eric W. Biederman61a28782006-10-02 02:18:49 -07003621static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3622 struct task_struct *task, int tid)
3623{
3624 char name[PROC_NUMBUF];
3625 int len = snprintf(name, sizeof(name), "%d", tid);
3626 return proc_fill_cache(filp, dirent, filldir, name, len,
3627 proc_task_instantiate, task, NULL);
3628}
3629
Linus Torvalds1da177e2005-04-16 15:20:36 -07003630/* for the /proc/TGID/task/ directories */
3631static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3632{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08003633 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003634 struct inode *inode = dentry->d_inode;
Guillaume Chazarain7d895242007-01-31 23:48:14 -08003635 struct task_struct *leader = NULL;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003636 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003637 int retval = -ENOENT;
3638 ino_t ino;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003639 int tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003640 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003641
Guillaume Chazarain7d895242007-01-31 23:48:14 -08003642 task = get_proc_task(inode);
3643 if (!task)
3644 goto out_no_task;
3645 rcu_read_lock();
3646 if (pid_alive(task)) {
3647 leader = task->group_leader;
3648 get_task_struct(leader);
3649 }
3650 rcu_read_unlock();
3651 put_task_struct(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003652 if (!leader)
3653 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003654 retval = 0;
3655
Linus Torvaldsee568b22009-03-17 10:02:35 -07003656 switch ((unsigned long)filp->f_pos) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003657 case 0:
3658 ino = inode->i_ino;
Zhang Leee6f7792009-03-16 14:44:31 +08003659 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003660 goto out;
Zhang Leee6f7792009-03-16 14:44:31 +08003661 filp->f_pos++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003662 /* fall through */
3663 case 1:
3664 ino = parent_ino(dentry);
Zhang Leee6f7792009-03-16 14:44:31 +08003665 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003666 goto out;
Zhang Leee6f7792009-03-16 14:44:31 +08003667 filp->f_pos++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003668 /* fall through */
3669 }
3670
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003671 /* f_version caches the tgid value that the last readdir call couldn't
3672 * return. lseek aka telldir automagically resets f_version to 0.
3673 */
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003674 ns = filp->f_dentry->d_sb->s_fs_info;
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07003675 tid = (int)filp->f_version;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003676 filp->f_version = 0;
Zhang Leee6f7792009-03-16 14:44:31 +08003677 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003678 task;
Zhang Leee6f7792009-03-16 14:44:31 +08003679 task = next_tid(task), filp->f_pos++) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003680 tid = task_pid_nr_ns(task, ns);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003681 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003682 /* returning this tgid failed, save it as the first
3683 * pid for the next readir call */
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07003684 filp->f_version = (u64)tid;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003685 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003686 break;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003687 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003688 }
3689out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07003690 put_task_struct(leader);
3691out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07003692 return retval;
3693}
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003694
3695static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3696{
3697 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07003698 struct task_struct *p = get_proc_task(inode);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003699 generic_fillattr(inode, stat);
3700
Eric W. Biederman99f89552006-06-26 00:25:55 -07003701 if (p) {
Eric W. Biederman99f89552006-06-26 00:25:55 -07003702 stat->nlink += get_nr_threads(p);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003703 put_task_struct(p);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003704 }
3705
3706 return 0;
3707}
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003708
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003709static const struct inode_operations proc_task_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003710 .lookup = proc_task_lookup,
3711 .getattr = proc_task_getattr,
3712 .setattr = proc_setattr,
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003713 .permission = proc_pid_permission,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003714};
3715
Arjan van de Ven00977a52007-02-12 00:55:34 -08003716static const struct file_operations proc_task_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003717 .read = generic_read_dir,
3718 .readdir = proc_task_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02003719 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003720};