blob: 3b42c1418f3118871d93275848f633217419129f [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/proc/base.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * proc base directory handling functions
7 *
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
Mauricio Line070ad42005-09-03 15:55:10 -070014 *
15 *
16 * Changelog:
17 * 17-Jan-2005
18 * Allan Bezerra
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
23 *
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
25 *
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
32 *
33 * Changelog:
34 * 21-Feb-2005
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
37 *
38 * ChangeLog:
39 * 10-Mar-2005
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
42 *
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
45 *
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
Linus Torvalds1da177e2005-04-16 15:20:36 -070048 */
49
50#include <asm/uaccess.h>
51
Linus Torvalds1da177e2005-04-16 15:20:36 -070052#include <linux/errno.h>
53#include <linux/time.h>
54#include <linux/proc_fs.h>
55#include <linux/stat.h>
Andrea Righi59954772008-07-27 17:29:15 +020056#include <linux/task_io_accounting_ops.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070057#include <linux/init.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080058#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070059#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040060#include <linux/fdtable.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#include <linux/string.h>
62#include <linux/seq_file.h>
63#include <linux/namei.h>
Kirill Korotaev6b3286e2006-12-08 02:37:56 -080064#include <linux/mnt_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070065#include <linux/mm.h>
David Rientjesa63d83f2010-08-09 17:19:46 -070066#include <linux/swap.h>
Dipankar Sarmab8359962005-09-09 13:04:14 -070067#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <linux/kallsyms.h>
Ken Chen2ec220e2008-11-10 11:26:08 +030069#include <linux/stacktrace.h>
Neil Hormand85f50d2007-10-18 23:40:37 -070070#include <linux/resource.h>
Kees Cook5096add2007-05-08 00:26:04 -070071#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070072#include <linux/mount.h>
73#include <linux/security.h>
74#include <linux/ptrace.h>
Roland McGrath0d094ef2008-07-25 19:45:49 -070075#include <linux/tracehook.h>
Paul Menagea4243162007-10-18 23:39:35 -070076#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070077#include <linux/cpuset.h>
78#include <linux/audit.h>
Al Viro5addc5d2005-11-07 17:15:49 -050079#include <linux/poll.h>
Serge E. Hallyn1651e142006-10-02 02:18:08 -070080#include <linux/nsproxy.h>
Alexey Dobriyan8ac773b2006-10-19 23:28:32 -070081#include <linux/oom.h>
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -070082#include <linux/elf.h>
Pavel Emelyanov60347f62007-10-18 23:40:03 -070083#include <linux/pid_namespace.h>
Al Viro5ad4e532009-03-29 19:50:06 -040084#include <linux/fs_struct.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090085#include <linux/slab.h>
Pavel Emelyanov640708a2012-01-10 15:11:23 -080086#include <linux/flex_array.h>
Chris Metcalff133ecc2011-05-26 12:40:09 -040087#ifdef CONFIG_HARDWALL
88#include <asm/hardwall.h>
89#endif
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -080090#include <trace/events/oom.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070091#include "internal.h"
92
Eric W. Biederman0f2fe202006-06-26 00:25:46 -070093/* NOTE:
94 * Implementing inode permission operations in /proc is almost
95 * certainly an error. Permission checks need to happen during
96 * each system call not at open time. The reason is that most of
97 * what we wish to check for permissions in /proc varies at runtime.
98 *
99 * The classic example of a problem is opening file descriptors
100 * in /proc for a task before it execs a suid executable.
101 */
102
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103struct pid_entry {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104 char *name;
Eric Dumazetc5141e62007-05-08 00:26:15 -0700105 int len;
Al Virod161a132011-07-24 03:36:29 -0400106 umode_t mode;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800107 const struct inode_operations *iop;
Arjan van de Ven00977a52007-02-12 00:55:34 -0800108 const struct file_operations *fop;
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700109 union proc_op op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110};
111
Eric W. Biederman61a28782006-10-02 02:18:49 -0700112#define NOD(NAME, MODE, IOP, FOP, OP) { \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700113 .name = (NAME), \
Eric Dumazetc5141e62007-05-08 00:26:15 -0700114 .len = sizeof(NAME) - 1, \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700115 .mode = MODE, \
116 .iop = IOP, \
117 .fop = FOP, \
118 .op = OP, \
119}
120
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300121#define DIR(NAME, MODE, iops, fops) \
122 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
123#define LNK(NAME, get_link) \
Eric W. Biederman61a28782006-10-02 02:18:49 -0700124 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700125 &proc_pid_link_inode_operations, NULL, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300126 { .proc_get_link = get_link } )
127#define REG(NAME, MODE, fops) \
128 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
129#define INF(NAME, MODE, read) \
Eric W. Biederman61a28782006-10-02 02:18:49 -0700130 NOD(NAME, (S_IFREG|(MODE)), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700131 NULL, &proc_info_file_operations, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300132 { .proc_read = read } )
133#define ONE(NAME, MODE, show) \
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800134 NOD(NAME, (S_IFREG|(MODE)), \
135 NULL, &proc_single_file_operations, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300136 { .proc_show = show } )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137
Pavel Emelyanov640708a2012-01-10 15:11:23 -0800138static int proc_fd_permission(struct inode *inode, int mask);
139
Vegard Nossumaed54172008-06-05 22:46:53 -0700140/*
141 * Count the number of hardlinks for the pid_entry table, excluding the .
142 * and .. links.
143 */
144static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
145 unsigned int n)
146{
147 unsigned int i;
148 unsigned int count;
149
150 count = 0;
151 for (i = 0; i < n; ++i) {
152 if (S_ISDIR(entries[i].mode))
153 ++count;
154 }
155
156 return count;
157}
158
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200159static int get_task_root(struct task_struct *task, struct path *root)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160{
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000161 int result = -ENOENT;
162
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700163 task_lock(task);
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200164 if (task->fs) {
165 get_fs_root(task->fs, root);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000166 result = 0;
167 }
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700168 task_unlock(task);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000169 return result;
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700170}
171
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800172static int proc_cwd_link(struct dentry *dentry, struct path *path)
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700173{
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800174 struct task_struct *task = get_proc_task(dentry->d_inode);
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700175 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700176
177 if (task) {
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200178 task_lock(task);
179 if (task->fs) {
180 get_fs_pwd(task->fs, path);
181 result = 0;
182 }
183 task_unlock(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700184 put_task_struct(task);
185 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 return result;
187}
188
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800189static int proc_root_link(struct dentry *dentry, struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190{
Cyrill Gorcunov7773fbc2012-01-10 15:11:20 -0800191 struct task_struct *task = get_proc_task(dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700193
194 if (task) {
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200195 result = get_task_root(task, path);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700196 put_task_struct(task);
197 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 return result;
199}
200
Linus Torvaldse2683372012-01-17 15:21:19 -0800201struct mm_struct *mm_for_maps(struct task_struct *task)
202{
203 return mm_access(task, PTRACE_MODE_READ);
204}
205
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206static int proc_pid_cmdline(struct task_struct *task, char * buffer)
207{
208 int res = 0;
209 unsigned int len;
210 struct mm_struct *mm = get_task_mm(task);
211 if (!mm)
212 goto out;
213 if (!mm->arg_end)
214 goto out_mm; /* Shh! No looking before we're done */
215
216 len = mm->arg_end - mm->arg_start;
217
218 if (len > PAGE_SIZE)
219 len = PAGE_SIZE;
220
221 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
222
223 // If the nul at the end of args has been overwritten, then
224 // assume application is using setproctitle(3).
225 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
226 len = strnlen(buffer, res);
227 if (len < res) {
228 res = len;
229 } else {
230 len = mm->env_end - mm->env_start;
231 if (len > PAGE_SIZE - res)
232 len = PAGE_SIZE - res;
233 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
234 res = strnlen(buffer, res);
235 }
236 }
237out_mm:
238 mmput(mm);
239out:
240 return res;
241}
242
243static int proc_pid_auxv(struct task_struct *task, char *buffer)
244{
Al Viro2fadaef2011-02-15 22:52:11 -0500245 struct mm_struct *mm = mm_for_maps(task);
246 int res = PTR_ERR(mm);
247 if (mm && !IS_ERR(mm)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 unsigned int nwords = 0;
Hannes Ederdfe6b7d2008-12-30 18:49:13 +0300249 do {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 nwords += 2;
Hannes Ederdfe6b7d2008-12-30 18:49:13 +0300251 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 res = nwords * sizeof(mm->saved_auxv[0]);
253 if (res > PAGE_SIZE)
254 res = PAGE_SIZE;
255 memcpy(buffer, mm->saved_auxv, res);
256 mmput(mm);
257 }
258 return res;
259}
260
261
262#ifdef CONFIG_KALLSYMS
263/*
264 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
265 * Returns the resolved symbol. If that fails, simply return the address.
266 */
267static int proc_pid_wchan(struct task_struct *task, char *buffer)
268{
Alexey Dobriyanffb45122007-05-08 00:28:41 -0700269 unsigned long wchan;
Tejun Heo9281ace2007-07-17 04:03:51 -0700270 char symname[KSYM_NAME_LEN];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271
272 wchan = get_wchan(task);
273
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700274 if (lookup_symbol_name(wchan, symname) < 0)
Jake Edgef83ce3e2009-05-04 12:51:14 -0600275 if (!ptrace_may_access(task, PTRACE_MODE_READ))
276 return 0;
277 else
278 return sprintf(buffer, "%lu", wchan);
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700279 else
280 return sprintf(buffer, "%s", symname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281}
282#endif /* CONFIG_KALLSYMS */
283
Al Viroa9712bc2011-03-23 15:52:50 -0400284static int lock_trace(struct task_struct *task)
285{
286 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
287 if (err)
288 return err;
289 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
290 mutex_unlock(&task->signal->cred_guard_mutex);
291 return -EPERM;
292 }
293 return 0;
294}
295
296static void unlock_trace(struct task_struct *task)
297{
298 mutex_unlock(&task->signal->cred_guard_mutex);
299}
300
Ken Chen2ec220e2008-11-10 11:26:08 +0300301#ifdef CONFIG_STACKTRACE
302
303#define MAX_STACK_TRACE_DEPTH 64
304
305static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
306 struct pid *pid, struct task_struct *task)
307{
308 struct stack_trace trace;
309 unsigned long *entries;
Al Viroa9712bc2011-03-23 15:52:50 -0400310 int err;
Ken Chen2ec220e2008-11-10 11:26:08 +0300311 int i;
312
313 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
314 if (!entries)
315 return -ENOMEM;
316
317 trace.nr_entries = 0;
318 trace.max_entries = MAX_STACK_TRACE_DEPTH;
319 trace.entries = entries;
320 trace.skip = 0;
Ken Chen2ec220e2008-11-10 11:26:08 +0300321
Al Viroa9712bc2011-03-23 15:52:50 -0400322 err = lock_trace(task);
323 if (!err) {
324 save_stack_trace_tsk(task, &trace);
325
326 for (i = 0; i < trace.nr_entries; i++) {
Linus Torvaldsb81a6182011-03-23 20:51:42 -0700327 seq_printf(m, "[<%pK>] %pS\n",
Al Viroa9712bc2011-03-23 15:52:50 -0400328 (void *)entries[i], (void *)entries[i]);
329 }
330 unlock_trace(task);
Ken Chen2ec220e2008-11-10 11:26:08 +0300331 }
332 kfree(entries);
333
Al Viroa9712bc2011-03-23 15:52:50 -0400334 return err;
Ken Chen2ec220e2008-11-10 11:26:08 +0300335}
336#endif
337
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338#ifdef CONFIG_SCHEDSTATS
339/*
340 * Provides /proc/PID/schedstat
341 */
342static int proc_pid_schedstat(struct task_struct *task, char *buffer)
343{
Balbir Singh172ba842007-07-09 18:52:00 +0200344 return sprintf(buffer, "%llu %llu %lu\n",
Ingo Molnar826e08b2008-12-22 07:37:41 +0100345 (unsigned long long)task->se.sum_exec_runtime,
346 (unsigned long long)task->sched_info.run_delay,
Ingo Molnar2d723762007-10-15 17:00:12 +0200347 task->sched_info.pcount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348}
349#endif
350
Arjan van de Ven97455122008-01-25 21:08:34 +0100351#ifdef CONFIG_LATENCYTOP
352static int lstats_show_proc(struct seq_file *m, void *v)
353{
354 int i;
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800355 struct inode *inode = m->private;
356 struct task_struct *task = get_proc_task(inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100357
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800358 if (!task)
359 return -ESRCH;
360 seq_puts(m, "Latency Top version : v0.1\n");
Arjan van de Ven97455122008-01-25 21:08:34 +0100361 for (i = 0; i < 32; i++) {
Joe Perches34e49d42011-01-12 17:00:30 -0800362 struct latency_record *lr = &task->latency_record[i];
363 if (lr->backtrace[0]) {
Arjan van de Ven97455122008-01-25 21:08:34 +0100364 int q;
Joe Perches34e49d42011-01-12 17:00:30 -0800365 seq_printf(m, "%i %li %li",
366 lr->count, lr->time, lr->max);
Arjan van de Ven97455122008-01-25 21:08:34 +0100367 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
Joe Perches34e49d42011-01-12 17:00:30 -0800368 unsigned long bt = lr->backtrace[q];
369 if (!bt)
Arjan van de Ven97455122008-01-25 21:08:34 +0100370 break;
Joe Perches34e49d42011-01-12 17:00:30 -0800371 if (bt == ULONG_MAX)
Arjan van de Ven97455122008-01-25 21:08:34 +0100372 break;
Joe Perches34e49d42011-01-12 17:00:30 -0800373 seq_printf(m, " %ps", (void *)bt);
Arjan van de Ven97455122008-01-25 21:08:34 +0100374 }
Alexey Dobriyan9d6de122011-01-12 17:00:32 -0800375 seq_putc(m, '\n');
Arjan van de Ven97455122008-01-25 21:08:34 +0100376 }
377
378 }
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800379 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100380 return 0;
381}
382
383static int lstats_open(struct inode *inode, struct file *file)
384{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800385 return single_open(file, lstats_show_proc, inode);
Hiroshi Shimamotod6643d12008-02-14 10:27:00 -0800386}
387
Arjan van de Ven97455122008-01-25 21:08:34 +0100388static ssize_t lstats_write(struct file *file, const char __user *buf,
389 size_t count, loff_t *offs)
390{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800391 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100392
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800393 if (!task)
394 return -ESRCH;
Arjan van de Ven97455122008-01-25 21:08:34 +0100395 clear_all_latency_tracing(task);
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800396 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100397
398 return count;
399}
400
401static const struct file_operations proc_lstats_operations = {
402 .open = lstats_open,
403 .read = seq_read,
404 .write = lstats_write,
405 .llseek = seq_lseek,
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800406 .release = single_release,
Arjan van de Ven97455122008-01-25 21:08:34 +0100407};
408
409#endif
410
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411static int proc_oom_score(struct task_struct *task, char *buffer)
412{
Oleg Nesterovb95c35e2010-04-01 15:13:57 +0200413 unsigned long points = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700415 read_lock(&tasklist_lock);
Oleg Nesterovb95c35e2010-04-01 15:13:57 +0200416 if (pid_alive(task))
David Rientjesa63d83f2010-08-09 17:19:46 -0700417 points = oom_badness(task, NULL, NULL,
418 totalram_pages + total_swap_pages);
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700419 read_unlock(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 return sprintf(buffer, "%lu\n", points);
421}
422
Neil Hormand85f50d2007-10-18 23:40:37 -0700423struct limit_names {
424 char *name;
425 char *unit;
426};
427
428static const struct limit_names lnames[RLIM_NLIMITS] = {
Kees Cookcff4edb2009-09-22 16:45:32 -0700429 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700430 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
431 [RLIMIT_DATA] = {"Max data size", "bytes"},
432 [RLIMIT_STACK] = {"Max stack size", "bytes"},
433 [RLIMIT_CORE] = {"Max core file size", "bytes"},
434 [RLIMIT_RSS] = {"Max resident set", "bytes"},
435 [RLIMIT_NPROC] = {"Max processes", "processes"},
436 [RLIMIT_NOFILE] = {"Max open files", "files"},
437 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
438 [RLIMIT_AS] = {"Max address space", "bytes"},
439 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
440 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
441 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
442 [RLIMIT_NICE] = {"Max nice priority", NULL},
443 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
Eugene Teo88081172008-02-23 15:23:52 -0800444 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700445};
446
447/* Display limits for a process */
448static int proc_pid_limits(struct task_struct *task, char *buffer)
449{
450 unsigned int i;
451 int count = 0;
452 unsigned long flags;
453 char *bufptr = buffer;
454
455 struct rlimit rlim[RLIM_NLIMITS];
456
Lai Jiangshana6bebbc2008-10-05 00:51:15 +0400457 if (!lock_task_sighand(task, &flags))
Neil Hormand85f50d2007-10-18 23:40:37 -0700458 return 0;
Neil Hormand85f50d2007-10-18 23:40:37 -0700459 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
460 unlock_task_sighand(task, &flags);
Neil Hormand85f50d2007-10-18 23:40:37 -0700461
462 /*
463 * print the file header
464 */
465 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
466 "Limit", "Soft Limit", "Hard Limit", "Units");
467
468 for (i = 0; i < RLIM_NLIMITS; i++) {
469 if (rlim[i].rlim_cur == RLIM_INFINITY)
470 count += sprintf(&bufptr[count], "%-25s %-20s ",
471 lnames[i].name, "unlimited");
472 else
473 count += sprintf(&bufptr[count], "%-25s %-20lu ",
474 lnames[i].name, rlim[i].rlim_cur);
475
476 if (rlim[i].rlim_max == RLIM_INFINITY)
477 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
478 else
479 count += sprintf(&bufptr[count], "%-20lu ",
480 rlim[i].rlim_max);
481
482 if (lnames[i].unit)
483 count += sprintf(&bufptr[count], "%-10s\n",
484 lnames[i].unit);
485 else
486 count += sprintf(&bufptr[count], "\n");
487 }
488
489 return count;
490}
491
Roland McGrathebcb6732008-07-25 19:46:00 -0700492#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
493static int proc_pid_syscall(struct task_struct *task, char *buffer)
494{
495 long nr;
496 unsigned long args[6], sp, pc;
Al Viroa9712bc2011-03-23 15:52:50 -0400497 int res = lock_trace(task);
498 if (res)
499 return res;
Roland McGrathebcb6732008-07-25 19:46:00 -0700500
501 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
Al Viroa9712bc2011-03-23 15:52:50 -0400502 res = sprintf(buffer, "running\n");
503 else if (nr < 0)
504 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
505 else
506 res = sprintf(buffer,
Roland McGrathebcb6732008-07-25 19:46:00 -0700507 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
508 nr,
509 args[0], args[1], args[2], args[3], args[4], args[5],
510 sp, pc);
Al Viroa9712bc2011-03-23 15:52:50 -0400511 unlock_trace(task);
512 return res;
Roland McGrathebcb6732008-07-25 19:46:00 -0700513}
514#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
515
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516/************************************************************************/
517/* Here the fs part begins */
518/************************************************************************/
519
520/* permission checks */
Eric W. Biederman778c1142006-06-26 00:25:58 -0700521static int proc_fd_access_allowed(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522{
Eric W. Biederman778c1142006-06-26 00:25:58 -0700523 struct task_struct *task;
524 int allowed = 0;
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700525 /* Allow access to a task's file descriptors if it is us or we
526 * may use ptrace attach to the process and find out that
527 * information.
Eric W. Biederman778c1142006-06-26 00:25:58 -0700528 */
529 task = get_proc_task(inode);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700530 if (task) {
Stephen Smalley006ebb42008-05-19 08:32:49 -0400531 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
Eric W. Biederman778c1142006-06-26 00:25:58 -0700532 put_task_struct(task);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700533 }
Eric W. Biederman778c1142006-06-26 00:25:58 -0700534 return allowed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535}
536
Eric W. Biederman6b4e3062010-03-07 16:41:34 -0800537int proc_setattr(struct dentry *dentry, struct iattr *attr)
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700538{
539 int error;
540 struct inode *inode = dentry->d_inode;
541
542 if (attr->ia_valid & ATTR_MODE)
543 return -EPERM;
544
545 error = inode_change_ok(inode, attr);
Christoph Hellwig10257742010-06-04 11:30:02 +0200546 if (error)
547 return error;
548
549 if ((attr->ia_valid & ATTR_SIZE) &&
550 attr->ia_size != i_size_read(inode)) {
551 error = vmtruncate(inode, attr->ia_size);
552 if (error)
553 return error;
554 }
555
556 setattr_copy(inode, attr);
557 mark_inode_dirty(inode);
558 return 0;
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700559}
560
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800561/*
562 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
563 * or euid/egid (for hide_pid_min=2)?
564 */
565static bool has_pid_permissions(struct pid_namespace *pid,
566 struct task_struct *task,
567 int hide_pid_min)
568{
569 if (pid->hide_pid < hide_pid_min)
570 return true;
571 if (in_group_p(pid->pid_gid))
572 return true;
573 return ptrace_may_access(task, PTRACE_MODE_READ);
574}
575
576
577static int proc_pid_permission(struct inode *inode, int mask)
578{
579 struct pid_namespace *pid = inode->i_sb->s_fs_info;
580 struct task_struct *task;
581 bool has_perms;
582
583 task = get_proc_task(inode);
Xiaotian Fenga2ef9902012-01-12 17:17:08 -0800584 if (!task)
585 return -ESRCH;
Vasiliy Kulikov04996802012-01-10 15:11:31 -0800586 has_perms = has_pid_permissions(pid, task, 1);
587 put_task_struct(task);
588
589 if (!has_perms) {
590 if (pid->hide_pid == 2) {
591 /*
592 * Let's make getdents(), stat(), and open()
593 * consistent with each other. If a process
594 * may not stat() a file, it shouldn't be seen
595 * in procfs at all.
596 */
597 return -ENOENT;
598 }
599
600 return -EPERM;
601 }
602 return generic_permission(inode, mask);
603}
604
605
606
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800607static const struct inode_operations proc_def_inode_operations = {
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700608 .setattr = proc_setattr,
609};
610
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
612
613static ssize_t proc_info_read(struct file * file, char __user * buf,
614 size_t count, loff_t *ppos)
615{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800616 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 unsigned long page;
618 ssize_t length;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700619 struct task_struct *task = get_proc_task(inode);
620
621 length = -ESRCH;
622 if (!task)
623 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624
625 if (count > PROC_BLOCK_SIZE)
626 count = PROC_BLOCK_SIZE;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700627
628 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700629 if (!(page = __get_free_page(GFP_TEMPORARY)))
Eric W. Biederman99f89552006-06-26 00:25:55 -0700630 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631
632 length = PROC_I(inode)->op.proc_read(task, (char*)page);
633
634 if (length >= 0)
635 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
636 free_page(page);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700637out:
638 put_task_struct(task);
639out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 return length;
641}
642
Arjan van de Ven00977a52007-02-12 00:55:34 -0800643static const struct file_operations proc_info_file_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 .read = proc_info_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +0100645 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646};
647
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800648static int proc_single_show(struct seq_file *m, void *v)
649{
650 struct inode *inode = m->private;
651 struct pid_namespace *ns;
652 struct pid *pid;
653 struct task_struct *task;
654 int ret;
655
656 ns = inode->i_sb->s_fs_info;
657 pid = proc_pid(inode);
658 task = get_pid_task(pid, PIDTYPE_PID);
659 if (!task)
660 return -ESRCH;
661
662 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
663
664 put_task_struct(task);
665 return ret;
666}
667
668static int proc_single_open(struct inode *inode, struct file *filp)
669{
Jovi Zhangc6a34052011-01-12 17:00:34 -0800670 return single_open(filp, proc_single_show, inode);
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800671}
672
673static const struct file_operations proc_single_file_operations = {
674 .open = proc_single_open,
675 .read = seq_read,
676 .llseek = seq_lseek,
677 .release = single_release,
678};
679
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680static int mem_open(struct inode* inode, struct file* file)
681{
Linus Torvaldse2683372012-01-17 15:21:19 -0800682 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
683 struct mm_struct *mm;
684
685 if (!task)
686 return -ESRCH;
687
688 mm = mm_access(task, PTRACE_MODE_ATTACH);
689 put_task_struct(task);
690
691 if (IS_ERR(mm))
692 return PTR_ERR(mm);
693
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100694 if (mm) {
695 /* ensure this mm_struct can't be freed */
696 atomic_inc(&mm->mm_count);
697 /* but do not pin its memory */
698 mmput(mm);
699 }
700
KAMEZAWA Hiroyuki4a3956c2010-10-01 14:20:22 -0700701 /* OK to pass negative loff_t, we can catch out-of-range */
702 file->f_mode |= FMODE_UNSIGNED_OFFSET;
Linus Torvaldse2683372012-01-17 15:21:19 -0800703 file->private_data = mm;
704
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 return 0;
706}
707
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100708static ssize_t mem_rw(struct file *file, char __user *buf,
709 size_t count, loff_t *ppos, int write)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710{
Linus Torvaldse2683372012-01-17 15:21:19 -0800711 struct mm_struct *mm = file->private_data;
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100712 unsigned long addr = *ppos;
713 ssize_t copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 char *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715
Linus Torvaldse2683372012-01-17 15:21:19 -0800716 if (!mm)
717 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718
Mel Gormane12ba742007-10-16 01:25:52 -0700719 page = (char *)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 if (!page)
Linus Torvaldse2683372012-01-17 15:21:19 -0800721 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700723 copied = 0;
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100724 if (!atomic_inc_not_zero(&mm->mm_users))
725 goto free;
726
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 while (count > 0) {
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100728 int this_len = min_t(int, count, PAGE_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100730 if (write && copy_from_user(page, buf, this_len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 copied = -EFAULT;
732 break;
733 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100734
735 this_len = access_remote_vm(mm, addr, page, this_len, write);
736 if (!this_len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737 if (!copied)
738 copied = -EIO;
739 break;
740 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100741
742 if (!write && copy_to_user(buf, page, this_len)) {
743 copied = -EFAULT;
744 break;
745 }
746
747 buf += this_len;
748 addr += this_len;
749 copied += this_len;
750 count -= this_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 }
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100752 *ppos = addr;
KOSAKI Motohiro30cd8902011-05-26 16:25:52 -0700753
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100754 mmput(mm);
755free:
KOSAKI Motohiro30cd8902011-05-26 16:25:52 -0700756 free_page((unsigned long) page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 return copied;
758}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759
Oleg Nesterov572d34b2012-01-31 17:14:54 +0100760static ssize_t mem_read(struct file *file, char __user *buf,
761 size_t count, loff_t *ppos)
762{
763 return mem_rw(file, buf, count, ppos, 0);
764}
765
766static ssize_t mem_write(struct file *file, const char __user *buf,
767 size_t count, loff_t *ppos)
768{
769 return mem_rw(file, (char __user*)buf, count, ppos, 1);
770}
771
Matt Mackall85863e42008-02-04 22:29:04 -0800772loff_t mem_lseek(struct file *file, loff_t offset, int orig)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773{
774 switch (orig) {
775 case 0:
776 file->f_pos = offset;
777 break;
778 case 1:
779 file->f_pos += offset;
780 break;
781 default:
782 return -EINVAL;
783 }
784 force_successful_syscall_return();
785 return file->f_pos;
786}
787
Linus Torvaldse2683372012-01-17 15:21:19 -0800788static int mem_release(struct inode *inode, struct file *file)
789{
790 struct mm_struct *mm = file->private_data;
Oleg Nesterov71879d32012-01-31 17:14:38 +0100791 if (mm)
Oleg Nesterov6d08f2c2012-01-31 17:15:11 +0100792 mmdrop(mm);
Linus Torvaldse2683372012-01-17 15:21:19 -0800793 return 0;
794}
795
Arjan van de Ven00977a52007-02-12 00:55:34 -0800796static const struct file_operations proc_mem_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 .llseek = mem_lseek,
798 .read = mem_read,
799 .write = mem_write,
800 .open = mem_open,
Linus Torvaldse2683372012-01-17 15:21:19 -0800801 .release = mem_release,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802};
803
James Pearson315e28c2007-10-16 23:30:17 -0700804static ssize_t environ_read(struct file *file, char __user *buf,
805 size_t count, loff_t *ppos)
806{
807 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
808 char *page;
809 unsigned long src = *ppos;
810 int ret = -ESRCH;
811 struct mm_struct *mm;
812
813 if (!task)
814 goto out_no_task;
815
James Pearson315e28c2007-10-16 23:30:17 -0700816 ret = -ENOMEM;
817 page = (char *)__get_free_page(GFP_TEMPORARY);
818 if (!page)
819 goto out;
820
James Pearson315e28c2007-10-16 23:30:17 -0700821
Al Virod6f64b82011-02-15 22:26:01 -0500822 mm = mm_for_maps(task);
823 ret = PTR_ERR(mm);
824 if (!mm || IS_ERR(mm))
James Pearson315e28c2007-10-16 23:30:17 -0700825 goto out_free;
826
Al Virod6f64b82011-02-15 22:26:01 -0500827 ret = 0;
James Pearson315e28c2007-10-16 23:30:17 -0700828 while (count > 0) {
829 int this_len, retval, max_len;
830
831 this_len = mm->env_end - (mm->env_start + src);
832
833 if (this_len <= 0)
834 break;
835
836 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
837 this_len = (this_len > max_len) ? max_len : this_len;
838
839 retval = access_process_vm(task, (mm->env_start + src),
840 page, this_len, 0);
841
842 if (retval <= 0) {
843 ret = retval;
844 break;
845 }
846
847 if (copy_to_user(buf, page, retval)) {
848 ret = -EFAULT;
849 break;
850 }
851
852 ret += retval;
853 src += retval;
854 buf += retval;
855 count -= retval;
856 }
857 *ppos = src;
858
859 mmput(mm);
860out_free:
861 free_page((unsigned long) page);
862out:
863 put_task_struct(task);
864out_no_task:
865 return ret;
866}
867
868static const struct file_operations proc_environ_operations = {
869 .read = environ_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +0100870 .llseek = generic_file_llseek,
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();
1564 stat->uid = 0;
1565 stat->gid = 0;
1566 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 {
1625 inode->i_uid = 0;
1626 inode->i_gid = 0;
1627 }
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;
1756 if (FD_ISSET(fd, fdt->close_on_exec))
1757 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) {
1802 rcu_read_lock();
1803 if (fcheck_files(files, fd)) {
1804 rcu_read_unlock();
1805 put_files_struct(files);
1806 if (task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001807 rcu_read_lock();
1808 cred = __task_cred(task);
1809 inode->i_uid = cred->euid;
1810 inode->i_gid = cred->egid;
1811 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001812 } else {
1813 inode->i_uid = 0;
1814 inode->i_gid = 0;
1815 }
1816 inode->i_mode &= ~(S_ISUID | S_ISGID);
1817 security_task_to_inode(task, inode);
1818 put_task_struct(task);
1819 return 1;
1820 }
1821 rcu_read_unlock();
1822 put_files_struct(files);
1823 }
1824 put_task_struct(task);
1825 }
1826 d_drop(dentry);
1827 return 0;
1828}
1829
Al Virod72f71e2009-02-20 05:58:47 +00001830static const struct dentry_operations tid_fd_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001831{
1832 .d_revalidate = tid_fd_revalidate,
1833 .d_delete = pid_delete_dentry,
1834};
1835
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001836static struct dentry *proc_fd_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001837 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001838{
Eric Dumazetc5141e62007-05-08 00:26:15 -07001839 unsigned fd = *(const unsigned *)ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001840 struct file *file;
1841 struct files_struct *files;
1842 struct inode *inode;
1843 struct proc_inode *ei;
1844 struct dentry *error = ERR_PTR(-ENOENT);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001845
Eric W. Biederman61a28782006-10-02 02:18:49 -07001846 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001847 if (!inode)
1848 goto out;
1849 ei = PROC_I(inode);
1850 ei->fd = fd;
1851 files = get_files_struct(task);
1852 if (!files)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001853 goto out_iput;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001854 inode->i_mode = S_IFLNK;
1855
1856 /*
1857 * We are not taking a ref to the file structure, so we must
1858 * hold ->file_lock.
1859 */
1860 spin_lock(&files->file_lock);
1861 file = fcheck_files(files, fd);
1862 if (!file)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001863 goto out_unlock;
Al Viroaeb5d722008-09-02 15:28:45 -04001864 if (file->f_mode & FMODE_READ)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001865 inode->i_mode |= S_IRUSR | S_IXUSR;
Al Viroaeb5d722008-09-02 15:28:45 -04001866 if (file->f_mode & FMODE_WRITE)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001867 inode->i_mode |= S_IWUSR | S_IXUSR;
1868 spin_unlock(&files->file_lock);
1869 put_files_struct(files);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001870
Linus Torvalds5e442a42011-11-09 18:16:00 -05001871 inode->i_op = &proc_pid_link_inode_operations;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001872 inode->i_size = 64;
1873 ei->op.proc_get_link = proc_fd_link;
Nick Pigginfb045ad2011-01-07 17:49:55 +11001874 d_set_d_op(dentry, &tid_fd_dentry_operations);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001875 d_add(dentry, inode);
1876 /* Close the race of the process dying before we return the dentry */
1877 if (tid_fd_revalidate(dentry, NULL))
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001878 error = NULL;
1879
1880 out:
1881 return error;
1882out_unlock:
1883 spin_unlock(&files->file_lock);
1884 put_files_struct(files);
1885out_iput:
1886 iput(inode);
1887 goto out;
1888}
1889
Miklos Szeredi27932742007-05-08 00:26:17 -07001890static struct dentry *proc_lookupfd_common(struct inode *dir,
1891 struct dentry *dentry,
1892 instantiate_t instantiate)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001893{
1894 struct task_struct *task = get_proc_task(dir);
1895 unsigned fd = name_to_int(dentry);
1896 struct dentry *result = ERR_PTR(-ENOENT);
1897
1898 if (!task)
1899 goto out_no_task;
1900 if (fd == ~0U)
1901 goto out;
1902
Miklos Szeredi27932742007-05-08 00:26:17 -07001903 result = instantiate(dir, dentry, task, &fd);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001904out:
1905 put_task_struct(task);
1906out_no_task:
1907 return result;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001908}
1909
Miklos Szeredi27932742007-05-08 00:26:17 -07001910static int proc_readfd_common(struct file * filp, void * dirent,
1911 filldir_t filldir, instantiate_t instantiate)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001913 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman56347082006-06-26 00:25:40 -07001914 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001915 struct task_struct *p = get_proc_task(inode);
Pavel Emelyanov457c2512007-10-18 23:40:43 -07001916 unsigned int fd, ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001918 struct files_struct * files;
1919
1920 retval = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001921 if (!p)
1922 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924
1925 fd = filp->f_pos;
1926 switch (fd) {
1927 case 0:
1928 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
Linus Torvalds5e442a42011-11-09 18:16:00 -05001929 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930 filp->f_pos++;
1931 case 1:
Eric W. Biederman56347082006-06-26 00:25:40 -07001932 ino = parent_ino(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
Linus Torvalds5e442a42011-11-09 18:16:00 -05001934 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 filp->f_pos++;
1936 default:
1937 files = get_files_struct(p);
1938 if (!files)
Linus Torvalds5e442a42011-11-09 18:16:00 -05001939 goto out;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001940 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 for (fd = filp->f_pos-2;
Al Viro9b4f5262008-04-22 01:32:44 -04001942 fd < files_fdtable(files)->max_fds;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943 fd++, filp->f_pos++) {
Miklos Szeredi27932742007-05-08 00:26:17 -07001944 char name[PROC_NUMBUF];
1945 int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001946
1947 if (!fcheck_files(files, fd))
1948 continue;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001949 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950
Miklos Szeredi27932742007-05-08 00:26:17 -07001951 len = snprintf(name, sizeof(name), "%d", fd);
1952 if (proc_fill_cache(filp, dirent, filldir,
1953 name, len, instantiate,
1954 p, &fd) < 0) {
Dipankar Sarmab8359962005-09-09 13:04:14 -07001955 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956 break;
1957 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07001958 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001959 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07001960 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961 put_files_struct(files);
1962 }
1963out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07001964 put_task_struct(p);
1965out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966 return retval;
1967}
1968
Miklos Szeredi27932742007-05-08 00:26:17 -07001969static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1970 struct nameidata *nd)
1971{
1972 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1973}
1974
1975static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1976{
1977 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1978}
1979
1980static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1981 size_t len, loff_t *ppos)
1982{
1983 char tmp[PROC_FDINFO_MAX];
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001984 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
Miklos Szeredi27932742007-05-08 00:26:17 -07001985 if (!err)
1986 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1987 return err;
1988}
1989
1990static const struct file_operations proc_fdinfo_file_operations = {
Arnd Bergmann87df8422010-03-17 23:06:02 +01001991 .open = nonseekable_open,
Miklos Szeredi27932742007-05-08 00:26:17 -07001992 .read = proc_fdinfo_read,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001993 .llseek = no_llseek,
Miklos Szeredi27932742007-05-08 00:26:17 -07001994};
1995
Arjan van de Ven00977a52007-02-12 00:55:34 -08001996static const struct file_operations proc_fd_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997 .read = generic_read_dir,
1998 .readdir = proc_readfd,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001999 .llseek = default_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000};
2001
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002002#ifdef CONFIG_CHECKPOINT_RESTORE
2003
2004/*
2005 * dname_to_vma_addr - maps a dentry name into two unsigned longs
2006 * which represent vma start and end addresses.
2007 */
2008static int dname_to_vma_addr(struct dentry *dentry,
2009 unsigned long *start, unsigned long *end)
2010{
2011 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
2012 return -EINVAL;
2013
2014 return 0;
2015}
2016
2017static int map_files_d_revalidate(struct dentry *dentry, struct nameidata *nd)
2018{
2019 unsigned long vm_start, vm_end;
2020 bool exact_vma_exists = false;
2021 struct mm_struct *mm = NULL;
2022 struct task_struct *task;
2023 const struct cred *cred;
2024 struct inode *inode;
2025 int status = 0;
2026
2027 if (nd && nd->flags & LOOKUP_RCU)
2028 return -ECHILD;
2029
2030 if (!capable(CAP_SYS_ADMIN)) {
2031 status = -EACCES;
2032 goto out_notask;
2033 }
2034
2035 inode = dentry->d_inode;
2036 task = get_proc_task(inode);
2037 if (!task)
2038 goto out_notask;
2039
2040 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2041 goto out;
2042
2043 mm = get_task_mm(task);
2044 if (!mm)
2045 goto out;
2046
2047 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2048 down_read(&mm->mmap_sem);
2049 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
2050 up_read(&mm->mmap_sem);
2051 }
2052
2053 mmput(mm);
2054
2055 if (exact_vma_exists) {
2056 if (task_dumpable(task)) {
2057 rcu_read_lock();
2058 cred = __task_cred(task);
2059 inode->i_uid = cred->euid;
2060 inode->i_gid = cred->egid;
2061 rcu_read_unlock();
2062 } else {
2063 inode->i_uid = 0;
2064 inode->i_gid = 0;
2065 }
2066 security_task_to_inode(task, inode);
2067 status = 1;
2068 }
2069
2070out:
2071 put_task_struct(task);
2072
2073out_notask:
2074 if (status <= 0)
2075 d_drop(dentry);
2076
2077 return status;
2078}
2079
2080static const struct dentry_operations tid_map_files_dentry_operations = {
2081 .d_revalidate = map_files_d_revalidate,
2082 .d_delete = pid_delete_dentry,
2083};
2084
2085static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
2086{
2087 unsigned long vm_start, vm_end;
2088 struct vm_area_struct *vma;
2089 struct task_struct *task;
2090 struct mm_struct *mm;
2091 int rc;
2092
2093 rc = -ENOENT;
2094 task = get_proc_task(dentry->d_inode);
2095 if (!task)
2096 goto out;
2097
2098 mm = get_task_mm(task);
2099 put_task_struct(task);
2100 if (!mm)
2101 goto out;
2102
2103 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2104 if (rc)
2105 goto out_mmput;
2106
2107 down_read(&mm->mmap_sem);
2108 vma = find_exact_vma(mm, vm_start, vm_end);
2109 if (vma && vma->vm_file) {
2110 *path = vma->vm_file->f_path;
2111 path_get(path);
2112 rc = 0;
2113 }
2114 up_read(&mm->mmap_sem);
2115
2116out_mmput:
2117 mmput(mm);
2118out:
2119 return rc;
2120}
2121
2122struct map_files_info {
2123 struct file *file;
2124 unsigned long len;
2125 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2126};
2127
2128static struct dentry *
2129proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2130 struct task_struct *task, const void *ptr)
2131{
2132 const struct file *file = ptr;
2133 struct proc_inode *ei;
2134 struct inode *inode;
2135
2136 if (!file)
2137 return ERR_PTR(-ENOENT);
2138
2139 inode = proc_pid_make_inode(dir->i_sb, task);
2140 if (!inode)
2141 return ERR_PTR(-ENOENT);
2142
2143 ei = PROC_I(inode);
2144 ei->op.proc_get_link = proc_map_files_get_link;
2145
2146 inode->i_op = &proc_pid_link_inode_operations;
2147 inode->i_size = 64;
2148 inode->i_mode = S_IFLNK;
2149
2150 if (file->f_mode & FMODE_READ)
2151 inode->i_mode |= S_IRUSR;
2152 if (file->f_mode & FMODE_WRITE)
2153 inode->i_mode |= S_IWUSR;
2154
2155 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2156 d_add(dentry, inode);
2157
2158 return NULL;
2159}
2160
2161static struct dentry *proc_map_files_lookup(struct inode *dir,
2162 struct dentry *dentry, struct nameidata *nd)
2163{
2164 unsigned long vm_start, vm_end;
2165 struct vm_area_struct *vma;
2166 struct task_struct *task;
2167 struct dentry *result;
2168 struct mm_struct *mm;
2169
2170 result = ERR_PTR(-EACCES);
2171 if (!capable(CAP_SYS_ADMIN))
2172 goto out;
2173
2174 result = ERR_PTR(-ENOENT);
2175 task = get_proc_task(dir);
2176 if (!task)
2177 goto out;
2178
2179 result = ERR_PTR(-EACCES);
2180 if (lock_trace(task))
2181 goto out_put_task;
2182
2183 result = ERR_PTR(-ENOENT);
2184 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2185 goto out_unlock;
2186
2187 mm = get_task_mm(task);
2188 if (!mm)
2189 goto out_unlock;
2190
2191 down_read(&mm->mmap_sem);
2192 vma = find_exact_vma(mm, vm_start, vm_end);
2193 if (!vma)
2194 goto out_no_vma;
2195
2196 result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
2197
2198out_no_vma:
2199 up_read(&mm->mmap_sem);
2200 mmput(mm);
2201out_unlock:
2202 unlock_trace(task);
2203out_put_task:
2204 put_task_struct(task);
2205out:
2206 return result;
2207}
2208
2209static const struct inode_operations proc_map_files_inode_operations = {
2210 .lookup = proc_map_files_lookup,
2211 .permission = proc_fd_permission,
2212 .setattr = proc_setattr,
2213};
2214
2215static int
2216proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2217{
2218 struct dentry *dentry = filp->f_path.dentry;
2219 struct inode *inode = dentry->d_inode;
2220 struct vm_area_struct *vma;
2221 struct task_struct *task;
2222 struct mm_struct *mm;
2223 ino_t ino;
2224 int ret;
2225
2226 ret = -EACCES;
2227 if (!capable(CAP_SYS_ADMIN))
2228 goto out;
2229
2230 ret = -ENOENT;
2231 task = get_proc_task(inode);
2232 if (!task)
2233 goto out;
2234
2235 ret = -EACCES;
2236 if (lock_trace(task))
2237 goto out_put_task;
2238
2239 ret = 0;
2240 switch (filp->f_pos) {
2241 case 0:
2242 ino = inode->i_ino;
2243 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
2244 goto out_unlock;
2245 filp->f_pos++;
2246 case 1:
2247 ino = parent_ino(dentry);
2248 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2249 goto out_unlock;
2250 filp->f_pos++;
2251 default:
2252 {
2253 unsigned long nr_files, pos, i;
2254 struct flex_array *fa = NULL;
2255 struct map_files_info info;
2256 struct map_files_info *p;
2257
2258 mm = get_task_mm(task);
2259 if (!mm)
2260 goto out_unlock;
2261 down_read(&mm->mmap_sem);
2262
2263 nr_files = 0;
2264
2265 /*
2266 * We need two passes here:
2267 *
2268 * 1) Collect vmas of mapped files with mmap_sem taken
2269 * 2) Release mmap_sem and instantiate entries
2270 *
2271 * otherwise we get lockdep complained, since filldir()
2272 * routine might require mmap_sem taken in might_fault().
2273 */
2274
2275 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2276 if (vma->vm_file && ++pos > filp->f_pos)
2277 nr_files++;
2278 }
2279
2280 if (nr_files) {
2281 fa = flex_array_alloc(sizeof(info), nr_files,
2282 GFP_KERNEL);
2283 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2284 GFP_KERNEL)) {
2285 ret = -ENOMEM;
2286 if (fa)
2287 flex_array_free(fa);
2288 up_read(&mm->mmap_sem);
2289 mmput(mm);
2290 goto out_unlock;
2291 }
2292 for (i = 0, vma = mm->mmap, pos = 2; vma;
2293 vma = vma->vm_next) {
2294 if (!vma->vm_file)
2295 continue;
2296 if (++pos <= filp->f_pos)
2297 continue;
2298
2299 get_file(vma->vm_file);
2300 info.file = vma->vm_file;
2301 info.len = snprintf(info.name,
2302 sizeof(info.name), "%lx-%lx",
2303 vma->vm_start, vma->vm_end);
2304 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2305 BUG();
2306 }
2307 }
2308 up_read(&mm->mmap_sem);
2309
2310 for (i = 0; i < nr_files; i++) {
2311 p = flex_array_get(fa, i);
2312 ret = proc_fill_cache(filp, dirent, filldir,
2313 p->name, p->len,
2314 proc_map_files_instantiate,
2315 task, p->file);
2316 if (ret)
2317 break;
2318 filp->f_pos++;
2319 fput(p->file);
2320 }
2321 for (; i < nr_files; i++) {
2322 /*
2323 * In case of error don't forget
2324 * to put rest of file refs.
2325 */
2326 p = flex_array_get(fa, i);
2327 fput(p->file);
2328 }
2329 if (fa)
2330 flex_array_free(fa);
2331 mmput(mm);
2332 }
2333 }
2334
2335out_unlock:
2336 unlock_trace(task);
2337out_put_task:
2338 put_task_struct(task);
2339out:
2340 return ret;
2341}
2342
2343static const struct file_operations proc_map_files_operations = {
2344 .read = generic_read_dir,
2345 .readdir = proc_map_files_readdir,
2346 .llseek = default_llseek,
2347};
2348
2349#endif /* CONFIG_CHECKPOINT_RESTORE */
2350
Linus Torvalds1da177e2005-04-16 15:20:36 -07002351/*
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002352 * /proc/pid/fd needs a special permission handler so that a process can still
2353 * access /proc/self/fd after it has executed a setuid().
2354 */
Al Viro10556cb2011-06-20 19:28:19 -04002355static int proc_fd_permission(struct inode *inode, int mask)
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002356{
Al Viro2830ba72011-06-20 19:16:29 -04002357 int rv = generic_permission(inode, mask);
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002358 if (rv == 0)
2359 return 0;
2360 if (task_pid(current) == proc_pid(inode))
2361 rv = 0;
2362 return rv;
2363}
2364
2365/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002366 * proc directories can do almost nothing..
2367 */
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002368static const struct inode_operations proc_fd_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002369 .lookup = proc_lookupfd,
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002370 .permission = proc_fd_permission,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07002371 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002372};
2373
Miklos Szeredi27932742007-05-08 00:26:17 -07002374static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2375 struct dentry *dentry, struct task_struct *task, const void *ptr)
2376{
2377 unsigned fd = *(unsigned *)ptr;
2378 struct inode *inode;
2379 struct proc_inode *ei;
2380 struct dentry *error = ERR_PTR(-ENOENT);
2381
2382 inode = proc_pid_make_inode(dir->i_sb, task);
2383 if (!inode)
2384 goto out;
2385 ei = PROC_I(inode);
2386 ei->fd = fd;
2387 inode->i_mode = S_IFREG | S_IRUSR;
2388 inode->i_fop = &proc_fdinfo_file_operations;
Nick Pigginfb045ad2011-01-07 17:49:55 +11002389 d_set_d_op(dentry, &tid_fd_dentry_operations);
Miklos Szeredi27932742007-05-08 00:26:17 -07002390 d_add(dentry, inode);
2391 /* Close the race of the process dying before we return the dentry */
2392 if (tid_fd_revalidate(dentry, NULL))
2393 error = NULL;
2394
2395 out:
2396 return error;
2397}
2398
2399static struct dentry *proc_lookupfdinfo(struct inode *dir,
2400 struct dentry *dentry,
2401 struct nameidata *nd)
2402{
2403 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2404}
2405
2406static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2407{
2408 return proc_readfd_common(filp, dirent, filldir,
2409 proc_fdinfo_instantiate);
2410}
2411
2412static const struct file_operations proc_fdinfo_operations = {
2413 .read = generic_read_dir,
2414 .readdir = proc_readfdinfo,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002415 .llseek = default_llseek,
Miklos Szeredi27932742007-05-08 00:26:17 -07002416};
2417
2418/*
2419 * proc directories can do almost nothing..
2420 */
2421static const struct inode_operations proc_fdinfo_inode_operations = {
2422 .lookup = proc_lookupfdinfo,
2423 .setattr = proc_setattr,
2424};
2425
2426
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002427static struct dentry *proc_pident_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002428 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002429{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002430 const struct pid_entry *p = ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002431 struct inode *inode;
2432 struct proc_inode *ei;
KOSAKI Motohirobd6daba2009-05-28 14:34:21 -07002433 struct dentry *error = ERR_PTR(-ENOENT);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002434
Eric W. Biederman61a28782006-10-02 02:18:49 -07002435 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002436 if (!inode)
2437 goto out;
2438
2439 ei = PROC_I(inode);
2440 inode->i_mode = p->mode;
2441 if (S_ISDIR(inode->i_mode))
Miklos Szeredibfe86842011-10-28 14:13:29 +02002442 set_nlink(inode, 2); /* Use getattr to fix if necessary */
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002443 if (p->iop)
2444 inode->i_op = p->iop;
2445 if (p->fop)
2446 inode->i_fop = p->fop;
2447 ei->op = p->op;
Nick Pigginfb045ad2011-01-07 17:49:55 +11002448 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002449 d_add(dentry, inode);
2450 /* Close the race of the process dying before we return the dentry */
2451 if (pid_revalidate(dentry, NULL))
2452 error = NULL;
2453out:
2454 return error;
2455}
2456
Linus Torvalds1da177e2005-04-16 15:20:36 -07002457static struct dentry *proc_pident_lookup(struct inode *dir,
2458 struct dentry *dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002459 const struct pid_entry *ents,
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002460 unsigned int nents)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002461{
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002462 struct dentry *error;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002463 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002464 const struct pid_entry *p, *last;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002465
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002466 error = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002467
Eric W. Biederman99f89552006-06-26 00:25:55 -07002468 if (!task)
2469 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002470
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002471 /*
2472 * Yes, it does not scale. And it should not. Don't add
2473 * new entries into /proc/<tgid>/ without very good reasons.
2474 */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002475 last = &ents[nents - 1];
2476 for (p = ents; p <= last; p++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002477 if (p->len != dentry->d_name.len)
2478 continue;
2479 if (!memcmp(dentry->d_name.name, p->name, p->len))
2480 break;
2481 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002482 if (p > last)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002483 goto out;
2484
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002485 error = proc_pident_instantiate(dir, dentry, task, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002486out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07002487 put_task_struct(task);
2488out_no_task:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002489 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002490}
2491
Eric Dumazetc5141e62007-05-08 00:26:15 -07002492static int proc_pident_fill_cache(struct file *filp, void *dirent,
2493 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002494{
2495 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2496 proc_pident_instantiate, task, p);
2497}
2498
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002499static int proc_pident_readdir(struct file *filp,
2500 void *dirent, filldir_t filldir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002501 const struct pid_entry *ents, unsigned int nents)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002502{
2503 int i;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002504 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002505 struct inode *inode = dentry->d_inode;
2506 struct task_struct *task = get_proc_task(inode);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002507 const struct pid_entry *p, *last;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002508 ino_t ino;
2509 int ret;
2510
2511 ret = -ENOENT;
2512 if (!task)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002513 goto out_no_task;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002514
2515 ret = 0;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002516 i = filp->f_pos;
2517 switch (i) {
2518 case 0:
2519 ino = inode->i_ino;
2520 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2521 goto out;
2522 i++;
2523 filp->f_pos++;
2524 /* fall through */
2525 case 1:
2526 ino = parent_ino(dentry);
2527 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2528 goto out;
2529 i++;
2530 filp->f_pos++;
2531 /* fall through */
2532 default:
2533 i -= 2;
2534 if (i >= nents) {
2535 ret = 1;
2536 goto out;
2537 }
2538 p = ents + i;
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002539 last = &ents[nents - 1];
2540 while (p <= last) {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002541 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002542 goto out;
2543 filp->f_pos++;
2544 p++;
2545 }
2546 }
2547
2548 ret = 1;
2549out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07002550 put_task_struct(task);
2551out_no_task:
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002552 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002553}
2554
Linus Torvalds1da177e2005-04-16 15:20:36 -07002555#ifdef CONFIG_SECURITY
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002556static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2557 size_t count, loff_t *ppos)
2558{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002559 struct inode * inode = file->f_path.dentry->d_inode;
Al Viro04ff9702007-03-12 16:17:58 +00002560 char *p = NULL;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002561 ssize_t length;
2562 struct task_struct *task = get_proc_task(inode);
2563
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002564 if (!task)
Al Viro04ff9702007-03-12 16:17:58 +00002565 return -ESRCH;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002566
2567 length = security_getprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002568 (char*)file->f_path.dentry->d_name.name,
Al Viro04ff9702007-03-12 16:17:58 +00002569 &p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002570 put_task_struct(task);
Al Viro04ff9702007-03-12 16:17:58 +00002571 if (length > 0)
2572 length = simple_read_from_buffer(buf, count, ppos, p, length);
2573 kfree(p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002574 return length;
2575}
2576
2577static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2578 size_t count, loff_t *ppos)
2579{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002580 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002581 char *page;
2582 ssize_t length;
2583 struct task_struct *task = get_proc_task(inode);
2584
2585 length = -ESRCH;
2586 if (!task)
2587 goto out_no_task;
2588 if (count > PAGE_SIZE)
2589 count = PAGE_SIZE;
2590
2591 /* No partial writes. */
2592 length = -EINVAL;
2593 if (*ppos != 0)
2594 goto out;
2595
2596 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -07002597 page = (char*)__get_free_page(GFP_TEMPORARY);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002598 if (!page)
2599 goto out;
2600
2601 length = -EFAULT;
2602 if (copy_from_user(page, buf, count))
2603 goto out_free;
2604
David Howells107db7c2009-05-08 13:55:27 +01002605 /* Guard against adverse ptrace interaction */
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07002606 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
David Howells107db7c2009-05-08 13:55:27 +01002607 if (length < 0)
2608 goto out_free;
2609
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002610 length = security_setprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002611 (char*)file->f_path.dentry->d_name.name,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002612 (void*)page, count);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07002613 mutex_unlock(&task->signal->cred_guard_mutex);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002614out_free:
2615 free_page((unsigned long) page);
2616out:
2617 put_task_struct(task);
2618out_no_task:
2619 return length;
2620}
2621
Arjan van de Ven00977a52007-02-12 00:55:34 -08002622static const struct file_operations proc_pid_attr_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002623 .read = proc_pid_attr_read,
2624 .write = proc_pid_attr_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01002625 .llseek = generic_file_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002626};
2627
Eric Dumazetc5141e62007-05-08 00:26:15 -07002628static const struct pid_entry attr_dir_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002629 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2630 REG("prev", S_IRUGO, proc_pid_attr_operations),
2631 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2632 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2633 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2634 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002635};
2636
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002637static int proc_attr_dir_readdir(struct file * filp,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002638 void * dirent, filldir_t filldir)
2639{
2640 return proc_pident_readdir(filp,dirent,filldir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002641 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002642}
2643
Arjan van de Ven00977a52007-02-12 00:55:34 -08002644static const struct file_operations proc_attr_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002645 .read = generic_read_dir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002646 .readdir = proc_attr_dir_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002647 .llseek = default_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002648};
2649
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002650static struct dentry *proc_attr_dir_lookup(struct inode *dir,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002651 struct dentry *dentry, struct nameidata *nd)
2652{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002653 return proc_pident_lookup(dir, dentry,
2654 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002655}
2656
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002657static const struct inode_operations proc_attr_dir_inode_operations = {
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002658 .lookup = proc_attr_dir_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07002659 .getattr = pid_getattr,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07002660 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002661};
2662
Linus Torvalds1da177e2005-04-16 15:20:36 -07002663#endif
2664
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08002665#ifdef CONFIG_ELF_CORE
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002666static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2667 size_t count, loff_t *ppos)
2668{
2669 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2670 struct mm_struct *mm;
2671 char buffer[PROC_NUMBUF];
2672 size_t len;
2673 int ret;
2674
2675 if (!task)
2676 return -ESRCH;
2677
2678 ret = 0;
2679 mm = get_task_mm(task);
2680 if (mm) {
2681 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2682 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2683 MMF_DUMP_FILTER_SHIFT));
2684 mmput(mm);
2685 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2686 }
2687
2688 put_task_struct(task);
2689
2690 return ret;
2691}
2692
2693static ssize_t proc_coredump_filter_write(struct file *file,
2694 const char __user *buf,
2695 size_t count,
2696 loff_t *ppos)
2697{
2698 struct task_struct *task;
2699 struct mm_struct *mm;
2700 char buffer[PROC_NUMBUF], *end;
2701 unsigned int val;
2702 int ret;
2703 int i;
2704 unsigned long mask;
2705
2706 ret = -EFAULT;
2707 memset(buffer, 0, sizeof(buffer));
2708 if (count > sizeof(buffer) - 1)
2709 count = sizeof(buffer) - 1;
2710 if (copy_from_user(buffer, buf, count))
2711 goto out_no_task;
2712
2713 ret = -EINVAL;
2714 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2715 if (*end == '\n')
2716 end++;
2717 if (end - buffer == 0)
2718 goto out_no_task;
2719
2720 ret = -ESRCH;
2721 task = get_proc_task(file->f_dentry->d_inode);
2722 if (!task)
2723 goto out_no_task;
2724
2725 ret = end - buffer;
2726 mm = get_task_mm(task);
2727 if (!mm)
2728 goto out_no_mm;
2729
2730 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2731 if (val & mask)
2732 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2733 else
2734 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2735 }
2736
2737 mmput(mm);
2738 out_no_mm:
2739 put_task_struct(task);
2740 out_no_task:
2741 return ret;
2742}
2743
2744static const struct file_operations proc_coredump_filter_operations = {
2745 .read = proc_coredump_filter_read,
2746 .write = proc_coredump_filter_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01002747 .llseek = generic_file_llseek,
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002748};
2749#endif
2750
Linus Torvalds1da177e2005-04-16 15:20:36 -07002751/*
2752 * /proc/self:
2753 */
2754static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2755 int buflen)
2756{
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002757 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002758 pid_t tgid = task_tgid_nr_ns(current, ns);
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002759 char tmp[PROC_NUMBUF];
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002760 if (!tgid)
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002761 return -ENOENT;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002762 sprintf(tmp, "%d", tgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002763 return vfs_readlink(dentry,buffer,buflen,tmp);
2764}
2765
Al Viro008b1502005-08-20 00:17:39 +01002766static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002767{
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002768 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002769 pid_t tgid = task_tgid_nr_ns(current, ns);
Al Viro7fee4862010-01-14 01:03:28 -05002770 char *name = ERR_PTR(-ENOENT);
2771 if (tgid) {
2772 name = __getname();
2773 if (!name)
2774 name = ERR_PTR(-ENOMEM);
2775 else
2776 sprintf(name, "%d", tgid);
2777 }
2778 nd_set_link(nd, name);
2779 return NULL;
2780}
2781
2782static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2783 void *cookie)
2784{
2785 char *s = nd_get_link(nd);
2786 if (!IS_ERR(s))
2787 __putname(s);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002788}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002789
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002790static const struct inode_operations proc_self_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002791 .readlink = proc_self_readlink,
2792 .follow_link = proc_self_follow_link,
Al Viro7fee4862010-01-14 01:03:28 -05002793 .put_link = proc_self_put_link,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002794};
2795
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002796/*
Eric W. Biederman801199c2006-10-02 02:18:48 -07002797 * proc base
2798 *
2799 * These are the directory entries in the root directory of /proc
2800 * that properly belong to the /proc filesystem, as they describe
2801 * describe something that is process related.
2802 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07002803static const struct pid_entry proc_base_stuff[] = {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002804 NOD("self", S_IFLNK|S_IRWXUGO,
Eric W. Biederman801199c2006-10-02 02:18:48 -07002805 &proc_self_inode_operations, NULL, {}),
Eric W. Biederman801199c2006-10-02 02:18:48 -07002806};
2807
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002808static struct dentry *proc_base_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002809 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman801199c2006-10-02 02:18:48 -07002810{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002811 const struct pid_entry *p = ptr;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002812 struct inode *inode;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002813 struct proc_inode *ei;
Dan Carpenter73d36462010-05-26 14:43:25 -07002814 struct dentry *error;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002815
2816 /* Allocate the inode */
2817 error = ERR_PTR(-ENOMEM);
2818 inode = new_inode(dir->i_sb);
2819 if (!inode)
2820 goto out;
2821
2822 /* Initialize the inode */
2823 ei = PROC_I(inode);
Christoph Hellwig85fe4022010-10-23 11:19:54 -04002824 inode->i_ino = get_next_ino();
Eric W. Biederman801199c2006-10-02 02:18:48 -07002825 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002826
2827 /*
2828 * grab the reference to the task.
2829 */
Oleg Nesterov1a657f782006-10-02 02:18:59 -07002830 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002831 if (!ei->pid)
2832 goto out_iput;
2833
Eric W. Biederman801199c2006-10-02 02:18:48 -07002834 inode->i_mode = p->mode;
2835 if (S_ISDIR(inode->i_mode))
Miklos Szeredibfe86842011-10-28 14:13:29 +02002836 set_nlink(inode, 2);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002837 if (S_ISLNK(inode->i_mode))
2838 inode->i_size = 64;
2839 if (p->iop)
2840 inode->i_op = p->iop;
2841 if (p->fop)
2842 inode->i_fop = p->fop;
2843 ei->op = p->op;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002844 d_add(dentry, inode);
2845 error = NULL;
2846out:
Eric W. Biederman801199c2006-10-02 02:18:48 -07002847 return error;
2848out_iput:
2849 iput(inode);
2850 goto out;
2851}
2852
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002853static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2854{
2855 struct dentry *error;
2856 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002857 const struct pid_entry *p, *last;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002858
2859 error = ERR_PTR(-ENOENT);
2860
2861 if (!task)
2862 goto out_no_task;
2863
2864 /* Lookup the directory entry */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002865 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2866 for (p = proc_base_stuff; p <= last; p++) {
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002867 if (p->len != dentry->d_name.len)
2868 continue;
2869 if (!memcmp(dentry->d_name.name, p->name, p->len))
2870 break;
2871 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002872 if (p > last)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002873 goto out;
2874
2875 error = proc_base_instantiate(dir, dentry, task, p);
2876
2877out:
2878 put_task_struct(task);
2879out_no_task:
2880 return error;
2881}
2882
Eric Dumazetc5141e62007-05-08 00:26:15 -07002883static int proc_base_fill_cache(struct file *filp, void *dirent,
2884 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002885{
2886 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2887 proc_base_instantiate, task, p);
2888}
2889
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002890#ifdef CONFIG_TASK_IO_ACCOUNTING
Andrea Righi297c5d92008-07-25 01:48:49 -07002891static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002892{
Andrea Righi940389b2008-07-28 00:48:12 +02002893 struct task_io_accounting acct = task->ioac;
Andrea Righi59954772008-07-27 17:29:15 +02002894 unsigned long flags;
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002895 int result;
Andrea Righi297c5d92008-07-25 01:48:49 -07002896
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002897 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2898 if (result)
2899 return result;
2900
2901 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2902 result = -EACCES;
2903 goto out_unlock;
2904 }
Vasiliy Kulikov1d1221f2011-06-24 16:08:38 +04002905
Andrea Righi59954772008-07-27 17:29:15 +02002906 if (whole && lock_task_sighand(task, &flags)) {
2907 struct task_struct *t = task;
Andrea Righib2d002d2008-07-26 15:22:27 -07002908
Andrea Righi59954772008-07-27 17:29:15 +02002909 task_io_accounting_add(&acct, &task->signal->ioac);
2910 while_each_thread(task, t)
2911 task_io_accounting_add(&acct, &t->ioac);
Andrea Righi297c5d92008-07-25 01:48:49 -07002912
Andrea Righi59954772008-07-27 17:29:15 +02002913 unlock_task_sighand(task, &flags);
Andrea Righi297c5d92008-07-25 01:48:49 -07002914 }
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002915 result = sprintf(buffer,
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002916 "rchar: %llu\n"
2917 "wchar: %llu\n"
2918 "syscr: %llu\n"
2919 "syscw: %llu\n"
2920 "read_bytes: %llu\n"
2921 "write_bytes: %llu\n"
2922 "cancelled_write_bytes: %llu\n",
Alexander Beregalov7c443192008-08-05 13:01:34 -07002923 (unsigned long long)acct.rchar,
2924 (unsigned long long)acct.wchar,
2925 (unsigned long long)acct.syscr,
2926 (unsigned long long)acct.syscw,
2927 (unsigned long long)acct.read_bytes,
2928 (unsigned long long)acct.write_bytes,
2929 (unsigned long long)acct.cancelled_write_bytes);
Vasiliy Kulikov293eb1e2011-07-26 16:08:38 -07002930out_unlock:
2931 mutex_unlock(&task->signal->cred_guard_mutex);
2932 return result;
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002933}
Andrea Righi297c5d92008-07-25 01:48:49 -07002934
2935static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2936{
2937 return do_io_accounting(task, buffer, 0);
2938}
2939
2940static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2941{
2942 return do_io_accounting(task, buffer, 1);
2943}
2944#endif /* CONFIG_TASK_IO_ACCOUNTING */
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002945
Kees Cook47830722008-10-06 03:11:58 +04002946static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2947 struct pid *pid, struct task_struct *task)
2948{
Al Viroa9712bc2011-03-23 15:52:50 -04002949 int err = lock_trace(task);
2950 if (!err) {
2951 seq_printf(m, "%08x\n", task->personality);
2952 unlock_trace(task);
2953 }
2954 return err;
Kees Cook47830722008-10-06 03:11:58 +04002955}
2956
Eric W. Biederman801199c2006-10-02 02:18:48 -07002957/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002958 * Thread groups
2959 */
Arjan van de Ven00977a52007-02-12 00:55:34 -08002960static const struct file_operations proc_task_operations;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002961static const struct inode_operations proc_task_inode_operations;
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002962
Eric Dumazetc5141e62007-05-08 00:26:15 -07002963static const struct pid_entry tgid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002964 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2965 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
Pavel Emelyanov640708a2012-01-10 15:11:23 -08002966#ifdef CONFIG_CHECKPOINT_RESTORE
2967 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
2968#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002969 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08002970 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002971#ifdef CONFIG_NET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002972 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002973#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002974 REG("environ", S_IRUSR, proc_environ_operations),
2975 INF("auxv", S_IRUSR, proc_pid_auxv),
2976 ONE("status", S_IRUGO, proc_pid_status),
Al Viroa9712bc2011-03-23 15:52:50 -04002977 ONE("personality", S_IRUGO, proc_pid_personality),
Jiri Olsa3036e7b2010-09-30 15:15:33 -07002978 INF("limits", S_IRUGO, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002979#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002980 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002981#endif
Mike Galbraith5091faa2010-11-30 14:18:03 +01002982#ifdef CONFIG_SCHED_AUTOGROUP
2983 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
2984#endif
john stultz4614a696b2009-12-14 18:00:05 -08002985 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
Roland McGrathebcb6732008-07-25 19:46:00 -07002986#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Al Viroa9712bc2011-03-23 15:52:50 -04002987 INF("syscall", S_IRUGO, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07002988#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002989 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2990 ONE("stat", S_IRUGO, proc_tgid_stat),
2991 ONE("statm", S_IRUGO, proc_pid_statm),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07002992 REG("maps", S_IRUGO, proc_pid_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002993#ifdef CONFIG_NUMA
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07002994 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002995#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002996 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2997 LNK("cwd", proc_cwd_link),
2998 LNK("root", proc_root_link),
2999 LNK("exe", proc_exe_link),
3000 REG("mounts", S_IRUGO, proc_mounts_operations),
3001 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3002 REG("mountstats", S_IRUSR, proc_mountstats_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08003003#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003004 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07003005 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
Al Viroca6b0bf2011-02-15 22:04:37 -05003006 REG("pagemap", S_IRUGO, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003007#endif
3008#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003009 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003010#endif
3011#ifdef CONFIG_KALLSYMS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003012 INF("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003013#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03003014#ifdef CONFIG_STACKTRACE
Al Viroa9712bc2011-03-23 15:52:50 -04003015 ONE("stack", S_IRUGO, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003016#endif
3017#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003018 INF("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003019#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01003020#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003021 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01003022#endif
Paul Menage8793d852007-10-18 23:39:39 -07003023#ifdef CONFIG_PROC_PID_CPUSET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003024 REG("cpuset", S_IRUGO, proc_cpuset_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003025#endif
Paul Menagea4243162007-10-18 23:39:35 -07003026#ifdef CONFIG_CGROUPS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003027 REG("cgroup", S_IRUGO, proc_cgroup_operations),
Paul Menagea4243162007-10-18 23:39:35 -07003028#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003029 INF("oom_score", S_IRUGO, proc_oom_score),
3030 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
David Rientjesa63d83f2010-08-09 17:19:46 -07003031 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003032#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003033 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3034 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003035#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003036#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003037 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003038#endif
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08003039#ifdef CONFIG_ELF_CORE
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003040 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07003041#endif
Andrew Mortonaba76fd2006-12-10 02:19:48 -08003042#ifdef CONFIG_TASK_IO_ACCOUNTING
Vasiliy Kulikov1d1221f2011-06-24 16:08:38 +04003043 INF("io", S_IRUSR, proc_tgid_io_accounting),
Andrew Mortonaba76fd2006-12-10 02:19:48 -08003044#endif
Chris Metcalff133ecc2011-05-26 12:40:09 -04003045#ifdef CONFIG_HARDWALL
3046 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3047#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003048};
3049
3050static int proc_tgid_base_readdir(struct file * filp,
3051 void * dirent, filldir_t filldir)
3052{
3053 return proc_pident_readdir(filp,dirent,filldir,
3054 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
3055}
3056
Arjan van de Ven00977a52007-02-12 00:55:34 -08003057static const struct file_operations proc_tgid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003058 .read = generic_read_dir,
3059 .readdir = proc_tgid_base_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02003060 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003061};
3062
3063static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003064 return proc_pident_lookup(dir, dentry,
3065 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003066}
3067
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003068static const struct inode_operations proc_tgid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003069 .lookup = proc_tgid_base_lookup,
3070 .getattr = pid_getattr,
3071 .setattr = proc_setattr,
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003072 .permission = proc_pid_permission,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003073};
3074
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003075static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003076{
Eric W. Biederman48e64842006-06-26 00:25:48 -07003077 struct dentry *dentry, *leader, *dir;
Eric W. Biederman8578cea2006-06-26 00:25:54 -07003078 char buf[PROC_NUMBUF];
Eric W. Biederman48e64842006-06-26 00:25:48 -07003079 struct qstr name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003080
Eric W. Biederman48e64842006-06-26 00:25:48 -07003081 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003082 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3083 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003084 if (dentry) {
Sukadev Bhattiprolu29f12ca2009-11-11 14:26:32 -08003085 shrink_dcache_parent(dentry);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003086 d_drop(dentry);
3087 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003088 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003089
Eric W. Biederman48e64842006-06-26 00:25:48 -07003090 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003091 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3092 leader = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003093 if (!leader)
3094 goto out;
3095
3096 name.name = "task";
3097 name.len = strlen(name.name);
3098 dir = d_hash_and_lookup(leader, &name);
3099 if (!dir)
3100 goto out_put_leader;
3101
3102 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003103 name.len = snprintf(buf, sizeof(buf), "%d", pid);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003104 dentry = d_hash_and_lookup(dir, &name);
3105 if (dentry) {
3106 shrink_dcache_parent(dentry);
3107 d_drop(dentry);
3108 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003109 }
Eric W. Biederman48e64842006-06-26 00:25:48 -07003110
3111 dput(dir);
3112out_put_leader:
3113 dput(leader);
3114out:
3115 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003116}
3117
Randy Dunlap0895e912007-10-21 21:00:10 -07003118/**
3119 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3120 * @task: task that should be flushed.
3121 *
3122 * When flushing dentries from proc, one needs to flush them from global
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003123 * proc (proc_mnt) and from all the namespaces' procs this task was seen
Randy Dunlap0895e912007-10-21 21:00:10 -07003124 * in. This call is supposed to do all of this job.
3125 *
3126 * Looks in the dcache for
3127 * /proc/@pid
3128 * /proc/@tgid/task/@pid
3129 * if either directory is present flushes it and all of it'ts children
3130 * from the dcache.
3131 *
3132 * It is safe and reasonable to cache /proc entries for a task until
3133 * that task exits. After that they just clog up the dcache with
3134 * useless entries, possibly causing useful dcache entries to be
3135 * flushed instead. This routine is proved to flush those useless
3136 * dcache entries at process exit time.
3137 *
3138 * NOTE: This routine is just an optimization so it does not guarantee
3139 * that no dcache entries will exist at process exit time it
3140 * just makes it very unlikely that any will persist.
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003141 */
3142
3143void proc_flush_task(struct task_struct *task)
3144{
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08003145 int i;
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07003146 struct pid *pid, *tgid;
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003147 struct upid *upid;
3148
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003149 pid = task_pid(task);
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07003150 tgid = task_tgid(task);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003151
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08003152 for (i = 0; i <= pid->level; i++) {
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003153 upid = &pid->numbers[i];
3154 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07003155 tgid->numbers[i].nr);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07003156 }
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -07003157
3158 upid = &pid->numbers[pid->level];
3159 if (upid->nr == 1)
3160 pid_ns_release_proc(upid->ns);
Pavel Emelyanov60347f62007-10-18 23:40:03 -07003161}
3162
Adrian Bunk9711ef92006-12-06 20:38:31 -08003163static struct dentry *proc_pid_instantiate(struct inode *dir,
3164 struct dentry * dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07003165 struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003166{
3167 struct dentry *error = ERR_PTR(-ENOENT);
3168 struct inode *inode;
3169
Eric W. Biederman61a28782006-10-02 02:18:49 -07003170 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003171 if (!inode)
3172 goto out;
3173
3174 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3175 inode->i_op = &proc_tgid_base_inode_operations;
3176 inode->i_fop = &proc_tgid_base_operations;
3177 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07003178
Miklos Szeredibfe86842011-10-28 14:13:29 +02003179 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3180 ARRAY_SIZE(tgid_base_stuff)));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003181
Nick Pigginfb045ad2011-01-07 17:49:55 +11003182 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003183
3184 d_add(dentry, inode);
3185 /* Close the race of the process dying before we return the dentry */
3186 if (pid_revalidate(dentry, NULL))
3187 error = NULL;
3188out:
3189 return error;
3190}
3191
Linus Torvalds1da177e2005-04-16 15:20:36 -07003192struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3193{
Dan Carpenter73d36462010-05-26 14:43:25 -07003194 struct dentry *result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003195 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003196 unsigned tgid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003197 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003198
Eric W. Biederman801199c2006-10-02 02:18:48 -07003199 result = proc_base_lookup(dir, dentry);
3200 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3201 goto out;
3202
Linus Torvalds1da177e2005-04-16 15:20:36 -07003203 tgid = name_to_int(dentry);
3204 if (tgid == ~0U)
3205 goto out;
3206
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003207 ns = dentry->d_sb->s_fs_info;
Eric W. Biedermande758732006-06-26 00:25:51 -07003208 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003209 task = find_task_by_pid_ns(tgid, ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003210 if (task)
3211 get_task_struct(task);
Eric W. Biedermande758732006-06-26 00:25:51 -07003212 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003213 if (!task)
3214 goto out;
3215
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003216 result = proc_pid_instantiate(dir, dentry, task, NULL);
Eric W. Biederman48e64842006-06-26 00:25:48 -07003217 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003218out:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07003219 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003220}
3221
Linus Torvalds1da177e2005-04-16 15:20:36 -07003222/*
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003223 * Find the first task with tgid >= tgid
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003224 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07003225 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003226struct tgid_iter {
3227 unsigned int tgid;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003228 struct task_struct *task;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003229};
3230static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3231{
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003232 struct pid *pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003233
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003234 if (iter.task)
3235 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003236 rcu_read_lock();
3237retry:
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003238 iter.task = NULL;
3239 pid = find_ge_pid(iter.tgid, ns);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003240 if (pid) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003241 iter.tgid = pid_nr_ns(pid, ns);
3242 iter.task = pid_task(pid, PIDTYPE_PID);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003243 /* What we to know is if the pid we have find is the
3244 * pid of a thread_group_leader. Testing for task
3245 * being a thread_group_leader is the obvious thing
3246 * todo but there is a window when it fails, due to
3247 * the pid transfer logic in de_thread.
3248 *
3249 * So we perform the straight forward test of seeing
3250 * if the pid we have found is the pid of a thread
3251 * group leader, and don't worry if the task we have
3252 * found doesn't happen to be a thread group leader.
3253 * As we don't care in the case of readdir.
3254 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003255 if (!iter.task || !has_group_leader_pid(iter.task)) {
3256 iter.tgid += 1;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003257 goto retry;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003258 }
3259 get_task_struct(iter.task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003260 }
Eric W. Biederman454cc102006-06-26 00:25:51 -07003261 rcu_read_unlock();
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003262 return iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003263}
3264
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003265#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003266
Eric W. Biederman61a28782006-10-02 02:18:49 -07003267static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003268 struct tgid_iter iter)
Eric W. Biederman61a28782006-10-02 02:18:49 -07003269{
3270 char name[PROC_NUMBUF];
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003271 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003272 return proc_fill_cache(filp, dirent, filldir, name, len,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003273 proc_pid_instantiate, iter.task, NULL);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003274}
3275
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003276static int fake_filldir(void *buf, const char *name, int namelen,
3277 loff_t offset, u64 ino, unsigned d_type)
3278{
3279 return 0;
3280}
3281
Linus Torvalds1da177e2005-04-16 15:20:36 -07003282/* for the /proc/ directory itself, after non-process stuff has been done */
3283int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3284{
Linus Torvaldsd8bdc592011-04-18 10:36:54 -07003285 unsigned int nr;
3286 struct task_struct *reaper;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003287 struct tgid_iter iter;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003288 struct pid_namespace *ns;
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003289 filldir_t __filldir;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003290
Linus Torvaldsd8bdc592011-04-18 10:36:54 -07003291 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3292 goto out_no_task;
3293 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3294
3295 reaper = get_proc_task(filp->f_path.dentry->d_inode);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003296 if (!reaper)
3297 goto out_no_task;
3298
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003299 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
Eric Dumazetc5141e62007-05-08 00:26:15 -07003300 const struct pid_entry *p = &proc_base_stuff[nr];
Eric W. Biederman61a28782006-10-02 02:18:49 -07003301 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
Eric W. Biederman801199c2006-10-02 02:18:48 -07003302 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003303 }
3304
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003305 ns = filp->f_dentry->d_sb->s_fs_info;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003306 iter.task = NULL;
3307 iter.tgid = filp->f_pos - TGID_OFFSET;
3308 for (iter = next_tgid(ns, iter);
3309 iter.task;
3310 iter.tgid += 1, iter = next_tgid(ns, iter)) {
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003311 if (has_pid_permissions(ns, iter.task, 2))
3312 __filldir = filldir;
3313 else
3314 __filldir = fake_filldir;
3315
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003316 filp->f_pos = iter.tgid + TGID_OFFSET;
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003317 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003318 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003319 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003320 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003321 }
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003322 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3323out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07003324 put_task_struct(reaper);
3325out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07003326 return 0;
3327}
3328
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003329/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003330 * Tasks
3331 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07003332static const struct pid_entry tid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003333 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
Jerome Marchand38355412010-04-27 13:13:06 -07003334 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Eric W. Biederman6b4e3062010-03-07 16:41:34 -08003335 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003336 REG("environ", S_IRUSR, proc_environ_operations),
3337 INF("auxv", S_IRUSR, proc_pid_auxv),
3338 ONE("status", S_IRUGO, proc_pid_status),
Al Viroa9712bc2011-03-23 15:52:50 -04003339 ONE("personality", S_IRUGO, proc_pid_personality),
Jiri Olsa3036e7b2010-09-30 15:15:33 -07003340 INF("limits", S_IRUGO, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02003341#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003342 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02003343#endif
john stultz4614a696b2009-12-14 18:00:05 -08003344 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
Roland McGrathebcb6732008-07-25 19:46:00 -07003345#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Al Viroa9712bc2011-03-23 15:52:50 -04003346 INF("syscall", S_IRUGO, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07003347#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003348 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3349 ONE("stat", S_IRUGO, proc_tid_stat),
3350 ONE("statm", S_IRUGO, proc_pid_statm),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07003351 REG("maps", S_IRUGO, proc_tid_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003352#ifdef CONFIG_NUMA
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07003353 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003354#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003355 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3356 LNK("cwd", proc_cwd_link),
3357 LNK("root", proc_root_link),
3358 LNK("exe", proc_exe_link),
3359 REG("mounts", S_IRUGO, proc_mounts_operations),
3360 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08003361#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003362 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
Siddhesh Poyarekarb7643752012-03-21 16:34:04 -07003363 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
Al Viroca6b0bf2011-02-15 22:04:37 -05003364 REG("pagemap", S_IRUGO, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003365#endif
3366#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003367 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003368#endif
3369#ifdef CONFIG_KALLSYMS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003370 INF("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003371#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03003372#ifdef CONFIG_STACKTRACE
Al Viroa9712bc2011-03-23 15:52:50 -04003373 ONE("stack", S_IRUGO, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003374#endif
3375#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003376 INF("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003377#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01003378#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003379 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01003380#endif
Paul Menage8793d852007-10-18 23:39:39 -07003381#ifdef CONFIG_PROC_PID_CPUSET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003382 REG("cpuset", S_IRUGO, proc_cpuset_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003383#endif
Paul Menagea4243162007-10-18 23:39:35 -07003384#ifdef CONFIG_CGROUPS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003385 REG("cgroup", S_IRUGO, proc_cgroup_operations),
Paul Menagea4243162007-10-18 23:39:35 -07003386#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003387 INF("oom_score", S_IRUGO, proc_oom_score),
3388 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
David Rientjesa63d83f2010-08-09 17:19:46 -07003389 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003390#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003391 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
Al Viro26ec3c62011-02-15 21:24:05 -05003392 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003393#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003394#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003395 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003396#endif
Andrea Righi297c5d92008-07-25 01:48:49 -07003397#ifdef CONFIG_TASK_IO_ACCOUNTING
Vasiliy Kulikov1d1221f2011-06-24 16:08:38 +04003398 INF("io", S_IRUSR, proc_tid_io_accounting),
Andrea Righi297c5d92008-07-25 01:48:49 -07003399#endif
Chris Metcalff133ecc2011-05-26 12:40:09 -04003400#ifdef CONFIG_HARDWALL
3401 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3402#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003403};
3404
3405static int proc_tid_base_readdir(struct file * filp,
3406 void * dirent, filldir_t filldir)
3407{
3408 return proc_pident_readdir(filp,dirent,filldir,
3409 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3410}
3411
3412static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003413 return proc_pident_lookup(dir, dentry,
3414 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003415}
3416
Arjan van de Ven00977a52007-02-12 00:55:34 -08003417static const struct file_operations proc_tid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003418 .read = generic_read_dir,
3419 .readdir = proc_tid_base_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02003420 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003421};
3422
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003423static const struct inode_operations proc_tid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003424 .lookup = proc_tid_base_lookup,
3425 .getattr = pid_getattr,
3426 .setattr = proc_setattr,
3427};
3428
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003429static struct dentry *proc_task_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07003430 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003431{
3432 struct dentry *error = ERR_PTR(-ENOENT);
3433 struct inode *inode;
Eric W. Biederman61a28782006-10-02 02:18:49 -07003434 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003435
3436 if (!inode)
3437 goto out;
3438 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3439 inode->i_op = &proc_tid_base_inode_operations;
3440 inode->i_fop = &proc_tid_base_operations;
3441 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07003442
Miklos Szeredibfe86842011-10-28 14:13:29 +02003443 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3444 ARRAY_SIZE(tid_base_stuff)));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003445
Nick Pigginfb045ad2011-01-07 17:49:55 +11003446 d_set_d_op(dentry, &pid_dentry_operations);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003447
3448 d_add(dentry, inode);
3449 /* Close the race of the process dying before we return the dentry */
3450 if (pid_revalidate(dentry, NULL))
3451 error = NULL;
3452out:
3453 return error;
3454}
3455
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003456static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3457{
3458 struct dentry *result = ERR_PTR(-ENOENT);
3459 struct task_struct *task;
3460 struct task_struct *leader = get_proc_task(dir);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003461 unsigned tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003462 struct pid_namespace *ns;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003463
3464 if (!leader)
3465 goto out_no_task;
3466
3467 tid = name_to_int(dentry);
3468 if (tid == ~0U)
3469 goto out;
3470
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003471 ns = dentry->d_sb->s_fs_info;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003472 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003473 task = find_task_by_pid_ns(tid, ns);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003474 if (task)
3475 get_task_struct(task);
3476 rcu_read_unlock();
3477 if (!task)
3478 goto out;
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07003479 if (!same_thread_group(leader, task))
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003480 goto out_drop_task;
3481
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003482 result = proc_task_instantiate(dir, dentry, task, NULL);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003483out_drop_task:
3484 put_task_struct(task);
3485out:
3486 put_task_struct(leader);
3487out_no_task:
3488 return result;
3489}
3490
3491/*
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003492 * Find the first tid of a thread group to return to user space.
3493 *
3494 * Usually this is just the thread group leader, but if the users
3495 * buffer was too small or there was a seek into the middle of the
3496 * directory we have more work todo.
3497 *
3498 * In the case of a short read we start with find_task_by_pid.
3499 *
3500 * In the case of a seek we start with the leader and walk nr
3501 * threads past it.
3502 */
Eric W. Biedermancc288732006-06-26 00:26:01 -07003503static struct task_struct *first_tid(struct task_struct *leader,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003504 int tid, int nr, struct pid_namespace *ns)
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003505{
Oleg Nesterova872ff02006-06-26 00:26:01 -07003506 struct task_struct *pos;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003507
Eric W. Biedermancc288732006-06-26 00:26:01 -07003508 rcu_read_lock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003509 /* Attempt to start with the pid of a thread */
3510 if (tid && (nr > 0)) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003511 pos = find_task_by_pid_ns(tid, ns);
Oleg Nesterova872ff02006-06-26 00:26:01 -07003512 if (pos && (pos->group_leader == leader))
3513 goto found;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003514 }
3515
3516 /* If nr exceeds the number of threads there is nothing todo */
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003517 pos = NULL;
Oleg Nesterova872ff02006-06-26 00:26:01 -07003518 if (nr && nr >= get_nr_threads(leader))
3519 goto out;
3520
3521 /* If we haven't found our starting place yet start
3522 * with the leader and walk nr threads forward.
3523 */
3524 for (pos = leader; nr > 0; --nr) {
3525 pos = next_thread(pos);
3526 if (pos == leader) {
3527 pos = NULL;
3528 goto out;
3529 }
3530 }
3531found:
3532 get_task_struct(pos);
3533out:
Eric W. Biedermancc288732006-06-26 00:26:01 -07003534 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003535 return pos;
3536}
3537
3538/*
3539 * Find the next thread in the thread list.
3540 * Return NULL if there is an error or no next thread.
3541 *
3542 * The reference to the input task_struct is released.
3543 */
3544static struct task_struct *next_tid(struct task_struct *start)
3545{
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003546 struct task_struct *pos = NULL;
Eric W. Biedermancc288732006-06-26 00:26:01 -07003547 rcu_read_lock();
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003548 if (pid_alive(start)) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003549 pos = next_thread(start);
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003550 if (thread_group_leader(pos))
3551 pos = NULL;
3552 else
3553 get_task_struct(pos);
3554 }
Eric W. Biedermancc288732006-06-26 00:26:01 -07003555 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003556 put_task_struct(start);
3557 return pos;
3558}
3559
Eric W. Biederman61a28782006-10-02 02:18:49 -07003560static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3561 struct task_struct *task, int tid)
3562{
3563 char name[PROC_NUMBUF];
3564 int len = snprintf(name, sizeof(name), "%d", tid);
3565 return proc_fill_cache(filp, dirent, filldir, name, len,
3566 proc_task_instantiate, task, NULL);
3567}
3568
Linus Torvalds1da177e2005-04-16 15:20:36 -07003569/* for the /proc/TGID/task/ directories */
3570static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3571{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08003572 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003573 struct inode *inode = dentry->d_inode;
Guillaume Chazarain7d895242007-01-31 23:48:14 -08003574 struct task_struct *leader = NULL;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003575 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003576 int retval = -ENOENT;
3577 ino_t ino;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003578 int tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003579 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003580
Guillaume Chazarain7d895242007-01-31 23:48:14 -08003581 task = get_proc_task(inode);
3582 if (!task)
3583 goto out_no_task;
3584 rcu_read_lock();
3585 if (pid_alive(task)) {
3586 leader = task->group_leader;
3587 get_task_struct(leader);
3588 }
3589 rcu_read_unlock();
3590 put_task_struct(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003591 if (!leader)
3592 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003593 retval = 0;
3594
Linus Torvaldsee568b22009-03-17 10:02:35 -07003595 switch ((unsigned long)filp->f_pos) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003596 case 0:
3597 ino = inode->i_ino;
Zhang Leee6f7792009-03-16 14:44:31 +08003598 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003599 goto out;
Zhang Leee6f7792009-03-16 14:44:31 +08003600 filp->f_pos++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003601 /* fall through */
3602 case 1:
3603 ino = parent_ino(dentry);
Zhang Leee6f7792009-03-16 14:44:31 +08003604 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003605 goto out;
Zhang Leee6f7792009-03-16 14:44:31 +08003606 filp->f_pos++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003607 /* fall through */
3608 }
3609
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003610 /* f_version caches the tgid value that the last readdir call couldn't
3611 * return. lseek aka telldir automagically resets f_version to 0.
3612 */
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003613 ns = filp->f_dentry->d_sb->s_fs_info;
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07003614 tid = (int)filp->f_version;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003615 filp->f_version = 0;
Zhang Leee6f7792009-03-16 14:44:31 +08003616 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003617 task;
Zhang Leee6f7792009-03-16 14:44:31 +08003618 task = next_tid(task), filp->f_pos++) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003619 tid = task_pid_nr_ns(task, ns);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003620 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003621 /* returning this tgid failed, save it as the first
3622 * pid for the next readir call */
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07003623 filp->f_version = (u64)tid;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003624 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003625 break;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003626 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003627 }
3628out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07003629 put_task_struct(leader);
3630out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07003631 return retval;
3632}
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003633
3634static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3635{
3636 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07003637 struct task_struct *p = get_proc_task(inode);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003638 generic_fillattr(inode, stat);
3639
Eric W. Biederman99f89552006-06-26 00:25:55 -07003640 if (p) {
Eric W. Biederman99f89552006-06-26 00:25:55 -07003641 stat->nlink += get_nr_threads(p);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003642 put_task_struct(p);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003643 }
3644
3645 return 0;
3646}
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003647
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003648static const struct inode_operations proc_task_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003649 .lookup = proc_task_lookup,
3650 .getattr = proc_task_getattr,
3651 .setattr = proc_setattr,
Vasiliy Kulikov04996802012-01-10 15:11:31 -08003652 .permission = proc_pid_permission,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003653};
3654
Arjan van de Ven00977a52007-02-12 00:55:34 -08003655static const struct file_operations proc_task_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003656 .read = generic_read_dir,
3657 .readdir = proc_task_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02003658 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003659};