blob: f3d02ca461ecfcc341eb783fd15da8df486fde11 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070086#include "internal.h"
87
Eric W. Biederman0f2fe202006-06-26 00:25:46 -070088/* NOTE:
89 * Implementing inode permission operations in /proc is almost
90 * certainly an error. Permission checks need to happen during
91 * each system call not at open time. The reason is that most of
92 * what we wish to check for permissions in /proc varies at runtime.
93 *
94 * The classic example of a problem is opening file descriptors
95 * in /proc for a task before it execs a suid executable.
96 */
97
Linus Torvalds1da177e2005-04-16 15:20:36 -070098struct pid_entry {
Linus Torvalds1da177e2005-04-16 15:20:36 -070099 char *name;
Eric Dumazetc5141e62007-05-08 00:26:15 -0700100 int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101 mode_t mode;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800102 const struct inode_operations *iop;
Arjan van de Ven00977a52007-02-12 00:55:34 -0800103 const struct file_operations *fop;
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700104 union proc_op op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105};
106
Eric W. Biederman61a28782006-10-02 02:18:49 -0700107#define NOD(NAME, MODE, IOP, FOP, OP) { \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700108 .name = (NAME), \
Eric Dumazetc5141e62007-05-08 00:26:15 -0700109 .len = sizeof(NAME) - 1, \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700110 .mode = MODE, \
111 .iop = IOP, \
112 .fop = FOP, \
113 .op = OP, \
114}
115
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300116#define DIR(NAME, MODE, iops, fops) \
117 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
118#define LNK(NAME, get_link) \
Eric W. Biederman61a28782006-10-02 02:18:49 -0700119 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700120 &proc_pid_link_inode_operations, NULL, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300121 { .proc_get_link = get_link } )
122#define REG(NAME, MODE, fops) \
123 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
124#define INF(NAME, MODE, read) \
Eric W. Biederman61a28782006-10-02 02:18:49 -0700125 NOD(NAME, (S_IFREG|(MODE)), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700126 NULL, &proc_info_file_operations, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300127 { .proc_read = read } )
128#define ONE(NAME, MODE, show) \
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800129 NOD(NAME, (S_IFREG|(MODE)), \
130 NULL, &proc_single_file_operations, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300131 { .proc_show = show } )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132
Vegard Nossumaed54172008-06-05 22:46:53 -0700133/*
134 * Count the number of hardlinks for the pid_entry table, excluding the .
135 * and .. links.
136 */
137static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
138 unsigned int n)
139{
140 unsigned int i;
141 unsigned int count;
142
143 count = 0;
144 for (i = 0; i < n; ++i) {
145 if (S_ISDIR(entries[i].mode))
146 ++count;
147 }
148
149 return count;
150}
151
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200152static int get_task_root(struct task_struct *task, struct path *root)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153{
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000154 int result = -ENOENT;
155
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700156 task_lock(task);
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200157 if (task->fs) {
158 get_fs_root(task->fs, root);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000159 result = 0;
160 }
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700161 task_unlock(task);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000162 return result;
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700163}
164
Jan Blunck3dcd25f2008-02-14 19:38:35 -0800165static int proc_cwd_link(struct inode *inode, struct path *path)
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700166{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700167 struct task_struct *task = get_proc_task(inode);
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700168 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700169
170 if (task) {
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200171 task_lock(task);
172 if (task->fs) {
173 get_fs_pwd(task->fs, path);
174 result = 0;
175 }
176 task_unlock(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700177 put_task_struct(task);
178 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 return result;
180}
181
Jan Blunck3dcd25f2008-02-14 19:38:35 -0800182static int proc_root_link(struct inode *inode, struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700184 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700186
187 if (task) {
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200188 result = get_task_root(task, path);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700189 put_task_struct(task);
190 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 return result;
192}
193
Roland McGrath638fa2022008-04-29 01:01:38 -0700194/*
195 * Return zero if current may access user memory in @task, -error if not.
196 */
197static int check_mem_permission(struct task_struct *task)
198{
199 /*
200 * A task can always look at itself, in case it chooses
201 * to use system calls instead of load instructions.
202 */
203 if (task == current)
204 return 0;
205
206 /*
207 * If current is actively ptrace'ing, and would also be
208 * permitted to freshly attach with ptrace now, permit it.
209 */
Roland McGrath0d094ef2008-07-25 19:45:49 -0700210 if (task_is_stopped_or_traced(task)) {
211 int match;
212 rcu_read_lock();
213 match = (tracehook_tracer_task(task) == current);
214 rcu_read_unlock();
215 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
216 return 0;
217 }
Roland McGrath638fa2022008-04-29 01:01:38 -0700218
219 /*
220 * Noone else is allowed.
221 */
222 return -EPERM;
223}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224
Al Viro831830b2008-01-02 14:09:57 +0000225struct mm_struct *mm_for_maps(struct task_struct *task)
226{
Oleg Nesterov704b8362009-07-10 03:27:40 +0200227 struct mm_struct *mm;
Oleg Nesterov00f89d22009-07-10 03:27:38 +0200228
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -0700229 if (mutex_lock_killable(&task->signal->cred_guard_mutex))
Oleg Nesterov704b8362009-07-10 03:27:40 +0200230 return NULL;
231
232 mm = get_task_mm(task);
233 if (mm && mm != current->mm &&
234 !ptrace_may_access(task, PTRACE_MODE_READ)) {
235 mmput(mm);
236 mm = NULL;
Oleg Nesterov13f0fea2009-06-23 21:25:32 +0200237 }
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -0700238 mutex_unlock(&task->signal->cred_guard_mutex);
Oleg Nesterov704b8362009-07-10 03:27:40 +0200239
Al Viro831830b2008-01-02 14:09:57 +0000240 return mm;
Al Viro831830b2008-01-02 14:09:57 +0000241}
242
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243static int proc_pid_cmdline(struct task_struct *task, char * buffer)
244{
245 int res = 0;
246 unsigned int len;
247 struct mm_struct *mm = get_task_mm(task);
248 if (!mm)
249 goto out;
250 if (!mm->arg_end)
251 goto out_mm; /* Shh! No looking before we're done */
252
253 len = mm->arg_end - mm->arg_start;
254
255 if (len > PAGE_SIZE)
256 len = PAGE_SIZE;
257
258 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
259
260 // If the nul at the end of args has been overwritten, then
261 // assume application is using setproctitle(3).
262 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
263 len = strnlen(buffer, res);
264 if (len < res) {
265 res = len;
266 } else {
267 len = mm->env_end - mm->env_start;
268 if (len > PAGE_SIZE - res)
269 len = PAGE_SIZE - res;
270 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
271 res = strnlen(buffer, res);
272 }
273 }
274out_mm:
275 mmput(mm);
276out:
277 return res;
278}
279
280static int proc_pid_auxv(struct task_struct *task, char *buffer)
281{
282 int res = 0;
283 struct mm_struct *mm = get_task_mm(task);
284 if (mm) {
285 unsigned int nwords = 0;
Hannes Ederdfe6b7d2008-12-30 18:49:13 +0300286 do {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 nwords += 2;
Hannes Ederdfe6b7d2008-12-30 18:49:13 +0300288 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 res = nwords * sizeof(mm->saved_auxv[0]);
290 if (res > PAGE_SIZE)
291 res = PAGE_SIZE;
292 memcpy(buffer, mm->saved_auxv, res);
293 mmput(mm);
294 }
295 return res;
296}
297
298
299#ifdef CONFIG_KALLSYMS
300/*
301 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
302 * Returns the resolved symbol. If that fails, simply return the address.
303 */
304static int proc_pid_wchan(struct task_struct *task, char *buffer)
305{
Alexey Dobriyanffb45122007-05-08 00:28:41 -0700306 unsigned long wchan;
Tejun Heo9281ace2007-07-17 04:03:51 -0700307 char symname[KSYM_NAME_LEN];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
309 wchan = get_wchan(task);
310
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700311 if (lookup_symbol_name(wchan, symname) < 0)
Jake Edgef83ce3e2009-05-04 12:51:14 -0600312 if (!ptrace_may_access(task, PTRACE_MODE_READ))
313 return 0;
314 else
315 return sprintf(buffer, "%lu", wchan);
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700316 else
317 return sprintf(buffer, "%s", symname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318}
319#endif /* CONFIG_KALLSYMS */
320
Ken Chen2ec220e2008-11-10 11:26:08 +0300321#ifdef CONFIG_STACKTRACE
322
323#define MAX_STACK_TRACE_DEPTH 64
324
325static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
326 struct pid *pid, struct task_struct *task)
327{
328 struct stack_trace trace;
329 unsigned long *entries;
330 int i;
331
332 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
333 if (!entries)
334 return -ENOMEM;
335
336 trace.nr_entries = 0;
337 trace.max_entries = MAX_STACK_TRACE_DEPTH;
338 trace.entries = entries;
339 trace.skip = 0;
340 save_stack_trace_tsk(task, &trace);
341
342 for (i = 0; i < trace.nr_entries; i++) {
343 seq_printf(m, "[<%p>] %pS\n",
344 (void *)entries[i], (void *)entries[i]);
345 }
346 kfree(entries);
347
348 return 0;
349}
350#endif
351
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352#ifdef CONFIG_SCHEDSTATS
353/*
354 * Provides /proc/PID/schedstat
355 */
356static int proc_pid_schedstat(struct task_struct *task, char *buffer)
357{
Balbir Singh172ba842007-07-09 18:52:00 +0200358 return sprintf(buffer, "%llu %llu %lu\n",
Ingo Molnar826e08b2008-12-22 07:37:41 +0100359 (unsigned long long)task->se.sum_exec_runtime,
360 (unsigned long long)task->sched_info.run_delay,
Ingo Molnar2d723762007-10-15 17:00:12 +0200361 task->sched_info.pcount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362}
363#endif
364
Arjan van de Ven97455122008-01-25 21:08:34 +0100365#ifdef CONFIG_LATENCYTOP
366static int lstats_show_proc(struct seq_file *m, void *v)
367{
368 int i;
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800369 struct inode *inode = m->private;
370 struct task_struct *task = get_proc_task(inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100371
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800372 if (!task)
373 return -ESRCH;
374 seq_puts(m, "Latency Top version : v0.1\n");
Arjan van de Ven97455122008-01-25 21:08:34 +0100375 for (i = 0; i < 32; i++) {
376 if (task->latency_record[i].backtrace[0]) {
377 int q;
378 seq_printf(m, "%i %li %li ",
379 task->latency_record[i].count,
380 task->latency_record[i].time,
381 task->latency_record[i].max);
382 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
Hugh Dickins9c246242008-12-09 13:14:27 -0800383 char sym[KSYM_SYMBOL_LEN];
Arjan van de Ven97455122008-01-25 21:08:34 +0100384 char *c;
385 if (!task->latency_record[i].backtrace[q])
386 break;
387 if (task->latency_record[i].backtrace[q] == ULONG_MAX)
388 break;
389 sprint_symbol(sym, task->latency_record[i].backtrace[q]);
390 c = strchr(sym, '+');
391 if (c)
392 *c = 0;
393 seq_printf(m, "%s ", sym);
394 }
395 seq_printf(m, "\n");
396 }
397
398 }
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800399 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100400 return 0;
401}
402
403static int lstats_open(struct inode *inode, struct file *file)
404{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800405 return single_open(file, lstats_show_proc, inode);
Hiroshi Shimamotod6643d12008-02-14 10:27:00 -0800406}
407
Arjan van de Ven97455122008-01-25 21:08:34 +0100408static ssize_t lstats_write(struct file *file, const char __user *buf,
409 size_t count, loff_t *offs)
410{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800411 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100412
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800413 if (!task)
414 return -ESRCH;
Arjan van de Ven97455122008-01-25 21:08:34 +0100415 clear_all_latency_tracing(task);
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800416 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100417
418 return count;
419}
420
421static const struct file_operations proc_lstats_operations = {
422 .open = lstats_open,
423 .read = seq_read,
424 .write = lstats_write,
425 .llseek = seq_lseek,
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800426 .release = single_release,
Arjan van de Ven97455122008-01-25 21:08:34 +0100427};
428
429#endif
430
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431static int proc_oom_score(struct task_struct *task, char *buffer)
432{
Oleg Nesterovb95c35e2010-04-01 15:13:57 +0200433 unsigned long points = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700435 read_lock(&tasklist_lock);
Oleg Nesterovb95c35e2010-04-01 15:13:57 +0200436 if (pid_alive(task))
David Rientjesa63d83f2010-08-09 17:19:46 -0700437 points = oom_badness(task, NULL, NULL,
438 totalram_pages + total_swap_pages);
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700439 read_unlock(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 return sprintf(buffer, "%lu\n", points);
441}
442
Neil Hormand85f50d2007-10-18 23:40:37 -0700443struct limit_names {
444 char *name;
445 char *unit;
446};
447
448static const struct limit_names lnames[RLIM_NLIMITS] = {
Kees Cookcff4edb2009-09-22 16:45:32 -0700449 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700450 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
451 [RLIMIT_DATA] = {"Max data size", "bytes"},
452 [RLIMIT_STACK] = {"Max stack size", "bytes"},
453 [RLIMIT_CORE] = {"Max core file size", "bytes"},
454 [RLIMIT_RSS] = {"Max resident set", "bytes"},
455 [RLIMIT_NPROC] = {"Max processes", "processes"},
456 [RLIMIT_NOFILE] = {"Max open files", "files"},
457 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
458 [RLIMIT_AS] = {"Max address space", "bytes"},
459 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
460 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
461 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
462 [RLIMIT_NICE] = {"Max nice priority", NULL},
463 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
Eugene Teo88081172008-02-23 15:23:52 -0800464 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700465};
466
467/* Display limits for a process */
468static int proc_pid_limits(struct task_struct *task, char *buffer)
469{
470 unsigned int i;
471 int count = 0;
472 unsigned long flags;
473 char *bufptr = buffer;
474
475 struct rlimit rlim[RLIM_NLIMITS];
476
Lai Jiangshana6bebbc2008-10-05 00:51:15 +0400477 if (!lock_task_sighand(task, &flags))
Neil Hormand85f50d2007-10-18 23:40:37 -0700478 return 0;
Neil Hormand85f50d2007-10-18 23:40:37 -0700479 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
480 unlock_task_sighand(task, &flags);
Neil Hormand85f50d2007-10-18 23:40:37 -0700481
482 /*
483 * print the file header
484 */
485 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
486 "Limit", "Soft Limit", "Hard Limit", "Units");
487
488 for (i = 0; i < RLIM_NLIMITS; i++) {
489 if (rlim[i].rlim_cur == RLIM_INFINITY)
490 count += sprintf(&bufptr[count], "%-25s %-20s ",
491 lnames[i].name, "unlimited");
492 else
493 count += sprintf(&bufptr[count], "%-25s %-20lu ",
494 lnames[i].name, rlim[i].rlim_cur);
495
496 if (rlim[i].rlim_max == RLIM_INFINITY)
497 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
498 else
499 count += sprintf(&bufptr[count], "%-20lu ",
500 rlim[i].rlim_max);
501
502 if (lnames[i].unit)
503 count += sprintf(&bufptr[count], "%-10s\n",
504 lnames[i].unit);
505 else
506 count += sprintf(&bufptr[count], "\n");
507 }
508
509 return count;
510}
511
Roland McGrathebcb6732008-07-25 19:46:00 -0700512#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
513static int proc_pid_syscall(struct task_struct *task, char *buffer)
514{
515 long nr;
516 unsigned long args[6], sp, pc;
517
518 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
519 return sprintf(buffer, "running\n");
520
521 if (nr < 0)
522 return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
523
524 return sprintf(buffer,
525 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
526 nr,
527 args[0], args[1], args[2], args[3], args[4], args[5],
528 sp, pc);
529}
530#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
531
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532/************************************************************************/
533/* Here the fs part begins */
534/************************************************************************/
535
536/* permission checks */
Eric W. Biederman778c1142006-06-26 00:25:58 -0700537static int proc_fd_access_allowed(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538{
Eric W. Biederman778c1142006-06-26 00:25:58 -0700539 struct task_struct *task;
540 int allowed = 0;
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700541 /* Allow access to a task's file descriptors if it is us or we
542 * may use ptrace attach to the process and find out that
543 * information.
Eric W. Biederman778c1142006-06-26 00:25:58 -0700544 */
545 task = get_proc_task(inode);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700546 if (task) {
Stephen Smalley006ebb42008-05-19 08:32:49 -0400547 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
Eric W. Biederman778c1142006-06-26 00:25:58 -0700548 put_task_struct(task);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700549 }
Eric W. Biederman778c1142006-06-26 00:25:58 -0700550 return allowed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551}
552
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700553static int proc_setattr(struct dentry *dentry, struct iattr *attr)
554{
555 int error;
556 struct inode *inode = dentry->d_inode;
557
558 if (attr->ia_valid & ATTR_MODE)
559 return -EPERM;
560
561 error = inode_change_ok(inode, attr);
Christoph Hellwig10257742010-06-04 11:30:02 +0200562 if (error)
563 return error;
564
565 if ((attr->ia_valid & ATTR_SIZE) &&
566 attr->ia_size != i_size_read(inode)) {
567 error = vmtruncate(inode, attr->ia_size);
568 if (error)
569 return error;
570 }
571
572 setattr_copy(inode, attr);
573 mark_inode_dirty(inode);
574 return 0;
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700575}
576
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800577static const struct inode_operations proc_def_inode_operations = {
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700578 .setattr = proc_setattr,
579};
580
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100581static int mounts_open_common(struct inode *inode, struct file *file,
582 const struct seq_operations *op)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700584 struct task_struct *task = get_proc_task(inode);
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700585 struct nsproxy *nsp;
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800586 struct mnt_namespace *ns = NULL;
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100587 struct path root;
Al Viro5addc5d2005-11-07 17:15:49 -0500588 struct proc_mounts *p;
589 int ret = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590
Eric W. Biederman99f89552006-06-26 00:25:55 -0700591 if (task) {
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700592 rcu_read_lock();
593 nsp = task_nsproxy(task);
594 if (nsp) {
595 ns = nsp->mnt_ns;
Alexey Dobriyan863c4702007-01-26 00:56:53 -0800596 if (ns)
597 get_mnt_ns(ns);
598 }
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700599 rcu_read_unlock();
Miklos Szeredif7ad3c62010-08-10 11:41:36 +0200600 if (ns && get_task_root(task, &root) == 0)
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000601 ret = 0;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700602 put_task_struct(task);
603 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100605 if (!ns)
606 goto err;
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000607 if (ret)
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100608 goto err_put_ns;
609
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100610 ret = -ENOMEM;
611 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
612 if (!p)
613 goto err_put_path;
614
615 file->private_data = &p->m;
616 ret = seq_open(file, op);
617 if (ret)
618 goto err_free;
619
620 p->m.private = p;
621 p->ns = ns;
622 p->root = root;
623 p->event = ns->event;
624
625 return 0;
626
627 err_free:
628 kfree(p);
629 err_put_path:
630 path_put(&root);
631 err_put_ns:
632 put_mnt_ns(ns);
633 err:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 return ret;
635}
636
637static int mounts_release(struct inode *inode, struct file *file)
638{
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100639 struct proc_mounts *p = file->private_data;
640 path_put(&p->root);
641 put_mnt_ns(p->ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 return seq_release(inode, file);
643}
644
Al Viro5addc5d2005-11-07 17:15:49 -0500645static unsigned mounts_poll(struct file *file, poll_table *wait)
646{
647 struct proc_mounts *p = file->private_data;
KOSAKI Motohiro31b07092009-04-09 13:57:59 +0900648 unsigned res = POLLIN | POLLRDNORM;
Al Viro5addc5d2005-11-07 17:15:49 -0500649
Al Viro9f5596a2010-02-05 00:40:25 -0500650 poll_wait(file, &p->ns->poll, wait);
651 if (mnt_had_events(p))
KOSAKI Motohiro31b07092009-04-09 13:57:59 +0900652 res |= POLLERR | POLLPRI;
Al Viro5addc5d2005-11-07 17:15:49 -0500653
654 return res;
655}
656
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100657static int mounts_open(struct inode *inode, struct file *file)
658{
659 return mounts_open_common(inode, file, &mounts_op);
660}
661
Arjan van de Ven00977a52007-02-12 00:55:34 -0800662static const struct file_operations proc_mounts_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 .open = mounts_open,
664 .read = seq_read,
665 .llseek = seq_lseek,
666 .release = mounts_release,
Al Viro5addc5d2005-11-07 17:15:49 -0500667 .poll = mounts_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668};
669
Ram Pai2d4d4862008-03-27 13:06:25 +0100670static int mountinfo_open(struct inode *inode, struct file *file)
671{
672 return mounts_open_common(inode, file, &mountinfo_op);
673}
674
675static const struct file_operations proc_mountinfo_operations = {
676 .open = mountinfo_open,
677 .read = seq_read,
678 .llseek = seq_lseek,
679 .release = mounts_release,
680 .poll = mounts_poll,
681};
682
Chuck Leverb4629fe2006-03-20 13:44:12 -0500683static int mountstats_open(struct inode *inode, struct file *file)
684{
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100685 return mounts_open_common(inode, file, &mountstats_op);
Chuck Leverb4629fe2006-03-20 13:44:12 -0500686}
687
Arjan van de Ven00977a52007-02-12 00:55:34 -0800688static const struct file_operations proc_mountstats_operations = {
Chuck Leverb4629fe2006-03-20 13:44:12 -0500689 .open = mountstats_open,
690 .read = seq_read,
691 .llseek = seq_lseek,
692 .release = mounts_release,
693};
694
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
696
697static ssize_t proc_info_read(struct file * file, char __user * buf,
698 size_t count, loff_t *ppos)
699{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800700 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 unsigned long page;
702 ssize_t length;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700703 struct task_struct *task = get_proc_task(inode);
704
705 length = -ESRCH;
706 if (!task)
707 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708
709 if (count > PROC_BLOCK_SIZE)
710 count = PROC_BLOCK_SIZE;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700711
712 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700713 if (!(page = __get_free_page(GFP_TEMPORARY)))
Eric W. Biederman99f89552006-06-26 00:25:55 -0700714 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715
716 length = PROC_I(inode)->op.proc_read(task, (char*)page);
717
718 if (length >= 0)
719 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
720 free_page(page);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700721out:
722 put_task_struct(task);
723out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 return length;
725}
726
Arjan van de Ven00977a52007-02-12 00:55:34 -0800727static const struct file_operations proc_info_file_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 .read = proc_info_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +0100729 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730};
731
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800732static int proc_single_show(struct seq_file *m, void *v)
733{
734 struct inode *inode = m->private;
735 struct pid_namespace *ns;
736 struct pid *pid;
737 struct task_struct *task;
738 int ret;
739
740 ns = inode->i_sb->s_fs_info;
741 pid = proc_pid(inode);
742 task = get_pid_task(pid, PIDTYPE_PID);
743 if (!task)
744 return -ESRCH;
745
746 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
747
748 put_task_struct(task);
749 return ret;
750}
751
752static int proc_single_open(struct inode *inode, struct file *filp)
753{
754 int ret;
755 ret = single_open(filp, proc_single_show, NULL);
756 if (!ret) {
757 struct seq_file *m = filp->private_data;
758
759 m->private = inode;
760 }
761 return ret;
762}
763
764static const struct file_operations proc_single_file_operations = {
765 .open = proc_single_open,
766 .read = seq_read,
767 .llseek = seq_lseek,
768 .release = single_release,
769};
770
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771static int mem_open(struct inode* inode, struct file* file)
772{
773 file->private_data = (void*)((long)current->self_exec_id);
KAMEZAWA Hiroyuki4a3956c2010-10-01 14:20:22 -0700774 /* OK to pass negative loff_t, we can catch out-of-range */
775 file->f_mode |= FMODE_UNSIGNED_OFFSET;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 return 0;
777}
778
779static ssize_t mem_read(struct file * file, char __user * buf,
780 size_t count, loff_t *ppos)
781{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800782 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 char *page;
784 unsigned long src = *ppos;
785 int ret = -ESRCH;
786 struct mm_struct *mm;
787
Eric W. Biederman99f89552006-06-26 00:25:55 -0700788 if (!task)
789 goto out_no_task;
790
Roland McGrath638fa2022008-04-29 01:01:38 -0700791 if (check_mem_permission(task))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 goto out;
793
794 ret = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700795 page = (char *)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 if (!page)
797 goto out;
798
799 ret = 0;
800
801 mm = get_task_mm(task);
802 if (!mm)
803 goto out_free;
804
805 ret = -EIO;
806
807 if (file->private_data != (void*)((long)current->self_exec_id))
808 goto out_put;
809
810 ret = 0;
811
812 while (count > 0) {
813 int this_len, retval;
814
815 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
816 retval = access_process_vm(task, src, page, this_len, 0);
Roland McGrath638fa2022008-04-29 01:01:38 -0700817 if (!retval || check_mem_permission(task)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 if (!ret)
819 ret = -EIO;
820 break;
821 }
822
823 if (copy_to_user(buf, page, retval)) {
824 ret = -EFAULT;
825 break;
826 }
827
828 ret += retval;
829 src += retval;
830 buf += retval;
831 count -= retval;
832 }
833 *ppos = src;
834
835out_put:
836 mmput(mm);
837out_free:
838 free_page((unsigned long) page);
839out:
Eric W. Biederman99f89552006-06-26 00:25:55 -0700840 put_task_struct(task);
841out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 return ret;
843}
844
845#define mem_write NULL
846
847#ifndef mem_write
848/* This is a security hazard */
Glauber de Oliveira Costa63967fa2007-02-20 13:58:12 -0800849static ssize_t mem_write(struct file * file, const char __user *buf,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 size_t count, loff_t *ppos)
851{
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700852 int copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 char *page;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800854 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 unsigned long dst = *ppos;
856
Eric W. Biederman99f89552006-06-26 00:25:55 -0700857 copied = -ESRCH;
858 if (!task)
859 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860
Roland McGrath638fa2022008-04-29 01:01:38 -0700861 if (check_mem_permission(task))
Eric W. Biederman99f89552006-06-26 00:25:55 -0700862 goto out;
863
864 copied = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700865 page = (char *)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 if (!page)
Eric W. Biederman99f89552006-06-26 00:25:55 -0700867 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700869 copied = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870 while (count > 0) {
871 int this_len, retval;
872
873 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
874 if (copy_from_user(page, buf, this_len)) {
875 copied = -EFAULT;
876 break;
877 }
878 retval = access_process_vm(task, dst, page, this_len, 1);
879 if (!retval) {
880 if (!copied)
881 copied = -EIO;
882 break;
883 }
884 copied += retval;
885 buf += retval;
886 dst += retval;
887 count -= retval;
888 }
889 *ppos = dst;
890 free_page((unsigned long) page);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700891out:
892 put_task_struct(task);
893out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 return copied;
895}
896#endif
897
Matt Mackall85863e42008-02-04 22:29:04 -0800898loff_t mem_lseek(struct file *file, loff_t offset, int orig)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899{
900 switch (orig) {
901 case 0:
902 file->f_pos = offset;
903 break;
904 case 1:
905 file->f_pos += offset;
906 break;
907 default:
908 return -EINVAL;
909 }
910 force_successful_syscall_return();
911 return file->f_pos;
912}
913
Arjan van de Ven00977a52007-02-12 00:55:34 -0800914static const struct file_operations proc_mem_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 .llseek = mem_lseek,
916 .read = mem_read,
917 .write = mem_write,
918 .open = mem_open,
919};
920
James Pearson315e28c2007-10-16 23:30:17 -0700921static ssize_t environ_read(struct file *file, char __user *buf,
922 size_t count, loff_t *ppos)
923{
924 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
925 char *page;
926 unsigned long src = *ppos;
927 int ret = -ESRCH;
928 struct mm_struct *mm;
929
930 if (!task)
931 goto out_no_task;
932
Stephen Smalley006ebb42008-05-19 08:32:49 -0400933 if (!ptrace_may_access(task, PTRACE_MODE_READ))
James Pearson315e28c2007-10-16 23:30:17 -0700934 goto out;
935
936 ret = -ENOMEM;
937 page = (char *)__get_free_page(GFP_TEMPORARY);
938 if (!page)
939 goto out;
940
941 ret = 0;
942
943 mm = get_task_mm(task);
944 if (!mm)
945 goto out_free;
946
947 while (count > 0) {
948 int this_len, retval, max_len;
949
950 this_len = mm->env_end - (mm->env_start + src);
951
952 if (this_len <= 0)
953 break;
954
955 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
956 this_len = (this_len > max_len) ? max_len : this_len;
957
958 retval = access_process_vm(task, (mm->env_start + src),
959 page, this_len, 0);
960
961 if (retval <= 0) {
962 ret = retval;
963 break;
964 }
965
966 if (copy_to_user(buf, page, retval)) {
967 ret = -EFAULT;
968 break;
969 }
970
971 ret += retval;
972 src += retval;
973 buf += retval;
974 count -= retval;
975 }
976 *ppos = src;
977
978 mmput(mm);
979out_free:
980 free_page((unsigned long) page);
981out:
982 put_task_struct(task);
983out_no_task:
984 return ret;
985}
986
987static const struct file_operations proc_environ_operations = {
988 .read = environ_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +0100989 .llseek = generic_file_llseek,
James Pearson315e28c2007-10-16 23:30:17 -0700990};
991
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992static ssize_t oom_adjust_read(struct file *file, char __user *buf,
993 size_t count, loff_t *ppos)
994{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800995 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Eric W. Biederman8578cea2006-06-26 00:25:54 -0700996 char buffer[PROC_NUMBUF];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 size_t len;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -0700998 int oom_adjust = OOM_DISABLE;
999 unsigned long flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000
Eric W. Biederman99f89552006-06-26 00:25:55 -07001001 if (!task)
1002 return -ESRCH;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001003
1004 if (lock_task_sighand(task, &flags)) {
1005 oom_adjust = task->signal->oom_adj;
1006 unlock_task_sighand(task, &flags);
1007 }
1008
Eric W. Biederman99f89552006-06-26 00:25:55 -07001009 put_task_struct(task);
1010
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001011 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001012
1013 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014}
1015
1016static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1017 size_t count, loff_t *ppos)
1018{
Eric W. Biederman99f89552006-06-26 00:25:55 -07001019 struct task_struct *task;
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -07001020 char buffer[PROC_NUMBUF];
1021 long oom_adjust;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001022 unsigned long flags;
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -07001023 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001025 memset(buffer, 0, sizeof(buffer));
1026 if (count > sizeof(buffer) - 1)
1027 count = sizeof(buffer) - 1;
David Rientjes723548b2010-10-26 14:21:25 -07001028 if (copy_from_user(buffer, buf, count)) {
1029 err = -EFAULT;
1030 goto out;
1031 }
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -07001032
1033 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1034 if (err)
David Rientjes723548b2010-10-26 14:21:25 -07001035 goto out;
Alexey Dobriyan8ac773b2006-10-19 23:28:32 -07001036 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
David Rientjes723548b2010-10-26 14:21:25 -07001037 oom_adjust != OOM_DISABLE) {
1038 err = -EINVAL;
1039 goto out;
1040 }
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -07001041
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001042 task = get_proc_task(file->f_path.dentry->d_inode);
David Rientjes723548b2010-10-26 14:21:25 -07001043 if (!task) {
1044 err = -ESRCH;
1045 goto out;
1046 }
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001047
Ying Han3d5992d2010-10-26 14:21:23 -07001048 task_lock(task);
1049 if (!task->mm) {
David Rientjes723548b2010-10-26 14:21:25 -07001050 err = -EINVAL;
1051 goto err_task_lock;
Ying Han3d5992d2010-10-26 14:21:23 -07001052 }
1053
David Rientjesd19d5472010-10-26 14:21:26 -07001054 if (!lock_task_sighand(task, &flags)) {
1055 err = -ESRCH;
1056 goto err_task_lock;
1057 }
1058
1059 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1060 err = -EACCES;
1061 goto err_sighand;
1062 }
1063
Ying Han3d5992d2010-10-26 14:21:23 -07001064 if (oom_adjust != task->signal->oom_adj) {
1065 if (oom_adjust == OOM_DISABLE)
1066 atomic_inc(&task->mm->oom_disable_count);
1067 if (task->signal->oom_adj == OOM_DISABLE)
1068 atomic_dec(&task->mm->oom_disable_count);
1069 }
1070
David Rientjes51b1bd22010-08-09 17:19:47 -07001071 /*
1072 * Warn that /proc/pid/oom_adj is deprecated, see
1073 * Documentation/feature-removal-schedule.txt.
1074 */
1075 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, "
1076 "please use /proc/%d/oom_score_adj instead.\n",
1077 current->comm, task_pid_nr(current),
1078 task_pid_nr(task), task_pid_nr(task));
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001079 task->signal->oom_adj = oom_adjust;
David Rientjesa63d83f2010-08-09 17:19:46 -07001080 /*
1081 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
1082 * value is always attainable.
1083 */
1084 if (task->signal->oom_adj == OOM_ADJUST_MAX)
1085 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
1086 else
1087 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
1088 -OOM_DISABLE;
David Rientjes723548b2010-10-26 14:21:25 -07001089err_sighand:
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001090 unlock_task_sighand(task, &flags);
David Rientjesd19d5472010-10-26 14:21:26 -07001091err_task_lock:
1092 task_unlock(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -07001093 put_task_struct(task);
David Rientjes723548b2010-10-26 14:21:25 -07001094out:
1095 return err < 0 ? err : count;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096}
1097
Arjan van de Ven00977a52007-02-12 00:55:34 -08001098static const struct file_operations proc_oom_adjust_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 .read = oom_adjust_read,
1100 .write = oom_adjust_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001101 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102};
1103
David Rientjesa63d83f2010-08-09 17:19:46 -07001104static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
1105 size_t count, loff_t *ppos)
1106{
1107 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
1108 char buffer[PROC_NUMBUF];
1109 int oom_score_adj = OOM_SCORE_ADJ_MIN;
1110 unsigned long flags;
1111 size_t len;
1112
1113 if (!task)
1114 return -ESRCH;
1115 if (lock_task_sighand(task, &flags)) {
1116 oom_score_adj = task->signal->oom_score_adj;
1117 unlock_task_sighand(task, &flags);
1118 }
1119 put_task_struct(task);
1120 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1121 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1122}
1123
1124static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1125 size_t count, loff_t *ppos)
1126{
1127 struct task_struct *task;
1128 char buffer[PROC_NUMBUF];
1129 unsigned long flags;
1130 long oom_score_adj;
1131 int err;
1132
1133 memset(buffer, 0, sizeof(buffer));
1134 if (count > sizeof(buffer) - 1)
1135 count = sizeof(buffer) - 1;
David Rientjes723548b2010-10-26 14:21:25 -07001136 if (copy_from_user(buffer, buf, count)) {
1137 err = -EFAULT;
1138 goto out;
1139 }
David Rientjesa63d83f2010-08-09 17:19:46 -07001140
1141 err = strict_strtol(strstrip(buffer), 0, &oom_score_adj);
1142 if (err)
David Rientjes723548b2010-10-26 14:21:25 -07001143 goto out;
David Rientjesa63d83f2010-08-09 17:19:46 -07001144 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
David Rientjes723548b2010-10-26 14:21:25 -07001145 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1146 err = -EINVAL;
1147 goto out;
1148 }
David Rientjesa63d83f2010-08-09 17:19:46 -07001149
1150 task = get_proc_task(file->f_path.dentry->d_inode);
David Rientjes723548b2010-10-26 14:21:25 -07001151 if (!task) {
1152 err = -ESRCH;
1153 goto out;
1154 }
David Rientjesa63d83f2010-08-09 17:19:46 -07001155
Ying Han3d5992d2010-10-26 14:21:23 -07001156 task_lock(task);
1157 if (!task->mm) {
David Rientjes723548b2010-10-26 14:21:25 -07001158 err = -EINVAL;
1159 goto err_task_lock;
Ying Han3d5992d2010-10-26 14:21:23 -07001160 }
David Rientjesd19d5472010-10-26 14:21:26 -07001161
1162 if (!lock_task_sighand(task, &flags)) {
1163 err = -ESRCH;
1164 goto err_task_lock;
1165 }
1166
1167 if (oom_score_adj < task->signal->oom_score_adj &&
1168 !capable(CAP_SYS_RESOURCE)) {
1169 err = -EACCES;
1170 goto err_sighand;
1171 }
1172
Ying Han3d5992d2010-10-26 14:21:23 -07001173 if (oom_score_adj != task->signal->oom_score_adj) {
1174 if (oom_score_adj == OOM_SCORE_ADJ_MIN)
1175 atomic_inc(&task->mm->oom_disable_count);
1176 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1177 atomic_dec(&task->mm->oom_disable_count);
1178 }
David Rientjesa63d83f2010-08-09 17:19:46 -07001179 task->signal->oom_score_adj = oom_score_adj;
1180 /*
1181 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1182 * always attainable.
1183 */
1184 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1185 task->signal->oom_adj = OOM_DISABLE;
1186 else
1187 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1188 OOM_SCORE_ADJ_MAX;
David Rientjes723548b2010-10-26 14:21:25 -07001189err_sighand:
David Rientjesa63d83f2010-08-09 17:19:46 -07001190 unlock_task_sighand(task, &flags);
David Rientjesd19d5472010-10-26 14:21:26 -07001191err_task_lock:
1192 task_unlock(task);
David Rientjesa63d83f2010-08-09 17:19:46 -07001193 put_task_struct(task);
David Rientjes723548b2010-10-26 14:21:25 -07001194out:
1195 return err < 0 ? err : count;
David Rientjesa63d83f2010-08-09 17:19:46 -07001196}
1197
1198static const struct file_operations proc_oom_score_adj_operations = {
1199 .read = oom_score_adj_read,
1200 .write = oom_score_adj_write,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001201 .llseek = default_llseek,
David Rientjesa63d83f2010-08-09 17:19:46 -07001202};
1203
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204#ifdef CONFIG_AUDITSYSCALL
1205#define TMPBUFLEN 21
1206static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1207 size_t count, loff_t *ppos)
1208{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001209 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001210 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 ssize_t length;
1212 char tmpbuf[TMPBUFLEN];
1213
Eric W. Biederman99f89552006-06-26 00:25:55 -07001214 if (!task)
1215 return -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
Al Viro0c11b942008-01-10 04:20:52 -05001217 audit_get_loginuid(task));
Eric W. Biederman99f89552006-06-26 00:25:55 -07001218 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1220}
1221
1222static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1223 size_t count, loff_t *ppos)
1224{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001225 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226 char *page, *tmp;
1227 ssize_t length;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228 uid_t loginuid;
1229
1230 if (!capable(CAP_AUDIT_CONTROL))
1231 return -EPERM;
1232
Paul E. McKenney7dc52152010-02-22 17:04:52 -08001233 rcu_read_lock();
1234 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1235 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 return -EPERM;
Paul E. McKenney7dc52152010-02-22 17:04:52 -08001237 }
1238 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239
Al Viroe0182902006-05-18 08:28:02 -04001240 if (count >= PAGE_SIZE)
1241 count = PAGE_SIZE - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242
1243 if (*ppos != 0) {
1244 /* No partial writes. */
1245 return -EINVAL;
1246 }
Mel Gormane12ba742007-10-16 01:25:52 -07001247 page = (char*)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 if (!page)
1249 return -ENOMEM;
1250 length = -EFAULT;
1251 if (copy_from_user(page, buf, count))
1252 goto out_free_page;
1253
Al Viroe0182902006-05-18 08:28:02 -04001254 page[count] = '\0';
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 loginuid = simple_strtoul(page, &tmp, 10);
1256 if (tmp == page) {
1257 length = -EINVAL;
1258 goto out_free_page;
1259
1260 }
Eric W. Biederman99f89552006-06-26 00:25:55 -07001261 length = audit_set_loginuid(current, loginuid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262 if (likely(length == 0))
1263 length = count;
1264
1265out_free_page:
1266 free_page((unsigned long) page);
1267 return length;
1268}
1269
Arjan van de Ven00977a52007-02-12 00:55:34 -08001270static const struct file_operations proc_loginuid_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271 .read = proc_loginuid_read,
1272 .write = proc_loginuid_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001273 .llseek = generic_file_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274};
Eric Paris1e0bd752008-03-13 08:15:31 -04001275
1276static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1277 size_t count, loff_t *ppos)
1278{
1279 struct inode * inode = file->f_path.dentry->d_inode;
1280 struct task_struct *task = get_proc_task(inode);
1281 ssize_t length;
1282 char tmpbuf[TMPBUFLEN];
1283
1284 if (!task)
1285 return -ESRCH;
1286 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1287 audit_get_sessionid(task));
1288 put_task_struct(task);
1289 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1290}
1291
1292static const struct file_operations proc_sessionid_operations = {
1293 .read = proc_sessionid_read,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001294 .llseek = generic_file_llseek,
Eric Paris1e0bd752008-03-13 08:15:31 -04001295};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296#endif
1297
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001298#ifdef CONFIG_FAULT_INJECTION
1299static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1300 size_t count, loff_t *ppos)
1301{
1302 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1303 char buffer[PROC_NUMBUF];
1304 size_t len;
1305 int make_it_fail;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001306
1307 if (!task)
1308 return -ESRCH;
1309 make_it_fail = task->make_it_fail;
1310 put_task_struct(task);
1311
1312 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001313
1314 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001315}
1316
1317static ssize_t proc_fault_inject_write(struct file * file,
1318 const char __user * buf, size_t count, loff_t *ppos)
1319{
1320 struct task_struct *task;
1321 char buffer[PROC_NUMBUF], *end;
1322 int make_it_fail;
1323
1324 if (!capable(CAP_SYS_RESOURCE))
1325 return -EPERM;
1326 memset(buffer, 0, sizeof(buffer));
1327 if (count > sizeof(buffer) - 1)
1328 count = sizeof(buffer) - 1;
1329 if (copy_from_user(buffer, buf, count))
1330 return -EFAULT;
Vincent Licba8aaf2009-09-22 16:45:38 -07001331 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1332 if (*end)
1333 return -EINVAL;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001334 task = get_proc_task(file->f_dentry->d_inode);
1335 if (!task)
1336 return -ESRCH;
1337 task->make_it_fail = make_it_fail;
1338 put_task_struct(task);
Vincent Licba8aaf2009-09-22 16:45:38 -07001339
1340 return count;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001341}
1342
Arjan van de Ven00977a52007-02-12 00:55:34 -08001343static const struct file_operations proc_fault_inject_operations = {
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001344 .read = proc_fault_inject_read,
1345 .write = proc_fault_inject_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01001346 .llseek = generic_file_llseek,
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001347};
1348#endif
1349
Arjan van de Ven97455122008-01-25 21:08:34 +01001350
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001351#ifdef CONFIG_SCHED_DEBUG
1352/*
1353 * Print out various scheduling related per-task fields:
1354 */
1355static int sched_show(struct seq_file *m, void *v)
1356{
1357 struct inode *inode = m->private;
1358 struct task_struct *p;
1359
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001360 p = get_proc_task(inode);
1361 if (!p)
1362 return -ESRCH;
1363 proc_sched_show_task(p, m);
1364
1365 put_task_struct(p);
1366
1367 return 0;
1368}
1369
1370static ssize_t
1371sched_write(struct file *file, const char __user *buf,
1372 size_t count, loff_t *offset)
1373{
1374 struct inode *inode = file->f_path.dentry->d_inode;
1375 struct task_struct *p;
1376
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001377 p = get_proc_task(inode);
1378 if (!p)
1379 return -ESRCH;
1380 proc_sched_set_task(p);
1381
1382 put_task_struct(p);
1383
1384 return count;
1385}
1386
1387static int sched_open(struct inode *inode, struct file *filp)
1388{
1389 int ret;
1390
1391 ret = single_open(filp, sched_show, NULL);
1392 if (!ret) {
1393 struct seq_file *m = filp->private_data;
1394
1395 m->private = inode;
1396 }
1397 return ret;
1398}
1399
1400static const struct file_operations proc_pid_sched_operations = {
1401 .open = sched_open,
1402 .read = seq_read,
1403 .write = sched_write,
1404 .llseek = seq_lseek,
Alexey Dobriyan5ea473a2007-07-31 00:38:50 -07001405 .release = single_release,
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001406};
1407
1408#endif
1409
john stultz4614a696b2009-12-14 18:00:05 -08001410static ssize_t comm_write(struct file *file, const char __user *buf,
1411 size_t count, loff_t *offset)
1412{
1413 struct inode *inode = file->f_path.dentry->d_inode;
1414 struct task_struct *p;
1415 char buffer[TASK_COMM_LEN];
1416
1417 memset(buffer, 0, sizeof(buffer));
1418 if (count > sizeof(buffer) - 1)
1419 count = sizeof(buffer) - 1;
1420 if (copy_from_user(buffer, buf, count))
1421 return -EFAULT;
1422
1423 p = get_proc_task(inode);
1424 if (!p)
1425 return -ESRCH;
1426
1427 if (same_thread_group(current, p))
1428 set_task_comm(p, buffer);
1429 else
1430 count = -EINVAL;
1431
1432 put_task_struct(p);
1433
1434 return count;
1435}
1436
1437static int comm_show(struct seq_file *m, void *v)
1438{
1439 struct inode *inode = m->private;
1440 struct task_struct *p;
1441
1442 p = get_proc_task(inode);
1443 if (!p)
1444 return -ESRCH;
1445
1446 task_lock(p);
1447 seq_printf(m, "%s\n", p->comm);
1448 task_unlock(p);
1449
1450 put_task_struct(p);
1451
1452 return 0;
1453}
1454
1455static int comm_open(struct inode *inode, struct file *filp)
1456{
1457 int ret;
1458
1459 ret = single_open(filp, comm_show, NULL);
1460 if (!ret) {
1461 struct seq_file *m = filp->private_data;
1462
1463 m->private = inode;
1464 }
1465 return ret;
1466}
1467
1468static const struct file_operations proc_pid_set_comm_operations = {
1469 .open = comm_open,
1470 .read = seq_read,
1471 .write = comm_write,
1472 .llseek = seq_lseek,
1473 .release = single_release,
1474};
1475
Matt Helsley925d1c42008-04-29 01:01:36 -07001476/*
1477 * We added or removed a vma mapping the executable. The vmas are only mapped
1478 * during exec and are not mapped with the mmap system call.
1479 * Callers must hold down_write() on the mm's mmap_sem for these
1480 */
1481void added_exe_file_vma(struct mm_struct *mm)
1482{
1483 mm->num_exe_file_vmas++;
1484}
1485
1486void removed_exe_file_vma(struct mm_struct *mm)
1487{
1488 mm->num_exe_file_vmas--;
1489 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1490 fput(mm->exe_file);
1491 mm->exe_file = NULL;
1492 }
1493
1494}
1495
1496void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1497{
1498 if (new_exe_file)
1499 get_file(new_exe_file);
1500 if (mm->exe_file)
1501 fput(mm->exe_file);
1502 mm->exe_file = new_exe_file;
1503 mm->num_exe_file_vmas = 0;
1504}
1505
1506struct file *get_mm_exe_file(struct mm_struct *mm)
1507{
1508 struct file *exe_file;
1509
1510 /* We need mmap_sem to protect against races with removal of
1511 * VM_EXECUTABLE vmas */
1512 down_read(&mm->mmap_sem);
1513 exe_file = mm->exe_file;
1514 if (exe_file)
1515 get_file(exe_file);
1516 up_read(&mm->mmap_sem);
1517 return exe_file;
1518}
1519
1520void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1521{
1522 /* It's safe to write the exe_file pointer without exe_file_lock because
1523 * this is called during fork when the task is not yet in /proc */
1524 newmm->exe_file = get_mm_exe_file(oldmm);
1525}
1526
1527static int proc_exe_link(struct inode *inode, struct path *exe_path)
1528{
1529 struct task_struct *task;
1530 struct mm_struct *mm;
1531 struct file *exe_file;
1532
1533 task = get_proc_task(inode);
1534 if (!task)
1535 return -ENOENT;
1536 mm = get_task_mm(task);
1537 put_task_struct(task);
1538 if (!mm)
1539 return -ENOENT;
1540 exe_file = get_mm_exe_file(mm);
1541 mmput(mm);
1542 if (exe_file) {
1543 *exe_path = exe_file->f_path;
1544 path_get(&exe_file->f_path);
1545 fput(exe_file);
1546 return 0;
1547 } else
1548 return -ENOENT;
1549}
1550
Al Viro008b1502005-08-20 00:17:39 +01001551static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552{
1553 struct inode *inode = dentry->d_inode;
1554 int error = -EACCES;
1555
1556 /* We don't need a base pointer in the /proc filesystem */
Jan Blunck1d957f92008-02-14 19:34:35 -08001557 path_put(&nd->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558
Eric W. Biederman778c1142006-06-26 00:25:58 -07001559 /* Are we allowed to snoop on the tasks file descriptors? */
1560 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001563 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564out:
Al Viro008b1502005-08-20 00:17:39 +01001565 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566}
1567
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001568static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569{
Mel Gormane12ba742007-10-16 01:25:52 -07001570 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001571 char *pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572 int len;
1573
1574 if (!tmp)
1575 return -ENOMEM;
Akinobu Mita0c28f282007-05-08 00:31:41 -07001576
Miklos Szeredi8df9d1a2010-08-10 11:41:41 +02001577 pathname = d_path_with_unreachable(path, tmp, PAGE_SIZE);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001578 len = PTR_ERR(pathname);
1579 if (IS_ERR(pathname))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580 goto out;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001581 len = tmp + PAGE_SIZE - 1 - pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582
1583 if (len > buflen)
1584 len = buflen;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001585 if (copy_to_user(buffer, pathname, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586 len = -EFAULT;
1587 out:
1588 free_page((unsigned long)tmp);
1589 return len;
1590}
1591
1592static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1593{
1594 int error = -EACCES;
1595 struct inode *inode = dentry->d_inode;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001596 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597
Eric W. Biederman778c1142006-06-26 00:25:58 -07001598 /* Are we allowed to snoop on the tasks file descriptors? */
1599 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001602 error = PROC_I(inode)->op.proc_get_link(inode, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 if (error)
1604 goto out;
1605
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001606 error = do_proc_readlink(&path, buffer, buflen);
1607 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609 return error;
1610}
1611
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08001612static const struct inode_operations proc_pid_link_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613 .readlink = proc_pid_readlink,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001614 .follow_link = proc_pid_follow_link,
1615 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616};
1617
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001618
1619/* building an inode */
1620
1621static int task_dumpable(struct task_struct *task)
1622{
1623 int dumpable = 0;
1624 struct mm_struct *mm;
1625
1626 task_lock(task);
1627 mm = task->mm;
1628 if (mm)
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001629 dumpable = get_dumpable(mm);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001630 task_unlock(task);
1631 if(dumpable == 1)
1632 return 1;
1633 return 0;
1634}
1635
1636
Eric W. Biederman61a28782006-10-02 02:18:49 -07001637static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001638{
1639 struct inode * inode;
1640 struct proc_inode *ei;
David Howellsc69e8d92008-11-14 10:39:19 +11001641 const struct cred *cred;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001642
1643 /* We need a new inode */
1644
1645 inode = new_inode(sb);
1646 if (!inode)
1647 goto out;
1648
1649 /* Common stuff */
1650 ei = PROC_I(inode);
Christoph Hellwig85fe4022010-10-23 11:19:54 -04001651 inode->i_ino = get_next_ino();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001652 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001653 inode->i_op = &proc_def_inode_operations;
1654
1655 /*
1656 * grab the reference to task.
1657 */
Oleg Nesterov1a657f782006-10-02 02:18:59 -07001658 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001659 if (!ei->pid)
1660 goto out_unlock;
1661
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001662 if (task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001663 rcu_read_lock();
1664 cred = __task_cred(task);
1665 inode->i_uid = cred->euid;
1666 inode->i_gid = cred->egid;
1667 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001668 }
1669 security_task_to_inode(task, inode);
1670
1671out:
1672 return inode;
1673
1674out_unlock:
1675 iput(inode);
1676 return NULL;
1677}
1678
1679static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1680{
1681 struct inode *inode = dentry->d_inode;
1682 struct task_struct *task;
David Howellsc69e8d92008-11-14 10:39:19 +11001683 const struct cred *cred;
1684
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001685 generic_fillattr(inode, stat);
1686
1687 rcu_read_lock();
1688 stat->uid = 0;
1689 stat->gid = 0;
1690 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1691 if (task) {
1692 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1693 task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001694 cred = __task_cred(task);
1695 stat->uid = cred->euid;
1696 stat->gid = cred->egid;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001697 }
1698 }
1699 rcu_read_unlock();
1700 return 0;
1701}
1702
1703/* dentry stuff */
1704
1705/*
1706 * Exceptional case: normally we are not allowed to unhash a busy
1707 * directory. In this case, however, we can do it - no aliasing problems
1708 * due to the way we treat inodes.
1709 *
1710 * Rewrite the inode's ownerships here because the owning task may have
1711 * performed a setuid(), etc.
1712 *
1713 * Before the /proc/pid/status file was created the only way to read
1714 * the effective uid of a /process was to stat /proc/pid. Reading
1715 * /proc/pid/status is slow enough that procps and other packages
1716 * kept stating /proc/pid. To keep the rules in /proc simple I have
1717 * made this apply to all per process world readable and executable
1718 * directories.
1719 */
1720static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1721{
1722 struct inode *inode = dentry->d_inode;
1723 struct task_struct *task = get_proc_task(inode);
David Howellsc69e8d92008-11-14 10:39:19 +11001724 const struct cred *cred;
1725
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001726 if (task) {
1727 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1728 task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001729 rcu_read_lock();
1730 cred = __task_cred(task);
1731 inode->i_uid = cred->euid;
1732 inode->i_gid = cred->egid;
1733 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001734 } else {
1735 inode->i_uid = 0;
1736 inode->i_gid = 0;
1737 }
1738 inode->i_mode &= ~(S_ISUID | S_ISGID);
1739 security_task_to_inode(task, inode);
1740 put_task_struct(task);
1741 return 1;
1742 }
1743 d_drop(dentry);
1744 return 0;
1745}
1746
1747static int pid_delete_dentry(struct dentry * dentry)
1748{
1749 /* Is the task we represent dead?
1750 * If so, then don't put the dentry on the lru list,
1751 * kill it immediately.
1752 */
1753 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1754}
1755
Al Virod72f71e2009-02-20 05:58:47 +00001756static const struct dentry_operations pid_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001757{
1758 .d_revalidate = pid_revalidate,
1759 .d_delete = pid_delete_dentry,
1760};
1761
1762/* Lookups */
1763
Eric Dumazetc5141e62007-05-08 00:26:15 -07001764typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1765 struct task_struct *, const void *);
Eric W. Biederman61a28782006-10-02 02:18:49 -07001766
Eric W. Biederman1c0d04c2006-10-02 02:18:57 -07001767/*
1768 * Fill a directory entry.
1769 *
1770 * If possible create the dcache entry and derive our inode number and
1771 * file type from dcache entry.
1772 *
1773 * Since all of the proc inode numbers are dynamically generated, the inode
1774 * numbers do not exist until the inode is cache. This means creating the
1775 * the dcache entry in readdir is necessary to keep the inode numbers
1776 * reported by readdir in sync with the inode numbers reported
1777 * by stat.
1778 */
Eric W. Biederman61a28782006-10-02 02:18:49 -07001779static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1780 char *name, int len,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001781 instantiate_t instantiate, struct task_struct *task, const void *ptr)
Eric W. Biederman61a28782006-10-02 02:18:49 -07001782{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001783 struct dentry *child, *dir = filp->f_path.dentry;
Eric W. Biederman61a28782006-10-02 02:18:49 -07001784 struct inode *inode;
1785 struct qstr qname;
1786 ino_t ino = 0;
1787 unsigned type = DT_UNKNOWN;
1788
1789 qname.name = name;
1790 qname.len = len;
1791 qname.hash = full_name_hash(name, len);
1792
1793 child = d_lookup(dir, &qname);
1794 if (!child) {
1795 struct dentry *new;
1796 new = d_alloc(dir, &qname);
1797 if (new) {
1798 child = instantiate(dir->d_inode, new, task, ptr);
1799 if (child)
1800 dput(new);
1801 else
1802 child = new;
1803 }
1804 }
1805 if (!child || IS_ERR(child) || !child->d_inode)
1806 goto end_instantiate;
1807 inode = child->d_inode;
1808 if (inode) {
1809 ino = inode->i_ino;
1810 type = inode->i_mode >> 12;
1811 }
1812 dput(child);
1813end_instantiate:
1814 if (!ino)
1815 ino = find_inode_number(dir, &qname);
1816 if (!ino)
1817 ino = 1;
1818 return filldir(dirent, name, len, filp->f_pos, ino, type);
1819}
1820
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001821static unsigned name_to_int(struct dentry *dentry)
1822{
1823 const char *name = dentry->d_name.name;
1824 int len = dentry->d_name.len;
1825 unsigned n = 0;
1826
1827 if (len > 1 && *name == '0')
1828 goto out;
1829 while (len-- > 0) {
1830 unsigned c = *name++ - '0';
1831 if (c > 9)
1832 goto out;
1833 if (n >= (~0U-9)/10)
1834 goto out;
1835 n *= 10;
1836 n += c;
1837 }
1838 return n;
1839out:
1840 return ~0U;
1841}
1842
Miklos Szeredi27932742007-05-08 00:26:17 -07001843#define PROC_FDINFO_MAX 64
1844
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001845static int proc_fd_info(struct inode *inode, struct path *path, char *info)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001846{
1847 struct task_struct *task = get_proc_task(inode);
1848 struct files_struct *files = NULL;
1849 struct file *file;
1850 int fd = proc_fd(inode);
1851
1852 if (task) {
1853 files = get_files_struct(task);
1854 put_task_struct(task);
1855 }
1856 if (files) {
1857 /*
1858 * We are not taking a ref to the file structure, so we must
1859 * hold ->file_lock.
1860 */
1861 spin_lock(&files->file_lock);
1862 file = fcheck_files(files, fd);
1863 if (file) {
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001864 if (path) {
1865 *path = file->f_path;
1866 path_get(&file->f_path);
1867 }
Miklos Szeredi27932742007-05-08 00:26:17 -07001868 if (info)
1869 snprintf(info, PROC_FDINFO_MAX,
1870 "pos:\t%lli\n"
1871 "flags:\t0%o\n",
1872 (long long) file->f_pos,
1873 file->f_flags);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001874 spin_unlock(&files->file_lock);
1875 put_files_struct(files);
1876 return 0;
1877 }
1878 spin_unlock(&files->file_lock);
1879 put_files_struct(files);
1880 }
1881 return -ENOENT;
1882}
1883
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001884static int proc_fd_link(struct inode *inode, struct path *path)
Miklos Szeredi27932742007-05-08 00:26:17 -07001885{
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001886 return proc_fd_info(inode, path, NULL);
Miklos Szeredi27932742007-05-08 00:26:17 -07001887}
1888
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001889static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1890{
1891 struct inode *inode = dentry->d_inode;
1892 struct task_struct *task = get_proc_task(inode);
1893 int fd = proc_fd(inode);
1894 struct files_struct *files;
David Howellsc69e8d92008-11-14 10:39:19 +11001895 const struct cred *cred;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001896
1897 if (task) {
1898 files = get_files_struct(task);
1899 if (files) {
1900 rcu_read_lock();
1901 if (fcheck_files(files, fd)) {
1902 rcu_read_unlock();
1903 put_files_struct(files);
1904 if (task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001905 rcu_read_lock();
1906 cred = __task_cred(task);
1907 inode->i_uid = cred->euid;
1908 inode->i_gid = cred->egid;
1909 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001910 } else {
1911 inode->i_uid = 0;
1912 inode->i_gid = 0;
1913 }
1914 inode->i_mode &= ~(S_ISUID | S_ISGID);
1915 security_task_to_inode(task, inode);
1916 put_task_struct(task);
1917 return 1;
1918 }
1919 rcu_read_unlock();
1920 put_files_struct(files);
1921 }
1922 put_task_struct(task);
1923 }
1924 d_drop(dentry);
1925 return 0;
1926}
1927
Al Virod72f71e2009-02-20 05:58:47 +00001928static const struct dentry_operations tid_fd_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001929{
1930 .d_revalidate = tid_fd_revalidate,
1931 .d_delete = pid_delete_dentry,
1932};
1933
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001934static struct dentry *proc_fd_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001935 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001936{
Eric Dumazetc5141e62007-05-08 00:26:15 -07001937 unsigned fd = *(const unsigned *)ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001938 struct file *file;
1939 struct files_struct *files;
1940 struct inode *inode;
1941 struct proc_inode *ei;
1942 struct dentry *error = ERR_PTR(-ENOENT);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001943
Eric W. Biederman61a28782006-10-02 02:18:49 -07001944 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001945 if (!inode)
1946 goto out;
1947 ei = PROC_I(inode);
1948 ei->fd = fd;
1949 files = get_files_struct(task);
1950 if (!files)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001951 goto out_iput;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001952 inode->i_mode = S_IFLNK;
1953
1954 /*
1955 * We are not taking a ref to the file structure, so we must
1956 * hold ->file_lock.
1957 */
1958 spin_lock(&files->file_lock);
1959 file = fcheck_files(files, fd);
1960 if (!file)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001961 goto out_unlock;
Al Viroaeb5d722008-09-02 15:28:45 -04001962 if (file->f_mode & FMODE_READ)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001963 inode->i_mode |= S_IRUSR | S_IXUSR;
Al Viroaeb5d722008-09-02 15:28:45 -04001964 if (file->f_mode & FMODE_WRITE)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001965 inode->i_mode |= S_IWUSR | S_IXUSR;
1966 spin_unlock(&files->file_lock);
1967 put_files_struct(files);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001968
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001969 inode->i_op = &proc_pid_link_inode_operations;
1970 inode->i_size = 64;
1971 ei->op.proc_get_link = proc_fd_link;
1972 dentry->d_op = &tid_fd_dentry_operations;
1973 d_add(dentry, inode);
1974 /* Close the race of the process dying before we return the dentry */
1975 if (tid_fd_revalidate(dentry, NULL))
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001976 error = NULL;
1977
1978 out:
1979 return error;
1980out_unlock:
1981 spin_unlock(&files->file_lock);
1982 put_files_struct(files);
1983out_iput:
1984 iput(inode);
1985 goto out;
1986}
1987
Miklos Szeredi27932742007-05-08 00:26:17 -07001988static struct dentry *proc_lookupfd_common(struct inode *dir,
1989 struct dentry *dentry,
1990 instantiate_t instantiate)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001991{
1992 struct task_struct *task = get_proc_task(dir);
1993 unsigned fd = name_to_int(dentry);
1994 struct dentry *result = ERR_PTR(-ENOENT);
1995
1996 if (!task)
1997 goto out_no_task;
1998 if (fd == ~0U)
1999 goto out;
2000
Miklos Szeredi27932742007-05-08 00:26:17 -07002001 result = instantiate(dir, dentry, task, &fd);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002002out:
2003 put_task_struct(task);
2004out_no_task:
2005 return result;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002006}
2007
Miklos Szeredi27932742007-05-08 00:26:17 -07002008static int proc_readfd_common(struct file * filp, void * dirent,
2009 filldir_t filldir, instantiate_t instantiate)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002010{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002011 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman56347082006-06-26 00:25:40 -07002012 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002013 struct task_struct *p = get_proc_task(inode);
Pavel Emelyanov457c2512007-10-18 23:40:43 -07002014 unsigned int fd, ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016 struct files_struct * files;
2017
2018 retval = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002019 if (!p)
2020 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022
2023 fd = filp->f_pos;
2024 switch (fd) {
2025 case 0:
2026 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
2027 goto out;
2028 filp->f_pos++;
2029 case 1:
Eric W. Biederman56347082006-06-26 00:25:40 -07002030 ino = parent_ino(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2032 goto out;
2033 filp->f_pos++;
2034 default:
2035 files = get_files_struct(p);
2036 if (!files)
2037 goto out;
Dipankar Sarmab8359962005-09-09 13:04:14 -07002038 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039 for (fd = filp->f_pos-2;
Al Viro9b4f5262008-04-22 01:32:44 -04002040 fd < files_fdtable(files)->max_fds;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041 fd++, filp->f_pos++) {
Miklos Szeredi27932742007-05-08 00:26:17 -07002042 char name[PROC_NUMBUF];
2043 int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044
2045 if (!fcheck_files(files, fd))
2046 continue;
Dipankar Sarmab8359962005-09-09 13:04:14 -07002047 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002048
Miklos Szeredi27932742007-05-08 00:26:17 -07002049 len = snprintf(name, sizeof(name), "%d", fd);
2050 if (proc_fill_cache(filp, dirent, filldir,
2051 name, len, instantiate,
2052 p, &fd) < 0) {
Dipankar Sarmab8359962005-09-09 13:04:14 -07002053 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002054 break;
2055 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07002056 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07002058 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059 put_files_struct(files);
2060 }
2061out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07002062 put_task_struct(p);
2063out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002064 return retval;
2065}
2066
Miklos Szeredi27932742007-05-08 00:26:17 -07002067static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
2068 struct nameidata *nd)
2069{
2070 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
2071}
2072
2073static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
2074{
2075 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
2076}
2077
2078static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
2079 size_t len, loff_t *ppos)
2080{
2081 char tmp[PROC_FDINFO_MAX];
Jan Blunck3dcd25f2008-02-14 19:38:35 -08002082 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
Miklos Szeredi27932742007-05-08 00:26:17 -07002083 if (!err)
2084 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
2085 return err;
2086}
2087
2088static const struct file_operations proc_fdinfo_file_operations = {
Arnd Bergmann87df8422010-03-17 23:06:02 +01002089 .open = nonseekable_open,
Miklos Szeredi27932742007-05-08 00:26:17 -07002090 .read = proc_fdinfo_read,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002091 .llseek = no_llseek,
Miklos Szeredi27932742007-05-08 00:26:17 -07002092};
2093
Arjan van de Ven00977a52007-02-12 00:55:34 -08002094static const struct file_operations proc_fd_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095 .read = generic_read_dir,
2096 .readdir = proc_readfd,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002097 .llseek = default_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098};
2099
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100/*
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002101 * /proc/pid/fd needs a special permission handler so that a process can still
2102 * access /proc/self/fd after it has executed a setuid().
2103 */
Al Viroe6305c42008-07-15 21:03:57 -04002104static int proc_fd_permission(struct inode *inode, int mask)
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002105{
2106 int rv;
2107
2108 rv = generic_permission(inode, mask, NULL);
2109 if (rv == 0)
2110 return 0;
2111 if (task_pid(current) == proc_pid(inode))
2112 rv = 0;
2113 return rv;
2114}
2115
2116/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117 * proc directories can do almost nothing..
2118 */
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002119static const struct inode_operations proc_fd_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120 .lookup = proc_lookupfd,
Alexey Dobriyan8948e112007-05-08 00:23:35 -07002121 .permission = proc_fd_permission,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07002122 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123};
2124
Miklos Szeredi27932742007-05-08 00:26:17 -07002125static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2126 struct dentry *dentry, struct task_struct *task, const void *ptr)
2127{
2128 unsigned fd = *(unsigned *)ptr;
2129 struct inode *inode;
2130 struct proc_inode *ei;
2131 struct dentry *error = ERR_PTR(-ENOENT);
2132
2133 inode = proc_pid_make_inode(dir->i_sb, task);
2134 if (!inode)
2135 goto out;
2136 ei = PROC_I(inode);
2137 ei->fd = fd;
2138 inode->i_mode = S_IFREG | S_IRUSR;
2139 inode->i_fop = &proc_fdinfo_file_operations;
2140 dentry->d_op = &tid_fd_dentry_operations;
2141 d_add(dentry, inode);
2142 /* Close the race of the process dying before we return the dentry */
2143 if (tid_fd_revalidate(dentry, NULL))
2144 error = NULL;
2145
2146 out:
2147 return error;
2148}
2149
2150static struct dentry *proc_lookupfdinfo(struct inode *dir,
2151 struct dentry *dentry,
2152 struct nameidata *nd)
2153{
2154 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2155}
2156
2157static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2158{
2159 return proc_readfd_common(filp, dirent, filldir,
2160 proc_fdinfo_instantiate);
2161}
2162
2163static const struct file_operations proc_fdinfo_operations = {
2164 .read = generic_read_dir,
2165 .readdir = proc_readfdinfo,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002166 .llseek = default_llseek,
Miklos Szeredi27932742007-05-08 00:26:17 -07002167};
2168
2169/*
2170 * proc directories can do almost nothing..
2171 */
2172static const struct inode_operations proc_fdinfo_inode_operations = {
2173 .lookup = proc_lookupfdinfo,
2174 .setattr = proc_setattr,
2175};
2176
2177
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002178static struct dentry *proc_pident_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002179 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002180{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002181 const struct pid_entry *p = ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002182 struct inode *inode;
2183 struct proc_inode *ei;
KOSAKI Motohirobd6daba2009-05-28 14:34:21 -07002184 struct dentry *error = ERR_PTR(-ENOENT);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002185
Eric W. Biederman61a28782006-10-02 02:18:49 -07002186 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002187 if (!inode)
2188 goto out;
2189
2190 ei = PROC_I(inode);
2191 inode->i_mode = p->mode;
2192 if (S_ISDIR(inode->i_mode))
2193 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2194 if (p->iop)
2195 inode->i_op = p->iop;
2196 if (p->fop)
2197 inode->i_fop = p->fop;
2198 ei->op = p->op;
2199 dentry->d_op = &pid_dentry_operations;
2200 d_add(dentry, inode);
2201 /* Close the race of the process dying before we return the dentry */
2202 if (pid_revalidate(dentry, NULL))
2203 error = NULL;
2204out:
2205 return error;
2206}
2207
Linus Torvalds1da177e2005-04-16 15:20:36 -07002208static struct dentry *proc_pident_lookup(struct inode *dir,
2209 struct dentry *dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002210 const struct pid_entry *ents,
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002211 unsigned int nents)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212{
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002213 struct dentry *error;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002214 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002215 const struct pid_entry *p, *last;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002216
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002217 error = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218
Eric W. Biederman99f89552006-06-26 00:25:55 -07002219 if (!task)
2220 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002222 /*
2223 * Yes, it does not scale. And it should not. Don't add
2224 * new entries into /proc/<tgid>/ without very good reasons.
2225 */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002226 last = &ents[nents - 1];
2227 for (p = ents; p <= last; p++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002228 if (p->len != dentry->d_name.len)
2229 continue;
2230 if (!memcmp(dentry->d_name.name, p->name, p->len))
2231 break;
2232 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002233 if (p > last)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002234 goto out;
2235
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002236 error = proc_pident_instantiate(dir, dentry, task, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002237out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07002238 put_task_struct(task);
2239out_no_task:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002240 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241}
2242
Eric Dumazetc5141e62007-05-08 00:26:15 -07002243static int proc_pident_fill_cache(struct file *filp, void *dirent,
2244 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002245{
2246 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2247 proc_pident_instantiate, task, p);
2248}
2249
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002250static int proc_pident_readdir(struct file *filp,
2251 void *dirent, filldir_t filldir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002252 const struct pid_entry *ents, unsigned int nents)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002253{
2254 int i;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002255 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002256 struct inode *inode = dentry->d_inode;
2257 struct task_struct *task = get_proc_task(inode);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002258 const struct pid_entry *p, *last;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002259 ino_t ino;
2260 int ret;
2261
2262 ret = -ENOENT;
2263 if (!task)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002264 goto out_no_task;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002265
2266 ret = 0;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002267 i = filp->f_pos;
2268 switch (i) {
2269 case 0:
2270 ino = inode->i_ino;
2271 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2272 goto out;
2273 i++;
2274 filp->f_pos++;
2275 /* fall through */
2276 case 1:
2277 ino = parent_ino(dentry);
2278 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2279 goto out;
2280 i++;
2281 filp->f_pos++;
2282 /* fall through */
2283 default:
2284 i -= 2;
2285 if (i >= nents) {
2286 ret = 1;
2287 goto out;
2288 }
2289 p = ents + i;
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002290 last = &ents[nents - 1];
2291 while (p <= last) {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002292 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002293 goto out;
2294 filp->f_pos++;
2295 p++;
2296 }
2297 }
2298
2299 ret = 1;
2300out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07002301 put_task_struct(task);
2302out_no_task:
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002303 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002304}
2305
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306#ifdef CONFIG_SECURITY
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002307static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2308 size_t count, loff_t *ppos)
2309{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002310 struct inode * inode = file->f_path.dentry->d_inode;
Al Viro04ff9702007-03-12 16:17:58 +00002311 char *p = NULL;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002312 ssize_t length;
2313 struct task_struct *task = get_proc_task(inode);
2314
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002315 if (!task)
Al Viro04ff9702007-03-12 16:17:58 +00002316 return -ESRCH;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002317
2318 length = security_getprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002319 (char*)file->f_path.dentry->d_name.name,
Al Viro04ff9702007-03-12 16:17:58 +00002320 &p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002321 put_task_struct(task);
Al Viro04ff9702007-03-12 16:17:58 +00002322 if (length > 0)
2323 length = simple_read_from_buffer(buf, count, ppos, p, length);
2324 kfree(p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002325 return length;
2326}
2327
2328static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2329 size_t count, loff_t *ppos)
2330{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002331 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002332 char *page;
2333 ssize_t length;
2334 struct task_struct *task = get_proc_task(inode);
2335
2336 length = -ESRCH;
2337 if (!task)
2338 goto out_no_task;
2339 if (count > PAGE_SIZE)
2340 count = PAGE_SIZE;
2341
2342 /* No partial writes. */
2343 length = -EINVAL;
2344 if (*ppos != 0)
2345 goto out;
2346
2347 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -07002348 page = (char*)__get_free_page(GFP_TEMPORARY);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002349 if (!page)
2350 goto out;
2351
2352 length = -EFAULT;
2353 if (copy_from_user(page, buf, count))
2354 goto out_free;
2355
David Howells107db7c2009-05-08 13:55:27 +01002356 /* Guard against adverse ptrace interaction */
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07002357 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
David Howells107db7c2009-05-08 13:55:27 +01002358 if (length < 0)
2359 goto out_free;
2360
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002361 length = security_setprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002362 (char*)file->f_path.dentry->d_name.name,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002363 (void*)page, count);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07002364 mutex_unlock(&task->signal->cred_guard_mutex);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002365out_free:
2366 free_page((unsigned long) page);
2367out:
2368 put_task_struct(task);
2369out_no_task:
2370 return length;
2371}
2372
Arjan van de Ven00977a52007-02-12 00:55:34 -08002373static const struct file_operations proc_pid_attr_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002374 .read = proc_pid_attr_read,
2375 .write = proc_pid_attr_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01002376 .llseek = generic_file_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002377};
2378
Eric Dumazetc5141e62007-05-08 00:26:15 -07002379static const struct pid_entry attr_dir_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002380 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2381 REG("prev", S_IRUGO, proc_pid_attr_operations),
2382 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2383 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2384 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2385 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002386};
2387
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002388static int proc_attr_dir_readdir(struct file * filp,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389 void * dirent, filldir_t filldir)
2390{
2391 return proc_pident_readdir(filp,dirent,filldir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002392 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002393}
2394
Arjan van de Ven00977a52007-02-12 00:55:34 -08002395static const struct file_operations proc_attr_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002396 .read = generic_read_dir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002397 .readdir = proc_attr_dir_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002398 .llseek = default_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002399};
2400
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002401static struct dentry *proc_attr_dir_lookup(struct inode *dir,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002402 struct dentry *dentry, struct nameidata *nd)
2403{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002404 return proc_pident_lookup(dir, dentry,
2405 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002406}
2407
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002408static const struct inode_operations proc_attr_dir_inode_operations = {
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002409 .lookup = proc_attr_dir_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07002410 .getattr = pid_getattr,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07002411 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002412};
2413
Linus Torvalds1da177e2005-04-16 15:20:36 -07002414#endif
2415
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08002416#ifdef CONFIG_ELF_CORE
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002417static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2418 size_t count, loff_t *ppos)
2419{
2420 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2421 struct mm_struct *mm;
2422 char buffer[PROC_NUMBUF];
2423 size_t len;
2424 int ret;
2425
2426 if (!task)
2427 return -ESRCH;
2428
2429 ret = 0;
2430 mm = get_task_mm(task);
2431 if (mm) {
2432 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2433 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2434 MMF_DUMP_FILTER_SHIFT));
2435 mmput(mm);
2436 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2437 }
2438
2439 put_task_struct(task);
2440
2441 return ret;
2442}
2443
2444static ssize_t proc_coredump_filter_write(struct file *file,
2445 const char __user *buf,
2446 size_t count,
2447 loff_t *ppos)
2448{
2449 struct task_struct *task;
2450 struct mm_struct *mm;
2451 char buffer[PROC_NUMBUF], *end;
2452 unsigned int val;
2453 int ret;
2454 int i;
2455 unsigned long mask;
2456
2457 ret = -EFAULT;
2458 memset(buffer, 0, sizeof(buffer));
2459 if (count > sizeof(buffer) - 1)
2460 count = sizeof(buffer) - 1;
2461 if (copy_from_user(buffer, buf, count))
2462 goto out_no_task;
2463
2464 ret = -EINVAL;
2465 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2466 if (*end == '\n')
2467 end++;
2468 if (end - buffer == 0)
2469 goto out_no_task;
2470
2471 ret = -ESRCH;
2472 task = get_proc_task(file->f_dentry->d_inode);
2473 if (!task)
2474 goto out_no_task;
2475
2476 ret = end - buffer;
2477 mm = get_task_mm(task);
2478 if (!mm)
2479 goto out_no_mm;
2480
2481 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2482 if (val & mask)
2483 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2484 else
2485 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2486 }
2487
2488 mmput(mm);
2489 out_no_mm:
2490 put_task_struct(task);
2491 out_no_task:
2492 return ret;
2493}
2494
2495static const struct file_operations proc_coredump_filter_operations = {
2496 .read = proc_coredump_filter_read,
2497 .write = proc_coredump_filter_write,
Arnd Bergmann87df8422010-03-17 23:06:02 +01002498 .llseek = generic_file_llseek,
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002499};
2500#endif
2501
Linus Torvalds1da177e2005-04-16 15:20:36 -07002502/*
2503 * /proc/self:
2504 */
2505static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2506 int buflen)
2507{
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002508 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002509 pid_t tgid = task_tgid_nr_ns(current, ns);
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002510 char tmp[PROC_NUMBUF];
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002511 if (!tgid)
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002512 return -ENOENT;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002513 sprintf(tmp, "%d", tgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002514 return vfs_readlink(dentry,buffer,buflen,tmp);
2515}
2516
Al Viro008b1502005-08-20 00:17:39 +01002517static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002518{
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002519 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002520 pid_t tgid = task_tgid_nr_ns(current, ns);
Al Viro7fee4862010-01-14 01:03:28 -05002521 char *name = ERR_PTR(-ENOENT);
2522 if (tgid) {
2523 name = __getname();
2524 if (!name)
2525 name = ERR_PTR(-ENOMEM);
2526 else
2527 sprintf(name, "%d", tgid);
2528 }
2529 nd_set_link(nd, name);
2530 return NULL;
2531}
2532
2533static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2534 void *cookie)
2535{
2536 char *s = nd_get_link(nd);
2537 if (!IS_ERR(s))
2538 __putname(s);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002539}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002540
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002541static const struct inode_operations proc_self_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002542 .readlink = proc_self_readlink,
2543 .follow_link = proc_self_follow_link,
Al Viro7fee4862010-01-14 01:03:28 -05002544 .put_link = proc_self_put_link,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002545};
2546
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002547/*
Eric W. Biederman801199c2006-10-02 02:18:48 -07002548 * proc base
2549 *
2550 * These are the directory entries in the root directory of /proc
2551 * that properly belong to the /proc filesystem, as they describe
2552 * describe something that is process related.
2553 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07002554static const struct pid_entry proc_base_stuff[] = {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002555 NOD("self", S_IFLNK|S_IRWXUGO,
Eric W. Biederman801199c2006-10-02 02:18:48 -07002556 &proc_self_inode_operations, NULL, {}),
Eric W. Biederman801199c2006-10-02 02:18:48 -07002557};
2558
2559/*
2560 * Exceptional case: normally we are not allowed to unhash a busy
2561 * directory. In this case, however, we can do it - no aliasing problems
2562 * due to the way we treat inodes.
2563 */
2564static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2565{
2566 struct inode *inode = dentry->d_inode;
2567 struct task_struct *task = get_proc_task(inode);
2568 if (task) {
2569 put_task_struct(task);
2570 return 1;
2571 }
2572 d_drop(dentry);
2573 return 0;
2574}
2575
Al Virod72f71e2009-02-20 05:58:47 +00002576static const struct dentry_operations proc_base_dentry_operations =
Eric W. Biederman801199c2006-10-02 02:18:48 -07002577{
2578 .d_revalidate = proc_base_revalidate,
2579 .d_delete = pid_delete_dentry,
2580};
2581
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002582static struct dentry *proc_base_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002583 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman801199c2006-10-02 02:18:48 -07002584{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002585 const struct pid_entry *p = ptr;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002586 struct inode *inode;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002587 struct proc_inode *ei;
Dan Carpenter73d36462010-05-26 14:43:25 -07002588 struct dentry *error;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002589
2590 /* Allocate the inode */
2591 error = ERR_PTR(-ENOMEM);
2592 inode = new_inode(dir->i_sb);
2593 if (!inode)
2594 goto out;
2595
2596 /* Initialize the inode */
2597 ei = PROC_I(inode);
Christoph Hellwig85fe4022010-10-23 11:19:54 -04002598 inode->i_ino = get_next_ino();
Eric W. Biederman801199c2006-10-02 02:18:48 -07002599 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002600
2601 /*
2602 * grab the reference to the task.
2603 */
Oleg Nesterov1a657f782006-10-02 02:18:59 -07002604 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002605 if (!ei->pid)
2606 goto out_iput;
2607
Eric W. Biederman801199c2006-10-02 02:18:48 -07002608 inode->i_mode = p->mode;
2609 if (S_ISDIR(inode->i_mode))
2610 inode->i_nlink = 2;
2611 if (S_ISLNK(inode->i_mode))
2612 inode->i_size = 64;
2613 if (p->iop)
2614 inode->i_op = p->iop;
2615 if (p->fop)
2616 inode->i_fop = p->fop;
2617 ei->op = p->op;
2618 dentry->d_op = &proc_base_dentry_operations;
2619 d_add(dentry, inode);
2620 error = NULL;
2621out:
Eric W. Biederman801199c2006-10-02 02:18:48 -07002622 return error;
2623out_iput:
2624 iput(inode);
2625 goto out;
2626}
2627
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002628static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2629{
2630 struct dentry *error;
2631 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002632 const struct pid_entry *p, *last;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002633
2634 error = ERR_PTR(-ENOENT);
2635
2636 if (!task)
2637 goto out_no_task;
2638
2639 /* Lookup the directory entry */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002640 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2641 for (p = proc_base_stuff; p <= last; p++) {
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002642 if (p->len != dentry->d_name.len)
2643 continue;
2644 if (!memcmp(dentry->d_name.name, p->name, p->len))
2645 break;
2646 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002647 if (p > last)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002648 goto out;
2649
2650 error = proc_base_instantiate(dir, dentry, task, p);
2651
2652out:
2653 put_task_struct(task);
2654out_no_task:
2655 return error;
2656}
2657
Eric Dumazetc5141e62007-05-08 00:26:15 -07002658static int proc_base_fill_cache(struct file *filp, void *dirent,
2659 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002660{
2661 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2662 proc_base_instantiate, task, p);
2663}
2664
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002665#ifdef CONFIG_TASK_IO_ACCOUNTING
Andrea Righi297c5d92008-07-25 01:48:49 -07002666static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002667{
Andrea Righi940389b2008-07-28 00:48:12 +02002668 struct task_io_accounting acct = task->ioac;
Andrea Righi59954772008-07-27 17:29:15 +02002669 unsigned long flags;
Andrea Righi297c5d92008-07-25 01:48:49 -07002670
Andrea Righi59954772008-07-27 17:29:15 +02002671 if (whole && lock_task_sighand(task, &flags)) {
2672 struct task_struct *t = task;
Andrea Righib2d002d2008-07-26 15:22:27 -07002673
Andrea Righi59954772008-07-27 17:29:15 +02002674 task_io_accounting_add(&acct, &task->signal->ioac);
2675 while_each_thread(task, t)
2676 task_io_accounting_add(&acct, &t->ioac);
Andrea Righi297c5d92008-07-25 01:48:49 -07002677
Andrea Righi59954772008-07-27 17:29:15 +02002678 unlock_task_sighand(task, &flags);
Andrea Righi297c5d92008-07-25 01:48:49 -07002679 }
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002680 return sprintf(buffer,
2681 "rchar: %llu\n"
2682 "wchar: %llu\n"
2683 "syscr: %llu\n"
2684 "syscw: %llu\n"
2685 "read_bytes: %llu\n"
2686 "write_bytes: %llu\n"
2687 "cancelled_write_bytes: %llu\n",
Alexander Beregalov7c443192008-08-05 13:01:34 -07002688 (unsigned long long)acct.rchar,
2689 (unsigned long long)acct.wchar,
2690 (unsigned long long)acct.syscr,
2691 (unsigned long long)acct.syscw,
2692 (unsigned long long)acct.read_bytes,
2693 (unsigned long long)acct.write_bytes,
2694 (unsigned long long)acct.cancelled_write_bytes);
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002695}
Andrea Righi297c5d92008-07-25 01:48:49 -07002696
2697static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2698{
2699 return do_io_accounting(task, buffer, 0);
2700}
2701
2702static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2703{
2704 return do_io_accounting(task, buffer, 1);
2705}
2706#endif /* CONFIG_TASK_IO_ACCOUNTING */
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002707
Kees Cook47830722008-10-06 03:11:58 +04002708static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2709 struct pid *pid, struct task_struct *task)
2710{
2711 seq_printf(m, "%08x\n", task->personality);
2712 return 0;
2713}
2714
Eric W. Biederman801199c2006-10-02 02:18:48 -07002715/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002716 * Thread groups
2717 */
Arjan van de Ven00977a52007-02-12 00:55:34 -08002718static const struct file_operations proc_task_operations;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002719static const struct inode_operations proc_task_inode_operations;
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002720
Eric Dumazetc5141e62007-05-08 00:26:15 -07002721static const struct pid_entry tgid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002722 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2723 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2724 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002725#ifdef CONFIG_NET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002726 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002727#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002728 REG("environ", S_IRUSR, proc_environ_operations),
2729 INF("auxv", S_IRUSR, proc_pid_auxv),
2730 ONE("status", S_IRUGO, proc_pid_status),
2731 ONE("personality", S_IRUSR, proc_pid_personality),
Jiri Olsa3036e7b2010-09-30 15:15:33 -07002732 INF("limits", S_IRUGO, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002733#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002734 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002735#endif
john stultz4614a696b2009-12-14 18:00:05 -08002736 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
Roland McGrathebcb6732008-07-25 19:46:00 -07002737#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002738 INF("syscall", S_IRUSR, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07002739#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002740 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2741 ONE("stat", S_IRUGO, proc_tgid_stat),
2742 ONE("statm", S_IRUGO, proc_pid_statm),
2743 REG("maps", S_IRUGO, proc_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002744#ifdef CONFIG_NUMA
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002745 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002746#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002747 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2748 LNK("cwd", proc_cwd_link),
2749 LNK("root", proc_root_link),
2750 LNK("exe", proc_exe_link),
2751 REG("mounts", S_IRUGO, proc_mounts_operations),
2752 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2753 REG("mountstats", S_IRUSR, proc_mountstats_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08002754#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002755 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2756 REG("smaps", S_IRUGO, proc_smaps_operations),
2757 REG("pagemap", S_IRUSR, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002758#endif
2759#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002760 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002761#endif
2762#ifdef CONFIG_KALLSYMS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002763 INF("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002764#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03002765#ifdef CONFIG_STACKTRACE
2766 ONE("stack", S_IRUSR, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002767#endif
2768#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002769 INF("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002770#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01002771#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002772 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01002773#endif
Paul Menage8793d852007-10-18 23:39:39 -07002774#ifdef CONFIG_PROC_PID_CPUSET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002775 REG("cpuset", S_IRUGO, proc_cpuset_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002776#endif
Paul Menagea4243162007-10-18 23:39:35 -07002777#ifdef CONFIG_CGROUPS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002778 REG("cgroup", S_IRUGO, proc_cgroup_operations),
Paul Menagea4243162007-10-18 23:39:35 -07002779#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002780 INF("oom_score", S_IRUGO, proc_oom_score),
2781 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
David Rientjesa63d83f2010-08-09 17:19:46 -07002782 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002783#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002784 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2785 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002786#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002787#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002788 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002789#endif
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08002790#ifdef CONFIG_ELF_CORE
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002791 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002792#endif
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002793#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002794 INF("io", S_IRUGO, proc_tgid_io_accounting),
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002795#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002796};
2797
2798static int proc_tgid_base_readdir(struct file * filp,
2799 void * dirent, filldir_t filldir)
2800{
2801 return proc_pident_readdir(filp,dirent,filldir,
2802 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2803}
2804
Arjan van de Ven00977a52007-02-12 00:55:34 -08002805static const struct file_operations proc_tgid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002806 .read = generic_read_dir,
2807 .readdir = proc_tgid_base_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002808 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002809};
2810
2811static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002812 return proc_pident_lookup(dir, dentry,
2813 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002814}
2815
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002816static const struct inode_operations proc_tgid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002817 .lookup = proc_tgid_base_lookup,
2818 .getattr = pid_getattr,
2819 .setattr = proc_setattr,
2820};
2821
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002822static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002823{
Eric W. Biederman48e64842006-06-26 00:25:48 -07002824 struct dentry *dentry, *leader, *dir;
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002825 char buf[PROC_NUMBUF];
Eric W. Biederman48e64842006-06-26 00:25:48 -07002826 struct qstr name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002827
Eric W. Biederman48e64842006-06-26 00:25:48 -07002828 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002829 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2830 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002831 if (dentry) {
Sukadev Bhattiprolu29f12ca2009-11-11 14:26:32 -08002832 shrink_dcache_parent(dentry);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002833 d_drop(dentry);
2834 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002835 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002836
Eric W. Biederman48e64842006-06-26 00:25:48 -07002837 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002838 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2839 leader = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002840 if (!leader)
2841 goto out;
2842
2843 name.name = "task";
2844 name.len = strlen(name.name);
2845 dir = d_hash_and_lookup(leader, &name);
2846 if (!dir)
2847 goto out_put_leader;
2848
2849 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002850 name.len = snprintf(buf, sizeof(buf), "%d", pid);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002851 dentry = d_hash_and_lookup(dir, &name);
2852 if (dentry) {
2853 shrink_dcache_parent(dentry);
2854 d_drop(dentry);
2855 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002856 }
Eric W. Biederman48e64842006-06-26 00:25:48 -07002857
2858 dput(dir);
2859out_put_leader:
2860 dput(leader);
2861out:
2862 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002863}
2864
Randy Dunlap0895e912007-10-21 21:00:10 -07002865/**
2866 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2867 * @task: task that should be flushed.
2868 *
2869 * When flushing dentries from proc, one needs to flush them from global
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002870 * proc (proc_mnt) and from all the namespaces' procs this task was seen
Randy Dunlap0895e912007-10-21 21:00:10 -07002871 * in. This call is supposed to do all of this job.
2872 *
2873 * Looks in the dcache for
2874 * /proc/@pid
2875 * /proc/@tgid/task/@pid
2876 * if either directory is present flushes it and all of it'ts children
2877 * from the dcache.
2878 *
2879 * It is safe and reasonable to cache /proc entries for a task until
2880 * that task exits. After that they just clog up the dcache with
2881 * useless entries, possibly causing useful dcache entries to be
2882 * flushed instead. This routine is proved to flush those useless
2883 * dcache entries at process exit time.
2884 *
2885 * NOTE: This routine is just an optimization so it does not guarantee
2886 * that no dcache entries will exist at process exit time it
2887 * just makes it very unlikely that any will persist.
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002888 */
2889
2890void proc_flush_task(struct task_struct *task)
2891{
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002892 int i;
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07002893 struct pid *pid, *tgid;
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002894 struct upid *upid;
2895
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002896 pid = task_pid(task);
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07002897 tgid = task_tgid(task);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002898
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002899 for (i = 0; i <= pid->level; i++) {
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002900 upid = &pid->numbers[i];
2901 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07002902 tgid->numbers[i].nr);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002903 }
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -07002904
2905 upid = &pid->numbers[pid->level];
2906 if (upid->nr == 1)
2907 pid_ns_release_proc(upid->ns);
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002908}
2909
Adrian Bunk9711ef92006-12-06 20:38:31 -08002910static struct dentry *proc_pid_instantiate(struct inode *dir,
2911 struct dentry * dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002912 struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002913{
2914 struct dentry *error = ERR_PTR(-ENOENT);
2915 struct inode *inode;
2916
Eric W. Biederman61a28782006-10-02 02:18:49 -07002917 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002918 if (!inode)
2919 goto out;
2920
2921 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2922 inode->i_op = &proc_tgid_base_inode_operations;
2923 inode->i_fop = &proc_tgid_base_operations;
2924 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07002925
2926 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2927 ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002928
2929 dentry->d_op = &pid_dentry_operations;
2930
2931 d_add(dentry, inode);
2932 /* Close the race of the process dying before we return the dentry */
2933 if (pid_revalidate(dentry, NULL))
2934 error = NULL;
2935out:
2936 return error;
2937}
2938
Linus Torvalds1da177e2005-04-16 15:20:36 -07002939struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2940{
Dan Carpenter73d36462010-05-26 14:43:25 -07002941 struct dentry *result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002942 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002943 unsigned tgid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002944 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002945
Eric W. Biederman801199c2006-10-02 02:18:48 -07002946 result = proc_base_lookup(dir, dentry);
2947 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2948 goto out;
2949
Linus Torvalds1da177e2005-04-16 15:20:36 -07002950 tgid = name_to_int(dentry);
2951 if (tgid == ~0U)
2952 goto out;
2953
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002954 ns = dentry->d_sb->s_fs_info;
Eric W. Biedermande758732006-06-26 00:25:51 -07002955 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002956 task = find_task_by_pid_ns(tgid, ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002957 if (task)
2958 get_task_struct(task);
Eric W. Biedermande758732006-06-26 00:25:51 -07002959 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002960 if (!task)
2961 goto out;
2962
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002963 result = proc_pid_instantiate(dir, dentry, task, NULL);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002964 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002965out:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002966 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002967}
2968
Linus Torvalds1da177e2005-04-16 15:20:36 -07002969/*
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002970 * Find the first task with tgid >= tgid
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002971 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002972 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002973struct tgid_iter {
2974 unsigned int tgid;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002975 struct task_struct *task;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002976};
2977static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2978{
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002979 struct pid *pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002980
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002981 if (iter.task)
2982 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002983 rcu_read_lock();
2984retry:
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002985 iter.task = NULL;
2986 pid = find_ge_pid(iter.tgid, ns);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002987 if (pid) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002988 iter.tgid = pid_nr_ns(pid, ns);
2989 iter.task = pid_task(pid, PIDTYPE_PID);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002990 /* What we to know is if the pid we have find is the
2991 * pid of a thread_group_leader. Testing for task
2992 * being a thread_group_leader is the obvious thing
2993 * todo but there is a window when it fails, due to
2994 * the pid transfer logic in de_thread.
2995 *
2996 * So we perform the straight forward test of seeing
2997 * if the pid we have found is the pid of a thread
2998 * group leader, and don't worry if the task we have
2999 * found doesn't happen to be a thread group leader.
3000 * As we don't care in the case of readdir.
3001 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003002 if (!iter.task || !has_group_leader_pid(iter.task)) {
3003 iter.tgid += 1;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003004 goto retry;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003005 }
3006 get_task_struct(iter.task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003007 }
Eric W. Biederman454cc102006-06-26 00:25:51 -07003008 rcu_read_unlock();
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003009 return iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003010}
3011
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003012#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003013
Eric W. Biederman61a28782006-10-02 02:18:49 -07003014static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003015 struct tgid_iter iter)
Eric W. Biederman61a28782006-10-02 02:18:49 -07003016{
3017 char name[PROC_NUMBUF];
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003018 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003019 return proc_fill_cache(filp, dirent, filldir, name, len,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003020 proc_pid_instantiate, iter.task, NULL);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003021}
3022
Linus Torvalds1da177e2005-04-16 15:20:36 -07003023/* for the /proc/ directory itself, after non-process stuff has been done */
3024int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3025{
Linus Torvalds1da177e2005-04-16 15:20:36 -07003026 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08003027 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003028 struct tgid_iter iter;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003029 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003030
Eric W. Biederman61a28782006-10-02 02:18:49 -07003031 if (!reaper)
3032 goto out_no_task;
3033
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003034 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
Eric Dumazetc5141e62007-05-08 00:26:15 -07003035 const struct pid_entry *p = &proc_base_stuff[nr];
Eric W. Biederman61a28782006-10-02 02:18:49 -07003036 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
Eric W. Biederman801199c2006-10-02 02:18:48 -07003037 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003038 }
3039
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003040 ns = filp->f_dentry->d_sb->s_fs_info;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08003041 iter.task = NULL;
3042 iter.tgid = filp->f_pos - TGID_OFFSET;
3043 for (iter = next_tgid(ns, iter);
3044 iter.task;
3045 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3046 filp->f_pos = iter.tgid + TGID_OFFSET;
3047 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
3048 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003049 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003050 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003051 }
Eric W. Biederman0804ef42006-10-02 02:17:04 -07003052 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3053out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07003054 put_task_struct(reaper);
3055out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07003056 return 0;
3057}
3058
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003059/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003060 * Tasks
3061 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07003062static const struct pid_entry tid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003063 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
Jerome Marchand38355412010-04-27 13:13:06 -07003064 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003065 REG("environ", S_IRUSR, proc_environ_operations),
3066 INF("auxv", S_IRUSR, proc_pid_auxv),
3067 ONE("status", S_IRUGO, proc_pid_status),
3068 ONE("personality", S_IRUSR, proc_pid_personality),
Jiri Olsa3036e7b2010-09-30 15:15:33 -07003069 INF("limits", S_IRUGO, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02003070#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003071 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02003072#endif
john stultz4614a696b2009-12-14 18:00:05 -08003073 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
Roland McGrathebcb6732008-07-25 19:46:00 -07003074#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003075 INF("syscall", S_IRUSR, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07003076#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003077 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3078 ONE("stat", S_IRUGO, proc_tid_stat),
3079 ONE("statm", S_IRUGO, proc_pid_statm),
3080 REG("maps", S_IRUGO, proc_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003081#ifdef CONFIG_NUMA
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003082 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003083#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003084 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3085 LNK("cwd", proc_cwd_link),
3086 LNK("root", proc_root_link),
3087 LNK("exe", proc_exe_link),
3088 REG("mounts", S_IRUGO, proc_mounts_operations),
3089 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08003090#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003091 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3092 REG("smaps", S_IRUGO, proc_smaps_operations),
3093 REG("pagemap", S_IRUSR, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003094#endif
3095#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003096 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003097#endif
3098#ifdef CONFIG_KALLSYMS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003099 INF("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003100#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03003101#ifdef CONFIG_STACKTRACE
3102 ONE("stack", S_IRUSR, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003103#endif
3104#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003105 INF("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003106#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01003107#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003108 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01003109#endif
Paul Menage8793d852007-10-18 23:39:39 -07003110#ifdef CONFIG_PROC_PID_CPUSET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003111 REG("cpuset", S_IRUGO, proc_cpuset_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003112#endif
Paul Menagea4243162007-10-18 23:39:35 -07003113#ifdef CONFIG_CGROUPS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003114 REG("cgroup", S_IRUGO, proc_cgroup_operations),
Paul Menagea4243162007-10-18 23:39:35 -07003115#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003116 INF("oom_score", S_IRUGO, proc_oom_score),
3117 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
David Rientjesa63d83f2010-08-09 17:19:46 -07003118 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003119#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003120 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3121 REG("sessionid", S_IRUSR, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003122#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003123#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003124 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08003125#endif
Andrea Righi297c5d92008-07-25 01:48:49 -07003126#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03003127 INF("io", S_IRUGO, proc_tid_io_accounting),
Andrea Righi297c5d92008-07-25 01:48:49 -07003128#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003129};
3130
3131static int proc_tid_base_readdir(struct file * filp,
3132 void * dirent, filldir_t filldir)
3133{
3134 return proc_pident_readdir(filp,dirent,filldir,
3135 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3136}
3137
3138static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07003139 return proc_pident_lookup(dir, dentry,
3140 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003141}
3142
Arjan van de Ven00977a52007-02-12 00:55:34 -08003143static const struct file_operations proc_tid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003144 .read = generic_read_dir,
3145 .readdir = proc_tid_base_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02003146 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003147};
3148
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003149static const struct inode_operations proc_tid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003150 .lookup = proc_tid_base_lookup,
3151 .getattr = pid_getattr,
3152 .setattr = proc_setattr,
3153};
3154
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003155static struct dentry *proc_task_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07003156 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003157{
3158 struct dentry *error = ERR_PTR(-ENOENT);
3159 struct inode *inode;
Eric W. Biederman61a28782006-10-02 02:18:49 -07003160 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003161
3162 if (!inode)
3163 goto out;
3164 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3165 inode->i_op = &proc_tid_base_inode_operations;
3166 inode->i_fop = &proc_tid_base_operations;
3167 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07003168
3169 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3170 ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003171
3172 dentry->d_op = &pid_dentry_operations;
3173
3174 d_add(dentry, inode);
3175 /* Close the race of the process dying before we return the dentry */
3176 if (pid_revalidate(dentry, NULL))
3177 error = NULL;
3178out:
3179 return error;
3180}
3181
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003182static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3183{
3184 struct dentry *result = ERR_PTR(-ENOENT);
3185 struct task_struct *task;
3186 struct task_struct *leader = get_proc_task(dir);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003187 unsigned tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003188 struct pid_namespace *ns;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003189
3190 if (!leader)
3191 goto out_no_task;
3192
3193 tid = name_to_int(dentry);
3194 if (tid == ~0U)
3195 goto out;
3196
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003197 ns = dentry->d_sb->s_fs_info;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003198 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003199 task = find_task_by_pid_ns(tid, ns);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003200 if (task)
3201 get_task_struct(task);
3202 rcu_read_unlock();
3203 if (!task)
3204 goto out;
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07003205 if (!same_thread_group(leader, task))
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003206 goto out_drop_task;
3207
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003208 result = proc_task_instantiate(dir, dentry, task, NULL);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003209out_drop_task:
3210 put_task_struct(task);
3211out:
3212 put_task_struct(leader);
3213out_no_task:
3214 return result;
3215}
3216
3217/*
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003218 * Find the first tid of a thread group to return to user space.
3219 *
3220 * Usually this is just the thread group leader, but if the users
3221 * buffer was too small or there was a seek into the middle of the
3222 * directory we have more work todo.
3223 *
3224 * In the case of a short read we start with find_task_by_pid.
3225 *
3226 * In the case of a seek we start with the leader and walk nr
3227 * threads past it.
3228 */
Eric W. Biedermancc288732006-06-26 00:26:01 -07003229static struct task_struct *first_tid(struct task_struct *leader,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003230 int tid, int nr, struct pid_namespace *ns)
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003231{
Oleg Nesterova872ff02006-06-26 00:26:01 -07003232 struct task_struct *pos;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003233
Eric W. Biedermancc288732006-06-26 00:26:01 -07003234 rcu_read_lock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003235 /* Attempt to start with the pid of a thread */
3236 if (tid && (nr > 0)) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003237 pos = find_task_by_pid_ns(tid, ns);
Oleg Nesterova872ff02006-06-26 00:26:01 -07003238 if (pos && (pos->group_leader == leader))
3239 goto found;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003240 }
3241
3242 /* If nr exceeds the number of threads there is nothing todo */
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003243 pos = NULL;
Oleg Nesterova872ff02006-06-26 00:26:01 -07003244 if (nr && nr >= get_nr_threads(leader))
3245 goto out;
3246
3247 /* If we haven't found our starting place yet start
3248 * with the leader and walk nr threads forward.
3249 */
3250 for (pos = leader; nr > 0; --nr) {
3251 pos = next_thread(pos);
3252 if (pos == leader) {
3253 pos = NULL;
3254 goto out;
3255 }
3256 }
3257found:
3258 get_task_struct(pos);
3259out:
Eric W. Biedermancc288732006-06-26 00:26:01 -07003260 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003261 return pos;
3262}
3263
3264/*
3265 * Find the next thread in the thread list.
3266 * Return NULL if there is an error or no next thread.
3267 *
3268 * The reference to the input task_struct is released.
3269 */
3270static struct task_struct *next_tid(struct task_struct *start)
3271{
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003272 struct task_struct *pos = NULL;
Eric W. Biedermancc288732006-06-26 00:26:01 -07003273 rcu_read_lock();
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003274 if (pid_alive(start)) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003275 pos = next_thread(start);
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003276 if (thread_group_leader(pos))
3277 pos = NULL;
3278 else
3279 get_task_struct(pos);
3280 }
Eric W. Biedermancc288732006-06-26 00:26:01 -07003281 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003282 put_task_struct(start);
3283 return pos;
3284}
3285
Eric W. Biederman61a28782006-10-02 02:18:49 -07003286static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3287 struct task_struct *task, int tid)
3288{
3289 char name[PROC_NUMBUF];
3290 int len = snprintf(name, sizeof(name), "%d", tid);
3291 return proc_fill_cache(filp, dirent, filldir, name, len,
3292 proc_task_instantiate, task, NULL);
3293}
3294
Linus Torvalds1da177e2005-04-16 15:20:36 -07003295/* for the /proc/TGID/task/ directories */
3296static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3297{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08003298 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003299 struct inode *inode = dentry->d_inode;
Guillaume Chazarain7d895242007-01-31 23:48:14 -08003300 struct task_struct *leader = NULL;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003301 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003302 int retval = -ENOENT;
3303 ino_t ino;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003304 int tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003305 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003306
Guillaume Chazarain7d895242007-01-31 23:48:14 -08003307 task = get_proc_task(inode);
3308 if (!task)
3309 goto out_no_task;
3310 rcu_read_lock();
3311 if (pid_alive(task)) {
3312 leader = task->group_leader;
3313 get_task_struct(leader);
3314 }
3315 rcu_read_unlock();
3316 put_task_struct(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003317 if (!leader)
3318 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003319 retval = 0;
3320
Linus Torvaldsee568b22009-03-17 10:02:35 -07003321 switch ((unsigned long)filp->f_pos) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003322 case 0:
3323 ino = inode->i_ino;
Zhang Leee6f7792009-03-16 14:44:31 +08003324 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003325 goto out;
Zhang Leee6f7792009-03-16 14:44:31 +08003326 filp->f_pos++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003327 /* fall through */
3328 case 1:
3329 ino = parent_ino(dentry);
Zhang Leee6f7792009-03-16 14:44:31 +08003330 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003331 goto out;
Zhang Leee6f7792009-03-16 14:44:31 +08003332 filp->f_pos++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003333 /* fall through */
3334 }
3335
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003336 /* f_version caches the tgid value that the last readdir call couldn't
3337 * return. lseek aka telldir automagically resets f_version to 0.
3338 */
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003339 ns = filp->f_dentry->d_sb->s_fs_info;
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07003340 tid = (int)filp->f_version;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003341 filp->f_version = 0;
Zhang Leee6f7792009-03-16 14:44:31 +08003342 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003343 task;
Zhang Leee6f7792009-03-16 14:44:31 +08003344 task = next_tid(task), filp->f_pos++) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003345 tid = task_pid_nr_ns(task, ns);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003346 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003347 /* returning this tgid failed, save it as the first
3348 * pid for the next readir call */
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07003349 filp->f_version = (u64)tid;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003350 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003351 break;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003352 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003353 }
3354out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07003355 put_task_struct(leader);
3356out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07003357 return retval;
3358}
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003359
3360static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3361{
3362 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07003363 struct task_struct *p = get_proc_task(inode);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003364 generic_fillattr(inode, stat);
3365
Eric W. Biederman99f89552006-06-26 00:25:55 -07003366 if (p) {
Eric W. Biederman99f89552006-06-26 00:25:55 -07003367 stat->nlink += get_nr_threads(p);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003368 put_task_struct(p);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003369 }
3370
3371 return 0;
3372}
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003373
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003374static const struct inode_operations proc_task_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003375 .lookup = proc_task_lookup,
3376 .getattr = proc_task_getattr,
3377 .setattr = proc_setattr,
3378};
3379
Arjan van de Ven00977a52007-02-12 00:55:34 -08003380static const struct file_operations proc_task_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003381 .read = generic_read_dir,
3382 .readdir = proc_task_readdir,
Arnd Bergmann6038f372010-08-15 18:52:59 +02003383 .llseek = default_llseek,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003384};