blob: 18d5cc62d8ed92ebefb2a85cb26642ecc78d796d [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>
Dipankar Sarmab8359962005-09-09 13:04:14 -070066#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070067#include <linux/kallsyms.h>
Ken Chen2ec220e2008-11-10 11:26:08 +030068#include <linux/stacktrace.h>
Neil Hormand85f50d2007-10-18 23:40:37 -070069#include <linux/resource.h>
Kees Cook5096add2007-05-08 00:26:04 -070070#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070071#include <linux/mount.h>
72#include <linux/security.h>
73#include <linux/ptrace.h>
Roland McGrath0d094ef2008-07-25 19:45:49 -070074#include <linux/tracehook.h>
Paul Menagea4243162007-10-18 23:39:35 -070075#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#include <linux/cpuset.h>
77#include <linux/audit.h>
Al Viro5addc5d2005-11-07 17:15:49 -050078#include <linux/poll.h>
Serge E. Hallyn1651e142006-10-02 02:18:08 -070079#include <linux/nsproxy.h>
Alexey Dobriyan8ac773b2006-10-19 23:28:32 -070080#include <linux/oom.h>
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -070081#include <linux/elf.h>
Pavel Emelyanov60347f62007-10-18 23:40:03 -070082#include <linux/pid_namespace.h>
Al Viro5ad4e532009-03-29 19:50:06 -040083#include <linux/fs_struct.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070084#include "internal.h"
85
Eric W. Biederman0f2fe202006-06-26 00:25:46 -070086/* NOTE:
87 * Implementing inode permission operations in /proc is almost
88 * certainly an error. Permission checks need to happen during
89 * each system call not at open time. The reason is that most of
90 * what we wish to check for permissions in /proc varies at runtime.
91 *
92 * The classic example of a problem is opening file descriptors
93 * in /proc for a task before it execs a suid executable.
94 */
95
Linus Torvalds1da177e2005-04-16 15:20:36 -070096struct pid_entry {
Linus Torvalds1da177e2005-04-16 15:20:36 -070097 char *name;
Eric Dumazetc5141e62007-05-08 00:26:15 -070098 int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -070099 mode_t mode;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800100 const struct inode_operations *iop;
Arjan van de Ven00977a52007-02-12 00:55:34 -0800101 const struct file_operations *fop;
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700102 union proc_op op;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103};
104
Eric W. Biederman61a28782006-10-02 02:18:49 -0700105#define NOD(NAME, MODE, IOP, FOP, OP) { \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700106 .name = (NAME), \
Eric Dumazetc5141e62007-05-08 00:26:15 -0700107 .len = sizeof(NAME) - 1, \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700108 .mode = MODE, \
109 .iop = IOP, \
110 .fop = FOP, \
111 .op = OP, \
112}
113
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300114#define DIR(NAME, MODE, iops, fops) \
115 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
116#define LNK(NAME, get_link) \
Eric W. Biederman61a28782006-10-02 02:18:49 -0700117 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700118 &proc_pid_link_inode_operations, NULL, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300119 { .proc_get_link = get_link } )
120#define REG(NAME, MODE, fops) \
121 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
122#define INF(NAME, MODE, read) \
Eric W. Biederman61a28782006-10-02 02:18:49 -0700123 NOD(NAME, (S_IFREG|(MODE)), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700124 NULL, &proc_info_file_operations, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300125 { .proc_read = read } )
126#define ONE(NAME, MODE, show) \
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800127 NOD(NAME, (S_IFREG|(MODE)), \
128 NULL, &proc_single_file_operations, \
Alexey Dobriyan631f9c12008-11-10 01:32:52 +0300129 { .proc_show = show } )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130
Vegard Nossumaed54172008-06-05 22:46:53 -0700131/*
132 * Count the number of hardlinks for the pid_entry table, excluding the .
133 * and .. links.
134 */
135static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
136 unsigned int n)
137{
138 unsigned int i;
139 unsigned int count;
140
141 count = 0;
142 for (i = 0; i < n; ++i) {
143 if (S_ISDIR(entries[i].mode))
144 ++count;
145 }
146
147 return count;
148}
149
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000150static int get_fs_path(struct task_struct *task, struct path *path, bool root)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151{
152 struct fs_struct *fs;
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000153 int result = -ENOENT;
154
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700155 task_lock(task);
156 fs = task->fs;
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000157 if (fs) {
158 read_lock(&fs->lock);
159 *path = root ? fs->root : fs->pwd;
160 path_get(path);
161 read_unlock(&fs->lock);
162 result = 0;
163 }
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700164 task_unlock(task);
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000165 return result;
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700166}
167
Eric W. Biederman99f89552006-06-26 00:25:55 -0700168static int get_nr_threads(struct task_struct *tsk)
169{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700170 unsigned long flags;
171 int count = 0;
172
173 if (lock_task_sighand(tsk, &flags)) {
174 count = atomic_read(&tsk->signal->count);
175 unlock_task_sighand(tsk, &flags);
176 }
177 return count;
178}
179
Jan Blunck3dcd25f2008-02-14 19:38:35 -0800180static int proc_cwd_link(struct inode *inode, struct path *path)
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700181{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700182 struct task_struct *task = get_proc_task(inode);
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700183 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700184
185 if (task) {
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000186 result = get_fs_path(task, path, 0);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700187 put_task_struct(task);
188 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 return result;
190}
191
Jan Blunck3dcd25f2008-02-14 19:38:35 -0800192static int proc_root_link(struct inode *inode, struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700194 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700196
197 if (task) {
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000198 result = get_fs_path(task, path, 1);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700199 put_task_struct(task);
200 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 return result;
202}
203
Roland McGrath638fa202008-04-29 01:01:38 -0700204/*
205 * Return zero if current may access user memory in @task, -error if not.
206 */
207static int check_mem_permission(struct task_struct *task)
208{
209 /*
210 * A task can always look at itself, in case it chooses
211 * to use system calls instead of load instructions.
212 */
213 if (task == current)
214 return 0;
215
216 /*
217 * If current is actively ptrace'ing, and would also be
218 * permitted to freshly attach with ptrace now, permit it.
219 */
Roland McGrath0d094ef2008-07-25 19:45:49 -0700220 if (task_is_stopped_or_traced(task)) {
221 int match;
222 rcu_read_lock();
223 match = (tracehook_tracer_task(task) == current);
224 rcu_read_unlock();
225 if (match && ptrace_may_access(task, PTRACE_MODE_ATTACH))
226 return 0;
227 }
Roland McGrath638fa202008-04-29 01:01:38 -0700228
229 /*
230 * Noone else is allowed.
231 */
232 return -EPERM;
233}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234
Al Viro831830b2008-01-02 14:09:57 +0000235struct mm_struct *mm_for_maps(struct task_struct *task)
236{
Oleg Nesterov704b8362009-07-10 03:27:40 +0200237 struct mm_struct *mm;
Oleg Nesterov00f89d22009-07-10 03:27:38 +0200238
Oleg Nesterov704b8362009-07-10 03:27:40 +0200239 if (mutex_lock_killable(&task->cred_guard_mutex))
240 return NULL;
241
242 mm = get_task_mm(task);
243 if (mm && mm != current->mm &&
244 !ptrace_may_access(task, PTRACE_MODE_READ)) {
245 mmput(mm);
246 mm = NULL;
Oleg Nesterov13f0fea2009-06-23 21:25:32 +0200247 }
Oleg Nesterov704b8362009-07-10 03:27:40 +0200248 mutex_unlock(&task->cred_guard_mutex);
249
Al Viro831830b2008-01-02 14:09:57 +0000250 return mm;
Al Viro831830b2008-01-02 14:09:57 +0000251}
252
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253static int proc_pid_cmdline(struct task_struct *task, char * buffer)
254{
255 int res = 0;
256 unsigned int len;
257 struct mm_struct *mm = get_task_mm(task);
258 if (!mm)
259 goto out;
260 if (!mm->arg_end)
261 goto out_mm; /* Shh! No looking before we're done */
262
263 len = mm->arg_end - mm->arg_start;
264
265 if (len > PAGE_SIZE)
266 len = PAGE_SIZE;
267
268 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
269
270 // If the nul at the end of args has been overwritten, then
271 // assume application is using setproctitle(3).
272 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
273 len = strnlen(buffer, res);
274 if (len < res) {
275 res = len;
276 } else {
277 len = mm->env_end - mm->env_start;
278 if (len > PAGE_SIZE - res)
279 len = PAGE_SIZE - res;
280 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
281 res = strnlen(buffer, res);
282 }
283 }
284out_mm:
285 mmput(mm);
286out:
287 return res;
288}
289
290static int proc_pid_auxv(struct task_struct *task, char *buffer)
291{
292 int res = 0;
293 struct mm_struct *mm = get_task_mm(task);
294 if (mm) {
295 unsigned int nwords = 0;
Hannes Ederdfe6b7d2008-12-30 18:49:13 +0300296 do {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 nwords += 2;
Hannes Ederdfe6b7d2008-12-30 18:49:13 +0300298 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 res = nwords * sizeof(mm->saved_auxv[0]);
300 if (res > PAGE_SIZE)
301 res = PAGE_SIZE;
302 memcpy(buffer, mm->saved_auxv, res);
303 mmput(mm);
304 }
305 return res;
306}
307
308
309#ifdef CONFIG_KALLSYMS
310/*
311 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
312 * Returns the resolved symbol. If that fails, simply return the address.
313 */
314static int proc_pid_wchan(struct task_struct *task, char *buffer)
315{
Alexey Dobriyanffb45122007-05-08 00:28:41 -0700316 unsigned long wchan;
Tejun Heo9281ace2007-07-17 04:03:51 -0700317 char symname[KSYM_NAME_LEN];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318
319 wchan = get_wchan(task);
320
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700321 if (lookup_symbol_name(wchan, symname) < 0)
Jake Edgef83ce3e2009-05-04 12:51:14 -0600322 if (!ptrace_may_access(task, PTRACE_MODE_READ))
323 return 0;
324 else
325 return sprintf(buffer, "%lu", wchan);
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700326 else
327 return sprintf(buffer, "%s", symname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328}
329#endif /* CONFIG_KALLSYMS */
330
Ken Chen2ec220e2008-11-10 11:26:08 +0300331#ifdef CONFIG_STACKTRACE
332
333#define MAX_STACK_TRACE_DEPTH 64
334
335static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
336 struct pid *pid, struct task_struct *task)
337{
338 struct stack_trace trace;
339 unsigned long *entries;
340 int i;
341
342 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
343 if (!entries)
344 return -ENOMEM;
345
346 trace.nr_entries = 0;
347 trace.max_entries = MAX_STACK_TRACE_DEPTH;
348 trace.entries = entries;
349 trace.skip = 0;
350 save_stack_trace_tsk(task, &trace);
351
352 for (i = 0; i < trace.nr_entries; i++) {
353 seq_printf(m, "[<%p>] %pS\n",
354 (void *)entries[i], (void *)entries[i]);
355 }
356 kfree(entries);
357
358 return 0;
359}
360#endif
361
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362#ifdef CONFIG_SCHEDSTATS
363/*
364 * Provides /proc/PID/schedstat
365 */
366static int proc_pid_schedstat(struct task_struct *task, char *buffer)
367{
Balbir Singh172ba842007-07-09 18:52:00 +0200368 return sprintf(buffer, "%llu %llu %lu\n",
Ingo Molnar826e08b2008-12-22 07:37:41 +0100369 (unsigned long long)task->se.sum_exec_runtime,
370 (unsigned long long)task->sched_info.run_delay,
Ingo Molnar2d723762007-10-15 17:00:12 +0200371 task->sched_info.pcount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372}
373#endif
374
Arjan van de Ven97455122008-01-25 21:08:34 +0100375#ifdef CONFIG_LATENCYTOP
376static int lstats_show_proc(struct seq_file *m, void *v)
377{
378 int i;
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800379 struct inode *inode = m->private;
380 struct task_struct *task = get_proc_task(inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100381
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800382 if (!task)
383 return -ESRCH;
384 seq_puts(m, "Latency Top version : v0.1\n");
Arjan van de Ven97455122008-01-25 21:08:34 +0100385 for (i = 0; i < 32; i++) {
386 if (task->latency_record[i].backtrace[0]) {
387 int q;
388 seq_printf(m, "%i %li %li ",
389 task->latency_record[i].count,
390 task->latency_record[i].time,
391 task->latency_record[i].max);
392 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
Hugh Dickins9c246242008-12-09 13:14:27 -0800393 char sym[KSYM_SYMBOL_LEN];
Arjan van de Ven97455122008-01-25 21:08:34 +0100394 char *c;
395 if (!task->latency_record[i].backtrace[q])
396 break;
397 if (task->latency_record[i].backtrace[q] == ULONG_MAX)
398 break;
399 sprint_symbol(sym, task->latency_record[i].backtrace[q]);
400 c = strchr(sym, '+');
401 if (c)
402 *c = 0;
403 seq_printf(m, "%s ", sym);
404 }
405 seq_printf(m, "\n");
406 }
407
408 }
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800409 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100410 return 0;
411}
412
413static int lstats_open(struct inode *inode, struct file *file)
414{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800415 return single_open(file, lstats_show_proc, inode);
Hiroshi Shimamotod6643d12008-02-14 10:27:00 -0800416}
417
Arjan van de Ven97455122008-01-25 21:08:34 +0100418static ssize_t lstats_write(struct file *file, const char __user *buf,
419 size_t count, loff_t *offs)
420{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800421 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100422
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800423 if (!task)
424 return -ESRCH;
Arjan van de Ven97455122008-01-25 21:08:34 +0100425 clear_all_latency_tracing(task);
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800426 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100427
428 return count;
429}
430
431static const struct file_operations proc_lstats_operations = {
432 .open = lstats_open,
433 .read = seq_read,
434 .write = lstats_write,
435 .llseek = seq_lseek,
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800436 .release = single_release,
Arjan van de Ven97455122008-01-25 21:08:34 +0100437};
438
439#endif
440
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441/* The badness from the OOM killer */
442unsigned long badness(struct task_struct *p, unsigned long uptime);
443static int proc_oom_score(struct task_struct *task, char *buffer)
444{
445 unsigned long points;
446 struct timespec uptime;
447
448 do_posix_clock_monotonic_gettime(&uptime);
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700449 read_lock(&tasklist_lock);
KOSAKI Motohiro495789a2009-09-21 17:03:14 -0700450 points = badness(task->group_leader, uptime.tv_sec);
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700451 read_unlock(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 return sprintf(buffer, "%lu\n", points);
453}
454
Neil Hormand85f50d2007-10-18 23:40:37 -0700455struct limit_names {
456 char *name;
457 char *unit;
458};
459
460static const struct limit_names lnames[RLIM_NLIMITS] = {
Kees Cookcff4edb2009-09-22 16:45:32 -0700461 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700462 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
463 [RLIMIT_DATA] = {"Max data size", "bytes"},
464 [RLIMIT_STACK] = {"Max stack size", "bytes"},
465 [RLIMIT_CORE] = {"Max core file size", "bytes"},
466 [RLIMIT_RSS] = {"Max resident set", "bytes"},
467 [RLIMIT_NPROC] = {"Max processes", "processes"},
468 [RLIMIT_NOFILE] = {"Max open files", "files"},
469 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
470 [RLIMIT_AS] = {"Max address space", "bytes"},
471 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
472 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
473 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
474 [RLIMIT_NICE] = {"Max nice priority", NULL},
475 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
Eugene Teo88081172008-02-23 15:23:52 -0800476 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700477};
478
479/* Display limits for a process */
480static int proc_pid_limits(struct task_struct *task, char *buffer)
481{
482 unsigned int i;
483 int count = 0;
484 unsigned long flags;
485 char *bufptr = buffer;
486
487 struct rlimit rlim[RLIM_NLIMITS];
488
Lai Jiangshana6bebbc2008-10-05 00:51:15 +0400489 if (!lock_task_sighand(task, &flags))
Neil Hormand85f50d2007-10-18 23:40:37 -0700490 return 0;
Neil Hormand85f50d2007-10-18 23:40:37 -0700491 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
492 unlock_task_sighand(task, &flags);
Neil Hormand85f50d2007-10-18 23:40:37 -0700493
494 /*
495 * print the file header
496 */
497 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
498 "Limit", "Soft Limit", "Hard Limit", "Units");
499
500 for (i = 0; i < RLIM_NLIMITS; i++) {
501 if (rlim[i].rlim_cur == RLIM_INFINITY)
502 count += sprintf(&bufptr[count], "%-25s %-20s ",
503 lnames[i].name, "unlimited");
504 else
505 count += sprintf(&bufptr[count], "%-25s %-20lu ",
506 lnames[i].name, rlim[i].rlim_cur);
507
508 if (rlim[i].rlim_max == RLIM_INFINITY)
509 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
510 else
511 count += sprintf(&bufptr[count], "%-20lu ",
512 rlim[i].rlim_max);
513
514 if (lnames[i].unit)
515 count += sprintf(&bufptr[count], "%-10s\n",
516 lnames[i].unit);
517 else
518 count += sprintf(&bufptr[count], "\n");
519 }
520
521 return count;
522}
523
Roland McGrathebcb6732008-07-25 19:46:00 -0700524#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
525static int proc_pid_syscall(struct task_struct *task, char *buffer)
526{
527 long nr;
528 unsigned long args[6], sp, pc;
529
530 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
531 return sprintf(buffer, "running\n");
532
533 if (nr < 0)
534 return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
535
536 return sprintf(buffer,
537 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
538 nr,
539 args[0], args[1], args[2], args[3], args[4], args[5],
540 sp, pc);
541}
542#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
543
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544/************************************************************************/
545/* Here the fs part begins */
546/************************************************************************/
547
548/* permission checks */
Eric W. Biederman778c1142006-06-26 00:25:58 -0700549static int proc_fd_access_allowed(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550{
Eric W. Biederman778c1142006-06-26 00:25:58 -0700551 struct task_struct *task;
552 int allowed = 0;
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700553 /* Allow access to a task's file descriptors if it is us or we
554 * may use ptrace attach to the process and find out that
555 * information.
Eric W. Biederman778c1142006-06-26 00:25:58 -0700556 */
557 task = get_proc_task(inode);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700558 if (task) {
Stephen Smalley006ebb42008-05-19 08:32:49 -0400559 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
Eric W. Biederman778c1142006-06-26 00:25:58 -0700560 put_task_struct(task);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700561 }
Eric W. Biederman778c1142006-06-26 00:25:58 -0700562 return allowed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563}
564
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700565static int proc_setattr(struct dentry *dentry, struct iattr *attr)
566{
567 int error;
568 struct inode *inode = dentry->d_inode;
569
570 if (attr->ia_valid & ATTR_MODE)
571 return -EPERM;
572
573 error = inode_change_ok(inode, attr);
John Johansen1e8123f2007-05-08 00:29:41 -0700574 if (!error)
575 error = inode_setattr(inode, attr);
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700576 return error;
577}
578
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800579static const struct inode_operations proc_def_inode_operations = {
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700580 .setattr = proc_setattr,
581};
582
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100583static int mounts_open_common(struct inode *inode, struct file *file,
584 const struct seq_operations *op)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700586 struct task_struct *task = get_proc_task(inode);
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700587 struct nsproxy *nsp;
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800588 struct mnt_namespace *ns = NULL;
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100589 struct path root;
Al Viro5addc5d2005-11-07 17:15:49 -0500590 struct proc_mounts *p;
591 int ret = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592
Eric W. Biederman99f89552006-06-26 00:25:55 -0700593 if (task) {
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700594 rcu_read_lock();
595 nsp = task_nsproxy(task);
596 if (nsp) {
597 ns = nsp->mnt_ns;
Alexey Dobriyan863c4702007-01-26 00:56:53 -0800598 if (ns)
599 get_mnt_ns(ns);
600 }
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700601 rcu_read_unlock();
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000602 if (ns && get_fs_path(task, &root, 1) == 0)
603 ret = 0;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700604 put_task_struct(task);
605 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100607 if (!ns)
608 goto err;
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000609 if (ret)
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100610 goto err_put_ns;
611
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100612 ret = -ENOMEM;
613 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
614 if (!p)
615 goto err_put_path;
616
617 file->private_data = &p->m;
618 ret = seq_open(file, op);
619 if (ret)
620 goto err_free;
621
622 p->m.private = p;
623 p->ns = ns;
624 p->root = root;
625 p->event = ns->event;
626
627 return 0;
628
629 err_free:
630 kfree(p);
631 err_put_path:
632 path_put(&root);
633 err_put_ns:
634 put_mnt_ns(ns);
635 err:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 return ret;
637}
638
639static int mounts_release(struct inode *inode, struct file *file)
640{
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100641 struct proc_mounts *p = file->private_data;
642 path_put(&p->root);
643 put_mnt_ns(p->ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 return seq_release(inode, file);
645}
646
Al Viro5addc5d2005-11-07 17:15:49 -0500647static unsigned mounts_poll(struct file *file, poll_table *wait)
648{
649 struct proc_mounts *p = file->private_data;
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100650 struct mnt_namespace *ns = p->ns;
KOSAKI Motohiro31b07092009-04-09 13:57:59 +0900651 unsigned res = POLLIN | POLLRDNORM;
Al Viro5addc5d2005-11-07 17:15:49 -0500652
653 poll_wait(file, &ns->poll, wait);
654
655 spin_lock(&vfsmount_lock);
656 if (p->event != ns->event) {
657 p->event = ns->event;
KOSAKI Motohiro31b07092009-04-09 13:57:59 +0900658 res |= POLLERR | POLLPRI;
Al Viro5addc5d2005-11-07 17:15:49 -0500659 }
660 spin_unlock(&vfsmount_lock);
661
662 return res;
663}
664
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100665static int mounts_open(struct inode *inode, struct file *file)
666{
667 return mounts_open_common(inode, file, &mounts_op);
668}
669
Arjan van de Ven00977a52007-02-12 00:55:34 -0800670static const struct file_operations proc_mounts_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 .open = mounts_open,
672 .read = seq_read,
673 .llseek = seq_lseek,
674 .release = mounts_release,
Al Viro5addc5d2005-11-07 17:15:49 -0500675 .poll = mounts_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676};
677
Ram Pai2d4d4862008-03-27 13:06:25 +0100678static int mountinfo_open(struct inode *inode, struct file *file)
679{
680 return mounts_open_common(inode, file, &mountinfo_op);
681}
682
683static const struct file_operations proc_mountinfo_operations = {
684 .open = mountinfo_open,
685 .read = seq_read,
686 .llseek = seq_lseek,
687 .release = mounts_release,
688 .poll = mounts_poll,
689};
690
Chuck Leverb4629fe2006-03-20 13:44:12 -0500691static int mountstats_open(struct inode *inode, struct file *file)
692{
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100693 return mounts_open_common(inode, file, &mountstats_op);
Chuck Leverb4629fe2006-03-20 13:44:12 -0500694}
695
Arjan van de Ven00977a52007-02-12 00:55:34 -0800696static const struct file_operations proc_mountstats_operations = {
Chuck Leverb4629fe2006-03-20 13:44:12 -0500697 .open = mountstats_open,
698 .read = seq_read,
699 .llseek = seq_lseek,
700 .release = mounts_release,
701};
702
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
704
705static ssize_t proc_info_read(struct file * file, char __user * buf,
706 size_t count, loff_t *ppos)
707{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800708 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 unsigned long page;
710 ssize_t length;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700711 struct task_struct *task = get_proc_task(inode);
712
713 length = -ESRCH;
714 if (!task)
715 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716
717 if (count > PROC_BLOCK_SIZE)
718 count = PROC_BLOCK_SIZE;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700719
720 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700721 if (!(page = __get_free_page(GFP_TEMPORARY)))
Eric W. Biederman99f89552006-06-26 00:25:55 -0700722 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723
724 length = PROC_I(inode)->op.proc_read(task, (char*)page);
725
726 if (length >= 0)
727 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
728 free_page(page);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700729out:
730 put_task_struct(task);
731out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 return length;
733}
734
Arjan van de Ven00977a52007-02-12 00:55:34 -0800735static const struct file_operations proc_info_file_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 .read = proc_info_read,
737};
738
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800739static int proc_single_show(struct seq_file *m, void *v)
740{
741 struct inode *inode = m->private;
742 struct pid_namespace *ns;
743 struct pid *pid;
744 struct task_struct *task;
745 int ret;
746
747 ns = inode->i_sb->s_fs_info;
748 pid = proc_pid(inode);
749 task = get_pid_task(pid, PIDTYPE_PID);
750 if (!task)
751 return -ESRCH;
752
753 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
754
755 put_task_struct(task);
756 return ret;
757}
758
759static int proc_single_open(struct inode *inode, struct file *filp)
760{
761 int ret;
762 ret = single_open(filp, proc_single_show, NULL);
763 if (!ret) {
764 struct seq_file *m = filp->private_data;
765
766 m->private = inode;
767 }
768 return ret;
769}
770
771static const struct file_operations proc_single_file_operations = {
772 .open = proc_single_open,
773 .read = seq_read,
774 .llseek = seq_lseek,
775 .release = single_release,
776};
777
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778static int mem_open(struct inode* inode, struct file* file)
779{
780 file->private_data = (void*)((long)current->self_exec_id);
781 return 0;
782}
783
784static ssize_t mem_read(struct file * file, char __user * buf,
785 size_t count, loff_t *ppos)
786{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800787 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 char *page;
789 unsigned long src = *ppos;
790 int ret = -ESRCH;
791 struct mm_struct *mm;
792
Eric W. Biederman99f89552006-06-26 00:25:55 -0700793 if (!task)
794 goto out_no_task;
795
Roland McGrath638fa202008-04-29 01:01:38 -0700796 if (check_mem_permission(task))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 goto out;
798
799 ret = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700800 page = (char *)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 if (!page)
802 goto out;
803
804 ret = 0;
805
806 mm = get_task_mm(task);
807 if (!mm)
808 goto out_free;
809
810 ret = -EIO;
811
812 if (file->private_data != (void*)((long)current->self_exec_id))
813 goto out_put;
814
815 ret = 0;
816
817 while (count > 0) {
818 int this_len, retval;
819
820 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
821 retval = access_process_vm(task, src, page, this_len, 0);
Roland McGrath638fa202008-04-29 01:01:38 -0700822 if (!retval || check_mem_permission(task)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 if (!ret)
824 ret = -EIO;
825 break;
826 }
827
828 if (copy_to_user(buf, page, retval)) {
829 ret = -EFAULT;
830 break;
831 }
832
833 ret += retval;
834 src += retval;
835 buf += retval;
836 count -= retval;
837 }
838 *ppos = src;
839
840out_put:
841 mmput(mm);
842out_free:
843 free_page((unsigned long) page);
844out:
Eric W. Biederman99f89552006-06-26 00:25:55 -0700845 put_task_struct(task);
846out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 return ret;
848}
849
850#define mem_write NULL
851
852#ifndef mem_write
853/* This is a security hazard */
Glauber de Oliveira Costa63967fa2007-02-20 13:58:12 -0800854static ssize_t mem_write(struct file * file, const char __user *buf,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 size_t count, loff_t *ppos)
856{
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700857 int copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 char *page;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800859 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 unsigned long dst = *ppos;
861
Eric W. Biederman99f89552006-06-26 00:25:55 -0700862 copied = -ESRCH;
863 if (!task)
864 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865
Roland McGrath638fa202008-04-29 01:01:38 -0700866 if (check_mem_permission(task))
Eric W. Biederman99f89552006-06-26 00:25:55 -0700867 goto out;
868
869 copied = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700870 page = (char *)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 if (!page)
Eric W. Biederman99f89552006-06-26 00:25:55 -0700872 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700874 copied = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 while (count > 0) {
876 int this_len, retval;
877
878 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
879 if (copy_from_user(page, buf, this_len)) {
880 copied = -EFAULT;
881 break;
882 }
883 retval = access_process_vm(task, dst, page, this_len, 1);
884 if (!retval) {
885 if (!copied)
886 copied = -EIO;
887 break;
888 }
889 copied += retval;
890 buf += retval;
891 dst += retval;
892 count -= retval;
893 }
894 *ppos = dst;
895 free_page((unsigned long) page);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700896out:
897 put_task_struct(task);
898out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 return copied;
900}
901#endif
902
Matt Mackall85863e42008-02-04 22:29:04 -0800903loff_t mem_lseek(struct file *file, loff_t offset, int orig)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904{
905 switch (orig) {
906 case 0:
907 file->f_pos = offset;
908 break;
909 case 1:
910 file->f_pos += offset;
911 break;
912 default:
913 return -EINVAL;
914 }
915 force_successful_syscall_return();
916 return file->f_pos;
917}
918
Arjan van de Ven00977a52007-02-12 00:55:34 -0800919static const struct file_operations proc_mem_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920 .llseek = mem_lseek,
921 .read = mem_read,
922 .write = mem_write,
923 .open = mem_open,
924};
925
James Pearson315e28c2007-10-16 23:30:17 -0700926static ssize_t environ_read(struct file *file, char __user *buf,
927 size_t count, loff_t *ppos)
928{
929 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
930 char *page;
931 unsigned long src = *ppos;
932 int ret = -ESRCH;
933 struct mm_struct *mm;
934
935 if (!task)
936 goto out_no_task;
937
Stephen Smalley006ebb42008-05-19 08:32:49 -0400938 if (!ptrace_may_access(task, PTRACE_MODE_READ))
James Pearson315e28c2007-10-16 23:30:17 -0700939 goto out;
940
941 ret = -ENOMEM;
942 page = (char *)__get_free_page(GFP_TEMPORARY);
943 if (!page)
944 goto out;
945
946 ret = 0;
947
948 mm = get_task_mm(task);
949 if (!mm)
950 goto out_free;
951
952 while (count > 0) {
953 int this_len, retval, max_len;
954
955 this_len = mm->env_end - (mm->env_start + src);
956
957 if (this_len <= 0)
958 break;
959
960 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
961 this_len = (this_len > max_len) ? max_len : this_len;
962
963 retval = access_process_vm(task, (mm->env_start + src),
964 page, this_len, 0);
965
966 if (retval <= 0) {
967 ret = retval;
968 break;
969 }
970
971 if (copy_to_user(buf, page, retval)) {
972 ret = -EFAULT;
973 break;
974 }
975
976 ret += retval;
977 src += retval;
978 buf += retval;
979 count -= retval;
980 }
981 *ppos = src;
982
983 mmput(mm);
984out_free:
985 free_page((unsigned long) page);
986out:
987 put_task_struct(task);
988out_no_task:
989 return ret;
990}
991
992static const struct file_operations proc_environ_operations = {
993 .read = environ_read,
994};
995
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996static ssize_t oom_adjust_read(struct file *file, char __user *buf,
997 size_t count, loff_t *ppos)
998{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800999 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001000 char buffer[PROC_NUMBUF];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 size_t len;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001002 int oom_adjust = OOM_DISABLE;
1003 unsigned long flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004
Eric W. Biederman99f89552006-06-26 00:25:55 -07001005 if (!task)
1006 return -ESRCH;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001007
1008 if (lock_task_sighand(task, &flags)) {
1009 oom_adjust = task->signal->oom_adj;
1010 unlock_task_sighand(task, &flags);
1011 }
1012
Eric W. Biederman99f89552006-06-26 00:25:55 -07001013 put_task_struct(task);
1014
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001015 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001016
1017 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018}
1019
1020static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1021 size_t count, loff_t *ppos)
1022{
Eric W. Biederman99f89552006-06-26 00:25:55 -07001023 struct task_struct *task;
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -07001024 char buffer[PROC_NUMBUF];
1025 long oom_adjust;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001026 unsigned long flags;
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -07001027 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001029 memset(buffer, 0, sizeof(buffer));
1030 if (count > sizeof(buffer) - 1)
1031 count = sizeof(buffer) - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 if (copy_from_user(buffer, buf, count))
1033 return -EFAULT;
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -07001034
1035 err = strict_strtol(strstrip(buffer), 0, &oom_adjust);
1036 if (err)
1037 return -EINVAL;
Alexey Dobriyan8ac773b2006-10-19 23:28:32 -07001038 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1039 oom_adjust != OOM_DISABLE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 return -EINVAL;
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);
Eric W. Biederman99f89552006-06-26 00:25:55 -07001043 if (!task)
1044 return -ESRCH;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001045 if (!lock_task_sighand(task, &flags)) {
1046 put_task_struct(task);
1047 return -ESRCH;
1048 }
1049
1050 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
1051 unlock_task_sighand(task, &flags);
Guillem Jover8fb4fc62006-12-06 20:32:24 -08001052 put_task_struct(task);
1053 return -EACCES;
1054 }
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001055
1056 task->signal->oom_adj = oom_adjust;
1057
1058 unlock_task_sighand(task, &flags);
Eric W. Biederman99f89552006-06-26 00:25:55 -07001059 put_task_struct(task);
KOSAKI Motohiro5d863b82009-09-21 17:03:16 -07001060
1061 return count;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062}
1063
Arjan van de Ven00977a52007-02-12 00:55:34 -08001064static const struct file_operations proc_oom_adjust_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 .read = oom_adjust_read,
1066 .write = oom_adjust_write,
1067};
1068
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069#ifdef CONFIG_AUDITSYSCALL
1070#define TMPBUFLEN 21
1071static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1072 size_t count, loff_t *ppos)
1073{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001074 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001075 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 ssize_t length;
1077 char tmpbuf[TMPBUFLEN];
1078
Eric W. Biederman99f89552006-06-26 00:25:55 -07001079 if (!task)
1080 return -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
Al Viro0c11b942008-01-10 04:20:52 -05001082 audit_get_loginuid(task));
Eric W. Biederman99f89552006-06-26 00:25:55 -07001083 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1085}
1086
1087static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1088 size_t count, loff_t *ppos)
1089{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001090 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 char *page, *tmp;
1092 ssize_t length;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093 uid_t loginuid;
1094
1095 if (!capable(CAP_AUDIT_CONTROL))
1096 return -EPERM;
1097
Eric W. Biederman13b41b02006-06-26 00:25:56 -07001098 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 return -EPERM;
1100
Al Viroe0182902006-05-18 08:28:02 -04001101 if (count >= PAGE_SIZE)
1102 count = PAGE_SIZE - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103
1104 if (*ppos != 0) {
1105 /* No partial writes. */
1106 return -EINVAL;
1107 }
Mel Gormane12ba742007-10-16 01:25:52 -07001108 page = (char*)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 if (!page)
1110 return -ENOMEM;
1111 length = -EFAULT;
1112 if (copy_from_user(page, buf, count))
1113 goto out_free_page;
1114
Al Viroe0182902006-05-18 08:28:02 -04001115 page[count] = '\0';
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 loginuid = simple_strtoul(page, &tmp, 10);
1117 if (tmp == page) {
1118 length = -EINVAL;
1119 goto out_free_page;
1120
1121 }
Eric W. Biederman99f89552006-06-26 00:25:55 -07001122 length = audit_set_loginuid(current, loginuid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 if (likely(length == 0))
1124 length = count;
1125
1126out_free_page:
1127 free_page((unsigned long) page);
1128 return length;
1129}
1130
Arjan van de Ven00977a52007-02-12 00:55:34 -08001131static const struct file_operations proc_loginuid_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 .read = proc_loginuid_read,
1133 .write = proc_loginuid_write,
1134};
Eric Paris1e0bd752008-03-13 08:15:31 -04001135
1136static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1137 size_t count, loff_t *ppos)
1138{
1139 struct inode * inode = file->f_path.dentry->d_inode;
1140 struct task_struct *task = get_proc_task(inode);
1141 ssize_t length;
1142 char tmpbuf[TMPBUFLEN];
1143
1144 if (!task)
1145 return -ESRCH;
1146 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1147 audit_get_sessionid(task));
1148 put_task_struct(task);
1149 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1150}
1151
1152static const struct file_operations proc_sessionid_operations = {
1153 .read = proc_sessionid_read,
1154};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155#endif
1156
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001157#ifdef CONFIG_FAULT_INJECTION
1158static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1159 size_t count, loff_t *ppos)
1160{
1161 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1162 char buffer[PROC_NUMBUF];
1163 size_t len;
1164 int make_it_fail;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001165
1166 if (!task)
1167 return -ESRCH;
1168 make_it_fail = task->make_it_fail;
1169 put_task_struct(task);
1170
1171 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001172
1173 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001174}
1175
1176static ssize_t proc_fault_inject_write(struct file * file,
1177 const char __user * buf, size_t count, loff_t *ppos)
1178{
1179 struct task_struct *task;
1180 char buffer[PROC_NUMBUF], *end;
1181 int make_it_fail;
1182
1183 if (!capable(CAP_SYS_RESOURCE))
1184 return -EPERM;
1185 memset(buffer, 0, sizeof(buffer));
1186 if (count > sizeof(buffer) - 1)
1187 count = sizeof(buffer) - 1;
1188 if (copy_from_user(buffer, buf, count))
1189 return -EFAULT;
Vincent Licba8aaf2009-09-22 16:45:38 -07001190 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1191 if (*end)
1192 return -EINVAL;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001193 task = get_proc_task(file->f_dentry->d_inode);
1194 if (!task)
1195 return -ESRCH;
1196 task->make_it_fail = make_it_fail;
1197 put_task_struct(task);
Vincent Licba8aaf2009-09-22 16:45:38 -07001198
1199 return count;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001200}
1201
Arjan van de Ven00977a52007-02-12 00:55:34 -08001202static const struct file_operations proc_fault_inject_operations = {
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001203 .read = proc_fault_inject_read,
1204 .write = proc_fault_inject_write,
1205};
1206#endif
1207
Arjan van de Ven97455122008-01-25 21:08:34 +01001208
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001209#ifdef CONFIG_SCHED_DEBUG
1210/*
1211 * Print out various scheduling related per-task fields:
1212 */
1213static int sched_show(struct seq_file *m, void *v)
1214{
1215 struct inode *inode = m->private;
1216 struct task_struct *p;
1217
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001218 p = get_proc_task(inode);
1219 if (!p)
1220 return -ESRCH;
1221 proc_sched_show_task(p, m);
1222
1223 put_task_struct(p);
1224
1225 return 0;
1226}
1227
1228static ssize_t
1229sched_write(struct file *file, const char __user *buf,
1230 size_t count, loff_t *offset)
1231{
1232 struct inode *inode = file->f_path.dentry->d_inode;
1233 struct task_struct *p;
1234
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001235 p = get_proc_task(inode);
1236 if (!p)
1237 return -ESRCH;
1238 proc_sched_set_task(p);
1239
1240 put_task_struct(p);
1241
1242 return count;
1243}
1244
1245static int sched_open(struct inode *inode, struct file *filp)
1246{
1247 int ret;
1248
1249 ret = single_open(filp, sched_show, NULL);
1250 if (!ret) {
1251 struct seq_file *m = filp->private_data;
1252
1253 m->private = inode;
1254 }
1255 return ret;
1256}
1257
1258static const struct file_operations proc_pid_sched_operations = {
1259 .open = sched_open,
1260 .read = seq_read,
1261 .write = sched_write,
1262 .llseek = seq_lseek,
Alexey Dobriyan5ea473a2007-07-31 00:38:50 -07001263 .release = single_release,
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001264};
1265
1266#endif
1267
john stultz4614a696b2009-12-14 18:00:05 -08001268static ssize_t comm_write(struct file *file, const char __user *buf,
1269 size_t count, loff_t *offset)
1270{
1271 struct inode *inode = file->f_path.dentry->d_inode;
1272 struct task_struct *p;
1273 char buffer[TASK_COMM_LEN];
1274
1275 memset(buffer, 0, sizeof(buffer));
1276 if (count > sizeof(buffer) - 1)
1277 count = sizeof(buffer) - 1;
1278 if (copy_from_user(buffer, buf, count))
1279 return -EFAULT;
1280
1281 p = get_proc_task(inode);
1282 if (!p)
1283 return -ESRCH;
1284
1285 if (same_thread_group(current, p))
1286 set_task_comm(p, buffer);
1287 else
1288 count = -EINVAL;
1289
1290 put_task_struct(p);
1291
1292 return count;
1293}
1294
1295static int comm_show(struct seq_file *m, void *v)
1296{
1297 struct inode *inode = m->private;
1298 struct task_struct *p;
1299
1300 p = get_proc_task(inode);
1301 if (!p)
1302 return -ESRCH;
1303
1304 task_lock(p);
1305 seq_printf(m, "%s\n", p->comm);
1306 task_unlock(p);
1307
1308 put_task_struct(p);
1309
1310 return 0;
1311}
1312
1313static int comm_open(struct inode *inode, struct file *filp)
1314{
1315 int ret;
1316
1317 ret = single_open(filp, comm_show, NULL);
1318 if (!ret) {
1319 struct seq_file *m = filp->private_data;
1320
1321 m->private = inode;
1322 }
1323 return ret;
1324}
1325
1326static const struct file_operations proc_pid_set_comm_operations = {
1327 .open = comm_open,
1328 .read = seq_read,
1329 .write = comm_write,
1330 .llseek = seq_lseek,
1331 .release = single_release,
1332};
1333
Matt Helsley925d1c42008-04-29 01:01:36 -07001334/*
1335 * We added or removed a vma mapping the executable. The vmas are only mapped
1336 * during exec and are not mapped with the mmap system call.
1337 * Callers must hold down_write() on the mm's mmap_sem for these
1338 */
1339void added_exe_file_vma(struct mm_struct *mm)
1340{
1341 mm->num_exe_file_vmas++;
1342}
1343
1344void removed_exe_file_vma(struct mm_struct *mm)
1345{
1346 mm->num_exe_file_vmas--;
1347 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1348 fput(mm->exe_file);
1349 mm->exe_file = NULL;
1350 }
1351
1352}
1353
1354void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1355{
1356 if (new_exe_file)
1357 get_file(new_exe_file);
1358 if (mm->exe_file)
1359 fput(mm->exe_file);
1360 mm->exe_file = new_exe_file;
1361 mm->num_exe_file_vmas = 0;
1362}
1363
1364struct file *get_mm_exe_file(struct mm_struct *mm)
1365{
1366 struct file *exe_file;
1367
1368 /* We need mmap_sem to protect against races with removal of
1369 * VM_EXECUTABLE vmas */
1370 down_read(&mm->mmap_sem);
1371 exe_file = mm->exe_file;
1372 if (exe_file)
1373 get_file(exe_file);
1374 up_read(&mm->mmap_sem);
1375 return exe_file;
1376}
1377
1378void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1379{
1380 /* It's safe to write the exe_file pointer without exe_file_lock because
1381 * this is called during fork when the task is not yet in /proc */
1382 newmm->exe_file = get_mm_exe_file(oldmm);
1383}
1384
1385static int proc_exe_link(struct inode *inode, struct path *exe_path)
1386{
1387 struct task_struct *task;
1388 struct mm_struct *mm;
1389 struct file *exe_file;
1390
1391 task = get_proc_task(inode);
1392 if (!task)
1393 return -ENOENT;
1394 mm = get_task_mm(task);
1395 put_task_struct(task);
1396 if (!mm)
1397 return -ENOENT;
1398 exe_file = get_mm_exe_file(mm);
1399 mmput(mm);
1400 if (exe_file) {
1401 *exe_path = exe_file->f_path;
1402 path_get(&exe_file->f_path);
1403 fput(exe_file);
1404 return 0;
1405 } else
1406 return -ENOENT;
1407}
1408
Al Viro008b1502005-08-20 00:17:39 +01001409static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410{
1411 struct inode *inode = dentry->d_inode;
1412 int error = -EACCES;
1413
1414 /* We don't need a base pointer in the /proc filesystem */
Jan Blunck1d957f92008-02-14 19:34:35 -08001415 path_put(&nd->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416
Eric W. Biederman778c1142006-06-26 00:25:58 -07001417 /* Are we allowed to snoop on the tasks file descriptors? */
1418 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001421 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 nd->last_type = LAST_BIND;
1423out:
Al Viro008b1502005-08-20 00:17:39 +01001424 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425}
1426
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001427static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428{
Mel Gormane12ba742007-10-16 01:25:52 -07001429 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001430 char *pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431 int len;
1432
1433 if (!tmp)
1434 return -ENOMEM;
Akinobu Mita0c28f282007-05-08 00:31:41 -07001435
Jan Blunckcf28b482008-02-14 19:38:44 -08001436 pathname = d_path(path, tmp, PAGE_SIZE);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001437 len = PTR_ERR(pathname);
1438 if (IS_ERR(pathname))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 goto out;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001440 len = tmp + PAGE_SIZE - 1 - pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441
1442 if (len > buflen)
1443 len = buflen;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001444 if (copy_to_user(buffer, pathname, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445 len = -EFAULT;
1446 out:
1447 free_page((unsigned long)tmp);
1448 return len;
1449}
1450
1451static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1452{
1453 int error = -EACCES;
1454 struct inode *inode = dentry->d_inode;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001455 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456
Eric W. Biederman778c1142006-06-26 00:25:58 -07001457 /* Are we allowed to snoop on the tasks file descriptors? */
1458 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001461 error = PROC_I(inode)->op.proc_get_link(inode, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 if (error)
1463 goto out;
1464
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001465 error = do_proc_readlink(&path, buffer, buflen);
1466 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468 return error;
1469}
1470
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08001471static const struct inode_operations proc_pid_link_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 .readlink = proc_pid_readlink,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001473 .follow_link = proc_pid_follow_link,
1474 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475};
1476
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001477
1478/* building an inode */
1479
1480static int task_dumpable(struct task_struct *task)
1481{
1482 int dumpable = 0;
1483 struct mm_struct *mm;
1484
1485 task_lock(task);
1486 mm = task->mm;
1487 if (mm)
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001488 dumpable = get_dumpable(mm);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001489 task_unlock(task);
1490 if(dumpable == 1)
1491 return 1;
1492 return 0;
1493}
1494
1495
Eric W. Biederman61a28782006-10-02 02:18:49 -07001496static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001497{
1498 struct inode * inode;
1499 struct proc_inode *ei;
David Howellsc69e8d92008-11-14 10:39:19 +11001500 const struct cred *cred;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001501
1502 /* We need a new inode */
1503
1504 inode = new_inode(sb);
1505 if (!inode)
1506 goto out;
1507
1508 /* Common stuff */
1509 ei = PROC_I(inode);
1510 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001511 inode->i_op = &proc_def_inode_operations;
1512
1513 /*
1514 * grab the reference to task.
1515 */
Oleg Nesterov1a657f782006-10-02 02:18:59 -07001516 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001517 if (!ei->pid)
1518 goto out_unlock;
1519
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001520 if (task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001521 rcu_read_lock();
1522 cred = __task_cred(task);
1523 inode->i_uid = cred->euid;
1524 inode->i_gid = cred->egid;
1525 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001526 }
1527 security_task_to_inode(task, inode);
1528
1529out:
1530 return inode;
1531
1532out_unlock:
1533 iput(inode);
1534 return NULL;
1535}
1536
1537static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1538{
1539 struct inode *inode = dentry->d_inode;
1540 struct task_struct *task;
David Howellsc69e8d92008-11-14 10:39:19 +11001541 const struct cred *cred;
1542
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001543 generic_fillattr(inode, stat);
1544
1545 rcu_read_lock();
1546 stat->uid = 0;
1547 stat->gid = 0;
1548 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1549 if (task) {
1550 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1551 task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001552 cred = __task_cred(task);
1553 stat->uid = cred->euid;
1554 stat->gid = cred->egid;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001555 }
1556 }
1557 rcu_read_unlock();
1558 return 0;
1559}
1560
1561/* dentry stuff */
1562
1563/*
1564 * Exceptional case: normally we are not allowed to unhash a busy
1565 * directory. In this case, however, we can do it - no aliasing problems
1566 * due to the way we treat inodes.
1567 *
1568 * Rewrite the inode's ownerships here because the owning task may have
1569 * performed a setuid(), etc.
1570 *
1571 * Before the /proc/pid/status file was created the only way to read
1572 * the effective uid of a /process was to stat /proc/pid. Reading
1573 * /proc/pid/status is slow enough that procps and other packages
1574 * kept stating /proc/pid. To keep the rules in /proc simple I have
1575 * made this apply to all per process world readable and executable
1576 * directories.
1577 */
1578static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1579{
1580 struct inode *inode = dentry->d_inode;
1581 struct task_struct *task = get_proc_task(inode);
David Howellsc69e8d92008-11-14 10:39:19 +11001582 const struct cred *cred;
1583
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001584 if (task) {
1585 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1586 task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001587 rcu_read_lock();
1588 cred = __task_cred(task);
1589 inode->i_uid = cred->euid;
1590 inode->i_gid = cred->egid;
1591 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001592 } else {
1593 inode->i_uid = 0;
1594 inode->i_gid = 0;
1595 }
1596 inode->i_mode &= ~(S_ISUID | S_ISGID);
1597 security_task_to_inode(task, inode);
1598 put_task_struct(task);
1599 return 1;
1600 }
1601 d_drop(dentry);
1602 return 0;
1603}
1604
1605static int pid_delete_dentry(struct dentry * dentry)
1606{
1607 /* Is the task we represent dead?
1608 * If so, then don't put the dentry on the lru list,
1609 * kill it immediately.
1610 */
1611 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1612}
1613
Al Virod72f71e2009-02-20 05:58:47 +00001614static const struct dentry_operations pid_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001615{
1616 .d_revalidate = pid_revalidate,
1617 .d_delete = pid_delete_dentry,
1618};
1619
1620/* Lookups */
1621
Eric Dumazetc5141e62007-05-08 00:26:15 -07001622typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1623 struct task_struct *, const void *);
Eric W. Biederman61a28782006-10-02 02:18:49 -07001624
Eric W. Biederman1c0d04c2006-10-02 02:18:57 -07001625/*
1626 * Fill a directory entry.
1627 *
1628 * If possible create the dcache entry and derive our inode number and
1629 * file type from dcache entry.
1630 *
1631 * Since all of the proc inode numbers are dynamically generated, the inode
1632 * numbers do not exist until the inode is cache. This means creating the
1633 * the dcache entry in readdir is necessary to keep the inode numbers
1634 * reported by readdir in sync with the inode numbers reported
1635 * by stat.
1636 */
Eric W. Biederman61a28782006-10-02 02:18:49 -07001637static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1638 char *name, int len,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001639 instantiate_t instantiate, struct task_struct *task, const void *ptr)
Eric W. Biederman61a28782006-10-02 02:18:49 -07001640{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001641 struct dentry *child, *dir = filp->f_path.dentry;
Eric W. Biederman61a28782006-10-02 02:18:49 -07001642 struct inode *inode;
1643 struct qstr qname;
1644 ino_t ino = 0;
1645 unsigned type = DT_UNKNOWN;
1646
1647 qname.name = name;
1648 qname.len = len;
1649 qname.hash = full_name_hash(name, len);
1650
1651 child = d_lookup(dir, &qname);
1652 if (!child) {
1653 struct dentry *new;
1654 new = d_alloc(dir, &qname);
1655 if (new) {
1656 child = instantiate(dir->d_inode, new, task, ptr);
1657 if (child)
1658 dput(new);
1659 else
1660 child = new;
1661 }
1662 }
1663 if (!child || IS_ERR(child) || !child->d_inode)
1664 goto end_instantiate;
1665 inode = child->d_inode;
1666 if (inode) {
1667 ino = inode->i_ino;
1668 type = inode->i_mode >> 12;
1669 }
1670 dput(child);
1671end_instantiate:
1672 if (!ino)
1673 ino = find_inode_number(dir, &qname);
1674 if (!ino)
1675 ino = 1;
1676 return filldir(dirent, name, len, filp->f_pos, ino, type);
1677}
1678
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001679static unsigned name_to_int(struct dentry *dentry)
1680{
1681 const char *name = dentry->d_name.name;
1682 int len = dentry->d_name.len;
1683 unsigned n = 0;
1684
1685 if (len > 1 && *name == '0')
1686 goto out;
1687 while (len-- > 0) {
1688 unsigned c = *name++ - '0';
1689 if (c > 9)
1690 goto out;
1691 if (n >= (~0U-9)/10)
1692 goto out;
1693 n *= 10;
1694 n += c;
1695 }
1696 return n;
1697out:
1698 return ~0U;
1699}
1700
Miklos Szeredi27932742007-05-08 00:26:17 -07001701#define PROC_FDINFO_MAX 64
1702
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001703static int proc_fd_info(struct inode *inode, struct path *path, char *info)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001704{
1705 struct task_struct *task = get_proc_task(inode);
1706 struct files_struct *files = NULL;
1707 struct file *file;
1708 int fd = proc_fd(inode);
1709
1710 if (task) {
1711 files = get_files_struct(task);
1712 put_task_struct(task);
1713 }
1714 if (files) {
1715 /*
1716 * We are not taking a ref to the file structure, so we must
1717 * hold ->file_lock.
1718 */
1719 spin_lock(&files->file_lock);
1720 file = fcheck_files(files, fd);
1721 if (file) {
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001722 if (path) {
1723 *path = file->f_path;
1724 path_get(&file->f_path);
1725 }
Miklos Szeredi27932742007-05-08 00:26:17 -07001726 if (info)
1727 snprintf(info, PROC_FDINFO_MAX,
1728 "pos:\t%lli\n"
1729 "flags:\t0%o\n",
1730 (long long) file->f_pos,
1731 file->f_flags);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001732 spin_unlock(&files->file_lock);
1733 put_files_struct(files);
1734 return 0;
1735 }
1736 spin_unlock(&files->file_lock);
1737 put_files_struct(files);
1738 }
1739 return -ENOENT;
1740}
1741
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001742static int proc_fd_link(struct inode *inode, struct path *path)
Miklos Szeredi27932742007-05-08 00:26:17 -07001743{
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001744 return proc_fd_info(inode, path, NULL);
Miklos Szeredi27932742007-05-08 00:26:17 -07001745}
1746
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001747static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1748{
1749 struct inode *inode = dentry->d_inode;
1750 struct task_struct *task = get_proc_task(inode);
1751 int fd = proc_fd(inode);
1752 struct files_struct *files;
David Howellsc69e8d92008-11-14 10:39:19 +11001753 const struct cred *cred;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001754
1755 if (task) {
1756 files = get_files_struct(task);
1757 if (files) {
1758 rcu_read_lock();
1759 if (fcheck_files(files, fd)) {
1760 rcu_read_unlock();
1761 put_files_struct(files);
1762 if (task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001763 rcu_read_lock();
1764 cred = __task_cred(task);
1765 inode->i_uid = cred->euid;
1766 inode->i_gid = cred->egid;
1767 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001768 } else {
1769 inode->i_uid = 0;
1770 inode->i_gid = 0;
1771 }
1772 inode->i_mode &= ~(S_ISUID | S_ISGID);
1773 security_task_to_inode(task, inode);
1774 put_task_struct(task);
1775 return 1;
1776 }
1777 rcu_read_unlock();
1778 put_files_struct(files);
1779 }
1780 put_task_struct(task);
1781 }
1782 d_drop(dentry);
1783 return 0;
1784}
1785
Al Virod72f71e2009-02-20 05:58:47 +00001786static const struct dentry_operations tid_fd_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001787{
1788 .d_revalidate = tid_fd_revalidate,
1789 .d_delete = pid_delete_dentry,
1790};
1791
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001792static struct dentry *proc_fd_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001793 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001794{
Eric Dumazetc5141e62007-05-08 00:26:15 -07001795 unsigned fd = *(const unsigned *)ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001796 struct file *file;
1797 struct files_struct *files;
1798 struct inode *inode;
1799 struct proc_inode *ei;
1800 struct dentry *error = ERR_PTR(-ENOENT);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001801
Eric W. Biederman61a28782006-10-02 02:18:49 -07001802 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001803 if (!inode)
1804 goto out;
1805 ei = PROC_I(inode);
1806 ei->fd = fd;
1807 files = get_files_struct(task);
1808 if (!files)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001809 goto out_iput;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001810 inode->i_mode = S_IFLNK;
1811
1812 /*
1813 * We are not taking a ref to the file structure, so we must
1814 * hold ->file_lock.
1815 */
1816 spin_lock(&files->file_lock);
1817 file = fcheck_files(files, fd);
1818 if (!file)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001819 goto out_unlock;
Al Viroaeb5d722008-09-02 15:28:45 -04001820 if (file->f_mode & FMODE_READ)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001821 inode->i_mode |= S_IRUSR | S_IXUSR;
Al Viroaeb5d722008-09-02 15:28:45 -04001822 if (file->f_mode & FMODE_WRITE)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001823 inode->i_mode |= S_IWUSR | S_IXUSR;
1824 spin_unlock(&files->file_lock);
1825 put_files_struct(files);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001826
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001827 inode->i_op = &proc_pid_link_inode_operations;
1828 inode->i_size = 64;
1829 ei->op.proc_get_link = proc_fd_link;
1830 dentry->d_op = &tid_fd_dentry_operations;
1831 d_add(dentry, inode);
1832 /* Close the race of the process dying before we return the dentry */
1833 if (tid_fd_revalidate(dentry, NULL))
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001834 error = NULL;
1835
1836 out:
1837 return error;
1838out_unlock:
1839 spin_unlock(&files->file_lock);
1840 put_files_struct(files);
1841out_iput:
1842 iput(inode);
1843 goto out;
1844}
1845
Miklos Szeredi27932742007-05-08 00:26:17 -07001846static struct dentry *proc_lookupfd_common(struct inode *dir,
1847 struct dentry *dentry,
1848 instantiate_t instantiate)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001849{
1850 struct task_struct *task = get_proc_task(dir);
1851 unsigned fd = name_to_int(dentry);
1852 struct dentry *result = ERR_PTR(-ENOENT);
1853
1854 if (!task)
1855 goto out_no_task;
1856 if (fd == ~0U)
1857 goto out;
1858
Miklos Szeredi27932742007-05-08 00:26:17 -07001859 result = instantiate(dir, dentry, task, &fd);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001860out:
1861 put_task_struct(task);
1862out_no_task:
1863 return result;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001864}
1865
Miklos Szeredi27932742007-05-08 00:26:17 -07001866static int proc_readfd_common(struct file * filp, void * dirent,
1867 filldir_t filldir, instantiate_t instantiate)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001869 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman56347082006-06-26 00:25:40 -07001870 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001871 struct task_struct *p = get_proc_task(inode);
Pavel Emelyanov457c2512007-10-18 23:40:43 -07001872 unsigned int fd, ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874 struct files_struct * files;
1875
1876 retval = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001877 if (!p)
1878 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880
1881 fd = filp->f_pos;
1882 switch (fd) {
1883 case 0:
1884 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1885 goto out;
1886 filp->f_pos++;
1887 case 1:
Eric W. Biederman56347082006-06-26 00:25:40 -07001888 ino = parent_ino(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1890 goto out;
1891 filp->f_pos++;
1892 default:
1893 files = get_files_struct(p);
1894 if (!files)
1895 goto out;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001896 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897 for (fd = filp->f_pos-2;
Al Viro9b4f5262008-04-22 01:32:44 -04001898 fd < files_fdtable(files)->max_fds;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899 fd++, filp->f_pos++) {
Miklos Szeredi27932742007-05-08 00:26:17 -07001900 char name[PROC_NUMBUF];
1901 int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001902
1903 if (!fcheck_files(files, fd))
1904 continue;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001905 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906
Miklos Szeredi27932742007-05-08 00:26:17 -07001907 len = snprintf(name, sizeof(name), "%d", fd);
1908 if (proc_fill_cache(filp, dirent, filldir,
1909 name, len, instantiate,
1910 p, &fd) < 0) {
Dipankar Sarmab8359962005-09-09 13:04:14 -07001911 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912 break;
1913 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07001914 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07001916 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917 put_files_struct(files);
1918 }
1919out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07001920 put_task_struct(p);
1921out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922 return retval;
1923}
1924
Miklos Szeredi27932742007-05-08 00:26:17 -07001925static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1926 struct nameidata *nd)
1927{
1928 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1929}
1930
1931static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1932{
1933 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1934}
1935
1936static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1937 size_t len, loff_t *ppos)
1938{
1939 char tmp[PROC_FDINFO_MAX];
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001940 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
Miklos Szeredi27932742007-05-08 00:26:17 -07001941 if (!err)
1942 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1943 return err;
1944}
1945
1946static const struct file_operations proc_fdinfo_file_operations = {
1947 .open = nonseekable_open,
1948 .read = proc_fdinfo_read,
1949};
1950
Arjan van de Ven00977a52007-02-12 00:55:34 -08001951static const struct file_operations proc_fd_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952 .read = generic_read_dir,
1953 .readdir = proc_readfd,
1954};
1955
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956/*
Alexey Dobriyan8948e112007-05-08 00:23:35 -07001957 * /proc/pid/fd needs a special permission handler so that a process can still
1958 * access /proc/self/fd after it has executed a setuid().
1959 */
Al Viroe6305c42008-07-15 21:03:57 -04001960static int proc_fd_permission(struct inode *inode, int mask)
Alexey Dobriyan8948e112007-05-08 00:23:35 -07001961{
1962 int rv;
1963
1964 rv = generic_permission(inode, mask, NULL);
1965 if (rv == 0)
1966 return 0;
1967 if (task_pid(current) == proc_pid(inode))
1968 rv = 0;
1969 return rv;
1970}
1971
1972/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973 * proc directories can do almost nothing..
1974 */
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08001975static const struct inode_operations proc_fd_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976 .lookup = proc_lookupfd,
Alexey Dobriyan8948e112007-05-08 00:23:35 -07001977 .permission = proc_fd_permission,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001978 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979};
1980
Miklos Szeredi27932742007-05-08 00:26:17 -07001981static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1982 struct dentry *dentry, struct task_struct *task, const void *ptr)
1983{
1984 unsigned fd = *(unsigned *)ptr;
1985 struct inode *inode;
1986 struct proc_inode *ei;
1987 struct dentry *error = ERR_PTR(-ENOENT);
1988
1989 inode = proc_pid_make_inode(dir->i_sb, task);
1990 if (!inode)
1991 goto out;
1992 ei = PROC_I(inode);
1993 ei->fd = fd;
1994 inode->i_mode = S_IFREG | S_IRUSR;
1995 inode->i_fop = &proc_fdinfo_file_operations;
1996 dentry->d_op = &tid_fd_dentry_operations;
1997 d_add(dentry, inode);
1998 /* Close the race of the process dying before we return the dentry */
1999 if (tid_fd_revalidate(dentry, NULL))
2000 error = NULL;
2001
2002 out:
2003 return error;
2004}
2005
2006static struct dentry *proc_lookupfdinfo(struct inode *dir,
2007 struct dentry *dentry,
2008 struct nameidata *nd)
2009{
2010 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2011}
2012
2013static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2014{
2015 return proc_readfd_common(filp, dirent, filldir,
2016 proc_fdinfo_instantiate);
2017}
2018
2019static const struct file_operations proc_fdinfo_operations = {
2020 .read = generic_read_dir,
2021 .readdir = proc_readfdinfo,
2022};
2023
2024/*
2025 * proc directories can do almost nothing..
2026 */
2027static const struct inode_operations proc_fdinfo_inode_operations = {
2028 .lookup = proc_lookupfdinfo,
2029 .setattr = proc_setattr,
2030};
2031
2032
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002033static struct dentry *proc_pident_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002034 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002035{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002036 const struct pid_entry *p = ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002037 struct inode *inode;
2038 struct proc_inode *ei;
KOSAKI Motohirobd6daba2009-05-28 14:34:21 -07002039 struct dentry *error = ERR_PTR(-ENOENT);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002040
Eric W. Biederman61a28782006-10-02 02:18:49 -07002041 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002042 if (!inode)
2043 goto out;
2044
2045 ei = PROC_I(inode);
2046 inode->i_mode = p->mode;
2047 if (S_ISDIR(inode->i_mode))
2048 inode->i_nlink = 2; /* Use getattr to fix if necessary */
2049 if (p->iop)
2050 inode->i_op = p->iop;
2051 if (p->fop)
2052 inode->i_fop = p->fop;
2053 ei->op = p->op;
2054 dentry->d_op = &pid_dentry_operations;
2055 d_add(dentry, inode);
2056 /* Close the race of the process dying before we return the dentry */
2057 if (pid_revalidate(dentry, NULL))
2058 error = NULL;
2059out:
2060 return error;
2061}
2062
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063static struct dentry *proc_pident_lookup(struct inode *dir,
2064 struct dentry *dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002065 const struct pid_entry *ents,
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002066 unsigned int nents)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002067{
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002068 struct dentry *error;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002069 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002070 const struct pid_entry *p, *last;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002072 error = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073
Eric W. Biederman99f89552006-06-26 00:25:55 -07002074 if (!task)
2075 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002077 /*
2078 * Yes, it does not scale. And it should not. Don't add
2079 * new entries into /proc/<tgid>/ without very good reasons.
2080 */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002081 last = &ents[nents - 1];
2082 for (p = ents; p <= last; p++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083 if (p->len != dentry->d_name.len)
2084 continue;
2085 if (!memcmp(dentry->d_name.name, p->name, p->len))
2086 break;
2087 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002088 if (p > last)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002089 goto out;
2090
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002091 error = proc_pident_instantiate(dir, dentry, task, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07002093 put_task_struct(task);
2094out_no_task:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002095 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096}
2097
Eric Dumazetc5141e62007-05-08 00:26:15 -07002098static int proc_pident_fill_cache(struct file *filp, void *dirent,
2099 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002100{
2101 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2102 proc_pident_instantiate, task, p);
2103}
2104
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002105static int proc_pident_readdir(struct file *filp,
2106 void *dirent, filldir_t filldir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002107 const struct pid_entry *ents, unsigned int nents)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002108{
2109 int i;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002110 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002111 struct inode *inode = dentry->d_inode;
2112 struct task_struct *task = get_proc_task(inode);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002113 const struct pid_entry *p, *last;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002114 ino_t ino;
2115 int ret;
2116
2117 ret = -ENOENT;
2118 if (!task)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002119 goto out_no_task;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002120
2121 ret = 0;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002122 i = filp->f_pos;
2123 switch (i) {
2124 case 0:
2125 ino = inode->i_ino;
2126 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2127 goto out;
2128 i++;
2129 filp->f_pos++;
2130 /* fall through */
2131 case 1:
2132 ino = parent_ino(dentry);
2133 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2134 goto out;
2135 i++;
2136 filp->f_pos++;
2137 /* fall through */
2138 default:
2139 i -= 2;
2140 if (i >= nents) {
2141 ret = 1;
2142 goto out;
2143 }
2144 p = ents + i;
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002145 last = &ents[nents - 1];
2146 while (p <= last) {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002147 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002148 goto out;
2149 filp->f_pos++;
2150 p++;
2151 }
2152 }
2153
2154 ret = 1;
2155out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07002156 put_task_struct(task);
2157out_no_task:
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002158 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159}
2160
Linus Torvalds1da177e2005-04-16 15:20:36 -07002161#ifdef CONFIG_SECURITY
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002162static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2163 size_t count, loff_t *ppos)
2164{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002165 struct inode * inode = file->f_path.dentry->d_inode;
Al Viro04ff9702007-03-12 16:17:58 +00002166 char *p = NULL;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002167 ssize_t length;
2168 struct task_struct *task = get_proc_task(inode);
2169
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002170 if (!task)
Al Viro04ff9702007-03-12 16:17:58 +00002171 return -ESRCH;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002172
2173 length = security_getprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002174 (char*)file->f_path.dentry->d_name.name,
Al Viro04ff9702007-03-12 16:17:58 +00002175 &p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002176 put_task_struct(task);
Al Viro04ff9702007-03-12 16:17:58 +00002177 if (length > 0)
2178 length = simple_read_from_buffer(buf, count, ppos, p, length);
2179 kfree(p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002180 return length;
2181}
2182
2183static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2184 size_t count, loff_t *ppos)
2185{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002186 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002187 char *page;
2188 ssize_t length;
2189 struct task_struct *task = get_proc_task(inode);
2190
2191 length = -ESRCH;
2192 if (!task)
2193 goto out_no_task;
2194 if (count > PAGE_SIZE)
2195 count = PAGE_SIZE;
2196
2197 /* No partial writes. */
2198 length = -EINVAL;
2199 if (*ppos != 0)
2200 goto out;
2201
2202 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -07002203 page = (char*)__get_free_page(GFP_TEMPORARY);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002204 if (!page)
2205 goto out;
2206
2207 length = -EFAULT;
2208 if (copy_from_user(page, buf, count))
2209 goto out_free;
2210
David Howells107db7c2009-05-08 13:55:27 +01002211 /* Guard against adverse ptrace interaction */
2212 length = mutex_lock_interruptible(&task->cred_guard_mutex);
2213 if (length < 0)
2214 goto out_free;
2215
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002216 length = security_setprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002217 (char*)file->f_path.dentry->d_name.name,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002218 (void*)page, count);
David Howells107db7c2009-05-08 13:55:27 +01002219 mutex_unlock(&task->cred_guard_mutex);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002220out_free:
2221 free_page((unsigned long) page);
2222out:
2223 put_task_struct(task);
2224out_no_task:
2225 return length;
2226}
2227
Arjan van de Ven00977a52007-02-12 00:55:34 -08002228static const struct file_operations proc_pid_attr_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002229 .read = proc_pid_attr_read,
2230 .write = proc_pid_attr_write,
2231};
2232
Eric Dumazetc5141e62007-05-08 00:26:15 -07002233static const struct pid_entry attr_dir_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002234 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2235 REG("prev", S_IRUGO, proc_pid_attr_operations),
2236 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2237 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2238 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2239 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002240};
2241
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002242static int proc_attr_dir_readdir(struct file * filp,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002243 void * dirent, filldir_t filldir)
2244{
2245 return proc_pident_readdir(filp,dirent,filldir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002246 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002247}
2248
Arjan van de Ven00977a52007-02-12 00:55:34 -08002249static const struct file_operations proc_attr_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002250 .read = generic_read_dir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002251 .readdir = proc_attr_dir_readdir,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002252};
2253
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002254static struct dentry *proc_attr_dir_lookup(struct inode *dir,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002255 struct dentry *dentry, struct nameidata *nd)
2256{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002257 return proc_pident_lookup(dir, dentry,
2258 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002259}
2260
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002261static const struct inode_operations proc_attr_dir_inode_operations = {
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002262 .lookup = proc_attr_dir_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07002263 .getattr = pid_getattr,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07002264 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002265};
2266
Linus Torvalds1da177e2005-04-16 15:20:36 -07002267#endif
2268
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08002269#ifdef CONFIG_ELF_CORE
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002270static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2271 size_t count, loff_t *ppos)
2272{
2273 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2274 struct mm_struct *mm;
2275 char buffer[PROC_NUMBUF];
2276 size_t len;
2277 int ret;
2278
2279 if (!task)
2280 return -ESRCH;
2281
2282 ret = 0;
2283 mm = get_task_mm(task);
2284 if (mm) {
2285 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2286 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2287 MMF_DUMP_FILTER_SHIFT));
2288 mmput(mm);
2289 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2290 }
2291
2292 put_task_struct(task);
2293
2294 return ret;
2295}
2296
2297static ssize_t proc_coredump_filter_write(struct file *file,
2298 const char __user *buf,
2299 size_t count,
2300 loff_t *ppos)
2301{
2302 struct task_struct *task;
2303 struct mm_struct *mm;
2304 char buffer[PROC_NUMBUF], *end;
2305 unsigned int val;
2306 int ret;
2307 int i;
2308 unsigned long mask;
2309
2310 ret = -EFAULT;
2311 memset(buffer, 0, sizeof(buffer));
2312 if (count > sizeof(buffer) - 1)
2313 count = sizeof(buffer) - 1;
2314 if (copy_from_user(buffer, buf, count))
2315 goto out_no_task;
2316
2317 ret = -EINVAL;
2318 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2319 if (*end == '\n')
2320 end++;
2321 if (end - buffer == 0)
2322 goto out_no_task;
2323
2324 ret = -ESRCH;
2325 task = get_proc_task(file->f_dentry->d_inode);
2326 if (!task)
2327 goto out_no_task;
2328
2329 ret = end - buffer;
2330 mm = get_task_mm(task);
2331 if (!mm)
2332 goto out_no_mm;
2333
2334 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2335 if (val & mask)
2336 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2337 else
2338 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2339 }
2340
2341 mmput(mm);
2342 out_no_mm:
2343 put_task_struct(task);
2344 out_no_task:
2345 return ret;
2346}
2347
2348static const struct file_operations proc_coredump_filter_operations = {
2349 .read = proc_coredump_filter_read,
2350 .write = proc_coredump_filter_write,
2351};
2352#endif
2353
Linus Torvalds1da177e2005-04-16 15:20:36 -07002354/*
2355 * /proc/self:
2356 */
2357static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2358 int buflen)
2359{
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002360 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002361 pid_t tgid = task_tgid_nr_ns(current, ns);
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002362 char tmp[PROC_NUMBUF];
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002363 if (!tgid)
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002364 return -ENOENT;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002365 sprintf(tmp, "%d", tgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002366 return vfs_readlink(dentry,buffer,buflen,tmp);
2367}
2368
Al Viro008b1502005-08-20 00:17:39 +01002369static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002370{
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002371 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002372 pid_t tgid = task_tgid_nr_ns(current, ns);
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002373 char tmp[PROC_NUMBUF];
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002374 if (!tgid)
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002375 return ERR_PTR(-ENOENT);
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002376 sprintf(tmp, "%d", task_tgid_nr_ns(current, ns));
Al Viro008b1502005-08-20 00:17:39 +01002377 return ERR_PTR(vfs_follow_link(nd,tmp));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002378}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002379
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002380static const struct inode_operations proc_self_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381 .readlink = proc_self_readlink,
2382 .follow_link = proc_self_follow_link,
2383};
2384
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002385/*
Eric W. Biederman801199c2006-10-02 02:18:48 -07002386 * proc base
2387 *
2388 * These are the directory entries in the root directory of /proc
2389 * that properly belong to the /proc filesystem, as they describe
2390 * describe something that is process related.
2391 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07002392static const struct pid_entry proc_base_stuff[] = {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002393 NOD("self", S_IFLNK|S_IRWXUGO,
Eric W. Biederman801199c2006-10-02 02:18:48 -07002394 &proc_self_inode_operations, NULL, {}),
Eric W. Biederman801199c2006-10-02 02:18:48 -07002395};
2396
2397/*
2398 * Exceptional case: normally we are not allowed to unhash a busy
2399 * directory. In this case, however, we can do it - no aliasing problems
2400 * due to the way we treat inodes.
2401 */
2402static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2403{
2404 struct inode *inode = dentry->d_inode;
2405 struct task_struct *task = get_proc_task(inode);
2406 if (task) {
2407 put_task_struct(task);
2408 return 1;
2409 }
2410 d_drop(dentry);
2411 return 0;
2412}
2413
Al Virod72f71e2009-02-20 05:58:47 +00002414static const struct dentry_operations proc_base_dentry_operations =
Eric W. Biederman801199c2006-10-02 02:18:48 -07002415{
2416 .d_revalidate = proc_base_revalidate,
2417 .d_delete = pid_delete_dentry,
2418};
2419
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002420static struct dentry *proc_base_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002421 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman801199c2006-10-02 02:18:48 -07002422{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002423 const struct pid_entry *p = ptr;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002424 struct inode *inode;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002425 struct proc_inode *ei;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002426 struct dentry *error = ERR_PTR(-EINVAL);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002427
2428 /* Allocate the inode */
2429 error = ERR_PTR(-ENOMEM);
2430 inode = new_inode(dir->i_sb);
2431 if (!inode)
2432 goto out;
2433
2434 /* Initialize the inode */
2435 ei = PROC_I(inode);
2436 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002437
2438 /*
2439 * grab the reference to the task.
2440 */
Oleg Nesterov1a657f782006-10-02 02:18:59 -07002441 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002442 if (!ei->pid)
2443 goto out_iput;
2444
Eric W. Biederman801199c2006-10-02 02:18:48 -07002445 inode->i_mode = p->mode;
2446 if (S_ISDIR(inode->i_mode))
2447 inode->i_nlink = 2;
2448 if (S_ISLNK(inode->i_mode))
2449 inode->i_size = 64;
2450 if (p->iop)
2451 inode->i_op = p->iop;
2452 if (p->fop)
2453 inode->i_fop = p->fop;
2454 ei->op = p->op;
2455 dentry->d_op = &proc_base_dentry_operations;
2456 d_add(dentry, inode);
2457 error = NULL;
2458out:
Eric W. Biederman801199c2006-10-02 02:18:48 -07002459 return error;
2460out_iput:
2461 iput(inode);
2462 goto out;
2463}
2464
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002465static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2466{
2467 struct dentry *error;
2468 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002469 const struct pid_entry *p, *last;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002470
2471 error = ERR_PTR(-ENOENT);
2472
2473 if (!task)
2474 goto out_no_task;
2475
2476 /* Lookup the directory entry */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002477 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2478 for (p = proc_base_stuff; p <= last; p++) {
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002479 if (p->len != dentry->d_name.len)
2480 continue;
2481 if (!memcmp(dentry->d_name.name, p->name, p->len))
2482 break;
2483 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002484 if (p > last)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002485 goto out;
2486
2487 error = proc_base_instantiate(dir, dentry, task, p);
2488
2489out:
2490 put_task_struct(task);
2491out_no_task:
2492 return error;
2493}
2494
Eric Dumazetc5141e62007-05-08 00:26:15 -07002495static int proc_base_fill_cache(struct file *filp, void *dirent,
2496 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002497{
2498 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2499 proc_base_instantiate, task, p);
2500}
2501
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002502#ifdef CONFIG_TASK_IO_ACCOUNTING
Andrea Righi297c5d92008-07-25 01:48:49 -07002503static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002504{
Andrea Righi940389b2008-07-28 00:48:12 +02002505 struct task_io_accounting acct = task->ioac;
Andrea Righi59954772008-07-27 17:29:15 +02002506 unsigned long flags;
Andrea Righi297c5d92008-07-25 01:48:49 -07002507
Andrea Righi59954772008-07-27 17:29:15 +02002508 if (whole && lock_task_sighand(task, &flags)) {
2509 struct task_struct *t = task;
Andrea Righib2d002d2008-07-26 15:22:27 -07002510
Andrea Righi59954772008-07-27 17:29:15 +02002511 task_io_accounting_add(&acct, &task->signal->ioac);
2512 while_each_thread(task, t)
2513 task_io_accounting_add(&acct, &t->ioac);
Andrea Righi297c5d92008-07-25 01:48:49 -07002514
Andrea Righi59954772008-07-27 17:29:15 +02002515 unlock_task_sighand(task, &flags);
Andrea Righi297c5d92008-07-25 01:48:49 -07002516 }
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002517 return sprintf(buffer,
2518 "rchar: %llu\n"
2519 "wchar: %llu\n"
2520 "syscr: %llu\n"
2521 "syscw: %llu\n"
2522 "read_bytes: %llu\n"
2523 "write_bytes: %llu\n"
2524 "cancelled_write_bytes: %llu\n",
Alexander Beregalov7c443192008-08-05 13:01:34 -07002525 (unsigned long long)acct.rchar,
2526 (unsigned long long)acct.wchar,
2527 (unsigned long long)acct.syscr,
2528 (unsigned long long)acct.syscw,
2529 (unsigned long long)acct.read_bytes,
2530 (unsigned long long)acct.write_bytes,
2531 (unsigned long long)acct.cancelled_write_bytes);
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002532}
Andrea Righi297c5d92008-07-25 01:48:49 -07002533
2534static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2535{
2536 return do_io_accounting(task, buffer, 0);
2537}
2538
2539static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2540{
2541 return do_io_accounting(task, buffer, 1);
2542}
2543#endif /* CONFIG_TASK_IO_ACCOUNTING */
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002544
Kees Cook47830722008-10-06 03:11:58 +04002545static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2546 struct pid *pid, struct task_struct *task)
2547{
2548 seq_printf(m, "%08x\n", task->personality);
2549 return 0;
2550}
2551
Eric W. Biederman801199c2006-10-02 02:18:48 -07002552/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002553 * Thread groups
2554 */
Arjan van de Ven00977a52007-02-12 00:55:34 -08002555static const struct file_operations proc_task_operations;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002556static const struct inode_operations proc_task_inode_operations;
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002557
Eric Dumazetc5141e62007-05-08 00:26:15 -07002558static const struct pid_entry tgid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002559 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2560 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2561 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002562#ifdef CONFIG_NET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002563 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002564#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002565 REG("environ", S_IRUSR, proc_environ_operations),
2566 INF("auxv", S_IRUSR, proc_pid_auxv),
2567 ONE("status", S_IRUGO, proc_pid_status),
2568 ONE("personality", S_IRUSR, proc_pid_personality),
2569 INF("limits", S_IRUSR, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002570#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002571 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002572#endif
john stultz4614a696b2009-12-14 18:00:05 -08002573 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
Roland McGrathebcb6732008-07-25 19:46:00 -07002574#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002575 INF("syscall", S_IRUSR, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07002576#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002577 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2578 ONE("stat", S_IRUGO, proc_tgid_stat),
2579 ONE("statm", S_IRUGO, proc_pid_statm),
2580 REG("maps", S_IRUGO, proc_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002581#ifdef CONFIG_NUMA
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002582 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002583#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002584 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2585 LNK("cwd", proc_cwd_link),
2586 LNK("root", proc_root_link),
2587 LNK("exe", proc_exe_link),
2588 REG("mounts", S_IRUGO, proc_mounts_operations),
2589 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2590 REG("mountstats", S_IRUSR, proc_mountstats_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08002591#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002592 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2593 REG("smaps", S_IRUGO, proc_smaps_operations),
2594 REG("pagemap", S_IRUSR, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002595#endif
2596#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002597 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002598#endif
2599#ifdef CONFIG_KALLSYMS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002600 INF("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002601#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03002602#ifdef CONFIG_STACKTRACE
2603 ONE("stack", S_IRUSR, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002604#endif
2605#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002606 INF("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002607#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01002608#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002609 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01002610#endif
Paul Menage8793d852007-10-18 23:39:39 -07002611#ifdef CONFIG_PROC_PID_CPUSET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002612 REG("cpuset", S_IRUGO, proc_cpuset_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002613#endif
Paul Menagea4243162007-10-18 23:39:35 -07002614#ifdef CONFIG_CGROUPS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002615 REG("cgroup", S_IRUGO, proc_cgroup_operations),
Paul Menagea4243162007-10-18 23:39:35 -07002616#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002617 INF("oom_score", S_IRUGO, proc_oom_score),
2618 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002619#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002620 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2621 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002622#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002623#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002624 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002625#endif
Christoph Hellwig698ba7b2009-12-15 16:47:37 -08002626#ifdef CONFIG_ELF_CORE
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002627 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002628#endif
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002629#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002630 INF("io", S_IRUGO, proc_tgid_io_accounting),
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002631#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002632};
2633
2634static int proc_tgid_base_readdir(struct file * filp,
2635 void * dirent, filldir_t filldir)
2636{
2637 return proc_pident_readdir(filp,dirent,filldir,
2638 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2639}
2640
Arjan van de Ven00977a52007-02-12 00:55:34 -08002641static const struct file_operations proc_tgid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002642 .read = generic_read_dir,
2643 .readdir = proc_tgid_base_readdir,
2644};
2645
2646static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002647 return proc_pident_lookup(dir, dentry,
2648 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002649}
2650
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002651static const struct inode_operations proc_tgid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002652 .lookup = proc_tgid_base_lookup,
2653 .getattr = pid_getattr,
2654 .setattr = proc_setattr,
2655};
2656
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002657static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002658{
Eric W. Biederman48e64842006-06-26 00:25:48 -07002659 struct dentry *dentry, *leader, *dir;
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002660 char buf[PROC_NUMBUF];
Eric W. Biederman48e64842006-06-26 00:25:48 -07002661 struct qstr name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002662
Eric W. Biederman48e64842006-06-26 00:25:48 -07002663 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002664 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2665 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002666 if (dentry) {
Sukadev Bhattiprolu29f12ca2009-11-11 14:26:32 -08002667 shrink_dcache_parent(dentry);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002668 d_drop(dentry);
2669 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002670 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002671
Eric W. Biederman48e64842006-06-26 00:25:48 -07002672 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002673 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2674 leader = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002675 if (!leader)
2676 goto out;
2677
2678 name.name = "task";
2679 name.len = strlen(name.name);
2680 dir = d_hash_and_lookup(leader, &name);
2681 if (!dir)
2682 goto out_put_leader;
2683
2684 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002685 name.len = snprintf(buf, sizeof(buf), "%d", pid);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002686 dentry = d_hash_and_lookup(dir, &name);
2687 if (dentry) {
2688 shrink_dcache_parent(dentry);
2689 d_drop(dentry);
2690 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002691 }
Eric W. Biederman48e64842006-06-26 00:25:48 -07002692
2693 dput(dir);
2694out_put_leader:
2695 dput(leader);
2696out:
2697 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002698}
2699
Randy Dunlap0895e912007-10-21 21:00:10 -07002700/**
2701 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2702 * @task: task that should be flushed.
2703 *
2704 * When flushing dentries from proc, one needs to flush them from global
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002705 * proc (proc_mnt) and from all the namespaces' procs this task was seen
Randy Dunlap0895e912007-10-21 21:00:10 -07002706 * in. This call is supposed to do all of this job.
2707 *
2708 * Looks in the dcache for
2709 * /proc/@pid
2710 * /proc/@tgid/task/@pid
2711 * if either directory is present flushes it and all of it'ts children
2712 * from the dcache.
2713 *
2714 * It is safe and reasonable to cache /proc entries for a task until
2715 * that task exits. After that they just clog up the dcache with
2716 * useless entries, possibly causing useful dcache entries to be
2717 * flushed instead. This routine is proved to flush those useless
2718 * dcache entries at process exit time.
2719 *
2720 * NOTE: This routine is just an optimization so it does not guarantee
2721 * that no dcache entries will exist at process exit time it
2722 * just makes it very unlikely that any will persist.
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002723 */
2724
2725void proc_flush_task(struct task_struct *task)
2726{
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002727 int i;
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07002728 struct pid *pid, *tgid;
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002729 struct upid *upid;
2730
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002731 pid = task_pid(task);
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07002732 tgid = task_tgid(task);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002733
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002734 for (i = 0; i <= pid->level; i++) {
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002735 upid = &pid->numbers[i];
2736 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
Oleg Nesterov9b4d1cbe2009-09-22 16:45:34 -07002737 tgid->numbers[i].nr);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002738 }
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -07002739
2740 upid = &pid->numbers[pid->level];
2741 if (upid->nr == 1)
2742 pid_ns_release_proc(upid->ns);
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002743}
2744
Adrian Bunk9711ef92006-12-06 20:38:31 -08002745static struct dentry *proc_pid_instantiate(struct inode *dir,
2746 struct dentry * dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002747 struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002748{
2749 struct dentry *error = ERR_PTR(-ENOENT);
2750 struct inode *inode;
2751
Eric W. Biederman61a28782006-10-02 02:18:49 -07002752 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002753 if (!inode)
2754 goto out;
2755
2756 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2757 inode->i_op = &proc_tgid_base_inode_operations;
2758 inode->i_fop = &proc_tgid_base_operations;
2759 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07002760
2761 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2762 ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002763
2764 dentry->d_op = &pid_dentry_operations;
2765
2766 d_add(dentry, inode);
2767 /* Close the race of the process dying before we return the dentry */
2768 if (pid_revalidate(dentry, NULL))
2769 error = NULL;
2770out:
2771 return error;
2772}
2773
Linus Torvalds1da177e2005-04-16 15:20:36 -07002774struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2775{
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002776 struct dentry *result = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002777 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002778 unsigned tgid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002779 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002780
Eric W. Biederman801199c2006-10-02 02:18:48 -07002781 result = proc_base_lookup(dir, dentry);
2782 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2783 goto out;
2784
Linus Torvalds1da177e2005-04-16 15:20:36 -07002785 tgid = name_to_int(dentry);
2786 if (tgid == ~0U)
2787 goto out;
2788
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002789 ns = dentry->d_sb->s_fs_info;
Eric W. Biedermande758732006-06-26 00:25:51 -07002790 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002791 task = find_task_by_pid_ns(tgid, ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002792 if (task)
2793 get_task_struct(task);
Eric W. Biedermande758732006-06-26 00:25:51 -07002794 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002795 if (!task)
2796 goto out;
2797
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002798 result = proc_pid_instantiate(dir, dentry, task, NULL);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002799 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002800out:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002801 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002802}
2803
Linus Torvalds1da177e2005-04-16 15:20:36 -07002804/*
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002805 * Find the first task with tgid >= tgid
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002806 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002807 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002808struct tgid_iter {
2809 unsigned int tgid;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002810 struct task_struct *task;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002811};
2812static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2813{
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002814 struct pid *pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002815
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002816 if (iter.task)
2817 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002818 rcu_read_lock();
2819retry:
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002820 iter.task = NULL;
2821 pid = find_ge_pid(iter.tgid, ns);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002822 if (pid) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002823 iter.tgid = pid_nr_ns(pid, ns);
2824 iter.task = pid_task(pid, PIDTYPE_PID);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002825 /* What we to know is if the pid we have find is the
2826 * pid of a thread_group_leader. Testing for task
2827 * being a thread_group_leader is the obvious thing
2828 * todo but there is a window when it fails, due to
2829 * the pid transfer logic in de_thread.
2830 *
2831 * So we perform the straight forward test of seeing
2832 * if the pid we have found is the pid of a thread
2833 * group leader, and don't worry if the task we have
2834 * found doesn't happen to be a thread group leader.
2835 * As we don't care in the case of readdir.
2836 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002837 if (!iter.task || !has_group_leader_pid(iter.task)) {
2838 iter.tgid += 1;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002839 goto retry;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002840 }
2841 get_task_struct(iter.task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002842 }
Eric W. Biederman454cc102006-06-26 00:25:51 -07002843 rcu_read_unlock();
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002844 return iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002845}
2846
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002847#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002848
Eric W. Biederman61a28782006-10-02 02:18:49 -07002849static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002850 struct tgid_iter iter)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002851{
2852 char name[PROC_NUMBUF];
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002853 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
Eric W. Biederman61a28782006-10-02 02:18:49 -07002854 return proc_fill_cache(filp, dirent, filldir, name, len,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002855 proc_pid_instantiate, iter.task, NULL);
Eric W. Biederman61a28782006-10-02 02:18:49 -07002856}
2857
Linus Torvalds1da177e2005-04-16 15:20:36 -07002858/* for the /proc/ directory itself, after non-process stuff has been done */
2859int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2860{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002861 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002862 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002863 struct tgid_iter iter;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002864 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002865
Eric W. Biederman61a28782006-10-02 02:18:49 -07002866 if (!reaper)
2867 goto out_no_task;
2868
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002869 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
Eric Dumazetc5141e62007-05-08 00:26:15 -07002870 const struct pid_entry *p = &proc_base_stuff[nr];
Eric W. Biederman61a28782006-10-02 02:18:49 -07002871 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
Eric W. Biederman801199c2006-10-02 02:18:48 -07002872 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002873 }
2874
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002875 ns = filp->f_dentry->d_sb->s_fs_info;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002876 iter.task = NULL;
2877 iter.tgid = filp->f_pos - TGID_OFFSET;
2878 for (iter = next_tgid(ns, iter);
2879 iter.task;
2880 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2881 filp->f_pos = iter.tgid + TGID_OFFSET;
2882 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
2883 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002884 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002885 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002886 }
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002887 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2888out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07002889 put_task_struct(reaper);
2890out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002891 return 0;
2892}
2893
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002894/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002895 * Tasks
2896 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07002897static const struct pid_entry tid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002898 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2899 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fd_operations),
2900 REG("environ", S_IRUSR, proc_environ_operations),
2901 INF("auxv", S_IRUSR, proc_pid_auxv),
2902 ONE("status", S_IRUGO, proc_pid_status),
2903 ONE("personality", S_IRUSR, proc_pid_personality),
2904 INF("limits", S_IRUSR, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002905#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002906 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002907#endif
john stultz4614a696b2009-12-14 18:00:05 -08002908 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
Roland McGrathebcb6732008-07-25 19:46:00 -07002909#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002910 INF("syscall", S_IRUSR, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07002911#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002912 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2913 ONE("stat", S_IRUGO, proc_tid_stat),
2914 ONE("statm", S_IRUGO, proc_pid_statm),
2915 REG("maps", S_IRUGO, proc_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002916#ifdef CONFIG_NUMA
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002917 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002918#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002919 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2920 LNK("cwd", proc_cwd_link),
2921 LNK("root", proc_root_link),
2922 LNK("exe", proc_exe_link),
2923 REG("mounts", S_IRUGO, proc_mounts_operations),
2924 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08002925#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002926 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2927 REG("smaps", S_IRUGO, proc_smaps_operations),
2928 REG("pagemap", S_IRUSR, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002929#endif
2930#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002931 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002932#endif
2933#ifdef CONFIG_KALLSYMS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002934 INF("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002935#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03002936#ifdef CONFIG_STACKTRACE
2937 ONE("stack", S_IRUSR, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002938#endif
2939#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002940 INF("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002941#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01002942#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002943 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01002944#endif
Paul Menage8793d852007-10-18 23:39:39 -07002945#ifdef CONFIG_PROC_PID_CPUSET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002946 REG("cpuset", S_IRUGO, proc_cpuset_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002947#endif
Paul Menagea4243162007-10-18 23:39:35 -07002948#ifdef CONFIG_CGROUPS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002949 REG("cgroup", S_IRUGO, proc_cgroup_operations),
Paul Menagea4243162007-10-18 23:39:35 -07002950#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002951 INF("oom_score", S_IRUGO, proc_oom_score),
2952 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002953#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002954 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2955 REG("sessionid", S_IRUSR, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002956#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002957#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002958 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002959#endif
Andrea Righi297c5d92008-07-25 01:48:49 -07002960#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002961 INF("io", S_IRUGO, proc_tid_io_accounting),
Andrea Righi297c5d92008-07-25 01:48:49 -07002962#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002963};
2964
2965static int proc_tid_base_readdir(struct file * filp,
2966 void * dirent, filldir_t filldir)
2967{
2968 return proc_pident_readdir(filp,dirent,filldir,
2969 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2970}
2971
2972static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002973 return proc_pident_lookup(dir, dentry,
2974 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002975}
2976
Arjan van de Ven00977a52007-02-12 00:55:34 -08002977static const struct file_operations proc_tid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002978 .read = generic_read_dir,
2979 .readdir = proc_tid_base_readdir,
2980};
2981
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002982static const struct inode_operations proc_tid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002983 .lookup = proc_tid_base_lookup,
2984 .getattr = pid_getattr,
2985 .setattr = proc_setattr,
2986};
2987
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002988static struct dentry *proc_task_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002989 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002990{
2991 struct dentry *error = ERR_PTR(-ENOENT);
2992 struct inode *inode;
Eric W. Biederman61a28782006-10-02 02:18:49 -07002993 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002994
2995 if (!inode)
2996 goto out;
2997 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2998 inode->i_op = &proc_tid_base_inode_operations;
2999 inode->i_fop = &proc_tid_base_operations;
3000 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07003001
3002 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
3003 ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003004
3005 dentry->d_op = &pid_dentry_operations;
3006
3007 d_add(dentry, inode);
3008 /* Close the race of the process dying before we return the dentry */
3009 if (pid_revalidate(dentry, NULL))
3010 error = NULL;
3011out:
3012 return error;
3013}
3014
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003015static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3016{
3017 struct dentry *result = ERR_PTR(-ENOENT);
3018 struct task_struct *task;
3019 struct task_struct *leader = get_proc_task(dir);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003020 unsigned tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003021 struct pid_namespace *ns;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003022
3023 if (!leader)
3024 goto out_no_task;
3025
3026 tid = name_to_int(dentry);
3027 if (tid == ~0U)
3028 goto out;
3029
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003030 ns = dentry->d_sb->s_fs_info;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003031 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003032 task = find_task_by_pid_ns(tid, ns);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003033 if (task)
3034 get_task_struct(task);
3035 rcu_read_unlock();
3036 if (!task)
3037 goto out;
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07003038 if (!same_thread_group(leader, task))
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003039 goto out_drop_task;
3040
Eric W. Biederman444ceed2006-10-02 02:18:49 -07003041 result = proc_task_instantiate(dir, dentry, task, NULL);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003042out_drop_task:
3043 put_task_struct(task);
3044out:
3045 put_task_struct(leader);
3046out_no_task:
3047 return result;
3048}
3049
3050/*
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003051 * Find the first tid of a thread group to return to user space.
3052 *
3053 * Usually this is just the thread group leader, but if the users
3054 * buffer was too small or there was a seek into the middle of the
3055 * directory we have more work todo.
3056 *
3057 * In the case of a short read we start with find_task_by_pid.
3058 *
3059 * In the case of a seek we start with the leader and walk nr
3060 * threads past it.
3061 */
Eric W. Biedermancc288732006-06-26 00:26:01 -07003062static struct task_struct *first_tid(struct task_struct *leader,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003063 int tid, int nr, struct pid_namespace *ns)
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003064{
Oleg Nesterova872ff02006-06-26 00:26:01 -07003065 struct task_struct *pos;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003066
Eric W. Biedermancc288732006-06-26 00:26:01 -07003067 rcu_read_lock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003068 /* Attempt to start with the pid of a thread */
3069 if (tid && (nr > 0)) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003070 pos = find_task_by_pid_ns(tid, ns);
Oleg Nesterova872ff02006-06-26 00:26:01 -07003071 if (pos && (pos->group_leader == leader))
3072 goto found;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003073 }
3074
3075 /* If nr exceeds the number of threads there is nothing todo */
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003076 pos = NULL;
Oleg Nesterova872ff02006-06-26 00:26:01 -07003077 if (nr && nr >= get_nr_threads(leader))
3078 goto out;
3079
3080 /* If we haven't found our starting place yet start
3081 * with the leader and walk nr threads forward.
3082 */
3083 for (pos = leader; nr > 0; --nr) {
3084 pos = next_thread(pos);
3085 if (pos == leader) {
3086 pos = NULL;
3087 goto out;
3088 }
3089 }
3090found:
3091 get_task_struct(pos);
3092out:
Eric W. Biedermancc288732006-06-26 00:26:01 -07003093 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003094 return pos;
3095}
3096
3097/*
3098 * Find the next thread in the thread list.
3099 * Return NULL if there is an error or no next thread.
3100 *
3101 * The reference to the input task_struct is released.
3102 */
3103static struct task_struct *next_tid(struct task_struct *start)
3104{
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003105 struct task_struct *pos = NULL;
Eric W. Biedermancc288732006-06-26 00:26:01 -07003106 rcu_read_lock();
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003107 if (pid_alive(start)) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003108 pos = next_thread(start);
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003109 if (thread_group_leader(pos))
3110 pos = NULL;
3111 else
3112 get_task_struct(pos);
3113 }
Eric W. Biedermancc288732006-06-26 00:26:01 -07003114 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003115 put_task_struct(start);
3116 return pos;
3117}
3118
Eric W. Biederman61a28782006-10-02 02:18:49 -07003119static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3120 struct task_struct *task, int tid)
3121{
3122 char name[PROC_NUMBUF];
3123 int len = snprintf(name, sizeof(name), "%d", tid);
3124 return proc_fill_cache(filp, dirent, filldir, name, len,
3125 proc_task_instantiate, task, NULL);
3126}
3127
Linus Torvalds1da177e2005-04-16 15:20:36 -07003128/* for the /proc/TGID/task/ directories */
3129static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3130{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08003131 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003132 struct inode *inode = dentry->d_inode;
Guillaume Chazarain7d895242007-01-31 23:48:14 -08003133 struct task_struct *leader = NULL;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003134 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003135 int retval = -ENOENT;
3136 ino_t ino;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003137 int tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003138 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003139
Guillaume Chazarain7d895242007-01-31 23:48:14 -08003140 task = get_proc_task(inode);
3141 if (!task)
3142 goto out_no_task;
3143 rcu_read_lock();
3144 if (pid_alive(task)) {
3145 leader = task->group_leader;
3146 get_task_struct(leader);
3147 }
3148 rcu_read_unlock();
3149 put_task_struct(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003150 if (!leader)
3151 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003152 retval = 0;
3153
Linus Torvaldsee568b22009-03-17 10:02:35 -07003154 switch ((unsigned long)filp->f_pos) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003155 case 0:
3156 ino = inode->i_ino;
Zhang Leee6f7792009-03-16 14:44:31 +08003157 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003158 goto out;
Zhang Leee6f7792009-03-16 14:44:31 +08003159 filp->f_pos++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003160 /* fall through */
3161 case 1:
3162 ino = parent_ino(dentry);
Zhang Leee6f7792009-03-16 14:44:31 +08003163 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003164 goto out;
Zhang Leee6f7792009-03-16 14:44:31 +08003165 filp->f_pos++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003166 /* fall through */
3167 }
3168
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003169 /* f_version caches the tgid value that the last readdir call couldn't
3170 * return. lseek aka telldir automagically resets f_version to 0.
3171 */
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003172 ns = filp->f_dentry->d_sb->s_fs_info;
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07003173 tid = (int)filp->f_version;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003174 filp->f_version = 0;
Zhang Leee6f7792009-03-16 14:44:31 +08003175 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003176 task;
Zhang Leee6f7792009-03-16 14:44:31 +08003177 task = next_tid(task), filp->f_pos++) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003178 tid = task_pid_nr_ns(task, ns);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003179 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003180 /* returning this tgid failed, save it as the first
3181 * pid for the next readir call */
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07003182 filp->f_version = (u64)tid;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003183 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003184 break;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003185 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003186 }
3187out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07003188 put_task_struct(leader);
3189out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07003190 return retval;
3191}
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003192
3193static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3194{
3195 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07003196 struct task_struct *p = get_proc_task(inode);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003197 generic_fillattr(inode, stat);
3198
Eric W. Biederman99f89552006-06-26 00:25:55 -07003199 if (p) {
Eric W. Biederman99f89552006-06-26 00:25:55 -07003200 stat->nlink += get_nr_threads(p);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003201 put_task_struct(p);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003202 }
3203
3204 return 0;
3205}
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003206
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003207static const struct inode_operations proc_task_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003208 .lookup = proc_task_lookup,
3209 .getattr = proc_task_getattr,
3210 .setattr = proc_setattr,
3211};
3212
Arjan van de Ven00977a52007-02-12 00:55:34 -08003213static const struct file_operations proc_task_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003214 .read = generic_read_dir,
3215 .readdir = proc_task_readdir,
3216};