blob: 6f742f6658a9897503568c3ea7b78407829e760f [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);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 points = badness(task, 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] = {
461 [RLIMIT_CPU] = {"Max cpu time", "ms"},
462 [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;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001002 int oom_adjust;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003
Eric W. Biederman99f89552006-06-26 00:25:55 -07001004 if (!task)
1005 return -ESRCH;
KOSAKI Motohiro0753ba02009-08-18 14:11:10 -07001006 oom_adjust = task->oomkilladj;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001007 put_task_struct(task);
1008
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001009 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001010
1011 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012}
1013
1014static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1015 size_t count, loff_t *ppos)
1016{
Eric W. Biederman99f89552006-06-26 00:25:55 -07001017 struct task_struct *task;
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001018 char buffer[PROC_NUMBUF], *end;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 int oom_adjust;
1020
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001021 memset(buffer, 0, sizeof(buffer));
1022 if (count > sizeof(buffer) - 1)
1023 count = sizeof(buffer) - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 if (copy_from_user(buffer, buf, count))
1025 return -EFAULT;
1026 oom_adjust = simple_strtol(buffer, &end, 0);
Alexey Dobriyan8ac773b2006-10-19 23:28:32 -07001027 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1028 oom_adjust != OOM_DISABLE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029 return -EINVAL;
1030 if (*end == '\n')
1031 end++;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001032 task = get_proc_task(file->f_path.dentry->d_inode);
Eric W. Biederman99f89552006-06-26 00:25:55 -07001033 if (!task)
1034 return -ESRCH;
KOSAKI Motohiro0753ba02009-08-18 14:11:10 -07001035 if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) {
Guillem Jover8fb4fc62006-12-06 20:32:24 -08001036 put_task_struct(task);
1037 return -EACCES;
1038 }
KOSAKI Motohiro0753ba02009-08-18 14:11:10 -07001039 task->oomkilladj = oom_adjust;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001040 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 if (end - buffer == 0)
1042 return -EIO;
1043 return end - buffer;
1044}
1045
Arjan van de Ven00977a52007-02-12 00:55:34 -08001046static const struct file_operations proc_oom_adjust_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 .read = oom_adjust_read,
1048 .write = oom_adjust_write,
1049};
1050
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051#ifdef CONFIG_AUDITSYSCALL
1052#define TMPBUFLEN 21
1053static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1054 size_t count, loff_t *ppos)
1055{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001056 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001057 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 ssize_t length;
1059 char tmpbuf[TMPBUFLEN];
1060
Eric W. Biederman99f89552006-06-26 00:25:55 -07001061 if (!task)
1062 return -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
Al Viro0c11b942008-01-10 04:20:52 -05001064 audit_get_loginuid(task));
Eric W. Biederman99f89552006-06-26 00:25:55 -07001065 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1067}
1068
1069static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1070 size_t count, loff_t *ppos)
1071{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001072 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 char *page, *tmp;
1074 ssize_t length;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 uid_t loginuid;
1076
1077 if (!capable(CAP_AUDIT_CONTROL))
1078 return -EPERM;
1079
Eric W. Biederman13b41b02006-06-26 00:25:56 -07001080 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 return -EPERM;
1082
Al Viroe0182902006-05-18 08:28:02 -04001083 if (count >= PAGE_SIZE)
1084 count = PAGE_SIZE - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085
1086 if (*ppos != 0) {
1087 /* No partial writes. */
1088 return -EINVAL;
1089 }
Mel Gormane12ba742007-10-16 01:25:52 -07001090 page = (char*)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 if (!page)
1092 return -ENOMEM;
1093 length = -EFAULT;
1094 if (copy_from_user(page, buf, count))
1095 goto out_free_page;
1096
Al Viroe0182902006-05-18 08:28:02 -04001097 page[count] = '\0';
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 loginuid = simple_strtoul(page, &tmp, 10);
1099 if (tmp == page) {
1100 length = -EINVAL;
1101 goto out_free_page;
1102
1103 }
Eric W. Biederman99f89552006-06-26 00:25:55 -07001104 length = audit_set_loginuid(current, loginuid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 if (likely(length == 0))
1106 length = count;
1107
1108out_free_page:
1109 free_page((unsigned long) page);
1110 return length;
1111}
1112
Arjan van de Ven00977a52007-02-12 00:55:34 -08001113static const struct file_operations proc_loginuid_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 .read = proc_loginuid_read,
1115 .write = proc_loginuid_write,
1116};
Eric Paris1e0bd752008-03-13 08:15:31 -04001117
1118static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1119 size_t count, loff_t *ppos)
1120{
1121 struct inode * inode = file->f_path.dentry->d_inode;
1122 struct task_struct *task = get_proc_task(inode);
1123 ssize_t length;
1124 char tmpbuf[TMPBUFLEN];
1125
1126 if (!task)
1127 return -ESRCH;
1128 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1129 audit_get_sessionid(task));
1130 put_task_struct(task);
1131 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1132}
1133
1134static const struct file_operations proc_sessionid_operations = {
1135 .read = proc_sessionid_read,
1136};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137#endif
1138
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001139#ifdef CONFIG_FAULT_INJECTION
1140static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1141 size_t count, loff_t *ppos)
1142{
1143 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1144 char buffer[PROC_NUMBUF];
1145 size_t len;
1146 int make_it_fail;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001147
1148 if (!task)
1149 return -ESRCH;
1150 make_it_fail = task->make_it_fail;
1151 put_task_struct(task);
1152
1153 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001154
1155 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001156}
1157
1158static ssize_t proc_fault_inject_write(struct file * file,
1159 const char __user * buf, size_t count, loff_t *ppos)
1160{
1161 struct task_struct *task;
1162 char buffer[PROC_NUMBUF], *end;
1163 int make_it_fail;
1164
1165 if (!capable(CAP_SYS_RESOURCE))
1166 return -EPERM;
1167 memset(buffer, 0, sizeof(buffer));
1168 if (count > sizeof(buffer) - 1)
1169 count = sizeof(buffer) - 1;
1170 if (copy_from_user(buffer, buf, count))
1171 return -EFAULT;
1172 make_it_fail = simple_strtol(buffer, &end, 0);
1173 if (*end == '\n')
1174 end++;
1175 task = get_proc_task(file->f_dentry->d_inode);
1176 if (!task)
1177 return -ESRCH;
1178 task->make_it_fail = make_it_fail;
1179 put_task_struct(task);
1180 if (end - buffer == 0)
1181 return -EIO;
1182 return end - buffer;
1183}
1184
Arjan van de Ven00977a52007-02-12 00:55:34 -08001185static const struct file_operations proc_fault_inject_operations = {
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001186 .read = proc_fault_inject_read,
1187 .write = proc_fault_inject_write,
1188};
1189#endif
1190
Arjan van de Ven97455122008-01-25 21:08:34 +01001191
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001192#ifdef CONFIG_SCHED_DEBUG
1193/*
1194 * Print out various scheduling related per-task fields:
1195 */
1196static int sched_show(struct seq_file *m, void *v)
1197{
1198 struct inode *inode = m->private;
1199 struct task_struct *p;
1200
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001201 p = get_proc_task(inode);
1202 if (!p)
1203 return -ESRCH;
1204 proc_sched_show_task(p, m);
1205
1206 put_task_struct(p);
1207
1208 return 0;
1209}
1210
1211static ssize_t
1212sched_write(struct file *file, const char __user *buf,
1213 size_t count, loff_t *offset)
1214{
1215 struct inode *inode = file->f_path.dentry->d_inode;
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_set_task(p);
1222
1223 put_task_struct(p);
1224
1225 return count;
1226}
1227
1228static int sched_open(struct inode *inode, struct file *filp)
1229{
1230 int ret;
1231
1232 ret = single_open(filp, sched_show, NULL);
1233 if (!ret) {
1234 struct seq_file *m = filp->private_data;
1235
1236 m->private = inode;
1237 }
1238 return ret;
1239}
1240
1241static const struct file_operations proc_pid_sched_operations = {
1242 .open = sched_open,
1243 .read = seq_read,
1244 .write = sched_write,
1245 .llseek = seq_lseek,
Alexey Dobriyan5ea473a2007-07-31 00:38:50 -07001246 .release = single_release,
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001247};
1248
1249#endif
1250
Matt Helsley925d1c42008-04-29 01:01:36 -07001251/*
1252 * We added or removed a vma mapping the executable. The vmas are only mapped
1253 * during exec and are not mapped with the mmap system call.
1254 * Callers must hold down_write() on the mm's mmap_sem for these
1255 */
1256void added_exe_file_vma(struct mm_struct *mm)
1257{
1258 mm->num_exe_file_vmas++;
1259}
1260
1261void removed_exe_file_vma(struct mm_struct *mm)
1262{
1263 mm->num_exe_file_vmas--;
1264 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1265 fput(mm->exe_file);
1266 mm->exe_file = NULL;
1267 }
1268
1269}
1270
1271void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1272{
1273 if (new_exe_file)
1274 get_file(new_exe_file);
1275 if (mm->exe_file)
1276 fput(mm->exe_file);
1277 mm->exe_file = new_exe_file;
1278 mm->num_exe_file_vmas = 0;
1279}
1280
1281struct file *get_mm_exe_file(struct mm_struct *mm)
1282{
1283 struct file *exe_file;
1284
1285 /* We need mmap_sem to protect against races with removal of
1286 * VM_EXECUTABLE vmas */
1287 down_read(&mm->mmap_sem);
1288 exe_file = mm->exe_file;
1289 if (exe_file)
1290 get_file(exe_file);
1291 up_read(&mm->mmap_sem);
1292 return exe_file;
1293}
1294
1295void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1296{
1297 /* It's safe to write the exe_file pointer without exe_file_lock because
1298 * this is called during fork when the task is not yet in /proc */
1299 newmm->exe_file = get_mm_exe_file(oldmm);
1300}
1301
1302static int proc_exe_link(struct inode *inode, struct path *exe_path)
1303{
1304 struct task_struct *task;
1305 struct mm_struct *mm;
1306 struct file *exe_file;
1307
1308 task = get_proc_task(inode);
1309 if (!task)
1310 return -ENOENT;
1311 mm = get_task_mm(task);
1312 put_task_struct(task);
1313 if (!mm)
1314 return -ENOENT;
1315 exe_file = get_mm_exe_file(mm);
1316 mmput(mm);
1317 if (exe_file) {
1318 *exe_path = exe_file->f_path;
1319 path_get(&exe_file->f_path);
1320 fput(exe_file);
1321 return 0;
1322 } else
1323 return -ENOENT;
1324}
1325
Al Viro008b1502005-08-20 00:17:39 +01001326static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327{
1328 struct inode *inode = dentry->d_inode;
1329 int error = -EACCES;
1330
1331 /* We don't need a base pointer in the /proc filesystem */
Jan Blunck1d957f92008-02-14 19:34:35 -08001332 path_put(&nd->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333
Eric W. Biederman778c1142006-06-26 00:25:58 -07001334 /* Are we allowed to snoop on the tasks file descriptors? */
1335 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001338 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 nd->last_type = LAST_BIND;
1340out:
Al Viro008b1502005-08-20 00:17:39 +01001341 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342}
1343
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001344static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345{
Mel Gormane12ba742007-10-16 01:25:52 -07001346 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001347 char *pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348 int len;
1349
1350 if (!tmp)
1351 return -ENOMEM;
Akinobu Mita0c28f282007-05-08 00:31:41 -07001352
Jan Blunckcf28b482008-02-14 19:38:44 -08001353 pathname = d_path(path, tmp, PAGE_SIZE);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001354 len = PTR_ERR(pathname);
1355 if (IS_ERR(pathname))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 goto out;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001357 len = tmp + PAGE_SIZE - 1 - pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358
1359 if (len > buflen)
1360 len = buflen;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001361 if (copy_to_user(buffer, pathname, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 len = -EFAULT;
1363 out:
1364 free_page((unsigned long)tmp);
1365 return len;
1366}
1367
1368static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1369{
1370 int error = -EACCES;
1371 struct inode *inode = dentry->d_inode;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001372 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373
Eric W. Biederman778c1142006-06-26 00:25:58 -07001374 /* Are we allowed to snoop on the tasks file descriptors? */
1375 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001378 error = PROC_I(inode)->op.proc_get_link(inode, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 if (error)
1380 goto out;
1381
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001382 error = do_proc_readlink(&path, buffer, buflen);
1383 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 return error;
1386}
1387
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08001388static const struct inode_operations proc_pid_link_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389 .readlink = proc_pid_readlink,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001390 .follow_link = proc_pid_follow_link,
1391 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392};
1393
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001394
1395/* building an inode */
1396
1397static int task_dumpable(struct task_struct *task)
1398{
1399 int dumpable = 0;
1400 struct mm_struct *mm;
1401
1402 task_lock(task);
1403 mm = task->mm;
1404 if (mm)
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001405 dumpable = get_dumpable(mm);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001406 task_unlock(task);
1407 if(dumpable == 1)
1408 return 1;
1409 return 0;
1410}
1411
1412
Eric W. Biederman61a28782006-10-02 02:18:49 -07001413static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001414{
1415 struct inode * inode;
1416 struct proc_inode *ei;
David Howellsc69e8d92008-11-14 10:39:19 +11001417 const struct cred *cred;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001418
1419 /* We need a new inode */
1420
1421 inode = new_inode(sb);
1422 if (!inode)
1423 goto out;
1424
1425 /* Common stuff */
1426 ei = PROC_I(inode);
1427 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001428 inode->i_op = &proc_def_inode_operations;
1429
1430 /*
1431 * grab the reference to task.
1432 */
Oleg Nesterov1a657f72006-10-02 02:18:59 -07001433 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001434 if (!ei->pid)
1435 goto out_unlock;
1436
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001437 if (task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001438 rcu_read_lock();
1439 cred = __task_cred(task);
1440 inode->i_uid = cred->euid;
1441 inode->i_gid = cred->egid;
1442 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001443 }
1444 security_task_to_inode(task, inode);
1445
1446out:
1447 return inode;
1448
1449out_unlock:
1450 iput(inode);
1451 return NULL;
1452}
1453
1454static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1455{
1456 struct inode *inode = dentry->d_inode;
1457 struct task_struct *task;
David Howellsc69e8d92008-11-14 10:39:19 +11001458 const struct cred *cred;
1459
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001460 generic_fillattr(inode, stat);
1461
1462 rcu_read_lock();
1463 stat->uid = 0;
1464 stat->gid = 0;
1465 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1466 if (task) {
1467 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1468 task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001469 cred = __task_cred(task);
1470 stat->uid = cred->euid;
1471 stat->gid = cred->egid;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001472 }
1473 }
1474 rcu_read_unlock();
1475 return 0;
1476}
1477
1478/* dentry stuff */
1479
1480/*
1481 * Exceptional case: normally we are not allowed to unhash a busy
1482 * directory. In this case, however, we can do it - no aliasing problems
1483 * due to the way we treat inodes.
1484 *
1485 * Rewrite the inode's ownerships here because the owning task may have
1486 * performed a setuid(), etc.
1487 *
1488 * Before the /proc/pid/status file was created the only way to read
1489 * the effective uid of a /process was to stat /proc/pid. Reading
1490 * /proc/pid/status is slow enough that procps and other packages
1491 * kept stating /proc/pid. To keep the rules in /proc simple I have
1492 * made this apply to all per process world readable and executable
1493 * directories.
1494 */
1495static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1496{
1497 struct inode *inode = dentry->d_inode;
1498 struct task_struct *task = get_proc_task(inode);
David Howellsc69e8d92008-11-14 10:39:19 +11001499 const struct cred *cred;
1500
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001501 if (task) {
1502 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1503 task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001504 rcu_read_lock();
1505 cred = __task_cred(task);
1506 inode->i_uid = cred->euid;
1507 inode->i_gid = cred->egid;
1508 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001509 } else {
1510 inode->i_uid = 0;
1511 inode->i_gid = 0;
1512 }
1513 inode->i_mode &= ~(S_ISUID | S_ISGID);
1514 security_task_to_inode(task, inode);
1515 put_task_struct(task);
1516 return 1;
1517 }
1518 d_drop(dentry);
1519 return 0;
1520}
1521
1522static int pid_delete_dentry(struct dentry * dentry)
1523{
1524 /* Is the task we represent dead?
1525 * If so, then don't put the dentry on the lru list,
1526 * kill it immediately.
1527 */
1528 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1529}
1530
Al Virod72f71e2009-02-20 05:58:47 +00001531static const struct dentry_operations pid_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001532{
1533 .d_revalidate = pid_revalidate,
1534 .d_delete = pid_delete_dentry,
1535};
1536
1537/* Lookups */
1538
Eric Dumazetc5141e62007-05-08 00:26:15 -07001539typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1540 struct task_struct *, const void *);
Eric W. Biederman61a28782006-10-02 02:18:49 -07001541
Eric W. Biederman1c0d04c2006-10-02 02:18:57 -07001542/*
1543 * Fill a directory entry.
1544 *
1545 * If possible create the dcache entry and derive our inode number and
1546 * file type from dcache entry.
1547 *
1548 * Since all of the proc inode numbers are dynamically generated, the inode
1549 * numbers do not exist until the inode is cache. This means creating the
1550 * the dcache entry in readdir is necessary to keep the inode numbers
1551 * reported by readdir in sync with the inode numbers reported
1552 * by stat.
1553 */
Eric W. Biederman61a28782006-10-02 02:18:49 -07001554static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1555 char *name, int len,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001556 instantiate_t instantiate, struct task_struct *task, const void *ptr)
Eric W. Biederman61a28782006-10-02 02:18:49 -07001557{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001558 struct dentry *child, *dir = filp->f_path.dentry;
Eric W. Biederman61a28782006-10-02 02:18:49 -07001559 struct inode *inode;
1560 struct qstr qname;
1561 ino_t ino = 0;
1562 unsigned type = DT_UNKNOWN;
1563
1564 qname.name = name;
1565 qname.len = len;
1566 qname.hash = full_name_hash(name, len);
1567
1568 child = d_lookup(dir, &qname);
1569 if (!child) {
1570 struct dentry *new;
1571 new = d_alloc(dir, &qname);
1572 if (new) {
1573 child = instantiate(dir->d_inode, new, task, ptr);
1574 if (child)
1575 dput(new);
1576 else
1577 child = new;
1578 }
1579 }
1580 if (!child || IS_ERR(child) || !child->d_inode)
1581 goto end_instantiate;
1582 inode = child->d_inode;
1583 if (inode) {
1584 ino = inode->i_ino;
1585 type = inode->i_mode >> 12;
1586 }
1587 dput(child);
1588end_instantiate:
1589 if (!ino)
1590 ino = find_inode_number(dir, &qname);
1591 if (!ino)
1592 ino = 1;
1593 return filldir(dirent, name, len, filp->f_pos, ino, type);
1594}
1595
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001596static unsigned name_to_int(struct dentry *dentry)
1597{
1598 const char *name = dentry->d_name.name;
1599 int len = dentry->d_name.len;
1600 unsigned n = 0;
1601
1602 if (len > 1 && *name == '0')
1603 goto out;
1604 while (len-- > 0) {
1605 unsigned c = *name++ - '0';
1606 if (c > 9)
1607 goto out;
1608 if (n >= (~0U-9)/10)
1609 goto out;
1610 n *= 10;
1611 n += c;
1612 }
1613 return n;
1614out:
1615 return ~0U;
1616}
1617
Miklos Szeredi27932742007-05-08 00:26:17 -07001618#define PROC_FDINFO_MAX 64
1619
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001620static int proc_fd_info(struct inode *inode, struct path *path, char *info)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001621{
1622 struct task_struct *task = get_proc_task(inode);
1623 struct files_struct *files = NULL;
1624 struct file *file;
1625 int fd = proc_fd(inode);
1626
1627 if (task) {
1628 files = get_files_struct(task);
1629 put_task_struct(task);
1630 }
1631 if (files) {
1632 /*
1633 * We are not taking a ref to the file structure, so we must
1634 * hold ->file_lock.
1635 */
1636 spin_lock(&files->file_lock);
1637 file = fcheck_files(files, fd);
1638 if (file) {
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001639 if (path) {
1640 *path = file->f_path;
1641 path_get(&file->f_path);
1642 }
Miklos Szeredi27932742007-05-08 00:26:17 -07001643 if (info)
1644 snprintf(info, PROC_FDINFO_MAX,
1645 "pos:\t%lli\n"
1646 "flags:\t0%o\n",
1647 (long long) file->f_pos,
1648 file->f_flags);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001649 spin_unlock(&files->file_lock);
1650 put_files_struct(files);
1651 return 0;
1652 }
1653 spin_unlock(&files->file_lock);
1654 put_files_struct(files);
1655 }
1656 return -ENOENT;
1657}
1658
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001659static int proc_fd_link(struct inode *inode, struct path *path)
Miklos Szeredi27932742007-05-08 00:26:17 -07001660{
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001661 return proc_fd_info(inode, path, NULL);
Miklos Szeredi27932742007-05-08 00:26:17 -07001662}
1663
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001664static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1665{
1666 struct inode *inode = dentry->d_inode;
1667 struct task_struct *task = get_proc_task(inode);
1668 int fd = proc_fd(inode);
1669 struct files_struct *files;
David Howellsc69e8d92008-11-14 10:39:19 +11001670 const struct cred *cred;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001671
1672 if (task) {
1673 files = get_files_struct(task);
1674 if (files) {
1675 rcu_read_lock();
1676 if (fcheck_files(files, fd)) {
1677 rcu_read_unlock();
1678 put_files_struct(files);
1679 if (task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001680 rcu_read_lock();
1681 cred = __task_cred(task);
1682 inode->i_uid = cred->euid;
1683 inode->i_gid = cred->egid;
1684 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001685 } else {
1686 inode->i_uid = 0;
1687 inode->i_gid = 0;
1688 }
1689 inode->i_mode &= ~(S_ISUID | S_ISGID);
1690 security_task_to_inode(task, inode);
1691 put_task_struct(task);
1692 return 1;
1693 }
1694 rcu_read_unlock();
1695 put_files_struct(files);
1696 }
1697 put_task_struct(task);
1698 }
1699 d_drop(dentry);
1700 return 0;
1701}
1702
Al Virod72f71e2009-02-20 05:58:47 +00001703static const struct dentry_operations tid_fd_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001704{
1705 .d_revalidate = tid_fd_revalidate,
1706 .d_delete = pid_delete_dentry,
1707};
1708
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001709static struct dentry *proc_fd_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001710 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001711{
Eric Dumazetc5141e62007-05-08 00:26:15 -07001712 unsigned fd = *(const unsigned *)ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001713 struct file *file;
1714 struct files_struct *files;
1715 struct inode *inode;
1716 struct proc_inode *ei;
1717 struct dentry *error = ERR_PTR(-ENOENT);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001718
Eric W. Biederman61a28782006-10-02 02:18:49 -07001719 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001720 if (!inode)
1721 goto out;
1722 ei = PROC_I(inode);
1723 ei->fd = fd;
1724 files = get_files_struct(task);
1725 if (!files)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001726 goto out_iput;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001727 inode->i_mode = S_IFLNK;
1728
1729 /*
1730 * We are not taking a ref to the file structure, so we must
1731 * hold ->file_lock.
1732 */
1733 spin_lock(&files->file_lock);
1734 file = fcheck_files(files, fd);
1735 if (!file)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001736 goto out_unlock;
Al Viroaeb5d722008-09-02 15:28:45 -04001737 if (file->f_mode & FMODE_READ)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001738 inode->i_mode |= S_IRUSR | S_IXUSR;
Al Viroaeb5d722008-09-02 15:28:45 -04001739 if (file->f_mode & FMODE_WRITE)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001740 inode->i_mode |= S_IWUSR | S_IXUSR;
1741 spin_unlock(&files->file_lock);
1742 put_files_struct(files);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001743
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001744 inode->i_op = &proc_pid_link_inode_operations;
1745 inode->i_size = 64;
1746 ei->op.proc_get_link = proc_fd_link;
1747 dentry->d_op = &tid_fd_dentry_operations;
1748 d_add(dentry, inode);
1749 /* Close the race of the process dying before we return the dentry */
1750 if (tid_fd_revalidate(dentry, NULL))
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001751 error = NULL;
1752
1753 out:
1754 return error;
1755out_unlock:
1756 spin_unlock(&files->file_lock);
1757 put_files_struct(files);
1758out_iput:
1759 iput(inode);
1760 goto out;
1761}
1762
Miklos Szeredi27932742007-05-08 00:26:17 -07001763static struct dentry *proc_lookupfd_common(struct inode *dir,
1764 struct dentry *dentry,
1765 instantiate_t instantiate)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001766{
1767 struct task_struct *task = get_proc_task(dir);
1768 unsigned fd = name_to_int(dentry);
1769 struct dentry *result = ERR_PTR(-ENOENT);
1770
1771 if (!task)
1772 goto out_no_task;
1773 if (fd == ~0U)
1774 goto out;
1775
Miklos Szeredi27932742007-05-08 00:26:17 -07001776 result = instantiate(dir, dentry, task, &fd);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001777out:
1778 put_task_struct(task);
1779out_no_task:
1780 return result;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001781}
1782
Miklos Szeredi27932742007-05-08 00:26:17 -07001783static int proc_readfd_common(struct file * filp, void * dirent,
1784 filldir_t filldir, instantiate_t instantiate)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001786 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman56347082006-06-26 00:25:40 -07001787 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001788 struct task_struct *p = get_proc_task(inode);
Pavel Emelyanov457c2512007-10-18 23:40:43 -07001789 unsigned int fd, ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791 struct files_struct * files;
1792
1793 retval = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001794 if (!p)
1795 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797
1798 fd = filp->f_pos;
1799 switch (fd) {
1800 case 0:
1801 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1802 goto out;
1803 filp->f_pos++;
1804 case 1:
Eric W. Biederman56347082006-06-26 00:25:40 -07001805 ino = parent_ino(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1807 goto out;
1808 filp->f_pos++;
1809 default:
1810 files = get_files_struct(p);
1811 if (!files)
1812 goto out;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001813 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814 for (fd = filp->f_pos-2;
Al Viro9b4f5262008-04-22 01:32:44 -04001815 fd < files_fdtable(files)->max_fds;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816 fd++, filp->f_pos++) {
Miklos Szeredi27932742007-05-08 00:26:17 -07001817 char name[PROC_NUMBUF];
1818 int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819
1820 if (!fcheck_files(files, fd))
1821 continue;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001822 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001823
Miklos Szeredi27932742007-05-08 00:26:17 -07001824 len = snprintf(name, sizeof(name), "%d", fd);
1825 if (proc_fill_cache(filp, dirent, filldir,
1826 name, len, instantiate,
1827 p, &fd) < 0) {
Dipankar Sarmab8359962005-09-09 13:04:14 -07001828 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829 break;
1830 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07001831 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001832 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07001833 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834 put_files_struct(files);
1835 }
1836out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07001837 put_task_struct(p);
1838out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839 return retval;
1840}
1841
Miklos Szeredi27932742007-05-08 00:26:17 -07001842static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1843 struct nameidata *nd)
1844{
1845 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1846}
1847
1848static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1849{
1850 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1851}
1852
1853static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1854 size_t len, loff_t *ppos)
1855{
1856 char tmp[PROC_FDINFO_MAX];
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001857 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
Miklos Szeredi27932742007-05-08 00:26:17 -07001858 if (!err)
1859 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1860 return err;
1861}
1862
1863static const struct file_operations proc_fdinfo_file_operations = {
1864 .open = nonseekable_open,
1865 .read = proc_fdinfo_read,
1866};
1867
Arjan van de Ven00977a52007-02-12 00:55:34 -08001868static const struct file_operations proc_fd_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869 .read = generic_read_dir,
1870 .readdir = proc_readfd,
1871};
1872
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873/*
Alexey Dobriyan8948e112007-05-08 00:23:35 -07001874 * /proc/pid/fd needs a special permission handler so that a process can still
1875 * access /proc/self/fd after it has executed a setuid().
1876 */
Al Viroe6305c42008-07-15 21:03:57 -04001877static int proc_fd_permission(struct inode *inode, int mask)
Alexey Dobriyan8948e112007-05-08 00:23:35 -07001878{
1879 int rv;
1880
1881 rv = generic_permission(inode, mask, NULL);
1882 if (rv == 0)
1883 return 0;
1884 if (task_pid(current) == proc_pid(inode))
1885 rv = 0;
1886 return rv;
1887}
1888
1889/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001890 * proc directories can do almost nothing..
1891 */
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08001892static const struct inode_operations proc_fd_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893 .lookup = proc_lookupfd,
Alexey Dobriyan8948e112007-05-08 00:23:35 -07001894 .permission = proc_fd_permission,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001895 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896};
1897
Miklos Szeredi27932742007-05-08 00:26:17 -07001898static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1899 struct dentry *dentry, struct task_struct *task, const void *ptr)
1900{
1901 unsigned fd = *(unsigned *)ptr;
1902 struct inode *inode;
1903 struct proc_inode *ei;
1904 struct dentry *error = ERR_PTR(-ENOENT);
1905
1906 inode = proc_pid_make_inode(dir->i_sb, task);
1907 if (!inode)
1908 goto out;
1909 ei = PROC_I(inode);
1910 ei->fd = fd;
1911 inode->i_mode = S_IFREG | S_IRUSR;
1912 inode->i_fop = &proc_fdinfo_file_operations;
1913 dentry->d_op = &tid_fd_dentry_operations;
1914 d_add(dentry, inode);
1915 /* Close the race of the process dying before we return the dentry */
1916 if (tid_fd_revalidate(dentry, NULL))
1917 error = NULL;
1918
1919 out:
1920 return error;
1921}
1922
1923static struct dentry *proc_lookupfdinfo(struct inode *dir,
1924 struct dentry *dentry,
1925 struct nameidata *nd)
1926{
1927 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
1928}
1929
1930static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
1931{
1932 return proc_readfd_common(filp, dirent, filldir,
1933 proc_fdinfo_instantiate);
1934}
1935
1936static const struct file_operations proc_fdinfo_operations = {
1937 .read = generic_read_dir,
1938 .readdir = proc_readfdinfo,
1939};
1940
1941/*
1942 * proc directories can do almost nothing..
1943 */
1944static const struct inode_operations proc_fdinfo_inode_operations = {
1945 .lookup = proc_lookupfdinfo,
1946 .setattr = proc_setattr,
1947};
1948
1949
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001950static struct dentry *proc_pident_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001951 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001952{
Eric Dumazetc5141e62007-05-08 00:26:15 -07001953 const struct pid_entry *p = ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001954 struct inode *inode;
1955 struct proc_inode *ei;
KOSAKI Motohirobd6daba2009-05-28 14:34:21 -07001956 struct dentry *error = ERR_PTR(-ENOENT);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001957
Eric W. Biederman61a28782006-10-02 02:18:49 -07001958 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001959 if (!inode)
1960 goto out;
1961
1962 ei = PROC_I(inode);
1963 inode->i_mode = p->mode;
1964 if (S_ISDIR(inode->i_mode))
1965 inode->i_nlink = 2; /* Use getattr to fix if necessary */
1966 if (p->iop)
1967 inode->i_op = p->iop;
1968 if (p->fop)
1969 inode->i_fop = p->fop;
1970 ei->op = p->op;
1971 dentry->d_op = &pid_dentry_operations;
1972 d_add(dentry, inode);
1973 /* Close the race of the process dying before we return the dentry */
1974 if (pid_revalidate(dentry, NULL))
1975 error = NULL;
1976out:
1977 return error;
1978}
1979
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980static struct dentry *proc_pident_lookup(struct inode *dir,
1981 struct dentry *dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001982 const struct pid_entry *ents,
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07001983 unsigned int nents)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984{
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001985 struct dentry *error;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001986 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07001987 const struct pid_entry *p, *last;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001989 error = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990
Eric W. Biederman99f89552006-06-26 00:25:55 -07001991 if (!task)
1992 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993
Eric W. Biederman20cdc892006-10-02 02:17:07 -07001994 /*
1995 * Yes, it does not scale. And it should not. Don't add
1996 * new entries into /proc/<tgid>/ without very good reasons.
1997 */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07001998 last = &ents[nents - 1];
1999 for (p = ents; p <= last; p++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000 if (p->len != dentry->d_name.len)
2001 continue;
2002 if (!memcmp(dentry->d_name.name, p->name, p->len))
2003 break;
2004 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002005 if (p > last)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006 goto out;
2007
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002008 error = proc_pident_instantiate(dir, dentry, task, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07002010 put_task_struct(task);
2011out_no_task:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002012 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013}
2014
Eric Dumazetc5141e62007-05-08 00:26:15 -07002015static int proc_pident_fill_cache(struct file *filp, void *dirent,
2016 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002017{
2018 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2019 proc_pident_instantiate, task, p);
2020}
2021
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002022static int proc_pident_readdir(struct file *filp,
2023 void *dirent, filldir_t filldir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002024 const struct pid_entry *ents, unsigned int nents)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002025{
2026 int i;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002027 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002028 struct inode *inode = dentry->d_inode;
2029 struct task_struct *task = get_proc_task(inode);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002030 const struct pid_entry *p, *last;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002031 ino_t ino;
2032 int ret;
2033
2034 ret = -ENOENT;
2035 if (!task)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002036 goto out_no_task;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002037
2038 ret = 0;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002039 i = filp->f_pos;
2040 switch (i) {
2041 case 0:
2042 ino = inode->i_ino;
2043 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2044 goto out;
2045 i++;
2046 filp->f_pos++;
2047 /* fall through */
2048 case 1:
2049 ino = parent_ino(dentry);
2050 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2051 goto out;
2052 i++;
2053 filp->f_pos++;
2054 /* fall through */
2055 default:
2056 i -= 2;
2057 if (i >= nents) {
2058 ret = 1;
2059 goto out;
2060 }
2061 p = ents + i;
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002062 last = &ents[nents - 1];
2063 while (p <= last) {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002064 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002065 goto out;
2066 filp->f_pos++;
2067 p++;
2068 }
2069 }
2070
2071 ret = 1;
2072out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07002073 put_task_struct(task);
2074out_no_task:
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002075 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076}
2077
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078#ifdef CONFIG_SECURITY
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002079static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2080 size_t count, loff_t *ppos)
2081{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002082 struct inode * inode = file->f_path.dentry->d_inode;
Al Viro04ff9702007-03-12 16:17:58 +00002083 char *p = NULL;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002084 ssize_t length;
2085 struct task_struct *task = get_proc_task(inode);
2086
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002087 if (!task)
Al Viro04ff9702007-03-12 16:17:58 +00002088 return -ESRCH;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002089
2090 length = security_getprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002091 (char*)file->f_path.dentry->d_name.name,
Al Viro04ff9702007-03-12 16:17:58 +00002092 &p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002093 put_task_struct(task);
Al Viro04ff9702007-03-12 16:17:58 +00002094 if (length > 0)
2095 length = simple_read_from_buffer(buf, count, ppos, p, length);
2096 kfree(p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002097 return length;
2098}
2099
2100static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2101 size_t count, loff_t *ppos)
2102{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002103 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002104 char *page;
2105 ssize_t length;
2106 struct task_struct *task = get_proc_task(inode);
2107
2108 length = -ESRCH;
2109 if (!task)
2110 goto out_no_task;
2111 if (count > PAGE_SIZE)
2112 count = PAGE_SIZE;
2113
2114 /* No partial writes. */
2115 length = -EINVAL;
2116 if (*ppos != 0)
2117 goto out;
2118
2119 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -07002120 page = (char*)__get_free_page(GFP_TEMPORARY);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002121 if (!page)
2122 goto out;
2123
2124 length = -EFAULT;
2125 if (copy_from_user(page, buf, count))
2126 goto out_free;
2127
David Howells107db7c2009-05-08 13:55:27 +01002128 /* Guard against adverse ptrace interaction */
2129 length = mutex_lock_interruptible(&task->cred_guard_mutex);
2130 if (length < 0)
2131 goto out_free;
2132
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002133 length = security_setprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002134 (char*)file->f_path.dentry->d_name.name,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002135 (void*)page, count);
David Howells107db7c2009-05-08 13:55:27 +01002136 mutex_unlock(&task->cred_guard_mutex);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002137out_free:
2138 free_page((unsigned long) page);
2139out:
2140 put_task_struct(task);
2141out_no_task:
2142 return length;
2143}
2144
Arjan van de Ven00977a52007-02-12 00:55:34 -08002145static const struct file_operations proc_pid_attr_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002146 .read = proc_pid_attr_read,
2147 .write = proc_pid_attr_write,
2148};
2149
Eric Dumazetc5141e62007-05-08 00:26:15 -07002150static const struct pid_entry attr_dir_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002151 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2152 REG("prev", S_IRUGO, proc_pid_attr_operations),
2153 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2154 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2155 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2156 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002157};
2158
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002159static int proc_attr_dir_readdir(struct file * filp,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002160 void * dirent, filldir_t filldir)
2161{
2162 return proc_pident_readdir(filp,dirent,filldir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002163 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002164}
2165
Arjan van de Ven00977a52007-02-12 00:55:34 -08002166static const struct file_operations proc_attr_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002167 .read = generic_read_dir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002168 .readdir = proc_attr_dir_readdir,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002169};
2170
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002171static struct dentry *proc_attr_dir_lookup(struct inode *dir,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002172 struct dentry *dentry, struct nameidata *nd)
2173{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002174 return proc_pident_lookup(dir, dentry,
2175 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002176}
2177
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002178static const struct inode_operations proc_attr_dir_inode_operations = {
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002179 .lookup = proc_attr_dir_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07002180 .getattr = pid_getattr,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07002181 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002182};
2183
Linus Torvalds1da177e2005-04-16 15:20:36 -07002184#endif
2185
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002186#if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2187static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2188 size_t count, loff_t *ppos)
2189{
2190 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2191 struct mm_struct *mm;
2192 char buffer[PROC_NUMBUF];
2193 size_t len;
2194 int ret;
2195
2196 if (!task)
2197 return -ESRCH;
2198
2199 ret = 0;
2200 mm = get_task_mm(task);
2201 if (mm) {
2202 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2203 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2204 MMF_DUMP_FILTER_SHIFT));
2205 mmput(mm);
2206 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2207 }
2208
2209 put_task_struct(task);
2210
2211 return ret;
2212}
2213
2214static ssize_t proc_coredump_filter_write(struct file *file,
2215 const char __user *buf,
2216 size_t count,
2217 loff_t *ppos)
2218{
2219 struct task_struct *task;
2220 struct mm_struct *mm;
2221 char buffer[PROC_NUMBUF], *end;
2222 unsigned int val;
2223 int ret;
2224 int i;
2225 unsigned long mask;
2226
2227 ret = -EFAULT;
2228 memset(buffer, 0, sizeof(buffer));
2229 if (count > sizeof(buffer) - 1)
2230 count = sizeof(buffer) - 1;
2231 if (copy_from_user(buffer, buf, count))
2232 goto out_no_task;
2233
2234 ret = -EINVAL;
2235 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2236 if (*end == '\n')
2237 end++;
2238 if (end - buffer == 0)
2239 goto out_no_task;
2240
2241 ret = -ESRCH;
2242 task = get_proc_task(file->f_dentry->d_inode);
2243 if (!task)
2244 goto out_no_task;
2245
2246 ret = end - buffer;
2247 mm = get_task_mm(task);
2248 if (!mm)
2249 goto out_no_mm;
2250
2251 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2252 if (val & mask)
2253 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2254 else
2255 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2256 }
2257
2258 mmput(mm);
2259 out_no_mm:
2260 put_task_struct(task);
2261 out_no_task:
2262 return ret;
2263}
2264
2265static const struct file_operations proc_coredump_filter_operations = {
2266 .read = proc_coredump_filter_read,
2267 .write = proc_coredump_filter_write,
2268};
2269#endif
2270
Linus Torvalds1da177e2005-04-16 15:20:36 -07002271/*
2272 * /proc/self:
2273 */
2274static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2275 int buflen)
2276{
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002277 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002278 pid_t tgid = task_tgid_nr_ns(current, ns);
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002279 char tmp[PROC_NUMBUF];
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002280 if (!tgid)
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002281 return -ENOENT;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002282 sprintf(tmp, "%d", tgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002283 return vfs_readlink(dentry,buffer,buflen,tmp);
2284}
2285
Al Viro008b1502005-08-20 00:17:39 +01002286static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002287{
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002288 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002289 pid_t tgid = task_tgid_nr_ns(current, ns);
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002290 char tmp[PROC_NUMBUF];
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002291 if (!tgid)
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002292 return ERR_PTR(-ENOENT);
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002293 sprintf(tmp, "%d", task_tgid_nr_ns(current, ns));
Al Viro008b1502005-08-20 00:17:39 +01002294 return ERR_PTR(vfs_follow_link(nd,tmp));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002295}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002297static const struct inode_operations proc_self_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002298 .readlink = proc_self_readlink,
2299 .follow_link = proc_self_follow_link,
2300};
2301
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002302/*
Eric W. Biederman801199c2006-10-02 02:18:48 -07002303 * proc base
2304 *
2305 * These are the directory entries in the root directory of /proc
2306 * that properly belong to the /proc filesystem, as they describe
2307 * describe something that is process related.
2308 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07002309static const struct pid_entry proc_base_stuff[] = {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002310 NOD("self", S_IFLNK|S_IRWXUGO,
Eric W. Biederman801199c2006-10-02 02:18:48 -07002311 &proc_self_inode_operations, NULL, {}),
Eric W. Biederman801199c2006-10-02 02:18:48 -07002312};
2313
2314/*
2315 * Exceptional case: normally we are not allowed to unhash a busy
2316 * directory. In this case, however, we can do it - no aliasing problems
2317 * due to the way we treat inodes.
2318 */
2319static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2320{
2321 struct inode *inode = dentry->d_inode;
2322 struct task_struct *task = get_proc_task(inode);
2323 if (task) {
2324 put_task_struct(task);
2325 return 1;
2326 }
2327 d_drop(dentry);
2328 return 0;
2329}
2330
Al Virod72f71e2009-02-20 05:58:47 +00002331static const struct dentry_operations proc_base_dentry_operations =
Eric W. Biederman801199c2006-10-02 02:18:48 -07002332{
2333 .d_revalidate = proc_base_revalidate,
2334 .d_delete = pid_delete_dentry,
2335};
2336
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002337static struct dentry *proc_base_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002338 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman801199c2006-10-02 02:18:48 -07002339{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002340 const struct pid_entry *p = ptr;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002341 struct inode *inode;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002342 struct proc_inode *ei;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002343 struct dentry *error = ERR_PTR(-EINVAL);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002344
2345 /* Allocate the inode */
2346 error = ERR_PTR(-ENOMEM);
2347 inode = new_inode(dir->i_sb);
2348 if (!inode)
2349 goto out;
2350
2351 /* Initialize the inode */
2352 ei = PROC_I(inode);
2353 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002354
2355 /*
2356 * grab the reference to the task.
2357 */
Oleg Nesterov1a657f72006-10-02 02:18:59 -07002358 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002359 if (!ei->pid)
2360 goto out_iput;
2361
Eric W. Biederman801199c2006-10-02 02:18:48 -07002362 inode->i_mode = p->mode;
2363 if (S_ISDIR(inode->i_mode))
2364 inode->i_nlink = 2;
2365 if (S_ISLNK(inode->i_mode))
2366 inode->i_size = 64;
2367 if (p->iop)
2368 inode->i_op = p->iop;
2369 if (p->fop)
2370 inode->i_fop = p->fop;
2371 ei->op = p->op;
2372 dentry->d_op = &proc_base_dentry_operations;
2373 d_add(dentry, inode);
2374 error = NULL;
2375out:
Eric W. Biederman801199c2006-10-02 02:18:48 -07002376 return error;
2377out_iput:
2378 iput(inode);
2379 goto out;
2380}
2381
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002382static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2383{
2384 struct dentry *error;
2385 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002386 const struct pid_entry *p, *last;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002387
2388 error = ERR_PTR(-ENOENT);
2389
2390 if (!task)
2391 goto out_no_task;
2392
2393 /* Lookup the directory entry */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002394 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2395 for (p = proc_base_stuff; p <= last; p++) {
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002396 if (p->len != dentry->d_name.len)
2397 continue;
2398 if (!memcmp(dentry->d_name.name, p->name, p->len))
2399 break;
2400 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002401 if (p > last)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002402 goto out;
2403
2404 error = proc_base_instantiate(dir, dentry, task, p);
2405
2406out:
2407 put_task_struct(task);
2408out_no_task:
2409 return error;
2410}
2411
Eric Dumazetc5141e62007-05-08 00:26:15 -07002412static int proc_base_fill_cache(struct file *filp, void *dirent,
2413 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002414{
2415 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2416 proc_base_instantiate, task, p);
2417}
2418
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002419#ifdef CONFIG_TASK_IO_ACCOUNTING
Andrea Righi297c5d92008-07-25 01:48:49 -07002420static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002421{
Andrea Righi940389b2008-07-28 00:48:12 +02002422 struct task_io_accounting acct = task->ioac;
Andrea Righi59954772008-07-27 17:29:15 +02002423 unsigned long flags;
Andrea Righi297c5d92008-07-25 01:48:49 -07002424
Andrea Righi59954772008-07-27 17:29:15 +02002425 if (whole && lock_task_sighand(task, &flags)) {
2426 struct task_struct *t = task;
Andrea Righib2d002d2008-07-26 15:22:27 -07002427
Andrea Righi59954772008-07-27 17:29:15 +02002428 task_io_accounting_add(&acct, &task->signal->ioac);
2429 while_each_thread(task, t)
2430 task_io_accounting_add(&acct, &t->ioac);
Andrea Righi297c5d92008-07-25 01:48:49 -07002431
Andrea Righi59954772008-07-27 17:29:15 +02002432 unlock_task_sighand(task, &flags);
Andrea Righi297c5d92008-07-25 01:48:49 -07002433 }
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002434 return sprintf(buffer,
2435 "rchar: %llu\n"
2436 "wchar: %llu\n"
2437 "syscr: %llu\n"
2438 "syscw: %llu\n"
2439 "read_bytes: %llu\n"
2440 "write_bytes: %llu\n"
2441 "cancelled_write_bytes: %llu\n",
Alexander Beregalov7c443192008-08-05 13:01:34 -07002442 (unsigned long long)acct.rchar,
2443 (unsigned long long)acct.wchar,
2444 (unsigned long long)acct.syscr,
2445 (unsigned long long)acct.syscw,
2446 (unsigned long long)acct.read_bytes,
2447 (unsigned long long)acct.write_bytes,
2448 (unsigned long long)acct.cancelled_write_bytes);
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002449}
Andrea Righi297c5d92008-07-25 01:48:49 -07002450
2451static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2452{
2453 return do_io_accounting(task, buffer, 0);
2454}
2455
2456static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2457{
2458 return do_io_accounting(task, buffer, 1);
2459}
2460#endif /* CONFIG_TASK_IO_ACCOUNTING */
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002461
Kees Cook47830722008-10-06 03:11:58 +04002462static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2463 struct pid *pid, struct task_struct *task)
2464{
2465 seq_printf(m, "%08x\n", task->personality);
2466 return 0;
2467}
2468
Eric W. Biederman801199c2006-10-02 02:18:48 -07002469/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002470 * Thread groups
2471 */
Arjan van de Ven00977a52007-02-12 00:55:34 -08002472static const struct file_operations proc_task_operations;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002473static const struct inode_operations proc_task_inode_operations;
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002474
Eric Dumazetc5141e62007-05-08 00:26:15 -07002475static const struct pid_entry tgid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002476 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2477 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2478 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002479#ifdef CONFIG_NET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002480 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002481#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002482 REG("environ", S_IRUSR, proc_environ_operations),
2483 INF("auxv", S_IRUSR, proc_pid_auxv),
2484 ONE("status", S_IRUGO, proc_pid_status),
2485 ONE("personality", S_IRUSR, proc_pid_personality),
2486 INF("limits", S_IRUSR, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002487#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002488 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002489#endif
Roland McGrathebcb6732008-07-25 19:46:00 -07002490#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002491 INF("syscall", S_IRUSR, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07002492#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002493 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2494 ONE("stat", S_IRUGO, proc_tgid_stat),
2495 ONE("statm", S_IRUGO, proc_pid_statm),
2496 REG("maps", S_IRUGO, proc_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002497#ifdef CONFIG_NUMA
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002498 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002499#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002500 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2501 LNK("cwd", proc_cwd_link),
2502 LNK("root", proc_root_link),
2503 LNK("exe", proc_exe_link),
2504 REG("mounts", S_IRUGO, proc_mounts_operations),
2505 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2506 REG("mountstats", S_IRUSR, proc_mountstats_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08002507#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002508 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2509 REG("smaps", S_IRUGO, proc_smaps_operations),
2510 REG("pagemap", S_IRUSR, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002511#endif
2512#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002513 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002514#endif
2515#ifdef CONFIG_KALLSYMS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002516 INF("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002517#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03002518#ifdef CONFIG_STACKTRACE
2519 ONE("stack", S_IRUSR, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002520#endif
2521#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002522 INF("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002523#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01002524#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002525 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01002526#endif
Paul Menage8793d852007-10-18 23:39:39 -07002527#ifdef CONFIG_PROC_PID_CPUSET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002528 REG("cpuset", S_IRUGO, proc_cpuset_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002529#endif
Paul Menagea4243162007-10-18 23:39:35 -07002530#ifdef CONFIG_CGROUPS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002531 REG("cgroup", S_IRUGO, proc_cgroup_operations),
Paul Menagea4243162007-10-18 23:39:35 -07002532#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002533 INF("oom_score", S_IRUGO, proc_oom_score),
2534 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002535#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002536 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2537 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002538#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002539#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002540 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002541#endif
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002542#if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002543 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002544#endif
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002545#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002546 INF("io", S_IRUGO, proc_tgid_io_accounting),
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002547#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002548};
2549
2550static int proc_tgid_base_readdir(struct file * filp,
2551 void * dirent, filldir_t filldir)
2552{
2553 return proc_pident_readdir(filp,dirent,filldir,
2554 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2555}
2556
Arjan van de Ven00977a52007-02-12 00:55:34 -08002557static const struct file_operations proc_tgid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002558 .read = generic_read_dir,
2559 .readdir = proc_tgid_base_readdir,
2560};
2561
2562static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002563 return proc_pident_lookup(dir, dentry,
2564 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002565}
2566
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002567static const struct inode_operations proc_tgid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002568 .lookup = proc_tgid_base_lookup,
2569 .getattr = pid_getattr,
2570 .setattr = proc_setattr,
2571};
2572
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002573static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002574{
Eric W. Biederman48e64842006-06-26 00:25:48 -07002575 struct dentry *dentry, *leader, *dir;
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002576 char buf[PROC_NUMBUF];
Eric W. Biederman48e64842006-06-26 00:25:48 -07002577 struct qstr name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002578
Eric W. Biederman48e64842006-06-26 00:25:48 -07002579 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002580 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2581 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002582 if (dentry) {
Andrea Arcangeli77667552008-02-04 22:29:21 -08002583 if (!(current->flags & PF_EXITING))
2584 shrink_dcache_parent(dentry);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002585 d_drop(dentry);
2586 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002587 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002588
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002589 if (tgid == 0)
Eric W. Biederman48e64842006-06-26 00:25:48 -07002590 goto out;
2591
2592 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002593 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2594 leader = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002595 if (!leader)
2596 goto out;
2597
2598 name.name = "task";
2599 name.len = strlen(name.name);
2600 dir = d_hash_and_lookup(leader, &name);
2601 if (!dir)
2602 goto out_put_leader;
2603
2604 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002605 name.len = snprintf(buf, sizeof(buf), "%d", pid);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002606 dentry = d_hash_and_lookup(dir, &name);
2607 if (dentry) {
2608 shrink_dcache_parent(dentry);
2609 d_drop(dentry);
2610 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002611 }
Eric W. Biederman48e64842006-06-26 00:25:48 -07002612
2613 dput(dir);
2614out_put_leader:
2615 dput(leader);
2616out:
2617 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002618}
2619
Randy Dunlap0895e912007-10-21 21:00:10 -07002620/**
2621 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2622 * @task: task that should be flushed.
2623 *
2624 * When flushing dentries from proc, one needs to flush them from global
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002625 * proc (proc_mnt) and from all the namespaces' procs this task was seen
Randy Dunlap0895e912007-10-21 21:00:10 -07002626 * in. This call is supposed to do all of this job.
2627 *
2628 * Looks in the dcache for
2629 * /proc/@pid
2630 * /proc/@tgid/task/@pid
2631 * if either directory is present flushes it and all of it'ts children
2632 * from the dcache.
2633 *
2634 * It is safe and reasonable to cache /proc entries for a task until
2635 * that task exits. After that they just clog up the dcache with
2636 * useless entries, possibly causing useful dcache entries to be
2637 * flushed instead. This routine is proved to flush those useless
2638 * dcache entries at process exit time.
2639 *
2640 * NOTE: This routine is just an optimization so it does not guarantee
2641 * that no dcache entries will exist at process exit time it
2642 * just makes it very unlikely that any will persist.
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002643 */
2644
2645void proc_flush_task(struct task_struct *task)
2646{
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002647 int i;
2648 struct pid *pid, *tgid = NULL;
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002649 struct upid *upid;
2650
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002651 pid = task_pid(task);
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002652 if (thread_group_leader(task))
2653 tgid = task_tgid(task);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002654
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002655 for (i = 0; i <= pid->level; i++) {
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002656 upid = &pid->numbers[i];
2657 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002658 tgid ? tgid->numbers[i].nr : 0);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002659 }
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -07002660
2661 upid = &pid->numbers[pid->level];
2662 if (upid->nr == 1)
2663 pid_ns_release_proc(upid->ns);
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002664}
2665
Adrian Bunk9711ef92006-12-06 20:38:31 -08002666static struct dentry *proc_pid_instantiate(struct inode *dir,
2667 struct dentry * dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002668 struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002669{
2670 struct dentry *error = ERR_PTR(-ENOENT);
2671 struct inode *inode;
2672
Eric W. Biederman61a28782006-10-02 02:18:49 -07002673 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002674 if (!inode)
2675 goto out;
2676
2677 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2678 inode->i_op = &proc_tgid_base_inode_operations;
2679 inode->i_fop = &proc_tgid_base_operations;
2680 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07002681
2682 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2683 ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002684
2685 dentry->d_op = &pid_dentry_operations;
2686
2687 d_add(dentry, inode);
2688 /* Close the race of the process dying before we return the dentry */
2689 if (pid_revalidate(dentry, NULL))
2690 error = NULL;
2691out:
2692 return error;
2693}
2694
Linus Torvalds1da177e2005-04-16 15:20:36 -07002695struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2696{
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002697 struct dentry *result = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002698 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002699 unsigned tgid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002700 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002701
Eric W. Biederman801199c2006-10-02 02:18:48 -07002702 result = proc_base_lookup(dir, dentry);
2703 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2704 goto out;
2705
Linus Torvalds1da177e2005-04-16 15:20:36 -07002706 tgid = name_to_int(dentry);
2707 if (tgid == ~0U)
2708 goto out;
2709
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002710 ns = dentry->d_sb->s_fs_info;
Eric W. Biedermande758732006-06-26 00:25:51 -07002711 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002712 task = find_task_by_pid_ns(tgid, ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002713 if (task)
2714 get_task_struct(task);
Eric W. Biedermande758732006-06-26 00:25:51 -07002715 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002716 if (!task)
2717 goto out;
2718
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002719 result = proc_pid_instantiate(dir, dentry, task, NULL);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002720 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002721out:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002722 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002723}
2724
Linus Torvalds1da177e2005-04-16 15:20:36 -07002725/*
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002726 * Find the first task with tgid >= tgid
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002727 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002728 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002729struct tgid_iter {
2730 unsigned int tgid;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002731 struct task_struct *task;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002732};
2733static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2734{
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002735 struct pid *pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002736
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002737 if (iter.task)
2738 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002739 rcu_read_lock();
2740retry:
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002741 iter.task = NULL;
2742 pid = find_ge_pid(iter.tgid, ns);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002743 if (pid) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002744 iter.tgid = pid_nr_ns(pid, ns);
2745 iter.task = pid_task(pid, PIDTYPE_PID);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002746 /* What we to know is if the pid we have find is the
2747 * pid of a thread_group_leader. Testing for task
2748 * being a thread_group_leader is the obvious thing
2749 * todo but there is a window when it fails, due to
2750 * the pid transfer logic in de_thread.
2751 *
2752 * So we perform the straight forward test of seeing
2753 * if the pid we have found is the pid of a thread
2754 * group leader, and don't worry if the task we have
2755 * found doesn't happen to be a thread group leader.
2756 * As we don't care in the case of readdir.
2757 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002758 if (!iter.task || !has_group_leader_pid(iter.task)) {
2759 iter.tgid += 1;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002760 goto retry;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002761 }
2762 get_task_struct(iter.task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002763 }
Eric W. Biederman454cc102006-06-26 00:25:51 -07002764 rcu_read_unlock();
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002765 return iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002766}
2767
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002768#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002769
Eric W. Biederman61a28782006-10-02 02:18:49 -07002770static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002771 struct tgid_iter iter)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002772{
2773 char name[PROC_NUMBUF];
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002774 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
Eric W. Biederman61a28782006-10-02 02:18:49 -07002775 return proc_fill_cache(filp, dirent, filldir, name, len,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002776 proc_pid_instantiate, iter.task, NULL);
Eric W. Biederman61a28782006-10-02 02:18:49 -07002777}
2778
Linus Torvalds1da177e2005-04-16 15:20:36 -07002779/* for the /proc/ directory itself, after non-process stuff has been done */
2780int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2781{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002782 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002783 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002784 struct tgid_iter iter;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002785 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002786
Eric W. Biederman61a28782006-10-02 02:18:49 -07002787 if (!reaper)
2788 goto out_no_task;
2789
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002790 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
Eric Dumazetc5141e62007-05-08 00:26:15 -07002791 const struct pid_entry *p = &proc_base_stuff[nr];
Eric W. Biederman61a28782006-10-02 02:18:49 -07002792 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
Eric W. Biederman801199c2006-10-02 02:18:48 -07002793 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002794 }
2795
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002796 ns = filp->f_dentry->d_sb->s_fs_info;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002797 iter.task = NULL;
2798 iter.tgid = filp->f_pos - TGID_OFFSET;
2799 for (iter = next_tgid(ns, iter);
2800 iter.task;
2801 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2802 filp->f_pos = iter.tgid + TGID_OFFSET;
2803 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
2804 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002805 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002806 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002807 }
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002808 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2809out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07002810 put_task_struct(reaper);
2811out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002812 return 0;
2813}
2814
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002815/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002816 * Tasks
2817 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07002818static const struct pid_entry tid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002819 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2820 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fd_operations),
2821 REG("environ", S_IRUSR, proc_environ_operations),
2822 INF("auxv", S_IRUSR, proc_pid_auxv),
2823 ONE("status", S_IRUGO, proc_pid_status),
2824 ONE("personality", S_IRUSR, proc_pid_personality),
2825 INF("limits", S_IRUSR, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002826#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002827 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002828#endif
Roland McGrathebcb6732008-07-25 19:46:00 -07002829#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002830 INF("syscall", S_IRUSR, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07002831#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002832 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2833 ONE("stat", S_IRUGO, proc_tid_stat),
2834 ONE("statm", S_IRUGO, proc_pid_statm),
2835 REG("maps", S_IRUGO, proc_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002836#ifdef CONFIG_NUMA
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002837 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002838#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002839 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2840 LNK("cwd", proc_cwd_link),
2841 LNK("root", proc_root_link),
2842 LNK("exe", proc_exe_link),
2843 REG("mounts", S_IRUGO, proc_mounts_operations),
2844 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08002845#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002846 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2847 REG("smaps", S_IRUGO, proc_smaps_operations),
2848 REG("pagemap", S_IRUSR, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002849#endif
2850#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002851 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002852#endif
2853#ifdef CONFIG_KALLSYMS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002854 INF("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002855#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03002856#ifdef CONFIG_STACKTRACE
2857 ONE("stack", S_IRUSR, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002858#endif
2859#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002860 INF("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002861#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01002862#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002863 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01002864#endif
Paul Menage8793d852007-10-18 23:39:39 -07002865#ifdef CONFIG_PROC_PID_CPUSET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002866 REG("cpuset", S_IRUGO, proc_cpuset_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002867#endif
Paul Menagea4243162007-10-18 23:39:35 -07002868#ifdef CONFIG_CGROUPS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002869 REG("cgroup", S_IRUGO, proc_cgroup_operations),
Paul Menagea4243162007-10-18 23:39:35 -07002870#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002871 INF("oom_score", S_IRUGO, proc_oom_score),
2872 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002873#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002874 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2875 REG("sessionid", S_IRUSR, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002876#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002877#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002878 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002879#endif
Andrea Righi297c5d92008-07-25 01:48:49 -07002880#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002881 INF("io", S_IRUGO, proc_tid_io_accounting),
Andrea Righi297c5d92008-07-25 01:48:49 -07002882#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002883};
2884
2885static int proc_tid_base_readdir(struct file * filp,
2886 void * dirent, filldir_t filldir)
2887{
2888 return proc_pident_readdir(filp,dirent,filldir,
2889 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2890}
2891
2892static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002893 return proc_pident_lookup(dir, dentry,
2894 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002895}
2896
Arjan van de Ven00977a52007-02-12 00:55:34 -08002897static const struct file_operations proc_tid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002898 .read = generic_read_dir,
2899 .readdir = proc_tid_base_readdir,
2900};
2901
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002902static const struct inode_operations proc_tid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002903 .lookup = proc_tid_base_lookup,
2904 .getattr = pid_getattr,
2905 .setattr = proc_setattr,
2906};
2907
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002908static struct dentry *proc_task_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002909 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002910{
2911 struct dentry *error = ERR_PTR(-ENOENT);
2912 struct inode *inode;
Eric W. Biederman61a28782006-10-02 02:18:49 -07002913 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002914
2915 if (!inode)
2916 goto out;
2917 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2918 inode->i_op = &proc_tid_base_inode_operations;
2919 inode->i_fop = &proc_tid_base_operations;
2920 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07002921
2922 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
2923 ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002924
2925 dentry->d_op = &pid_dentry_operations;
2926
2927 d_add(dentry, inode);
2928 /* Close the race of the process dying before we return the dentry */
2929 if (pid_revalidate(dentry, NULL))
2930 error = NULL;
2931out:
2932 return error;
2933}
2934
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002935static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2936{
2937 struct dentry *result = ERR_PTR(-ENOENT);
2938 struct task_struct *task;
2939 struct task_struct *leader = get_proc_task(dir);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002940 unsigned tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002941 struct pid_namespace *ns;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002942
2943 if (!leader)
2944 goto out_no_task;
2945
2946 tid = name_to_int(dentry);
2947 if (tid == ~0U)
2948 goto out;
2949
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002950 ns = dentry->d_sb->s_fs_info;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002951 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002952 task = find_task_by_pid_ns(tid, ns);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002953 if (task)
2954 get_task_struct(task);
2955 rcu_read_unlock();
2956 if (!task)
2957 goto out;
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07002958 if (!same_thread_group(leader, task))
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002959 goto out_drop_task;
2960
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002961 result = proc_task_instantiate(dir, dentry, task, NULL);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002962out_drop_task:
2963 put_task_struct(task);
2964out:
2965 put_task_struct(leader);
2966out_no_task:
2967 return result;
2968}
2969
2970/*
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002971 * Find the first tid of a thread group to return to user space.
2972 *
2973 * Usually this is just the thread group leader, but if the users
2974 * buffer was too small or there was a seek into the middle of the
2975 * directory we have more work todo.
2976 *
2977 * In the case of a short read we start with find_task_by_pid.
2978 *
2979 * In the case of a seek we start with the leader and walk nr
2980 * threads past it.
2981 */
Eric W. Biedermancc288732006-06-26 00:26:01 -07002982static struct task_struct *first_tid(struct task_struct *leader,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002983 int tid, int nr, struct pid_namespace *ns)
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002984{
Oleg Nesterova872ff02006-06-26 00:26:01 -07002985 struct task_struct *pos;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002986
Eric W. Biedermancc288732006-06-26 00:26:01 -07002987 rcu_read_lock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002988 /* Attempt to start with the pid of a thread */
2989 if (tid && (nr > 0)) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002990 pos = find_task_by_pid_ns(tid, ns);
Oleg Nesterova872ff02006-06-26 00:26:01 -07002991 if (pos && (pos->group_leader == leader))
2992 goto found;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002993 }
2994
2995 /* If nr exceeds the number of threads there is nothing todo */
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002996 pos = NULL;
Oleg Nesterova872ff02006-06-26 00:26:01 -07002997 if (nr && nr >= get_nr_threads(leader))
2998 goto out;
2999
3000 /* If we haven't found our starting place yet start
3001 * with the leader and walk nr threads forward.
3002 */
3003 for (pos = leader; nr > 0; --nr) {
3004 pos = next_thread(pos);
3005 if (pos == leader) {
3006 pos = NULL;
3007 goto out;
3008 }
3009 }
3010found:
3011 get_task_struct(pos);
3012out:
Eric W. Biedermancc288732006-06-26 00:26:01 -07003013 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003014 return pos;
3015}
3016
3017/*
3018 * Find the next thread in the thread list.
3019 * Return NULL if there is an error or no next thread.
3020 *
3021 * The reference to the input task_struct is released.
3022 */
3023static struct task_struct *next_tid(struct task_struct *start)
3024{
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003025 struct task_struct *pos = NULL;
Eric W. Biedermancc288732006-06-26 00:26:01 -07003026 rcu_read_lock();
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003027 if (pid_alive(start)) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003028 pos = next_thread(start);
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003029 if (thread_group_leader(pos))
3030 pos = NULL;
3031 else
3032 get_task_struct(pos);
3033 }
Eric W. Biedermancc288732006-06-26 00:26:01 -07003034 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003035 put_task_struct(start);
3036 return pos;
3037}
3038
Eric W. Biederman61a28782006-10-02 02:18:49 -07003039static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3040 struct task_struct *task, int tid)
3041{
3042 char name[PROC_NUMBUF];
3043 int len = snprintf(name, sizeof(name), "%d", tid);
3044 return proc_fill_cache(filp, dirent, filldir, name, len,
3045 proc_task_instantiate, task, NULL);
3046}
3047
Linus Torvalds1da177e2005-04-16 15:20:36 -07003048/* for the /proc/TGID/task/ directories */
3049static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3050{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08003051 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003052 struct inode *inode = dentry->d_inode;
Guillaume Chazarain7d895242007-01-31 23:48:14 -08003053 struct task_struct *leader = NULL;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003054 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003055 int retval = -ENOENT;
3056 ino_t ino;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003057 int tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003058 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003059
Guillaume Chazarain7d895242007-01-31 23:48:14 -08003060 task = get_proc_task(inode);
3061 if (!task)
3062 goto out_no_task;
3063 rcu_read_lock();
3064 if (pid_alive(task)) {
3065 leader = task->group_leader;
3066 get_task_struct(leader);
3067 }
3068 rcu_read_unlock();
3069 put_task_struct(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003070 if (!leader)
3071 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003072 retval = 0;
3073
Linus Torvaldsee568b22009-03-17 10:02:35 -07003074 switch ((unsigned long)filp->f_pos) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003075 case 0:
3076 ino = inode->i_ino;
Zhang Leee6f7792009-03-16 14:44:31 +08003077 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003078 goto out;
Zhang Leee6f7792009-03-16 14:44:31 +08003079 filp->f_pos++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003080 /* fall through */
3081 case 1:
3082 ino = parent_ino(dentry);
Zhang Leee6f7792009-03-16 14:44:31 +08003083 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003084 goto out;
Zhang Leee6f7792009-03-16 14:44:31 +08003085 filp->f_pos++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003086 /* fall through */
3087 }
3088
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003089 /* f_version caches the tgid value that the last readdir call couldn't
3090 * return. lseek aka telldir automagically resets f_version to 0.
3091 */
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003092 ns = filp->f_dentry->d_sb->s_fs_info;
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07003093 tid = (int)filp->f_version;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003094 filp->f_version = 0;
Zhang Leee6f7792009-03-16 14:44:31 +08003095 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003096 task;
Zhang Leee6f7792009-03-16 14:44:31 +08003097 task = next_tid(task), filp->f_pos++) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003098 tid = task_pid_nr_ns(task, ns);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003099 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003100 /* returning this tgid failed, save it as the first
3101 * pid for the next readir call */
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07003102 filp->f_version = (u64)tid;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003103 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003104 break;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003105 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003106 }
3107out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07003108 put_task_struct(leader);
3109out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07003110 return retval;
3111}
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003112
3113static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3114{
3115 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07003116 struct task_struct *p = get_proc_task(inode);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003117 generic_fillattr(inode, stat);
3118
Eric W. Biederman99f89552006-06-26 00:25:55 -07003119 if (p) {
Eric W. Biederman99f89552006-06-26 00:25:55 -07003120 stat->nlink += get_nr_threads(p);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003121 put_task_struct(p);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003122 }
3123
3124 return 0;
3125}
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003126
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003127static const struct inode_operations proc_task_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003128 .lookup = proc_task_lookup,
3129 .getattr = proc_task_getattr,
3130 .setattr = proc_setattr,
3131};
3132
Arjan van de Ven00977a52007-02-12 00:55:34 -08003133static const struct file_operations proc_task_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003134 .read = generic_read_dir,
3135 .readdir = proc_task_readdir,
3136};