blob: fb45615943c2adba5e630fd1ea14a1d9e1c31fff [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{
237 struct mm_struct *mm = get_task_mm(task);
238 if (!mm)
239 return NULL;
240 down_read(&mm->mmap_sem);
241 task_lock(task);
242 if (task->mm != mm)
243 goto out;
Stephen Smalley006ebb42008-05-19 08:32:49 -0400244 if (task->mm != current->mm &&
245 __ptrace_may_access(task, PTRACE_MODE_READ) < 0)
Al Viro831830b2008-01-02 14:09:57 +0000246 goto out;
247 task_unlock(task);
248 return mm;
249out:
250 task_unlock(task);
251 up_read(&mm->mmap_sem);
252 mmput(mm);
253 return NULL;
254}
255
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256static int proc_pid_cmdline(struct task_struct *task, char * buffer)
257{
258 int res = 0;
259 unsigned int len;
260 struct mm_struct *mm = get_task_mm(task);
261 if (!mm)
262 goto out;
263 if (!mm->arg_end)
264 goto out_mm; /* Shh! No looking before we're done */
265
266 len = mm->arg_end - mm->arg_start;
267
268 if (len > PAGE_SIZE)
269 len = PAGE_SIZE;
270
271 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
272
273 // If the nul at the end of args has been overwritten, then
274 // assume application is using setproctitle(3).
275 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
276 len = strnlen(buffer, res);
277 if (len < res) {
278 res = len;
279 } else {
280 len = mm->env_end - mm->env_start;
281 if (len > PAGE_SIZE - res)
282 len = PAGE_SIZE - res;
283 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
284 res = strnlen(buffer, res);
285 }
286 }
287out_mm:
288 mmput(mm);
289out:
290 return res;
291}
292
293static int proc_pid_auxv(struct task_struct *task, char *buffer)
294{
295 int res = 0;
296 struct mm_struct *mm = get_task_mm(task);
297 if (mm) {
298 unsigned int nwords = 0;
Hannes Ederdfe6b7d2008-12-30 18:49:13 +0300299 do {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 nwords += 2;
Hannes Ederdfe6b7d2008-12-30 18:49:13 +0300301 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 res = nwords * sizeof(mm->saved_auxv[0]);
303 if (res > PAGE_SIZE)
304 res = PAGE_SIZE;
305 memcpy(buffer, mm->saved_auxv, res);
306 mmput(mm);
307 }
308 return res;
309}
310
311
312#ifdef CONFIG_KALLSYMS
313/*
314 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
315 * Returns the resolved symbol. If that fails, simply return the address.
316 */
317static int proc_pid_wchan(struct task_struct *task, char *buffer)
318{
Alexey Dobriyanffb45122007-05-08 00:28:41 -0700319 unsigned long wchan;
Tejun Heo9281ace2007-07-17 04:03:51 -0700320 char symname[KSYM_NAME_LEN];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321
322 wchan = get_wchan(task);
323
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700324 if (lookup_symbol_name(wchan, symname) < 0)
Jake Edgef83ce3e2009-05-04 12:51:14 -0600325 if (!ptrace_may_access(task, PTRACE_MODE_READ))
326 return 0;
327 else
328 return sprintf(buffer, "%lu", wchan);
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700329 else
330 return sprintf(buffer, "%s", symname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331}
332#endif /* CONFIG_KALLSYMS */
333
Ken Chen2ec220e2008-11-10 11:26:08 +0300334#ifdef CONFIG_STACKTRACE
335
336#define MAX_STACK_TRACE_DEPTH 64
337
338static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
339 struct pid *pid, struct task_struct *task)
340{
341 struct stack_trace trace;
342 unsigned long *entries;
343 int i;
344
345 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
346 if (!entries)
347 return -ENOMEM;
348
349 trace.nr_entries = 0;
350 trace.max_entries = MAX_STACK_TRACE_DEPTH;
351 trace.entries = entries;
352 trace.skip = 0;
353 save_stack_trace_tsk(task, &trace);
354
355 for (i = 0; i < trace.nr_entries; i++) {
356 seq_printf(m, "[<%p>] %pS\n",
357 (void *)entries[i], (void *)entries[i]);
358 }
359 kfree(entries);
360
361 return 0;
362}
363#endif
364
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365#ifdef CONFIG_SCHEDSTATS
366/*
367 * Provides /proc/PID/schedstat
368 */
369static int proc_pid_schedstat(struct task_struct *task, char *buffer)
370{
Balbir Singh172ba842007-07-09 18:52:00 +0200371 return sprintf(buffer, "%llu %llu %lu\n",
Ingo Molnar826e08b2008-12-22 07:37:41 +0100372 (unsigned long long)task->se.sum_exec_runtime,
373 (unsigned long long)task->sched_info.run_delay,
Ingo Molnar2d723762007-10-15 17:00:12 +0200374 task->sched_info.pcount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375}
376#endif
377
Arjan van de Ven97455122008-01-25 21:08:34 +0100378#ifdef CONFIG_LATENCYTOP
379static int lstats_show_proc(struct seq_file *m, void *v)
380{
381 int i;
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800382 struct inode *inode = m->private;
383 struct task_struct *task = get_proc_task(inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100384
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800385 if (!task)
386 return -ESRCH;
387 seq_puts(m, "Latency Top version : v0.1\n");
Arjan van de Ven97455122008-01-25 21:08:34 +0100388 for (i = 0; i < 32; i++) {
389 if (task->latency_record[i].backtrace[0]) {
390 int q;
391 seq_printf(m, "%i %li %li ",
392 task->latency_record[i].count,
393 task->latency_record[i].time,
394 task->latency_record[i].max);
395 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
Hugh Dickins9c246242008-12-09 13:14:27 -0800396 char sym[KSYM_SYMBOL_LEN];
Arjan van de Ven97455122008-01-25 21:08:34 +0100397 char *c;
398 if (!task->latency_record[i].backtrace[q])
399 break;
400 if (task->latency_record[i].backtrace[q] == ULONG_MAX)
401 break;
402 sprint_symbol(sym, task->latency_record[i].backtrace[q]);
403 c = strchr(sym, '+');
404 if (c)
405 *c = 0;
406 seq_printf(m, "%s ", sym);
407 }
408 seq_printf(m, "\n");
409 }
410
411 }
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800412 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100413 return 0;
414}
415
416static int lstats_open(struct inode *inode, struct file *file)
417{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800418 return single_open(file, lstats_show_proc, inode);
Hiroshi Shimamotod6643d12008-02-14 10:27:00 -0800419}
420
Arjan van de Ven97455122008-01-25 21:08:34 +0100421static ssize_t lstats_write(struct file *file, const char __user *buf,
422 size_t count, loff_t *offs)
423{
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800424 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
Arjan van de Ven97455122008-01-25 21:08:34 +0100425
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800426 if (!task)
427 return -ESRCH;
Arjan van de Ven97455122008-01-25 21:08:34 +0100428 clear_all_latency_tracing(task);
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800429 put_task_struct(task);
Arjan van de Ven97455122008-01-25 21:08:34 +0100430
431 return count;
432}
433
434static const struct file_operations proc_lstats_operations = {
435 .open = lstats_open,
436 .read = seq_read,
437 .write = lstats_write,
438 .llseek = seq_lseek,
Hiroshi Shimamoto13d77c32008-02-20 16:53:29 -0800439 .release = single_release,
Arjan van de Ven97455122008-01-25 21:08:34 +0100440};
441
442#endif
443
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444/* The badness from the OOM killer */
445unsigned long badness(struct task_struct *p, unsigned long uptime);
446static int proc_oom_score(struct task_struct *task, char *buffer)
447{
448 unsigned long points;
449 struct timespec uptime;
450
451 do_posix_clock_monotonic_gettime(&uptime);
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700452 read_lock(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 points = badness(task, uptime.tv_sec);
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700454 read_unlock(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 return sprintf(buffer, "%lu\n", points);
456}
457
Neil Hormand85f50d2007-10-18 23:40:37 -0700458struct limit_names {
459 char *name;
460 char *unit;
461};
462
463static const struct limit_names lnames[RLIM_NLIMITS] = {
464 [RLIMIT_CPU] = {"Max cpu time", "ms"},
465 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
466 [RLIMIT_DATA] = {"Max data size", "bytes"},
467 [RLIMIT_STACK] = {"Max stack size", "bytes"},
468 [RLIMIT_CORE] = {"Max core file size", "bytes"},
469 [RLIMIT_RSS] = {"Max resident set", "bytes"},
470 [RLIMIT_NPROC] = {"Max processes", "processes"},
471 [RLIMIT_NOFILE] = {"Max open files", "files"},
472 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
473 [RLIMIT_AS] = {"Max address space", "bytes"},
474 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
475 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
476 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
477 [RLIMIT_NICE] = {"Max nice priority", NULL},
478 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
Eugene Teo88081172008-02-23 15:23:52 -0800479 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
Neil Hormand85f50d2007-10-18 23:40:37 -0700480};
481
482/* Display limits for a process */
483static int proc_pid_limits(struct task_struct *task, char *buffer)
484{
485 unsigned int i;
486 int count = 0;
487 unsigned long flags;
488 char *bufptr = buffer;
489
490 struct rlimit rlim[RLIM_NLIMITS];
491
Lai Jiangshana6bebbc2008-10-05 00:51:15 +0400492 if (!lock_task_sighand(task, &flags))
Neil Hormand85f50d2007-10-18 23:40:37 -0700493 return 0;
Neil Hormand85f50d2007-10-18 23:40:37 -0700494 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
495 unlock_task_sighand(task, &flags);
Neil Hormand85f50d2007-10-18 23:40:37 -0700496
497 /*
498 * print the file header
499 */
500 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
501 "Limit", "Soft Limit", "Hard Limit", "Units");
502
503 for (i = 0; i < RLIM_NLIMITS; i++) {
504 if (rlim[i].rlim_cur == RLIM_INFINITY)
505 count += sprintf(&bufptr[count], "%-25s %-20s ",
506 lnames[i].name, "unlimited");
507 else
508 count += sprintf(&bufptr[count], "%-25s %-20lu ",
509 lnames[i].name, rlim[i].rlim_cur);
510
511 if (rlim[i].rlim_max == RLIM_INFINITY)
512 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
513 else
514 count += sprintf(&bufptr[count], "%-20lu ",
515 rlim[i].rlim_max);
516
517 if (lnames[i].unit)
518 count += sprintf(&bufptr[count], "%-10s\n",
519 lnames[i].unit);
520 else
521 count += sprintf(&bufptr[count], "\n");
522 }
523
524 return count;
525}
526
Roland McGrathebcb6732008-07-25 19:46:00 -0700527#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
528static int proc_pid_syscall(struct task_struct *task, char *buffer)
529{
530 long nr;
531 unsigned long args[6], sp, pc;
532
533 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
534 return sprintf(buffer, "running\n");
535
536 if (nr < 0)
537 return sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
538
539 return sprintf(buffer,
540 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
541 nr,
542 args[0], args[1], args[2], args[3], args[4], args[5],
543 sp, pc);
544}
545#endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
546
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547/************************************************************************/
548/* Here the fs part begins */
549/************************************************************************/
550
551/* permission checks */
Eric W. Biederman778c1142006-06-26 00:25:58 -0700552static int proc_fd_access_allowed(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553{
Eric W. Biederman778c1142006-06-26 00:25:58 -0700554 struct task_struct *task;
555 int allowed = 0;
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700556 /* Allow access to a task's file descriptors if it is us or we
557 * may use ptrace attach to the process and find out that
558 * information.
Eric W. Biederman778c1142006-06-26 00:25:58 -0700559 */
560 task = get_proc_task(inode);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700561 if (task) {
Stephen Smalley006ebb42008-05-19 08:32:49 -0400562 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
Eric W. Biederman778c1142006-06-26 00:25:58 -0700563 put_task_struct(task);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700564 }
Eric W. Biederman778c1142006-06-26 00:25:58 -0700565 return allowed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566}
567
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700568static int proc_setattr(struct dentry *dentry, struct iattr *attr)
569{
570 int error;
571 struct inode *inode = dentry->d_inode;
572
573 if (attr->ia_valid & ATTR_MODE)
574 return -EPERM;
575
576 error = inode_change_ok(inode, attr);
John Johansen1e8123f2007-05-08 00:29:41 -0700577 if (!error)
578 error = inode_setattr(inode, attr);
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700579 return error;
580}
581
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800582static const struct inode_operations proc_def_inode_operations = {
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700583 .setattr = proc_setattr,
584};
585
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100586static int mounts_open_common(struct inode *inode, struct file *file,
587 const struct seq_operations *op)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700589 struct task_struct *task = get_proc_task(inode);
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700590 struct nsproxy *nsp;
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800591 struct mnt_namespace *ns = NULL;
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100592 struct path root;
Al Viro5addc5d2005-11-07 17:15:49 -0500593 struct proc_mounts *p;
594 int ret = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595
Eric W. Biederman99f89552006-06-26 00:25:55 -0700596 if (task) {
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700597 rcu_read_lock();
598 nsp = task_nsproxy(task);
599 if (nsp) {
600 ns = nsp->mnt_ns;
Alexey Dobriyan863c4702007-01-26 00:56:53 -0800601 if (ns)
602 get_mnt_ns(ns);
603 }
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700604 rcu_read_unlock();
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000605 if (ns && get_fs_path(task, &root, 1) == 0)
606 ret = 0;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700607 put_task_struct(task);
608 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100610 if (!ns)
611 goto err;
Hugh Dickins7c2c7d92009-03-28 23:21:27 +0000612 if (ret)
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100613 goto err_put_ns;
614
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100615 ret = -ENOMEM;
616 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
617 if (!p)
618 goto err_put_path;
619
620 file->private_data = &p->m;
621 ret = seq_open(file, op);
622 if (ret)
623 goto err_free;
624
625 p->m.private = p;
626 p->ns = ns;
627 p->root = root;
628 p->event = ns->event;
629
630 return 0;
631
632 err_free:
633 kfree(p);
634 err_put_path:
635 path_put(&root);
636 err_put_ns:
637 put_mnt_ns(ns);
638 err:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639 return ret;
640}
641
642static int mounts_release(struct inode *inode, struct file *file)
643{
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100644 struct proc_mounts *p = file->private_data;
645 path_put(&p->root);
646 put_mnt_ns(p->ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 return seq_release(inode, file);
648}
649
Al Viro5addc5d2005-11-07 17:15:49 -0500650static unsigned mounts_poll(struct file *file, poll_table *wait)
651{
652 struct proc_mounts *p = file->private_data;
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100653 struct mnt_namespace *ns = p->ns;
KOSAKI Motohiro31b07092009-04-09 13:57:59 +0900654 unsigned res = POLLIN | POLLRDNORM;
Al Viro5addc5d2005-11-07 17:15:49 -0500655
656 poll_wait(file, &ns->poll, wait);
657
658 spin_lock(&vfsmount_lock);
659 if (p->event != ns->event) {
660 p->event = ns->event;
KOSAKI Motohiro31b07092009-04-09 13:57:59 +0900661 res |= POLLERR | POLLPRI;
Al Viro5addc5d2005-11-07 17:15:49 -0500662 }
663 spin_unlock(&vfsmount_lock);
664
665 return res;
666}
667
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100668static int mounts_open(struct inode *inode, struct file *file)
669{
670 return mounts_open_common(inode, file, &mounts_op);
671}
672
Arjan van de Ven00977a52007-02-12 00:55:34 -0800673static const struct file_operations proc_mounts_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 .open = mounts_open,
675 .read = seq_read,
676 .llseek = seq_lseek,
677 .release = mounts_release,
Al Viro5addc5d2005-11-07 17:15:49 -0500678 .poll = mounts_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679};
680
Ram Pai2d4d4862008-03-27 13:06:25 +0100681static int mountinfo_open(struct inode *inode, struct file *file)
682{
683 return mounts_open_common(inode, file, &mountinfo_op);
684}
685
686static const struct file_operations proc_mountinfo_operations = {
687 .open = mountinfo_open,
688 .read = seq_read,
689 .llseek = seq_lseek,
690 .release = mounts_release,
691 .poll = mounts_poll,
692};
693
Chuck Leverb4629fe2006-03-20 13:44:12 -0500694static int mountstats_open(struct inode *inode, struct file *file)
695{
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100696 return mounts_open_common(inode, file, &mountstats_op);
Chuck Leverb4629fe2006-03-20 13:44:12 -0500697}
698
Arjan van de Ven00977a52007-02-12 00:55:34 -0800699static const struct file_operations proc_mountstats_operations = {
Chuck Leverb4629fe2006-03-20 13:44:12 -0500700 .open = mountstats_open,
701 .read = seq_read,
702 .llseek = seq_lseek,
703 .release = mounts_release,
704};
705
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
707
708static ssize_t proc_info_read(struct file * file, char __user * buf,
709 size_t count, loff_t *ppos)
710{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800711 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712 unsigned long page;
713 ssize_t length;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700714 struct task_struct *task = get_proc_task(inode);
715
716 length = -ESRCH;
717 if (!task)
718 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719
720 if (count > PROC_BLOCK_SIZE)
721 count = PROC_BLOCK_SIZE;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700722
723 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700724 if (!(page = __get_free_page(GFP_TEMPORARY)))
Eric W. Biederman99f89552006-06-26 00:25:55 -0700725 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726
727 length = PROC_I(inode)->op.proc_read(task, (char*)page);
728
729 if (length >= 0)
730 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
731 free_page(page);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700732out:
733 put_task_struct(task);
734out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 return length;
736}
737
Arjan van de Ven00977a52007-02-12 00:55:34 -0800738static const struct file_operations proc_info_file_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739 .read = proc_info_read,
740};
741
Eric W. Biedermanbe614082008-02-08 04:18:30 -0800742static int proc_single_show(struct seq_file *m, void *v)
743{
744 struct inode *inode = m->private;
745 struct pid_namespace *ns;
746 struct pid *pid;
747 struct task_struct *task;
748 int ret;
749
750 ns = inode->i_sb->s_fs_info;
751 pid = proc_pid(inode);
752 task = get_pid_task(pid, PIDTYPE_PID);
753 if (!task)
754 return -ESRCH;
755
756 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
757
758 put_task_struct(task);
759 return ret;
760}
761
762static int proc_single_open(struct inode *inode, struct file *filp)
763{
764 int ret;
765 ret = single_open(filp, proc_single_show, NULL);
766 if (!ret) {
767 struct seq_file *m = filp->private_data;
768
769 m->private = inode;
770 }
771 return ret;
772}
773
774static const struct file_operations proc_single_file_operations = {
775 .open = proc_single_open,
776 .read = seq_read,
777 .llseek = seq_lseek,
778 .release = single_release,
779};
780
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781static int mem_open(struct inode* inode, struct file* file)
782{
783 file->private_data = (void*)((long)current->self_exec_id);
784 return 0;
785}
786
787static ssize_t mem_read(struct file * file, char __user * buf,
788 size_t count, loff_t *ppos)
789{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800790 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 char *page;
792 unsigned long src = *ppos;
793 int ret = -ESRCH;
794 struct mm_struct *mm;
795
Eric W. Biederman99f89552006-06-26 00:25:55 -0700796 if (!task)
797 goto out_no_task;
798
Roland McGrath638fa202008-04-29 01:01:38 -0700799 if (check_mem_permission(task))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 goto out;
801
802 ret = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700803 page = (char *)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 if (!page)
805 goto out;
806
807 ret = 0;
808
809 mm = get_task_mm(task);
810 if (!mm)
811 goto out_free;
812
813 ret = -EIO;
814
815 if (file->private_data != (void*)((long)current->self_exec_id))
816 goto out_put;
817
818 ret = 0;
819
820 while (count > 0) {
821 int this_len, retval;
822
823 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
824 retval = access_process_vm(task, src, page, this_len, 0);
Roland McGrath638fa202008-04-29 01:01:38 -0700825 if (!retval || check_mem_permission(task)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 if (!ret)
827 ret = -EIO;
828 break;
829 }
830
831 if (copy_to_user(buf, page, retval)) {
832 ret = -EFAULT;
833 break;
834 }
835
836 ret += retval;
837 src += retval;
838 buf += retval;
839 count -= retval;
840 }
841 *ppos = src;
842
843out_put:
844 mmput(mm);
845out_free:
846 free_page((unsigned long) page);
847out:
Eric W. Biederman99f89552006-06-26 00:25:55 -0700848 put_task_struct(task);
849out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 return ret;
851}
852
853#define mem_write NULL
854
855#ifndef mem_write
856/* This is a security hazard */
Glauber de Oliveira Costa63967fa2007-02-20 13:58:12 -0800857static ssize_t mem_write(struct file * file, const char __user *buf,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 size_t count, loff_t *ppos)
859{
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700860 int copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 char *page;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800862 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 unsigned long dst = *ppos;
864
Eric W. Biederman99f89552006-06-26 00:25:55 -0700865 copied = -ESRCH;
866 if (!task)
867 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868
Roland McGrath638fa202008-04-29 01:01:38 -0700869 if (check_mem_permission(task))
Eric W. Biederman99f89552006-06-26 00:25:55 -0700870 goto out;
871
872 copied = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700873 page = (char *)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 if (!page)
Eric W. Biederman99f89552006-06-26 00:25:55 -0700875 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700877 copied = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 while (count > 0) {
879 int this_len, retval;
880
881 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
882 if (copy_from_user(page, buf, this_len)) {
883 copied = -EFAULT;
884 break;
885 }
886 retval = access_process_vm(task, dst, page, this_len, 1);
887 if (!retval) {
888 if (!copied)
889 copied = -EIO;
890 break;
891 }
892 copied += retval;
893 buf += retval;
894 dst += retval;
895 count -= retval;
896 }
897 *ppos = dst;
898 free_page((unsigned long) page);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700899out:
900 put_task_struct(task);
901out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 return copied;
903}
904#endif
905
Matt Mackall85863e42008-02-04 22:29:04 -0800906loff_t mem_lseek(struct file *file, loff_t offset, int orig)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907{
908 switch (orig) {
909 case 0:
910 file->f_pos = offset;
911 break;
912 case 1:
913 file->f_pos += offset;
914 break;
915 default:
916 return -EINVAL;
917 }
918 force_successful_syscall_return();
919 return file->f_pos;
920}
921
Arjan van de Ven00977a52007-02-12 00:55:34 -0800922static const struct file_operations proc_mem_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 .llseek = mem_lseek,
924 .read = mem_read,
925 .write = mem_write,
926 .open = mem_open,
927};
928
James Pearson315e28c2007-10-16 23:30:17 -0700929static ssize_t environ_read(struct file *file, char __user *buf,
930 size_t count, loff_t *ppos)
931{
932 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
933 char *page;
934 unsigned long src = *ppos;
935 int ret = -ESRCH;
936 struct mm_struct *mm;
937
938 if (!task)
939 goto out_no_task;
940
Stephen Smalley006ebb42008-05-19 08:32:49 -0400941 if (!ptrace_may_access(task, PTRACE_MODE_READ))
James Pearson315e28c2007-10-16 23:30:17 -0700942 goto out;
943
944 ret = -ENOMEM;
945 page = (char *)__get_free_page(GFP_TEMPORARY);
946 if (!page)
947 goto out;
948
949 ret = 0;
950
951 mm = get_task_mm(task);
952 if (!mm)
953 goto out_free;
954
955 while (count > 0) {
956 int this_len, retval, max_len;
957
958 this_len = mm->env_end - (mm->env_start + src);
959
960 if (this_len <= 0)
961 break;
962
963 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
964 this_len = (this_len > max_len) ? max_len : this_len;
965
966 retval = access_process_vm(task, (mm->env_start + src),
967 page, this_len, 0);
968
969 if (retval <= 0) {
970 ret = retval;
971 break;
972 }
973
974 if (copy_to_user(buf, page, retval)) {
975 ret = -EFAULT;
976 break;
977 }
978
979 ret += retval;
980 src += retval;
981 buf += retval;
982 count -= retval;
983 }
984 *ppos = src;
985
986 mmput(mm);
987out_free:
988 free_page((unsigned long) page);
989out:
990 put_task_struct(task);
991out_no_task:
992 return ret;
993}
994
995static const struct file_operations proc_environ_operations = {
996 .read = environ_read,
997};
998
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999static ssize_t oom_adjust_read(struct file *file, char __user *buf,
1000 size_t count, loff_t *ppos)
1001{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001002 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001003 char buffer[PROC_NUMBUF];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 size_t len;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001005 int oom_adjust;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006
Eric W. Biederman99f89552006-06-26 00:25:55 -07001007 if (!task)
1008 return -ESRCH;
1009 oom_adjust = task->oomkilladj;
1010 put_task_struct(task);
1011
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001012 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001013
1014 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015}
1016
1017static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
1018 size_t count, loff_t *ppos)
1019{
Eric W. Biederman99f89552006-06-26 00:25:55 -07001020 struct task_struct *task;
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001021 char buffer[PROC_NUMBUF], *end;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 int oom_adjust;
1023
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001024 memset(buffer, 0, sizeof(buffer));
1025 if (count > sizeof(buffer) - 1)
1026 count = sizeof(buffer) - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 if (copy_from_user(buffer, buf, count))
1028 return -EFAULT;
1029 oom_adjust = simple_strtol(buffer, &end, 0);
Alexey Dobriyan8ac773b2006-10-19 23:28:32 -07001030 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
1031 oom_adjust != OOM_DISABLE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 return -EINVAL;
1033 if (*end == '\n')
1034 end++;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001035 task = get_proc_task(file->f_path.dentry->d_inode);
Eric W. Biederman99f89552006-06-26 00:25:55 -07001036 if (!task)
1037 return -ESRCH;
Guillem Jover8fb4fc62006-12-06 20:32:24 -08001038 if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) {
1039 put_task_struct(task);
1040 return -EACCES;
1041 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 task->oomkilladj = oom_adjust;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001043 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 if (end - buffer == 0)
1045 return -EIO;
1046 return end - buffer;
1047}
1048
Arjan van de Ven00977a52007-02-12 00:55:34 -08001049static const struct file_operations proc_oom_adjust_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 .read = oom_adjust_read,
1051 .write = oom_adjust_write,
1052};
1053
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054#ifdef CONFIG_AUDITSYSCALL
1055#define TMPBUFLEN 21
1056static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1057 size_t count, loff_t *ppos)
1058{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001059 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001060 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 ssize_t length;
1062 char tmpbuf[TMPBUFLEN];
1063
Eric W. Biederman99f89552006-06-26 00:25:55 -07001064 if (!task)
1065 return -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
Al Viro0c11b942008-01-10 04:20:52 -05001067 audit_get_loginuid(task));
Eric W. Biederman99f89552006-06-26 00:25:55 -07001068 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1070}
1071
1072static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1073 size_t count, loff_t *ppos)
1074{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001075 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 char *page, *tmp;
1077 ssize_t length;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 uid_t loginuid;
1079
1080 if (!capable(CAP_AUDIT_CONTROL))
1081 return -EPERM;
1082
Eric W. Biederman13b41b02006-06-26 00:25:56 -07001083 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 return -EPERM;
1085
Al Viroe0182902006-05-18 08:28:02 -04001086 if (count >= PAGE_SIZE)
1087 count = PAGE_SIZE - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088
1089 if (*ppos != 0) {
1090 /* No partial writes. */
1091 return -EINVAL;
1092 }
Mel Gormane12ba742007-10-16 01:25:52 -07001093 page = (char*)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 if (!page)
1095 return -ENOMEM;
1096 length = -EFAULT;
1097 if (copy_from_user(page, buf, count))
1098 goto out_free_page;
1099
Al Viroe0182902006-05-18 08:28:02 -04001100 page[count] = '\0';
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 loginuid = simple_strtoul(page, &tmp, 10);
1102 if (tmp == page) {
1103 length = -EINVAL;
1104 goto out_free_page;
1105
1106 }
Eric W. Biederman99f89552006-06-26 00:25:55 -07001107 length = audit_set_loginuid(current, loginuid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 if (likely(length == 0))
1109 length = count;
1110
1111out_free_page:
1112 free_page((unsigned long) page);
1113 return length;
1114}
1115
Arjan van de Ven00977a52007-02-12 00:55:34 -08001116static const struct file_operations proc_loginuid_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 .read = proc_loginuid_read,
1118 .write = proc_loginuid_write,
1119};
Eric Paris1e0bd752008-03-13 08:15:31 -04001120
1121static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1122 size_t count, loff_t *ppos)
1123{
1124 struct inode * inode = file->f_path.dentry->d_inode;
1125 struct task_struct *task = get_proc_task(inode);
1126 ssize_t length;
1127 char tmpbuf[TMPBUFLEN];
1128
1129 if (!task)
1130 return -ESRCH;
1131 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1132 audit_get_sessionid(task));
1133 put_task_struct(task);
1134 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1135}
1136
1137static const struct file_operations proc_sessionid_operations = {
1138 .read = proc_sessionid_read,
1139};
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140#endif
1141
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001142#ifdef CONFIG_FAULT_INJECTION
1143static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1144 size_t count, loff_t *ppos)
1145{
1146 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1147 char buffer[PROC_NUMBUF];
1148 size_t len;
1149 int make_it_fail;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001150
1151 if (!task)
1152 return -ESRCH;
1153 make_it_fail = task->make_it_fail;
1154 put_task_struct(task);
1155
1156 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001157
1158 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001159}
1160
1161static ssize_t proc_fault_inject_write(struct file * file,
1162 const char __user * buf, size_t count, loff_t *ppos)
1163{
1164 struct task_struct *task;
1165 char buffer[PROC_NUMBUF], *end;
1166 int make_it_fail;
1167
1168 if (!capable(CAP_SYS_RESOURCE))
1169 return -EPERM;
1170 memset(buffer, 0, sizeof(buffer));
1171 if (count > sizeof(buffer) - 1)
1172 count = sizeof(buffer) - 1;
1173 if (copy_from_user(buffer, buf, count))
1174 return -EFAULT;
1175 make_it_fail = simple_strtol(buffer, &end, 0);
1176 if (*end == '\n')
1177 end++;
1178 task = get_proc_task(file->f_dentry->d_inode);
1179 if (!task)
1180 return -ESRCH;
1181 task->make_it_fail = make_it_fail;
1182 put_task_struct(task);
1183 if (end - buffer == 0)
1184 return -EIO;
1185 return end - buffer;
1186}
1187
Arjan van de Ven00977a52007-02-12 00:55:34 -08001188static const struct file_operations proc_fault_inject_operations = {
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001189 .read = proc_fault_inject_read,
1190 .write = proc_fault_inject_write,
1191};
1192#endif
1193
Arjan van de Ven97455122008-01-25 21:08:34 +01001194
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001195#ifdef CONFIG_SCHED_DEBUG
1196/*
1197 * Print out various scheduling related per-task fields:
1198 */
1199static int sched_show(struct seq_file *m, void *v)
1200{
1201 struct inode *inode = m->private;
1202 struct task_struct *p;
1203
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001204 p = get_proc_task(inode);
1205 if (!p)
1206 return -ESRCH;
1207 proc_sched_show_task(p, m);
1208
1209 put_task_struct(p);
1210
1211 return 0;
1212}
1213
1214static ssize_t
1215sched_write(struct file *file, const char __user *buf,
1216 size_t count, loff_t *offset)
1217{
1218 struct inode *inode = file->f_path.dentry->d_inode;
1219 struct task_struct *p;
1220
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001221 p = get_proc_task(inode);
1222 if (!p)
1223 return -ESRCH;
1224 proc_sched_set_task(p);
1225
1226 put_task_struct(p);
1227
1228 return count;
1229}
1230
1231static int sched_open(struct inode *inode, struct file *filp)
1232{
1233 int ret;
1234
1235 ret = single_open(filp, sched_show, NULL);
1236 if (!ret) {
1237 struct seq_file *m = filp->private_data;
1238
1239 m->private = inode;
1240 }
1241 return ret;
1242}
1243
1244static const struct file_operations proc_pid_sched_operations = {
1245 .open = sched_open,
1246 .read = seq_read,
1247 .write = sched_write,
1248 .llseek = seq_lseek,
Alexey Dobriyan5ea473a2007-07-31 00:38:50 -07001249 .release = single_release,
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001250};
1251
1252#endif
1253
Matt Helsley925d1c42008-04-29 01:01:36 -07001254/*
1255 * We added or removed a vma mapping the executable. The vmas are only mapped
1256 * during exec and are not mapped with the mmap system call.
1257 * Callers must hold down_write() on the mm's mmap_sem for these
1258 */
1259void added_exe_file_vma(struct mm_struct *mm)
1260{
1261 mm->num_exe_file_vmas++;
1262}
1263
1264void removed_exe_file_vma(struct mm_struct *mm)
1265{
1266 mm->num_exe_file_vmas--;
1267 if ((mm->num_exe_file_vmas == 0) && mm->exe_file){
1268 fput(mm->exe_file);
1269 mm->exe_file = NULL;
1270 }
1271
1272}
1273
1274void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1275{
1276 if (new_exe_file)
1277 get_file(new_exe_file);
1278 if (mm->exe_file)
1279 fput(mm->exe_file);
1280 mm->exe_file = new_exe_file;
1281 mm->num_exe_file_vmas = 0;
1282}
1283
1284struct file *get_mm_exe_file(struct mm_struct *mm)
1285{
1286 struct file *exe_file;
1287
1288 /* We need mmap_sem to protect against races with removal of
1289 * VM_EXECUTABLE vmas */
1290 down_read(&mm->mmap_sem);
1291 exe_file = mm->exe_file;
1292 if (exe_file)
1293 get_file(exe_file);
1294 up_read(&mm->mmap_sem);
1295 return exe_file;
1296}
1297
1298void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
1299{
1300 /* It's safe to write the exe_file pointer without exe_file_lock because
1301 * this is called during fork when the task is not yet in /proc */
1302 newmm->exe_file = get_mm_exe_file(oldmm);
1303}
1304
1305static int proc_exe_link(struct inode *inode, struct path *exe_path)
1306{
1307 struct task_struct *task;
1308 struct mm_struct *mm;
1309 struct file *exe_file;
1310
1311 task = get_proc_task(inode);
1312 if (!task)
1313 return -ENOENT;
1314 mm = get_task_mm(task);
1315 put_task_struct(task);
1316 if (!mm)
1317 return -ENOENT;
1318 exe_file = get_mm_exe_file(mm);
1319 mmput(mm);
1320 if (exe_file) {
1321 *exe_path = exe_file->f_path;
1322 path_get(&exe_file->f_path);
1323 fput(exe_file);
1324 return 0;
1325 } else
1326 return -ENOENT;
1327}
1328
Al Viro008b1502005-08-20 00:17:39 +01001329static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330{
1331 struct inode *inode = dentry->d_inode;
1332 int error = -EACCES;
1333
1334 /* We don't need a base pointer in the /proc filesystem */
Jan Blunck1d957f92008-02-14 19:34:35 -08001335 path_put(&nd->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336
Eric W. Biederman778c1142006-06-26 00:25:58 -07001337 /* Are we allowed to snoop on the tasks file descriptors? */
1338 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001341 error = PROC_I(inode)->op.proc_get_link(inode, &nd->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 nd->last_type = LAST_BIND;
1343out:
Al Viro008b1502005-08-20 00:17:39 +01001344 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345}
1346
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001347static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348{
Mel Gormane12ba742007-10-16 01:25:52 -07001349 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001350 char *pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351 int len;
1352
1353 if (!tmp)
1354 return -ENOMEM;
Akinobu Mita0c28f282007-05-08 00:31:41 -07001355
Jan Blunckcf28b482008-02-14 19:38:44 -08001356 pathname = d_path(path, tmp, PAGE_SIZE);
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001357 len = PTR_ERR(pathname);
1358 if (IS_ERR(pathname))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359 goto out;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001360 len = tmp + PAGE_SIZE - 1 - pathname;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361
1362 if (len > buflen)
1363 len = buflen;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001364 if (copy_to_user(buffer, pathname, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 len = -EFAULT;
1366 out:
1367 free_page((unsigned long)tmp);
1368 return len;
1369}
1370
1371static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1372{
1373 int error = -EACCES;
1374 struct inode *inode = dentry->d_inode;
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001375 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376
Eric W. Biederman778c1142006-06-26 00:25:58 -07001377 /* Are we allowed to snoop on the tasks file descriptors? */
1378 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001381 error = PROC_I(inode)->op.proc_get_link(inode, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 if (error)
1383 goto out;
1384
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001385 error = do_proc_readlink(&path, buffer, buflen);
1386 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 return error;
1389}
1390
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08001391static const struct inode_operations proc_pid_link_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392 .readlink = proc_pid_readlink,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001393 .follow_link = proc_pid_follow_link,
1394 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395};
1396
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001397
1398/* building an inode */
1399
1400static int task_dumpable(struct task_struct *task)
1401{
1402 int dumpable = 0;
1403 struct mm_struct *mm;
1404
1405 task_lock(task);
1406 mm = task->mm;
1407 if (mm)
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001408 dumpable = get_dumpable(mm);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001409 task_unlock(task);
1410 if(dumpable == 1)
1411 return 1;
1412 return 0;
1413}
1414
1415
Eric W. Biederman61a28782006-10-02 02:18:49 -07001416static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001417{
1418 struct inode * inode;
1419 struct proc_inode *ei;
David Howellsc69e8d92008-11-14 10:39:19 +11001420 const struct cred *cred;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001421
1422 /* We need a new inode */
1423
1424 inode = new_inode(sb);
1425 if (!inode)
1426 goto out;
1427
1428 /* Common stuff */
1429 ei = PROC_I(inode);
1430 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001431 inode->i_op = &proc_def_inode_operations;
1432
1433 /*
1434 * grab the reference to task.
1435 */
Oleg Nesterov1a657f782006-10-02 02:18:59 -07001436 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001437 if (!ei->pid)
1438 goto out_unlock;
1439
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001440 if (task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001441 rcu_read_lock();
1442 cred = __task_cred(task);
1443 inode->i_uid = cred->euid;
1444 inode->i_gid = cred->egid;
1445 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001446 }
1447 security_task_to_inode(task, inode);
1448
1449out:
1450 return inode;
1451
1452out_unlock:
1453 iput(inode);
1454 return NULL;
1455}
1456
1457static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1458{
1459 struct inode *inode = dentry->d_inode;
1460 struct task_struct *task;
David Howellsc69e8d92008-11-14 10:39:19 +11001461 const struct cred *cred;
1462
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001463 generic_fillattr(inode, stat);
1464
1465 rcu_read_lock();
1466 stat->uid = 0;
1467 stat->gid = 0;
1468 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1469 if (task) {
1470 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1471 task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001472 cred = __task_cred(task);
1473 stat->uid = cred->euid;
1474 stat->gid = cred->egid;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001475 }
1476 }
1477 rcu_read_unlock();
1478 return 0;
1479}
1480
1481/* dentry stuff */
1482
1483/*
1484 * Exceptional case: normally we are not allowed to unhash a busy
1485 * directory. In this case, however, we can do it - no aliasing problems
1486 * due to the way we treat inodes.
1487 *
1488 * Rewrite the inode's ownerships here because the owning task may have
1489 * performed a setuid(), etc.
1490 *
1491 * Before the /proc/pid/status file was created the only way to read
1492 * the effective uid of a /process was to stat /proc/pid. Reading
1493 * /proc/pid/status is slow enough that procps and other packages
1494 * kept stating /proc/pid. To keep the rules in /proc simple I have
1495 * made this apply to all per process world readable and executable
1496 * directories.
1497 */
1498static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1499{
1500 struct inode *inode = dentry->d_inode;
1501 struct task_struct *task = get_proc_task(inode);
David Howellsc69e8d92008-11-14 10:39:19 +11001502 const struct cred *cred;
1503
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001504 if (task) {
1505 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1506 task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001507 rcu_read_lock();
1508 cred = __task_cred(task);
1509 inode->i_uid = cred->euid;
1510 inode->i_gid = cred->egid;
1511 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001512 } else {
1513 inode->i_uid = 0;
1514 inode->i_gid = 0;
1515 }
1516 inode->i_mode &= ~(S_ISUID | S_ISGID);
1517 security_task_to_inode(task, inode);
1518 put_task_struct(task);
1519 return 1;
1520 }
1521 d_drop(dentry);
1522 return 0;
1523}
1524
1525static int pid_delete_dentry(struct dentry * dentry)
1526{
1527 /* Is the task we represent dead?
1528 * If so, then don't put the dentry on the lru list,
1529 * kill it immediately.
1530 */
1531 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1532}
1533
Al Virod72f71e2009-02-20 05:58:47 +00001534static const struct dentry_operations pid_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001535{
1536 .d_revalidate = pid_revalidate,
1537 .d_delete = pid_delete_dentry,
1538};
1539
1540/* Lookups */
1541
Eric Dumazetc5141e62007-05-08 00:26:15 -07001542typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1543 struct task_struct *, const void *);
Eric W. Biederman61a28782006-10-02 02:18:49 -07001544
Eric W. Biederman1c0d04c2006-10-02 02:18:57 -07001545/*
1546 * Fill a directory entry.
1547 *
1548 * If possible create the dcache entry and derive our inode number and
1549 * file type from dcache entry.
1550 *
1551 * Since all of the proc inode numbers are dynamically generated, the inode
1552 * numbers do not exist until the inode is cache. This means creating the
1553 * the dcache entry in readdir is necessary to keep the inode numbers
1554 * reported by readdir in sync with the inode numbers reported
1555 * by stat.
1556 */
Eric W. Biederman61a28782006-10-02 02:18:49 -07001557static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1558 char *name, int len,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001559 instantiate_t instantiate, struct task_struct *task, const void *ptr)
Eric W. Biederman61a28782006-10-02 02:18:49 -07001560{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001561 struct dentry *child, *dir = filp->f_path.dentry;
Eric W. Biederman61a28782006-10-02 02:18:49 -07001562 struct inode *inode;
1563 struct qstr qname;
1564 ino_t ino = 0;
1565 unsigned type = DT_UNKNOWN;
1566
1567 qname.name = name;
1568 qname.len = len;
1569 qname.hash = full_name_hash(name, len);
1570
1571 child = d_lookup(dir, &qname);
1572 if (!child) {
1573 struct dentry *new;
1574 new = d_alloc(dir, &qname);
1575 if (new) {
1576 child = instantiate(dir->d_inode, new, task, ptr);
1577 if (child)
1578 dput(new);
1579 else
1580 child = new;
1581 }
1582 }
1583 if (!child || IS_ERR(child) || !child->d_inode)
1584 goto end_instantiate;
1585 inode = child->d_inode;
1586 if (inode) {
1587 ino = inode->i_ino;
1588 type = inode->i_mode >> 12;
1589 }
1590 dput(child);
1591end_instantiate:
1592 if (!ino)
1593 ino = find_inode_number(dir, &qname);
1594 if (!ino)
1595 ino = 1;
1596 return filldir(dirent, name, len, filp->f_pos, ino, type);
1597}
1598
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001599static unsigned name_to_int(struct dentry *dentry)
1600{
1601 const char *name = dentry->d_name.name;
1602 int len = dentry->d_name.len;
1603 unsigned n = 0;
1604
1605 if (len > 1 && *name == '0')
1606 goto out;
1607 while (len-- > 0) {
1608 unsigned c = *name++ - '0';
1609 if (c > 9)
1610 goto out;
1611 if (n >= (~0U-9)/10)
1612 goto out;
1613 n *= 10;
1614 n += c;
1615 }
1616 return n;
1617out:
1618 return ~0U;
1619}
1620
Miklos Szeredi27932742007-05-08 00:26:17 -07001621#define PROC_FDINFO_MAX 64
1622
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001623static int proc_fd_info(struct inode *inode, struct path *path, char *info)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001624{
1625 struct task_struct *task = get_proc_task(inode);
1626 struct files_struct *files = NULL;
1627 struct file *file;
1628 int fd = proc_fd(inode);
1629
1630 if (task) {
1631 files = get_files_struct(task);
1632 put_task_struct(task);
1633 }
1634 if (files) {
1635 /*
1636 * We are not taking a ref to the file structure, so we must
1637 * hold ->file_lock.
1638 */
1639 spin_lock(&files->file_lock);
1640 file = fcheck_files(files, fd);
1641 if (file) {
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001642 if (path) {
1643 *path = file->f_path;
1644 path_get(&file->f_path);
1645 }
Miklos Szeredi27932742007-05-08 00:26:17 -07001646 if (info)
1647 snprintf(info, PROC_FDINFO_MAX,
1648 "pos:\t%lli\n"
1649 "flags:\t0%o\n",
1650 (long long) file->f_pos,
1651 file->f_flags);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001652 spin_unlock(&files->file_lock);
1653 put_files_struct(files);
1654 return 0;
1655 }
1656 spin_unlock(&files->file_lock);
1657 put_files_struct(files);
1658 }
1659 return -ENOENT;
1660}
1661
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001662static int proc_fd_link(struct inode *inode, struct path *path)
Miklos Szeredi27932742007-05-08 00:26:17 -07001663{
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001664 return proc_fd_info(inode, path, NULL);
Miklos Szeredi27932742007-05-08 00:26:17 -07001665}
1666
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001667static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1668{
1669 struct inode *inode = dentry->d_inode;
1670 struct task_struct *task = get_proc_task(inode);
1671 int fd = proc_fd(inode);
1672 struct files_struct *files;
David Howellsc69e8d92008-11-14 10:39:19 +11001673 const struct cred *cred;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001674
1675 if (task) {
1676 files = get_files_struct(task);
1677 if (files) {
1678 rcu_read_lock();
1679 if (fcheck_files(files, fd)) {
1680 rcu_read_unlock();
1681 put_files_struct(files);
1682 if (task_dumpable(task)) {
David Howellsc69e8d92008-11-14 10:39:19 +11001683 rcu_read_lock();
1684 cred = __task_cred(task);
1685 inode->i_uid = cred->euid;
1686 inode->i_gid = cred->egid;
1687 rcu_read_unlock();
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001688 } else {
1689 inode->i_uid = 0;
1690 inode->i_gid = 0;
1691 }
1692 inode->i_mode &= ~(S_ISUID | S_ISGID);
1693 security_task_to_inode(task, inode);
1694 put_task_struct(task);
1695 return 1;
1696 }
1697 rcu_read_unlock();
1698 put_files_struct(files);
1699 }
1700 put_task_struct(task);
1701 }
1702 d_drop(dentry);
1703 return 0;
1704}
1705
Al Virod72f71e2009-02-20 05:58:47 +00001706static const struct dentry_operations tid_fd_dentry_operations =
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001707{
1708 .d_revalidate = tid_fd_revalidate,
1709 .d_delete = pid_delete_dentry,
1710};
1711
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001712static struct dentry *proc_fd_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001713 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001714{
Eric Dumazetc5141e62007-05-08 00:26:15 -07001715 unsigned fd = *(const unsigned *)ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001716 struct file *file;
1717 struct files_struct *files;
1718 struct inode *inode;
1719 struct proc_inode *ei;
1720 struct dentry *error = ERR_PTR(-ENOENT);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001721
Eric W. Biederman61a28782006-10-02 02:18:49 -07001722 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001723 if (!inode)
1724 goto out;
1725 ei = PROC_I(inode);
1726 ei->fd = fd;
1727 files = get_files_struct(task);
1728 if (!files)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001729 goto out_iput;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001730 inode->i_mode = S_IFLNK;
1731
1732 /*
1733 * We are not taking a ref to the file structure, so we must
1734 * hold ->file_lock.
1735 */
1736 spin_lock(&files->file_lock);
1737 file = fcheck_files(files, fd);
1738 if (!file)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001739 goto out_unlock;
Al Viroaeb5d722008-09-02 15:28:45 -04001740 if (file->f_mode & FMODE_READ)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001741 inode->i_mode |= S_IRUSR | S_IXUSR;
Al Viroaeb5d722008-09-02 15:28:45 -04001742 if (file->f_mode & FMODE_WRITE)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001743 inode->i_mode |= S_IWUSR | S_IXUSR;
1744 spin_unlock(&files->file_lock);
1745 put_files_struct(files);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001746
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001747 inode->i_op = &proc_pid_link_inode_operations;
1748 inode->i_size = 64;
1749 ei->op.proc_get_link = proc_fd_link;
1750 dentry->d_op = &tid_fd_dentry_operations;
1751 d_add(dentry, inode);
1752 /* Close the race of the process dying before we return the dentry */
1753 if (tid_fd_revalidate(dentry, NULL))
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001754 error = NULL;
1755
1756 out:
1757 return error;
1758out_unlock:
1759 spin_unlock(&files->file_lock);
1760 put_files_struct(files);
1761out_iput:
1762 iput(inode);
1763 goto out;
1764}
1765
Miklos Szeredi27932742007-05-08 00:26:17 -07001766static struct dentry *proc_lookupfd_common(struct inode *dir,
1767 struct dentry *dentry,
1768 instantiate_t instantiate)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001769{
1770 struct task_struct *task = get_proc_task(dir);
1771 unsigned fd = name_to_int(dentry);
1772 struct dentry *result = ERR_PTR(-ENOENT);
1773
1774 if (!task)
1775 goto out_no_task;
1776 if (fd == ~0U)
1777 goto out;
1778
Miklos Szeredi27932742007-05-08 00:26:17 -07001779 result = instantiate(dir, dentry, task, &fd);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001780out:
1781 put_task_struct(task);
1782out_no_task:
1783 return result;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001784}
1785
Miklos Szeredi27932742007-05-08 00:26:17 -07001786static int proc_readfd_common(struct file * filp, void * dirent,
1787 filldir_t filldir, instantiate_t instantiate)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001789 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman56347082006-06-26 00:25:40 -07001790 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001791 struct task_struct *p = get_proc_task(inode);
Pavel Emelyanov457c2512007-10-18 23:40:43 -07001792 unsigned int fd, ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794 struct files_struct * files;
1795
1796 retval = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001797 if (!p)
1798 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800
1801 fd = filp->f_pos;
1802 switch (fd) {
1803 case 0:
1804 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1805 goto out;
1806 filp->f_pos++;
1807 case 1:
Eric W. Biederman56347082006-06-26 00:25:40 -07001808 ino = parent_ino(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1810 goto out;
1811 filp->f_pos++;
1812 default:
1813 files = get_files_struct(p);
1814 if (!files)
1815 goto out;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001816 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817 for (fd = filp->f_pos-2;
Al Viro9b4f5262008-04-22 01:32:44 -04001818 fd < files_fdtable(files)->max_fds;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819 fd++, filp->f_pos++) {
Miklos Szeredi27932742007-05-08 00:26:17 -07001820 char name[PROC_NUMBUF];
1821 int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822
1823 if (!fcheck_files(files, fd))
1824 continue;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001825 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826
Miklos Szeredi27932742007-05-08 00:26:17 -07001827 len = snprintf(name, sizeof(name), "%d", fd);
1828 if (proc_fill_cache(filp, dirent, filldir,
1829 name, len, instantiate,
1830 p, &fd) < 0) {
Dipankar Sarmab8359962005-09-09 13:04:14 -07001831 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001832 break;
1833 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07001834 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001835 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07001836 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837 put_files_struct(files);
1838 }
1839out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07001840 put_task_struct(p);
1841out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842 return retval;
1843}
1844
Miklos Szeredi27932742007-05-08 00:26:17 -07001845static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1846 struct nameidata *nd)
1847{
1848 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1849}
1850
1851static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1852{
1853 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1854}
1855
1856static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1857 size_t len, loff_t *ppos)
1858{
1859 char tmp[PROC_FDINFO_MAX];
Jan Blunck3dcd25f2008-02-14 19:38:35 -08001860 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
Miklos Szeredi27932742007-05-08 00:26:17 -07001861 if (!err)
1862 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1863 return err;
1864}
1865
1866static const struct file_operations proc_fdinfo_file_operations = {
1867 .open = nonseekable_open,
1868 .read = proc_fdinfo_read,
1869};
1870
Arjan van de Ven00977a52007-02-12 00:55:34 -08001871static const struct file_operations proc_fd_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872 .read = generic_read_dir,
1873 .readdir = proc_readfd,
1874};
1875
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876/*
Alexey Dobriyan8948e112007-05-08 00:23:35 -07001877 * /proc/pid/fd needs a special permission handler so that a process can still
1878 * access /proc/self/fd after it has executed a setuid().
1879 */
Al Viroe6305c42008-07-15 21:03:57 -04001880static int proc_fd_permission(struct inode *inode, int mask)
Alexey Dobriyan8948e112007-05-08 00:23:35 -07001881{
1882 int rv;
1883
1884 rv = generic_permission(inode, mask, NULL);
1885 if (rv == 0)
1886 return 0;
1887 if (task_pid(current) == proc_pid(inode))
1888 rv = 0;
1889 return rv;
1890}
1891
1892/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893 * proc directories can do almost nothing..
1894 */
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08001895static const struct inode_operations proc_fd_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896 .lookup = proc_lookupfd,
Alexey Dobriyan8948e112007-05-08 00:23:35 -07001897 .permission = proc_fd_permission,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001898 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899};
1900
Miklos Szeredi27932742007-05-08 00:26:17 -07001901static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1902 struct dentry *dentry, struct task_struct *task, const void *ptr)
1903{
1904 unsigned fd = *(unsigned *)ptr;
1905 struct inode *inode;
1906 struct proc_inode *ei;
1907 struct dentry *error = ERR_PTR(-ENOENT);
1908
1909 inode = proc_pid_make_inode(dir->i_sb, task);
1910 if (!inode)
1911 goto out;
1912 ei = PROC_I(inode);
1913 ei->fd = fd;
1914 inode->i_mode = S_IFREG | S_IRUSR;
1915 inode->i_fop = &proc_fdinfo_file_operations;
1916 dentry->d_op = &tid_fd_dentry_operations;
1917 d_add(dentry, inode);
1918 /* Close the race of the process dying before we return the dentry */
1919 if (tid_fd_revalidate(dentry, NULL))
1920 error = NULL;
1921
1922 out:
1923 return error;
1924}
1925
1926static struct dentry *proc_lookupfdinfo(struct inode *dir,
1927 struct dentry *dentry,
1928 struct nameidata *nd)
1929{
1930 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
1931}
1932
1933static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
1934{
1935 return proc_readfd_common(filp, dirent, filldir,
1936 proc_fdinfo_instantiate);
1937}
1938
1939static const struct file_operations proc_fdinfo_operations = {
1940 .read = generic_read_dir,
1941 .readdir = proc_readfdinfo,
1942};
1943
1944/*
1945 * proc directories can do almost nothing..
1946 */
1947static const struct inode_operations proc_fdinfo_inode_operations = {
1948 .lookup = proc_lookupfdinfo,
1949 .setattr = proc_setattr,
1950};
1951
1952
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001953static struct dentry *proc_pident_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001954 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001955{
Eric Dumazetc5141e62007-05-08 00:26:15 -07001956 const struct pid_entry *p = ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001957 struct inode *inode;
1958 struct proc_inode *ei;
1959 struct dentry *error = ERR_PTR(-EINVAL);
1960
Eric W. Biederman61a28782006-10-02 02:18:49 -07001961 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001962 if (!inode)
1963 goto out;
1964
1965 ei = PROC_I(inode);
1966 inode->i_mode = p->mode;
1967 if (S_ISDIR(inode->i_mode))
1968 inode->i_nlink = 2; /* Use getattr to fix if necessary */
1969 if (p->iop)
1970 inode->i_op = p->iop;
1971 if (p->fop)
1972 inode->i_fop = p->fop;
1973 ei->op = p->op;
1974 dentry->d_op = &pid_dentry_operations;
1975 d_add(dentry, inode);
1976 /* Close the race of the process dying before we return the dentry */
1977 if (pid_revalidate(dentry, NULL))
1978 error = NULL;
1979out:
1980 return error;
1981}
1982
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983static struct dentry *proc_pident_lookup(struct inode *dir,
1984 struct dentry *dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001985 const struct pid_entry *ents,
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07001986 unsigned int nents)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987{
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001988 struct dentry *error;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001989 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07001990 const struct pid_entry *p, *last;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001992 error = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993
Eric W. Biederman99f89552006-06-26 00:25:55 -07001994 if (!task)
1995 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996
Eric W. Biederman20cdc892006-10-02 02:17:07 -07001997 /*
1998 * Yes, it does not scale. And it should not. Don't add
1999 * new entries into /proc/<tgid>/ without very good reasons.
2000 */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002001 last = &ents[nents - 1];
2002 for (p = ents; p <= last; p++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003 if (p->len != dentry->d_name.len)
2004 continue;
2005 if (!memcmp(dentry->d_name.name, p->name, p->len))
2006 break;
2007 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002008 if (p > last)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 goto out;
2010
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002011 error = proc_pident_instantiate(dir, dentry, task, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002012out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07002013 put_task_struct(task);
2014out_no_task:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002015 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016}
2017
Eric Dumazetc5141e62007-05-08 00:26:15 -07002018static int proc_pident_fill_cache(struct file *filp, void *dirent,
2019 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002020{
2021 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2022 proc_pident_instantiate, task, p);
2023}
2024
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002025static int proc_pident_readdir(struct file *filp,
2026 void *dirent, filldir_t filldir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002027 const struct pid_entry *ents, unsigned int nents)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002028{
2029 int i;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002030 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002031 struct inode *inode = dentry->d_inode;
2032 struct task_struct *task = get_proc_task(inode);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002033 const struct pid_entry *p, *last;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002034 ino_t ino;
2035 int ret;
2036
2037 ret = -ENOENT;
2038 if (!task)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002039 goto out_no_task;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002040
2041 ret = 0;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002042 i = filp->f_pos;
2043 switch (i) {
2044 case 0:
2045 ino = inode->i_ino;
2046 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2047 goto out;
2048 i++;
2049 filp->f_pos++;
2050 /* fall through */
2051 case 1:
2052 ino = parent_ino(dentry);
2053 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2054 goto out;
2055 i++;
2056 filp->f_pos++;
2057 /* fall through */
2058 default:
2059 i -= 2;
2060 if (i >= nents) {
2061 ret = 1;
2062 goto out;
2063 }
2064 p = ents + i;
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002065 last = &ents[nents - 1];
2066 while (p <= last) {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002067 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002068 goto out;
2069 filp->f_pos++;
2070 p++;
2071 }
2072 }
2073
2074 ret = 1;
2075out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07002076 put_task_struct(task);
2077out_no_task:
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002078 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079}
2080
Linus Torvalds1da177e2005-04-16 15:20:36 -07002081#ifdef CONFIG_SECURITY
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002082static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2083 size_t count, loff_t *ppos)
2084{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002085 struct inode * inode = file->f_path.dentry->d_inode;
Al Viro04ff9702007-03-12 16:17:58 +00002086 char *p = NULL;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002087 ssize_t length;
2088 struct task_struct *task = get_proc_task(inode);
2089
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002090 if (!task)
Al Viro04ff9702007-03-12 16:17:58 +00002091 return -ESRCH;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002092
2093 length = security_getprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002094 (char*)file->f_path.dentry->d_name.name,
Al Viro04ff9702007-03-12 16:17:58 +00002095 &p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002096 put_task_struct(task);
Al Viro04ff9702007-03-12 16:17:58 +00002097 if (length > 0)
2098 length = simple_read_from_buffer(buf, count, ppos, p, length);
2099 kfree(p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002100 return length;
2101}
2102
2103static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2104 size_t count, loff_t *ppos)
2105{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002106 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002107 char *page;
2108 ssize_t length;
2109 struct task_struct *task = get_proc_task(inode);
2110
2111 length = -ESRCH;
2112 if (!task)
2113 goto out_no_task;
2114 if (count > PAGE_SIZE)
2115 count = PAGE_SIZE;
2116
2117 /* No partial writes. */
2118 length = -EINVAL;
2119 if (*ppos != 0)
2120 goto out;
2121
2122 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -07002123 page = (char*)__get_free_page(GFP_TEMPORARY);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002124 if (!page)
2125 goto out;
2126
2127 length = -EFAULT;
2128 if (copy_from_user(page, buf, count))
2129 goto out_free;
2130
2131 length = security_setprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002132 (char*)file->f_path.dentry->d_name.name,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002133 (void*)page, count);
2134out_free:
2135 free_page((unsigned long) page);
2136out:
2137 put_task_struct(task);
2138out_no_task:
2139 return length;
2140}
2141
Arjan van de Ven00977a52007-02-12 00:55:34 -08002142static const struct file_operations proc_pid_attr_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002143 .read = proc_pid_attr_read,
2144 .write = proc_pid_attr_write,
2145};
2146
Eric Dumazetc5141e62007-05-08 00:26:15 -07002147static const struct pid_entry attr_dir_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002148 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2149 REG("prev", S_IRUGO, proc_pid_attr_operations),
2150 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2151 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2152 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2153 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002154};
2155
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002156static int proc_attr_dir_readdir(struct file * filp,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157 void * dirent, filldir_t filldir)
2158{
2159 return proc_pident_readdir(filp,dirent,filldir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002160 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002161}
2162
Arjan van de Ven00977a52007-02-12 00:55:34 -08002163static const struct file_operations proc_attr_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002164 .read = generic_read_dir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002165 .readdir = proc_attr_dir_readdir,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002166};
2167
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002168static struct dentry *proc_attr_dir_lookup(struct inode *dir,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002169 struct dentry *dentry, struct nameidata *nd)
2170{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002171 return proc_pident_lookup(dir, dentry,
2172 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002173}
2174
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002175static const struct inode_operations proc_attr_dir_inode_operations = {
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002176 .lookup = proc_attr_dir_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07002177 .getattr = pid_getattr,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07002178 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002179};
2180
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181#endif
2182
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002183#if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2184static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2185 size_t count, loff_t *ppos)
2186{
2187 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2188 struct mm_struct *mm;
2189 char buffer[PROC_NUMBUF];
2190 size_t len;
2191 int ret;
2192
2193 if (!task)
2194 return -ESRCH;
2195
2196 ret = 0;
2197 mm = get_task_mm(task);
2198 if (mm) {
2199 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2200 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2201 MMF_DUMP_FILTER_SHIFT));
2202 mmput(mm);
2203 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2204 }
2205
2206 put_task_struct(task);
2207
2208 return ret;
2209}
2210
2211static ssize_t proc_coredump_filter_write(struct file *file,
2212 const char __user *buf,
2213 size_t count,
2214 loff_t *ppos)
2215{
2216 struct task_struct *task;
2217 struct mm_struct *mm;
2218 char buffer[PROC_NUMBUF], *end;
2219 unsigned int val;
2220 int ret;
2221 int i;
2222 unsigned long mask;
2223
2224 ret = -EFAULT;
2225 memset(buffer, 0, sizeof(buffer));
2226 if (count > sizeof(buffer) - 1)
2227 count = sizeof(buffer) - 1;
2228 if (copy_from_user(buffer, buf, count))
2229 goto out_no_task;
2230
2231 ret = -EINVAL;
2232 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2233 if (*end == '\n')
2234 end++;
2235 if (end - buffer == 0)
2236 goto out_no_task;
2237
2238 ret = -ESRCH;
2239 task = get_proc_task(file->f_dentry->d_inode);
2240 if (!task)
2241 goto out_no_task;
2242
2243 ret = end - buffer;
2244 mm = get_task_mm(task);
2245 if (!mm)
2246 goto out_no_mm;
2247
2248 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2249 if (val & mask)
2250 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2251 else
2252 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2253 }
2254
2255 mmput(mm);
2256 out_no_mm:
2257 put_task_struct(task);
2258 out_no_task:
2259 return ret;
2260}
2261
2262static const struct file_operations proc_coredump_filter_operations = {
2263 .read = proc_coredump_filter_read,
2264 .write = proc_coredump_filter_write,
2265};
2266#endif
2267
Linus Torvalds1da177e2005-04-16 15:20:36 -07002268/*
2269 * /proc/self:
2270 */
2271static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2272 int buflen)
2273{
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002274 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002275 pid_t tgid = task_tgid_nr_ns(current, ns);
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002276 char tmp[PROC_NUMBUF];
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002277 if (!tgid)
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002278 return -ENOENT;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002279 sprintf(tmp, "%d", tgid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002280 return vfs_readlink(dentry,buffer,buflen,tmp);
2281}
2282
Al Viro008b1502005-08-20 00:17:39 +01002283static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284{
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002285 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002286 pid_t tgid = task_tgid_nr_ns(current, ns);
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002287 char tmp[PROC_NUMBUF];
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002288 if (!tgid)
Eric W. Biederman488e5bc2008-02-08 04:18:34 -08002289 return ERR_PTR(-ENOENT);
Andrew Mortonb55fcb22008-02-08 15:00:43 -08002290 sprintf(tmp, "%d", task_tgid_nr_ns(current, ns));
Al Viro008b1502005-08-20 00:17:39 +01002291 return ERR_PTR(vfs_follow_link(nd,tmp));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002292}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002293
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002294static const struct inode_operations proc_self_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002295 .readlink = proc_self_readlink,
2296 .follow_link = proc_self_follow_link,
2297};
2298
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002299/*
Eric W. Biederman801199c2006-10-02 02:18:48 -07002300 * proc base
2301 *
2302 * These are the directory entries in the root directory of /proc
2303 * that properly belong to the /proc filesystem, as they describe
2304 * describe something that is process related.
2305 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07002306static const struct pid_entry proc_base_stuff[] = {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002307 NOD("self", S_IFLNK|S_IRWXUGO,
Eric W. Biederman801199c2006-10-02 02:18:48 -07002308 &proc_self_inode_operations, NULL, {}),
Eric W. Biederman801199c2006-10-02 02:18:48 -07002309};
2310
2311/*
2312 * Exceptional case: normally we are not allowed to unhash a busy
2313 * directory. In this case, however, we can do it - no aliasing problems
2314 * due to the way we treat inodes.
2315 */
2316static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2317{
2318 struct inode *inode = dentry->d_inode;
2319 struct task_struct *task = get_proc_task(inode);
2320 if (task) {
2321 put_task_struct(task);
2322 return 1;
2323 }
2324 d_drop(dentry);
2325 return 0;
2326}
2327
Al Virod72f71e2009-02-20 05:58:47 +00002328static const struct dentry_operations proc_base_dentry_operations =
Eric W. Biederman801199c2006-10-02 02:18:48 -07002329{
2330 .d_revalidate = proc_base_revalidate,
2331 .d_delete = pid_delete_dentry,
2332};
2333
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002334static struct dentry *proc_base_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002335 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman801199c2006-10-02 02:18:48 -07002336{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002337 const struct pid_entry *p = ptr;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002338 struct inode *inode;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002339 struct proc_inode *ei;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002340 struct dentry *error = ERR_PTR(-EINVAL);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002341
2342 /* Allocate the inode */
2343 error = ERR_PTR(-ENOMEM);
2344 inode = new_inode(dir->i_sb);
2345 if (!inode)
2346 goto out;
2347
2348 /* Initialize the inode */
2349 ei = PROC_I(inode);
2350 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002351
2352 /*
2353 * grab the reference to the task.
2354 */
Oleg Nesterov1a657f782006-10-02 02:18:59 -07002355 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002356 if (!ei->pid)
2357 goto out_iput;
2358
Eric W. Biederman801199c2006-10-02 02:18:48 -07002359 inode->i_mode = p->mode;
2360 if (S_ISDIR(inode->i_mode))
2361 inode->i_nlink = 2;
2362 if (S_ISLNK(inode->i_mode))
2363 inode->i_size = 64;
2364 if (p->iop)
2365 inode->i_op = p->iop;
2366 if (p->fop)
2367 inode->i_fop = p->fop;
2368 ei->op = p->op;
2369 dentry->d_op = &proc_base_dentry_operations;
2370 d_add(dentry, inode);
2371 error = NULL;
2372out:
Eric W. Biederman801199c2006-10-02 02:18:48 -07002373 return error;
2374out_iput:
2375 iput(inode);
2376 goto out;
2377}
2378
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002379static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2380{
2381 struct dentry *error;
2382 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002383 const struct pid_entry *p, *last;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002384
2385 error = ERR_PTR(-ENOENT);
2386
2387 if (!task)
2388 goto out_no_task;
2389
2390 /* Lookup the directory entry */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002391 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2392 for (p = proc_base_stuff; p <= last; p++) {
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002393 if (p->len != dentry->d_name.len)
2394 continue;
2395 if (!memcmp(dentry->d_name.name, p->name, p->len))
2396 break;
2397 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002398 if (p > last)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002399 goto out;
2400
2401 error = proc_base_instantiate(dir, dentry, task, p);
2402
2403out:
2404 put_task_struct(task);
2405out_no_task:
2406 return error;
2407}
2408
Eric Dumazetc5141e62007-05-08 00:26:15 -07002409static int proc_base_fill_cache(struct file *filp, void *dirent,
2410 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002411{
2412 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2413 proc_base_instantiate, task, p);
2414}
2415
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002416#ifdef CONFIG_TASK_IO_ACCOUNTING
Andrea Righi297c5d92008-07-25 01:48:49 -07002417static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002418{
Andrea Righi940389b2008-07-28 00:48:12 +02002419 struct task_io_accounting acct = task->ioac;
Andrea Righi59954772008-07-27 17:29:15 +02002420 unsigned long flags;
Andrea Righi297c5d92008-07-25 01:48:49 -07002421
Andrea Righi59954772008-07-27 17:29:15 +02002422 if (whole && lock_task_sighand(task, &flags)) {
2423 struct task_struct *t = task;
Andrea Righib2d002d2008-07-26 15:22:27 -07002424
Andrea Righi59954772008-07-27 17:29:15 +02002425 task_io_accounting_add(&acct, &task->signal->ioac);
2426 while_each_thread(task, t)
2427 task_io_accounting_add(&acct, &t->ioac);
Andrea Righi297c5d92008-07-25 01:48:49 -07002428
Andrea Righi59954772008-07-27 17:29:15 +02002429 unlock_task_sighand(task, &flags);
Andrea Righi297c5d92008-07-25 01:48:49 -07002430 }
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002431 return sprintf(buffer,
2432 "rchar: %llu\n"
2433 "wchar: %llu\n"
2434 "syscr: %llu\n"
2435 "syscw: %llu\n"
2436 "read_bytes: %llu\n"
2437 "write_bytes: %llu\n"
2438 "cancelled_write_bytes: %llu\n",
Alexander Beregalov7c443192008-08-05 13:01:34 -07002439 (unsigned long long)acct.rchar,
2440 (unsigned long long)acct.wchar,
2441 (unsigned long long)acct.syscr,
2442 (unsigned long long)acct.syscw,
2443 (unsigned long long)acct.read_bytes,
2444 (unsigned long long)acct.write_bytes,
2445 (unsigned long long)acct.cancelled_write_bytes);
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002446}
Andrea Righi297c5d92008-07-25 01:48:49 -07002447
2448static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2449{
2450 return do_io_accounting(task, buffer, 0);
2451}
2452
2453static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2454{
2455 return do_io_accounting(task, buffer, 1);
2456}
2457#endif /* CONFIG_TASK_IO_ACCOUNTING */
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002458
Kees Cook47830722008-10-06 03:11:58 +04002459static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
2460 struct pid *pid, struct task_struct *task)
2461{
2462 seq_printf(m, "%08x\n", task->personality);
2463 return 0;
2464}
2465
Eric W. Biederman801199c2006-10-02 02:18:48 -07002466/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002467 * Thread groups
2468 */
Arjan van de Ven00977a52007-02-12 00:55:34 -08002469static const struct file_operations proc_task_operations;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002470static const struct inode_operations proc_task_inode_operations;
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002471
Eric Dumazetc5141e62007-05-08 00:26:15 -07002472static const struct pid_entry tgid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002473 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
2474 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2475 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002476#ifdef CONFIG_NET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002477 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
Andrew Mortonb2211a32008-03-11 18:03:35 -07002478#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002479 REG("environ", S_IRUSR, proc_environ_operations),
2480 INF("auxv", S_IRUSR, proc_pid_auxv),
2481 ONE("status", S_IRUGO, proc_pid_status),
2482 ONE("personality", S_IRUSR, proc_pid_personality),
2483 INF("limits", S_IRUSR, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002484#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002485 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002486#endif
Roland McGrathebcb6732008-07-25 19:46:00 -07002487#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002488 INF("syscall", S_IRUSR, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07002489#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002490 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2491 ONE("stat", S_IRUGO, proc_tgid_stat),
2492 ONE("statm", S_IRUGO, proc_pid_statm),
2493 REG("maps", S_IRUGO, proc_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002494#ifdef CONFIG_NUMA
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002495 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002496#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002497 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2498 LNK("cwd", proc_cwd_link),
2499 LNK("root", proc_root_link),
2500 LNK("exe", proc_exe_link),
2501 REG("mounts", S_IRUGO, proc_mounts_operations),
2502 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
2503 REG("mountstats", S_IRUSR, proc_mountstats_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08002504#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002505 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2506 REG("smaps", S_IRUGO, proc_smaps_operations),
2507 REG("pagemap", S_IRUSR, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002508#endif
2509#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002510 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002511#endif
2512#ifdef CONFIG_KALLSYMS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002513 INF("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002514#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03002515#ifdef CONFIG_STACKTRACE
2516 ONE("stack", S_IRUSR, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002517#endif
2518#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002519 INF("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002520#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01002521#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002522 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01002523#endif
Paul Menage8793d852007-10-18 23:39:39 -07002524#ifdef CONFIG_PROC_PID_CPUSET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002525 REG("cpuset", S_IRUGO, proc_cpuset_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002526#endif
Paul Menagea4243162007-10-18 23:39:35 -07002527#ifdef CONFIG_CGROUPS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002528 REG("cgroup", S_IRUGO, proc_cgroup_operations),
Paul Menagea4243162007-10-18 23:39:35 -07002529#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002530 INF("oom_score", S_IRUGO, proc_oom_score),
2531 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002532#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002533 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2534 REG("sessionid", S_IRUGO, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002535#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002536#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002537 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002538#endif
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002539#if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002540 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002541#endif
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002542#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002543 INF("io", S_IRUGO, proc_tgid_io_accounting),
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002544#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002545};
2546
2547static int proc_tgid_base_readdir(struct file * filp,
2548 void * dirent, filldir_t filldir)
2549{
2550 return proc_pident_readdir(filp,dirent,filldir,
2551 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2552}
2553
Arjan van de Ven00977a52007-02-12 00:55:34 -08002554static const struct file_operations proc_tgid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002555 .read = generic_read_dir,
2556 .readdir = proc_tgid_base_readdir,
2557};
2558
2559static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002560 return proc_pident_lookup(dir, dentry,
2561 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002562}
2563
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002564static const struct inode_operations proc_tgid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002565 .lookup = proc_tgid_base_lookup,
2566 .getattr = pid_getattr,
2567 .setattr = proc_setattr,
2568};
2569
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002570static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002571{
Eric W. Biederman48e64842006-06-26 00:25:48 -07002572 struct dentry *dentry, *leader, *dir;
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002573 char buf[PROC_NUMBUF];
Eric W. Biederman48e64842006-06-26 00:25:48 -07002574 struct qstr name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002575
Eric W. Biederman48e64842006-06-26 00:25:48 -07002576 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002577 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2578 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002579 if (dentry) {
Andrea Arcangeli77667552008-02-04 22:29:21 -08002580 if (!(current->flags & PF_EXITING))
2581 shrink_dcache_parent(dentry);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002582 d_drop(dentry);
2583 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002584 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002585
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002586 if (tgid == 0)
Eric W. Biederman48e64842006-06-26 00:25:48 -07002587 goto out;
2588
2589 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002590 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2591 leader = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002592 if (!leader)
2593 goto out;
2594
2595 name.name = "task";
2596 name.len = strlen(name.name);
2597 dir = d_hash_and_lookup(leader, &name);
2598 if (!dir)
2599 goto out_put_leader;
2600
2601 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002602 name.len = snprintf(buf, sizeof(buf), "%d", pid);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002603 dentry = d_hash_and_lookup(dir, &name);
2604 if (dentry) {
2605 shrink_dcache_parent(dentry);
2606 d_drop(dentry);
2607 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002608 }
Eric W. Biederman48e64842006-06-26 00:25:48 -07002609
2610 dput(dir);
2611out_put_leader:
2612 dput(leader);
2613out:
2614 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002615}
2616
Randy Dunlap0895e912007-10-21 21:00:10 -07002617/**
2618 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2619 * @task: task that should be flushed.
2620 *
2621 * When flushing dentries from proc, one needs to flush them from global
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002622 * proc (proc_mnt) and from all the namespaces' procs this task was seen
Randy Dunlap0895e912007-10-21 21:00:10 -07002623 * in. This call is supposed to do all of this job.
2624 *
2625 * Looks in the dcache for
2626 * /proc/@pid
2627 * /proc/@tgid/task/@pid
2628 * if either directory is present flushes it and all of it'ts children
2629 * from the dcache.
2630 *
2631 * It is safe and reasonable to cache /proc entries for a task until
2632 * that task exits. After that they just clog up the dcache with
2633 * useless entries, possibly causing useful dcache entries to be
2634 * flushed instead. This routine is proved to flush those useless
2635 * dcache entries at process exit time.
2636 *
2637 * NOTE: This routine is just an optimization so it does not guarantee
2638 * that no dcache entries will exist at process exit time it
2639 * just makes it very unlikely that any will persist.
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002640 */
2641
2642void proc_flush_task(struct task_struct *task)
2643{
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002644 int i;
2645 struct pid *pid, *tgid = NULL;
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002646 struct upid *upid;
2647
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002648 pid = task_pid(task);
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002649 if (thread_group_leader(task))
2650 tgid = task_tgid(task);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002651
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002652 for (i = 0; i <= pid->level; i++) {
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002653 upid = &pid->numbers[i];
2654 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002655 tgid ? tgid->numbers[i].nr : 0);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002656 }
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -07002657
2658 upid = &pid->numbers[pid->level];
2659 if (upid->nr == 1)
2660 pid_ns_release_proc(upid->ns);
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002661}
2662
Adrian Bunk9711ef92006-12-06 20:38:31 -08002663static struct dentry *proc_pid_instantiate(struct inode *dir,
2664 struct dentry * dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002665 struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002666{
2667 struct dentry *error = ERR_PTR(-ENOENT);
2668 struct inode *inode;
2669
Eric W. Biederman61a28782006-10-02 02:18:49 -07002670 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002671 if (!inode)
2672 goto out;
2673
2674 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2675 inode->i_op = &proc_tgid_base_inode_operations;
2676 inode->i_fop = &proc_tgid_base_operations;
2677 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07002678
2679 inode->i_nlink = 2 + pid_entry_count_dirs(tgid_base_stuff,
2680 ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002681
2682 dentry->d_op = &pid_dentry_operations;
2683
2684 d_add(dentry, inode);
2685 /* Close the race of the process dying before we return the dentry */
2686 if (pid_revalidate(dentry, NULL))
2687 error = NULL;
2688out:
2689 return error;
2690}
2691
Linus Torvalds1da177e2005-04-16 15:20:36 -07002692struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2693{
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002694 struct dentry *result = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002695 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002696 unsigned tgid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002697 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002698
Eric W. Biederman801199c2006-10-02 02:18:48 -07002699 result = proc_base_lookup(dir, dentry);
2700 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2701 goto out;
2702
Linus Torvalds1da177e2005-04-16 15:20:36 -07002703 tgid = name_to_int(dentry);
2704 if (tgid == ~0U)
2705 goto out;
2706
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002707 ns = dentry->d_sb->s_fs_info;
Eric W. Biedermande758732006-06-26 00:25:51 -07002708 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002709 task = find_task_by_pid_ns(tgid, ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002710 if (task)
2711 get_task_struct(task);
Eric W. Biedermande758732006-06-26 00:25:51 -07002712 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002713 if (!task)
2714 goto out;
2715
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002716 result = proc_pid_instantiate(dir, dentry, task, NULL);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002717 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002718out:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002719 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002720}
2721
Linus Torvalds1da177e2005-04-16 15:20:36 -07002722/*
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002723 * Find the first task with tgid >= tgid
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002724 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002725 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002726struct tgid_iter {
2727 unsigned int tgid;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002728 struct task_struct *task;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002729};
2730static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2731{
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002732 struct pid *pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002733
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002734 if (iter.task)
2735 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002736 rcu_read_lock();
2737retry:
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002738 iter.task = NULL;
2739 pid = find_ge_pid(iter.tgid, ns);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002740 if (pid) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002741 iter.tgid = pid_nr_ns(pid, ns);
2742 iter.task = pid_task(pid, PIDTYPE_PID);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002743 /* What we to know is if the pid we have find is the
2744 * pid of a thread_group_leader. Testing for task
2745 * being a thread_group_leader is the obvious thing
2746 * todo but there is a window when it fails, due to
2747 * the pid transfer logic in de_thread.
2748 *
2749 * So we perform the straight forward test of seeing
2750 * if the pid we have found is the pid of a thread
2751 * group leader, and don't worry if the task we have
2752 * found doesn't happen to be a thread group leader.
2753 * As we don't care in the case of readdir.
2754 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002755 if (!iter.task || !has_group_leader_pid(iter.task)) {
2756 iter.tgid += 1;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002757 goto retry;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002758 }
2759 get_task_struct(iter.task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002760 }
Eric W. Biederman454cc102006-06-26 00:25:51 -07002761 rcu_read_unlock();
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002762 return iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002763}
2764
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002765#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002766
Eric W. Biederman61a28782006-10-02 02:18:49 -07002767static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002768 struct tgid_iter iter)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002769{
2770 char name[PROC_NUMBUF];
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002771 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
Eric W. Biederman61a28782006-10-02 02:18:49 -07002772 return proc_fill_cache(filp, dirent, filldir, name, len,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002773 proc_pid_instantiate, iter.task, NULL);
Eric W. Biederman61a28782006-10-02 02:18:49 -07002774}
2775
Linus Torvalds1da177e2005-04-16 15:20:36 -07002776/* for the /proc/ directory itself, after non-process stuff has been done */
2777int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2778{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002779 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002780 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002781 struct tgid_iter iter;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002782 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002783
Eric W. Biederman61a28782006-10-02 02:18:49 -07002784 if (!reaper)
2785 goto out_no_task;
2786
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002787 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
Eric Dumazetc5141e62007-05-08 00:26:15 -07002788 const struct pid_entry *p = &proc_base_stuff[nr];
Eric W. Biederman61a28782006-10-02 02:18:49 -07002789 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
Eric W. Biederman801199c2006-10-02 02:18:48 -07002790 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002791 }
2792
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002793 ns = filp->f_dentry->d_sb->s_fs_info;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002794 iter.task = NULL;
2795 iter.tgid = filp->f_pos - TGID_OFFSET;
2796 for (iter = next_tgid(ns, iter);
2797 iter.task;
2798 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2799 filp->f_pos = iter.tgid + TGID_OFFSET;
2800 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
2801 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002802 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002803 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002804 }
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002805 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2806out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07002807 put_task_struct(reaper);
2808out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002809 return 0;
2810}
2811
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002812/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002813 * Tasks
2814 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07002815static const struct pid_entry tid_base_stuff[] = {
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002816 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
2817 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fd_operations),
2818 REG("environ", S_IRUSR, proc_environ_operations),
2819 INF("auxv", S_IRUSR, proc_pid_auxv),
2820 ONE("status", S_IRUGO, proc_pid_status),
2821 ONE("personality", S_IRUSR, proc_pid_personality),
2822 INF("limits", S_IRUSR, proc_pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002823#ifdef CONFIG_SCHED_DEBUG
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002824 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002825#endif
Roland McGrathebcb6732008-07-25 19:46:00 -07002826#ifdef CONFIG_HAVE_ARCH_TRACEHOOK
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002827 INF("syscall", S_IRUSR, proc_pid_syscall),
Roland McGrathebcb6732008-07-25 19:46:00 -07002828#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002829 INF("cmdline", S_IRUGO, proc_pid_cmdline),
2830 ONE("stat", S_IRUGO, proc_tid_stat),
2831 ONE("statm", S_IRUGO, proc_pid_statm),
2832 REG("maps", S_IRUGO, proc_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002833#ifdef CONFIG_NUMA
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002834 REG("numa_maps", S_IRUGO, proc_numa_maps_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002835#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002836 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
2837 LNK("cwd", proc_cwd_link),
2838 LNK("root", proc_root_link),
2839 LNK("exe", proc_exe_link),
2840 REG("mounts", S_IRUGO, proc_mounts_operations),
2841 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
Matt Mackall1e883282008-02-04 22:29:07 -08002842#ifdef CONFIG_PROC_PAGE_MONITOR
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002843 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
2844 REG("smaps", S_IRUGO, proc_smaps_operations),
2845 REG("pagemap", S_IRUSR, proc_pagemap_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002846#endif
2847#ifdef CONFIG_SECURITY
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002848 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002849#endif
2850#ifdef CONFIG_KALLSYMS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002851 INF("wchan", S_IRUGO, proc_pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002852#endif
Ken Chen2ec220e2008-11-10 11:26:08 +03002853#ifdef CONFIG_STACKTRACE
2854 ONE("stack", S_IRUSR, proc_pid_stack),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002855#endif
2856#ifdef CONFIG_SCHEDSTATS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002857 INF("schedstat", S_IRUGO, proc_pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002858#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01002859#ifdef CONFIG_LATENCYTOP
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002860 REG("latency", S_IRUGO, proc_lstats_operations),
Arjan van de Ven97455122008-01-25 21:08:34 +01002861#endif
Paul Menage8793d852007-10-18 23:39:39 -07002862#ifdef CONFIG_PROC_PID_CPUSET
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002863 REG("cpuset", S_IRUGO, proc_cpuset_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002864#endif
Paul Menagea4243162007-10-18 23:39:35 -07002865#ifdef CONFIG_CGROUPS
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002866 REG("cgroup", S_IRUGO, proc_cgroup_operations),
Paul Menagea4243162007-10-18 23:39:35 -07002867#endif
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002868 INF("oom_score", S_IRUGO, proc_oom_score),
2869 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002870#ifdef CONFIG_AUDITSYSCALL
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002871 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
2872 REG("sessionid", S_IRUSR, proc_sessionid_operations),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002873#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002874#ifdef CONFIG_FAULT_INJECTION
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002875 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002876#endif
Andrea Righi297c5d92008-07-25 01:48:49 -07002877#ifdef CONFIG_TASK_IO_ACCOUNTING
Alexey Dobriyan631f9c12008-11-10 01:32:52 +03002878 INF("io", S_IRUGO, proc_tid_io_accounting),
Andrea Righi297c5d92008-07-25 01:48:49 -07002879#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002880};
2881
2882static int proc_tid_base_readdir(struct file * filp,
2883 void * dirent, filldir_t filldir)
2884{
2885 return proc_pident_readdir(filp,dirent,filldir,
2886 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2887}
2888
2889static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002890 return proc_pident_lookup(dir, dentry,
2891 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002892}
2893
Arjan van de Ven00977a52007-02-12 00:55:34 -08002894static const struct file_operations proc_tid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002895 .read = generic_read_dir,
2896 .readdir = proc_tid_base_readdir,
2897};
2898
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002899static const struct inode_operations proc_tid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002900 .lookup = proc_tid_base_lookup,
2901 .getattr = pid_getattr,
2902 .setattr = proc_setattr,
2903};
2904
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002905static struct dentry *proc_task_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002906 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002907{
2908 struct dentry *error = ERR_PTR(-ENOENT);
2909 struct inode *inode;
Eric W. Biederman61a28782006-10-02 02:18:49 -07002910 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002911
2912 if (!inode)
2913 goto out;
2914 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2915 inode->i_op = &proc_tid_base_inode_operations;
2916 inode->i_fop = &proc_tid_base_operations;
2917 inode->i_flags|=S_IMMUTABLE;
Vegard Nossumaed54172008-06-05 22:46:53 -07002918
2919 inode->i_nlink = 2 + pid_entry_count_dirs(tid_base_stuff,
2920 ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002921
2922 dentry->d_op = &pid_dentry_operations;
2923
2924 d_add(dentry, inode);
2925 /* Close the race of the process dying before we return the dentry */
2926 if (pid_revalidate(dentry, NULL))
2927 error = NULL;
2928out:
2929 return error;
2930}
2931
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002932static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2933{
2934 struct dentry *result = ERR_PTR(-ENOENT);
2935 struct task_struct *task;
2936 struct task_struct *leader = get_proc_task(dir);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002937 unsigned tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002938 struct pid_namespace *ns;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002939
2940 if (!leader)
2941 goto out_no_task;
2942
2943 tid = name_to_int(dentry);
2944 if (tid == ~0U)
2945 goto out;
2946
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002947 ns = dentry->d_sb->s_fs_info;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002948 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002949 task = find_task_by_pid_ns(tid, ns);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002950 if (task)
2951 get_task_struct(task);
2952 rcu_read_unlock();
2953 if (!task)
2954 goto out;
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07002955 if (!same_thread_group(leader, task))
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002956 goto out_drop_task;
2957
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002958 result = proc_task_instantiate(dir, dentry, task, NULL);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002959out_drop_task:
2960 put_task_struct(task);
2961out:
2962 put_task_struct(leader);
2963out_no_task:
2964 return result;
2965}
2966
2967/*
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002968 * Find the first tid of a thread group to return to user space.
2969 *
2970 * Usually this is just the thread group leader, but if the users
2971 * buffer was too small or there was a seek into the middle of the
2972 * directory we have more work todo.
2973 *
2974 * In the case of a short read we start with find_task_by_pid.
2975 *
2976 * In the case of a seek we start with the leader and walk nr
2977 * threads past it.
2978 */
Eric W. Biedermancc288732006-06-26 00:26:01 -07002979static struct task_struct *first_tid(struct task_struct *leader,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002980 int tid, int nr, struct pid_namespace *ns)
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002981{
Oleg Nesterova872ff02006-06-26 00:26:01 -07002982 struct task_struct *pos;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002983
Eric W. Biedermancc288732006-06-26 00:26:01 -07002984 rcu_read_lock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002985 /* Attempt to start with the pid of a thread */
2986 if (tid && (nr > 0)) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002987 pos = find_task_by_pid_ns(tid, ns);
Oleg Nesterova872ff02006-06-26 00:26:01 -07002988 if (pos && (pos->group_leader == leader))
2989 goto found;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002990 }
2991
2992 /* If nr exceeds the number of threads there is nothing todo */
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002993 pos = NULL;
Oleg Nesterova872ff02006-06-26 00:26:01 -07002994 if (nr && nr >= get_nr_threads(leader))
2995 goto out;
2996
2997 /* If we haven't found our starting place yet start
2998 * with the leader and walk nr threads forward.
2999 */
3000 for (pos = leader; nr > 0; --nr) {
3001 pos = next_thread(pos);
3002 if (pos == leader) {
3003 pos = NULL;
3004 goto out;
3005 }
3006 }
3007found:
3008 get_task_struct(pos);
3009out:
Eric W. Biedermancc288732006-06-26 00:26:01 -07003010 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003011 return pos;
3012}
3013
3014/*
3015 * Find the next thread in the thread list.
3016 * Return NULL if there is an error or no next thread.
3017 *
3018 * The reference to the input task_struct is released.
3019 */
3020static struct task_struct *next_tid(struct task_struct *start)
3021{
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003022 struct task_struct *pos = NULL;
Eric W. Biedermancc288732006-06-26 00:26:01 -07003023 rcu_read_lock();
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003024 if (pid_alive(start)) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003025 pos = next_thread(start);
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07003026 if (thread_group_leader(pos))
3027 pos = NULL;
3028 else
3029 get_task_struct(pos);
3030 }
Eric W. Biedermancc288732006-06-26 00:26:01 -07003031 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003032 put_task_struct(start);
3033 return pos;
3034}
3035
Eric W. Biederman61a28782006-10-02 02:18:49 -07003036static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3037 struct task_struct *task, int tid)
3038{
3039 char name[PROC_NUMBUF];
3040 int len = snprintf(name, sizeof(name), "%d", tid);
3041 return proc_fill_cache(filp, dirent, filldir, name, len,
3042 proc_task_instantiate, task, NULL);
3043}
3044
Linus Torvalds1da177e2005-04-16 15:20:36 -07003045/* for the /proc/TGID/task/ directories */
3046static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3047{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08003048 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003049 struct inode *inode = dentry->d_inode;
Guillaume Chazarain7d895242007-01-31 23:48:14 -08003050 struct task_struct *leader = NULL;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003051 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003052 int retval = -ENOENT;
3053 ino_t ino;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003054 int tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003055 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003056
Guillaume Chazarain7d895242007-01-31 23:48:14 -08003057 task = get_proc_task(inode);
3058 if (!task)
3059 goto out_no_task;
3060 rcu_read_lock();
3061 if (pid_alive(task)) {
3062 leader = task->group_leader;
3063 get_task_struct(leader);
3064 }
3065 rcu_read_unlock();
3066 put_task_struct(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003067 if (!leader)
3068 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003069 retval = 0;
3070
Linus Torvaldsee568b22009-03-17 10:02:35 -07003071 switch ((unsigned long)filp->f_pos) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003072 case 0:
3073 ino = inode->i_ino;
Zhang Leee6f7792009-03-16 14:44:31 +08003074 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003075 goto out;
Zhang Leee6f7792009-03-16 14:44:31 +08003076 filp->f_pos++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003077 /* fall through */
3078 case 1:
3079 ino = parent_ino(dentry);
Zhang Leee6f7792009-03-16 14:44:31 +08003080 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003081 goto out;
Zhang Leee6f7792009-03-16 14:44:31 +08003082 filp->f_pos++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003083 /* fall through */
3084 }
3085
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003086 /* f_version caches the tgid value that the last readdir call couldn't
3087 * return. lseek aka telldir automagically resets f_version to 0.
3088 */
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003089 ns = filp->f_dentry->d_sb->s_fs_info;
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07003090 tid = (int)filp->f_version;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003091 filp->f_version = 0;
Zhang Leee6f7792009-03-16 14:44:31 +08003092 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003093 task;
Zhang Leee6f7792009-03-16 14:44:31 +08003094 task = next_tid(task), filp->f_pos++) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07003095 tid = task_pid_nr_ns(task, ns);
Eric W. Biederman61a28782006-10-02 02:18:49 -07003096 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003097 /* returning this tgid failed, save it as the first
3098 * pid for the next readir call */
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07003099 filp->f_version = (u64)tid;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003100 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003101 break;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07003102 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07003103 }
3104out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07003105 put_task_struct(leader);
3106out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07003107 return retval;
3108}
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003109
3110static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3111{
3112 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07003113 struct task_struct *p = get_proc_task(inode);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003114 generic_fillattr(inode, stat);
3115
Eric W. Biederman99f89552006-06-26 00:25:55 -07003116 if (p) {
Eric W. Biederman99f89552006-06-26 00:25:55 -07003117 stat->nlink += get_nr_threads(p);
Eric W. Biederman99f89552006-06-26 00:25:55 -07003118 put_task_struct(p);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07003119 }
3120
3121 return 0;
3122}
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003123
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08003124static const struct inode_operations proc_task_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003125 .lookup = proc_task_lookup,
3126 .getattr = proc_task_getattr,
3127 .setattr = proc_setattr,
3128};
3129
Arjan van de Ven00977a52007-02-12 00:55:34 -08003130static const struct file_operations proc_task_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07003131 .read = generic_read_dir,
3132 .readdir = proc_task_readdir,
3133};