blob: c59852b38787382a9399de7c69ae787ce22d9bf4 [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>
56#include <linux/init.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080057#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070058#include <linux/file.h>
59#include <linux/string.h>
60#include <linux/seq_file.h>
61#include <linux/namei.h>
Kirill Korotaev6b3286e2006-12-08 02:37:56 -080062#include <linux/mnt_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070063#include <linux/mm.h>
Dipankar Sarmab8359962005-09-09 13:04:14 -070064#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070065#include <linux/kallsyms.h>
Neil Hormand85f50d2007-10-18 23:40:37 -070066#include <linux/resource.h>
Kees Cook5096add2007-05-08 00:26:04 -070067#include <linux/module.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <linux/mount.h>
69#include <linux/security.h>
70#include <linux/ptrace.h>
Paul Menagea4243162007-10-18 23:39:35 -070071#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070072#include <linux/cpuset.h>
73#include <linux/audit.h>
Al Viro5addc5d2005-11-07 17:15:49 -050074#include <linux/poll.h>
Serge E. Hallyn1651e142006-10-02 02:18:08 -070075#include <linux/nsproxy.h>
Alexey Dobriyan8ac773b2006-10-19 23:28:32 -070076#include <linux/oom.h>
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -070077#include <linux/elf.h>
Pavel Emelyanov60347f62007-10-18 23:40:03 -070078#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070079#include "internal.h"
80
Eric W. Biederman0f2fe202006-06-26 00:25:46 -070081/* NOTE:
82 * Implementing inode permission operations in /proc is almost
83 * certainly an error. Permission checks need to happen during
84 * each system call not at open time. The reason is that most of
85 * what we wish to check for permissions in /proc varies at runtime.
86 *
87 * The classic example of a problem is opening file descriptors
88 * in /proc for a task before it execs a suid executable.
89 */
90
Linus Torvalds1da177e2005-04-16 15:20:36 -070091struct pid_entry {
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 char *name;
Eric Dumazetc5141e62007-05-08 00:26:15 -070093 int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -070094 mode_t mode;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -080095 const struct inode_operations *iop;
Arjan van de Ven00977a52007-02-12 00:55:34 -080096 const struct file_operations *fop;
Eric W. Biederman20cdc892006-10-02 02:17:07 -070097 union proc_op op;
Linus Torvalds1da177e2005-04-16 15:20:36 -070098};
99
Eric W. Biederman61a28782006-10-02 02:18:49 -0700100#define NOD(NAME, MODE, IOP, FOP, OP) { \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700101 .name = (NAME), \
Eric Dumazetc5141e62007-05-08 00:26:15 -0700102 .len = sizeof(NAME) - 1, \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700103 .mode = MODE, \
104 .iop = IOP, \
105 .fop = FOP, \
106 .op = OP, \
107}
108
Eric W. Biederman61a28782006-10-02 02:18:49 -0700109#define DIR(NAME, MODE, OTYPE) \
110 NOD(NAME, (S_IFDIR|(MODE)), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700111 &proc_##OTYPE##_inode_operations, &proc_##OTYPE##_operations, \
112 {} )
Eric W. Biederman61a28782006-10-02 02:18:49 -0700113#define LNK(NAME, OTYPE) \
114 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700115 &proc_pid_link_inode_operations, NULL, \
116 { .proc_get_link = &proc_##OTYPE##_link } )
Eric W. Biederman61a28782006-10-02 02:18:49 -0700117#define REG(NAME, MODE, OTYPE) \
118 NOD(NAME, (S_IFREG|(MODE)), NULL, \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700119 &proc_##OTYPE##_operations, {})
Eric W. Biederman61a28782006-10-02 02:18:49 -0700120#define INF(NAME, MODE, OTYPE) \
121 NOD(NAME, (S_IFREG|(MODE)), \
Eric W. Biederman20cdc892006-10-02 02:17:07 -0700122 NULL, &proc_info_file_operations, \
123 { .proc_read = &proc_##OTYPE } )
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124
Kees Cook5096add2007-05-08 00:26:04 -0700125int maps_protect;
126EXPORT_SYMBOL(maps_protect);
127
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700128static struct fs_struct *get_fs_struct(struct task_struct *task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129{
130 struct fs_struct *fs;
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700131 task_lock(task);
132 fs = task->fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133 if(fs)
134 atomic_inc(&fs->count);
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700135 task_unlock(task);
136 return fs;
137}
138
Eric W. Biederman99f89552006-06-26 00:25:55 -0700139static int get_nr_threads(struct task_struct *tsk)
140{
141 /* Must be called with the rcu_read_lock held */
142 unsigned long flags;
143 int count = 0;
144
145 if (lock_task_sighand(tsk, &flags)) {
146 count = atomic_read(&tsk->signal->count);
147 unlock_task_sighand(tsk, &flags);
148 }
149 return count;
150}
151
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700152static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
153{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700154 struct task_struct *task = get_proc_task(inode);
155 struct fs_struct *fs = NULL;
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700156 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700157
158 if (task) {
159 fs = get_fs_struct(task);
160 put_task_struct(task);
161 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 if (fs) {
163 read_lock(&fs->lock);
164 *mnt = mntget(fs->pwdmnt);
165 *dentry = dget(fs->pwd);
166 read_unlock(&fs->lock);
167 result = 0;
168 put_fs_struct(fs);
169 }
170 return result;
171}
172
173static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
174{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700175 struct task_struct *task = get_proc_task(inode);
176 struct fs_struct *fs = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700178
179 if (task) {
180 fs = get_fs_struct(task);
181 put_task_struct(task);
182 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 if (fs) {
184 read_lock(&fs->lock);
185 *mnt = mntget(fs->rootmnt);
186 *dentry = dget(fs->root);
187 read_unlock(&fs->lock);
188 result = 0;
189 put_fs_struct(fs);
190 }
191 return result;
192}
193
194#define MAY_PTRACE(task) \
195 (task == current || \
196 (task->parent == current && \
197 (task->ptrace & PT_PTRACED) && \
Matthew Wilcox6d8982d2007-12-06 11:04:01 -0500198 (task_is_stopped_or_traced(task)) && \
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 security_ptrace(current,task) == 0))
200
Al Viro831830b2008-01-02 14:09:57 +0000201struct mm_struct *mm_for_maps(struct task_struct *task)
202{
203 struct mm_struct *mm = get_task_mm(task);
204 if (!mm)
205 return NULL;
206 down_read(&mm->mmap_sem);
207 task_lock(task);
208 if (task->mm != mm)
209 goto out;
210 if (task->mm != current->mm && __ptrace_may_attach(task) < 0)
211 goto out;
212 task_unlock(task);
213 return mm;
214out:
215 task_unlock(task);
216 up_read(&mm->mmap_sem);
217 mmput(mm);
218 return NULL;
219}
220
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221static int proc_pid_cmdline(struct task_struct *task, char * buffer)
222{
223 int res = 0;
224 unsigned int len;
225 struct mm_struct *mm = get_task_mm(task);
226 if (!mm)
227 goto out;
228 if (!mm->arg_end)
229 goto out_mm; /* Shh! No looking before we're done */
230
231 len = mm->arg_end - mm->arg_start;
232
233 if (len > PAGE_SIZE)
234 len = PAGE_SIZE;
235
236 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
237
238 // If the nul at the end of args has been overwritten, then
239 // assume application is using setproctitle(3).
240 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
241 len = strnlen(buffer, res);
242 if (len < res) {
243 res = len;
244 } else {
245 len = mm->env_end - mm->env_start;
246 if (len > PAGE_SIZE - res)
247 len = PAGE_SIZE - res;
248 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
249 res = strnlen(buffer, res);
250 }
251 }
252out_mm:
253 mmput(mm);
254out:
255 return res;
256}
257
258static int proc_pid_auxv(struct task_struct *task, char *buffer)
259{
260 int res = 0;
261 struct mm_struct *mm = get_task_mm(task);
262 if (mm) {
263 unsigned int nwords = 0;
264 do
265 nwords += 2;
266 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
267 res = nwords * sizeof(mm->saved_auxv[0]);
268 if (res > PAGE_SIZE)
269 res = PAGE_SIZE;
270 memcpy(buffer, mm->saved_auxv, res);
271 mmput(mm);
272 }
273 return res;
274}
275
276
277#ifdef CONFIG_KALLSYMS
278/*
279 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
280 * Returns the resolved symbol. If that fails, simply return the address.
281 */
282static int proc_pid_wchan(struct task_struct *task, char *buffer)
283{
Alexey Dobriyanffb45122007-05-08 00:28:41 -0700284 unsigned long wchan;
Tejun Heo9281ace2007-07-17 04:03:51 -0700285 char symname[KSYM_NAME_LEN];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286
287 wchan = get_wchan(task);
288
Alexey Dobriyan9d65cb42007-05-08 00:28:43 -0700289 if (lookup_symbol_name(wchan, symname) < 0)
290 return sprintf(buffer, "%lu", wchan);
291 else
292 return sprintf(buffer, "%s", symname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293}
294#endif /* CONFIG_KALLSYMS */
295
296#ifdef CONFIG_SCHEDSTATS
297/*
298 * Provides /proc/PID/schedstat
299 */
300static int proc_pid_schedstat(struct task_struct *task, char *buffer)
301{
Balbir Singh172ba842007-07-09 18:52:00 +0200302 return sprintf(buffer, "%llu %llu %lu\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 task->sched_info.cpu_time,
304 task->sched_info.run_delay,
Ingo Molnar2d723762007-10-15 17:00:12 +0200305 task->sched_info.pcount);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306}
307#endif
308
Arjan van de Ven97455122008-01-25 21:08:34 +0100309#ifdef CONFIG_LATENCYTOP
310static int lstats_show_proc(struct seq_file *m, void *v)
311{
312 int i;
313 struct task_struct *task = m->private;
314 seq_puts(m, "Latency Top version : v0.1\n");
315
316 for (i = 0; i < 32; i++) {
317 if (task->latency_record[i].backtrace[0]) {
318 int q;
319 seq_printf(m, "%i %li %li ",
320 task->latency_record[i].count,
321 task->latency_record[i].time,
322 task->latency_record[i].max);
323 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
324 char sym[KSYM_NAME_LEN];
325 char *c;
326 if (!task->latency_record[i].backtrace[q])
327 break;
328 if (task->latency_record[i].backtrace[q] == ULONG_MAX)
329 break;
330 sprint_symbol(sym, task->latency_record[i].backtrace[q]);
331 c = strchr(sym, '+');
332 if (c)
333 *c = 0;
334 seq_printf(m, "%s ", sym);
335 }
336 seq_printf(m, "\n");
337 }
338
339 }
340 return 0;
341}
342
343static int lstats_open(struct inode *inode, struct file *file)
344{
345 int ret;
346 struct seq_file *m;
347 struct task_struct *task = get_proc_task(inode);
348
349 ret = single_open(file, lstats_show_proc, NULL);
350 if (!ret) {
351 m = file->private_data;
352 m->private = task;
353 }
354 return ret;
355}
356
357static ssize_t lstats_write(struct file *file, const char __user *buf,
358 size_t count, loff_t *offs)
359{
360 struct seq_file *m;
361 struct task_struct *task;
362
363 m = file->private_data;
364 task = m->private;
365 clear_all_latency_tracing(task);
366
367 return count;
368}
369
370static const struct file_operations proc_lstats_operations = {
371 .open = lstats_open,
372 .read = seq_read,
373 .write = lstats_write,
374 .llseek = seq_lseek,
375 .release = single_release,
376};
377
378#endif
379
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380/* The badness from the OOM killer */
381unsigned long badness(struct task_struct *p, unsigned long uptime);
382static int proc_oom_score(struct task_struct *task, char *buffer)
383{
384 unsigned long points;
385 struct timespec uptime;
386
387 do_posix_clock_monotonic_gettime(&uptime);
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700388 read_lock(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 points = badness(task, uptime.tv_sec);
Alexey Dobriyan19c5d452007-05-08 00:26:46 -0700390 read_unlock(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 return sprintf(buffer, "%lu\n", points);
392}
393
Neil Hormand85f50d2007-10-18 23:40:37 -0700394struct limit_names {
395 char *name;
396 char *unit;
397};
398
399static const struct limit_names lnames[RLIM_NLIMITS] = {
400 [RLIMIT_CPU] = {"Max cpu time", "ms"},
401 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
402 [RLIMIT_DATA] = {"Max data size", "bytes"},
403 [RLIMIT_STACK] = {"Max stack size", "bytes"},
404 [RLIMIT_CORE] = {"Max core file size", "bytes"},
405 [RLIMIT_RSS] = {"Max resident set", "bytes"},
406 [RLIMIT_NPROC] = {"Max processes", "processes"},
407 [RLIMIT_NOFILE] = {"Max open files", "files"},
408 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
409 [RLIMIT_AS] = {"Max address space", "bytes"},
410 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
411 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
412 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
413 [RLIMIT_NICE] = {"Max nice priority", NULL},
414 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
415};
416
417/* Display limits for a process */
418static int proc_pid_limits(struct task_struct *task, char *buffer)
419{
420 unsigned int i;
421 int count = 0;
422 unsigned long flags;
423 char *bufptr = buffer;
424
425 struct rlimit rlim[RLIM_NLIMITS];
426
427 rcu_read_lock();
428 if (!lock_task_sighand(task,&flags)) {
429 rcu_read_unlock();
430 return 0;
431 }
432 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
433 unlock_task_sighand(task, &flags);
434 rcu_read_unlock();
435
436 /*
437 * print the file header
438 */
439 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
440 "Limit", "Soft Limit", "Hard Limit", "Units");
441
442 for (i = 0; i < RLIM_NLIMITS; i++) {
443 if (rlim[i].rlim_cur == RLIM_INFINITY)
444 count += sprintf(&bufptr[count], "%-25s %-20s ",
445 lnames[i].name, "unlimited");
446 else
447 count += sprintf(&bufptr[count], "%-25s %-20lu ",
448 lnames[i].name, rlim[i].rlim_cur);
449
450 if (rlim[i].rlim_max == RLIM_INFINITY)
451 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
452 else
453 count += sprintf(&bufptr[count], "%-20lu ",
454 rlim[i].rlim_max);
455
456 if (lnames[i].unit)
457 count += sprintf(&bufptr[count], "%-10s\n",
458 lnames[i].unit);
459 else
460 count += sprintf(&bufptr[count], "\n");
461 }
462
463 return count;
464}
465
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466/************************************************************************/
467/* Here the fs part begins */
468/************************************************************************/
469
470/* permission checks */
Eric W. Biederman778c1142006-06-26 00:25:58 -0700471static int proc_fd_access_allowed(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472{
Eric W. Biederman778c1142006-06-26 00:25:58 -0700473 struct task_struct *task;
474 int allowed = 0;
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700475 /* Allow access to a task's file descriptors if it is us or we
476 * may use ptrace attach to the process and find out that
477 * information.
Eric W. Biederman778c1142006-06-26 00:25:58 -0700478 */
479 task = get_proc_task(inode);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700480 if (task) {
481 allowed = ptrace_may_attach(task);
Eric W. Biederman778c1142006-06-26 00:25:58 -0700482 put_task_struct(task);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700483 }
Eric W. Biederman778c1142006-06-26 00:25:58 -0700484 return allowed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485}
486
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700487static int proc_setattr(struct dentry *dentry, struct iattr *attr)
488{
489 int error;
490 struct inode *inode = dentry->d_inode;
491
492 if (attr->ia_valid & ATTR_MODE)
493 return -EPERM;
494
495 error = inode_change_ok(inode, attr);
John Johansen1e8123f2007-05-08 00:29:41 -0700496 if (!error)
497 error = inode_setattr(inode, attr);
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700498 return error;
499}
500
Arjan van de Venc5ef1c42007-02-12 00:55:40 -0800501static const struct inode_operations proc_def_inode_operations = {
Linus Torvalds6d76fa52006-07-15 12:26:45 -0700502 .setattr = proc_setattr,
503};
504
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505extern struct seq_operations mounts_op;
Al Viro5addc5d2005-11-07 17:15:49 -0500506struct proc_mounts {
507 struct seq_file m;
508 int event;
509};
510
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511static int mounts_open(struct inode *inode, struct file *file)
512{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700513 struct task_struct *task = get_proc_task(inode);
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700514 struct nsproxy *nsp;
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800515 struct mnt_namespace *ns = NULL;
Al Viro5addc5d2005-11-07 17:15:49 -0500516 struct proc_mounts *p;
517 int ret = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518
Eric W. Biederman99f89552006-06-26 00:25:55 -0700519 if (task) {
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700520 rcu_read_lock();
521 nsp = task_nsproxy(task);
522 if (nsp) {
523 ns = nsp->mnt_ns;
Alexey Dobriyan863c4702007-01-26 00:56:53 -0800524 if (ns)
525 get_mnt_ns(ns);
526 }
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700527 rcu_read_unlock();
528
Eric W. Biederman99f89552006-06-26 00:25:55 -0700529 put_task_struct(task);
530 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800532 if (ns) {
Al Viro5addc5d2005-11-07 17:15:49 -0500533 ret = -ENOMEM;
534 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
535 if (p) {
536 file->private_data = &p->m;
537 ret = seq_open(file, &mounts_op);
538 if (!ret) {
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800539 p->m.private = ns;
540 p->event = ns->event;
Al Viro5addc5d2005-11-07 17:15:49 -0500541 return 0;
542 }
543 kfree(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 }
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800545 put_mnt_ns(ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 }
547 return ret;
548}
549
550static int mounts_release(struct inode *inode, struct file *file)
551{
552 struct seq_file *m = file->private_data;
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800553 struct mnt_namespace *ns = m->private;
554 put_mnt_ns(ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 return seq_release(inode, file);
556}
557
Al Viro5addc5d2005-11-07 17:15:49 -0500558static unsigned mounts_poll(struct file *file, poll_table *wait)
559{
560 struct proc_mounts *p = file->private_data;
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800561 struct mnt_namespace *ns = p->m.private;
Al Viro5addc5d2005-11-07 17:15:49 -0500562 unsigned res = 0;
563
564 poll_wait(file, &ns->poll, wait);
565
566 spin_lock(&vfsmount_lock);
567 if (p->event != ns->event) {
568 p->event = ns->event;
569 res = POLLERR;
570 }
571 spin_unlock(&vfsmount_lock);
572
573 return res;
574}
575
Arjan van de Ven00977a52007-02-12 00:55:34 -0800576static const struct file_operations proc_mounts_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 .open = mounts_open,
578 .read = seq_read,
579 .llseek = seq_lseek,
580 .release = mounts_release,
Al Viro5addc5d2005-11-07 17:15:49 -0500581 .poll = mounts_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582};
583
Chuck Leverb4629fe2006-03-20 13:44:12 -0500584extern struct seq_operations mountstats_op;
585static int mountstats_open(struct inode *inode, struct file *file)
586{
Chuck Leverb4629fe2006-03-20 13:44:12 -0500587 int ret = seq_open(file, &mountstats_op);
588
589 if (!ret) {
590 struct seq_file *m = file->private_data;
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700591 struct nsproxy *nsp;
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800592 struct mnt_namespace *mnt_ns = NULL;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700593 struct task_struct *task = get_proc_task(inode);
594
595 if (task) {
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -0700596 rcu_read_lock();
597 nsp = task_nsproxy(task);
598 if (nsp) {
599 mnt_ns = nsp->mnt_ns;
600 if (mnt_ns)
601 get_mnt_ns(mnt_ns);
602 }
603 rcu_read_unlock();
604
Eric W. Biederman99f89552006-06-26 00:25:55 -0700605 put_task_struct(task);
606 }
Chuck Leverb4629fe2006-03-20 13:44:12 -0500607
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800608 if (mnt_ns)
609 m->private = mnt_ns;
Chuck Leverb4629fe2006-03-20 13:44:12 -0500610 else {
611 seq_release(inode, file);
612 ret = -EINVAL;
613 }
614 }
615 return ret;
616}
617
Arjan van de Ven00977a52007-02-12 00:55:34 -0800618static const struct file_operations proc_mountstats_operations = {
Chuck Leverb4629fe2006-03-20 13:44:12 -0500619 .open = mountstats_open,
620 .read = seq_read,
621 .llseek = seq_lseek,
622 .release = mounts_release,
623};
624
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
626
627static ssize_t proc_info_read(struct file * file, char __user * buf,
628 size_t count, loff_t *ppos)
629{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800630 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 unsigned long page;
632 ssize_t length;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700633 struct task_struct *task = get_proc_task(inode);
634
635 length = -ESRCH;
636 if (!task)
637 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638
639 if (count > PROC_BLOCK_SIZE)
640 count = PROC_BLOCK_SIZE;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700641
642 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700643 if (!(page = __get_free_page(GFP_TEMPORARY)))
Eric W. Biederman99f89552006-06-26 00:25:55 -0700644 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645
646 length = PROC_I(inode)->op.proc_read(task, (char*)page);
647
648 if (length >= 0)
649 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
650 free_page(page);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700651out:
652 put_task_struct(task);
653out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 return length;
655}
656
Arjan van de Ven00977a52007-02-12 00:55:34 -0800657static const struct file_operations proc_info_file_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 .read = proc_info_read,
659};
660
661static int mem_open(struct inode* inode, struct file* file)
662{
663 file->private_data = (void*)((long)current->self_exec_id);
664 return 0;
665}
666
667static ssize_t mem_read(struct file * file, char __user * buf,
668 size_t count, loff_t *ppos)
669{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800670 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 char *page;
672 unsigned long src = *ppos;
673 int ret = -ESRCH;
674 struct mm_struct *mm;
675
Eric W. Biederman99f89552006-06-26 00:25:55 -0700676 if (!task)
677 goto out_no_task;
678
Miklos Szerediab8d11b2005-09-06 15:18:24 -0700679 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 goto out;
681
682 ret = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700683 page = (char *)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 if (!page)
685 goto out;
686
687 ret = 0;
688
689 mm = get_task_mm(task);
690 if (!mm)
691 goto out_free;
692
693 ret = -EIO;
694
695 if (file->private_data != (void*)((long)current->self_exec_id))
696 goto out_put;
697
698 ret = 0;
699
700 while (count > 0) {
701 int this_len, retval;
702
703 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
704 retval = access_process_vm(task, src, page, this_len, 0);
Miklos Szerediab8d11b2005-09-06 15:18:24 -0700705 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 if (!ret)
707 ret = -EIO;
708 break;
709 }
710
711 if (copy_to_user(buf, page, retval)) {
712 ret = -EFAULT;
713 break;
714 }
715
716 ret += retval;
717 src += retval;
718 buf += retval;
719 count -= retval;
720 }
721 *ppos = src;
722
723out_put:
724 mmput(mm);
725out_free:
726 free_page((unsigned long) page);
727out:
Eric W. Biederman99f89552006-06-26 00:25:55 -0700728 put_task_struct(task);
729out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 return ret;
731}
732
733#define mem_write NULL
734
735#ifndef mem_write
736/* This is a security hazard */
Glauber de Oliveira Costa63967fa2007-02-20 13:58:12 -0800737static ssize_t mem_write(struct file * file, const char __user *buf,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 size_t count, loff_t *ppos)
739{
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700740 int copied;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741 char *page;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800742 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 unsigned long dst = *ppos;
744
Eric W. Biederman99f89552006-06-26 00:25:55 -0700745 copied = -ESRCH;
746 if (!task)
747 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748
Eric W. Biederman99f89552006-06-26 00:25:55 -0700749 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
750 goto out;
751
752 copied = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -0700753 page = (char *)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 if (!page)
Eric W. Biederman99f89552006-06-26 00:25:55 -0700755 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756
Frederik Deweerdtf7ca54f2006-09-29 02:01:02 -0700757 copied = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 while (count > 0) {
759 int this_len, retval;
760
761 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
762 if (copy_from_user(page, buf, this_len)) {
763 copied = -EFAULT;
764 break;
765 }
766 retval = access_process_vm(task, dst, page, this_len, 1);
767 if (!retval) {
768 if (!copied)
769 copied = -EIO;
770 break;
771 }
772 copied += retval;
773 buf += retval;
774 dst += retval;
775 count -= retval;
776 }
777 *ppos = dst;
778 free_page((unsigned long) page);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700779out:
780 put_task_struct(task);
781out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 return copied;
783}
784#endif
785
Matt Mackall85863e42008-02-04 22:29:04 -0800786loff_t mem_lseek(struct file *file, loff_t offset, int orig)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787{
788 switch (orig) {
789 case 0:
790 file->f_pos = offset;
791 break;
792 case 1:
793 file->f_pos += offset;
794 break;
795 default:
796 return -EINVAL;
797 }
798 force_successful_syscall_return();
799 return file->f_pos;
800}
801
Arjan van de Ven00977a52007-02-12 00:55:34 -0800802static const struct file_operations proc_mem_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 .llseek = mem_lseek,
804 .read = mem_read,
805 .write = mem_write,
806 .open = mem_open,
807};
808
James Pearson315e28c2007-10-16 23:30:17 -0700809static ssize_t environ_read(struct file *file, char __user *buf,
810 size_t count, loff_t *ppos)
811{
812 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
813 char *page;
814 unsigned long src = *ppos;
815 int ret = -ESRCH;
816 struct mm_struct *mm;
817
818 if (!task)
819 goto out_no_task;
820
821 if (!ptrace_may_attach(task))
822 goto out;
823
824 ret = -ENOMEM;
825 page = (char *)__get_free_page(GFP_TEMPORARY);
826 if (!page)
827 goto out;
828
829 ret = 0;
830
831 mm = get_task_mm(task);
832 if (!mm)
833 goto out_free;
834
835 while (count > 0) {
836 int this_len, retval, max_len;
837
838 this_len = mm->env_end - (mm->env_start + src);
839
840 if (this_len <= 0)
841 break;
842
843 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
844 this_len = (this_len > max_len) ? max_len : this_len;
845
846 retval = access_process_vm(task, (mm->env_start + src),
847 page, this_len, 0);
848
849 if (retval <= 0) {
850 ret = retval;
851 break;
852 }
853
854 if (copy_to_user(buf, page, retval)) {
855 ret = -EFAULT;
856 break;
857 }
858
859 ret += retval;
860 src += retval;
861 buf += retval;
862 count -= retval;
863 }
864 *ppos = src;
865
866 mmput(mm);
867out_free:
868 free_page((unsigned long) page);
869out:
870 put_task_struct(task);
871out_no_task:
872 return ret;
873}
874
875static const struct file_operations proc_environ_operations = {
876 .read = environ_read,
877};
878
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879static ssize_t oom_adjust_read(struct file *file, char __user *buf,
880 size_t count, loff_t *ppos)
881{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800882 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
Eric W. Biederman8578cea2006-06-26 00:25:54 -0700883 char buffer[PROC_NUMBUF];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 size_t len;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700885 int oom_adjust;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886
Eric W. Biederman99f89552006-06-26 00:25:55 -0700887 if (!task)
888 return -ESRCH;
889 oom_adjust = task->oomkilladj;
890 put_task_struct(task);
891
Eric W. Biederman8578cea2006-06-26 00:25:54 -0700892 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
Akinobu Mita0c28f282007-05-08 00:31:41 -0700893
894 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895}
896
897static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
898 size_t count, loff_t *ppos)
899{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700900 struct task_struct *task;
Eric W. Biederman8578cea2006-06-26 00:25:54 -0700901 char buffer[PROC_NUMBUF], *end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 int oom_adjust;
903
Eric W. Biederman8578cea2006-06-26 00:25:54 -0700904 memset(buffer, 0, sizeof(buffer));
905 if (count > sizeof(buffer) - 1)
906 count = sizeof(buffer) - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 if (copy_from_user(buffer, buf, count))
908 return -EFAULT;
909 oom_adjust = simple_strtol(buffer, &end, 0);
Alexey Dobriyan8ac773b2006-10-19 23:28:32 -0700910 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
911 oom_adjust != OOM_DISABLE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 return -EINVAL;
913 if (*end == '\n')
914 end++;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800915 task = get_proc_task(file->f_path.dentry->d_inode);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700916 if (!task)
917 return -ESRCH;
Guillem Jover8fb4fc62006-12-06 20:32:24 -0800918 if (oom_adjust < task->oomkilladj && !capable(CAP_SYS_RESOURCE)) {
919 put_task_struct(task);
920 return -EACCES;
921 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 task->oomkilladj = oom_adjust;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700923 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 if (end - buffer == 0)
925 return -EIO;
926 return end - buffer;
927}
928
Arjan van de Ven00977a52007-02-12 00:55:34 -0800929static const struct file_operations proc_oom_adjust_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 .read = oom_adjust_read,
931 .write = oom_adjust_write,
932};
933
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934#ifdef CONFIG_AUDITSYSCALL
935#define TMPBUFLEN 21
936static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
937 size_t count, loff_t *ppos)
938{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800939 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700940 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 ssize_t length;
942 char tmpbuf[TMPBUFLEN];
943
Eric W. Biederman99f89552006-06-26 00:25:55 -0700944 if (!task)
945 return -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
Al Viro0c11b942008-01-10 04:20:52 -0500947 audit_get_loginuid(task));
Eric W. Biederman99f89552006-06-26 00:25:55 -0700948 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
950}
951
952static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
953 size_t count, loff_t *ppos)
954{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -0800955 struct inode * inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 char *page, *tmp;
957 ssize_t length;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 uid_t loginuid;
959
960 if (!capable(CAP_AUDIT_CONTROL))
961 return -EPERM;
962
Eric W. Biederman13b41b02006-06-26 00:25:56 -0700963 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 return -EPERM;
965
Al Viroe0182902006-05-18 08:28:02 -0400966 if (count >= PAGE_SIZE)
967 count = PAGE_SIZE - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968
969 if (*ppos != 0) {
970 /* No partial writes. */
971 return -EINVAL;
972 }
Mel Gormane12ba742007-10-16 01:25:52 -0700973 page = (char*)__get_free_page(GFP_TEMPORARY);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 if (!page)
975 return -ENOMEM;
976 length = -EFAULT;
977 if (copy_from_user(page, buf, count))
978 goto out_free_page;
979
Al Viroe0182902006-05-18 08:28:02 -0400980 page[count] = '\0';
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 loginuid = simple_strtoul(page, &tmp, 10);
982 if (tmp == page) {
983 length = -EINVAL;
984 goto out_free_page;
985
986 }
Eric W. Biederman99f89552006-06-26 00:25:55 -0700987 length = audit_set_loginuid(current, loginuid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988 if (likely(length == 0))
989 length = count;
990
991out_free_page:
992 free_page((unsigned long) page);
993 return length;
994}
995
Arjan van de Ven00977a52007-02-12 00:55:34 -0800996static const struct file_operations proc_loginuid_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 .read = proc_loginuid_read,
998 .write = proc_loginuid_write,
999};
1000#endif
1001
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001002#ifdef CONFIG_FAULT_INJECTION
1003static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1004 size_t count, loff_t *ppos)
1005{
1006 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1007 char buffer[PROC_NUMBUF];
1008 size_t len;
1009 int make_it_fail;
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001010
1011 if (!task)
1012 return -ESRCH;
1013 make_it_fail = task->make_it_fail;
1014 put_task_struct(task);
1015
1016 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
Akinobu Mita0c28f282007-05-08 00:31:41 -07001017
1018 return simple_read_from_buffer(buf, count, ppos, buffer, len);
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001019}
1020
1021static ssize_t proc_fault_inject_write(struct file * file,
1022 const char __user * buf, size_t count, loff_t *ppos)
1023{
1024 struct task_struct *task;
1025 char buffer[PROC_NUMBUF], *end;
1026 int make_it_fail;
1027
1028 if (!capable(CAP_SYS_RESOURCE))
1029 return -EPERM;
1030 memset(buffer, 0, sizeof(buffer));
1031 if (count > sizeof(buffer) - 1)
1032 count = sizeof(buffer) - 1;
1033 if (copy_from_user(buffer, buf, count))
1034 return -EFAULT;
1035 make_it_fail = simple_strtol(buffer, &end, 0);
1036 if (*end == '\n')
1037 end++;
1038 task = get_proc_task(file->f_dentry->d_inode);
1039 if (!task)
1040 return -ESRCH;
1041 task->make_it_fail = make_it_fail;
1042 put_task_struct(task);
1043 if (end - buffer == 0)
1044 return -EIO;
1045 return end - buffer;
1046}
1047
Arjan van de Ven00977a52007-02-12 00:55:34 -08001048static const struct file_operations proc_fault_inject_operations = {
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08001049 .read = proc_fault_inject_read,
1050 .write = proc_fault_inject_write,
1051};
1052#endif
1053
Arjan van de Ven97455122008-01-25 21:08:34 +01001054
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001055#ifdef CONFIG_SCHED_DEBUG
1056/*
1057 * Print out various scheduling related per-task fields:
1058 */
1059static int sched_show(struct seq_file *m, void *v)
1060{
1061 struct inode *inode = m->private;
1062 struct task_struct *p;
1063
1064 WARN_ON(!inode);
1065
1066 p = get_proc_task(inode);
1067 if (!p)
1068 return -ESRCH;
1069 proc_sched_show_task(p, m);
1070
1071 put_task_struct(p);
1072
1073 return 0;
1074}
1075
1076static ssize_t
1077sched_write(struct file *file, const char __user *buf,
1078 size_t count, loff_t *offset)
1079{
1080 struct inode *inode = file->f_path.dentry->d_inode;
1081 struct task_struct *p;
1082
1083 WARN_ON(!inode);
1084
1085 p = get_proc_task(inode);
1086 if (!p)
1087 return -ESRCH;
1088 proc_sched_set_task(p);
1089
1090 put_task_struct(p);
1091
1092 return count;
1093}
1094
1095static int sched_open(struct inode *inode, struct file *filp)
1096{
1097 int ret;
1098
1099 ret = single_open(filp, sched_show, NULL);
1100 if (!ret) {
1101 struct seq_file *m = filp->private_data;
1102
1103 m->private = inode;
1104 }
1105 return ret;
1106}
1107
1108static const struct file_operations proc_pid_sched_operations = {
1109 .open = sched_open,
1110 .read = seq_read,
1111 .write = sched_write,
1112 .llseek = seq_lseek,
Alexey Dobriyan5ea473a2007-07-31 00:38:50 -07001113 .release = single_release,
Ingo Molnar43ae34c2007-07-09 18:52:00 +02001114};
1115
1116#endif
1117
Al Viro008b1502005-08-20 00:17:39 +01001118static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119{
1120 struct inode *inode = dentry->d_inode;
1121 int error = -EACCES;
1122
1123 /* We don't need a base pointer in the /proc filesystem */
1124 path_release(nd);
1125
Eric W. Biederman778c1142006-06-26 00:25:58 -07001126 /* Are we allowed to snoop on the tasks file descriptors? */
1127 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129
1130 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
1131 nd->last_type = LAST_BIND;
1132out:
Al Viro008b1502005-08-20 00:17:39 +01001133 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134}
1135
1136static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
1137 char __user *buffer, int buflen)
1138{
1139 struct inode * inode;
Mel Gormane12ba742007-10-16 01:25:52 -07001140 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1141 char *path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 int len;
1143
1144 if (!tmp)
1145 return -ENOMEM;
Akinobu Mita0c28f282007-05-08 00:31:41 -07001146
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 inode = dentry->d_inode;
1148 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
1149 len = PTR_ERR(path);
1150 if (IS_ERR(path))
1151 goto out;
1152 len = tmp + PAGE_SIZE - 1 - path;
1153
1154 if (len > buflen)
1155 len = buflen;
1156 if (copy_to_user(buffer, path, len))
1157 len = -EFAULT;
1158 out:
1159 free_page((unsigned long)tmp);
1160 return len;
1161}
1162
1163static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1164{
1165 int error = -EACCES;
1166 struct inode *inode = dentry->d_inode;
1167 struct dentry *de;
1168 struct vfsmount *mnt = NULL;
1169
Eric W. Biederman778c1142006-06-26 00:25:58 -07001170 /* Are we allowed to snoop on the tasks file descriptors? */
1171 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173
1174 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
1175 if (error)
1176 goto out;
1177
1178 error = do_proc_readlink(de, mnt, buffer, buflen);
1179 dput(de);
1180 mntput(mnt);
1181out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 return error;
1183}
1184
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08001185static const struct inode_operations proc_pid_link_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186 .readlink = proc_pid_readlink,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001187 .follow_link = proc_pid_follow_link,
1188 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189};
1190
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001191
1192/* building an inode */
1193
1194static int task_dumpable(struct task_struct *task)
1195{
1196 int dumpable = 0;
1197 struct mm_struct *mm;
1198
1199 task_lock(task);
1200 mm = task->mm;
1201 if (mm)
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001202 dumpable = get_dumpable(mm);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001203 task_unlock(task);
1204 if(dumpable == 1)
1205 return 1;
1206 return 0;
1207}
1208
1209
Eric W. Biederman61a28782006-10-02 02:18:49 -07001210static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001211{
1212 struct inode * inode;
1213 struct proc_inode *ei;
1214
1215 /* We need a new inode */
1216
1217 inode = new_inode(sb);
1218 if (!inode)
1219 goto out;
1220
1221 /* Common stuff */
1222 ei = PROC_I(inode);
1223 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001224 inode->i_op = &proc_def_inode_operations;
1225
1226 /*
1227 * grab the reference to task.
1228 */
Oleg Nesterov1a657f782006-10-02 02:18:59 -07001229 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001230 if (!ei->pid)
1231 goto out_unlock;
1232
1233 inode->i_uid = 0;
1234 inode->i_gid = 0;
1235 if (task_dumpable(task)) {
1236 inode->i_uid = task->euid;
1237 inode->i_gid = task->egid;
1238 }
1239 security_task_to_inode(task, inode);
1240
1241out:
1242 return inode;
1243
1244out_unlock:
1245 iput(inode);
1246 return NULL;
1247}
1248
1249static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1250{
1251 struct inode *inode = dentry->d_inode;
1252 struct task_struct *task;
1253 generic_fillattr(inode, stat);
1254
1255 rcu_read_lock();
1256 stat->uid = 0;
1257 stat->gid = 0;
1258 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1259 if (task) {
1260 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1261 task_dumpable(task)) {
1262 stat->uid = task->euid;
1263 stat->gid = task->egid;
1264 }
1265 }
1266 rcu_read_unlock();
1267 return 0;
1268}
1269
1270/* dentry stuff */
1271
1272/*
1273 * Exceptional case: normally we are not allowed to unhash a busy
1274 * directory. In this case, however, we can do it - no aliasing problems
1275 * due to the way we treat inodes.
1276 *
1277 * Rewrite the inode's ownerships here because the owning task may have
1278 * performed a setuid(), etc.
1279 *
1280 * Before the /proc/pid/status file was created the only way to read
1281 * the effective uid of a /process was to stat /proc/pid. Reading
1282 * /proc/pid/status is slow enough that procps and other packages
1283 * kept stating /proc/pid. To keep the rules in /proc simple I have
1284 * made this apply to all per process world readable and executable
1285 * directories.
1286 */
1287static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1288{
1289 struct inode *inode = dentry->d_inode;
1290 struct task_struct *task = get_proc_task(inode);
1291 if (task) {
1292 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1293 task_dumpable(task)) {
1294 inode->i_uid = task->euid;
1295 inode->i_gid = task->egid;
1296 } else {
1297 inode->i_uid = 0;
1298 inode->i_gid = 0;
1299 }
1300 inode->i_mode &= ~(S_ISUID | S_ISGID);
1301 security_task_to_inode(task, inode);
1302 put_task_struct(task);
1303 return 1;
1304 }
1305 d_drop(dentry);
1306 return 0;
1307}
1308
1309static int pid_delete_dentry(struct dentry * dentry)
1310{
1311 /* Is the task we represent dead?
1312 * If so, then don't put the dentry on the lru list,
1313 * kill it immediately.
1314 */
1315 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1316}
1317
1318static struct dentry_operations pid_dentry_operations =
1319{
1320 .d_revalidate = pid_revalidate,
1321 .d_delete = pid_delete_dentry,
1322};
1323
1324/* Lookups */
1325
Eric Dumazetc5141e62007-05-08 00:26:15 -07001326typedef struct dentry *instantiate_t(struct inode *, struct dentry *,
1327 struct task_struct *, const void *);
Eric W. Biederman61a28782006-10-02 02:18:49 -07001328
Eric W. Biederman1c0d04c2006-10-02 02:18:57 -07001329/*
1330 * Fill a directory entry.
1331 *
1332 * If possible create the dcache entry and derive our inode number and
1333 * file type from dcache entry.
1334 *
1335 * Since all of the proc inode numbers are dynamically generated, the inode
1336 * numbers do not exist until the inode is cache. This means creating the
1337 * the dcache entry in readdir is necessary to keep the inode numbers
1338 * reported by readdir in sync with the inode numbers reported
1339 * by stat.
1340 */
Eric W. Biederman61a28782006-10-02 02:18:49 -07001341static int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1342 char *name, int len,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001343 instantiate_t instantiate, struct task_struct *task, const void *ptr)
Eric W. Biederman61a28782006-10-02 02:18:49 -07001344{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001345 struct dentry *child, *dir = filp->f_path.dentry;
Eric W. Biederman61a28782006-10-02 02:18:49 -07001346 struct inode *inode;
1347 struct qstr qname;
1348 ino_t ino = 0;
1349 unsigned type = DT_UNKNOWN;
1350
1351 qname.name = name;
1352 qname.len = len;
1353 qname.hash = full_name_hash(name, len);
1354
1355 child = d_lookup(dir, &qname);
1356 if (!child) {
1357 struct dentry *new;
1358 new = d_alloc(dir, &qname);
1359 if (new) {
1360 child = instantiate(dir->d_inode, new, task, ptr);
1361 if (child)
1362 dput(new);
1363 else
1364 child = new;
1365 }
1366 }
1367 if (!child || IS_ERR(child) || !child->d_inode)
1368 goto end_instantiate;
1369 inode = child->d_inode;
1370 if (inode) {
1371 ino = inode->i_ino;
1372 type = inode->i_mode >> 12;
1373 }
1374 dput(child);
1375end_instantiate:
1376 if (!ino)
1377 ino = find_inode_number(dir, &qname);
1378 if (!ino)
1379 ino = 1;
1380 return filldir(dirent, name, len, filp->f_pos, ino, type);
1381}
1382
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001383static unsigned name_to_int(struct dentry *dentry)
1384{
1385 const char *name = dentry->d_name.name;
1386 int len = dentry->d_name.len;
1387 unsigned n = 0;
1388
1389 if (len > 1 && *name == '0')
1390 goto out;
1391 while (len-- > 0) {
1392 unsigned c = *name++ - '0';
1393 if (c > 9)
1394 goto out;
1395 if (n >= (~0U-9)/10)
1396 goto out;
1397 n *= 10;
1398 n += c;
1399 }
1400 return n;
1401out:
1402 return ~0U;
1403}
1404
Miklos Szeredi27932742007-05-08 00:26:17 -07001405#define PROC_FDINFO_MAX 64
1406
1407static int proc_fd_info(struct inode *inode, struct dentry **dentry,
1408 struct vfsmount **mnt, char *info)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001409{
1410 struct task_struct *task = get_proc_task(inode);
1411 struct files_struct *files = NULL;
1412 struct file *file;
1413 int fd = proc_fd(inode);
1414
1415 if (task) {
1416 files = get_files_struct(task);
1417 put_task_struct(task);
1418 }
1419 if (files) {
1420 /*
1421 * We are not taking a ref to the file structure, so we must
1422 * hold ->file_lock.
1423 */
1424 spin_lock(&files->file_lock);
1425 file = fcheck_files(files, fd);
1426 if (file) {
Miklos Szeredi27932742007-05-08 00:26:17 -07001427 if (mnt)
1428 *mnt = mntget(file->f_path.mnt);
1429 if (dentry)
1430 *dentry = dget(file->f_path.dentry);
1431 if (info)
1432 snprintf(info, PROC_FDINFO_MAX,
1433 "pos:\t%lli\n"
1434 "flags:\t0%o\n",
1435 (long long) file->f_pos,
1436 file->f_flags);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001437 spin_unlock(&files->file_lock);
1438 put_files_struct(files);
1439 return 0;
1440 }
1441 spin_unlock(&files->file_lock);
1442 put_files_struct(files);
1443 }
1444 return -ENOENT;
1445}
1446
Miklos Szeredi27932742007-05-08 00:26:17 -07001447static int proc_fd_link(struct inode *inode, struct dentry **dentry,
1448 struct vfsmount **mnt)
1449{
1450 return proc_fd_info(inode, dentry, mnt, NULL);
1451}
1452
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001453static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1454{
1455 struct inode *inode = dentry->d_inode;
1456 struct task_struct *task = get_proc_task(inode);
1457 int fd = proc_fd(inode);
1458 struct files_struct *files;
1459
1460 if (task) {
1461 files = get_files_struct(task);
1462 if (files) {
1463 rcu_read_lock();
1464 if (fcheck_files(files, fd)) {
1465 rcu_read_unlock();
1466 put_files_struct(files);
1467 if (task_dumpable(task)) {
1468 inode->i_uid = task->euid;
1469 inode->i_gid = task->egid;
1470 } else {
1471 inode->i_uid = 0;
1472 inode->i_gid = 0;
1473 }
1474 inode->i_mode &= ~(S_ISUID | S_ISGID);
1475 security_task_to_inode(task, inode);
1476 put_task_struct(task);
1477 return 1;
1478 }
1479 rcu_read_unlock();
1480 put_files_struct(files);
1481 }
1482 put_task_struct(task);
1483 }
1484 d_drop(dentry);
1485 return 0;
1486}
1487
1488static struct dentry_operations tid_fd_dentry_operations =
1489{
1490 .d_revalidate = tid_fd_revalidate,
1491 .d_delete = pid_delete_dentry,
1492};
1493
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001494static struct dentry *proc_fd_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001495 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001496{
Eric Dumazetc5141e62007-05-08 00:26:15 -07001497 unsigned fd = *(const unsigned *)ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001498 struct file *file;
1499 struct files_struct *files;
1500 struct inode *inode;
1501 struct proc_inode *ei;
1502 struct dentry *error = ERR_PTR(-ENOENT);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001503
Eric W. Biederman61a28782006-10-02 02:18:49 -07001504 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001505 if (!inode)
1506 goto out;
1507 ei = PROC_I(inode);
1508 ei->fd = fd;
1509 files = get_files_struct(task);
1510 if (!files)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001511 goto out_iput;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001512 inode->i_mode = S_IFLNK;
1513
1514 /*
1515 * We are not taking a ref to the file structure, so we must
1516 * hold ->file_lock.
1517 */
1518 spin_lock(&files->file_lock);
1519 file = fcheck_files(files, fd);
1520 if (!file)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001521 goto out_unlock;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001522 if (file->f_mode & 1)
1523 inode->i_mode |= S_IRUSR | S_IXUSR;
1524 if (file->f_mode & 2)
1525 inode->i_mode |= S_IWUSR | S_IXUSR;
1526 spin_unlock(&files->file_lock);
1527 put_files_struct(files);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001528
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001529 inode->i_op = &proc_pid_link_inode_operations;
1530 inode->i_size = 64;
1531 ei->op.proc_get_link = proc_fd_link;
1532 dentry->d_op = &tid_fd_dentry_operations;
1533 d_add(dentry, inode);
1534 /* Close the race of the process dying before we return the dentry */
1535 if (tid_fd_revalidate(dentry, NULL))
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001536 error = NULL;
1537
1538 out:
1539 return error;
1540out_unlock:
1541 spin_unlock(&files->file_lock);
1542 put_files_struct(files);
1543out_iput:
1544 iput(inode);
1545 goto out;
1546}
1547
Miklos Szeredi27932742007-05-08 00:26:17 -07001548static struct dentry *proc_lookupfd_common(struct inode *dir,
1549 struct dentry *dentry,
1550 instantiate_t instantiate)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001551{
1552 struct task_struct *task = get_proc_task(dir);
1553 unsigned fd = name_to_int(dentry);
1554 struct dentry *result = ERR_PTR(-ENOENT);
1555
1556 if (!task)
1557 goto out_no_task;
1558 if (fd == ~0U)
1559 goto out;
1560
Miklos Szeredi27932742007-05-08 00:26:17 -07001561 result = instantiate(dir, dentry, task, &fd);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001562out:
1563 put_task_struct(task);
1564out_no_task:
1565 return result;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001566}
1567
Miklos Szeredi27932742007-05-08 00:26:17 -07001568static int proc_readfd_common(struct file * filp, void * dirent,
1569 filldir_t filldir, instantiate_t instantiate)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001571 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman56347082006-06-26 00:25:40 -07001572 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001573 struct task_struct *p = get_proc_task(inode);
Pavel Emelyanov457c2512007-10-18 23:40:43 -07001574 unsigned int fd, ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 struct files_struct * files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001577 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578
1579 retval = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001580 if (!p)
1581 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583
1584 fd = filp->f_pos;
1585 switch (fd) {
1586 case 0:
1587 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1588 goto out;
1589 filp->f_pos++;
1590 case 1:
Eric W. Biederman56347082006-06-26 00:25:40 -07001591 ino = parent_ino(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1593 goto out;
1594 filp->f_pos++;
1595 default:
1596 files = get_files_struct(p);
1597 if (!files)
1598 goto out;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001599 rcu_read_lock();
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001600 fdt = files_fdtable(files);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 for (fd = filp->f_pos-2;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001602 fd < fdt->max_fds;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 fd++, filp->f_pos++) {
Miklos Szeredi27932742007-05-08 00:26:17 -07001604 char name[PROC_NUMBUF];
1605 int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606
1607 if (!fcheck_files(files, fd))
1608 continue;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001609 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610
Miklos Szeredi27932742007-05-08 00:26:17 -07001611 len = snprintf(name, sizeof(name), "%d", fd);
1612 if (proc_fill_cache(filp, dirent, filldir,
1613 name, len, instantiate,
1614 p, &fd) < 0) {
Dipankar Sarmab8359962005-09-09 13:04:14 -07001615 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616 break;
1617 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07001618 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07001620 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621 put_files_struct(files);
1622 }
1623out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07001624 put_task_struct(p);
1625out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626 return retval;
1627}
1628
Miklos Szeredi27932742007-05-08 00:26:17 -07001629static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1630 struct nameidata *nd)
1631{
1632 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1633}
1634
1635static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1636{
1637 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1638}
1639
1640static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1641 size_t len, loff_t *ppos)
1642{
1643 char tmp[PROC_FDINFO_MAX];
1644 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, NULL, tmp);
1645 if (!err)
1646 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1647 return err;
1648}
1649
1650static const struct file_operations proc_fdinfo_file_operations = {
1651 .open = nonseekable_open,
1652 .read = proc_fdinfo_read,
1653};
1654
Arjan van de Ven00977a52007-02-12 00:55:34 -08001655static const struct file_operations proc_fd_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656 .read = generic_read_dir,
1657 .readdir = proc_readfd,
1658};
1659
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660/*
Alexey Dobriyan8948e112007-05-08 00:23:35 -07001661 * /proc/pid/fd needs a special permission handler so that a process can still
1662 * access /proc/self/fd after it has executed a setuid().
1663 */
1664static int proc_fd_permission(struct inode *inode, int mask,
1665 struct nameidata *nd)
1666{
1667 int rv;
1668
1669 rv = generic_permission(inode, mask, NULL);
1670 if (rv == 0)
1671 return 0;
1672 if (task_pid(current) == proc_pid(inode))
1673 rv = 0;
1674 return rv;
1675}
1676
1677/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 * proc directories can do almost nothing..
1679 */
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08001680static const struct inode_operations proc_fd_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681 .lookup = proc_lookupfd,
Alexey Dobriyan8948e112007-05-08 00:23:35 -07001682 .permission = proc_fd_permission,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001683 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684};
1685
Miklos Szeredi27932742007-05-08 00:26:17 -07001686static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
1687 struct dentry *dentry, struct task_struct *task, const void *ptr)
1688{
1689 unsigned fd = *(unsigned *)ptr;
1690 struct inode *inode;
1691 struct proc_inode *ei;
1692 struct dentry *error = ERR_PTR(-ENOENT);
1693
1694 inode = proc_pid_make_inode(dir->i_sb, task);
1695 if (!inode)
1696 goto out;
1697 ei = PROC_I(inode);
1698 ei->fd = fd;
1699 inode->i_mode = S_IFREG | S_IRUSR;
1700 inode->i_fop = &proc_fdinfo_file_operations;
1701 dentry->d_op = &tid_fd_dentry_operations;
1702 d_add(dentry, inode);
1703 /* Close the race of the process dying before we return the dentry */
1704 if (tid_fd_revalidate(dentry, NULL))
1705 error = NULL;
1706
1707 out:
1708 return error;
1709}
1710
1711static struct dentry *proc_lookupfdinfo(struct inode *dir,
1712 struct dentry *dentry,
1713 struct nameidata *nd)
1714{
1715 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
1716}
1717
1718static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
1719{
1720 return proc_readfd_common(filp, dirent, filldir,
1721 proc_fdinfo_instantiate);
1722}
1723
1724static const struct file_operations proc_fdinfo_operations = {
1725 .read = generic_read_dir,
1726 .readdir = proc_readfdinfo,
1727};
1728
1729/*
1730 * proc directories can do almost nothing..
1731 */
1732static const struct inode_operations proc_fdinfo_inode_operations = {
1733 .lookup = proc_lookupfdinfo,
1734 .setattr = proc_setattr,
1735};
1736
1737
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001738static struct dentry *proc_pident_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001739 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001740{
Eric Dumazetc5141e62007-05-08 00:26:15 -07001741 const struct pid_entry *p = ptr;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001742 struct inode *inode;
1743 struct proc_inode *ei;
1744 struct dentry *error = ERR_PTR(-EINVAL);
1745
Eric W. Biederman61a28782006-10-02 02:18:49 -07001746 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001747 if (!inode)
1748 goto out;
1749
1750 ei = PROC_I(inode);
1751 inode->i_mode = p->mode;
1752 if (S_ISDIR(inode->i_mode))
1753 inode->i_nlink = 2; /* Use getattr to fix if necessary */
1754 if (p->iop)
1755 inode->i_op = p->iop;
1756 if (p->fop)
1757 inode->i_fop = p->fop;
1758 ei->op = p->op;
1759 dentry->d_op = &pid_dentry_operations;
1760 d_add(dentry, inode);
1761 /* Close the race of the process dying before we return the dentry */
1762 if (pid_revalidate(dentry, NULL))
1763 error = NULL;
1764out:
1765 return error;
1766}
1767
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768static struct dentry *proc_pident_lookup(struct inode *dir,
1769 struct dentry *dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001770 const struct pid_entry *ents,
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07001771 unsigned int nents)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772{
1773 struct inode *inode;
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001774 struct dentry *error;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001775 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07001776 const struct pid_entry *p, *last;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001778 error = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779 inode = NULL;
1780
Eric W. Biederman99f89552006-06-26 00:25:55 -07001781 if (!task)
1782 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783
Eric W. Biederman20cdc892006-10-02 02:17:07 -07001784 /*
1785 * Yes, it does not scale. And it should not. Don't add
1786 * new entries into /proc/<tgid>/ without very good reasons.
1787 */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07001788 last = &ents[nents - 1];
1789 for (p = ents; p <= last; p++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 if (p->len != dentry->d_name.len)
1791 continue;
1792 if (!memcmp(dentry->d_name.name, p->name, p->len))
1793 break;
1794 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07001795 if (p > last)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 goto out;
1797
Eric W. Biederman444ceed2006-10-02 02:18:49 -07001798 error = proc_pident_instantiate(dir, dentry, task, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07001800 put_task_struct(task);
1801out_no_task:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001802 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803}
1804
Eric Dumazetc5141e62007-05-08 00:26:15 -07001805static int proc_pident_fill_cache(struct file *filp, void *dirent,
1806 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07001807{
1808 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
1809 proc_pident_instantiate, task, p);
1810}
1811
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001812static int proc_pident_readdir(struct file *filp,
1813 void *dirent, filldir_t filldir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07001814 const struct pid_entry *ents, unsigned int nents)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001815{
1816 int i;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001817 struct dentry *dentry = filp->f_path.dentry;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001818 struct inode *inode = dentry->d_inode;
1819 struct task_struct *task = get_proc_task(inode);
Eric Dumazetc5141e62007-05-08 00:26:15 -07001820 const struct pid_entry *p, *last;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001821 ino_t ino;
1822 int ret;
1823
1824 ret = -ENOENT;
1825 if (!task)
Eric W. Biederman61a28782006-10-02 02:18:49 -07001826 goto out_no_task;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001827
1828 ret = 0;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001829 i = filp->f_pos;
1830 switch (i) {
1831 case 0:
1832 ino = inode->i_ino;
1833 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1834 goto out;
1835 i++;
1836 filp->f_pos++;
1837 /* fall through */
1838 case 1:
1839 ino = parent_ino(dentry);
1840 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1841 goto out;
1842 i++;
1843 filp->f_pos++;
1844 /* fall through */
1845 default:
1846 i -= 2;
1847 if (i >= nents) {
1848 ret = 1;
1849 goto out;
1850 }
1851 p = ents + i;
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07001852 last = &ents[nents - 1];
1853 while (p <= last) {
Eric W. Biederman61a28782006-10-02 02:18:49 -07001854 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001855 goto out;
1856 filp->f_pos++;
1857 p++;
1858 }
1859 }
1860
1861 ret = 1;
1862out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07001863 put_task_struct(task);
1864out_no_task:
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001865 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866}
1867
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868#ifdef CONFIG_SECURITY
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001869static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1870 size_t count, loff_t *ppos)
1871{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001872 struct inode * inode = file->f_path.dentry->d_inode;
Al Viro04ff9702007-03-12 16:17:58 +00001873 char *p = NULL;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001874 ssize_t length;
1875 struct task_struct *task = get_proc_task(inode);
1876
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001877 if (!task)
Al Viro04ff9702007-03-12 16:17:58 +00001878 return -ESRCH;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001879
1880 length = security_getprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001881 (char*)file->f_path.dentry->d_name.name,
Al Viro04ff9702007-03-12 16:17:58 +00001882 &p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001883 put_task_struct(task);
Al Viro04ff9702007-03-12 16:17:58 +00001884 if (length > 0)
1885 length = simple_read_from_buffer(buf, count, ppos, p, length);
1886 kfree(p);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001887 return length;
1888}
1889
1890static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1891 size_t count, loff_t *ppos)
1892{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001893 struct inode * inode = file->f_path.dentry->d_inode;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001894 char *page;
1895 ssize_t length;
1896 struct task_struct *task = get_proc_task(inode);
1897
1898 length = -ESRCH;
1899 if (!task)
1900 goto out_no_task;
1901 if (count > PAGE_SIZE)
1902 count = PAGE_SIZE;
1903
1904 /* No partial writes. */
1905 length = -EINVAL;
1906 if (*ppos != 0)
1907 goto out;
1908
1909 length = -ENOMEM;
Mel Gormane12ba742007-10-16 01:25:52 -07001910 page = (char*)__get_free_page(GFP_TEMPORARY);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001911 if (!page)
1912 goto out;
1913
1914 length = -EFAULT;
1915 if (copy_from_user(page, buf, count))
1916 goto out_free;
1917
1918 length = security_setprocattr(task,
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08001919 (char*)file->f_path.dentry->d_name.name,
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001920 (void*)page, count);
1921out_free:
1922 free_page((unsigned long) page);
1923out:
1924 put_task_struct(task);
1925out_no_task:
1926 return length;
1927}
1928
Arjan van de Ven00977a52007-02-12 00:55:34 -08001929static const struct file_operations proc_pid_attr_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001930 .read = proc_pid_attr_read,
1931 .write = proc_pid_attr_write,
1932};
1933
Eric Dumazetc5141e62007-05-08 00:26:15 -07001934static const struct pid_entry attr_dir_stuff[] = {
Eric W. Biederman61a28782006-10-02 02:18:49 -07001935 REG("current", S_IRUGO|S_IWUGO, pid_attr),
1936 REG("prev", S_IRUGO, pid_attr),
1937 REG("exec", S_IRUGO|S_IWUGO, pid_attr),
1938 REG("fscreate", S_IRUGO|S_IWUGO, pid_attr),
1939 REG("keycreate", S_IRUGO|S_IWUGO, pid_attr),
1940 REG("sockcreate", S_IRUGO|S_IWUGO, pid_attr),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07001941};
1942
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07001943static int proc_attr_dir_readdir(struct file * filp,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944 void * dirent, filldir_t filldir)
1945{
1946 return proc_pident_readdir(filp,dirent,filldir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07001947 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948}
1949
Arjan van de Ven00977a52007-02-12 00:55:34 -08001950static const struct file_operations proc_attr_dir_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951 .read = generic_read_dir,
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07001952 .readdir = proc_attr_dir_readdir,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953};
1954
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07001955static struct dentry *proc_attr_dir_lookup(struct inode *dir,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956 struct dentry *dentry, struct nameidata *nd)
1957{
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07001958 return proc_pident_lookup(dir, dentry,
1959 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960}
1961
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08001962static const struct inode_operations proc_attr_dir_inode_operations = {
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07001963 .lookup = proc_attr_dir_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07001964 .getattr = pid_getattr,
Linus Torvalds6d76fa52006-07-15 12:26:45 -07001965 .setattr = proc_setattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966};
1967
Linus Torvalds1da177e2005-04-16 15:20:36 -07001968#endif
1969
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07001970#if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
1971static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
1972 size_t count, loff_t *ppos)
1973{
1974 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1975 struct mm_struct *mm;
1976 char buffer[PROC_NUMBUF];
1977 size_t len;
1978 int ret;
1979
1980 if (!task)
1981 return -ESRCH;
1982
1983 ret = 0;
1984 mm = get_task_mm(task);
1985 if (mm) {
1986 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
1987 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
1988 MMF_DUMP_FILTER_SHIFT));
1989 mmput(mm);
1990 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
1991 }
1992
1993 put_task_struct(task);
1994
1995 return ret;
1996}
1997
1998static ssize_t proc_coredump_filter_write(struct file *file,
1999 const char __user *buf,
2000 size_t count,
2001 loff_t *ppos)
2002{
2003 struct task_struct *task;
2004 struct mm_struct *mm;
2005 char buffer[PROC_NUMBUF], *end;
2006 unsigned int val;
2007 int ret;
2008 int i;
2009 unsigned long mask;
2010
2011 ret = -EFAULT;
2012 memset(buffer, 0, sizeof(buffer));
2013 if (count > sizeof(buffer) - 1)
2014 count = sizeof(buffer) - 1;
2015 if (copy_from_user(buffer, buf, count))
2016 goto out_no_task;
2017
2018 ret = -EINVAL;
2019 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2020 if (*end == '\n')
2021 end++;
2022 if (end - buffer == 0)
2023 goto out_no_task;
2024
2025 ret = -ESRCH;
2026 task = get_proc_task(file->f_dentry->d_inode);
2027 if (!task)
2028 goto out_no_task;
2029
2030 ret = end - buffer;
2031 mm = get_task_mm(task);
2032 if (!mm)
2033 goto out_no_mm;
2034
2035 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2036 if (val & mask)
2037 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2038 else
2039 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2040 }
2041
2042 mmput(mm);
2043 out_no_mm:
2044 put_task_struct(task);
2045 out_no_task:
2046 return ret;
2047}
2048
2049static const struct file_operations proc_coredump_filter_operations = {
2050 .read = proc_coredump_filter_read,
2051 .write = proc_coredump_filter_write,
2052};
2053#endif
2054
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055/*
2056 * /proc/self:
2057 */
2058static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2059 int buflen)
2060{
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002061 char tmp[PROC_NUMBUF];
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002062 sprintf(tmp, "%d", task_tgid_vnr(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063 return vfs_readlink(dentry,buffer,buflen,tmp);
2064}
2065
Al Viro008b1502005-08-20 00:17:39 +01002066static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002067{
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002068 char tmp[PROC_NUMBUF];
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002069 sprintf(tmp, "%d", task_tgid_vnr(current));
Al Viro008b1502005-08-20 00:17:39 +01002070 return ERR_PTR(vfs_follow_link(nd,tmp));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002071}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002073static const struct inode_operations proc_self_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074 .readlink = proc_self_readlink,
2075 .follow_link = proc_self_follow_link,
2076};
2077
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002078/*
Eric W. Biederman801199c2006-10-02 02:18:48 -07002079 * proc base
2080 *
2081 * These are the directory entries in the root directory of /proc
2082 * that properly belong to the /proc filesystem, as they describe
2083 * describe something that is process related.
2084 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07002085static const struct pid_entry proc_base_stuff[] = {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002086 NOD("self", S_IFLNK|S_IRWXUGO,
Eric W. Biederman801199c2006-10-02 02:18:48 -07002087 &proc_self_inode_operations, NULL, {}),
Eric W. Biederman801199c2006-10-02 02:18:48 -07002088};
2089
2090/*
2091 * Exceptional case: normally we are not allowed to unhash a busy
2092 * directory. In this case, however, we can do it - no aliasing problems
2093 * due to the way we treat inodes.
2094 */
2095static int proc_base_revalidate(struct dentry *dentry, struct nameidata *nd)
2096{
2097 struct inode *inode = dentry->d_inode;
2098 struct task_struct *task = get_proc_task(inode);
2099 if (task) {
2100 put_task_struct(task);
2101 return 1;
2102 }
2103 d_drop(dentry);
2104 return 0;
2105}
2106
2107static struct dentry_operations proc_base_dentry_operations =
2108{
2109 .d_revalidate = proc_base_revalidate,
2110 .d_delete = pid_delete_dentry,
2111};
2112
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002113static struct dentry *proc_base_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002114 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman801199c2006-10-02 02:18:48 -07002115{
Eric Dumazetc5141e62007-05-08 00:26:15 -07002116 const struct pid_entry *p = ptr;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002117 struct inode *inode;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002118 struct proc_inode *ei;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002119 struct dentry *error = ERR_PTR(-EINVAL);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002120
2121 /* Allocate the inode */
2122 error = ERR_PTR(-ENOMEM);
2123 inode = new_inode(dir->i_sb);
2124 if (!inode)
2125 goto out;
2126
2127 /* Initialize the inode */
2128 ei = PROC_I(inode);
2129 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
Eric W. Biederman801199c2006-10-02 02:18:48 -07002130
2131 /*
2132 * grab the reference to the task.
2133 */
Oleg Nesterov1a657f782006-10-02 02:18:59 -07002134 ei->pid = get_task_pid(task, PIDTYPE_PID);
Eric W. Biederman801199c2006-10-02 02:18:48 -07002135 if (!ei->pid)
2136 goto out_iput;
2137
2138 inode->i_uid = 0;
2139 inode->i_gid = 0;
2140 inode->i_mode = p->mode;
2141 if (S_ISDIR(inode->i_mode))
2142 inode->i_nlink = 2;
2143 if (S_ISLNK(inode->i_mode))
2144 inode->i_size = 64;
2145 if (p->iop)
2146 inode->i_op = p->iop;
2147 if (p->fop)
2148 inode->i_fop = p->fop;
2149 ei->op = p->op;
2150 dentry->d_op = &proc_base_dentry_operations;
2151 d_add(dentry, inode);
2152 error = NULL;
2153out:
Eric W. Biederman801199c2006-10-02 02:18:48 -07002154 return error;
2155out_iput:
2156 iput(inode);
2157 goto out;
2158}
2159
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002160static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2161{
2162 struct dentry *error;
2163 struct task_struct *task = get_proc_task(dir);
Eric Dumazetc5141e62007-05-08 00:26:15 -07002164 const struct pid_entry *p, *last;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002165
2166 error = ERR_PTR(-ENOENT);
2167
2168 if (!task)
2169 goto out_no_task;
2170
2171 /* Lookup the directory entry */
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002172 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2173 for (p = proc_base_stuff; p <= last; p++) {
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002174 if (p->len != dentry->d_name.len)
2175 continue;
2176 if (!memcmp(dentry->d_name.name, p->name, p->len))
2177 break;
2178 }
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002179 if (p > last)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002180 goto out;
2181
2182 error = proc_base_instantiate(dir, dentry, task, p);
2183
2184out:
2185 put_task_struct(task);
2186out_no_task:
2187 return error;
2188}
2189
Eric Dumazetc5141e62007-05-08 00:26:15 -07002190static int proc_base_fill_cache(struct file *filp, void *dirent,
2191 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002192{
2193 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2194 proc_base_instantiate, task, p);
2195}
2196
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002197#ifdef CONFIG_TASK_IO_ACCOUNTING
2198static int proc_pid_io_accounting(struct task_struct *task, char *buffer)
2199{
2200 return sprintf(buffer,
Alexey Dobriyan4b98d112007-02-10 01:46:45 -08002201#ifdef CONFIG_TASK_XACCT
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002202 "rchar: %llu\n"
2203 "wchar: %llu\n"
2204 "syscr: %llu\n"
2205 "syscw: %llu\n"
Alexey Dobriyan4b98d112007-02-10 01:46:45 -08002206#endif
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002207 "read_bytes: %llu\n"
2208 "write_bytes: %llu\n"
2209 "cancelled_write_bytes: %llu\n",
Alexey Dobriyan4b98d112007-02-10 01:46:45 -08002210#ifdef CONFIG_TASK_XACCT
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002211 (unsigned long long)task->rchar,
2212 (unsigned long long)task->wchar,
2213 (unsigned long long)task->syscr,
2214 (unsigned long long)task->syscw,
Alexey Dobriyan4b98d112007-02-10 01:46:45 -08002215#endif
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002216 (unsigned long long)task->ioac.read_bytes,
2217 (unsigned long long)task->ioac.write_bytes,
2218 (unsigned long long)task->ioac.cancelled_write_bytes);
2219}
2220#endif
2221
Eric W. Biederman801199c2006-10-02 02:18:48 -07002222/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002223 * Thread groups
2224 */
Arjan van de Ven00977a52007-02-12 00:55:34 -08002225static const struct file_operations proc_task_operations;
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002226static const struct inode_operations proc_task_inode_operations;
Eric W. Biederman20cdc892006-10-02 02:17:07 -07002227
Eric Dumazetc5141e62007-05-08 00:26:15 -07002228static const struct pid_entry tgid_base_stuff[] = {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002229 DIR("task", S_IRUGO|S_IXUGO, task),
2230 DIR("fd", S_IRUSR|S_IXUSR, fd),
Miklos Szeredi27932742007-05-08 00:26:17 -07002231 DIR("fdinfo", S_IRUSR|S_IXUSR, fdinfo),
James Pearson315e28c2007-10-16 23:30:17 -07002232 REG("environ", S_IRUSR, environ),
Eric W. Biederman61a28782006-10-02 02:18:49 -07002233 INF("auxv", S_IRUSR, pid_auxv),
2234 INF("status", S_IRUGO, pid_status),
Neil Hormand85f50d2007-10-18 23:40:37 -07002235 INF("limits", S_IRUSR, pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002236#ifdef CONFIG_SCHED_DEBUG
2237 REG("sched", S_IRUGO|S_IWUSR, pid_sched),
2238#endif
Eric W. Biederman61a28782006-10-02 02:18:49 -07002239 INF("cmdline", S_IRUGO, pid_cmdline),
2240 INF("stat", S_IRUGO, tgid_stat),
2241 INF("statm", S_IRUGO, pid_statm),
2242 REG("maps", S_IRUGO, maps),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002243#ifdef CONFIG_NUMA
Eric W. Biederman61a28782006-10-02 02:18:49 -07002244 REG("numa_maps", S_IRUGO, numa_maps),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002245#endif
Eric W. Biederman61a28782006-10-02 02:18:49 -07002246 REG("mem", S_IRUSR|S_IWUSR, mem),
Eric W. Biederman61a28782006-10-02 02:18:49 -07002247 LNK("cwd", cwd),
2248 LNK("root", root),
2249 LNK("exe", exe),
2250 REG("mounts", S_IRUGO, mounts),
2251 REG("mountstats", S_IRUSR, mountstats),
Matt Mackall1e883282008-02-04 22:29:07 -08002252#ifdef CONFIG_PROC_PAGE_MONITOR
David Rientjesb813e932007-05-06 14:49:24 -07002253 REG("clear_refs", S_IWUSR, clear_refs),
Eric W. Biederman61a28782006-10-02 02:18:49 -07002254 REG("smaps", S_IRUGO, smaps),
Matt Mackall85863e42008-02-04 22:29:04 -08002255 REG("pagemap", S_IRUSR, pagemap),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002256#endif
2257#ifdef CONFIG_SECURITY
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002258 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002259#endif
2260#ifdef CONFIG_KALLSYMS
Eric W. Biederman61a28782006-10-02 02:18:49 -07002261 INF("wchan", S_IRUGO, pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002262#endif
2263#ifdef CONFIG_SCHEDSTATS
Eric W. Biederman61a28782006-10-02 02:18:49 -07002264 INF("schedstat", S_IRUGO, pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002265#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01002266#ifdef CONFIG_LATENCYTOP
2267 REG("latency", S_IRUGO, lstats),
2268#endif
Paul Menage8793d852007-10-18 23:39:39 -07002269#ifdef CONFIG_PROC_PID_CPUSET
Eric W. Biederman61a28782006-10-02 02:18:49 -07002270 REG("cpuset", S_IRUGO, cpuset),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002271#endif
Paul Menagea4243162007-10-18 23:39:35 -07002272#ifdef CONFIG_CGROUPS
2273 REG("cgroup", S_IRUGO, cgroup),
2274#endif
Eric W. Biederman61a28782006-10-02 02:18:49 -07002275 INF("oom_score", S_IRUGO, oom_score),
2276 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002277#ifdef CONFIG_AUDITSYSCALL
Eric W. Biederman61a28782006-10-02 02:18:49 -07002278 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002279#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002280#ifdef CONFIG_FAULT_INJECTION
2281 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2282#endif
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -07002283#if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
2284 REG("coredump_filter", S_IRUGO|S_IWUSR, coredump_filter),
2285#endif
Andrew Mortonaba76fd2006-12-10 02:19:48 -08002286#ifdef CONFIG_TASK_IO_ACCOUNTING
2287 INF("io", S_IRUGO, pid_io_accounting),
2288#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002289};
2290
2291static int proc_tgid_base_readdir(struct file * filp,
2292 void * dirent, filldir_t filldir)
2293{
2294 return proc_pident_readdir(filp,dirent,filldir,
2295 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
2296}
2297
Arjan van de Ven00977a52007-02-12 00:55:34 -08002298static const struct file_operations proc_tgid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002299 .read = generic_read_dir,
2300 .readdir = proc_tgid_base_readdir,
2301};
2302
2303static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002304 return proc_pident_lookup(dir, dentry,
2305 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002306}
2307
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002308static const struct inode_operations proc_tgid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002309 .lookup = proc_tgid_base_lookup,
2310 .getattr = pid_getattr,
2311 .setattr = proc_setattr,
2312};
2313
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002314static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002315{
Eric W. Biederman48e64842006-06-26 00:25:48 -07002316 struct dentry *dentry, *leader, *dir;
Eric W. Biederman8578cea2006-06-26 00:25:54 -07002317 char buf[PROC_NUMBUF];
Eric W. Biederman48e64842006-06-26 00:25:48 -07002318 struct qstr name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319
Eric W. Biederman48e64842006-06-26 00:25:48 -07002320 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002321 name.len = snprintf(buf, sizeof(buf), "%d", pid);
2322 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002323 if (dentry) {
Andrea Arcangeli77667552008-02-04 22:29:21 -08002324 if (!(current->flags & PF_EXITING))
2325 shrink_dcache_parent(dentry);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002326 d_drop(dentry);
2327 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002328 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002330 if (tgid == 0)
Eric W. Biederman48e64842006-06-26 00:25:48 -07002331 goto out;
2332
2333 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002334 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
2335 leader = d_hash_and_lookup(mnt->mnt_root, &name);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002336 if (!leader)
2337 goto out;
2338
2339 name.name = "task";
2340 name.len = strlen(name.name);
2341 dir = d_hash_and_lookup(leader, &name);
2342 if (!dir)
2343 goto out_put_leader;
2344
2345 name.name = buf;
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002346 name.len = snprintf(buf, sizeof(buf), "%d", pid);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002347 dentry = d_hash_and_lookup(dir, &name);
2348 if (dentry) {
2349 shrink_dcache_parent(dentry);
2350 d_drop(dentry);
2351 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002352 }
Eric W. Biederman48e64842006-06-26 00:25:48 -07002353
2354 dput(dir);
2355out_put_leader:
2356 dput(leader);
2357out:
2358 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002359}
2360
Randy Dunlap0895e912007-10-21 21:00:10 -07002361/**
2362 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
2363 * @task: task that should be flushed.
2364 *
2365 * When flushing dentries from proc, one needs to flush them from global
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002366 * proc (proc_mnt) and from all the namespaces' procs this task was seen
Randy Dunlap0895e912007-10-21 21:00:10 -07002367 * in. This call is supposed to do all of this job.
2368 *
2369 * Looks in the dcache for
2370 * /proc/@pid
2371 * /proc/@tgid/task/@pid
2372 * if either directory is present flushes it and all of it'ts children
2373 * from the dcache.
2374 *
2375 * It is safe and reasonable to cache /proc entries for a task until
2376 * that task exits. After that they just clog up the dcache with
2377 * useless entries, possibly causing useful dcache entries to be
2378 * flushed instead. This routine is proved to flush those useless
2379 * dcache entries at process exit time.
2380 *
2381 * NOTE: This routine is just an optimization so it does not guarantee
2382 * that no dcache entries will exist at process exit time it
2383 * just makes it very unlikely that any will persist.
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002384 */
2385
2386void proc_flush_task(struct task_struct *task)
2387{
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002388 int i;
2389 struct pid *pid, *tgid = NULL;
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002390 struct upid *upid;
2391
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002392 pid = task_pid(task);
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002393 if (thread_group_leader(task))
2394 tgid = task_tgid(task);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002395
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002396 for (i = 0; i <= pid->level; i++) {
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002397 upid = &pid->numbers[i];
2398 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
Eric W. Biederman9fcc2d12007-11-14 17:00:07 -08002399 tgid ? tgid->numbers[i].nr : 0);
Pavel Emelyanov130f77e2007-10-18 23:40:11 -07002400 }
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -07002401
2402 upid = &pid->numbers[pid->level];
2403 if (upid->nr == 1)
2404 pid_ns_release_proc(upid->ns);
Pavel Emelyanov60347f62007-10-18 23:40:03 -07002405}
2406
Adrian Bunk9711ef92006-12-06 20:38:31 -08002407static struct dentry *proc_pid_instantiate(struct inode *dir,
2408 struct dentry * dentry,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002409 struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002410{
2411 struct dentry *error = ERR_PTR(-ENOENT);
2412 struct inode *inode;
2413
Eric W. Biederman61a28782006-10-02 02:18:49 -07002414 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002415 if (!inode)
2416 goto out;
2417
2418 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2419 inode->i_op = &proc_tgid_base_inode_operations;
2420 inode->i_fop = &proc_tgid_base_operations;
2421 inode->i_flags|=S_IMMUTABLE;
Miklos Szeredi27932742007-05-08 00:26:17 -07002422 inode->i_nlink = 5;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002423#ifdef CONFIG_SECURITY
2424 inode->i_nlink += 1;
2425#endif
2426
2427 dentry->d_op = &pid_dentry_operations;
2428
2429 d_add(dentry, inode);
2430 /* Close the race of the process dying before we return the dentry */
2431 if (pid_revalidate(dentry, NULL))
2432 error = NULL;
2433out:
2434 return error;
2435}
2436
Linus Torvalds1da177e2005-04-16 15:20:36 -07002437struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2438{
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002439 struct dentry *result = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002440 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002441 unsigned tgid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002442 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002443
Eric W. Biederman801199c2006-10-02 02:18:48 -07002444 result = proc_base_lookup(dir, dentry);
2445 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
2446 goto out;
2447
Linus Torvalds1da177e2005-04-16 15:20:36 -07002448 tgid = name_to_int(dentry);
2449 if (tgid == ~0U)
2450 goto out;
2451
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002452 ns = dentry->d_sb->s_fs_info;
Eric W. Biedermande758732006-06-26 00:25:51 -07002453 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002454 task = find_task_by_pid_ns(tgid, ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002455 if (task)
2456 get_task_struct(task);
Eric W. Biedermande758732006-06-26 00:25:51 -07002457 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002458 if (!task)
2459 goto out;
2460
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002461 result = proc_pid_instantiate(dir, dentry, task, NULL);
Eric W. Biederman48e64842006-06-26 00:25:48 -07002462 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002463out:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002464 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002465}
2466
Linus Torvalds1da177e2005-04-16 15:20:36 -07002467/*
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002468 * Find the first task with tgid >= tgid
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002469 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002470 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002471struct tgid_iter {
2472 unsigned int tgid;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002473 struct task_struct *task;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002474};
2475static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
2476{
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002477 struct pid *pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002478
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002479 if (iter.task)
2480 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002481 rcu_read_lock();
2482retry:
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002483 iter.task = NULL;
2484 pid = find_ge_pid(iter.tgid, ns);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002485 if (pid) {
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002486 iter.tgid = pid_nr_ns(pid, ns);
2487 iter.task = pid_task(pid, PIDTYPE_PID);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002488 /* What we to know is if the pid we have find is the
2489 * pid of a thread_group_leader. Testing for task
2490 * being a thread_group_leader is the obvious thing
2491 * todo but there is a window when it fails, due to
2492 * the pid transfer logic in de_thread.
2493 *
2494 * So we perform the straight forward test of seeing
2495 * if the pid we have found is the pid of a thread
2496 * group leader, and don't worry if the task we have
2497 * found doesn't happen to be a thread group leader.
2498 * As we don't care in the case of readdir.
2499 */
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002500 if (!iter.task || !has_group_leader_pid(iter.task)) {
2501 iter.tgid += 1;
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002502 goto retry;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002503 }
2504 get_task_struct(iter.task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002505 }
Eric W. Biederman454cc102006-06-26 00:25:51 -07002506 rcu_read_unlock();
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002507 return iter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002508}
2509
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002510#define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002511
Eric W. Biederman61a28782006-10-02 02:18:49 -07002512static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002513 struct tgid_iter iter)
Eric W. Biederman61a28782006-10-02 02:18:49 -07002514{
2515 char name[PROC_NUMBUF];
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002516 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
Eric W. Biederman61a28782006-10-02 02:18:49 -07002517 return proc_fill_cache(filp, dirent, filldir, name, len,
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002518 proc_pid_instantiate, iter.task, NULL);
Eric W. Biederman61a28782006-10-02 02:18:49 -07002519}
2520
Linus Torvalds1da177e2005-04-16 15:20:36 -07002521/* for the /proc/ directory itself, after non-process stuff has been done */
2522int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2523{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002524 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002525 struct task_struct *reaper = get_proc_task(filp->f_path.dentry->d_inode);
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002526 struct tgid_iter iter;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002527 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002528
Eric W. Biederman61a28782006-10-02 02:18:49 -07002529 if (!reaper)
2530 goto out_no_task;
2531
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002532 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
Eric Dumazetc5141e62007-05-08 00:26:15 -07002533 const struct pid_entry *p = &proc_base_stuff[nr];
Eric W. Biederman61a28782006-10-02 02:18:49 -07002534 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
Eric W. Biederman801199c2006-10-02 02:18:48 -07002535 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002536 }
2537
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002538 ns = filp->f_dentry->d_sb->s_fs_info;
Eric W. Biederman19fd4bb2007-11-28 16:21:26 -08002539 iter.task = NULL;
2540 iter.tgid = filp->f_pos - TGID_OFFSET;
2541 for (iter = next_tgid(ns, iter);
2542 iter.task;
2543 iter.tgid += 1, iter = next_tgid(ns, iter)) {
2544 filp->f_pos = iter.tgid + TGID_OFFSET;
2545 if (proc_pid_fill_cache(filp, dirent, filldir, iter) < 0) {
2546 put_task_struct(iter.task);
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002547 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002548 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002549 }
Eric W. Biederman0804ef42006-10-02 02:17:04 -07002550 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
2551out:
Eric W. Biederman61a28782006-10-02 02:18:49 -07002552 put_task_struct(reaper);
2553out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002554 return 0;
2555}
2556
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002557/*
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002558 * Tasks
2559 */
Eric Dumazetc5141e62007-05-08 00:26:15 -07002560static const struct pid_entry tid_base_stuff[] = {
Eric W. Biederman61a28782006-10-02 02:18:49 -07002561 DIR("fd", S_IRUSR|S_IXUSR, fd),
Miklos Szeredi27932742007-05-08 00:26:17 -07002562 DIR("fdinfo", S_IRUSR|S_IXUSR, fdinfo),
James Pearson315e28c2007-10-16 23:30:17 -07002563 REG("environ", S_IRUSR, environ),
Eric W. Biederman61a28782006-10-02 02:18:49 -07002564 INF("auxv", S_IRUSR, pid_auxv),
2565 INF("status", S_IRUGO, pid_status),
Neil Hormand85f50d2007-10-18 23:40:37 -07002566 INF("limits", S_IRUSR, pid_limits),
Ingo Molnar43ae34c2007-07-09 18:52:00 +02002567#ifdef CONFIG_SCHED_DEBUG
2568 REG("sched", S_IRUGO|S_IWUSR, pid_sched),
2569#endif
Eric W. Biederman61a28782006-10-02 02:18:49 -07002570 INF("cmdline", S_IRUGO, pid_cmdline),
2571 INF("stat", S_IRUGO, tid_stat),
2572 INF("statm", S_IRUGO, pid_statm),
2573 REG("maps", S_IRUGO, maps),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002574#ifdef CONFIG_NUMA
Eric W. Biederman61a28782006-10-02 02:18:49 -07002575 REG("numa_maps", S_IRUGO, numa_maps),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002576#endif
Eric W. Biederman61a28782006-10-02 02:18:49 -07002577 REG("mem", S_IRUSR|S_IWUSR, mem),
Eric W. Biederman61a28782006-10-02 02:18:49 -07002578 LNK("cwd", cwd),
2579 LNK("root", root),
2580 LNK("exe", exe),
2581 REG("mounts", S_IRUGO, mounts),
Matt Mackall1e883282008-02-04 22:29:07 -08002582#ifdef CONFIG_PROC_PAGE_MONITOR
David Rientjesb813e932007-05-06 14:49:24 -07002583 REG("clear_refs", S_IWUSR, clear_refs),
Eric W. Biederman61a28782006-10-02 02:18:49 -07002584 REG("smaps", S_IRUGO, smaps),
Matt Mackall85863e42008-02-04 22:29:04 -08002585 REG("pagemap", S_IRUSR, pagemap),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002586#endif
2587#ifdef CONFIG_SECURITY
Eric W. Biederman72d9dcf2006-10-02 02:18:50 -07002588 DIR("attr", S_IRUGO|S_IXUGO, attr_dir),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002589#endif
2590#ifdef CONFIG_KALLSYMS
Eric W. Biederman61a28782006-10-02 02:18:49 -07002591 INF("wchan", S_IRUGO, pid_wchan),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002592#endif
2593#ifdef CONFIG_SCHEDSTATS
Eric W. Biederman61a28782006-10-02 02:18:49 -07002594 INF("schedstat", S_IRUGO, pid_schedstat),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002595#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01002596#ifdef CONFIG_LATENCYTOP
2597 REG("latency", S_IRUGO, lstats),
2598#endif
Paul Menage8793d852007-10-18 23:39:39 -07002599#ifdef CONFIG_PROC_PID_CPUSET
Eric W. Biederman61a28782006-10-02 02:18:49 -07002600 REG("cpuset", S_IRUGO, cpuset),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002601#endif
Paul Menagea4243162007-10-18 23:39:35 -07002602#ifdef CONFIG_CGROUPS
2603 REG("cgroup", S_IRUGO, cgroup),
2604#endif
Eric W. Biederman61a28782006-10-02 02:18:49 -07002605 INF("oom_score", S_IRUGO, oom_score),
2606 REG("oom_adj", S_IRUGO|S_IWUSR, oom_adjust),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002607#ifdef CONFIG_AUDITSYSCALL
Eric W. Biederman61a28782006-10-02 02:18:49 -07002608 REG("loginuid", S_IWUSR|S_IRUGO, loginuid),
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002609#endif
Akinobu Mitaf4f154f2006-12-08 02:39:47 -08002610#ifdef CONFIG_FAULT_INJECTION
2611 REG("make-it-fail", S_IRUGO|S_IWUSR, fault_inject),
2612#endif
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002613};
2614
2615static int proc_tid_base_readdir(struct file * filp,
2616 void * dirent, filldir_t filldir)
2617{
2618 return proc_pident_readdir(filp,dirent,filldir,
2619 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
2620}
2621
2622static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
Eric W. Biederman7bcd6b02006-10-02 02:18:56 -07002623 return proc_pident_lookup(dir, dentry,
2624 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002625}
2626
Arjan van de Ven00977a52007-02-12 00:55:34 -08002627static const struct file_operations proc_tid_base_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002628 .read = generic_read_dir,
2629 .readdir = proc_tid_base_readdir,
2630};
2631
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002632static const struct inode_operations proc_tid_base_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002633 .lookup = proc_tid_base_lookup,
2634 .getattr = pid_getattr,
2635 .setattr = proc_setattr,
2636};
2637
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002638static struct dentry *proc_task_instantiate(struct inode *dir,
Eric Dumazetc5141e62007-05-08 00:26:15 -07002639 struct dentry *dentry, struct task_struct *task, const void *ptr)
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002640{
2641 struct dentry *error = ERR_PTR(-ENOENT);
2642 struct inode *inode;
Eric W. Biederman61a28782006-10-02 02:18:49 -07002643 inode = proc_pid_make_inode(dir->i_sb, task);
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002644
2645 if (!inode)
2646 goto out;
2647 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2648 inode->i_op = &proc_tid_base_inode_operations;
2649 inode->i_fop = &proc_tid_base_operations;
2650 inode->i_flags|=S_IMMUTABLE;
Miklos Szeredi27932742007-05-08 00:26:17 -07002651 inode->i_nlink = 4;
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002652#ifdef CONFIG_SECURITY
2653 inode->i_nlink += 1;
2654#endif
2655
2656 dentry->d_op = &pid_dentry_operations;
2657
2658 d_add(dentry, inode);
2659 /* Close the race of the process dying before we return the dentry */
2660 if (pid_revalidate(dentry, NULL))
2661 error = NULL;
2662out:
2663 return error;
2664}
2665
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002666static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2667{
2668 struct dentry *result = ERR_PTR(-ENOENT);
2669 struct task_struct *task;
2670 struct task_struct *leader = get_proc_task(dir);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002671 unsigned tid;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002672 struct pid_namespace *ns;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002673
2674 if (!leader)
2675 goto out_no_task;
2676
2677 tid = name_to_int(dentry);
2678 if (tid == ~0U)
2679 goto out;
2680
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002681 ns = dentry->d_sb->s_fs_info;
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002682 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002683 task = find_task_by_pid_ns(tid, ns);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002684 if (task)
2685 get_task_struct(task);
2686 rcu_read_unlock();
2687 if (!task)
2688 goto out;
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07002689 if (!same_thread_group(leader, task))
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002690 goto out_drop_task;
2691
Eric W. Biederman444ceed2006-10-02 02:18:49 -07002692 result = proc_task_instantiate(dir, dentry, task, NULL);
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002693out_drop_task:
2694 put_task_struct(task);
2695out:
2696 put_task_struct(leader);
2697out_no_task:
2698 return result;
2699}
2700
2701/*
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002702 * Find the first tid of a thread group to return to user space.
2703 *
2704 * Usually this is just the thread group leader, but if the users
2705 * buffer was too small or there was a seek into the middle of the
2706 * directory we have more work todo.
2707 *
2708 * In the case of a short read we start with find_task_by_pid.
2709 *
2710 * In the case of a seek we start with the leader and walk nr
2711 * threads past it.
2712 */
Eric W. Biedermancc288732006-06-26 00:26:01 -07002713static struct task_struct *first_tid(struct task_struct *leader,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002714 int tid, int nr, struct pid_namespace *ns)
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002715{
Oleg Nesterova872ff02006-06-26 00:26:01 -07002716 struct task_struct *pos;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002717
Eric W. Biedermancc288732006-06-26 00:26:01 -07002718 rcu_read_lock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002719 /* Attempt to start with the pid of a thread */
2720 if (tid && (nr > 0)) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002721 pos = find_task_by_pid_ns(tid, ns);
Oleg Nesterova872ff02006-06-26 00:26:01 -07002722 if (pos && (pos->group_leader == leader))
2723 goto found;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002724 }
2725
2726 /* If nr exceeds the number of threads there is nothing todo */
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002727 pos = NULL;
Oleg Nesterova872ff02006-06-26 00:26:01 -07002728 if (nr && nr >= get_nr_threads(leader))
2729 goto out;
2730
2731 /* If we haven't found our starting place yet start
2732 * with the leader and walk nr threads forward.
2733 */
2734 for (pos = leader; nr > 0; --nr) {
2735 pos = next_thread(pos);
2736 if (pos == leader) {
2737 pos = NULL;
2738 goto out;
2739 }
2740 }
2741found:
2742 get_task_struct(pos);
2743out:
Eric W. Biedermancc288732006-06-26 00:26:01 -07002744 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002745 return pos;
2746}
2747
2748/*
2749 * Find the next thread in the thread list.
2750 * Return NULL if there is an error or no next thread.
2751 *
2752 * The reference to the input task_struct is released.
2753 */
2754static struct task_struct *next_tid(struct task_struct *start)
2755{
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07002756 struct task_struct *pos = NULL;
Eric W. Biedermancc288732006-06-26 00:26:01 -07002757 rcu_read_lock();
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07002758 if (pid_alive(start)) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002759 pos = next_thread(start);
Oleg Nesterovc1df7fb2006-06-26 00:26:02 -07002760 if (thread_group_leader(pos))
2761 pos = NULL;
2762 else
2763 get_task_struct(pos);
2764 }
Eric W. Biedermancc288732006-06-26 00:26:01 -07002765 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002766 put_task_struct(start);
2767 return pos;
2768}
2769
Eric W. Biederman61a28782006-10-02 02:18:49 -07002770static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
2771 struct task_struct *task, int tid)
2772{
2773 char name[PROC_NUMBUF];
2774 int len = snprintf(name, sizeof(name), "%d", tid);
2775 return proc_fill_cache(filp, dirent, filldir, name, len,
2776 proc_task_instantiate, task, NULL);
2777}
2778
Linus Torvalds1da177e2005-04-16 15:20:36 -07002779/* for the /proc/TGID/task/ directories */
2780static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2781{
Josef "Jeff" Sipek2fddfee2006-12-08 02:36:36 -08002782 struct dentry *dentry = filp->f_path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002783 struct inode *inode = dentry->d_inode;
Guillaume Chazarain7d895242007-01-31 23:48:14 -08002784 struct task_struct *leader = NULL;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002785 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002786 int retval = -ENOENT;
2787 ino_t ino;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002788 int tid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002789 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002790 struct pid_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002791
Guillaume Chazarain7d895242007-01-31 23:48:14 -08002792 task = get_proc_task(inode);
2793 if (!task)
2794 goto out_no_task;
2795 rcu_read_lock();
2796 if (pid_alive(task)) {
2797 leader = task->group_leader;
2798 get_task_struct(leader);
2799 }
2800 rcu_read_unlock();
2801 put_task_struct(task);
Eric W. Biederman99f89552006-06-26 00:25:55 -07002802 if (!leader)
2803 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002804 retval = 0;
2805
2806 switch (pos) {
2807 case 0:
2808 ino = inode->i_ino;
2809 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2810 goto out;
2811 pos++;
2812 /* fall through */
2813 case 1:
2814 ino = parent_ino(dentry);
2815 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2816 goto out;
2817 pos++;
2818 /* fall through */
2819 }
2820
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002821 /* f_version caches the tgid value that the last readdir call couldn't
2822 * return. lseek aka telldir automagically resets f_version to 0.
2823 */
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002824 ns = filp->f_dentry->d_sb->s_fs_info;
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07002825 tid = (int)filp->f_version;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002826 filp->f_version = 0;
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002827 for (task = first_tid(leader, tid, pos - 2, ns);
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002828 task;
2829 task = next_tid(task), pos++) {
Pavel Emelyanovb4888932007-10-18 23:40:14 -07002830 tid = task_pid_nr_ns(task, ns);
Eric W. Biederman61a28782006-10-02 02:18:49 -07002831 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002832 /* returning this tgid failed, save it as the first
2833 * pid for the next readir call */
Mathieu Desnoyers2b47c362007-10-16 23:27:21 -07002834 filp->f_version = (u64)tid;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002835 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002836 break;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002837 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002838 }
2839out:
2840 filp->f_pos = pos;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002841 put_task_struct(leader);
2842out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002843 return retval;
2844}
Eric W. Biederman6e66b522006-06-26 00:25:47 -07002845
2846static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2847{
2848 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002849 struct task_struct *p = get_proc_task(inode);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07002850 generic_fillattr(inode, stat);
2851
Eric W. Biederman99f89552006-06-26 00:25:55 -07002852 if (p) {
2853 rcu_read_lock();
2854 stat->nlink += get_nr_threads(p);
2855 rcu_read_unlock();
2856 put_task_struct(p);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07002857 }
2858
2859 return 0;
2860}
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002861
Arjan van de Venc5ef1c42007-02-12 00:55:40 -08002862static const struct inode_operations proc_task_inode_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002863 .lookup = proc_task_lookup,
2864 .getattr = proc_task_getattr,
2865 .setattr = proc_setattr,
2866};
2867
Arjan van de Ven00977a52007-02-12 00:55:34 -08002868static const struct file_operations proc_task_operations = {
Eric W. Biederman28a6d672006-10-02 02:17:05 -07002869 .read = generic_read_dir,
2870 .readdir = proc_task_readdir,
2871};