blob: 773469703c62b62c222490e624c1168d02fb5fdc [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
52#include <linux/config.h>
53#include <linux/errno.h>
54#include <linux/time.h>
55#include <linux/proc_fs.h>
56#include <linux/stat.h>
57#include <linux/init.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080058#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070059#include <linux/file.h>
60#include <linux/string.h>
61#include <linux/seq_file.h>
62#include <linux/namei.h>
63#include <linux/namespace.h>
64#include <linux/mm.h>
65#include <linux/smp_lock.h>
Dipankar Sarmab8359962005-09-09 13:04:14 -070066#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070067#include <linux/kallsyms.h>
68#include <linux/mount.h>
69#include <linux/security.h>
70#include <linux/ptrace.h>
71#include <linux/seccomp.h>
72#include <linux/cpuset.h>
73#include <linux/audit.h>
Al Viro5addc5d2005-11-07 17:15:49 -050074#include <linux/poll.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070075#include "internal.h"
76
Eric W. Biederman0f2fe202006-06-26 00:25:46 -070077/* NOTE:
78 * Implementing inode permission operations in /proc is almost
79 * certainly an error. Permission checks need to happen during
80 * each system call not at open time. The reason is that most of
81 * what we wish to check for permissions in /proc varies at runtime.
82 *
83 * The classic example of a problem is opening file descriptors
84 * in /proc for a task before it execs a suid executable.
85 */
86
Linus Torvalds1da177e2005-04-16 15:20:36 -070087/*
88 * For hysterical raisins we keep the same inumbers as in the old procfs.
89 * Feel free to change the macro below - just keep the range distinct from
90 * inumbers of the rest of procfs (currently those are in 0x0000--0xffff).
91 * As soon as we'll get a separate superblock we will be able to forget
92 * about magical ranges too.
93 */
94
95#define fake_ino(pid,ino) (((pid)<<16)|(ino))
96
97enum pid_directory_inos {
98 PROC_TGID_INO = 2,
99 PROC_TGID_TASK,
100 PROC_TGID_STATUS,
101 PROC_TGID_MEM,
102#ifdef CONFIG_SECCOMP
103 PROC_TGID_SECCOMP,
104#endif
105 PROC_TGID_CWD,
106 PROC_TGID_ROOT,
107 PROC_TGID_EXE,
108 PROC_TGID_FD,
109 PROC_TGID_ENVIRON,
110 PROC_TGID_AUXV,
111 PROC_TGID_CMDLINE,
112 PROC_TGID_STAT,
113 PROC_TGID_STATM,
114 PROC_TGID_MAPS,
Christoph Lameter6e21c8f2005-09-03 15:54:45 -0700115 PROC_TGID_NUMA_MAPS,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116 PROC_TGID_MOUNTS,
Chuck Leverb4629fe2006-03-20 13:44:12 -0500117 PROC_TGID_MOUNTSTATS,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118 PROC_TGID_WCHAN,
Yoshinori Sato63c67642005-10-14 15:59:11 -0700119#ifdef CONFIG_MMU
Mauricio Line070ad42005-09-03 15:55:10 -0700120 PROC_TGID_SMAPS,
Yoshinori Sato63c67642005-10-14 15:59:11 -0700121#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122#ifdef CONFIG_SCHEDSTATS
123 PROC_TGID_SCHEDSTAT,
124#endif
125#ifdef CONFIG_CPUSETS
126 PROC_TGID_CPUSET,
127#endif
128#ifdef CONFIG_SECURITY
129 PROC_TGID_ATTR,
130 PROC_TGID_ATTR_CURRENT,
131 PROC_TGID_ATTR_PREV,
132 PROC_TGID_ATTR_EXEC,
133 PROC_TGID_ATTR_FSCREATE,
Michael LeMay4eb582c2006-06-26 00:24:57 -0700134 PROC_TGID_ATTR_KEYCREATE,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135#endif
136#ifdef CONFIG_AUDITSYSCALL
137 PROC_TGID_LOGINUID,
138#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139 PROC_TGID_OOM_SCORE,
140 PROC_TGID_OOM_ADJUST,
141 PROC_TID_INO,
142 PROC_TID_STATUS,
143 PROC_TID_MEM,
144#ifdef CONFIG_SECCOMP
145 PROC_TID_SECCOMP,
146#endif
147 PROC_TID_CWD,
148 PROC_TID_ROOT,
149 PROC_TID_EXE,
150 PROC_TID_FD,
151 PROC_TID_ENVIRON,
152 PROC_TID_AUXV,
153 PROC_TID_CMDLINE,
154 PROC_TID_STAT,
155 PROC_TID_STATM,
156 PROC_TID_MAPS,
Christoph Lameter6e21c8f2005-09-03 15:54:45 -0700157 PROC_TID_NUMA_MAPS,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 PROC_TID_MOUNTS,
Chuck Leverb4629fe2006-03-20 13:44:12 -0500159 PROC_TID_MOUNTSTATS,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160 PROC_TID_WCHAN,
Yoshinori Sato63c67642005-10-14 15:59:11 -0700161#ifdef CONFIG_MMU
Mauricio Line070ad42005-09-03 15:55:10 -0700162 PROC_TID_SMAPS,
Yoshinori Sato63c67642005-10-14 15:59:11 -0700163#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164#ifdef CONFIG_SCHEDSTATS
165 PROC_TID_SCHEDSTAT,
166#endif
167#ifdef CONFIG_CPUSETS
168 PROC_TID_CPUSET,
169#endif
170#ifdef CONFIG_SECURITY
171 PROC_TID_ATTR,
172 PROC_TID_ATTR_CURRENT,
173 PROC_TID_ATTR_PREV,
174 PROC_TID_ATTR_EXEC,
175 PROC_TID_ATTR_FSCREATE,
Michael LeMay4eb582c2006-06-26 00:24:57 -0700176 PROC_TID_ATTR_KEYCREATE,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177#endif
178#ifdef CONFIG_AUDITSYSCALL
179 PROC_TID_LOGINUID,
180#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 PROC_TID_OOM_SCORE,
182 PROC_TID_OOM_ADJUST,
Miklos Szeredi5e21ccb2005-09-06 15:18:23 -0700183
184 /* Add new entries before this */
185 PROC_TID_FD_DIR = 0x8000, /* 0x8000-0xffff */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186};
187
Eric W. Biederman8578cea2006-06-26 00:25:54 -0700188/* Worst case buffer size needed for holding an integer. */
189#define PROC_NUMBUF 10
190
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191struct pid_entry {
192 int type;
193 int len;
194 char *name;
195 mode_t mode;
196};
197
198#define E(type,name,mode) {(type),sizeof(name)-1,(name),(mode)}
199
200static struct pid_entry tgid_base_stuff[] = {
201 E(PROC_TGID_TASK, "task", S_IFDIR|S_IRUGO|S_IXUGO),
202 E(PROC_TGID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
203 E(PROC_TGID_ENVIRON, "environ", S_IFREG|S_IRUSR),
204 E(PROC_TGID_AUXV, "auxv", S_IFREG|S_IRUSR),
205 E(PROC_TGID_STATUS, "status", S_IFREG|S_IRUGO),
206 E(PROC_TGID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
207 E(PROC_TGID_STAT, "stat", S_IFREG|S_IRUGO),
208 E(PROC_TGID_STATM, "statm", S_IFREG|S_IRUGO),
209 E(PROC_TGID_MAPS, "maps", S_IFREG|S_IRUGO),
Christoph Lameter6e21c8f2005-09-03 15:54:45 -0700210#ifdef CONFIG_NUMA
211 E(PROC_TGID_NUMA_MAPS, "numa_maps", S_IFREG|S_IRUGO),
212#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 E(PROC_TGID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
214#ifdef CONFIG_SECCOMP
215 E(PROC_TGID_SECCOMP, "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
216#endif
217 E(PROC_TGID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
218 E(PROC_TGID_ROOT, "root", S_IFLNK|S_IRWXUGO),
219 E(PROC_TGID_EXE, "exe", S_IFLNK|S_IRWXUGO),
220 E(PROC_TGID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
Chuck Leverb4629fe2006-03-20 13:44:12 -0500221 E(PROC_TGID_MOUNTSTATS, "mountstats", S_IFREG|S_IRUSR),
Yoshinori Sato63c67642005-10-14 15:59:11 -0700222#ifdef CONFIG_MMU
Mauricio Line070ad42005-09-03 15:55:10 -0700223 E(PROC_TGID_SMAPS, "smaps", S_IFREG|S_IRUGO),
Yoshinori Sato63c67642005-10-14 15:59:11 -0700224#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225#ifdef CONFIG_SECURITY
226 E(PROC_TGID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
227#endif
228#ifdef CONFIG_KALLSYMS
229 E(PROC_TGID_WCHAN, "wchan", S_IFREG|S_IRUGO),
230#endif
231#ifdef CONFIG_SCHEDSTATS
232 E(PROC_TGID_SCHEDSTAT, "schedstat", S_IFREG|S_IRUGO),
233#endif
234#ifdef CONFIG_CPUSETS
235 E(PROC_TGID_CPUSET, "cpuset", S_IFREG|S_IRUGO),
236#endif
237 E(PROC_TGID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
238 E(PROC_TGID_OOM_ADJUST,"oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
239#ifdef CONFIG_AUDITSYSCALL
240 E(PROC_TGID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
241#endif
242 {0,0,NULL,0}
243};
244static struct pid_entry tid_base_stuff[] = {
245 E(PROC_TID_FD, "fd", S_IFDIR|S_IRUSR|S_IXUSR),
246 E(PROC_TID_ENVIRON, "environ", S_IFREG|S_IRUSR),
247 E(PROC_TID_AUXV, "auxv", S_IFREG|S_IRUSR),
248 E(PROC_TID_STATUS, "status", S_IFREG|S_IRUGO),
249 E(PROC_TID_CMDLINE, "cmdline", S_IFREG|S_IRUGO),
250 E(PROC_TID_STAT, "stat", S_IFREG|S_IRUGO),
251 E(PROC_TID_STATM, "statm", S_IFREG|S_IRUGO),
252 E(PROC_TID_MAPS, "maps", S_IFREG|S_IRUGO),
Christoph Lameter6e21c8f2005-09-03 15:54:45 -0700253#ifdef CONFIG_NUMA
254 E(PROC_TID_NUMA_MAPS, "numa_maps", S_IFREG|S_IRUGO),
255#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256 E(PROC_TID_MEM, "mem", S_IFREG|S_IRUSR|S_IWUSR),
257#ifdef CONFIG_SECCOMP
258 E(PROC_TID_SECCOMP, "seccomp", S_IFREG|S_IRUSR|S_IWUSR),
259#endif
260 E(PROC_TID_CWD, "cwd", S_IFLNK|S_IRWXUGO),
261 E(PROC_TID_ROOT, "root", S_IFLNK|S_IRWXUGO),
262 E(PROC_TID_EXE, "exe", S_IFLNK|S_IRWXUGO),
263 E(PROC_TID_MOUNTS, "mounts", S_IFREG|S_IRUGO),
Yoshinori Sato63c67642005-10-14 15:59:11 -0700264#ifdef CONFIG_MMU
Mauricio Line070ad42005-09-03 15:55:10 -0700265 E(PROC_TID_SMAPS, "smaps", S_IFREG|S_IRUGO),
Yoshinori Sato63c67642005-10-14 15:59:11 -0700266#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267#ifdef CONFIG_SECURITY
268 E(PROC_TID_ATTR, "attr", S_IFDIR|S_IRUGO|S_IXUGO),
269#endif
270#ifdef CONFIG_KALLSYMS
271 E(PROC_TID_WCHAN, "wchan", S_IFREG|S_IRUGO),
272#endif
273#ifdef CONFIG_SCHEDSTATS
274 E(PROC_TID_SCHEDSTAT, "schedstat",S_IFREG|S_IRUGO),
275#endif
276#ifdef CONFIG_CPUSETS
277 E(PROC_TID_CPUSET, "cpuset", S_IFREG|S_IRUGO),
278#endif
279 E(PROC_TID_OOM_SCORE, "oom_score",S_IFREG|S_IRUGO),
280 E(PROC_TID_OOM_ADJUST, "oom_adj", S_IFREG|S_IRUGO|S_IWUSR),
281#ifdef CONFIG_AUDITSYSCALL
282 E(PROC_TID_LOGINUID, "loginuid", S_IFREG|S_IWUSR|S_IRUGO),
283#endif
284 {0,0,NULL,0}
285};
286
287#ifdef CONFIG_SECURITY
288static struct pid_entry tgid_attr_stuff[] = {
289 E(PROC_TGID_ATTR_CURRENT, "current", S_IFREG|S_IRUGO|S_IWUGO),
290 E(PROC_TGID_ATTR_PREV, "prev", S_IFREG|S_IRUGO),
291 E(PROC_TGID_ATTR_EXEC, "exec", S_IFREG|S_IRUGO|S_IWUGO),
292 E(PROC_TGID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
Michael LeMay4eb582c2006-06-26 00:24:57 -0700293 E(PROC_TGID_ATTR_KEYCREATE, "keycreate", S_IFREG|S_IRUGO|S_IWUGO),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 {0,0,NULL,0}
295};
296static struct pid_entry tid_attr_stuff[] = {
297 E(PROC_TID_ATTR_CURRENT, "current", S_IFREG|S_IRUGO|S_IWUGO),
298 E(PROC_TID_ATTR_PREV, "prev", S_IFREG|S_IRUGO),
299 E(PROC_TID_ATTR_EXEC, "exec", S_IFREG|S_IRUGO|S_IWUGO),
300 E(PROC_TID_ATTR_FSCREATE, "fscreate", S_IFREG|S_IRUGO|S_IWUGO),
Michael LeMay4eb582c2006-06-26 00:24:57 -0700301 E(PROC_TID_ATTR_KEYCREATE, "keycreate", S_IFREG|S_IRUGO|S_IWUGO),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 {0,0,NULL,0}
303};
304#endif
305
306#undef E
307
308static int proc_fd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
309{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700310 struct task_struct *task = get_proc_task(inode);
311 struct files_struct *files = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 struct file *file;
Eric W. Biedermanaed7a6c2006-06-26 00:25:44 -0700313 int fd = proc_fd(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314
Eric W. Biederman99f89552006-06-26 00:25:55 -0700315 if (task) {
316 files = get_files_struct(task);
317 put_task_struct(task);
318 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 if (files) {
Dipankar Sarmaca99c1d2006-04-18 22:21:46 -0700320 /*
321 * We are not taking a ref to the file structure, so we must
322 * hold ->file_lock.
323 */
324 spin_lock(&files->file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325 file = fcheck_files(files, fd);
326 if (file) {
327 *mnt = mntget(file->f_vfsmnt);
328 *dentry = dget(file->f_dentry);
Dipankar Sarmaca99c1d2006-04-18 22:21:46 -0700329 spin_unlock(&files->file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 put_files_struct(files);
331 return 0;
332 }
Dipankar Sarmaca99c1d2006-04-18 22:21:46 -0700333 spin_unlock(&files->file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334 put_files_struct(files);
335 }
336 return -ENOENT;
337}
338
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700339static struct fs_struct *get_fs_struct(struct task_struct *task)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340{
341 struct fs_struct *fs;
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700342 task_lock(task);
343 fs = task->fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 if(fs)
345 atomic_inc(&fs->count);
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700346 task_unlock(task);
347 return fs;
348}
349
Eric W. Biederman99f89552006-06-26 00:25:55 -0700350static int get_nr_threads(struct task_struct *tsk)
351{
352 /* Must be called with the rcu_read_lock held */
353 unsigned long flags;
354 int count = 0;
355
356 if (lock_task_sighand(tsk, &flags)) {
357 count = atomic_read(&tsk->signal->count);
358 unlock_task_sighand(tsk, &flags);
359 }
360 return count;
361}
362
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700363static int proc_cwd_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
364{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700365 struct task_struct *task = get_proc_task(inode);
366 struct fs_struct *fs = NULL;
Miklos Szeredi0494f6e2005-09-06 15:18:22 -0700367 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700368
369 if (task) {
370 fs = get_fs_struct(task);
371 put_task_struct(task);
372 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 if (fs) {
374 read_lock(&fs->lock);
375 *mnt = mntget(fs->pwdmnt);
376 *dentry = dget(fs->pwd);
377 read_unlock(&fs->lock);
378 result = 0;
379 put_fs_struct(fs);
380 }
381 return result;
382}
383
384static int proc_root_link(struct inode *inode, struct dentry **dentry, struct vfsmount **mnt)
385{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700386 struct task_struct *task = get_proc_task(inode);
387 struct fs_struct *fs = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 int result = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700389
390 if (task) {
391 fs = get_fs_struct(task);
392 put_task_struct(task);
393 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 if (fs) {
395 read_lock(&fs->lock);
396 *mnt = mntget(fs->rootmnt);
397 *dentry = dget(fs->root);
398 read_unlock(&fs->lock);
399 result = 0;
400 put_fs_struct(fs);
401 }
402 return result;
403}
404
405#define MAY_PTRACE(task) \
406 (task == current || \
407 (task->parent == current && \
408 (task->ptrace & PT_PTRACED) && \
409 (task->state == TASK_STOPPED || task->state == TASK_TRACED) && \
410 security_ptrace(current,task) == 0))
411
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412static int proc_pid_environ(struct task_struct *task, char * buffer)
413{
414 int res = 0;
415 struct mm_struct *mm = get_task_mm(task);
416 if (mm) {
417 unsigned int len = mm->env_end - mm->env_start;
418 if (len > PAGE_SIZE)
419 len = PAGE_SIZE;
420 res = access_process_vm(task, mm->env_start, buffer, len, 0);
Miklos Szerediab8d11b2005-09-06 15:18:24 -0700421 if (!ptrace_may_attach(task))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 res = -ESRCH;
423 mmput(mm);
424 }
425 return res;
426}
427
428static int proc_pid_cmdline(struct task_struct *task, char * buffer)
429{
430 int res = 0;
431 unsigned int len;
432 struct mm_struct *mm = get_task_mm(task);
433 if (!mm)
434 goto out;
435 if (!mm->arg_end)
436 goto out_mm; /* Shh! No looking before we're done */
437
438 len = mm->arg_end - mm->arg_start;
439
440 if (len > PAGE_SIZE)
441 len = PAGE_SIZE;
442
443 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
444
445 // If the nul at the end of args has been overwritten, then
446 // assume application is using setproctitle(3).
447 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
448 len = strnlen(buffer, res);
449 if (len < res) {
450 res = len;
451 } else {
452 len = mm->env_end - mm->env_start;
453 if (len > PAGE_SIZE - res)
454 len = PAGE_SIZE - res;
455 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
456 res = strnlen(buffer, res);
457 }
458 }
459out_mm:
460 mmput(mm);
461out:
462 return res;
463}
464
465static int proc_pid_auxv(struct task_struct *task, char *buffer)
466{
467 int res = 0;
468 struct mm_struct *mm = get_task_mm(task);
469 if (mm) {
470 unsigned int nwords = 0;
471 do
472 nwords += 2;
473 while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
474 res = nwords * sizeof(mm->saved_auxv[0]);
475 if (res > PAGE_SIZE)
476 res = PAGE_SIZE;
477 memcpy(buffer, mm->saved_auxv, res);
478 mmput(mm);
479 }
480 return res;
481}
482
483
484#ifdef CONFIG_KALLSYMS
485/*
486 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
487 * Returns the resolved symbol. If that fails, simply return the address.
488 */
489static int proc_pid_wchan(struct task_struct *task, char *buffer)
490{
491 char *modname;
492 const char *sym_name;
493 unsigned long wchan, size, offset;
494 char namebuf[KSYM_NAME_LEN+1];
495
496 wchan = get_wchan(task);
497
498 sym_name = kallsyms_lookup(wchan, &size, &offset, &modname, namebuf);
499 if (sym_name)
500 return sprintf(buffer, "%s", sym_name);
501 return sprintf(buffer, "%lu", wchan);
502}
503#endif /* CONFIG_KALLSYMS */
504
505#ifdef CONFIG_SCHEDSTATS
506/*
507 * Provides /proc/PID/schedstat
508 */
509static int proc_pid_schedstat(struct task_struct *task, char *buffer)
510{
511 return sprintf(buffer, "%lu %lu %lu\n",
512 task->sched_info.cpu_time,
513 task->sched_info.run_delay,
514 task->sched_info.pcnt);
515}
516#endif
517
518/* The badness from the OOM killer */
519unsigned long badness(struct task_struct *p, unsigned long uptime);
520static int proc_oom_score(struct task_struct *task, char *buffer)
521{
522 unsigned long points;
523 struct timespec uptime;
524
525 do_posix_clock_monotonic_gettime(&uptime);
526 points = badness(task, uptime.tv_sec);
527 return sprintf(buffer, "%lu\n", points);
528}
529
530/************************************************************************/
531/* Here the fs part begins */
532/************************************************************************/
533
534/* permission checks */
Eric W. Biederman778c1142006-06-26 00:25:58 -0700535static int proc_fd_access_allowed(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536{
Eric W. Biederman778c1142006-06-26 00:25:58 -0700537 struct task_struct *task;
538 int allowed = 0;
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700539 /* Allow access to a task's file descriptors if it is us or we
540 * may use ptrace attach to the process and find out that
541 * information.
Eric W. Biederman778c1142006-06-26 00:25:58 -0700542 */
543 task = get_proc_task(inode);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700544 if (task) {
545 allowed = ptrace_may_attach(task);
Eric W. Biederman778c1142006-06-26 00:25:58 -0700546 put_task_struct(task);
Eric W. Biedermandf26c402006-06-26 00:25:59 -0700547 }
Eric W. Biederman778c1142006-06-26 00:25:58 -0700548 return allowed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549}
550
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551extern struct seq_operations mounts_op;
Al Viro5addc5d2005-11-07 17:15:49 -0500552struct proc_mounts {
553 struct seq_file m;
554 int event;
555};
556
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557static int mounts_open(struct inode *inode, struct file *file)
558{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700559 struct task_struct *task = get_proc_task(inode);
560 struct namespace *namespace = NULL;
Al Viro5addc5d2005-11-07 17:15:49 -0500561 struct proc_mounts *p;
562 int ret = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563
Eric W. Biederman99f89552006-06-26 00:25:55 -0700564 if (task) {
565 task_lock(task);
566 namespace = task->namespace;
567 if (namespace)
568 get_namespace(namespace);
569 task_unlock(task);
570 put_task_struct(task);
571 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572
Al Viro5addc5d2005-11-07 17:15:49 -0500573 if (namespace) {
574 ret = -ENOMEM;
575 p = kmalloc(sizeof(struct proc_mounts), GFP_KERNEL);
576 if (p) {
577 file->private_data = &p->m;
578 ret = seq_open(file, &mounts_op);
579 if (!ret) {
580 p->m.private = namespace;
581 p->event = namespace->event;
582 return 0;
583 }
584 kfree(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 }
Al Viro5addc5d2005-11-07 17:15:49 -0500586 put_namespace(namespace);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 }
588 return ret;
589}
590
591static int mounts_release(struct inode *inode, struct file *file)
592{
593 struct seq_file *m = file->private_data;
594 struct namespace *namespace = m->private;
595 put_namespace(namespace);
596 return seq_release(inode, file);
597}
598
Al Viro5addc5d2005-11-07 17:15:49 -0500599static unsigned mounts_poll(struct file *file, poll_table *wait)
600{
601 struct proc_mounts *p = file->private_data;
602 struct namespace *ns = p->m.private;
603 unsigned res = 0;
604
605 poll_wait(file, &ns->poll, wait);
606
607 spin_lock(&vfsmount_lock);
608 if (p->event != ns->event) {
609 p->event = ns->event;
610 res = POLLERR;
611 }
612 spin_unlock(&vfsmount_lock);
613
614 return res;
615}
616
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617static struct file_operations proc_mounts_operations = {
618 .open = mounts_open,
619 .read = seq_read,
620 .llseek = seq_lseek,
621 .release = mounts_release,
Al Viro5addc5d2005-11-07 17:15:49 -0500622 .poll = mounts_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623};
624
Chuck Leverb4629fe2006-03-20 13:44:12 -0500625extern struct seq_operations mountstats_op;
626static int mountstats_open(struct inode *inode, struct file *file)
627{
Chuck Leverb4629fe2006-03-20 13:44:12 -0500628 int ret = seq_open(file, &mountstats_op);
629
630 if (!ret) {
631 struct seq_file *m = file->private_data;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700632 struct namespace *namespace = NULL;
633 struct task_struct *task = get_proc_task(inode);
634
635 if (task) {
636 task_lock(task);
637 namespace = task->namespace;
638 if (namespace)
639 get_namespace(namespace);
640 task_unlock(task);
641 put_task_struct(task);
642 }
Chuck Leverb4629fe2006-03-20 13:44:12 -0500643
644 if (namespace)
645 m->private = namespace;
646 else {
647 seq_release(inode, file);
648 ret = -EINVAL;
649 }
650 }
651 return ret;
652}
653
654static struct file_operations proc_mountstats_operations = {
655 .open = mountstats_open,
656 .read = seq_read,
657 .llseek = seq_lseek,
658 .release = mounts_release,
659};
660
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661#define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
662
663static ssize_t proc_info_read(struct file * file, char __user * buf,
664 size_t count, loff_t *ppos)
665{
666 struct inode * inode = file->f_dentry->d_inode;
667 unsigned long page;
668 ssize_t length;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700669 struct task_struct *task = get_proc_task(inode);
670
671 length = -ESRCH;
672 if (!task)
673 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674
675 if (count > PROC_BLOCK_SIZE)
676 count = PROC_BLOCK_SIZE;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700677
678 length = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 if (!(page = __get_free_page(GFP_KERNEL)))
Eric W. Biederman99f89552006-06-26 00:25:55 -0700680 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681
682 length = PROC_I(inode)->op.proc_read(task, (char*)page);
683
684 if (length >= 0)
685 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
686 free_page(page);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700687out:
688 put_task_struct(task);
689out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690 return length;
691}
692
693static struct file_operations proc_info_file_operations = {
694 .read = proc_info_read,
695};
696
697static int mem_open(struct inode* inode, struct file* file)
698{
699 file->private_data = (void*)((long)current->self_exec_id);
700 return 0;
701}
702
703static ssize_t mem_read(struct file * file, char __user * buf,
704 size_t count, loff_t *ppos)
705{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700706 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707 char *page;
708 unsigned long src = *ppos;
709 int ret = -ESRCH;
710 struct mm_struct *mm;
711
Eric W. Biederman99f89552006-06-26 00:25:55 -0700712 if (!task)
713 goto out_no_task;
714
Miklos Szerediab8d11b2005-09-06 15:18:24 -0700715 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 goto out;
717
718 ret = -ENOMEM;
719 page = (char *)__get_free_page(GFP_USER);
720 if (!page)
721 goto out;
722
723 ret = 0;
724
725 mm = get_task_mm(task);
726 if (!mm)
727 goto out_free;
728
729 ret = -EIO;
730
731 if (file->private_data != (void*)((long)current->self_exec_id))
732 goto out_put;
733
734 ret = 0;
735
736 while (count > 0) {
737 int this_len, retval;
738
739 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
740 retval = access_process_vm(task, src, page, this_len, 0);
Miklos Szerediab8d11b2005-09-06 15:18:24 -0700741 if (!retval || !MAY_PTRACE(task) || !ptrace_may_attach(task)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742 if (!ret)
743 ret = -EIO;
744 break;
745 }
746
747 if (copy_to_user(buf, page, retval)) {
748 ret = -EFAULT;
749 break;
750 }
751
752 ret += retval;
753 src += retval;
754 buf += retval;
755 count -= retval;
756 }
757 *ppos = src;
758
759out_put:
760 mmput(mm);
761out_free:
762 free_page((unsigned long) page);
763out:
Eric W. Biederman99f89552006-06-26 00:25:55 -0700764 put_task_struct(task);
765out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 return ret;
767}
768
769#define mem_write NULL
770
771#ifndef mem_write
772/* This is a security hazard */
773static ssize_t mem_write(struct file * file, const char * buf,
774 size_t count, loff_t *ppos)
775{
776 int copied = 0;
777 char *page;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700778 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 unsigned long dst = *ppos;
780
Eric W. Biederman99f89552006-06-26 00:25:55 -0700781 copied = -ESRCH;
782 if (!task)
783 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784
Eric W. Biederman99f89552006-06-26 00:25:55 -0700785 if (!MAY_PTRACE(task) || !ptrace_may_attach(task))
786 goto out;
787
788 copied = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 page = (char *)__get_free_page(GFP_USER);
790 if (!page)
Eric W. Biederman99f89552006-06-26 00:25:55 -0700791 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792
793 while (count > 0) {
794 int this_len, retval;
795
796 this_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
797 if (copy_from_user(page, buf, this_len)) {
798 copied = -EFAULT;
799 break;
800 }
801 retval = access_process_vm(task, dst, page, this_len, 1);
802 if (!retval) {
803 if (!copied)
804 copied = -EIO;
805 break;
806 }
807 copied += retval;
808 buf += retval;
809 dst += retval;
810 count -= retval;
811 }
812 *ppos = dst;
813 free_page((unsigned long) page);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700814out:
815 put_task_struct(task);
816out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 return copied;
818}
819#endif
820
821static loff_t mem_lseek(struct file * file, loff_t offset, int orig)
822{
823 switch (orig) {
824 case 0:
825 file->f_pos = offset;
826 break;
827 case 1:
828 file->f_pos += offset;
829 break;
830 default:
831 return -EINVAL;
832 }
833 force_successful_syscall_return();
834 return file->f_pos;
835}
836
837static struct file_operations proc_mem_operations = {
838 .llseek = mem_lseek,
839 .read = mem_read,
840 .write = mem_write,
841 .open = mem_open,
842};
843
844static ssize_t oom_adjust_read(struct file *file, char __user *buf,
845 size_t count, loff_t *ppos)
846{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700847 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
Eric W. Biederman8578cea2006-06-26 00:25:54 -0700848 char buffer[PROC_NUMBUF];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 size_t len;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700850 int oom_adjust;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 loff_t __ppos = *ppos;
852
Eric W. Biederman99f89552006-06-26 00:25:55 -0700853 if (!task)
854 return -ESRCH;
855 oom_adjust = task->oomkilladj;
856 put_task_struct(task);
857
Eric W. Biederman8578cea2006-06-26 00:25:54 -0700858 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 if (__ppos >= len)
860 return 0;
861 if (count > len-__ppos)
862 count = len-__ppos;
863 if (copy_to_user(buf, buffer + __ppos, count))
864 return -EFAULT;
865 *ppos = __ppos + count;
866 return count;
867}
868
869static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
870 size_t count, loff_t *ppos)
871{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700872 struct task_struct *task;
Eric W. Biederman8578cea2006-06-26 00:25:54 -0700873 char buffer[PROC_NUMBUF], *end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 int oom_adjust;
875
876 if (!capable(CAP_SYS_RESOURCE))
877 return -EPERM;
Eric W. Biederman8578cea2006-06-26 00:25:54 -0700878 memset(buffer, 0, sizeof(buffer));
879 if (count > sizeof(buffer) - 1)
880 count = sizeof(buffer) - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881 if (copy_from_user(buffer, buf, count))
882 return -EFAULT;
883 oom_adjust = simple_strtol(buffer, &end, 0);
Andrea Arcangeli79befd02005-04-16 15:24:05 -0700884 if ((oom_adjust < -16 || oom_adjust > 15) && oom_adjust != OOM_DISABLE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885 return -EINVAL;
886 if (*end == '\n')
887 end++;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700888 task = get_proc_task(file->f_dentry->d_inode);
889 if (!task)
890 return -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 task->oomkilladj = oom_adjust;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700892 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 if (end - buffer == 0)
894 return -EIO;
895 return end - buffer;
896}
897
898static struct file_operations proc_oom_adjust_operations = {
899 .read = oom_adjust_read,
900 .write = oom_adjust_write,
901};
902
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903#ifdef CONFIG_AUDITSYSCALL
904#define TMPBUFLEN 21
905static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
906 size_t count, loff_t *ppos)
907{
908 struct inode * inode = file->f_dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -0700909 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 ssize_t length;
911 char tmpbuf[TMPBUFLEN];
912
Eric W. Biederman99f89552006-06-26 00:25:55 -0700913 if (!task)
914 return -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
916 audit_get_loginuid(task->audit_context));
Eric W. Biederman99f89552006-06-26 00:25:55 -0700917 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
919}
920
921static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
922 size_t count, loff_t *ppos)
923{
924 struct inode * inode = file->f_dentry->d_inode;
925 char *page, *tmp;
926 ssize_t length;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 uid_t loginuid;
928
929 if (!capable(CAP_AUDIT_CONTROL))
930 return -EPERM;
931
Eric W. Biederman13b41b02006-06-26 00:25:56 -0700932 if (current != pid_task(proc_pid(inode), PIDTYPE_PID))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933 return -EPERM;
934
Al Viroe0182902006-05-18 08:28:02 -0400935 if (count >= PAGE_SIZE)
936 count = PAGE_SIZE - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937
938 if (*ppos != 0) {
939 /* No partial writes. */
940 return -EINVAL;
941 }
942 page = (char*)__get_free_page(GFP_USER);
943 if (!page)
944 return -ENOMEM;
945 length = -EFAULT;
946 if (copy_from_user(page, buf, count))
947 goto out_free_page;
948
Al Viroe0182902006-05-18 08:28:02 -0400949 page[count] = '\0';
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950 loginuid = simple_strtoul(page, &tmp, 10);
951 if (tmp == page) {
952 length = -EINVAL;
953 goto out_free_page;
954
955 }
Eric W. Biederman99f89552006-06-26 00:25:55 -0700956 length = audit_set_loginuid(current, loginuid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 if (likely(length == 0))
958 length = count;
959
960out_free_page:
961 free_page((unsigned long) page);
962 return length;
963}
964
965static struct file_operations proc_loginuid_operations = {
966 .read = proc_loginuid_read,
967 .write = proc_loginuid_write,
968};
969#endif
970
971#ifdef CONFIG_SECCOMP
972static ssize_t seccomp_read(struct file *file, char __user *buf,
973 size_t count, loff_t *ppos)
974{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700975 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976 char __buf[20];
977 loff_t __ppos = *ppos;
978 size_t len;
979
Eric W. Biederman99f89552006-06-26 00:25:55 -0700980 if (!tsk)
981 return -ESRCH;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 /* no need to print the trailing zero, so use only len */
983 len = sprintf(__buf, "%u\n", tsk->seccomp.mode);
Eric W. Biederman99f89552006-06-26 00:25:55 -0700984 put_task_struct(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 if (__ppos >= len)
986 return 0;
987 if (count > len - __ppos)
988 count = len - __ppos;
989 if (copy_to_user(buf, __buf + __ppos, count))
990 return -EFAULT;
991 *ppos = __ppos + count;
992 return count;
993}
994
995static ssize_t seccomp_write(struct file *file, const char __user *buf,
996 size_t count, loff_t *ppos)
997{
Eric W. Biederman99f89552006-06-26 00:25:55 -0700998 struct task_struct *tsk = get_proc_task(file->f_dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 char __buf[20], *end;
1000 unsigned int seccomp_mode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001001 ssize_t result;
1002
1003 result = -ESRCH;
1004 if (!tsk)
1005 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006
1007 /* can set it only once to be even more secure */
Eric W. Biederman99f89552006-06-26 00:25:55 -07001008 result = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 if (unlikely(tsk->seccomp.mode))
Eric W. Biederman99f89552006-06-26 00:25:55 -07001010 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011
Eric W. Biederman99f89552006-06-26 00:25:55 -07001012 result = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 memset(__buf, 0, sizeof(__buf));
1014 count = min(count, sizeof(__buf) - 1);
1015 if (copy_from_user(__buf, buf, count))
Eric W. Biederman99f89552006-06-26 00:25:55 -07001016 goto out;
1017
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 seccomp_mode = simple_strtoul(__buf, &end, 0);
1019 if (*end == '\n')
1020 end++;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001021 result = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 if (seccomp_mode && seccomp_mode <= NR_SECCOMP_MODES) {
1023 tsk->seccomp.mode = seccomp_mode;
1024 set_tsk_thread_flag(tsk, TIF_SECCOMP);
1025 } else
Eric W. Biederman99f89552006-06-26 00:25:55 -07001026 goto out;
1027 result = -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 if (unlikely(!(end - __buf)))
Eric W. Biederman99f89552006-06-26 00:25:55 -07001029 goto out;
1030 result = end - __buf;
1031out:
1032 put_task_struct(tsk);
1033out_no_task:
1034 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035}
1036
1037static struct file_operations proc_seccomp_operations = {
1038 .read = seccomp_read,
1039 .write = seccomp_write,
1040};
1041#endif /* CONFIG_SECCOMP */
1042
Al Viro008b1502005-08-20 00:17:39 +01001043static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044{
1045 struct inode *inode = dentry->d_inode;
1046 int error = -EACCES;
1047
1048 /* We don't need a base pointer in the /proc filesystem */
1049 path_release(nd);
1050
Eric W. Biederman778c1142006-06-26 00:25:58 -07001051 /* Are we allowed to snoop on the tasks file descriptors? */
1052 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054
1055 error = PROC_I(inode)->op.proc_get_link(inode, &nd->dentry, &nd->mnt);
1056 nd->last_type = LAST_BIND;
1057out:
Al Viro008b1502005-08-20 00:17:39 +01001058 return ERR_PTR(error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059}
1060
1061static int do_proc_readlink(struct dentry *dentry, struct vfsmount *mnt,
1062 char __user *buffer, int buflen)
1063{
1064 struct inode * inode;
1065 char *tmp = (char*)__get_free_page(GFP_KERNEL), *path;
1066 int len;
1067
1068 if (!tmp)
1069 return -ENOMEM;
1070
1071 inode = dentry->d_inode;
1072 path = d_path(dentry, mnt, tmp, PAGE_SIZE);
1073 len = PTR_ERR(path);
1074 if (IS_ERR(path))
1075 goto out;
1076 len = tmp + PAGE_SIZE - 1 - path;
1077
1078 if (len > buflen)
1079 len = buflen;
1080 if (copy_to_user(buffer, path, len))
1081 len = -EFAULT;
1082 out:
1083 free_page((unsigned long)tmp);
1084 return len;
1085}
1086
1087static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1088{
1089 int error = -EACCES;
1090 struct inode *inode = dentry->d_inode;
1091 struct dentry *de;
1092 struct vfsmount *mnt = NULL;
1093
Eric W. Biederman778c1142006-06-26 00:25:58 -07001094 /* Are we allowed to snoop on the tasks file descriptors? */
1095 if (!proc_fd_access_allowed(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097
1098 error = PROC_I(inode)->op.proc_get_link(inode, &de, &mnt);
1099 if (error)
1100 goto out;
1101
1102 error = do_proc_readlink(de, mnt, buffer, buflen);
1103 dput(de);
1104 mntput(mnt);
1105out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 return error;
1107}
1108
1109static struct inode_operations proc_pid_link_inode_operations = {
1110 .readlink = proc_pid_readlink,
1111 .follow_link = proc_pid_follow_link
1112};
1113
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114static int proc_readfd(struct file * filp, void * dirent, filldir_t filldir)
1115{
Eric W. Biederman56347082006-06-26 00:25:40 -07001116 struct dentry *dentry = filp->f_dentry;
1117 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001118 struct task_struct *p = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 unsigned int fd, tid, ino;
1120 int retval;
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001121 char buf[PROC_NUMBUF];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122 struct files_struct * files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001123 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124
1125 retval = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001126 if (!p)
1127 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 retval = 0;
1129 tid = p->pid;
1130
1131 fd = filp->f_pos;
1132 switch (fd) {
1133 case 0:
1134 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1135 goto out;
1136 filp->f_pos++;
1137 case 1:
Eric W. Biederman56347082006-06-26 00:25:40 -07001138 ino = parent_ino(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1140 goto out;
1141 filp->f_pos++;
1142 default:
1143 files = get_files_struct(p);
1144 if (!files)
1145 goto out;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001146 rcu_read_lock();
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001147 fdt = files_fdtable(files);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 for (fd = filp->f_pos-2;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001149 fd < fdt->max_fds;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150 fd++, filp->f_pos++) {
1151 unsigned int i,j;
1152
1153 if (!fcheck_files(files, fd))
1154 continue;
Dipankar Sarmab8359962005-09-09 13:04:14 -07001155 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001157 j = PROC_NUMBUF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 i = fd;
1159 do {
1160 j--;
1161 buf[j] = '0' + (i % 10);
1162 i /= 10;
1163 } while (i);
1164
1165 ino = fake_ino(tid, PROC_TID_FD_DIR + fd);
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001166 if (filldir(dirent, buf+j, PROC_NUMBUF-j, fd+2, ino, DT_LNK) < 0) {
Dipankar Sarmab8359962005-09-09 13:04:14 -07001167 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168 break;
1169 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07001170 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07001172 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 put_files_struct(files);
1174 }
1175out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07001176 put_task_struct(p);
1177out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 return retval;
1179}
1180
1181static int proc_pident_readdir(struct file *filp,
1182 void *dirent, filldir_t filldir,
1183 struct pid_entry *ents, unsigned int nents)
1184{
1185 int i;
1186 int pid;
1187 struct dentry *dentry = filp->f_dentry;
1188 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001189 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190 struct pid_entry *p;
1191 ino_t ino;
1192 int ret;
1193
1194 ret = -ENOENT;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001195 if (!task)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 goto out;
1197
1198 ret = 0;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001199 pid = task->pid;
1200 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 i = filp->f_pos;
1202 switch (i) {
1203 case 0:
1204 ino = inode->i_ino;
1205 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
1206 goto out;
1207 i++;
1208 filp->f_pos++;
1209 /* fall through */
1210 case 1:
1211 ino = parent_ino(dentry);
1212 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
1213 goto out;
1214 i++;
1215 filp->f_pos++;
1216 /* fall through */
1217 default:
1218 i -= 2;
1219 if (i >= nents) {
1220 ret = 1;
1221 goto out;
1222 }
1223 p = ents + i;
1224 while (p->name) {
1225 if (filldir(dirent, p->name, p->len, filp->f_pos,
1226 fake_ino(pid, p->type), p->mode >> 12) < 0)
1227 goto out;
1228 filp->f_pos++;
1229 p++;
1230 }
1231 }
1232
1233 ret = 1;
1234out:
1235 return ret;
1236}
1237
1238static int proc_tgid_base_readdir(struct file * filp,
1239 void * dirent, filldir_t filldir)
1240{
1241 return proc_pident_readdir(filp,dirent,filldir,
1242 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
1243}
1244
1245static int proc_tid_base_readdir(struct file * filp,
1246 void * dirent, filldir_t filldir)
1247{
1248 return proc_pident_readdir(filp,dirent,filldir,
1249 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
1250}
1251
1252/* building an inode */
1253
1254static int task_dumpable(struct task_struct *task)
1255{
1256 int dumpable = 0;
1257 struct mm_struct *mm;
1258
1259 task_lock(task);
1260 mm = task->mm;
1261 if (mm)
1262 dumpable = mm->dumpable;
1263 task_unlock(task);
Alan Coxd6e71142005-06-23 00:09:43 -07001264 if(dumpable == 1)
1265 return 1;
1266 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267}
1268
1269
1270static struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task, int ino)
1271{
1272 struct inode * inode;
1273 struct proc_inode *ei;
1274
1275 /* We need a new inode */
1276
1277 inode = new_inode(sb);
1278 if (!inode)
1279 goto out;
1280
1281 /* Common stuff */
1282 ei = PROC_I(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1284 inode->i_ino = fake_ino(task->pid, ino);
1285
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286 /*
1287 * grab the reference to task.
1288 */
Eric W. Biederman13b41b02006-06-26 00:25:56 -07001289 ei->pid = get_pid(task->pids[PIDTYPE_PID].pid);
1290 if (!ei->pid)
Eric W. Biederman99f89552006-06-26 00:25:55 -07001291 goto out_unlock;
1292
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293 inode->i_uid = 0;
1294 inode->i_gid = 0;
Eric W. Biederman87bfbf62006-06-26 00:25:43 -07001295 if (task_dumpable(task)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 inode->i_uid = task->euid;
1297 inode->i_gid = task->egid;
1298 }
1299 security_task_to_inode(task, inode);
1300
1301out:
1302 return inode;
1303
1304out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305 iput(inode);
1306 return NULL;
1307}
1308
1309/* dentry stuff */
1310
1311/*
1312 * Exceptional case: normally we are not allowed to unhash a busy
1313 * directory. In this case, however, we can do it - no aliasing problems
1314 * due to the way we treat inodes.
1315 *
1316 * Rewrite the inode's ownerships here because the owning task may have
1317 * performed a setuid(), etc.
Eric W. Biederman99f89552006-06-26 00:25:55 -07001318 *
1319 * Before the /proc/pid/status file was created the only way to read
1320 * the effective uid of a /process was to stat /proc/pid. Reading
1321 * /proc/pid/status is slow enough that procps and other packages
1322 * kept stating /proc/pid. To keep the rules in /proc simple I have
1323 * made this apply to all per process world readable and executable
1324 * directories.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 */
1326static int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1327{
1328 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001329 struct task_struct *task = get_proc_task(inode);
1330 if (task) {
1331 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1332 task_dumpable(task)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333 inode->i_uid = task->euid;
1334 inode->i_gid = task->egid;
1335 } else {
1336 inode->i_uid = 0;
1337 inode->i_gid = 0;
1338 }
1339 security_task_to_inode(task, inode);
Eric W. Biederman99f89552006-06-26 00:25:55 -07001340 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341 return 1;
1342 }
1343 d_drop(dentry);
1344 return 0;
1345}
1346
Eric W. Biederman99f89552006-06-26 00:25:55 -07001347static int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1348{
1349 struct inode *inode = dentry->d_inode;
1350 struct task_struct *task;
1351 generic_fillattr(inode, stat);
1352
1353 rcu_read_lock();
1354 stat->uid = 0;
1355 stat->gid = 0;
1356 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1357 if (task) {
1358 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1359 task_dumpable(task)) {
1360 stat->uid = task->euid;
1361 stat->gid = task->egid;
1362 }
1363 }
1364 rcu_read_unlock();
1365 return 0;
1366}
1367
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1369{
1370 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001371 struct task_struct *task = get_proc_task(inode);
Eric W. Biedermanaed7a6c2006-06-26 00:25:44 -07001372 int fd = proc_fd(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 struct files_struct *files;
1374
Eric W. Biederman99f89552006-06-26 00:25:55 -07001375 if (task) {
1376 files = get_files_struct(task);
1377 if (files) {
1378 rcu_read_lock();
1379 if (fcheck_files(files, fd)) {
1380 rcu_read_unlock();
1381 put_files_struct(files);
1382 if (task_dumpable(task)) {
1383 inode->i_uid = task->euid;
1384 inode->i_gid = task->egid;
1385 } else {
1386 inode->i_uid = 0;
1387 inode->i_gid = 0;
1388 }
1389 security_task_to_inode(task, inode);
1390 put_task_struct(task);
1391 return 1;
1392 }
Dipankar Sarmab8359962005-09-09 13:04:14 -07001393 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 put_files_struct(files);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395 }
Eric W. Biederman99f89552006-06-26 00:25:55 -07001396 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397 }
1398 d_drop(dentry);
1399 return 0;
1400}
1401
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402static int pid_delete_dentry(struct dentry * dentry)
1403{
1404 /* Is the task we represent dead?
1405 * If so, then don't put the dentry on the lru list,
1406 * kill it immediately.
1407 */
Eric W. Biederman13b41b02006-06-26 00:25:56 -07001408 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409}
1410
1411static struct dentry_operations tid_fd_dentry_operations =
1412{
1413 .d_revalidate = tid_fd_revalidate,
1414 .d_delete = pid_delete_dentry,
1415};
1416
1417static struct dentry_operations pid_dentry_operations =
1418{
1419 .d_revalidate = pid_revalidate,
1420 .d_delete = pid_delete_dentry,
1421};
1422
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423/* Lookups */
1424
1425static unsigned name_to_int(struct dentry *dentry)
1426{
1427 const char *name = dentry->d_name.name;
1428 int len = dentry->d_name.len;
1429 unsigned n = 0;
1430
1431 if (len > 1 && *name == '0')
1432 goto out;
1433 while (len-- > 0) {
1434 unsigned c = *name++ - '0';
1435 if (c > 9)
1436 goto out;
1437 if (n >= (~0U-9)/10)
1438 goto out;
1439 n *= 10;
1440 n += c;
1441 }
1442 return n;
1443out:
1444 return ~0U;
1445}
1446
1447/* SMP-safe */
1448static struct dentry *proc_lookupfd(struct inode * dir, struct dentry * dentry, struct nameidata *nd)
1449{
Eric W. Biederman99f89552006-06-26 00:25:55 -07001450 struct task_struct *task = get_proc_task(dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451 unsigned fd = name_to_int(dentry);
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001452 struct dentry *result = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 struct file * file;
1454 struct files_struct * files;
1455 struct inode *inode;
1456 struct proc_inode *ei;
1457
Eric W. Biederman99f89552006-06-26 00:25:55 -07001458 if (!task)
1459 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460 if (fd == ~0U)
1461 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462
1463 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_FD_DIR+fd);
1464 if (!inode)
1465 goto out;
1466 ei = PROC_I(inode);
Eric W. Biedermanaed7a6c2006-06-26 00:25:44 -07001467 ei->fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468 files = get_files_struct(task);
1469 if (!files)
1470 goto out_unlock;
1471 inode->i_mode = S_IFLNK;
Dipankar Sarmaca99c1d2006-04-18 22:21:46 -07001472
1473 /*
1474 * We are not taking a ref to the file structure, so we must
1475 * hold ->file_lock.
1476 */
1477 spin_lock(&files->file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 file = fcheck_files(files, fd);
1479 if (!file)
1480 goto out_unlock2;
1481 if (file->f_mode & 1)
1482 inode->i_mode |= S_IRUSR | S_IXUSR;
1483 if (file->f_mode & 2)
1484 inode->i_mode |= S_IWUSR | S_IXUSR;
Dipankar Sarmaca99c1d2006-04-18 22:21:46 -07001485 spin_unlock(&files->file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 put_files_struct(files);
1487 inode->i_op = &proc_pid_link_inode_operations;
1488 inode->i_size = 64;
1489 ei->op.proc_get_link = proc_fd_link;
1490 dentry->d_op = &tid_fd_dentry_operations;
1491 d_add(dentry, inode);
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001492 /* Close the race of the process dying before we return the dentry */
1493 if (tid_fd_revalidate(dentry, NULL))
1494 result = NULL;
1495out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07001496 put_task_struct(task);
1497out_no_task:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001498 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499
1500out_unlock2:
Dipankar Sarmaca99c1d2006-04-18 22:21:46 -07001501 spin_unlock(&files->file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502 put_files_struct(files);
1503out_unlock:
1504 iput(inode);
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001505 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506}
1507
1508static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir);
1509static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07001510static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511
1512static struct file_operations proc_fd_operations = {
1513 .read = generic_read_dir,
1514 .readdir = proc_readfd,
1515};
1516
1517static struct file_operations proc_task_operations = {
1518 .read = generic_read_dir,
1519 .readdir = proc_task_readdir,
1520};
1521
1522/*
1523 * proc directories can do almost nothing..
1524 */
1525static struct inode_operations proc_fd_inode_operations = {
1526 .lookup = proc_lookupfd,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527};
1528
1529static struct inode_operations proc_task_inode_operations = {
1530 .lookup = proc_task_lookup,
Eric W. Biederman6e66b522006-06-26 00:25:47 -07001531 .getattr = proc_task_getattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532};
1533
1534#ifdef CONFIG_SECURITY
1535static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
1536 size_t count, loff_t *ppos)
1537{
1538 struct inode * inode = file->f_dentry->d_inode;
1539 unsigned long page;
1540 ssize_t length;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001541 struct task_struct *task = get_proc_task(inode);
1542
1543 length = -ESRCH;
1544 if (!task)
1545 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546
1547 if (count > PAGE_SIZE)
1548 count = PAGE_SIZE;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001549 length = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 if (!(page = __get_free_page(GFP_KERNEL)))
Eric W. Biederman99f89552006-06-26 00:25:55 -07001551 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552
1553 length = security_getprocattr(task,
1554 (char*)file->f_dentry->d_name.name,
1555 (void*)page, count);
1556 if (length >= 0)
1557 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
1558 free_page(page);
Eric W. Biederman99f89552006-06-26 00:25:55 -07001559out:
1560 put_task_struct(task);
1561out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562 return length;
1563}
1564
1565static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
1566 size_t count, loff_t *ppos)
1567{
1568 struct inode * inode = file->f_dentry->d_inode;
1569 char *page;
1570 ssize_t length;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001571 struct task_struct *task = get_proc_task(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572
Eric W. Biederman99f89552006-06-26 00:25:55 -07001573 length = -ESRCH;
1574 if (!task)
1575 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 if (count > PAGE_SIZE)
1577 count = PAGE_SIZE;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001578
1579 /* No partial writes. */
1580 length = -EINVAL;
1581 if (*ppos != 0)
1582 goto out;
1583
1584 length = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585 page = (char*)__get_free_page(GFP_USER);
1586 if (!page)
Eric W. Biederman99f89552006-06-26 00:25:55 -07001587 goto out;
1588
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 length = -EFAULT;
1590 if (copy_from_user(page, buf, count))
Eric W. Biederman99f89552006-06-26 00:25:55 -07001591 goto out_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592
1593 length = security_setprocattr(task,
1594 (char*)file->f_dentry->d_name.name,
1595 (void*)page, count);
Eric W. Biederman99f89552006-06-26 00:25:55 -07001596out_free:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 free_page((unsigned long) page);
Eric W. Biederman99f89552006-06-26 00:25:55 -07001598out:
1599 put_task_struct(task);
1600out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 return length;
1602}
1603
1604static struct file_operations proc_pid_attr_operations = {
1605 .read = proc_pid_attr_read,
1606 .write = proc_pid_attr_write,
1607};
1608
1609static struct file_operations proc_tid_attr_operations;
1610static struct inode_operations proc_tid_attr_inode_operations;
1611static struct file_operations proc_tgid_attr_operations;
1612static struct inode_operations proc_tgid_attr_inode_operations;
1613#endif
1614
1615/* SMP-safe */
1616static struct dentry *proc_pident_lookup(struct inode *dir,
1617 struct dentry *dentry,
1618 struct pid_entry *ents)
1619{
1620 struct inode *inode;
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001621 struct dentry *error;
Eric W. Biederman99f89552006-06-26 00:25:55 -07001622 struct task_struct *task = get_proc_task(dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623 struct pid_entry *p;
1624 struct proc_inode *ei;
1625
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001626 error = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627 inode = NULL;
1628
Eric W. Biederman99f89552006-06-26 00:25:55 -07001629 if (!task)
1630 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631
1632 for (p = ents; p->name; p++) {
1633 if (p->len != dentry->d_name.len)
1634 continue;
1635 if (!memcmp(dentry->d_name.name, p->name, p->len))
1636 break;
1637 }
1638 if (!p->name)
1639 goto out;
1640
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001641 error = ERR_PTR(-EINVAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 inode = proc_pid_make_inode(dir->i_sb, task, p->type);
1643 if (!inode)
1644 goto out;
1645
1646 ei = PROC_I(inode);
1647 inode->i_mode = p->mode;
1648 /*
1649 * Yes, it does not scale. And it should not. Don't add
1650 * new entries into /proc/<tgid>/ without very good reasons.
1651 */
1652 switch(p->type) {
1653 case PROC_TGID_TASK:
Eric W. Biederman6e66b522006-06-26 00:25:47 -07001654 inode->i_nlink = 2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655 inode->i_op = &proc_task_inode_operations;
1656 inode->i_fop = &proc_task_operations;
1657 break;
1658 case PROC_TID_FD:
1659 case PROC_TGID_FD:
1660 inode->i_nlink = 2;
1661 inode->i_op = &proc_fd_inode_operations;
1662 inode->i_fop = &proc_fd_operations;
1663 break;
1664 case PROC_TID_EXE:
1665 case PROC_TGID_EXE:
1666 inode->i_op = &proc_pid_link_inode_operations;
1667 ei->op.proc_get_link = proc_exe_link;
1668 break;
1669 case PROC_TID_CWD:
1670 case PROC_TGID_CWD:
1671 inode->i_op = &proc_pid_link_inode_operations;
1672 ei->op.proc_get_link = proc_cwd_link;
1673 break;
1674 case PROC_TID_ROOT:
1675 case PROC_TGID_ROOT:
1676 inode->i_op = &proc_pid_link_inode_operations;
1677 ei->op.proc_get_link = proc_root_link;
1678 break;
1679 case PROC_TID_ENVIRON:
1680 case PROC_TGID_ENVIRON:
1681 inode->i_fop = &proc_info_file_operations;
1682 ei->op.proc_read = proc_pid_environ;
1683 break;
1684 case PROC_TID_AUXV:
1685 case PROC_TGID_AUXV:
1686 inode->i_fop = &proc_info_file_operations;
1687 ei->op.proc_read = proc_pid_auxv;
1688 break;
1689 case PROC_TID_STATUS:
1690 case PROC_TGID_STATUS:
1691 inode->i_fop = &proc_info_file_operations;
1692 ei->op.proc_read = proc_pid_status;
1693 break;
1694 case PROC_TID_STAT:
1695 inode->i_fop = &proc_info_file_operations;
1696 ei->op.proc_read = proc_tid_stat;
1697 break;
1698 case PROC_TGID_STAT:
1699 inode->i_fop = &proc_info_file_operations;
1700 ei->op.proc_read = proc_tgid_stat;
1701 break;
1702 case PROC_TID_CMDLINE:
1703 case PROC_TGID_CMDLINE:
1704 inode->i_fop = &proc_info_file_operations;
1705 ei->op.proc_read = proc_pid_cmdline;
1706 break;
1707 case PROC_TID_STATM:
1708 case PROC_TGID_STATM:
1709 inode->i_fop = &proc_info_file_operations;
1710 ei->op.proc_read = proc_pid_statm;
1711 break;
1712 case PROC_TID_MAPS:
1713 case PROC_TGID_MAPS:
1714 inode->i_fop = &proc_maps_operations;
1715 break;
Christoph Lameter6e21c8f2005-09-03 15:54:45 -07001716#ifdef CONFIG_NUMA
1717 case PROC_TID_NUMA_MAPS:
1718 case PROC_TGID_NUMA_MAPS:
1719 inode->i_fop = &proc_numa_maps_operations;
1720 break;
1721#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722 case PROC_TID_MEM:
1723 case PROC_TGID_MEM:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 inode->i_fop = &proc_mem_operations;
1725 break;
1726#ifdef CONFIG_SECCOMP
1727 case PROC_TID_SECCOMP:
1728 case PROC_TGID_SECCOMP:
1729 inode->i_fop = &proc_seccomp_operations;
1730 break;
1731#endif /* CONFIG_SECCOMP */
1732 case PROC_TID_MOUNTS:
1733 case PROC_TGID_MOUNTS:
1734 inode->i_fop = &proc_mounts_operations;
1735 break;
Yoshinori Sato63c67642005-10-14 15:59:11 -07001736#ifdef CONFIG_MMU
Mauricio Line070ad42005-09-03 15:55:10 -07001737 case PROC_TID_SMAPS:
1738 case PROC_TGID_SMAPS:
1739 inode->i_fop = &proc_smaps_operations;
1740 break;
Yoshinori Sato63c67642005-10-14 15:59:11 -07001741#endif
Chuck Leverb4629fe2006-03-20 13:44:12 -05001742 case PROC_TID_MOUNTSTATS:
1743 case PROC_TGID_MOUNTSTATS:
1744 inode->i_fop = &proc_mountstats_operations;
1745 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746#ifdef CONFIG_SECURITY
1747 case PROC_TID_ATTR:
1748 inode->i_nlink = 2;
1749 inode->i_op = &proc_tid_attr_inode_operations;
1750 inode->i_fop = &proc_tid_attr_operations;
1751 break;
1752 case PROC_TGID_ATTR:
1753 inode->i_nlink = 2;
1754 inode->i_op = &proc_tgid_attr_inode_operations;
1755 inode->i_fop = &proc_tgid_attr_operations;
1756 break;
1757 case PROC_TID_ATTR_CURRENT:
1758 case PROC_TGID_ATTR_CURRENT:
1759 case PROC_TID_ATTR_PREV:
1760 case PROC_TGID_ATTR_PREV:
1761 case PROC_TID_ATTR_EXEC:
1762 case PROC_TGID_ATTR_EXEC:
1763 case PROC_TID_ATTR_FSCREATE:
1764 case PROC_TGID_ATTR_FSCREATE:
Michael LeMay4eb582c2006-06-26 00:24:57 -07001765 case PROC_TID_ATTR_KEYCREATE:
1766 case PROC_TGID_ATTR_KEYCREATE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767 inode->i_fop = &proc_pid_attr_operations;
1768 break;
1769#endif
1770#ifdef CONFIG_KALLSYMS
1771 case PROC_TID_WCHAN:
1772 case PROC_TGID_WCHAN:
1773 inode->i_fop = &proc_info_file_operations;
1774 ei->op.proc_read = proc_pid_wchan;
1775 break;
1776#endif
1777#ifdef CONFIG_SCHEDSTATS
1778 case PROC_TID_SCHEDSTAT:
1779 case PROC_TGID_SCHEDSTAT:
1780 inode->i_fop = &proc_info_file_operations;
1781 ei->op.proc_read = proc_pid_schedstat;
1782 break;
1783#endif
1784#ifdef CONFIG_CPUSETS
1785 case PROC_TID_CPUSET:
1786 case PROC_TGID_CPUSET:
1787 inode->i_fop = &proc_cpuset_operations;
1788 break;
1789#endif
1790 case PROC_TID_OOM_SCORE:
1791 case PROC_TGID_OOM_SCORE:
1792 inode->i_fop = &proc_info_file_operations;
1793 ei->op.proc_read = proc_oom_score;
1794 break;
1795 case PROC_TID_OOM_ADJUST:
1796 case PROC_TGID_OOM_ADJUST:
1797 inode->i_fop = &proc_oom_adjust_operations;
1798 break;
1799#ifdef CONFIG_AUDITSYSCALL
1800 case PROC_TID_LOGINUID:
1801 case PROC_TGID_LOGINUID:
1802 inode->i_fop = &proc_loginuid_operations;
1803 break;
1804#endif
1805 default:
1806 printk("procfs: impossible type (%d)",p->type);
1807 iput(inode);
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001808 error = ERR_PTR(-EINVAL);
1809 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810 }
1811 dentry->d_op = &pid_dentry_operations;
1812 d_add(dentry, inode);
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001813 /* Close the race of the process dying before we return the dentry */
1814 if (pid_revalidate(dentry, NULL))
1815 error = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07001817 put_task_struct(task);
1818out_no_task:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001819 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820}
1821
1822static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1823 return proc_pident_lookup(dir, dentry, tgid_base_stuff);
1824}
1825
1826static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
1827 return proc_pident_lookup(dir, dentry, tid_base_stuff);
1828}
1829
1830static struct file_operations proc_tgid_base_operations = {
1831 .read = generic_read_dir,
1832 .readdir = proc_tgid_base_readdir,
1833};
1834
1835static struct file_operations proc_tid_base_operations = {
1836 .read = generic_read_dir,
1837 .readdir = proc_tid_base_readdir,
1838};
1839
1840static struct inode_operations proc_tgid_base_inode_operations = {
1841 .lookup = proc_tgid_base_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07001842 .getattr = pid_getattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843};
1844
1845static struct inode_operations proc_tid_base_inode_operations = {
1846 .lookup = proc_tid_base_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07001847 .getattr = pid_getattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848};
1849
1850#ifdef CONFIG_SECURITY
1851static int proc_tgid_attr_readdir(struct file * filp,
1852 void * dirent, filldir_t filldir)
1853{
1854 return proc_pident_readdir(filp,dirent,filldir,
1855 tgid_attr_stuff,ARRAY_SIZE(tgid_attr_stuff));
1856}
1857
1858static int proc_tid_attr_readdir(struct file * filp,
1859 void * dirent, filldir_t filldir)
1860{
1861 return proc_pident_readdir(filp,dirent,filldir,
1862 tid_attr_stuff,ARRAY_SIZE(tid_attr_stuff));
1863}
1864
1865static struct file_operations proc_tgid_attr_operations = {
1866 .read = generic_read_dir,
1867 .readdir = proc_tgid_attr_readdir,
1868};
1869
1870static struct file_operations proc_tid_attr_operations = {
1871 .read = generic_read_dir,
1872 .readdir = proc_tid_attr_readdir,
1873};
1874
1875static struct dentry *proc_tgid_attr_lookup(struct inode *dir,
1876 struct dentry *dentry, struct nameidata *nd)
1877{
1878 return proc_pident_lookup(dir, dentry, tgid_attr_stuff);
1879}
1880
1881static struct dentry *proc_tid_attr_lookup(struct inode *dir,
1882 struct dentry *dentry, struct nameidata *nd)
1883{
1884 return proc_pident_lookup(dir, dentry, tid_attr_stuff);
1885}
1886
1887static struct inode_operations proc_tgid_attr_inode_operations = {
1888 .lookup = proc_tgid_attr_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07001889 .getattr = pid_getattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001890};
1891
1892static struct inode_operations proc_tid_attr_inode_operations = {
1893 .lookup = proc_tid_attr_lookup,
Eric W. Biederman99f89552006-06-26 00:25:55 -07001894 .getattr = pid_getattr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895};
1896#endif
1897
1898/*
1899 * /proc/self:
1900 */
1901static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
1902 int buflen)
1903{
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001904 char tmp[PROC_NUMBUF];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905 sprintf(tmp, "%d", current->tgid);
1906 return vfs_readlink(dentry,buffer,buflen,tmp);
1907}
1908
Al Viro008b1502005-08-20 00:17:39 +01001909static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910{
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001911 char tmp[PROC_NUMBUF];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912 sprintf(tmp, "%d", current->tgid);
Al Viro008b1502005-08-20 00:17:39 +01001913 return ERR_PTR(vfs_follow_link(nd,tmp));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914}
1915
1916static struct inode_operations proc_self_inode_operations = {
1917 .readlink = proc_self_readlink,
1918 .follow_link = proc_self_follow_link,
1919};
1920
1921/**
Eric W. Biederman48e64842006-06-26 00:25:48 -07001922 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 *
Eric W. Biederman48e64842006-06-26 00:25:48 -07001924 * @task: task that should be flushed.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925 *
Eric W. Biederman48e64842006-06-26 00:25:48 -07001926 * Looks in the dcache for
1927 * /proc/@pid
1928 * /proc/@tgid/task/@pid
1929 * if either directory is present flushes it and all of it'ts children
1930 * from the dcache.
1931 *
1932 * It is safe and reasonable to cache /proc entries for a task until
1933 * that task exits. After that they just clog up the dcache with
1934 * useless entries, possibly causing useful dcache entries to be
1935 * flushed instead. This routine is proved to flush those useless
1936 * dcache entries at process exit time.
1937 *
1938 * NOTE: This routine is just an optimization so it does not guarantee
1939 * that no dcache entries will exist at process exit time it
1940 * just makes it very unlikely that any will persist.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 */
Eric W. Biederman48e64842006-06-26 00:25:48 -07001942void proc_flush_task(struct task_struct *task)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943{
Eric W. Biederman48e64842006-06-26 00:25:48 -07001944 struct dentry *dentry, *leader, *dir;
Eric W. Biederman8578cea2006-06-26 00:25:54 -07001945 char buf[PROC_NUMBUF];
Eric W. Biederman48e64842006-06-26 00:25:48 -07001946 struct qstr name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947
Eric W. Biederman48e64842006-06-26 00:25:48 -07001948 name.name = buf;
1949 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
1950 dentry = d_hash_and_lookup(proc_mnt->mnt_root, &name);
1951 if (dentry) {
1952 shrink_dcache_parent(dentry);
1953 d_drop(dentry);
1954 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956
Eric W. Biederman48e64842006-06-26 00:25:48 -07001957 if (thread_group_leader(task))
1958 goto out;
1959
1960 name.name = buf;
1961 name.len = snprintf(buf, sizeof(buf), "%d", task->tgid);
1962 leader = d_hash_and_lookup(proc_mnt->mnt_root, &name);
1963 if (!leader)
1964 goto out;
1965
1966 name.name = "task";
1967 name.len = strlen(name.name);
1968 dir = d_hash_and_lookup(leader, &name);
1969 if (!dir)
1970 goto out_put_leader;
1971
1972 name.name = buf;
1973 name.len = snprintf(buf, sizeof(buf), "%d", task->pid);
1974 dentry = d_hash_and_lookup(dir, &name);
1975 if (dentry) {
1976 shrink_dcache_parent(dentry);
1977 d_drop(dentry);
1978 dput(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979 }
Eric W. Biederman48e64842006-06-26 00:25:48 -07001980
1981 dput(dir);
1982out_put_leader:
1983 dput(leader);
1984out:
1985 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986}
1987
1988/* SMP-safe */
1989struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
1990{
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07001991 struct dentry *result = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992 struct task_struct *task;
1993 struct inode *inode;
1994 struct proc_inode *ei;
1995 unsigned tgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996
1997 if (dentry->d_name.len == 4 && !memcmp(dentry->d_name.name,"self",4)) {
1998 inode = new_inode(dir->i_sb);
1999 if (!inode)
2000 return ERR_PTR(-ENOMEM);
2001 ei = PROC_I(inode);
2002 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2003 inode->i_ino = fake_ino(0, PROC_TGID_INO);
2004 ei->pde = NULL;
2005 inode->i_mode = S_IFLNK|S_IRWXUGO;
2006 inode->i_uid = inode->i_gid = 0;
2007 inode->i_size = 64;
2008 inode->i_op = &proc_self_inode_operations;
2009 d_add(dentry, inode);
2010 return NULL;
2011 }
2012 tgid = name_to_int(dentry);
2013 if (tgid == ~0U)
2014 goto out;
2015
Eric W. Biedermande758732006-06-26 00:25:51 -07002016 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017 task = find_task_by_pid(tgid);
2018 if (task)
2019 get_task_struct(task);
Eric W. Biedermande758732006-06-26 00:25:51 -07002020 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 if (!task)
2022 goto out;
2023
2024 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TGID_INO);
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002025 if (!inode)
2026 goto out_put_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2029 inode->i_op = &proc_tgid_base_inode_operations;
2030 inode->i_fop = &proc_tgid_base_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 inode->i_flags|=S_IMMUTABLE;
Daniel Drakebcf88e12005-05-01 08:59:03 -07002032#ifdef CONFIG_SECURITY
2033 inode->i_nlink = 5;
2034#else
2035 inode->i_nlink = 4;
2036#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037
Eric W. Biederman48e64842006-06-26 00:25:48 -07002038 dentry->d_op = &pid_dentry_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040 d_add(dentry, inode);
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002041 /* Close the race of the process dying before we return the dentry */
2042 if (pid_revalidate(dentry, NULL))
2043 result = NULL;
Eric W. Biederman48e64842006-06-26 00:25:48 -07002044
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002045out_put_task:
Eric W. Biederman48e64842006-06-26 00:25:48 -07002046 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047out:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002048 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049}
2050
2051/* SMP-safe */
2052static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
2053{
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002054 struct dentry *result = ERR_PTR(-ENOENT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055 struct task_struct *task;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002056 struct task_struct *leader = get_proc_task(dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057 struct inode *inode;
2058 unsigned tid;
2059
Eric W. Biederman99f89552006-06-26 00:25:55 -07002060 if (!leader)
2061 goto out_no_task;
2062
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063 tid = name_to_int(dentry);
2064 if (tid == ~0U)
2065 goto out;
2066
Eric W. Biedermande758732006-06-26 00:25:51 -07002067 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002068 task = find_task_by_pid(tid);
2069 if (task)
2070 get_task_struct(task);
Eric W. Biedermande758732006-06-26 00:25:51 -07002071 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072 if (!task)
2073 goto out;
2074 if (leader->tgid != task->tgid)
2075 goto out_drop_task;
2076
2077 inode = proc_pid_make_inode(dir->i_sb, task, PROC_TID_INO);
2078
2079
2080 if (!inode)
2081 goto out_drop_task;
2082 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
2083 inode->i_op = &proc_tid_base_inode_operations;
2084 inode->i_fop = &proc_tid_base_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085 inode->i_flags|=S_IMMUTABLE;
Daniel Drakebcf88e12005-05-01 08:59:03 -07002086#ifdef CONFIG_SECURITY
2087 inode->i_nlink = 4;
2088#else
2089 inode->i_nlink = 3;
2090#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091
Eric W. Biederman48e64842006-06-26 00:25:48 -07002092 dentry->d_op = &pid_dentry_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093
2094 d_add(dentry, inode);
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002095 /* Close the race of the process dying before we return the dentry */
2096 if (pid_revalidate(dentry, NULL))
2097 result = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099out_drop_task:
2100 put_task_struct(task);
2101out:
Eric W. Biederman99f89552006-06-26 00:25:55 -07002102 put_task_struct(leader);
2103out_no_task:
Eric W. Biedermancd6a3ce2006-06-26 00:25:49 -07002104 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105}
2106
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107/*
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002108 * Find the first tgid to return to user space.
2109 *
2110 * Usually this is just whatever follows &init_task, but if the users
2111 * buffer was too small to hold the full list or there was a seek into
2112 * the middle of the directory we have more work to do.
2113 *
2114 * In the case of a short read we start with find_task_by_pid.
2115 *
2116 * In the case of a seek we start with &init_task and walk nr
2117 * threads past it.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118 */
Eric W. Biederman9cc8cbc2006-06-26 00:25:52 -07002119static struct task_struct *first_tgid(int tgid, unsigned int nr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120{
Eric W. Biederman9cc8cbc2006-06-26 00:25:52 -07002121 struct task_struct *pos;
Eric W. Biederman454cc102006-06-26 00:25:51 -07002122 rcu_read_lock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002123 if (tgid && nr) {
2124 pos = find_task_by_pid(tgid);
Eric W. Biederman9cc8cbc2006-06-26 00:25:52 -07002125 if (pos && thread_group_leader(pos))
2126 goto found;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127 }
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002128 /* If nr exceeds the number of processes get out quickly */
Eric W. Biederman9cc8cbc2006-06-26 00:25:52 -07002129 pos = NULL;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002130 if (nr && nr >= nr_processes())
2131 goto done;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002132
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002133 /* If we haven't found our starting place yet start with
2134 * the init_task and walk nr tasks forward.
2135 */
Eric W. Biederman9cc8cbc2006-06-26 00:25:52 -07002136 for (pos = next_task(&init_task); nr > 0; --nr) {
2137 pos = next_task(pos);
2138 if (pos == &init_task) {
2139 pos = NULL;
2140 goto done;
2141 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002142 }
Eric W. Biederman9cc8cbc2006-06-26 00:25:52 -07002143found:
2144 get_task_struct(pos);
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002145done:
Eric W. Biederman454cc102006-06-26 00:25:51 -07002146 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002147 return pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002148}
2149
2150/*
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002151 * Find the next task in the task list.
2152 * Return NULL if we loop or there is any error.
2153 *
2154 * The reference to the input task_struct is released.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155 */
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002156static struct task_struct *next_tgid(struct task_struct *start)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157{
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002158 struct task_struct *pos;
Eric W. Biederman454cc102006-06-26 00:25:51 -07002159 rcu_read_lock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002160 pos = start;
2161 if (pid_alive(start))
2162 pos = next_task(start);
2163 if (pid_alive(pos) && (pos != &init_task)) {
2164 get_task_struct(pos);
2165 goto done;
2166 }
2167 pos = NULL;
2168done:
Eric W. Biederman454cc102006-06-26 00:25:51 -07002169 rcu_read_unlock();
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002170 put_task_struct(start);
2171 return pos;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002172}
2173
2174/* for the /proc/ directory itself, after non-process stuff has been done */
2175int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
2176{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002177 char buf[PROC_NUMBUF];
2178 unsigned int nr = filp->f_pos - FIRST_PROCESS_ENTRY;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002179 struct task_struct *task;
2180 int tgid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181
2182 if (!nr) {
2183 ino_t ino = fake_ino(0,PROC_TGID_INO);
2184 if (filldir(dirent, "self", 4, filp->f_pos, ino, DT_LNK) < 0)
2185 return 0;
2186 filp->f_pos++;
2187 nr++;
2188 }
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002189 nr -= 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002190
2191 /* f_version caches the tgid value that the last readdir call couldn't
2192 * return. lseek aka telldir automagically resets f_version to 0.
2193 */
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002194 tgid = filp->f_version;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195 filp->f_version = 0;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002196 for (task = first_tgid(tgid, nr);
2197 task;
2198 task = next_tgid(task), filp->f_pos++) {
2199 int len;
2200 ino_t ino;
2201 tgid = task->pid;
2202 len = snprintf(buf, sizeof(buf), "%d", tgid);
2203 ino = fake_ino(tgid, PROC_TGID_INO);
2204 if (filldir(dirent, buf, len, filp->f_pos, ino, DT_DIR) < 0) {
2205 /* returning this tgid failed, save it as the first
2206 * pid for the next readir call */
2207 filp->f_version = tgid;
2208 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002209 break;
2210 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212 return 0;
2213}
2214
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002215/*
2216 * Find the first tid of a thread group to return to user space.
2217 *
2218 * Usually this is just the thread group leader, but if the users
2219 * buffer was too small or there was a seek into the middle of the
2220 * directory we have more work todo.
2221 *
2222 * In the case of a short read we start with find_task_by_pid.
2223 *
2224 * In the case of a seek we start with the leader and walk nr
2225 * threads past it.
2226 */
2227static struct task_struct *first_tid(struct task_struct *leader, int tid, int nr)
2228{
2229 struct task_struct *pos = NULL;
2230 read_lock(&tasklist_lock);
2231
2232 /* Attempt to start with the pid of a thread */
2233 if (tid && (nr > 0)) {
2234 pos = find_task_by_pid(tid);
2235 if (pos && (pos->group_leader != leader))
2236 pos = NULL;
2237 if (pos)
2238 nr = 0;
2239 }
2240
2241 /* If nr exceeds the number of threads there is nothing todo */
2242 if (nr) {
Eric W. Biederman99f89552006-06-26 00:25:55 -07002243 if (nr >= get_nr_threads(leader))
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002244 goto done;
2245 }
2246
2247 /* If we haven't found our starting place yet start with the
2248 * leader and walk nr threads forward.
2249 */
2250 if (!pos && (nr >= 0))
2251 pos = leader;
2252
2253 for (; pos && pid_alive(pos); pos = next_thread(pos)) {
2254 if (--nr > 0)
2255 continue;
2256 get_task_struct(pos);
2257 goto done;
2258 }
2259 pos = NULL;
2260done:
2261 read_unlock(&tasklist_lock);
2262 return pos;
2263}
2264
2265/*
2266 * Find the next thread in the thread list.
2267 * Return NULL if there is an error or no next thread.
2268 *
2269 * The reference to the input task_struct is released.
2270 */
2271static struct task_struct *next_tid(struct task_struct *start)
2272{
2273 struct task_struct *pos;
2274 read_lock(&tasklist_lock);
2275 pos = start;
2276 if (pid_alive(start))
2277 pos = next_thread(start);
2278 if (pid_alive(pos) && (pos != start->group_leader))
2279 get_task_struct(pos);
2280 else
2281 pos = NULL;
2282 read_unlock(&tasklist_lock);
2283 put_task_struct(start);
2284 return pos;
2285}
2286
Linus Torvalds1da177e2005-04-16 15:20:36 -07002287/* for the /proc/TGID/task/ directories */
2288static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
2289{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002290 char buf[PROC_NUMBUF];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002291 struct dentry *dentry = filp->f_dentry;
2292 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002293 struct task_struct *leader = get_proc_task(inode);
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002294 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002295 int retval = -ENOENT;
2296 ino_t ino;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002297 int tid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002298 unsigned long pos = filp->f_pos; /* avoiding "long long" filp->f_pos */
2299
Eric W. Biederman99f89552006-06-26 00:25:55 -07002300 if (!leader)
2301 goto out_no_task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002302 retval = 0;
2303
2304 switch (pos) {
2305 case 0:
2306 ino = inode->i_ino;
2307 if (filldir(dirent, ".", 1, pos, ino, DT_DIR) < 0)
2308 goto out;
2309 pos++;
2310 /* fall through */
2311 case 1:
2312 ino = parent_ino(dentry);
2313 if (filldir(dirent, "..", 2, pos, ino, DT_DIR) < 0)
2314 goto out;
2315 pos++;
2316 /* fall through */
2317 }
2318
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002319 /* f_version caches the tgid value that the last readdir call couldn't
2320 * return. lseek aka telldir automagically resets f_version to 0.
2321 */
2322 tid = filp->f_version;
2323 filp->f_version = 0;
2324 for (task = first_tid(leader, tid, pos - 2);
2325 task;
2326 task = next_tid(task), pos++) {
2327 int len;
2328 tid = task->pid;
2329 len = snprintf(buf, sizeof(buf), "%d", tid);
2330 ino = fake_ino(tid, PROC_TID_INO);
2331 if (filldir(dirent, buf, len, pos, ino, DT_DIR < 0)) {
2332 /* returning this tgid failed, save it as the first
2333 * pid for the next readir call */
2334 filp->f_version = tid;
2335 put_task_struct(task);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002336 break;
Eric W. Biederman0bc58a92006-06-26 00:25:50 -07002337 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002338 }
2339out:
2340 filp->f_pos = pos;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002341 put_task_struct(leader);
2342out_no_task:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343 return retval;
2344}
Eric W. Biederman6e66b522006-06-26 00:25:47 -07002345
2346static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
2347{
2348 struct inode *inode = dentry->d_inode;
Eric W. Biederman99f89552006-06-26 00:25:55 -07002349 struct task_struct *p = get_proc_task(inode);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07002350 generic_fillattr(inode, stat);
2351
Eric W. Biederman99f89552006-06-26 00:25:55 -07002352 if (p) {
2353 rcu_read_lock();
2354 stat->nlink += get_nr_threads(p);
2355 rcu_read_unlock();
2356 put_task_struct(p);
Eric W. Biederman6e66b522006-06-26 00:25:47 -07002357 }
2358
2359 return 0;
2360}