blob: 72c82f69b01b28594e56bb9518df6f211f0d51a9 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/fcntl.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7#include <linux/syscalls.h>
8#include <linux/init.h>
9#include <linux/mm.h>
10#include <linux/fs.h>
11#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040012#include <linux/fdtable.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080013#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/dnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015#include <linux/slab.h>
16#include <linux/module.h>
Jens Axboe35f3d142010-05-20 10:43:18 +020017#include <linux/pipe_fs_i.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include <linux/security.h>
19#include <linux/ptrace.h>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070020#include <linux/signal.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070021#include <linux/rcupdate.h>
Pavel Emelyanovb4888932007-10-18 23:40:14 -070022#include <linux/pid_namespace.h>
Cyrill Gorcunov1d151c32012-07-30 14:43:00 -070023#include <linux/user_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024
25#include <asm/poll.h>
26#include <asm/siginfo.h>
27#include <asm/uaccess.h>
28
Jonathan Corbet76398422009-02-01 14:26:59 -070029#define SETFL_MASK (O_APPEND | O_NONBLOCK | O_NDELAY | O_DIRECT | O_NOATIME)
Linus Torvalds1da177e2005-04-16 15:20:36 -070030
31static int setfl(int fd, struct file * filp, unsigned long arg)
32{
Al Viro496ad9a2013-01-23 17:07:38 -050033 struct inode * inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -070034 int error = 0;
35
dean gaudet7d95c8f2006-02-03 03:04:30 -080036 /*
37 * O_APPEND cannot be cleared if the file is marked as append-only
38 * and the file is open for write.
39 */
40 if (((arg ^ filp->f_flags) & O_APPEND) && IS_APPEND(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -070041 return -EPERM;
42
43 /* O_NOATIME can only be set by the owner or superuser */
44 if ((arg & O_NOATIME) && !(filp->f_flags & O_NOATIME))
Serge E. Hallyn2e149672011-03-23 16:43:26 -070045 if (!inode_owner_or_capable(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -070046 return -EPERM;
47
48 /* required for strict SunOS emulation */
49 if (O_NONBLOCK != O_NDELAY)
50 if (arg & O_NDELAY)
51 arg |= O_NONBLOCK;
52
53 if (arg & O_DIRECT) {
54 if (!filp->f_mapping || !filp->f_mapping->a_ops ||
55 !filp->f_mapping->a_ops->direct_IO)
56 return -EINVAL;
57 }
58
Al Viro72c2d532013-09-22 16:27:52 -040059 if (filp->f_op->check_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -070060 error = filp->f_op->check_flags(arg);
61 if (error)
62 return error;
63
Jonathan Corbet218d11a2008-12-05 16:12:48 -070064 /*
Jonathan Corbet76398422009-02-01 14:26:59 -070065 * ->fasync() is responsible for setting the FASYNC bit.
Jonathan Corbet218d11a2008-12-05 16:12:48 -070066 */
Al Viro72c2d532013-09-22 16:27:52 -040067 if (((arg ^ filp->f_flags) & FASYNC) && filp->f_op->fasync) {
Jonathan Corbet76398422009-02-01 14:26:59 -070068 error = filp->f_op->fasync(fd, filp, (arg & FASYNC) != 0);
69 if (error < 0)
70 goto out;
Jonathan Corbet60aa4922009-02-01 14:52:56 -070071 if (error > 0)
72 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070073 }
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -070074 spin_lock(&filp->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070075 filp->f_flags = (arg & SETFL_MASK) | (filp->f_flags & ~SETFL_MASK);
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -070076 spin_unlock(&filp->f_lock);
Jonathan Corbet76398422009-02-01 14:26:59 -070077
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 out:
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 return error;
80}
81
Eric W. Biederman609d7fa2006-10-02 02:17:15 -070082static void f_modown(struct file *filp, struct pid *pid, enum pid_type type,
Oleg Nesterov2f38d702009-06-16 22:07:46 +020083 int force)
Linus Torvalds1da177e2005-04-16 15:20:36 -070084{
Linus Torvalds80e1e822010-02-07 10:11:23 -080085 write_lock_irq(&filp->f_owner.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070086 if (force || !filp->f_owner.pid) {
Eric W. Biederman609d7fa2006-10-02 02:17:15 -070087 put_pid(filp->f_owner.pid);
88 filp->f_owner.pid = get_pid(pid);
89 filp->f_owner.pid_type = type;
Oleg Nesterov2f38d702009-06-16 22:07:46 +020090
91 if (pid) {
92 const struct cred *cred = current_cred();
93 filp->f_owner.uid = cred->uid;
94 filp->f_owner.euid = cred->euid;
95 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070096 }
Linus Torvalds80e1e822010-02-07 10:11:23 -080097 write_unlock_irq(&filp->f_owner.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070098}
99
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700100int __f_setown(struct file *filp, struct pid *pid, enum pid_type type,
101 int force)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102{
103 int err;
Oleg Nesterov2f38d702009-06-16 22:07:46 +0200104
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 err = security_file_set_fowner(filp);
106 if (err)
107 return err;
108
Oleg Nesterov2f38d702009-06-16 22:07:46 +0200109 f_modown(filp, pid, type, force);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110 return 0;
111}
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700112EXPORT_SYMBOL(__f_setown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700114int f_setown(struct file *filp, unsigned long arg, int force)
115{
116 enum pid_type type;
117 struct pid *pid;
118 int who = arg;
119 int result;
120 type = PIDTYPE_PID;
121 if (who < 0) {
122 type = PIDTYPE_PGID;
123 who = -who;
124 }
125 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700126 pid = find_vpid(who);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700127 result = __f_setown(filp, pid, type, force);
128 rcu_read_unlock();
129 return result;
130}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131EXPORT_SYMBOL(f_setown);
132
133void f_delown(struct file *filp)
134{
Oleg Nesterov2f38d702009-06-16 22:07:46 +0200135 f_modown(filp, NULL, PIDTYPE_PID, 1);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700136}
137
138pid_t f_getown(struct file *filp)
139{
140 pid_t pid;
Eric W. Biederman43fa1ad2006-10-02 02:17:27 -0700141 read_lock(&filp->f_owner.lock);
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800142 pid = pid_vnr(filp->f_owner.pid);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700143 if (filp->f_owner.pid_type == PIDTYPE_PGID)
144 pid = -pid;
Eric W. Biederman43fa1ad2006-10-02 02:17:27 -0700145 read_unlock(&filp->f_owner.lock);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700146 return pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147}
148
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700149static int f_setown_ex(struct file *filp, unsigned long arg)
150{
Al Viro63784dd2012-09-26 21:43:05 -0400151 struct f_owner_ex __user *owner_p = (void __user *)arg;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700152 struct f_owner_ex owner;
153 struct pid *pid;
154 int type;
155 int ret;
156
157 ret = copy_from_user(&owner, owner_p, sizeof(owner));
158 if (ret)
Dan Carpenter5b544702010-06-03 12:35:42 +0200159 return -EFAULT;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700160
161 switch (owner.type) {
162 case F_OWNER_TID:
163 type = PIDTYPE_MAX;
164 break;
165
166 case F_OWNER_PID:
167 type = PIDTYPE_PID;
168 break;
169
Peter Zijlstra978b4052009-11-17 14:06:24 -0800170 case F_OWNER_PGRP:
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700171 type = PIDTYPE_PGID;
172 break;
173
174 default:
175 return -EINVAL;
176 }
177
178 rcu_read_lock();
179 pid = find_vpid(owner.pid);
180 if (owner.pid && !pid)
181 ret = -ESRCH;
182 else
183 ret = __f_setown(filp, pid, type, 1);
184 rcu_read_unlock();
185
186 return ret;
187}
188
189static int f_getown_ex(struct file *filp, unsigned long arg)
190{
Al Viro63784dd2012-09-26 21:43:05 -0400191 struct f_owner_ex __user *owner_p = (void __user *)arg;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700192 struct f_owner_ex owner;
193 int ret = 0;
194
195 read_lock(&filp->f_owner.lock);
196 owner.pid = pid_vnr(filp->f_owner.pid);
197 switch (filp->f_owner.pid_type) {
198 case PIDTYPE_MAX:
199 owner.type = F_OWNER_TID;
200 break;
201
202 case PIDTYPE_PID:
203 owner.type = F_OWNER_PID;
204 break;
205
206 case PIDTYPE_PGID:
Peter Zijlstra978b4052009-11-17 14:06:24 -0800207 owner.type = F_OWNER_PGRP;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700208 break;
209
210 default:
211 WARN_ON(1);
212 ret = -EINVAL;
213 break;
214 }
215 read_unlock(&filp->f_owner.lock);
216
Dan Carpenter5b544702010-06-03 12:35:42 +0200217 if (!ret) {
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700218 ret = copy_to_user(owner_p, &owner, sizeof(owner));
Dan Carpenter5b544702010-06-03 12:35:42 +0200219 if (ret)
220 ret = -EFAULT;
221 }
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700222 return ret;
223}
224
Cyrill Gorcunov1d151c32012-07-30 14:43:00 -0700225#ifdef CONFIG_CHECKPOINT_RESTORE
226static int f_getowner_uids(struct file *filp, unsigned long arg)
227{
228 struct user_namespace *user_ns = current_user_ns();
Al Viro63784dd2012-09-26 21:43:05 -0400229 uid_t __user *dst = (void __user *)arg;
Cyrill Gorcunov1d151c32012-07-30 14:43:00 -0700230 uid_t src[2];
231 int err;
232
233 read_lock(&filp->f_owner.lock);
234 src[0] = from_kuid(user_ns, filp->f_owner.uid);
235 src[1] = from_kuid(user_ns, filp->f_owner.euid);
236 read_unlock(&filp->f_owner.lock);
237
238 err = put_user(src[0], &dst[0]);
239 err |= put_user(src[1], &dst[1]);
240
241 return err;
242}
243#else
244static int f_getowner_uids(struct file *filp, unsigned long arg)
245{
246 return -EINVAL;
247}
248#endif
249
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250static long do_fcntl(int fd, unsigned int cmd, unsigned long arg,
251 struct file *filp)
252{
253 long err = -EINVAL;
254
255 switch (cmd) {
256 case F_DUPFD:
Al Virofe17f222012-08-21 11:48:11 -0400257 err = f_dupfd(arg, filp, 0);
258 break;
Ulrich Drepper22d2b352007-10-16 23:30:26 -0700259 case F_DUPFD_CLOEXEC:
Al Viro12197712012-10-08 23:21:58 +0100260 err = f_dupfd(arg, filp, O_CLOEXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 break;
262 case F_GETFD:
263 err = get_close_on_exec(fd) ? FD_CLOEXEC : 0;
264 break;
265 case F_SETFD:
266 err = 0;
267 set_close_on_exec(fd, arg & FD_CLOEXEC);
268 break;
269 case F_GETFL:
270 err = filp->f_flags;
271 break;
272 case F_SETFL:
273 err = setfl(fd, filp, arg);
274 break;
Jeff Layton5d50ffd2014-02-03 12:13:10 -0500275#if BITS_PER_LONG != 32
276 /* 32-bit arches must use fcntl64() */
Jeff Layton0d3f7a22014-04-22 08:23:58 -0400277 case F_OFD_GETLK:
Jeff Layton5d50ffd2014-02-03 12:13:10 -0500278#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 case F_GETLK:
Jeff Laytonc1e62b82014-02-03 12:13:09 -0500280 err = fcntl_getlk(filp, cmd, (struct flock __user *) arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 break;
Jeff Layton5d50ffd2014-02-03 12:13:10 -0500282#if BITS_PER_LONG != 32
283 /* 32-bit arches must use fcntl64() */
Jeff Layton0d3f7a22014-04-22 08:23:58 -0400284 case F_OFD_SETLK:
285 case F_OFD_SETLKW:
Jeff Layton5d50ffd2014-02-03 12:13:10 -0500286#endif
287 /* Fallthrough */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 case F_SETLK:
289 case F_SETLKW:
Peter Staubachc2936212005-07-27 11:45:09 -0700290 err = fcntl_setlk(fd, filp, cmd, (struct flock __user *) arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 break;
292 case F_GETOWN:
293 /*
294 * XXX If f_owner is a process group, the
295 * negative return value will get converted
296 * into an error. Oops. If we keep the
297 * current syscall conventions, the only way
298 * to fix this will be in libc.
299 */
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700300 err = f_getown(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301 force_successful_syscall_return();
302 break;
303 case F_SETOWN:
304 err = f_setown(filp, arg, 1);
305 break;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700306 case F_GETOWN_EX:
307 err = f_getown_ex(filp, arg);
308 break;
309 case F_SETOWN_EX:
310 err = f_setown_ex(filp, arg);
311 break;
Cyrill Gorcunov1d151c32012-07-30 14:43:00 -0700312 case F_GETOWNER_UIDS:
313 err = f_getowner_uids(filp, arg);
314 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 case F_GETSIG:
316 err = filp->f_owner.signum;
317 break;
318 case F_SETSIG:
319 /* arg == 0 restores default behaviour. */
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700320 if (!valid_signal(arg)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 break;
322 }
323 err = 0;
324 filp->f_owner.signum = arg;
325 break;
326 case F_GETLEASE:
327 err = fcntl_getlease(filp);
328 break;
329 case F_SETLEASE:
330 err = fcntl_setlease(fd, filp, arg);
331 break;
332 case F_NOTIFY:
333 err = fcntl_dirnotify(fd, filp, arg);
334 break;
Jens Axboe35f3d142010-05-20 10:43:18 +0200335 case F_SETPIPE_SZ:
336 case F_GETPIPE_SZ:
337 err = pipe_fcntl(filp, cmd, arg);
338 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339 default:
340 break;
341 }
342 return err;
343}
344
Al Viro1abf0c72011-03-13 03:51:11 -0400345static int check_fcntl_cmd(unsigned cmd)
346{
347 switch (cmd) {
348 case F_DUPFD:
349 case F_DUPFD_CLOEXEC:
350 case F_GETFD:
351 case F_SETFD:
352 case F_GETFL:
353 return 1;
354 }
355 return 0;
356}
357
Heiko Carstensa26eab22009-01-14 14:14:17 +0100358SYSCALL_DEFINE3(fcntl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359{
Al Viro2903ff02012-08-28 12:52:22 -0400360 struct fd f = fdget_raw(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 long err = -EBADF;
362
Al Viro2903ff02012-08-28 12:52:22 -0400363 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 goto out;
365
Al Viro2903ff02012-08-28 12:52:22 -0400366 if (unlikely(f.file->f_mode & FMODE_PATH)) {
Al Viro545ec2c2012-04-21 18:42:19 -0400367 if (!check_fcntl_cmd(cmd))
368 goto out1;
Al Viro1abf0c72011-03-13 03:51:11 -0400369 }
370
Al Viro2903ff02012-08-28 12:52:22 -0400371 err = security_file_fcntl(f.file, cmd, arg);
Al Viro545ec2c2012-04-21 18:42:19 -0400372 if (!err)
Al Viro2903ff02012-08-28 12:52:22 -0400373 err = do_fcntl(fd, cmd, arg, f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374
Al Viro545ec2c2012-04-21 18:42:19 -0400375out1:
Al Viro2903ff02012-08-28 12:52:22 -0400376 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377out:
378 return err;
379}
380
381#if BITS_PER_LONG == 32
Heiko Carstensa26eab22009-01-14 14:14:17 +0100382SYSCALL_DEFINE3(fcntl64, unsigned int, fd, unsigned int, cmd,
383 unsigned long, arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384{
Al Viro2903ff02012-08-28 12:52:22 -0400385 struct fd f = fdget_raw(fd);
Al Viro545ec2c2012-04-21 18:42:19 -0400386 long err = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387
Al Viro2903ff02012-08-28 12:52:22 -0400388 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 goto out;
390
Al Viro2903ff02012-08-28 12:52:22 -0400391 if (unlikely(f.file->f_mode & FMODE_PATH)) {
Al Viro545ec2c2012-04-21 18:42:19 -0400392 if (!check_fcntl_cmd(cmd))
393 goto out1;
Al Viro1abf0c72011-03-13 03:51:11 -0400394 }
395
Al Viro2903ff02012-08-28 12:52:22 -0400396 err = security_file_fcntl(f.file, cmd, arg);
Al Viro545ec2c2012-04-21 18:42:19 -0400397 if (err)
398 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399
400 switch (cmd) {
Jeff Layton5d50ffd2014-02-03 12:13:10 -0500401 case F_GETLK64:
Jeff Layton0d3f7a22014-04-22 08:23:58 -0400402 case F_OFD_GETLK:
Jeff Layton5d50ffd2014-02-03 12:13:10 -0500403 err = fcntl_getlk64(f.file, cmd, (struct flock64 __user *) arg);
404 break;
405 case F_SETLK64:
406 case F_SETLKW64:
Jeff Layton0d3f7a22014-04-22 08:23:58 -0400407 case F_OFD_SETLK:
408 case F_OFD_SETLKW:
Jeff Layton5d50ffd2014-02-03 12:13:10 -0500409 err = fcntl_setlk64(fd, f.file, cmd,
410 (struct flock64 __user *) arg);
411 break;
412 default:
413 err = do_fcntl(fd, cmd, arg, f.file);
414 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 }
Al Viro545ec2c2012-04-21 18:42:19 -0400416out1:
Al Viro2903ff02012-08-28 12:52:22 -0400417 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418out:
419 return err;
420}
421#endif
422
423/* Table to convert sigio signal codes into poll band bitmaps */
424
Eric Dumazetfa3536c2006-03-26 01:37:24 -0800425static const long band_table[NSIGPOLL] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 POLLIN | POLLRDNORM, /* POLL_IN */
427 POLLOUT | POLLWRNORM | POLLWRBAND, /* POLL_OUT */
428 POLLIN | POLLRDNORM | POLLMSG, /* POLL_MSG */
429 POLLERR, /* POLL_ERR */
430 POLLPRI | POLLRDBAND, /* POLL_PRI */
431 POLLHUP | POLLERR /* POLL_HUP */
432};
433
434static inline int sigio_perm(struct task_struct *p,
435 struct fown_struct *fown, int sig)
436{
David Howellsc69e8d92008-11-14 10:39:19 +1100437 const struct cred *cred;
438 int ret;
439
440 rcu_read_lock();
441 cred = __task_cred(p);
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -0800442 ret = ((uid_eq(fown->euid, GLOBAL_ROOT_UID) ||
443 uid_eq(fown->euid, cred->suid) || uid_eq(fown->euid, cred->uid) ||
444 uid_eq(fown->uid, cred->suid) || uid_eq(fown->uid, cred->uid)) &&
David Howellsc69e8d92008-11-14 10:39:19 +1100445 !security_file_send_sigiotask(p, fown, sig));
446 rcu_read_unlock();
447 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448}
449
450static void send_sigio_to_task(struct task_struct *p,
Oleg Nesterov8eeee4e2009-06-17 00:27:10 +0200451 struct fown_struct *fown,
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700452 int fd, int reason, int group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453{
Oleg Nesterov8eeee4e2009-06-17 00:27:10 +0200454 /*
455 * F_SETSIG can change ->signum lockless in parallel, make
456 * sure we read it once and use the same value throughout.
457 */
458 int signum = ACCESS_ONCE(fown->signum);
459
460 if (!sigio_perm(p, fown, signum))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 return;
462
Oleg Nesterov8eeee4e2009-06-17 00:27:10 +0200463 switch (signum) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 siginfo_t si;
465 default:
466 /* Queue a rt signal with the appropriate fd as its
467 value. We use SI_SIGIO as the source, not
468 SI_KERNEL, since kernel signals always get
469 delivered even if we can't queue. Failure to
470 queue in this case _should_ be reported; we fall
471 back to SIGIO in that case. --sct */
Oleg Nesterov8eeee4e2009-06-17 00:27:10 +0200472 si.si_signo = signum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 si.si_errno = 0;
474 si.si_code = reason;
475 /* Make sure we are called with one of the POLL_*
476 reasons, otherwise we could leak kernel stack into
477 userspace. */
Eric Sesterhennf6298aa2006-04-02 13:37:19 +0200478 BUG_ON((reason & __SI_MASK) != __SI_POLL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 if (reason - POLL_IN >= NSIGPOLL)
480 si.si_band = ~0L;
481 else
482 si.si_band = band_table[reason - POLL_IN];
483 si.si_fd = fd;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700484 if (!do_send_sig_info(signum, &si, p, group))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 break;
486 /* fall-through: fall back on the old plain SIGIO signal */
487 case 0:
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700488 do_send_sig_info(SIGIO, SEND_SIG_PRIV, p, group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 }
490}
491
492void send_sigio(struct fown_struct *fown, int fd, int band)
493{
494 struct task_struct *p;
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700495 enum pid_type type;
496 struct pid *pid;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700497 int group = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498
499 read_lock(&fown->lock);
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700500
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700501 type = fown->pid_type;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700502 if (type == PIDTYPE_MAX) {
503 group = 0;
504 type = PIDTYPE_PID;
505 }
506
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 pid = fown->pid;
508 if (!pid)
509 goto out_unlock_fown;
510
511 read_lock(&tasklist_lock);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700512 do_each_pid_task(pid, type, p) {
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700513 send_sigio_to_task(p, fown, fd, band, group);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700514 } while_each_pid_task(pid, type, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 read_unlock(&tasklist_lock);
516 out_unlock_fown:
517 read_unlock(&fown->lock);
518}
519
520static void send_sigurg_to_task(struct task_struct *p,
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700521 struct fown_struct *fown, int group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522{
523 if (sigio_perm(p, fown, SIGURG))
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700524 do_send_sig_info(SIGURG, SEND_SIG_PRIV, p, group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525}
526
527int send_sigurg(struct fown_struct *fown)
528{
529 struct task_struct *p;
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700530 enum pid_type type;
531 struct pid *pid;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700532 int group = 1;
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700533 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534
535 read_lock(&fown->lock);
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700536
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700537 type = fown->pid_type;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700538 if (type == PIDTYPE_MAX) {
539 group = 0;
540 type = PIDTYPE_PID;
541 }
542
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 pid = fown->pid;
544 if (!pid)
545 goto out_unlock_fown;
546
547 ret = 1;
548
549 read_lock(&tasklist_lock);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700550 do_each_pid_task(pid, type, p) {
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700551 send_sigurg_to_task(p, fown, group);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700552 } while_each_pid_task(pid, type, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 read_unlock(&tasklist_lock);
554 out_unlock_fown:
555 read_unlock(&fown->lock);
556 return ret;
557}
558
Eric Dumazet989a2972010-04-14 09:55:35 +0000559static DEFINE_SPINLOCK(fasync_lock);
Christoph Lametere18b8902006-12-06 20:33:20 -0800560static struct kmem_cache *fasync_cache __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561
Eric Dumazet989a2972010-04-14 09:55:35 +0000562static void fasync_free_rcu(struct rcu_head *head)
563{
564 kmem_cache_free(fasync_cache,
565 container_of(head, struct fasync_struct, fa_rcu));
566}
567
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568/*
Linus Torvalds53281b62009-12-16 08:23:37 -0800569 * Remove a fasync entry. If successfully removed, return
570 * positive and clear the FASYNC flag. If no entry exists,
571 * do nothing and return 0.
572 *
573 * NOTE! It is very important that the FASYNC flag always
574 * match the state "is the filp on a fasync list".
575 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 */
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400577int fasync_remove_entry(struct file *filp, struct fasync_struct **fapp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578{
579 struct fasync_struct *fa, **fp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 int result = 0;
581
Jonathan Corbet4a6a4492009-03-27 12:24:31 -0600582 spin_lock(&filp->f_lock);
Eric Dumazet989a2972010-04-14 09:55:35 +0000583 spin_lock(&fasync_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 for (fp = fapp; (fa = *fp) != NULL; fp = &fa->fa_next) {
Linus Torvalds53281b62009-12-16 08:23:37 -0800585 if (fa->fa_file != filp)
586 continue;
Eric Dumazet989a2972010-04-14 09:55:35 +0000587
588 spin_lock_irq(&fa->fa_lock);
589 fa->fa_file = NULL;
590 spin_unlock_irq(&fa->fa_lock);
591
Linus Torvalds53281b62009-12-16 08:23:37 -0800592 *fp = fa->fa_next;
Eric Dumazet989a2972010-04-14 09:55:35 +0000593 call_rcu(&fa->fa_rcu, fasync_free_rcu);
Jonathan Corbet76398422009-02-01 14:26:59 -0700594 filp->f_flags &= ~FASYNC;
Linus Torvalds53281b62009-12-16 08:23:37 -0800595 result = 1;
596 break;
597 }
Eric Dumazet989a2972010-04-14 09:55:35 +0000598 spin_unlock(&fasync_lock);
Jonathan Corbet4a6a4492009-03-27 12:24:31 -0600599 spin_unlock(&filp->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 return result;
601}
602
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400603struct fasync_struct *fasync_alloc(void)
Linus Torvalds53281b62009-12-16 08:23:37 -0800604{
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400605 return kmem_cache_alloc(fasync_cache, GFP_KERNEL);
606}
Linus Torvalds53281b62009-12-16 08:23:37 -0800607
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400608/*
609 * NOTE! This can be used only for unused fasync entries:
610 * entries that actually got inserted on the fasync list
611 * need to be released by rcu - see fasync_remove_entry.
612 */
613void fasync_free(struct fasync_struct *new)
614{
615 kmem_cache_free(fasync_cache, new);
616}
617
618/*
619 * Insert a new entry into the fasync list. Return the pointer to the
620 * old one if we didn't use the new one.
Linus Torvalds55f335a2010-10-27 18:17:02 -0700621 *
622 * NOTE! It is very important that the FASYNC flag always
623 * match the state "is the filp on a fasync list".
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400624 */
625struct fasync_struct *fasync_insert_entry(int fd, struct file *filp, struct fasync_struct **fapp, struct fasync_struct *new)
626{
627 struct fasync_struct *fa, **fp;
Linus Torvalds53281b62009-12-16 08:23:37 -0800628
629 spin_lock(&filp->f_lock);
Eric Dumazet989a2972010-04-14 09:55:35 +0000630 spin_lock(&fasync_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800631 for (fp = fapp; (fa = *fp) != NULL; fp = &fa->fa_next) {
632 if (fa->fa_file != filp)
633 continue;
Eric Dumazet989a2972010-04-14 09:55:35 +0000634
635 spin_lock_irq(&fa->fa_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800636 fa->fa_fd = fd;
Eric Dumazet989a2972010-04-14 09:55:35 +0000637 spin_unlock_irq(&fa->fa_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800638 goto out;
639 }
640
Eric Dumazet989a2972010-04-14 09:55:35 +0000641 spin_lock_init(&new->fa_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800642 new->magic = FASYNC_MAGIC;
643 new->fa_file = filp;
644 new->fa_fd = fd;
645 new->fa_next = *fapp;
Eric Dumazet989a2972010-04-14 09:55:35 +0000646 rcu_assign_pointer(*fapp, new);
Linus Torvalds53281b62009-12-16 08:23:37 -0800647 filp->f_flags |= FASYNC;
648
649out:
Eric Dumazet989a2972010-04-14 09:55:35 +0000650 spin_unlock(&fasync_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800651 spin_unlock(&filp->f_lock);
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400652 return fa;
653}
654
655/*
656 * Add a fasync entry. Return negative on error, positive if
657 * added, and zero if did nothing but change an existing one.
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400658 */
659static int fasync_add_entry(int fd, struct file *filp, struct fasync_struct **fapp)
660{
661 struct fasync_struct *new;
662
663 new = fasync_alloc();
664 if (!new)
665 return -ENOMEM;
666
667 /*
668 * fasync_insert_entry() returns the old (update) entry if
669 * it existed.
670 *
671 * So free the (unused) new entry and return 0 to let the
672 * caller know that we didn't add any new fasync entries.
673 */
674 if (fasync_insert_entry(fd, filp, fapp, new)) {
675 fasync_free(new);
676 return 0;
677 }
678
679 return 1;
Linus Torvalds53281b62009-12-16 08:23:37 -0800680}
681
682/*
683 * fasync_helper() is used by almost all character device drivers
684 * to set up the fasync queue, and for regular files by the file
685 * lease code. It returns negative on error, 0 if it did no changes
686 * and positive if it added/deleted the entry.
687 */
688int fasync_helper(int fd, struct file * filp, int on, struct fasync_struct **fapp)
689{
690 if (!on)
691 return fasync_remove_entry(filp, fapp);
692 return fasync_add_entry(fd, filp, fapp);
693}
694
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695EXPORT_SYMBOL(fasync_helper);
696
Eric Dumazet989a2972010-04-14 09:55:35 +0000697/*
698 * rcu_read_lock() is held
699 */
700static void kill_fasync_rcu(struct fasync_struct *fa, int sig, int band)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701{
702 while (fa) {
Eric Dumazet989a2972010-04-14 09:55:35 +0000703 struct fown_struct *fown;
Andrew Mortonf4985dc2010-06-29 15:05:42 -0700704 unsigned long flags;
705
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 if (fa->magic != FASYNC_MAGIC) {
707 printk(KERN_ERR "kill_fasync: bad magic number in "
708 "fasync_struct!\n");
709 return;
710 }
Andrew Mortonf4985dc2010-06-29 15:05:42 -0700711 spin_lock_irqsave(&fa->fa_lock, flags);
Eric Dumazet989a2972010-04-14 09:55:35 +0000712 if (fa->fa_file) {
713 fown = &fa->fa_file->f_owner;
714 /* Don't send SIGURG to processes which have not set a
715 queued signum: SIGURG has its own default signalling
716 mechanism. */
717 if (!(sig == SIGURG && fown->signum == 0))
718 send_sigio(fown, fa->fa_fd, band);
719 }
Andrew Mortonf4985dc2010-06-29 15:05:42 -0700720 spin_unlock_irqrestore(&fa->fa_lock, flags);
Eric Dumazet989a2972010-04-14 09:55:35 +0000721 fa = rcu_dereference(fa->fa_next);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722 }
723}
724
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725void kill_fasync(struct fasync_struct **fp, int sig, int band)
726{
727 /* First a quick test without locking: usually
728 * the list is empty.
729 */
730 if (*fp) {
Eric Dumazet989a2972010-04-14 09:55:35 +0000731 rcu_read_lock();
732 kill_fasync_rcu(rcu_dereference(*fp), sig, band);
733 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 }
735}
736EXPORT_SYMBOL(kill_fasync);
737
Wu Fengguang454eedb2010-08-10 18:01:29 -0700738static int __init fcntl_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739{
James Bottomley3ab04d52010-09-09 16:38:12 -0700740 /*
741 * Please add new bits here to ensure allocation uniqueness.
742 * Exceptions: O_NONBLOCK is a two bit define on parisc; O_NDELAY
743 * is defined as O_NONBLOCK on some platforms and not on others.
744 */
Zheng Liu3d62c452013-07-25 08:13:19 +0800745 BUILD_BUG_ON(20 - 1 /* for O_RDONLY being 0 */ != HWEIGHT32(
Wu Fengguang454eedb2010-08-10 18:01:29 -0700746 O_RDONLY | O_WRONLY | O_RDWR |
747 O_CREAT | O_EXCL | O_NOCTTY |
James Bottomley3ab04d52010-09-09 16:38:12 -0700748 O_TRUNC | O_APPEND | /* O_NONBLOCK | */
Wu Fengguang454eedb2010-08-10 18:01:29 -0700749 __O_SYNC | O_DSYNC | FASYNC |
750 O_DIRECT | O_LARGEFILE | O_DIRECTORY |
751 O_NOFOLLOW | O_NOATIME | O_CLOEXEC |
Zheng Liu3d62c452013-07-25 08:13:19 +0800752 __FMODE_EXEC | O_PATH | __O_TMPFILE
Wu Fengguang454eedb2010-08-10 18:01:29 -0700753 ));
754
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 fasync_cache = kmem_cache_create("fasync_cache",
Paul Mundt20c2df82007-07-20 10:11:58 +0900756 sizeof(struct fasync_struct), 0, SLAB_PANIC, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 return 0;
758}
759
Wu Fengguang454eedb2010-08-10 18:01:29 -0700760module_init(fcntl_init)