blob: 12273b6ea56dbdeb2a2bff0c6846030e4fc9bcd5 [file] [log] [blame]
Greg Kroah-Hartmanb2441312017-11-01 15:07:57 +01001// SPDX-License-Identifier: GPL-2.0
Linus Torvalds1da177e2005-04-16 15:20:36 -07002/*
3 * linux/fs/fcntl.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 */
7
8#include <linux/syscalls.h>
9#include <linux/init.h>
10#include <linux/mm.h>
Ingo Molnar29930022017-02-08 18:51:36 +010011#include <linux/sched/task.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/fs.h>
13#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040014#include <linux/fdtable.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080015#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/dnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/slab.h>
18#include <linux/module.h>
Jens Axboe35f3d142010-05-20 10:43:18 +020019#include <linux/pipe_fs_i.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <linux/security.h>
21#include <linux/ptrace.h>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070022#include <linux/signal.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070023#include <linux/rcupdate.h>
Pavel Emelyanovb4888932007-10-18 23:40:14 -070024#include <linux/pid_namespace.h>
Cyrill Gorcunov1d151c32012-07-30 14:43:00 -070025#include <linux/user_namespace.h>
Mike Kravetz5d752602018-06-07 17:06:01 -070026#include <linux/memfd.h>
Al Viro80f0cce2017-04-08 18:10:56 -040027#include <linux/compat.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070028
Al Virocfe39442018-02-01 12:14:57 -050029#include <linux/poll.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#include <asm/siginfo.h>
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080031#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032
Jonathan Corbet76398422009-02-01 14:26:59 -070033#define SETFL_MASK (O_APPEND | O_NONBLOCK | O_NDELAY | O_DIRECT | O_NOATIME)
Linus Torvalds1da177e2005-04-16 15:20:36 -070034
35static int setfl(int fd, struct file * filp, unsigned long arg)
36{
Al Viro496ad9a2013-01-23 17:07:38 -050037 struct inode * inode = file_inode(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -070038 int error = 0;
39
dean gaudet7d95c8f2006-02-03 03:04:30 -080040 /*
41 * O_APPEND cannot be cleared if the file is marked as append-only
42 * and the file is open for write.
43 */
44 if (((arg ^ filp->f_flags) & O_APPEND) && IS_APPEND(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -070045 return -EPERM;
46
47 /* O_NOATIME can only be set by the owner or superuser */
48 if ((arg & O_NOATIME) && !(filp->f_flags & O_NOATIME))
Serge E. Hallyn2e149672011-03-23 16:43:26 -070049 if (!inode_owner_or_capable(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -070050 return -EPERM;
51
52 /* required for strict SunOS emulation */
53 if (O_NONBLOCK != O_NDELAY)
54 if (arg & O_NDELAY)
55 arg |= O_NONBLOCK;
56
Stanislav Kinsburskiy0dbf5f22015-12-15 19:41:31 +040057 /* Pipe packetized mode is controlled by O_DIRECT flag */
Al Viro45063092016-12-04 18:24:56 -050058 if (!S_ISFIFO(inode->i_mode) && (arg & O_DIRECT)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070059 if (!filp->f_mapping || !filp->f_mapping->a_ops ||
60 !filp->f_mapping->a_ops->direct_IO)
61 return -EINVAL;
62 }
63
Al Viro72c2d532013-09-22 16:27:52 -040064 if (filp->f_op->check_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -070065 error = filp->f_op->check_flags(arg);
66 if (error)
67 return error;
68
Jonathan Corbet218d11a2008-12-05 16:12:48 -070069 /*
Jonathan Corbet76398422009-02-01 14:26:59 -070070 * ->fasync() is responsible for setting the FASYNC bit.
Jonathan Corbet218d11a2008-12-05 16:12:48 -070071 */
Al Viro72c2d532013-09-22 16:27:52 -040072 if (((arg ^ filp->f_flags) & FASYNC) && filp->f_op->fasync) {
Jonathan Corbet76398422009-02-01 14:26:59 -070073 error = filp->f_op->fasync(fd, filp, (arg & FASYNC) != 0);
74 if (error < 0)
75 goto out;
Jonathan Corbet60aa4922009-02-01 14:52:56 -070076 if (error > 0)
77 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 }
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -070079 spin_lock(&filp->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 filp->f_flags = (arg & SETFL_MASK) | (filp->f_flags & ~SETFL_MASK);
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -070081 spin_unlock(&filp->f_lock);
Jonathan Corbet76398422009-02-01 14:26:59 -070082
Linus Torvalds1da177e2005-04-16 15:20:36 -070083 out:
Linus Torvalds1da177e2005-04-16 15:20:36 -070084 return error;
85}
86
Eric W. Biederman609d7fa2006-10-02 02:17:15 -070087static void f_modown(struct file *filp, struct pid *pid, enum pid_type type,
Oleg Nesterov2f38d702009-06-16 22:07:46 +020088 int force)
Linus Torvalds1da177e2005-04-16 15:20:36 -070089{
Linus Torvalds80e1e822010-02-07 10:11:23 -080090 write_lock_irq(&filp->f_owner.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070091 if (force || !filp->f_owner.pid) {
Eric W. Biederman609d7fa2006-10-02 02:17:15 -070092 put_pid(filp->f_owner.pid);
93 filp->f_owner.pid = get_pid(pid);
94 filp->f_owner.pid_type = type;
Oleg Nesterov2f38d702009-06-16 22:07:46 +020095
96 if (pid) {
97 const struct cred *cred = current_cred();
98 filp->f_owner.uid = cred->uid;
99 filp->f_owner.euid = cred->euid;
100 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101 }
Linus Torvalds80e1e822010-02-07 10:11:23 -0800102 write_unlock_irq(&filp->f_owner.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103}
104
Jeff Laytone0b93ed2014-08-22 11:27:32 -0400105void __f_setown(struct file *filp, struct pid *pid, enum pid_type type,
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700106 int force)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107{
Jeff Laytone0b93ed2014-08-22 11:27:32 -0400108 security_file_set_fowner(filp);
Oleg Nesterov2f38d702009-06-16 22:07:46 +0200109 f_modown(filp, pid, type, force);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110}
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700111EXPORT_SYMBOL(__f_setown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112
Jiri Slaby393cc3f2017-06-13 13:35:50 +0200113int f_setown(struct file *filp, unsigned long arg, int force)
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700114{
115 enum pid_type type;
Jeff Laytonf7312732017-06-14 09:11:54 -0400116 struct pid *pid = NULL;
117 int who = arg, ret = 0;
118
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700119 type = PIDTYPE_PID;
120 if (who < 0) {
Jiri Slabyfc3dc672017-06-13 13:35:51 +0200121 /* avoid overflow below */
122 if (who == INT_MIN)
123 return -EINVAL;
124
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700125 type = PIDTYPE_PGID;
126 who = -who;
127 }
Jeff Laytonf7312732017-06-14 09:11:54 -0400128
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700129 rcu_read_lock();
Jeff Laytonf7312732017-06-14 09:11:54 -0400130 if (who) {
131 pid = find_vpid(who);
132 if (!pid)
133 ret = -ESRCH;
134 }
135
136 if (!ret)
137 __f_setown(filp, pid, type, force);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700138 rcu_read_unlock();
Jiri Slaby393cc3f2017-06-13 13:35:50 +0200139
Jeff Laytonf7312732017-06-14 09:11:54 -0400140 return ret;
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700141}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142EXPORT_SYMBOL(f_setown);
143
144void f_delown(struct file *filp)
145{
Oleg Nesterov2f38d702009-06-16 22:07:46 +0200146 f_modown(filp, NULL, PIDTYPE_PID, 1);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700147}
148
149pid_t f_getown(struct file *filp)
150{
151 pid_t pid;
Eric W. Biederman43fa1ad2006-10-02 02:17:27 -0700152 read_lock(&filp->f_owner.lock);
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800153 pid = pid_vnr(filp->f_owner.pid);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700154 if (filp->f_owner.pid_type == PIDTYPE_PGID)
155 pid = -pid;
Eric W. Biederman43fa1ad2006-10-02 02:17:27 -0700156 read_unlock(&filp->f_owner.lock);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700157 return pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158}
159
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700160static int f_setown_ex(struct file *filp, unsigned long arg)
161{
Al Viro63784dd2012-09-26 21:43:05 -0400162 struct f_owner_ex __user *owner_p = (void __user *)arg;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700163 struct f_owner_ex owner;
164 struct pid *pid;
165 int type;
166 int ret;
167
168 ret = copy_from_user(&owner, owner_p, sizeof(owner));
169 if (ret)
Dan Carpenter5b544702010-06-03 12:35:42 +0200170 return -EFAULT;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700171
172 switch (owner.type) {
173 case F_OWNER_TID:
174 type = PIDTYPE_MAX;
175 break;
176
177 case F_OWNER_PID:
178 type = PIDTYPE_PID;
179 break;
180
Peter Zijlstra978b4052009-11-17 14:06:24 -0800181 case F_OWNER_PGRP:
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700182 type = PIDTYPE_PGID;
183 break;
184
185 default:
186 return -EINVAL;
187 }
188
189 rcu_read_lock();
190 pid = find_vpid(owner.pid);
191 if (owner.pid && !pid)
192 ret = -ESRCH;
193 else
Jeff Laytone0b93ed2014-08-22 11:27:32 -0400194 __f_setown(filp, pid, type, 1);
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700195 rcu_read_unlock();
196
197 return ret;
198}
199
200static int f_getown_ex(struct file *filp, unsigned long arg)
201{
Al Viro63784dd2012-09-26 21:43:05 -0400202 struct f_owner_ex __user *owner_p = (void __user *)arg;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700203 struct f_owner_ex owner;
204 int ret = 0;
205
206 read_lock(&filp->f_owner.lock);
207 owner.pid = pid_vnr(filp->f_owner.pid);
208 switch (filp->f_owner.pid_type) {
209 case PIDTYPE_MAX:
210 owner.type = F_OWNER_TID;
211 break;
212
213 case PIDTYPE_PID:
214 owner.type = F_OWNER_PID;
215 break;
216
217 case PIDTYPE_PGID:
Peter Zijlstra978b4052009-11-17 14:06:24 -0800218 owner.type = F_OWNER_PGRP;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700219 break;
220
221 default:
222 WARN_ON(1);
223 ret = -EINVAL;
224 break;
225 }
226 read_unlock(&filp->f_owner.lock);
227
Dan Carpenter5b544702010-06-03 12:35:42 +0200228 if (!ret) {
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700229 ret = copy_to_user(owner_p, &owner, sizeof(owner));
Dan Carpenter5b544702010-06-03 12:35:42 +0200230 if (ret)
231 ret = -EFAULT;
232 }
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700233 return ret;
234}
235
Cyrill Gorcunov1d151c32012-07-30 14:43:00 -0700236#ifdef CONFIG_CHECKPOINT_RESTORE
237static int f_getowner_uids(struct file *filp, unsigned long arg)
238{
239 struct user_namespace *user_ns = current_user_ns();
Al Viro63784dd2012-09-26 21:43:05 -0400240 uid_t __user *dst = (void __user *)arg;
Cyrill Gorcunov1d151c32012-07-30 14:43:00 -0700241 uid_t src[2];
242 int err;
243
244 read_lock(&filp->f_owner.lock);
245 src[0] = from_kuid(user_ns, filp->f_owner.uid);
246 src[1] = from_kuid(user_ns, filp->f_owner.euid);
247 read_unlock(&filp->f_owner.lock);
248
249 err = put_user(src[0], &dst[0]);
250 err |= put_user(src[1], &dst[1]);
251
252 return err;
253}
254#else
255static int f_getowner_uids(struct file *filp, unsigned long arg)
256{
257 return -EINVAL;
258}
259#endif
260
Jens Axboec75b1d92017-06-27 11:47:04 -0600261static bool rw_hint_valid(enum rw_hint hint)
262{
263 switch (hint) {
264 case RWF_WRITE_LIFE_NOT_SET:
265 case RWH_WRITE_LIFE_NONE:
266 case RWH_WRITE_LIFE_SHORT:
267 case RWH_WRITE_LIFE_MEDIUM:
268 case RWH_WRITE_LIFE_LONG:
269 case RWH_WRITE_LIFE_EXTREME:
270 return true;
271 default:
272 return false;
273 }
274}
275
276static long fcntl_rw_hint(struct file *file, unsigned int cmd,
277 unsigned long arg)
278{
279 struct inode *inode = file_inode(file);
280 u64 *argp = (u64 __user *)arg;
281 enum rw_hint hint;
Jens Axboe5657cb02017-06-28 08:09:45 -0600282 u64 h;
Jens Axboec75b1d92017-06-27 11:47:04 -0600283
284 switch (cmd) {
285 case F_GET_FILE_RW_HINT:
Jens Axboe5657cb02017-06-28 08:09:45 -0600286 h = file_write_hint(file);
287 if (copy_to_user(argp, &h, sizeof(*argp)))
Jens Axboec75b1d92017-06-27 11:47:04 -0600288 return -EFAULT;
289 return 0;
290 case F_SET_FILE_RW_HINT:
Jens Axboe5657cb02017-06-28 08:09:45 -0600291 if (copy_from_user(&h, argp, sizeof(h)))
Jens Axboec75b1d92017-06-27 11:47:04 -0600292 return -EFAULT;
Jens Axboe5657cb02017-06-28 08:09:45 -0600293 hint = (enum rw_hint) h;
Jens Axboec75b1d92017-06-27 11:47:04 -0600294 if (!rw_hint_valid(hint))
295 return -EINVAL;
296
297 spin_lock(&file->f_lock);
298 file->f_write_hint = hint;
299 spin_unlock(&file->f_lock);
300 return 0;
301 case F_GET_RW_HINT:
Jens Axboe5657cb02017-06-28 08:09:45 -0600302 h = inode->i_write_hint;
303 if (copy_to_user(argp, &h, sizeof(*argp)))
Jens Axboec75b1d92017-06-27 11:47:04 -0600304 return -EFAULT;
305 return 0;
306 case F_SET_RW_HINT:
Jens Axboe5657cb02017-06-28 08:09:45 -0600307 if (copy_from_user(&h, argp, sizeof(h)))
Jens Axboec75b1d92017-06-27 11:47:04 -0600308 return -EFAULT;
Jens Axboe5657cb02017-06-28 08:09:45 -0600309 hint = (enum rw_hint) h;
Jens Axboec75b1d92017-06-27 11:47:04 -0600310 if (!rw_hint_valid(hint))
311 return -EINVAL;
312
313 inode_lock(inode);
314 inode->i_write_hint = hint;
315 inode_unlock(inode);
316 return 0;
317 default:
318 return -EINVAL;
319 }
320}
321
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322static long do_fcntl(int fd, unsigned int cmd, unsigned long arg,
323 struct file *filp)
324{
Christoph Hellwiga75d30c2017-05-27 06:07:19 -0400325 void __user *argp = (void __user *)arg;
326 struct flock flock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 long err = -EINVAL;
328
329 switch (cmd) {
330 case F_DUPFD:
Al Virofe17f222012-08-21 11:48:11 -0400331 err = f_dupfd(arg, filp, 0);
332 break;
Ulrich Drepper22d2b352007-10-16 23:30:26 -0700333 case F_DUPFD_CLOEXEC:
Al Viro12197712012-10-08 23:21:58 +0100334 err = f_dupfd(arg, filp, O_CLOEXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 break;
336 case F_GETFD:
337 err = get_close_on_exec(fd) ? FD_CLOEXEC : 0;
338 break;
339 case F_SETFD:
340 err = 0;
341 set_close_on_exec(fd, arg & FD_CLOEXEC);
342 break;
343 case F_GETFL:
344 err = filp->f_flags;
345 break;
346 case F_SETFL:
347 err = setfl(fd, filp, arg);
348 break;
Jeff Layton5d50ffd2014-02-03 12:13:10 -0500349#if BITS_PER_LONG != 32
350 /* 32-bit arches must use fcntl64() */
Jeff Layton0d3f7a22014-04-22 08:23:58 -0400351 case F_OFD_GETLK:
Jeff Layton5d50ffd2014-02-03 12:13:10 -0500352#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 case F_GETLK:
Christoph Hellwiga75d30c2017-05-27 06:07:19 -0400354 if (copy_from_user(&flock, argp, sizeof(flock)))
355 return -EFAULT;
356 err = fcntl_getlk(filp, cmd, &flock);
357 if (!err && copy_to_user(argp, &flock, sizeof(flock)))
358 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 break;
Jeff Layton5d50ffd2014-02-03 12:13:10 -0500360#if BITS_PER_LONG != 32
361 /* 32-bit arches must use fcntl64() */
Jeff Layton0d3f7a22014-04-22 08:23:58 -0400362 case F_OFD_SETLK:
363 case F_OFD_SETLKW:
Jeff Layton5d50ffd2014-02-03 12:13:10 -0500364#endif
365 /* Fallthrough */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 case F_SETLK:
367 case F_SETLKW:
Christoph Hellwiga75d30c2017-05-27 06:07:19 -0400368 if (copy_from_user(&flock, argp, sizeof(flock)))
369 return -EFAULT;
370 err = fcntl_setlk(fd, filp, cmd, &flock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 break;
372 case F_GETOWN:
373 /*
374 * XXX If f_owner is a process group, the
375 * negative return value will get converted
376 * into an error. Oops. If we keep the
377 * current syscall conventions, the only way
378 * to fix this will be in libc.
379 */
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700380 err = f_getown(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 force_successful_syscall_return();
382 break;
383 case F_SETOWN:
Jiri Slaby393cc3f2017-06-13 13:35:50 +0200384 err = f_setown(filp, arg, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 break;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700386 case F_GETOWN_EX:
387 err = f_getown_ex(filp, arg);
388 break;
389 case F_SETOWN_EX:
390 err = f_setown_ex(filp, arg);
391 break;
Cyrill Gorcunov1d151c32012-07-30 14:43:00 -0700392 case F_GETOWNER_UIDS:
393 err = f_getowner_uids(filp, arg);
394 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 case F_GETSIG:
396 err = filp->f_owner.signum;
397 break;
398 case F_SETSIG:
399 /* arg == 0 restores default behaviour. */
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700400 if (!valid_signal(arg)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 break;
402 }
403 err = 0;
404 filp->f_owner.signum = arg;
405 break;
406 case F_GETLEASE:
407 err = fcntl_getlease(filp);
408 break;
409 case F_SETLEASE:
410 err = fcntl_setlease(fd, filp, arg);
411 break;
412 case F_NOTIFY:
413 err = fcntl_dirnotify(fd, filp, arg);
414 break;
Jens Axboe35f3d142010-05-20 10:43:18 +0200415 case F_SETPIPE_SZ:
416 case F_GETPIPE_SZ:
417 err = pipe_fcntl(filp, cmd, arg);
418 break;
David Herrmann40e041a2014-08-08 14:25:27 -0700419 case F_ADD_SEALS:
420 case F_GET_SEALS:
Marc-André Lureau5aadc432018-01-31 16:19:18 -0800421 err = memfd_fcntl(filp, cmd, arg);
David Herrmann40e041a2014-08-08 14:25:27 -0700422 break;
Jens Axboec75b1d92017-06-27 11:47:04 -0600423 case F_GET_RW_HINT:
424 case F_SET_RW_HINT:
425 case F_GET_FILE_RW_HINT:
426 case F_SET_FILE_RW_HINT:
427 err = fcntl_rw_hint(filp, cmd, arg);
428 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 default:
430 break;
431 }
432 return err;
433}
434
Al Viro1abf0c72011-03-13 03:51:11 -0400435static int check_fcntl_cmd(unsigned cmd)
436{
437 switch (cmd) {
438 case F_DUPFD:
439 case F_DUPFD_CLOEXEC:
440 case F_GETFD:
441 case F_SETFD:
442 case F_GETFL:
443 return 1;
444 }
445 return 0;
446}
447
Heiko Carstensa26eab22009-01-14 14:14:17 +0100448SYSCALL_DEFINE3(fcntl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449{
Al Viro2903ff02012-08-28 12:52:22 -0400450 struct fd f = fdget_raw(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 long err = -EBADF;
452
Al Viro2903ff02012-08-28 12:52:22 -0400453 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 goto out;
455
Al Viro2903ff02012-08-28 12:52:22 -0400456 if (unlikely(f.file->f_mode & FMODE_PATH)) {
Al Viro545ec2c2012-04-21 18:42:19 -0400457 if (!check_fcntl_cmd(cmd))
458 goto out1;
Al Viro1abf0c72011-03-13 03:51:11 -0400459 }
460
Al Viro2903ff02012-08-28 12:52:22 -0400461 err = security_file_fcntl(f.file, cmd, arg);
Al Viro545ec2c2012-04-21 18:42:19 -0400462 if (!err)
Al Viro2903ff02012-08-28 12:52:22 -0400463 err = do_fcntl(fd, cmd, arg, f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464
Al Viro545ec2c2012-04-21 18:42:19 -0400465out1:
Al Viro2903ff02012-08-28 12:52:22 -0400466 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467out:
468 return err;
469}
470
471#if BITS_PER_LONG == 32
Heiko Carstensa26eab22009-01-14 14:14:17 +0100472SYSCALL_DEFINE3(fcntl64, unsigned int, fd, unsigned int, cmd,
473 unsigned long, arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474{
Christoph Hellwiga75d30c2017-05-27 06:07:19 -0400475 void __user *argp = (void __user *)arg;
Al Viro2903ff02012-08-28 12:52:22 -0400476 struct fd f = fdget_raw(fd);
Christoph Hellwiga75d30c2017-05-27 06:07:19 -0400477 struct flock64 flock;
Al Viro545ec2c2012-04-21 18:42:19 -0400478 long err = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479
Al Viro2903ff02012-08-28 12:52:22 -0400480 if (!f.file)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 goto out;
482
Al Viro2903ff02012-08-28 12:52:22 -0400483 if (unlikely(f.file->f_mode & FMODE_PATH)) {
Al Viro545ec2c2012-04-21 18:42:19 -0400484 if (!check_fcntl_cmd(cmd))
485 goto out1;
Al Viro1abf0c72011-03-13 03:51:11 -0400486 }
487
Al Viro2903ff02012-08-28 12:52:22 -0400488 err = security_file_fcntl(f.file, cmd, arg);
Al Viro545ec2c2012-04-21 18:42:19 -0400489 if (err)
490 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491
492 switch (cmd) {
Jeff Layton5d50ffd2014-02-03 12:13:10 -0500493 case F_GETLK64:
Jeff Layton0d3f7a22014-04-22 08:23:58 -0400494 case F_OFD_GETLK:
Christoph Hellwiga75d30c2017-05-27 06:07:19 -0400495 err = -EFAULT;
496 if (copy_from_user(&flock, argp, sizeof(flock)))
497 break;
498 err = fcntl_getlk64(f.file, cmd, &flock);
499 if (!err && copy_to_user(argp, &flock, sizeof(flock)))
500 err = -EFAULT;
Jeff Layton5d50ffd2014-02-03 12:13:10 -0500501 break;
502 case F_SETLK64:
503 case F_SETLKW64:
Jeff Layton0d3f7a22014-04-22 08:23:58 -0400504 case F_OFD_SETLK:
505 case F_OFD_SETLKW:
Christoph Hellwiga75d30c2017-05-27 06:07:19 -0400506 err = -EFAULT;
507 if (copy_from_user(&flock, argp, sizeof(flock)))
508 break;
509 err = fcntl_setlk64(fd, f.file, cmd, &flock);
Jeff Layton5d50ffd2014-02-03 12:13:10 -0500510 break;
511 default:
512 err = do_fcntl(fd, cmd, arg, f.file);
513 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 }
Al Viro545ec2c2012-04-21 18:42:19 -0400515out1:
Al Viro2903ff02012-08-28 12:52:22 -0400516 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517out:
518 return err;
519}
520#endif
521
Al Viro80f0cce2017-04-08 18:10:56 -0400522#ifdef CONFIG_COMPAT
Al Viro8c6657c2017-06-26 23:51:31 -0400523/* careful - don't use anywhere else */
Linus Torvaldsb59eea52017-07-07 13:48:18 -0700524#define copy_flock_fields(dst, src) \
525 (dst)->l_type = (src)->l_type; \
526 (dst)->l_whence = (src)->l_whence; \
527 (dst)->l_start = (src)->l_start; \
528 (dst)->l_len = (src)->l_len; \
529 (dst)->l_pid = (src)->l_pid;
Al Viro8c6657c2017-06-26 23:51:31 -0400530
Linus Torvaldsb59eea52017-07-07 13:48:18 -0700531static int get_compat_flock(struct flock *kfl, const struct compat_flock __user *ufl)
Al Viro80f0cce2017-04-08 18:10:56 -0400532{
Al Viro8c6657c2017-06-26 23:51:31 -0400533 struct compat_flock fl;
534
535 if (copy_from_user(&fl, ufl, sizeof(struct compat_flock)))
Al Viro80f0cce2017-04-08 18:10:56 -0400536 return -EFAULT;
Linus Torvaldsb59eea52017-07-07 13:48:18 -0700537 copy_flock_fields(kfl, &fl);
Al Viro8c6657c2017-06-26 23:51:31 -0400538 return 0;
539}
540
Linus Torvaldsb59eea52017-07-07 13:48:18 -0700541static int get_compat_flock64(struct flock *kfl, const struct compat_flock64 __user *ufl)
Al Viro8c6657c2017-06-26 23:51:31 -0400542{
543 struct compat_flock64 fl;
544
545 if (copy_from_user(&fl, ufl, sizeof(struct compat_flock64)))
546 return -EFAULT;
Linus Torvaldsb59eea52017-07-07 13:48:18 -0700547 copy_flock_fields(kfl, &fl);
Al Viro80f0cce2017-04-08 18:10:56 -0400548 return 0;
549}
550
Linus Torvaldsb59eea52017-07-07 13:48:18 -0700551static int put_compat_flock(const struct flock *kfl, struct compat_flock __user *ufl)
Al Viro80f0cce2017-04-08 18:10:56 -0400552{
Al Viro8c6657c2017-06-26 23:51:31 -0400553 struct compat_flock fl;
554
555 memset(&fl, 0, sizeof(struct compat_flock));
Linus Torvaldsb59eea52017-07-07 13:48:18 -0700556 copy_flock_fields(&fl, kfl);
Al Viro8c6657c2017-06-26 23:51:31 -0400557 if (copy_to_user(ufl, &fl, sizeof(struct compat_flock)))
Al Viro80f0cce2017-04-08 18:10:56 -0400558 return -EFAULT;
559 return 0;
560}
561
Linus Torvaldsb59eea52017-07-07 13:48:18 -0700562static int put_compat_flock64(const struct flock *kfl, struct compat_flock64 __user *ufl)
Al Viro80f0cce2017-04-08 18:10:56 -0400563{
Al Viro8c6657c2017-06-26 23:51:31 -0400564 struct compat_flock64 fl;
565
Jeff Layton4d2dc2c2017-11-14 14:42:57 -0500566 BUILD_BUG_ON(sizeof(kfl->l_start) > sizeof(ufl->l_start));
567 BUILD_BUG_ON(sizeof(kfl->l_len) > sizeof(ufl->l_len));
568
Al Viro8c6657c2017-06-26 23:51:31 -0400569 memset(&fl, 0, sizeof(struct compat_flock64));
Linus Torvaldsb59eea52017-07-07 13:48:18 -0700570 copy_flock_fields(&fl, kfl);
Al Viro8c6657c2017-06-26 23:51:31 -0400571 if (copy_to_user(ufl, &fl, sizeof(struct compat_flock64)))
Al Viro80f0cce2017-04-08 18:10:56 -0400572 return -EFAULT;
573 return 0;
574}
Al Viro8c6657c2017-06-26 23:51:31 -0400575#undef copy_flock_fields
Al Viro80f0cce2017-04-08 18:10:56 -0400576
577static unsigned int
578convert_fcntl_cmd(unsigned int cmd)
579{
580 switch (cmd) {
581 case F_GETLK64:
582 return F_GETLK;
583 case F_SETLK64:
584 return F_SETLK;
585 case F_SETLKW64:
586 return F_SETLKW;
587 }
588
589 return cmd;
590}
591
Christoph Hellwig94073ad2017-05-27 06:07:20 -0400592/*
593 * GETLK was successful and we need to return the data, but it needs to fit in
594 * the compat structure.
595 * l_start shouldn't be too big, unless the original start + end is greater than
596 * COMPAT_OFF_T_MAX, in which case the app was asking for trouble, so we return
597 * -EOVERFLOW in that case. l_len could be too big, in which case we just
598 * truncate it, and only allow the app to see that part of the conflicting lock
599 * that might make sense to it anyway
600 */
601static int fixup_compat_flock(struct flock *flock)
602{
603 if (flock->l_start > COMPAT_OFF_T_MAX)
604 return -EOVERFLOW;
605 if (flock->l_len > COMPAT_OFF_T_MAX)
606 flock->l_len = COMPAT_OFF_T_MAX;
607 return 0;
608}
609
Dominik Brodowskie02af2f2018-03-20 19:29:53 +0100610static long do_compat_fcntl64(unsigned int fd, unsigned int cmd,
611 compat_ulong_t arg)
Al Viro80f0cce2017-04-08 18:10:56 -0400612{
Christoph Hellwig94073ad2017-05-27 06:07:20 -0400613 struct fd f = fdget_raw(fd);
614 struct flock flock;
615 long err = -EBADF;
616
617 if (!f.file)
618 return err;
619
620 if (unlikely(f.file->f_mode & FMODE_PATH)) {
621 if (!check_fcntl_cmd(cmd))
622 goto out_put;
623 }
624
625 err = security_file_fcntl(f.file, cmd, arg);
626 if (err)
627 goto out_put;
Al Viro80f0cce2017-04-08 18:10:56 -0400628
629 switch (cmd) {
630 case F_GETLK:
Christoph Hellwig94073ad2017-05-27 06:07:20 -0400631 err = get_compat_flock(&flock, compat_ptr(arg));
632 if (err)
633 break;
634 err = fcntl_getlk(f.file, convert_fcntl_cmd(cmd), &flock);
635 if (err)
636 break;
637 err = fixup_compat_flock(&flock);
Jeff Layton9280a602017-11-14 14:43:56 -0500638 if (!err)
639 err = put_compat_flock(&flock, compat_ptr(arg));
Christoph Hellwig94073ad2017-05-27 06:07:20 -0400640 break;
641 case F_GETLK64:
642 case F_OFD_GETLK:
643 err = get_compat_flock64(&flock, compat_ptr(arg));
644 if (err)
645 break;
646 err = fcntl_getlk(f.file, convert_fcntl_cmd(cmd), &flock);
Jeff Layton4d2dc2c2017-11-14 14:42:57 -0500647 if (!err)
648 err = put_compat_flock64(&flock, compat_ptr(arg));
Christoph Hellwig94073ad2017-05-27 06:07:20 -0400649 break;
Al Viro80f0cce2017-04-08 18:10:56 -0400650 case F_SETLK:
651 case F_SETLKW:
Christoph Hellwig94073ad2017-05-27 06:07:20 -0400652 err = get_compat_flock(&flock, compat_ptr(arg));
653 if (err)
Al Viro80f0cce2017-04-08 18:10:56 -0400654 break;
Christoph Hellwig94073ad2017-05-27 06:07:20 -0400655 err = fcntl_setlk(fd, f.file, convert_fcntl_cmd(cmd), &flock);
Al Viro80f0cce2017-04-08 18:10:56 -0400656 break;
Al Viro80f0cce2017-04-08 18:10:56 -0400657 case F_SETLK64:
658 case F_SETLKW64:
Al Viro80f0cce2017-04-08 18:10:56 -0400659 case F_OFD_SETLK:
660 case F_OFD_SETLKW:
Christoph Hellwig94073ad2017-05-27 06:07:20 -0400661 err = get_compat_flock64(&flock, compat_ptr(arg));
662 if (err)
Al Viro80f0cce2017-04-08 18:10:56 -0400663 break;
Christoph Hellwig94073ad2017-05-27 06:07:20 -0400664 err = fcntl_setlk(fd, f.file, convert_fcntl_cmd(cmd), &flock);
Al Viro80f0cce2017-04-08 18:10:56 -0400665 break;
Al Viro80f0cce2017-04-08 18:10:56 -0400666 default:
Christoph Hellwig94073ad2017-05-27 06:07:20 -0400667 err = do_fcntl(fd, cmd, arg, f.file);
Al Viro80f0cce2017-04-08 18:10:56 -0400668 break;
669 }
Christoph Hellwig94073ad2017-05-27 06:07:20 -0400670out_put:
671 fdput(f);
672 return err;
Al Viro80f0cce2017-04-08 18:10:56 -0400673}
674
Dominik Brodowskie02af2f2018-03-20 19:29:53 +0100675COMPAT_SYSCALL_DEFINE3(fcntl64, unsigned int, fd, unsigned int, cmd,
676 compat_ulong_t, arg)
677{
678 return do_compat_fcntl64(fd, cmd, arg);
679}
680
Al Viro80f0cce2017-04-08 18:10:56 -0400681COMPAT_SYSCALL_DEFINE3(fcntl, unsigned int, fd, unsigned int, cmd,
682 compat_ulong_t, arg)
683{
684 switch (cmd) {
685 case F_GETLK64:
686 case F_SETLK64:
687 case F_SETLKW64:
688 case F_OFD_GETLK:
689 case F_OFD_SETLK:
690 case F_OFD_SETLKW:
691 return -EINVAL;
692 }
Dominik Brodowskie02af2f2018-03-20 19:29:53 +0100693 return do_compat_fcntl64(fd, cmd, arg);
Al Viro80f0cce2017-04-08 18:10:56 -0400694}
695#endif
696
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697/* Table to convert sigio signal codes into poll band bitmaps */
698
Al Viro5dc533c2017-07-16 22:14:00 -0400699static const __poll_t band_table[NSIGPOLL] = {
Linus Torvaldsa9a08842018-02-11 14:34:03 -0800700 EPOLLIN | EPOLLRDNORM, /* POLL_IN */
701 EPOLLOUT | EPOLLWRNORM | EPOLLWRBAND, /* POLL_OUT */
702 EPOLLIN | EPOLLRDNORM | EPOLLMSG, /* POLL_MSG */
703 EPOLLERR, /* POLL_ERR */
704 EPOLLPRI | EPOLLRDBAND, /* POLL_PRI */
705 EPOLLHUP | EPOLLERR /* POLL_HUP */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706};
707
708static inline int sigio_perm(struct task_struct *p,
709 struct fown_struct *fown, int sig)
710{
David Howellsc69e8d92008-11-14 10:39:19 +1100711 const struct cred *cred;
712 int ret;
713
714 rcu_read_lock();
715 cred = __task_cred(p);
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -0800716 ret = ((uid_eq(fown->euid, GLOBAL_ROOT_UID) ||
717 uid_eq(fown->euid, cred->suid) || uid_eq(fown->euid, cred->uid) ||
718 uid_eq(fown->uid, cred->suid) || uid_eq(fown->uid, cred->uid)) &&
David Howellsc69e8d92008-11-14 10:39:19 +1100719 !security_file_send_sigiotask(p, fown, sig));
720 rcu_read_unlock();
721 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722}
723
724static void send_sigio_to_task(struct task_struct *p,
Oleg Nesterov8eeee4e2009-06-17 00:27:10 +0200725 struct fown_struct *fown,
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700726 int fd, int reason, int group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727{
Oleg Nesterov8eeee4e2009-06-17 00:27:10 +0200728 /*
729 * F_SETSIG can change ->signum lockless in parallel, make
730 * sure we read it once and use the same value throughout.
731 */
Mark Rutland6aa7de02017-10-23 14:07:29 -0700732 int signum = READ_ONCE(fown->signum);
Oleg Nesterov8eeee4e2009-06-17 00:27:10 +0200733
734 if (!sigio_perm(p, fown, signum))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 return;
736
Oleg Nesterov8eeee4e2009-06-17 00:27:10 +0200737 switch (signum) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 siginfo_t si;
739 default:
740 /* Queue a rt signal with the appropriate fd as its
741 value. We use SI_SIGIO as the source, not
742 SI_KERNEL, since kernel signals always get
743 delivered even if we can't queue. Failure to
744 queue in this case _should_ be reported; we fall
745 back to SIGIO in that case. --sct */
Eric W. Biedermanfaf1f222018-01-05 17:27:42 -0600746 clear_siginfo(&si);
Oleg Nesterov8eeee4e2009-06-17 00:27:10 +0200747 si.si_signo = signum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 si.si_errno = 0;
749 si.si_code = reason;
Eric W. Biedermand08477a2017-06-29 09:28:50 -0500750 /*
751 * Posix definies POLL_IN and friends to be signal
752 * specific si_codes for SIG_POLL. Linux extended
753 * these si_codes to other signals in a way that is
754 * ambiguous if other signals also have signal
755 * specific si_codes. In that case use SI_SIGIO instead
756 * to remove the ambiguity.
757 */
Eric W. Biederman54640d22017-09-18 22:51:14 -0500758 if ((signum != SIGPOLL) && sig_specific_sicodes(signum))
Eric W. Biedermand08477a2017-06-29 09:28:50 -0500759 si.si_code = SI_SIGIO;
760
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 /* Make sure we are called with one of the POLL_*
762 reasons, otherwise we could leak kernel stack into
763 userspace. */
Eric W. Biedermand08477a2017-06-29 09:28:50 -0500764 BUG_ON((reason < POLL_IN) || ((reason - POLL_IN) >= NSIGPOLL));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 if (reason - POLL_IN >= NSIGPOLL)
766 si.si_band = ~0L;
767 else
Al Viroc71d2272017-11-29 19:00:41 -0500768 si.si_band = mangle_poll(band_table[reason - POLL_IN]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 si.si_fd = fd;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700770 if (!do_send_sig_info(signum, &si, p, group))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 break;
772 /* fall-through: fall back on the old plain SIGIO signal */
773 case 0:
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700774 do_send_sig_info(SIGIO, SEND_SIG_PRIV, p, group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 }
776}
777
778void send_sigio(struct fown_struct *fown, int fd, int band)
779{
780 struct task_struct *p;
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700781 enum pid_type type;
782 struct pid *pid;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700783 int group = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784
785 read_lock(&fown->lock);
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700786
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700787 type = fown->pid_type;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700788 if (type == PIDTYPE_MAX) {
789 group = 0;
790 type = PIDTYPE_PID;
791 }
792
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 pid = fown->pid;
794 if (!pid)
795 goto out_unlock_fown;
796
797 read_lock(&tasklist_lock);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700798 do_each_pid_task(pid, type, p) {
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700799 send_sigio_to_task(p, fown, fd, band, group);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700800 } while_each_pid_task(pid, type, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 read_unlock(&tasklist_lock);
802 out_unlock_fown:
803 read_unlock(&fown->lock);
804}
805
806static void send_sigurg_to_task(struct task_struct *p,
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700807 struct fown_struct *fown, int group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808{
809 if (sigio_perm(p, fown, SIGURG))
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700810 do_send_sig_info(SIGURG, SEND_SIG_PRIV, p, group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811}
812
813int send_sigurg(struct fown_struct *fown)
814{
815 struct task_struct *p;
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700816 enum pid_type type;
817 struct pid *pid;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700818 int group = 1;
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700819 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820
821 read_lock(&fown->lock);
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700822
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700823 type = fown->pid_type;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700824 if (type == PIDTYPE_MAX) {
825 group = 0;
826 type = PIDTYPE_PID;
827 }
828
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 pid = fown->pid;
830 if (!pid)
831 goto out_unlock_fown;
832
833 ret = 1;
834
835 read_lock(&tasklist_lock);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700836 do_each_pid_task(pid, type, p) {
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700837 send_sigurg_to_task(p, fown, group);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700838 } while_each_pid_task(pid, type, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 read_unlock(&tasklist_lock);
840 out_unlock_fown:
841 read_unlock(&fown->lock);
842 return ret;
843}
844
Eric Dumazet989a2972010-04-14 09:55:35 +0000845static DEFINE_SPINLOCK(fasync_lock);
Christoph Lametere18b8902006-12-06 20:33:20 -0800846static struct kmem_cache *fasync_cache __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847
Eric Dumazet989a2972010-04-14 09:55:35 +0000848static void fasync_free_rcu(struct rcu_head *head)
849{
850 kmem_cache_free(fasync_cache,
851 container_of(head, struct fasync_struct, fa_rcu));
852}
853
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854/*
Linus Torvalds53281b62009-12-16 08:23:37 -0800855 * Remove a fasync entry. If successfully removed, return
856 * positive and clear the FASYNC flag. If no entry exists,
857 * do nothing and return 0.
858 *
859 * NOTE! It is very important that the FASYNC flag always
860 * match the state "is the filp on a fasync list".
861 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 */
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400863int fasync_remove_entry(struct file *filp, struct fasync_struct **fapp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864{
865 struct fasync_struct *fa, **fp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 int result = 0;
867
Jonathan Corbet4a6a4492009-03-27 12:24:31 -0600868 spin_lock(&filp->f_lock);
Eric Dumazet989a2972010-04-14 09:55:35 +0000869 spin_lock(&fasync_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870 for (fp = fapp; (fa = *fp) != NULL; fp = &fa->fa_next) {
Linus Torvalds53281b62009-12-16 08:23:37 -0800871 if (fa->fa_file != filp)
872 continue;
Eric Dumazet989a2972010-04-14 09:55:35 +0000873
Kirill Tkhai7a107c02018-04-05 14:58:06 +0300874 write_lock_irq(&fa->fa_lock);
Eric Dumazet989a2972010-04-14 09:55:35 +0000875 fa->fa_file = NULL;
Kirill Tkhai7a107c02018-04-05 14:58:06 +0300876 write_unlock_irq(&fa->fa_lock);
Eric Dumazet989a2972010-04-14 09:55:35 +0000877
Linus Torvalds53281b62009-12-16 08:23:37 -0800878 *fp = fa->fa_next;
Eric Dumazet989a2972010-04-14 09:55:35 +0000879 call_rcu(&fa->fa_rcu, fasync_free_rcu);
Jonathan Corbet76398422009-02-01 14:26:59 -0700880 filp->f_flags &= ~FASYNC;
Linus Torvalds53281b62009-12-16 08:23:37 -0800881 result = 1;
882 break;
883 }
Eric Dumazet989a2972010-04-14 09:55:35 +0000884 spin_unlock(&fasync_lock);
Jonathan Corbet4a6a4492009-03-27 12:24:31 -0600885 spin_unlock(&filp->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 return result;
887}
888
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400889struct fasync_struct *fasync_alloc(void)
Linus Torvalds53281b62009-12-16 08:23:37 -0800890{
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400891 return kmem_cache_alloc(fasync_cache, GFP_KERNEL);
892}
Linus Torvalds53281b62009-12-16 08:23:37 -0800893
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400894/*
895 * NOTE! This can be used only for unused fasync entries:
896 * entries that actually got inserted on the fasync list
897 * need to be released by rcu - see fasync_remove_entry.
898 */
899void fasync_free(struct fasync_struct *new)
900{
901 kmem_cache_free(fasync_cache, new);
902}
903
904/*
905 * Insert a new entry into the fasync list. Return the pointer to the
906 * old one if we didn't use the new one.
Linus Torvalds55f335a2010-10-27 18:17:02 -0700907 *
908 * NOTE! It is very important that the FASYNC flag always
909 * match the state "is the filp on a fasync list".
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400910 */
911struct fasync_struct *fasync_insert_entry(int fd, struct file *filp, struct fasync_struct **fapp, struct fasync_struct *new)
912{
913 struct fasync_struct *fa, **fp;
Linus Torvalds53281b62009-12-16 08:23:37 -0800914
915 spin_lock(&filp->f_lock);
Eric Dumazet989a2972010-04-14 09:55:35 +0000916 spin_lock(&fasync_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800917 for (fp = fapp; (fa = *fp) != NULL; fp = &fa->fa_next) {
918 if (fa->fa_file != filp)
919 continue;
Eric Dumazet989a2972010-04-14 09:55:35 +0000920
Kirill Tkhai7a107c02018-04-05 14:58:06 +0300921 write_lock_irq(&fa->fa_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800922 fa->fa_fd = fd;
Kirill Tkhai7a107c02018-04-05 14:58:06 +0300923 write_unlock_irq(&fa->fa_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800924 goto out;
925 }
926
Kirill Tkhai7a107c02018-04-05 14:58:06 +0300927 rwlock_init(&new->fa_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800928 new->magic = FASYNC_MAGIC;
929 new->fa_file = filp;
930 new->fa_fd = fd;
931 new->fa_next = *fapp;
Eric Dumazet989a2972010-04-14 09:55:35 +0000932 rcu_assign_pointer(*fapp, new);
Linus Torvalds53281b62009-12-16 08:23:37 -0800933 filp->f_flags |= FASYNC;
934
935out:
Eric Dumazet989a2972010-04-14 09:55:35 +0000936 spin_unlock(&fasync_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800937 spin_unlock(&filp->f_lock);
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400938 return fa;
939}
940
941/*
942 * Add a fasync entry. Return negative on error, positive if
943 * added, and zero if did nothing but change an existing one.
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400944 */
945static int fasync_add_entry(int fd, struct file *filp, struct fasync_struct **fapp)
946{
947 struct fasync_struct *new;
948
949 new = fasync_alloc();
950 if (!new)
951 return -ENOMEM;
952
953 /*
954 * fasync_insert_entry() returns the old (update) entry if
955 * it existed.
956 *
957 * So free the (unused) new entry and return 0 to let the
958 * caller know that we didn't add any new fasync entries.
959 */
960 if (fasync_insert_entry(fd, filp, fapp, new)) {
961 fasync_free(new);
962 return 0;
963 }
964
965 return 1;
Linus Torvalds53281b62009-12-16 08:23:37 -0800966}
967
968/*
969 * fasync_helper() is used by almost all character device drivers
970 * to set up the fasync queue, and for regular files by the file
971 * lease code. It returns negative on error, 0 if it did no changes
972 * and positive if it added/deleted the entry.
973 */
974int fasync_helper(int fd, struct file * filp, int on, struct fasync_struct **fapp)
975{
976 if (!on)
977 return fasync_remove_entry(filp, fapp);
978 return fasync_add_entry(fd, filp, fapp);
979}
980
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981EXPORT_SYMBOL(fasync_helper);
982
Eric Dumazet989a2972010-04-14 09:55:35 +0000983/*
984 * rcu_read_lock() is held
985 */
986static void kill_fasync_rcu(struct fasync_struct *fa, int sig, int band)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987{
988 while (fa) {
Eric Dumazet989a2972010-04-14 09:55:35 +0000989 struct fown_struct *fown;
Andrew Mortonf4985dc2010-06-29 15:05:42 -0700990
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 if (fa->magic != FASYNC_MAGIC) {
992 printk(KERN_ERR "kill_fasync: bad magic number in "
993 "fasync_struct!\n");
994 return;
995 }
Kirill Tkhai7a107c02018-04-05 14:58:06 +0300996 read_lock(&fa->fa_lock);
Eric Dumazet989a2972010-04-14 09:55:35 +0000997 if (fa->fa_file) {
998 fown = &fa->fa_file->f_owner;
999 /* Don't send SIGURG to processes which have not set a
1000 queued signum: SIGURG has its own default signalling
1001 mechanism. */
1002 if (!(sig == SIGURG && fown->signum == 0))
1003 send_sigio(fown, fa->fa_fd, band);
1004 }
Kirill Tkhai7a107c02018-04-05 14:58:06 +03001005 read_unlock(&fa->fa_lock);
Eric Dumazet989a2972010-04-14 09:55:35 +00001006 fa = rcu_dereference(fa->fa_next);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 }
1008}
1009
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010void kill_fasync(struct fasync_struct **fp, int sig, int band)
1011{
1012 /* First a quick test without locking: usually
1013 * the list is empty.
1014 */
1015 if (*fp) {
Eric Dumazet989a2972010-04-14 09:55:35 +00001016 rcu_read_lock();
1017 kill_fasync_rcu(rcu_dereference(*fp), sig, band);
1018 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 }
1020}
1021EXPORT_SYMBOL(kill_fasync);
1022
Wu Fengguang454eedb2010-08-10 18:01:29 -07001023static int __init fcntl_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024{
James Bottomley3ab04d52010-09-09 16:38:12 -07001025 /*
1026 * Please add new bits here to ensure allocation uniqueness.
1027 * Exceptions: O_NONBLOCK is a two bit define on parisc; O_NDELAY
1028 * is defined as O_NONBLOCK on some platforms and not on others.
1029 */
Christoph Hellwig80f18372017-04-27 09:42:24 +02001030 BUILD_BUG_ON(21 - 1 /* for O_RDONLY being 0 */ !=
1031 HWEIGHT32(
1032 (VALID_OPEN_FLAGS & ~(O_NONBLOCK | O_NDELAY)) |
1033 __FMODE_EXEC | __FMODE_NONOTIFY));
Wu Fengguang454eedb2010-08-10 18:01:29 -07001034
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 fasync_cache = kmem_cache_create("fasync_cache",
Paul Mundt20c2df82007-07-20 10:11:58 +09001036 sizeof(struct fasync_struct), 0, SLAB_PANIC, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 return 0;
1038}
1039
Wu Fengguang454eedb2010-08-10 18:01:29 -07001040module_init(fcntl_init)