blob: 3b0bf82aab1ed0473d947f6e3b5ce28b8d944c6f [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>
David Herrmannc9ccca22014-08-08 14:25:27 -070023#include <linux/shmem_fs.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
Harvey Harrisonfc9b52c2008-02-08 04:19:52 -080029void set_close_on_exec(unsigned int fd, int flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -070030{
31 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -070032 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -070033 spin_lock(&files->file_lock);
Dipankar Sarmabadf1662005-09-09 13:04:10 -070034 fdt = files_fdtable(files);
Linus Torvalds1da177e2005-04-16 15:20:36 -070035 if (flag)
David Howells1dce27c2012-02-16 17:49:42 +000036 __set_close_on_exec(fd, fdt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070037 else
David Howells1dce27c2012-02-16 17:49:42 +000038 __clear_close_on_exec(fd, fdt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070039 spin_unlock(&files->file_lock);
40}
41
David Howells1dce27c2012-02-16 17:49:42 +000042static bool get_close_on_exec(unsigned int fd)
Linus Torvalds1da177e2005-04-16 15:20:36 -070043{
44 struct files_struct *files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -070045 struct fdtable *fdt;
David Howells1dce27c2012-02-16 17:49:42 +000046 bool res;
Dipankar Sarmab8359962005-09-09 13:04:14 -070047 rcu_read_lock();
Dipankar Sarmabadf1662005-09-09 13:04:10 -070048 fdt = files_fdtable(files);
David Howells1dce27c2012-02-16 17:49:42 +000049 res = close_on_exec(fd, fdt);
Dipankar Sarmab8359962005-09-09 13:04:14 -070050 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -070051 return res;
52}
53
Heiko Carstensa26eab22009-01-14 14:14:17 +010054SYSCALL_DEFINE3(dup3, unsigned int, oldfd, unsigned int, newfd, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -070055{
56 int err = -EBADF;
57 struct file * file, *tofree;
58 struct files_struct * files = current->files;
Dipankar Sarmabadf1662005-09-09 13:04:10 -070059 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -070060
Ulrich Drepper336dd1f2008-07-23 21:29:29 -070061 if ((flags & ~O_CLOEXEC) != 0)
62 return -EINVAL;
63
Al Viro6c5d0512008-07-26 13:38:19 -040064 if (unlikely(oldfd == newfd))
65 return -EINVAL;
66
Linus Torvalds1da177e2005-04-16 15:20:36 -070067 spin_lock(&files->file_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 err = expand_files(files, newfd);
Al Viro1b7e1902008-07-30 06:18:03 -040069 file = fcheck(oldfd);
70 if (unlikely(!file))
71 goto Ebadf;
Al Viro4e1e0182008-07-26 16:01:20 -040072 if (unlikely(err < 0)) {
73 if (err == -EMFILE)
Al Viro1b7e1902008-07-30 06:18:03 -040074 goto Ebadf;
75 goto out_unlock;
Al Viro4e1e0182008-07-26 16:01:20 -040076 }
Al Viro1b7e1902008-07-30 06:18:03 -040077 /*
78 * We need to detect attempts to do dup2() over allocated but still
79 * not finished descriptor. NB: OpenBSD avoids that at the price of
80 * extra work in their equivalent of fget() - they insert struct
81 * file immediately after grabbing descriptor, mark it larval if
82 * more work (e.g. actual opening) is needed and make sure that
83 * fget() treats larval files as absent. Potentially interesting,
84 * but while extra work in fget() is trivial, locking implications
85 * and amount of surgery on open()-related paths in VFS are not.
86 * FreeBSD fails with -EBADF in the same situation, NetBSD "solution"
87 * deadlocks in rather amusing ways, AFAICS. All of that is out of
88 * scope of POSIX or SUS, since neither considers shared descriptor
89 * tables and this condition does not arise without those.
90 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070091 err = -EBUSY;
Dipankar Sarmabadf1662005-09-09 13:04:10 -070092 fdt = files_fdtable(files);
93 tofree = fdt->fd[newfd];
David Howells1dce27c2012-02-16 17:49:42 +000094 if (!tofree && fd_is_open(newfd, fdt))
Al Viro1b7e1902008-07-30 06:18:03 -040095 goto out_unlock;
96 get_file(file);
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070097 rcu_assign_pointer(fdt->fd[newfd], file);
David Howells1dce27c2012-02-16 17:49:42 +000098 __set_open_fd(newfd, fdt);
Ulrich Drepper336dd1f2008-07-23 21:29:29 -070099 if (flags & O_CLOEXEC)
David Howells1dce27c2012-02-16 17:49:42 +0000100 __set_close_on_exec(newfd, fdt);
Ulrich Drepper336dd1f2008-07-23 21:29:29 -0700101 else
David Howells1dce27c2012-02-16 17:49:42 +0000102 __clear_close_on_exec(newfd, fdt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103 spin_unlock(&files->file_lock);
104
105 if (tofree)
106 filp_close(tofree, files);
Al Viro1b7e1902008-07-30 06:18:03 -0400107
108 return newfd;
109
110Ebadf:
111 err = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112out_unlock:
113 spin_unlock(&files->file_lock);
Al Viro1b7e1902008-07-30 06:18:03 -0400114 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115}
116
Heiko Carstensa26eab22009-01-14 14:14:17 +0100117SYSCALL_DEFINE2(dup2, unsigned int, oldfd, unsigned int, newfd)
Ulrich Drepper336dd1f2008-07-23 21:29:29 -0700118{
Al Viro6c5d0512008-07-26 13:38:19 -0400119 if (unlikely(newfd == oldfd)) { /* corner case */
120 struct files_struct *files = current->files;
Jeff Mahoney2b79bc42009-05-11 14:25:34 -0400121 int retval = oldfd;
122
Al Viro6c5d0512008-07-26 13:38:19 -0400123 rcu_read_lock();
124 if (!fcheck_files(files, oldfd))
Jeff Mahoney2b79bc42009-05-11 14:25:34 -0400125 retval = -EBADF;
Al Viro6c5d0512008-07-26 13:38:19 -0400126 rcu_read_unlock();
Jeff Mahoney2b79bc42009-05-11 14:25:34 -0400127 return retval;
Al Viro6c5d0512008-07-26 13:38:19 -0400128 }
Ulrich Drepper336dd1f2008-07-23 21:29:29 -0700129 return sys_dup3(oldfd, newfd, 0);
130}
131
Heiko Carstensa26eab22009-01-14 14:14:17 +0100132SYSCALL_DEFINE1(dup, unsigned int, fildes)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133{
134 int ret = -EBADF;
Al Viro1abf0c72011-03-13 03:51:11 -0400135 struct file *file = fget_raw(fildes);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136
Al Viro1027abe2008-07-30 04:13:04 -0400137 if (file) {
138 ret = get_unused_fd();
139 if (ret >= 0)
140 fd_install(ret, file);
141 else
142 fput(file);
143 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144 return ret;
145}
146
Jonathan Corbet76398422009-02-01 14:26:59 -0700147#define SETFL_MASK (O_APPEND | O_NONBLOCK | O_NDELAY | O_DIRECT | O_NOATIME)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148
149static int setfl(int fd, struct file * filp, unsigned long arg)
150{
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -0800151 struct inode * inode = filp->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 int error = 0;
153
dean gaudet7d95c8f2006-02-03 03:04:30 -0800154 /*
155 * O_APPEND cannot be cleared if the file is marked as append-only
156 * and the file is open for write.
157 */
158 if (((arg ^ filp->f_flags) & O_APPEND) && IS_APPEND(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159 return -EPERM;
160
161 /* O_NOATIME can only be set by the owner or superuser */
162 if ((arg & O_NOATIME) && !(filp->f_flags & O_NOATIME))
Serge E. Hallyn2e149672011-03-23 16:43:26 -0700163 if (!inode_owner_or_capable(inode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164 return -EPERM;
165
166 /* required for strict SunOS emulation */
167 if (O_NONBLOCK != O_NDELAY)
168 if (arg & O_NDELAY)
169 arg |= O_NONBLOCK;
170
171 if (arg & O_DIRECT) {
172 if (!filp->f_mapping || !filp->f_mapping->a_ops ||
173 !filp->f_mapping->a_ops->direct_IO)
174 return -EINVAL;
175 }
176
177 if (filp->f_op && filp->f_op->check_flags)
178 error = filp->f_op->check_flags(arg);
179 if (error)
180 return error;
181
Jonathan Corbet218d11a2008-12-05 16:12:48 -0700182 /*
Jonathan Corbet76398422009-02-01 14:26:59 -0700183 * ->fasync() is responsible for setting the FASYNC bit.
Jonathan Corbet218d11a2008-12-05 16:12:48 -0700184 */
Jonathan Corbet76398422009-02-01 14:26:59 -0700185 if (((arg ^ filp->f_flags) & FASYNC) && filp->f_op &&
186 filp->f_op->fasync) {
187 error = filp->f_op->fasync(fd, filp, (arg & FASYNC) != 0);
188 if (error < 0)
189 goto out;
Jonathan Corbet60aa4922009-02-01 14:52:56 -0700190 if (error > 0)
191 error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 }
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -0700193 spin_lock(&filp->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 filp->f_flags = (arg & SETFL_MASK) | (filp->f_flags & ~SETFL_MASK);
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -0700195 spin_unlock(&filp->f_lock);
Jonathan Corbet76398422009-02-01 14:26:59 -0700196
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 return error;
199}
200
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700201static void f_modown(struct file *filp, struct pid *pid, enum pid_type type,
Oleg Nesterov2f38d702009-06-16 22:07:46 +0200202 int force)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203{
Linus Torvalds80e1e822010-02-07 10:11:23 -0800204 write_lock_irq(&filp->f_owner.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 if (force || !filp->f_owner.pid) {
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700206 put_pid(filp->f_owner.pid);
207 filp->f_owner.pid = get_pid(pid);
208 filp->f_owner.pid_type = type;
Oleg Nesterov2f38d702009-06-16 22:07:46 +0200209
210 if (pid) {
211 const struct cred *cred = current_cred();
212 filp->f_owner.uid = cred->uid;
213 filp->f_owner.euid = cred->euid;
214 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215 }
Linus Torvalds80e1e822010-02-07 10:11:23 -0800216 write_unlock_irq(&filp->f_owner.lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217}
218
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700219int __f_setown(struct file *filp, struct pid *pid, enum pid_type type,
220 int force)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221{
222 int err;
Oleg Nesterov2f38d702009-06-16 22:07:46 +0200223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 err = security_file_set_fowner(filp);
225 if (err)
226 return err;
227
Oleg Nesterov2f38d702009-06-16 22:07:46 +0200228 f_modown(filp, pid, type, force);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 return 0;
230}
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700231EXPORT_SYMBOL(__f_setown);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700233int f_setown(struct file *filp, unsigned long arg, int force)
234{
235 enum pid_type type;
236 struct pid *pid;
237 int who = arg;
238 int result;
239 type = PIDTYPE_PID;
240 if (who < 0) {
241 type = PIDTYPE_PGID;
242 who = -who;
243 }
244 rcu_read_lock();
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700245 pid = find_vpid(who);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700246 result = __f_setown(filp, pid, type, force);
247 rcu_read_unlock();
248 return result;
249}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250EXPORT_SYMBOL(f_setown);
251
252void f_delown(struct file *filp)
253{
Oleg Nesterov2f38d702009-06-16 22:07:46 +0200254 f_modown(filp, NULL, PIDTYPE_PID, 1);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700255}
256
257pid_t f_getown(struct file *filp)
258{
259 pid_t pid;
Eric W. Biederman43fa1ad2006-10-02 02:17:27 -0700260 read_lock(&filp->f_owner.lock);
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800261 pid = pid_vnr(filp->f_owner.pid);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700262 if (filp->f_owner.pid_type == PIDTYPE_PGID)
263 pid = -pid;
Eric W. Biederman43fa1ad2006-10-02 02:17:27 -0700264 read_unlock(&filp->f_owner.lock);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700265 return pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266}
267
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700268static int f_setown_ex(struct file *filp, unsigned long arg)
269{
270 struct f_owner_ex * __user owner_p = (void * __user)arg;
271 struct f_owner_ex owner;
272 struct pid *pid;
273 int type;
274 int ret;
275
276 ret = copy_from_user(&owner, owner_p, sizeof(owner));
277 if (ret)
Dan Carpenter5b544702010-06-03 12:35:42 +0200278 return -EFAULT;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700279
280 switch (owner.type) {
281 case F_OWNER_TID:
282 type = PIDTYPE_MAX;
283 break;
284
285 case F_OWNER_PID:
286 type = PIDTYPE_PID;
287 break;
288
Peter Zijlstra978b4052009-11-17 14:06:24 -0800289 case F_OWNER_PGRP:
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700290 type = PIDTYPE_PGID;
291 break;
292
293 default:
294 return -EINVAL;
295 }
296
297 rcu_read_lock();
298 pid = find_vpid(owner.pid);
299 if (owner.pid && !pid)
300 ret = -ESRCH;
301 else
302 ret = __f_setown(filp, pid, type, 1);
303 rcu_read_unlock();
304
305 return ret;
306}
307
308static int f_getown_ex(struct file *filp, unsigned long arg)
309{
310 struct f_owner_ex * __user owner_p = (void * __user)arg;
311 struct f_owner_ex owner;
312 int ret = 0;
313
314 read_lock(&filp->f_owner.lock);
315 owner.pid = pid_vnr(filp->f_owner.pid);
316 switch (filp->f_owner.pid_type) {
317 case PIDTYPE_MAX:
318 owner.type = F_OWNER_TID;
319 break;
320
321 case PIDTYPE_PID:
322 owner.type = F_OWNER_PID;
323 break;
324
325 case PIDTYPE_PGID:
Peter Zijlstra978b4052009-11-17 14:06:24 -0800326 owner.type = F_OWNER_PGRP;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700327 break;
328
329 default:
330 WARN_ON(1);
331 ret = -EINVAL;
332 break;
333 }
334 read_unlock(&filp->f_owner.lock);
335
Dan Carpenter5b544702010-06-03 12:35:42 +0200336 if (!ret) {
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700337 ret = copy_to_user(owner_p, &owner, sizeof(owner));
Dan Carpenter5b544702010-06-03 12:35:42 +0200338 if (ret)
339 ret = -EFAULT;
340 }
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700341 return ret;
342}
343
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344static long do_fcntl(int fd, unsigned int cmd, unsigned long arg,
345 struct file *filp)
346{
347 long err = -EINVAL;
348
349 switch (cmd) {
350 case F_DUPFD:
Ulrich Drepper22d2b352007-10-16 23:30:26 -0700351 case F_DUPFD_CLOEXEC:
Jiri Slabyd554ed892010-03-05 13:42:42 -0800352 if (arg >= rlimit(RLIMIT_NOFILE))
Al Viro4e1e0182008-07-26 16:01:20 -0400353 break;
Al Viro1027abe2008-07-30 04:13:04 -0400354 err = alloc_fd(arg, cmd == F_DUPFD_CLOEXEC ? O_CLOEXEC : 0);
355 if (err >= 0) {
356 get_file(filp);
357 fd_install(err, filp);
358 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 break;
360 case F_GETFD:
361 err = get_close_on_exec(fd) ? FD_CLOEXEC : 0;
362 break;
363 case F_SETFD:
364 err = 0;
365 set_close_on_exec(fd, arg & FD_CLOEXEC);
366 break;
367 case F_GETFL:
368 err = filp->f_flags;
369 break;
370 case F_SETFL:
371 err = setfl(fd, filp, arg);
372 break;
Jeff Laytonf68ae182014-02-03 12:13:10 -0500373#if BITS_PER_LONG != 32
374 /* 32-bit arches must use fcntl64() */
Jeff Layton744c99d2014-04-22 08:23:58 -0400375 case F_OFD_GETLK:
Jeff Laytonf68ae182014-02-03 12:13:10 -0500376#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377 case F_GETLK:
Jeff Laytona3b0c802014-02-03 12:13:09 -0500378 err = fcntl_getlk(filp, cmd, (struct flock __user *) arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 break;
Jeff Laytonf68ae182014-02-03 12:13:10 -0500380#if BITS_PER_LONG != 32
381 /* 32-bit arches must use fcntl64() */
Jeff Layton744c99d2014-04-22 08:23:58 -0400382 case F_OFD_SETLK:
383 case F_OFD_SETLKW:
Jeff Laytonf68ae182014-02-03 12:13:10 -0500384#endif
385 /* Fallthrough */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 case F_SETLK:
387 case F_SETLKW:
Peter Staubachc2936212005-07-27 11:45:09 -0700388 err = fcntl_setlk(fd, filp, cmd, (struct flock __user *) arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 break;
390 case F_GETOWN:
391 /*
392 * XXX If f_owner is a process group, the
393 * negative return value will get converted
394 * into an error. Oops. If we keep the
395 * current syscall conventions, the only way
396 * to fix this will be in libc.
397 */
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700398 err = f_getown(filp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 force_successful_syscall_return();
400 break;
401 case F_SETOWN:
402 err = f_setown(filp, arg, 1);
403 break;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700404 case F_GETOWN_EX:
405 err = f_getown_ex(filp, arg);
406 break;
407 case F_SETOWN_EX:
408 err = f_setown_ex(filp, arg);
409 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 case F_GETSIG:
411 err = filp->f_owner.signum;
412 break;
413 case F_SETSIG:
414 /* arg == 0 restores default behaviour. */
Jesper Juhl7ed20e12005-05-01 08:59:14 -0700415 if (!valid_signal(arg)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 break;
417 }
418 err = 0;
419 filp->f_owner.signum = arg;
420 break;
421 case F_GETLEASE:
422 err = fcntl_getlease(filp);
423 break;
424 case F_SETLEASE:
425 err = fcntl_setlease(fd, filp, arg);
426 break;
427 case F_NOTIFY:
428 err = fcntl_dirnotify(fd, filp, arg);
429 break;
Jens Axboe35f3d142010-05-20 10:43:18 +0200430 case F_SETPIPE_SZ:
431 case F_GETPIPE_SZ:
432 err = pipe_fcntl(filp, cmd, arg);
433 break;
David Herrmannc9ccca22014-08-08 14:25:27 -0700434 case F_ADD_SEALS:
435 case F_GET_SEALS:
436 err = shmem_fcntl(filp, cmd, arg);
437 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 default:
439 break;
440 }
441 return err;
442}
443
Al Viro1abf0c72011-03-13 03:51:11 -0400444static int check_fcntl_cmd(unsigned cmd)
445{
446 switch (cmd) {
447 case F_DUPFD:
448 case F_DUPFD_CLOEXEC:
449 case F_GETFD:
450 case F_SETFD:
451 case F_GETFL:
452 return 1;
453 }
454 return 0;
455}
456
Heiko Carstensa26eab22009-01-14 14:14:17 +0100457SYSCALL_DEFINE3(fcntl, unsigned int, fd, unsigned int, cmd, unsigned long, arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458{
459 struct file *filp;
460 long err = -EBADF;
461
Al Viro1abf0c72011-03-13 03:51:11 -0400462 filp = fget_raw(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 if (!filp)
464 goto out;
465
Al Viro1abf0c72011-03-13 03:51:11 -0400466 if (unlikely(filp->f_mode & FMODE_PATH)) {
467 if (!check_fcntl_cmd(cmd)) {
468 fput(filp);
469 goto out;
470 }
471 }
472
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 err = security_file_fcntl(filp, cmd, arg);
474 if (err) {
475 fput(filp);
476 return err;
477 }
478
479 err = do_fcntl(fd, cmd, arg, filp);
480
481 fput(filp);
482out:
483 return err;
484}
485
486#if BITS_PER_LONG == 32
Heiko Carstensa26eab22009-01-14 14:14:17 +0100487SYSCALL_DEFINE3(fcntl64, unsigned int, fd, unsigned int, cmd,
488 unsigned long, arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489{
490 struct file * filp;
491 long err;
492
493 err = -EBADF;
Al Viro1abf0c72011-03-13 03:51:11 -0400494 filp = fget_raw(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 if (!filp)
496 goto out;
497
Al Viro1abf0c72011-03-13 03:51:11 -0400498 if (unlikely(filp->f_mode & FMODE_PATH)) {
499 if (!check_fcntl_cmd(cmd)) {
500 fput(filp);
501 goto out;
502 }
503 }
504
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 err = security_file_fcntl(filp, cmd, arg);
506 if (err) {
507 fput(filp);
508 return err;
509 }
510 err = -EBADF;
511
512 switch (cmd) {
Jeff Laytonf68ae182014-02-03 12:13:10 -0500513 case F_GETLK64:
Jeff Layton744c99d2014-04-22 08:23:58 -0400514 case F_OFD_GETLK:
Jeff Laytonf68ae182014-02-03 12:13:10 -0500515 err = fcntl_getlk64(filp, cmd, (struct flock64 __user *) arg);
516 break;
517 case F_SETLK64:
518 case F_SETLKW64:
Jeff Layton744c99d2014-04-22 08:23:58 -0400519 case F_OFD_SETLK:
520 case F_OFD_SETLKW:
Jeff Laytonf68ae182014-02-03 12:13:10 -0500521 err = fcntl_setlk64(fd, filp, cmd,
522 (struct flock64 __user *) arg);
523 break;
524 default:
525 err = do_fcntl(fd, cmd, arg, filp);
526 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 }
528 fput(filp);
529out:
530 return err;
531}
532#endif
533
534/* Table to convert sigio signal codes into poll band bitmaps */
535
Eric Dumazetfa3536c2006-03-26 01:37:24 -0800536static const long band_table[NSIGPOLL] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 POLLIN | POLLRDNORM, /* POLL_IN */
538 POLLOUT | POLLWRNORM | POLLWRBAND, /* POLL_OUT */
539 POLLIN | POLLRDNORM | POLLMSG, /* POLL_MSG */
540 POLLERR, /* POLL_ERR */
541 POLLPRI | POLLRDBAND, /* POLL_PRI */
542 POLLHUP | POLLERR /* POLL_HUP */
543};
544
545static inline int sigio_perm(struct task_struct *p,
546 struct fown_struct *fown, int sig)
547{
David Howellsc69e8d92008-11-14 10:39:19 +1100548 const struct cred *cred;
549 int ret;
550
551 rcu_read_lock();
552 cred = __task_cred(p);
553 ret = ((fown->euid == 0 ||
554 fown->euid == cred->suid || fown->euid == cred->uid ||
555 fown->uid == cred->suid || fown->uid == cred->uid) &&
556 !security_file_send_sigiotask(p, fown, sig));
557 rcu_read_unlock();
558 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559}
560
561static void send_sigio_to_task(struct task_struct *p,
Oleg Nesterov8eeee4e2009-06-17 00:27:10 +0200562 struct fown_struct *fown,
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700563 int fd, int reason, int group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564{
Oleg Nesterov8eeee4e2009-06-17 00:27:10 +0200565 /*
566 * F_SETSIG can change ->signum lockless in parallel, make
567 * sure we read it once and use the same value throughout.
568 */
569 int signum = ACCESS_ONCE(fown->signum);
570
571 if (!sigio_perm(p, fown, signum))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 return;
573
Oleg Nesterov8eeee4e2009-06-17 00:27:10 +0200574 switch (signum) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 siginfo_t si;
576 default:
577 /* Queue a rt signal with the appropriate fd as its
578 value. We use SI_SIGIO as the source, not
579 SI_KERNEL, since kernel signals always get
580 delivered even if we can't queue. Failure to
581 queue in this case _should_ be reported; we fall
582 back to SIGIO in that case. --sct */
Oleg Nesterov8eeee4e2009-06-17 00:27:10 +0200583 si.si_signo = signum;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 si.si_errno = 0;
585 si.si_code = reason;
586 /* Make sure we are called with one of the POLL_*
587 reasons, otherwise we could leak kernel stack into
588 userspace. */
Eric Sesterhennf6298aa2006-04-02 13:37:19 +0200589 BUG_ON((reason & __SI_MASK) != __SI_POLL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 if (reason - POLL_IN >= NSIGPOLL)
591 si.si_band = ~0L;
592 else
593 si.si_band = band_table[reason - POLL_IN];
594 si.si_fd = fd;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700595 if (!do_send_sig_info(signum, &si, p, group))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 break;
597 /* fall-through: fall back on the old plain SIGIO signal */
598 case 0:
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700599 do_send_sig_info(SIGIO, SEND_SIG_PRIV, p, group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 }
601}
602
603void send_sigio(struct fown_struct *fown, int fd, int band)
604{
605 struct task_struct *p;
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700606 enum pid_type type;
607 struct pid *pid;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700608 int group = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609
610 read_lock(&fown->lock);
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700611
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700612 type = fown->pid_type;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700613 if (type == PIDTYPE_MAX) {
614 group = 0;
615 type = PIDTYPE_PID;
616 }
617
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 pid = fown->pid;
619 if (!pid)
620 goto out_unlock_fown;
621
622 read_lock(&tasklist_lock);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700623 do_each_pid_task(pid, type, p) {
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700624 send_sigio_to_task(p, fown, fd, band, group);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700625 } while_each_pid_task(pid, type, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 read_unlock(&tasklist_lock);
627 out_unlock_fown:
628 read_unlock(&fown->lock);
629}
630
631static void send_sigurg_to_task(struct task_struct *p,
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700632 struct fown_struct *fown, int group)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633{
634 if (sigio_perm(p, fown, SIGURG))
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700635 do_send_sig_info(SIGURG, SEND_SIG_PRIV, p, group);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636}
637
638int send_sigurg(struct fown_struct *fown)
639{
640 struct task_struct *p;
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700641 enum pid_type type;
642 struct pid *pid;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700643 int group = 1;
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700644 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645
646 read_lock(&fown->lock);
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700647
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700648 type = fown->pid_type;
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700649 if (type == PIDTYPE_MAX) {
650 group = 0;
651 type = PIDTYPE_PID;
652 }
653
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 pid = fown->pid;
655 if (!pid)
656 goto out_unlock_fown;
657
658 ret = 1;
659
660 read_lock(&tasklist_lock);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700661 do_each_pid_task(pid, type, p) {
Peter Zijlstraba0a6c92009-09-23 15:57:03 -0700662 send_sigurg_to_task(p, fown, group);
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700663 } while_each_pid_task(pid, type, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 read_unlock(&tasklist_lock);
665 out_unlock_fown:
666 read_unlock(&fown->lock);
667 return ret;
668}
669
Eric Dumazet989a2972010-04-14 09:55:35 +0000670static DEFINE_SPINLOCK(fasync_lock);
Christoph Lametere18b8902006-12-06 20:33:20 -0800671static struct kmem_cache *fasync_cache __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672
Eric Dumazet989a2972010-04-14 09:55:35 +0000673static void fasync_free_rcu(struct rcu_head *head)
674{
675 kmem_cache_free(fasync_cache,
676 container_of(head, struct fasync_struct, fa_rcu));
677}
678
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679/*
Linus Torvalds53281b62009-12-16 08:23:37 -0800680 * Remove a fasync entry. If successfully removed, return
681 * positive and clear the FASYNC flag. If no entry exists,
682 * do nothing and return 0.
683 *
684 * NOTE! It is very important that the FASYNC flag always
685 * match the state "is the filp on a fasync list".
686 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 */
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400688int fasync_remove_entry(struct file *filp, struct fasync_struct **fapp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689{
690 struct fasync_struct *fa, **fp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 int result = 0;
692
Jonathan Corbet4a6a4492009-03-27 12:24:31 -0600693 spin_lock(&filp->f_lock);
Eric Dumazet989a2972010-04-14 09:55:35 +0000694 spin_lock(&fasync_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 for (fp = fapp; (fa = *fp) != NULL; fp = &fa->fa_next) {
Linus Torvalds53281b62009-12-16 08:23:37 -0800696 if (fa->fa_file != filp)
697 continue;
Eric Dumazet989a2972010-04-14 09:55:35 +0000698
699 spin_lock_irq(&fa->fa_lock);
700 fa->fa_file = NULL;
701 spin_unlock_irq(&fa->fa_lock);
702
Linus Torvalds53281b62009-12-16 08:23:37 -0800703 *fp = fa->fa_next;
Eric Dumazet989a2972010-04-14 09:55:35 +0000704 call_rcu(&fa->fa_rcu, fasync_free_rcu);
Jonathan Corbet76398422009-02-01 14:26:59 -0700705 filp->f_flags &= ~FASYNC;
Linus Torvalds53281b62009-12-16 08:23:37 -0800706 result = 1;
707 break;
708 }
Eric Dumazet989a2972010-04-14 09:55:35 +0000709 spin_unlock(&fasync_lock);
Jonathan Corbet4a6a4492009-03-27 12:24:31 -0600710 spin_unlock(&filp->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 return result;
712}
713
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400714struct fasync_struct *fasync_alloc(void)
Linus Torvalds53281b62009-12-16 08:23:37 -0800715{
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400716 return kmem_cache_alloc(fasync_cache, GFP_KERNEL);
717}
Linus Torvalds53281b62009-12-16 08:23:37 -0800718
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400719/*
720 * NOTE! This can be used only for unused fasync entries:
721 * entries that actually got inserted on the fasync list
722 * need to be released by rcu - see fasync_remove_entry.
723 */
724void fasync_free(struct fasync_struct *new)
725{
726 kmem_cache_free(fasync_cache, new);
727}
728
729/*
730 * Insert a new entry into the fasync list. Return the pointer to the
731 * old one if we didn't use the new one.
Linus Torvalds55f335a2010-10-27 18:17:02 -0700732 *
733 * NOTE! It is very important that the FASYNC flag always
734 * match the state "is the filp on a fasync list".
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400735 */
736struct fasync_struct *fasync_insert_entry(int fd, struct file *filp, struct fasync_struct **fapp, struct fasync_struct *new)
737{
738 struct fasync_struct *fa, **fp;
Linus Torvalds53281b62009-12-16 08:23:37 -0800739
740 spin_lock(&filp->f_lock);
Eric Dumazet989a2972010-04-14 09:55:35 +0000741 spin_lock(&fasync_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800742 for (fp = fapp; (fa = *fp) != NULL; fp = &fa->fa_next) {
743 if (fa->fa_file != filp)
744 continue;
Eric Dumazet989a2972010-04-14 09:55:35 +0000745
746 spin_lock_irq(&fa->fa_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800747 fa->fa_fd = fd;
Eric Dumazet989a2972010-04-14 09:55:35 +0000748 spin_unlock_irq(&fa->fa_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800749 goto out;
750 }
751
Eric Dumazet989a2972010-04-14 09:55:35 +0000752 spin_lock_init(&new->fa_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800753 new->magic = FASYNC_MAGIC;
754 new->fa_file = filp;
755 new->fa_fd = fd;
756 new->fa_next = *fapp;
Eric Dumazet989a2972010-04-14 09:55:35 +0000757 rcu_assign_pointer(*fapp, new);
Linus Torvalds53281b62009-12-16 08:23:37 -0800758 filp->f_flags |= FASYNC;
759
760out:
Eric Dumazet989a2972010-04-14 09:55:35 +0000761 spin_unlock(&fasync_lock);
Linus Torvalds53281b62009-12-16 08:23:37 -0800762 spin_unlock(&filp->f_lock);
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400763 return fa;
764}
765
766/*
767 * Add a fasync entry. Return negative on error, positive if
768 * added, and zero if did nothing but change an existing one.
Linus Torvaldsf7347ce2010-10-27 12:38:12 -0400769 */
770static int fasync_add_entry(int fd, struct file *filp, struct fasync_struct **fapp)
771{
772 struct fasync_struct *new;
773
774 new = fasync_alloc();
775 if (!new)
776 return -ENOMEM;
777
778 /*
779 * fasync_insert_entry() returns the old (update) entry if
780 * it existed.
781 *
782 * So free the (unused) new entry and return 0 to let the
783 * caller know that we didn't add any new fasync entries.
784 */
785 if (fasync_insert_entry(fd, filp, fapp, new)) {
786 fasync_free(new);
787 return 0;
788 }
789
790 return 1;
Linus Torvalds53281b62009-12-16 08:23:37 -0800791}
792
793/*
794 * fasync_helper() is used by almost all character device drivers
795 * to set up the fasync queue, and for regular files by the file
796 * lease code. It returns negative on error, 0 if it did no changes
797 * and positive if it added/deleted the entry.
798 */
799int fasync_helper(int fd, struct file * filp, int on, struct fasync_struct **fapp)
800{
801 if (!on)
802 return fasync_remove_entry(filp, fapp);
803 return fasync_add_entry(fd, filp, fapp);
804}
805
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806EXPORT_SYMBOL(fasync_helper);
807
Eric Dumazet989a2972010-04-14 09:55:35 +0000808/*
809 * rcu_read_lock() is held
810 */
811static void kill_fasync_rcu(struct fasync_struct *fa, int sig, int band)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812{
813 while (fa) {
Eric Dumazet989a2972010-04-14 09:55:35 +0000814 struct fown_struct *fown;
Andrew Mortonf4985dc2010-06-29 15:05:42 -0700815 unsigned long flags;
816
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 if (fa->magic != FASYNC_MAGIC) {
818 printk(KERN_ERR "kill_fasync: bad magic number in "
819 "fasync_struct!\n");
820 return;
821 }
Andrew Mortonf4985dc2010-06-29 15:05:42 -0700822 spin_lock_irqsave(&fa->fa_lock, flags);
Eric Dumazet989a2972010-04-14 09:55:35 +0000823 if (fa->fa_file) {
824 fown = &fa->fa_file->f_owner;
825 /* Don't send SIGURG to processes which have not set a
826 queued signum: SIGURG has its own default signalling
827 mechanism. */
828 if (!(sig == SIGURG && fown->signum == 0))
829 send_sigio(fown, fa->fa_fd, band);
830 }
Andrew Mortonf4985dc2010-06-29 15:05:42 -0700831 spin_unlock_irqrestore(&fa->fa_lock, flags);
Eric Dumazet989a2972010-04-14 09:55:35 +0000832 fa = rcu_dereference(fa->fa_next);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 }
834}
835
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836void kill_fasync(struct fasync_struct **fp, int sig, int band)
837{
838 /* First a quick test without locking: usually
839 * the list is empty.
840 */
841 if (*fp) {
Eric Dumazet989a2972010-04-14 09:55:35 +0000842 rcu_read_lock();
843 kill_fasync_rcu(rcu_dereference(*fp), sig, band);
844 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 }
846}
847EXPORT_SYMBOL(kill_fasync);
848
Wu Fengguang454eedb2010-08-10 18:01:29 -0700849static int __init fcntl_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850{
James Bottomley3ab04d52010-09-09 16:38:12 -0700851 /*
852 * Please add new bits here to ensure allocation uniqueness.
853 * Exceptions: O_NONBLOCK is a two bit define on parisc; O_NDELAY
854 * is defined as O_NONBLOCK on some platforms and not on others.
855 */
Zheng Liuf421b792013-07-25 08:13:19 +0800856 BUILD_BUG_ON(20 - 1 /* for O_RDONLY being 0 */ != HWEIGHT32(
Wu Fengguang454eedb2010-08-10 18:01:29 -0700857 O_RDONLY | O_WRONLY | O_RDWR |
858 O_CREAT | O_EXCL | O_NOCTTY |
James Bottomley3ab04d52010-09-09 16:38:12 -0700859 O_TRUNC | O_APPEND | /* O_NONBLOCK | */
Wu Fengguang454eedb2010-08-10 18:01:29 -0700860 __O_SYNC | O_DSYNC | FASYNC |
861 O_DIRECT | O_LARGEFILE | O_DIRECTORY |
862 O_NOFOLLOW | O_NOATIME | O_CLOEXEC |
Zheng Liuf421b792013-07-25 08:13:19 +0800863 __FMODE_EXEC | O_PATH | __O_TMPFILE
Wu Fengguang454eedb2010-08-10 18:01:29 -0700864 ));
865
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 fasync_cache = kmem_cache_create("fasync_cache",
Paul Mundt20c2df82007-07-20 10:11:58 +0900867 sizeof(struct fasync_struct), 0, SLAB_PANIC, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 return 0;
869}
870
Wu Fengguang454eedb2010-08-10 18:01:29 -0700871module_init(fcntl_init)