]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
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> | |
9f3acc31 | 12 | #include <linux/fdtable.h> |
16f7e0fe | 13 | #include <linux/capability.h> |
1da177e4 | 14 | #include <linux/dnotify.h> |
1da177e4 LT |
15 | #include <linux/slab.h> |
16 | #include <linux/module.h> | |
35f3d14d | 17 | #include <linux/pipe_fs_i.h> |
1da177e4 LT |
18 | #include <linux/security.h> |
19 | #include <linux/ptrace.h> | |
7ed20e1a | 20 | #include <linux/signal.h> |
ab2af1f5 | 21 | #include <linux/rcupdate.h> |
b488893a | 22 | #include <linux/pid_namespace.h> |
1d151c33 | 23 | #include <linux/user_namespace.h> |
40e041a2 | 24 | #include <linux/shmem_fs.h> |
1da177e4 LT |
25 | |
26 | #include <asm/poll.h> | |
27 | #include <asm/siginfo.h> | |
28 | #include <asm/uaccess.h> | |
29 | ||
76398425 | 30 | #define SETFL_MASK (O_APPEND | O_NONBLOCK | O_NDELAY | O_DIRECT | O_NOATIME) |
1da177e4 LT |
31 | |
32 | static int setfl(int fd, struct file * filp, unsigned long arg) | |
33 | { | |
496ad9aa | 34 | struct inode * inode = file_inode(filp); |
1da177e4 LT |
35 | int error = 0; |
36 | ||
7d95c8f2 | 37 | /* |
38 | * O_APPEND cannot be cleared if the file is marked as append-only | |
39 | * and the file is open for write. | |
40 | */ | |
41 | if (((arg ^ filp->f_flags) & O_APPEND) && IS_APPEND(inode)) | |
1da177e4 LT |
42 | return -EPERM; |
43 | ||
44 | /* O_NOATIME can only be set by the owner or superuser */ | |
45 | if ((arg & O_NOATIME) && !(filp->f_flags & O_NOATIME)) | |
2e149670 | 46 | if (!inode_owner_or_capable(inode)) |
1da177e4 LT |
47 | return -EPERM; |
48 | ||
49 | /* required for strict SunOS emulation */ | |
50 | if (O_NONBLOCK != O_NDELAY) | |
51 | if (arg & O_NDELAY) | |
52 | arg |= O_NONBLOCK; | |
53 | ||
0dbf5f20 SK |
54 | /* Pipe packetized mode is controlled by O_DIRECT flag */ |
55 | if (!S_ISFIFO(filp->f_inode->i_mode) && (arg & O_DIRECT)) { | |
1da177e4 LT |
56 | if (!filp->f_mapping || !filp->f_mapping->a_ops || |
57 | !filp->f_mapping->a_ops->direct_IO) | |
58 | return -EINVAL; | |
59 | } | |
60 | ||
72c2d531 | 61 | if (filp->f_op->check_flags) |
1da177e4 LT |
62 | error = filp->f_op->check_flags(arg); |
63 | if (error) | |
64 | return error; | |
65 | ||
218d11a8 | 66 | /* |
76398425 | 67 | * ->fasync() is responsible for setting the FASYNC bit. |
218d11a8 | 68 | */ |
72c2d531 | 69 | if (((arg ^ filp->f_flags) & FASYNC) && filp->f_op->fasync) { |
76398425 JC |
70 | error = filp->f_op->fasync(fd, filp, (arg & FASYNC) != 0); |
71 | if (error < 0) | |
72 | goto out; | |
60aa4924 JC |
73 | if (error > 0) |
74 | error = 0; | |
1da177e4 | 75 | } |
db1dd4d3 | 76 | spin_lock(&filp->f_lock); |
1da177e4 | 77 | filp->f_flags = (arg & SETFL_MASK) | (filp->f_flags & ~SETFL_MASK); |
db1dd4d3 | 78 | spin_unlock(&filp->f_lock); |
76398425 | 79 | |
1da177e4 | 80 | out: |
1da177e4 LT |
81 | return error; |
82 | } | |
83 | ||
609d7fa9 | 84 | static void f_modown(struct file *filp, struct pid *pid, enum pid_type type, |
2f38d70f | 85 | int force) |
1da177e4 | 86 | { |
80e1e823 | 87 | write_lock_irq(&filp->f_owner.lock); |
1da177e4 | 88 | if (force || !filp->f_owner.pid) { |
609d7fa9 EB |
89 | put_pid(filp->f_owner.pid); |
90 | filp->f_owner.pid = get_pid(pid); | |
91 | filp->f_owner.pid_type = type; | |
2f38d70f ON |
92 | |
93 | if (pid) { | |
94 | const struct cred *cred = current_cred(); | |
95 | filp->f_owner.uid = cred->uid; | |
96 | filp->f_owner.euid = cred->euid; | |
97 | } | |
1da177e4 | 98 | } |
80e1e823 | 99 | write_unlock_irq(&filp->f_owner.lock); |
1da177e4 LT |
100 | } |
101 | ||
e0b93edd | 102 | void __f_setown(struct file *filp, struct pid *pid, enum pid_type type, |
609d7fa9 | 103 | int force) |
1da177e4 | 104 | { |
e0b93edd | 105 | security_file_set_fowner(filp); |
2f38d70f | 106 | f_modown(filp, pid, type, force); |
1da177e4 | 107 | } |
609d7fa9 | 108 | EXPORT_SYMBOL(__f_setown); |
1da177e4 | 109 | |
e0b93edd | 110 | void f_setown(struct file *filp, unsigned long arg, int force) |
609d7fa9 EB |
111 | { |
112 | enum pid_type type; | |
113 | struct pid *pid; | |
114 | int who = arg; | |
609d7fa9 EB |
115 | type = PIDTYPE_PID; |
116 | if (who < 0) { | |
7b50205c JS |
117 | /* avoid overflow below */ |
118 | if (who == INT_MIN) | |
119 | return; | |
120 | ||
609d7fa9 EB |
121 | type = PIDTYPE_PGID; |
122 | who = -who; | |
123 | } | |
124 | rcu_read_lock(); | |
b488893a | 125 | pid = find_vpid(who); |
e0b93edd | 126 | __f_setown(filp, pid, type, force); |
609d7fa9 | 127 | rcu_read_unlock(); |
609d7fa9 | 128 | } |
1da177e4 LT |
129 | EXPORT_SYMBOL(f_setown); |
130 | ||
131 | void f_delown(struct file *filp) | |
132 | { | |
2f38d70f | 133 | f_modown(filp, NULL, PIDTYPE_PID, 1); |
609d7fa9 EB |
134 | } |
135 | ||
136 | pid_t f_getown(struct file *filp) | |
137 | { | |
138 | pid_t pid; | |
43fa1adb | 139 | read_lock(&filp->f_owner.lock); |
6c5f3e7b | 140 | pid = pid_vnr(filp->f_owner.pid); |
609d7fa9 EB |
141 | if (filp->f_owner.pid_type == PIDTYPE_PGID) |
142 | pid = -pid; | |
43fa1adb | 143 | read_unlock(&filp->f_owner.lock); |
609d7fa9 | 144 | return pid; |
1da177e4 LT |
145 | } |
146 | ||
ba0a6c9f PZ |
147 | static int f_setown_ex(struct file *filp, unsigned long arg) |
148 | { | |
63784dd0 | 149 | struct f_owner_ex __user *owner_p = (void __user *)arg; |
ba0a6c9f PZ |
150 | struct f_owner_ex owner; |
151 | struct pid *pid; | |
152 | int type; | |
153 | int ret; | |
154 | ||
155 | ret = copy_from_user(&owner, owner_p, sizeof(owner)); | |
156 | if (ret) | |
5b54470d | 157 | return -EFAULT; |
ba0a6c9f PZ |
158 | |
159 | switch (owner.type) { | |
160 | case F_OWNER_TID: | |
161 | type = PIDTYPE_MAX; | |
162 | break; | |
163 | ||
164 | case F_OWNER_PID: | |
165 | type = PIDTYPE_PID; | |
166 | break; | |
167 | ||
978b4053 | 168 | case F_OWNER_PGRP: |
ba0a6c9f PZ |
169 | type = PIDTYPE_PGID; |
170 | break; | |
171 | ||
172 | default: | |
173 | return -EINVAL; | |
174 | } | |
175 | ||
176 | rcu_read_lock(); | |
177 | pid = find_vpid(owner.pid); | |
178 | if (owner.pid && !pid) | |
179 | ret = -ESRCH; | |
180 | else | |
e0b93edd | 181 | __f_setown(filp, pid, type, 1); |
ba0a6c9f PZ |
182 | rcu_read_unlock(); |
183 | ||
184 | return ret; | |
185 | } | |
186 | ||
187 | static int f_getown_ex(struct file *filp, unsigned long arg) | |
188 | { | |
63784dd0 | 189 | struct f_owner_ex __user *owner_p = (void __user *)arg; |
ba0a6c9f PZ |
190 | struct f_owner_ex owner; |
191 | int ret = 0; | |
192 | ||
193 | read_lock(&filp->f_owner.lock); | |
194 | owner.pid = pid_vnr(filp->f_owner.pid); | |
195 | switch (filp->f_owner.pid_type) { | |
196 | case PIDTYPE_MAX: | |
197 | owner.type = F_OWNER_TID; | |
198 | break; | |
199 | ||
200 | case PIDTYPE_PID: | |
201 | owner.type = F_OWNER_PID; | |
202 | break; | |
203 | ||
204 | case PIDTYPE_PGID: | |
978b4053 | 205 | owner.type = F_OWNER_PGRP; |
ba0a6c9f PZ |
206 | break; |
207 | ||
208 | default: | |
209 | WARN_ON(1); | |
210 | ret = -EINVAL; | |
211 | break; | |
212 | } | |
213 | read_unlock(&filp->f_owner.lock); | |
214 | ||
5b54470d | 215 | if (!ret) { |
ba0a6c9f | 216 | ret = copy_to_user(owner_p, &owner, sizeof(owner)); |
5b54470d DC |
217 | if (ret) |
218 | ret = -EFAULT; | |
219 | } | |
ba0a6c9f PZ |
220 | return ret; |
221 | } | |
222 | ||
1d151c33 CG |
223 | #ifdef CONFIG_CHECKPOINT_RESTORE |
224 | static int f_getowner_uids(struct file *filp, unsigned long arg) | |
225 | { | |
226 | struct user_namespace *user_ns = current_user_ns(); | |
63784dd0 | 227 | uid_t __user *dst = (void __user *)arg; |
1d151c33 CG |
228 | uid_t src[2]; |
229 | int err; | |
230 | ||
231 | read_lock(&filp->f_owner.lock); | |
232 | src[0] = from_kuid(user_ns, filp->f_owner.uid); | |
233 | src[1] = from_kuid(user_ns, filp->f_owner.euid); | |
234 | read_unlock(&filp->f_owner.lock); | |
235 | ||
236 | err = put_user(src[0], &dst[0]); | |
237 | err |= put_user(src[1], &dst[1]); | |
238 | ||
239 | return err; | |
240 | } | |
241 | #else | |
242 | static int f_getowner_uids(struct file *filp, unsigned long arg) | |
243 | { | |
244 | return -EINVAL; | |
245 | } | |
246 | #endif | |
247 | ||
1da177e4 LT |
248 | static long do_fcntl(int fd, unsigned int cmd, unsigned long arg, |
249 | struct file *filp) | |
250 | { | |
251 | long err = -EINVAL; | |
252 | ||
253 | switch (cmd) { | |
254 | case F_DUPFD: | |
fe17f22d AV |
255 | err = f_dupfd(arg, filp, 0); |
256 | break; | |
22d2b35b | 257 | case F_DUPFD_CLOEXEC: |
12197718 | 258 | err = f_dupfd(arg, filp, O_CLOEXEC); |
1da177e4 LT |
259 | break; |
260 | case F_GETFD: | |
261 | err = get_close_on_exec(fd) ? FD_CLOEXEC : 0; | |
262 | break; | |
263 | case F_SETFD: | |
264 | err = 0; | |
265 | set_close_on_exec(fd, arg & FD_CLOEXEC); | |
266 | break; | |
267 | case F_GETFL: | |
268 | err = filp->f_flags; | |
269 | break; | |
270 | case F_SETFL: | |
271 | err = setfl(fd, filp, arg); | |
272 | break; | |
5d50ffd7 JL |
273 | #if BITS_PER_LONG != 32 |
274 | /* 32-bit arches must use fcntl64() */ | |
0d3f7a2d | 275 | case F_OFD_GETLK: |
5d50ffd7 | 276 | #endif |
1da177e4 | 277 | case F_GETLK: |
c1e62b8f | 278 | err = fcntl_getlk(filp, cmd, (struct flock __user *) arg); |
1da177e4 | 279 | break; |
5d50ffd7 JL |
280 | #if BITS_PER_LONG != 32 |
281 | /* 32-bit arches must use fcntl64() */ | |
0d3f7a2d JL |
282 | case F_OFD_SETLK: |
283 | case F_OFD_SETLKW: | |
5d50ffd7 JL |
284 | #endif |
285 | /* Fallthrough */ | |
1da177e4 LT |
286 | case F_SETLK: |
287 | case F_SETLKW: | |
c293621b | 288 | err = fcntl_setlk(fd, filp, cmd, (struct flock __user *) arg); |
1da177e4 LT |
289 | break; |
290 | case F_GETOWN: | |
291 | /* | |
292 | * XXX If f_owner is a process group, the | |
293 | * negative return value will get converted | |
294 | * into an error. Oops. If we keep the | |
295 | * current syscall conventions, the only way | |
296 | * to fix this will be in libc. | |
297 | */ | |
609d7fa9 | 298 | err = f_getown(filp); |
1da177e4 LT |
299 | force_successful_syscall_return(); |
300 | break; | |
301 | case F_SETOWN: | |
e0b93edd JL |
302 | f_setown(filp, arg, 1); |
303 | err = 0; | |
1da177e4 | 304 | break; |
ba0a6c9f PZ |
305 | case F_GETOWN_EX: |
306 | err = f_getown_ex(filp, arg); | |
307 | break; | |
308 | case F_SETOWN_EX: | |
309 | err = f_setown_ex(filp, arg); | |
310 | break; | |
1d151c33 CG |
311 | case F_GETOWNER_UIDS: |
312 | err = f_getowner_uids(filp, arg); | |
313 | break; | |
1da177e4 LT |
314 | case F_GETSIG: |
315 | err = filp->f_owner.signum; | |
316 | break; | |
317 | case F_SETSIG: | |
318 | /* arg == 0 restores default behaviour. */ | |
7ed20e1a | 319 | if (!valid_signal(arg)) { |
1da177e4 LT |
320 | break; |
321 | } | |
322 | err = 0; | |
323 | filp->f_owner.signum = arg; | |
324 | break; | |
325 | case F_GETLEASE: | |
326 | err = fcntl_getlease(filp); | |
327 | break; | |
328 | case F_SETLEASE: | |
329 | err = fcntl_setlease(fd, filp, arg); | |
330 | break; | |
331 | case F_NOTIFY: | |
332 | err = fcntl_dirnotify(fd, filp, arg); | |
333 | break; | |
35f3d14d JA |
334 | case F_SETPIPE_SZ: |
335 | case F_GETPIPE_SZ: | |
336 | err = pipe_fcntl(filp, cmd, arg); | |
337 | break; | |
40e041a2 DH |
338 | case F_ADD_SEALS: |
339 | case F_GET_SEALS: | |
340 | err = shmem_fcntl(filp, cmd, arg); | |
341 | break; | |
1da177e4 LT |
342 | default: |
343 | break; | |
344 | } | |
345 | return err; | |
346 | } | |
347 | ||
1abf0c71 AV |
348 | static int check_fcntl_cmd(unsigned cmd) |
349 | { | |
350 | switch (cmd) { | |
351 | case F_DUPFD: | |
352 | case F_DUPFD_CLOEXEC: | |
353 | case F_GETFD: | |
354 | case F_SETFD: | |
355 | case F_GETFL: | |
356 | return 1; | |
357 | } | |
358 | return 0; | |
359 | } | |
360 | ||
a26eab24 | 361 | SYSCALL_DEFINE3(fcntl, unsigned int, fd, unsigned int, cmd, unsigned long, arg) |
1da177e4 | 362 | { |
2903ff01 | 363 | struct fd f = fdget_raw(fd); |
1da177e4 LT |
364 | long err = -EBADF; |
365 | ||
2903ff01 | 366 | if (!f.file) |
1da177e4 LT |
367 | goto out; |
368 | ||
2903ff01 | 369 | if (unlikely(f.file->f_mode & FMODE_PATH)) { |
545ec2c7 AV |
370 | if (!check_fcntl_cmd(cmd)) |
371 | goto out1; | |
1abf0c71 AV |
372 | } |
373 | ||
2903ff01 | 374 | err = security_file_fcntl(f.file, cmd, arg); |
545ec2c7 | 375 | if (!err) |
2903ff01 | 376 | err = do_fcntl(fd, cmd, arg, f.file); |
1da177e4 | 377 | |
545ec2c7 | 378 | out1: |
2903ff01 | 379 | fdput(f); |
1da177e4 LT |
380 | out: |
381 | return err; | |
382 | } | |
383 | ||
384 | #if BITS_PER_LONG == 32 | |
a26eab24 HC |
385 | SYSCALL_DEFINE3(fcntl64, unsigned int, fd, unsigned int, cmd, |
386 | unsigned long, arg) | |
1da177e4 | 387 | { |
2903ff01 | 388 | struct fd f = fdget_raw(fd); |
545ec2c7 | 389 | long err = -EBADF; |
1da177e4 | 390 | |
2903ff01 | 391 | if (!f.file) |
1da177e4 LT |
392 | goto out; |
393 | ||
2903ff01 | 394 | if (unlikely(f.file->f_mode & FMODE_PATH)) { |
545ec2c7 AV |
395 | if (!check_fcntl_cmd(cmd)) |
396 | goto out1; | |
1abf0c71 AV |
397 | } |
398 | ||
2903ff01 | 399 | err = security_file_fcntl(f.file, cmd, arg); |
545ec2c7 AV |
400 | if (err) |
401 | goto out1; | |
1da177e4 LT |
402 | |
403 | switch (cmd) { | |
5d50ffd7 | 404 | case F_GETLK64: |
0d3f7a2d | 405 | case F_OFD_GETLK: |
5d50ffd7 JL |
406 | err = fcntl_getlk64(f.file, cmd, (struct flock64 __user *) arg); |
407 | break; | |
408 | case F_SETLK64: | |
409 | case F_SETLKW64: | |
0d3f7a2d JL |
410 | case F_OFD_SETLK: |
411 | case F_OFD_SETLKW: | |
5d50ffd7 JL |
412 | err = fcntl_setlk64(fd, f.file, cmd, |
413 | (struct flock64 __user *) arg); | |
414 | break; | |
415 | default: | |
416 | err = do_fcntl(fd, cmd, arg, f.file); | |
417 | break; | |
1da177e4 | 418 | } |
545ec2c7 | 419 | out1: |
2903ff01 | 420 | fdput(f); |
1da177e4 LT |
421 | out: |
422 | return err; | |
423 | } | |
424 | #endif | |
425 | ||
426 | /* Table to convert sigio signal codes into poll band bitmaps */ | |
427 | ||
fa3536cc | 428 | static const long band_table[NSIGPOLL] = { |
1da177e4 LT |
429 | POLLIN | POLLRDNORM, /* POLL_IN */ |
430 | POLLOUT | POLLWRNORM | POLLWRBAND, /* POLL_OUT */ | |
431 | POLLIN | POLLRDNORM | POLLMSG, /* POLL_MSG */ | |
432 | POLLERR, /* POLL_ERR */ | |
433 | POLLPRI | POLLRDBAND, /* POLL_PRI */ | |
434 | POLLHUP | POLLERR /* POLL_HUP */ | |
435 | }; | |
436 | ||
437 | static inline int sigio_perm(struct task_struct *p, | |
438 | struct fown_struct *fown, int sig) | |
439 | { | |
c69e8d9c DH |
440 | const struct cred *cred; |
441 | int ret; | |
442 | ||
443 | rcu_read_lock(); | |
444 | cred = __task_cred(p); | |
8e96e3b7 EB |
445 | ret = ((uid_eq(fown->euid, GLOBAL_ROOT_UID) || |
446 | uid_eq(fown->euid, cred->suid) || uid_eq(fown->euid, cred->uid) || | |
447 | uid_eq(fown->uid, cred->suid) || uid_eq(fown->uid, cred->uid)) && | |
c69e8d9c DH |
448 | !security_file_send_sigiotask(p, fown, sig)); |
449 | rcu_read_unlock(); | |
450 | return ret; | |
1da177e4 LT |
451 | } |
452 | ||
453 | static void send_sigio_to_task(struct task_struct *p, | |
8eeee4e2 | 454 | struct fown_struct *fown, |
ba0a6c9f | 455 | int fd, int reason, int group) |
1da177e4 | 456 | { |
8eeee4e2 ON |
457 | /* |
458 | * F_SETSIG can change ->signum lockless in parallel, make | |
459 | * sure we read it once and use the same value throughout. | |
460 | */ | |
461 | int signum = ACCESS_ONCE(fown->signum); | |
462 | ||
463 | if (!sigio_perm(p, fown, signum)) | |
1da177e4 LT |
464 | return; |
465 | ||
8eeee4e2 | 466 | switch (signum) { |
1da177e4 LT |
467 | siginfo_t si; |
468 | default: | |
469 | /* Queue a rt signal with the appropriate fd as its | |
470 | value. We use SI_SIGIO as the source, not | |
471 | SI_KERNEL, since kernel signals always get | |
472 | delivered even if we can't queue. Failure to | |
473 | queue in this case _should_ be reported; we fall | |
474 | back to SIGIO in that case. --sct */ | |
8eeee4e2 | 475 | si.si_signo = signum; |
1da177e4 LT |
476 | si.si_errno = 0; |
477 | si.si_code = reason; | |
478 | /* Make sure we are called with one of the POLL_* | |
479 | reasons, otherwise we could leak kernel stack into | |
480 | userspace. */ | |
f6298aab | 481 | BUG_ON((reason & __SI_MASK) != __SI_POLL); |
1da177e4 LT |
482 | if (reason - POLL_IN >= NSIGPOLL) |
483 | si.si_band = ~0L; | |
484 | else | |
485 | si.si_band = band_table[reason - POLL_IN]; | |
486 | si.si_fd = fd; | |
ba0a6c9f | 487 | if (!do_send_sig_info(signum, &si, p, group)) |
1da177e4 LT |
488 | break; |
489 | /* fall-through: fall back on the old plain SIGIO signal */ | |
490 | case 0: | |
ba0a6c9f | 491 | do_send_sig_info(SIGIO, SEND_SIG_PRIV, p, group); |
1da177e4 LT |
492 | } |
493 | } | |
494 | ||
495 | void send_sigio(struct fown_struct *fown, int fd, int band) | |
496 | { | |
497 | struct task_struct *p; | |
609d7fa9 EB |
498 | enum pid_type type; |
499 | struct pid *pid; | |
ba0a6c9f | 500 | int group = 1; |
1da177e4 LT |
501 | |
502 | read_lock(&fown->lock); | |
ba0a6c9f | 503 | |
609d7fa9 | 504 | type = fown->pid_type; |
ba0a6c9f PZ |
505 | if (type == PIDTYPE_MAX) { |
506 | group = 0; | |
507 | type = PIDTYPE_PID; | |
508 | } | |
509 | ||
1da177e4 LT |
510 | pid = fown->pid; |
511 | if (!pid) | |
512 | goto out_unlock_fown; | |
513 | ||
514 | read_lock(&tasklist_lock); | |
609d7fa9 | 515 | do_each_pid_task(pid, type, p) { |
ba0a6c9f | 516 | send_sigio_to_task(p, fown, fd, band, group); |
609d7fa9 | 517 | } while_each_pid_task(pid, type, p); |
1da177e4 LT |
518 | read_unlock(&tasklist_lock); |
519 | out_unlock_fown: | |
520 | read_unlock(&fown->lock); | |
521 | } | |
522 | ||
523 | static void send_sigurg_to_task(struct task_struct *p, | |
ba0a6c9f | 524 | struct fown_struct *fown, int group) |
1da177e4 LT |
525 | { |
526 | if (sigio_perm(p, fown, SIGURG)) | |
ba0a6c9f | 527 | do_send_sig_info(SIGURG, SEND_SIG_PRIV, p, group); |
1da177e4 LT |
528 | } |
529 | ||
530 | int send_sigurg(struct fown_struct *fown) | |
531 | { | |
532 | struct task_struct *p; | |
609d7fa9 EB |
533 | enum pid_type type; |
534 | struct pid *pid; | |
ba0a6c9f | 535 | int group = 1; |
609d7fa9 | 536 | int ret = 0; |
1da177e4 LT |
537 | |
538 | read_lock(&fown->lock); | |
ba0a6c9f | 539 | |
609d7fa9 | 540 | type = fown->pid_type; |
ba0a6c9f PZ |
541 | if (type == PIDTYPE_MAX) { |
542 | group = 0; | |
543 | type = PIDTYPE_PID; | |
544 | } | |
545 | ||
1da177e4 LT |
546 | pid = fown->pid; |
547 | if (!pid) | |
548 | goto out_unlock_fown; | |
549 | ||
550 | ret = 1; | |
551 | ||
552 | read_lock(&tasklist_lock); | |
609d7fa9 | 553 | do_each_pid_task(pid, type, p) { |
ba0a6c9f | 554 | send_sigurg_to_task(p, fown, group); |
609d7fa9 | 555 | } while_each_pid_task(pid, type, p); |
1da177e4 LT |
556 | read_unlock(&tasklist_lock); |
557 | out_unlock_fown: | |
558 | read_unlock(&fown->lock); | |
559 | return ret; | |
560 | } | |
561 | ||
989a2979 | 562 | static DEFINE_SPINLOCK(fasync_lock); |
e18b890b | 563 | static struct kmem_cache *fasync_cache __read_mostly; |
1da177e4 | 564 | |
989a2979 ED |
565 | static void fasync_free_rcu(struct rcu_head *head) |
566 | { | |
567 | kmem_cache_free(fasync_cache, | |
568 | container_of(head, struct fasync_struct, fa_rcu)); | |
569 | } | |
570 | ||
1da177e4 | 571 | /* |
53281b6d LT |
572 | * Remove a fasync entry. If successfully removed, return |
573 | * positive and clear the FASYNC flag. If no entry exists, | |
574 | * do nothing and return 0. | |
575 | * | |
576 | * NOTE! It is very important that the FASYNC flag always | |
577 | * match the state "is the filp on a fasync list". | |
578 | * | |
1da177e4 | 579 | */ |
f7347ce4 | 580 | int fasync_remove_entry(struct file *filp, struct fasync_struct **fapp) |
1da177e4 LT |
581 | { |
582 | struct fasync_struct *fa, **fp; | |
1da177e4 LT |
583 | int result = 0; |
584 | ||
53281b6d | 585 | spin_lock(&filp->f_lock); |
989a2979 | 586 | spin_lock(&fasync_lock); |
53281b6d LT |
587 | for (fp = fapp; (fa = *fp) != NULL; fp = &fa->fa_next) { |
588 | if (fa->fa_file != filp) | |
589 | continue; | |
989a2979 ED |
590 | |
591 | spin_lock_irq(&fa->fa_lock); | |
592 | fa->fa_file = NULL; | |
593 | spin_unlock_irq(&fa->fa_lock); | |
594 | ||
53281b6d | 595 | *fp = fa->fa_next; |
989a2979 | 596 | call_rcu(&fa->fa_rcu, fasync_free_rcu); |
53281b6d LT |
597 | filp->f_flags &= ~FASYNC; |
598 | result = 1; | |
599 | break; | |
1da177e4 | 600 | } |
989a2979 | 601 | spin_unlock(&fasync_lock); |
53281b6d LT |
602 | spin_unlock(&filp->f_lock); |
603 | return result; | |
604 | } | |
605 | ||
f7347ce4 LT |
606 | struct fasync_struct *fasync_alloc(void) |
607 | { | |
608 | return kmem_cache_alloc(fasync_cache, GFP_KERNEL); | |
609 | } | |
610 | ||
53281b6d | 611 | /* |
f7347ce4 LT |
612 | * NOTE! This can be used only for unused fasync entries: |
613 | * entries that actually got inserted on the fasync list | |
614 | * need to be released by rcu - see fasync_remove_entry. | |
53281b6d | 615 | */ |
f7347ce4 | 616 | void fasync_free(struct fasync_struct *new) |
53281b6d | 617 | { |
f7347ce4 LT |
618 | kmem_cache_free(fasync_cache, new); |
619 | } | |
53281b6d | 620 | |
f7347ce4 LT |
621 | /* |
622 | * Insert a new entry into the fasync list. Return the pointer to the | |
623 | * old one if we didn't use the new one. | |
55f335a8 LT |
624 | * |
625 | * NOTE! It is very important that the FASYNC flag always | |
626 | * match the state "is the filp on a fasync list". | |
f7347ce4 LT |
627 | */ |
628 | struct fasync_struct *fasync_insert_entry(int fd, struct file *filp, struct fasync_struct **fapp, struct fasync_struct *new) | |
629 | { | |
630 | struct fasync_struct *fa, **fp; | |
4a6a4499 | 631 | |
4a6a4499 | 632 | spin_lock(&filp->f_lock); |
989a2979 | 633 | spin_lock(&fasync_lock); |
1da177e4 | 634 | for (fp = fapp; (fa = *fp) != NULL; fp = &fa->fa_next) { |
53281b6d LT |
635 | if (fa->fa_file != filp) |
636 | continue; | |
989a2979 ED |
637 | |
638 | spin_lock_irq(&fa->fa_lock); | |
53281b6d | 639 | fa->fa_fd = fd; |
989a2979 | 640 | spin_unlock_irq(&fa->fa_lock); |
53281b6d | 641 | goto out; |
1da177e4 LT |
642 | } |
643 | ||
989a2979 | 644 | spin_lock_init(&new->fa_lock); |
53281b6d LT |
645 | new->magic = FASYNC_MAGIC; |
646 | new->fa_file = filp; | |
647 | new->fa_fd = fd; | |
648 | new->fa_next = *fapp; | |
989a2979 | 649 | rcu_assign_pointer(*fapp, new); |
53281b6d LT |
650 | filp->f_flags |= FASYNC; |
651 | ||
1da177e4 | 652 | out: |
989a2979 | 653 | spin_unlock(&fasync_lock); |
4a6a4499 | 654 | spin_unlock(&filp->f_lock); |
f7347ce4 LT |
655 | return fa; |
656 | } | |
657 | ||
658 | /* | |
659 | * Add a fasync entry. Return negative on error, positive if | |
660 | * added, and zero if did nothing but change an existing one. | |
f7347ce4 LT |
661 | */ |
662 | static int fasync_add_entry(int fd, struct file *filp, struct fasync_struct **fapp) | |
663 | { | |
664 | struct fasync_struct *new; | |
665 | ||
666 | new = fasync_alloc(); | |
667 | if (!new) | |
668 | return -ENOMEM; | |
669 | ||
670 | /* | |
671 | * fasync_insert_entry() returns the old (update) entry if | |
672 | * it existed. | |
673 | * | |
674 | * So free the (unused) new entry and return 0 to let the | |
675 | * caller know that we didn't add any new fasync entries. | |
676 | */ | |
677 | if (fasync_insert_entry(fd, filp, fapp, new)) { | |
678 | fasync_free(new); | |
679 | return 0; | |
680 | } | |
681 | ||
682 | return 1; | |
1da177e4 LT |
683 | } |
684 | ||
53281b6d LT |
685 | /* |
686 | * fasync_helper() is used by almost all character device drivers | |
687 | * to set up the fasync queue, and for regular files by the file | |
688 | * lease code. It returns negative on error, 0 if it did no changes | |
689 | * and positive if it added/deleted the entry. | |
690 | */ | |
691 | int fasync_helper(int fd, struct file * filp, int on, struct fasync_struct **fapp) | |
692 | { | |
693 | if (!on) | |
694 | return fasync_remove_entry(filp, fapp); | |
695 | return fasync_add_entry(fd, filp, fapp); | |
696 | } | |
697 | ||
1da177e4 LT |
698 | EXPORT_SYMBOL(fasync_helper); |
699 | ||
989a2979 ED |
700 | /* |
701 | * rcu_read_lock() is held | |
702 | */ | |
703 | static void kill_fasync_rcu(struct fasync_struct *fa, int sig, int band) | |
1da177e4 LT |
704 | { |
705 | while (fa) { | |
989a2979 | 706 | struct fown_struct *fown; |
f4985dc7 AM |
707 | unsigned long flags; |
708 | ||
1da177e4 LT |
709 | if (fa->magic != FASYNC_MAGIC) { |
710 | printk(KERN_ERR "kill_fasync: bad magic number in " | |
711 | "fasync_struct!\n"); | |
712 | return; | |
713 | } | |
f4985dc7 | 714 | spin_lock_irqsave(&fa->fa_lock, flags); |
989a2979 ED |
715 | if (fa->fa_file) { |
716 | fown = &fa->fa_file->f_owner; | |
717 | /* Don't send SIGURG to processes which have not set a | |
718 | queued signum: SIGURG has its own default signalling | |
719 | mechanism. */ | |
720 | if (!(sig == SIGURG && fown->signum == 0)) | |
721 | send_sigio(fown, fa->fa_fd, band); | |
722 | } | |
f4985dc7 | 723 | spin_unlock_irqrestore(&fa->fa_lock, flags); |
989a2979 | 724 | fa = rcu_dereference(fa->fa_next); |
1da177e4 LT |
725 | } |
726 | } | |
727 | ||
1da177e4 LT |
728 | void kill_fasync(struct fasync_struct **fp, int sig, int band) |
729 | { | |
730 | /* First a quick test without locking: usually | |
731 | * the list is empty. | |
732 | */ | |
733 | if (*fp) { | |
989a2979 ED |
734 | rcu_read_lock(); |
735 | kill_fasync_rcu(rcu_dereference(*fp), sig, band); | |
736 | rcu_read_unlock(); | |
1da177e4 LT |
737 | } |
738 | } | |
739 | EXPORT_SYMBOL(kill_fasync); | |
740 | ||
454eedb8 | 741 | static int __init fcntl_init(void) |
1da177e4 | 742 | { |
3ab04d5c JB |
743 | /* |
744 | * Please add new bits here to ensure allocation uniqueness. | |
745 | * Exceptions: O_NONBLOCK is a two bit define on parisc; O_NDELAY | |
746 | * is defined as O_NONBLOCK on some platforms and not on others. | |
747 | */ | |
6efb1b0b CH |
748 | BUILD_BUG_ON(21 - 1 /* for O_RDONLY being 0 */ != |
749 | HWEIGHT32( | |
750 | (VALID_OPEN_FLAGS & ~(O_NONBLOCK | O_NDELAY)) | | |
751 | __FMODE_EXEC | __FMODE_NONOTIFY)); | |
454eedb8 | 752 | |
1da177e4 | 753 | fasync_cache = kmem_cache_create("fasync_cache", |
20c2df83 | 754 | sizeof(struct fasync_struct), 0, SLAB_PANIC, NULL); |
1da177e4 LT |
755 | return 0; |
756 | } | |
757 | ||
454eedb8 | 758 | module_init(fcntl_init) |