]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * POSIX message queues filesystem for Linux. | |
3 | * | |
4 | * Copyright (C) 2003,2004 Krzysztof Benedyczak (golbi@mat.uni.torun.pl) | |
f66e928b | 5 | * Michal Wronski (michal.wronski@gmail.com) |
1da177e4 LT |
6 | * |
7 | * Spinlocks: Mohamed Abbas (abbas.mohamed@intel.com) | |
8 | * Lockless receive & send, fd based notify: | |
9 | * Manfred Spraul (manfred@colorfullife.com) | |
10 | * | |
20ca73bc GW |
11 | * Audit: George Wilson (ltcgcw@us.ibm.com) |
12 | * | |
1da177e4 LT |
13 | * This file is released under the GPL. |
14 | */ | |
15 | ||
c59ede7b | 16 | #include <linux/capability.h> |
1da177e4 LT |
17 | #include <linux/init.h> |
18 | #include <linux/pagemap.h> | |
19 | #include <linux/file.h> | |
20 | #include <linux/mount.h> | |
21 | #include <linux/namei.h> | |
22 | #include <linux/sysctl.h> | |
23 | #include <linux/poll.h> | |
24 | #include <linux/mqueue.h> | |
25 | #include <linux/msg.h> | |
26 | #include <linux/skbuff.h> | |
27 | #include <linux/netlink.h> | |
28 | #include <linux/syscalls.h> | |
20ca73bc | 29 | #include <linux/audit.h> |
7ed20e1a | 30 | #include <linux/signal.h> |
5f921ae9 | 31 | #include <linux/mutex.h> |
b488893a PE |
32 | #include <linux/nsproxy.h> |
33 | #include <linux/pid.h> | |
614b84cf | 34 | #include <linux/ipc_namespace.h> |
6b550f94 | 35 | #include <linux/user_namespace.h> |
5a0e3ad6 | 36 | #include <linux/slab.h> |
5f921ae9 | 37 | |
1da177e4 LT |
38 | #include <net/sock.h> |
39 | #include "util.h" | |
40 | ||
41 | #define MQUEUE_MAGIC 0x19800202 | |
42 | #define DIRENT_SIZE 20 | |
43 | #define FILENT_SIZE 80 | |
44 | ||
45 | #define SEND 0 | |
46 | #define RECV 1 | |
47 | ||
48 | #define STATE_NONE 0 | |
49 | #define STATE_PENDING 1 | |
50 | #define STATE_READY 2 | |
51 | ||
1da177e4 LT |
52 | struct ext_wait_queue { /* queue of sleeping tasks */ |
53 | struct task_struct *task; | |
54 | struct list_head list; | |
55 | struct msg_msg *msg; /* ptr of loaded message */ | |
56 | int state; /* one of STATE_* values */ | |
57 | }; | |
58 | ||
59 | struct mqueue_inode_info { | |
60 | spinlock_t lock; | |
61 | struct inode vfs_inode; | |
62 | wait_queue_head_t wait_q; | |
63 | ||
64 | struct msg_msg **messages; | |
65 | struct mq_attr attr; | |
66 | ||
67 | struct sigevent notify; | |
a03fcb73 | 68 | struct pid* notify_owner; |
338cec32 | 69 | struct user_struct *user; /* user who created, for accounting */ |
1da177e4 LT |
70 | struct sock *notify_sock; |
71 | struct sk_buff *notify_cookie; | |
72 | ||
73 | /* for tasks waiting for free space and messages, respectively */ | |
74 | struct ext_wait_queue e_wait_q[2]; | |
75 | ||
76 | unsigned long qsize; /* size of queue in memory (sum of all msgs) */ | |
77 | }; | |
78 | ||
92e1d5be | 79 | static const struct inode_operations mqueue_dir_inode_operations; |
9a32144e | 80 | static const struct file_operations mqueue_file_operations; |
b87221de | 81 | static const struct super_operations mqueue_super_ops; |
1da177e4 LT |
82 | static void remove_notification(struct mqueue_inode_info *info); |
83 | ||
e18b890b | 84 | static struct kmem_cache *mqueue_inode_cachep; |
1da177e4 LT |
85 | |
86 | static struct ctl_table_header * mq_sysctl_table; | |
87 | ||
88 | static inline struct mqueue_inode_info *MQUEUE_I(struct inode *inode) | |
89 | { | |
90 | return container_of(inode, struct mqueue_inode_info, vfs_inode); | |
91 | } | |
92 | ||
7eafd7c7 SH |
93 | /* |
94 | * This routine should be called with the mq_lock held. | |
95 | */ | |
96 | static inline struct ipc_namespace *__get_ns_from_inode(struct inode *inode) | |
614b84cf | 97 | { |
7eafd7c7 | 98 | return get_ipc_ns(inode->i_sb->s_fs_info); |
614b84cf SH |
99 | } |
100 | ||
7eafd7c7 | 101 | static struct ipc_namespace *get_ns_from_inode(struct inode *inode) |
614b84cf | 102 | { |
7eafd7c7 SH |
103 | struct ipc_namespace *ns; |
104 | ||
105 | spin_lock(&mq_lock); | |
106 | ns = __get_ns_from_inode(inode); | |
107 | spin_unlock(&mq_lock); | |
108 | return ns; | |
614b84cf SH |
109 | } |
110 | ||
7eafd7c7 | 111 | static struct inode *mqueue_get_inode(struct super_block *sb, |
1b9d5ff7 | 112 | struct ipc_namespace *ipc_ns, umode_t mode, |
7eafd7c7 | 113 | struct mq_attr *attr) |
1da177e4 | 114 | { |
86a264ab | 115 | struct user_struct *u = current_user(); |
1da177e4 | 116 | struct inode *inode; |
d40dcdb0 | 117 | int ret = -ENOMEM; |
1da177e4 LT |
118 | |
119 | inode = new_inode(sb); | |
04715206 JS |
120 | if (!inode) |
121 | goto err; | |
122 | ||
123 | inode->i_ino = get_next_ino(); | |
124 | inode->i_mode = mode; | |
125 | inode->i_uid = current_fsuid(); | |
126 | inode->i_gid = current_fsgid(); | |
127 | inode->i_mtime = inode->i_ctime = inode->i_atime = CURRENT_TIME; | |
128 | ||
129 | if (S_ISREG(mode)) { | |
130 | struct mqueue_inode_info *info; | |
04715206 JS |
131 | unsigned long mq_bytes, mq_msg_tblsz; |
132 | ||
133 | inode->i_fop = &mqueue_file_operations; | |
134 | inode->i_size = FILENT_SIZE; | |
135 | /* mqueue specific info */ | |
136 | info = MQUEUE_I(inode); | |
137 | spin_lock_init(&info->lock); | |
138 | init_waitqueue_head(&info->wait_q); | |
139 | INIT_LIST_HEAD(&info->e_wait_q[0].list); | |
140 | INIT_LIST_HEAD(&info->e_wait_q[1].list); | |
141 | info->notify_owner = NULL; | |
142 | info->qsize = 0; | |
143 | info->user = NULL; /* set when all is ok */ | |
144 | memset(&info->attr, 0, sizeof(info->attr)); | |
145 | info->attr.mq_maxmsg = ipc_ns->mq_msg_max; | |
146 | info->attr.mq_msgsize = ipc_ns->mq_msgsize_max; | |
147 | if (attr) { | |
148 | info->attr.mq_maxmsg = attr->mq_maxmsg; | |
149 | info->attr.mq_msgsize = attr->mq_msgsize; | |
150 | } | |
151 | mq_msg_tblsz = info->attr.mq_maxmsg * sizeof(struct msg_msg *); | |
152 | info->messages = kmalloc(mq_msg_tblsz, GFP_KERNEL); | |
153 | if (!info->messages) | |
154 | goto out_inode; | |
155 | ||
156 | mq_bytes = (mq_msg_tblsz + | |
157 | (info->attr.mq_maxmsg * info->attr.mq_msgsize)); | |
1da177e4 | 158 | |
04715206 JS |
159 | spin_lock(&mq_lock); |
160 | if (u->mq_bytes + mq_bytes < u->mq_bytes || | |
2a4e64b8 | 161 | u->mq_bytes + mq_bytes > rlimit(RLIMIT_MSGQUEUE)) { |
04715206 JS |
162 | spin_unlock(&mq_lock); |
163 | /* mqueue_evict_inode() releases info->messages */ | |
d40dcdb0 | 164 | ret = -EMFILE; |
04715206 | 165 | goto out_inode; |
1da177e4 | 166 | } |
04715206 JS |
167 | u->mq_bytes += mq_bytes; |
168 | spin_unlock(&mq_lock); | |
169 | ||
170 | /* all is ok */ | |
171 | info->user = get_uid(u); | |
172 | } else if (S_ISDIR(mode)) { | |
173 | inc_nlink(inode); | |
174 | /* Some things misbehave if size == 0 on a directory */ | |
175 | inode->i_size = 2 * DIRENT_SIZE; | |
176 | inode->i_op = &mqueue_dir_inode_operations; | |
177 | inode->i_fop = &simple_dir_operations; | |
1da177e4 | 178 | } |
04715206 | 179 | |
1da177e4 LT |
180 | return inode; |
181 | out_inode: | |
1da177e4 | 182 | iput(inode); |
04715206 | 183 | err: |
d40dcdb0 | 184 | return ERR_PTR(ret); |
1da177e4 LT |
185 | } |
186 | ||
187 | static int mqueue_fill_super(struct super_block *sb, void *data, int silent) | |
188 | { | |
189 | struct inode *inode; | |
7eafd7c7 | 190 | struct ipc_namespace *ns = data; |
8d8ffefa | 191 | int error; |
1da177e4 LT |
192 | |
193 | sb->s_blocksize = PAGE_CACHE_SIZE; | |
194 | sb->s_blocksize_bits = PAGE_CACHE_SHIFT; | |
195 | sb->s_magic = MQUEUE_MAGIC; | |
196 | sb->s_op = &mqueue_super_ops; | |
197 | ||
7eafd7c7 SH |
198 | inode = mqueue_get_inode(sb, ns, S_IFDIR | S_ISVTX | S_IRWXUGO, |
199 | NULL); | |
d40dcdb0 JS |
200 | if (IS_ERR(inode)) { |
201 | error = PTR_ERR(inode); | |
7eafd7c7 SH |
202 | goto out; |
203 | } | |
1da177e4 LT |
204 | |
205 | sb->s_root = d_alloc_root(inode); | |
206 | if (!sb->s_root) { | |
207 | iput(inode); | |
7eafd7c7 | 208 | error = -ENOMEM; |
8d8ffefa | 209 | goto out; |
1da177e4 | 210 | } |
8d8ffefa | 211 | error = 0; |
1da177e4 | 212 | |
7eafd7c7 SH |
213 | out: |
214 | return error; | |
1da177e4 LT |
215 | } |
216 | ||
ceefda69 | 217 | static struct dentry *mqueue_mount(struct file_system_type *fs_type, |
454e2398 | 218 | int flags, const char *dev_name, |
ceefda69 | 219 | void *data) |
1da177e4 | 220 | { |
7eafd7c7 SH |
221 | if (!(flags & MS_KERNMOUNT)) |
222 | data = current->nsproxy->ipc_ns; | |
ceefda69 | 223 | return mount_ns(fs_type, flags, data, mqueue_fill_super); |
1da177e4 LT |
224 | } |
225 | ||
51cc5068 | 226 | static void init_once(void *foo) |
1da177e4 LT |
227 | { |
228 | struct mqueue_inode_info *p = (struct mqueue_inode_info *) foo; | |
229 | ||
a35afb83 | 230 | inode_init_once(&p->vfs_inode); |
1da177e4 LT |
231 | } |
232 | ||
233 | static struct inode *mqueue_alloc_inode(struct super_block *sb) | |
234 | { | |
235 | struct mqueue_inode_info *ei; | |
236 | ||
e94b1766 | 237 | ei = kmem_cache_alloc(mqueue_inode_cachep, GFP_KERNEL); |
1da177e4 LT |
238 | if (!ei) |
239 | return NULL; | |
240 | return &ei->vfs_inode; | |
241 | } | |
242 | ||
fa0d7e3d | 243 | static void mqueue_i_callback(struct rcu_head *head) |
1da177e4 | 244 | { |
fa0d7e3d | 245 | struct inode *inode = container_of(head, struct inode, i_rcu); |
1da177e4 LT |
246 | kmem_cache_free(mqueue_inode_cachep, MQUEUE_I(inode)); |
247 | } | |
248 | ||
fa0d7e3d NP |
249 | static void mqueue_destroy_inode(struct inode *inode) |
250 | { | |
251 | call_rcu(&inode->i_rcu, mqueue_i_callback); | |
252 | } | |
253 | ||
6d8af64c | 254 | static void mqueue_evict_inode(struct inode *inode) |
1da177e4 LT |
255 | { |
256 | struct mqueue_inode_info *info; | |
257 | struct user_struct *user; | |
258 | unsigned long mq_bytes; | |
259 | int i; | |
7eafd7c7 | 260 | struct ipc_namespace *ipc_ns; |
1da177e4 | 261 | |
6d8af64c AV |
262 | end_writeback(inode); |
263 | ||
264 | if (S_ISDIR(inode->i_mode)) | |
1da177e4 | 265 | return; |
6d8af64c | 266 | |
7eafd7c7 | 267 | ipc_ns = get_ns_from_inode(inode); |
1da177e4 LT |
268 | info = MQUEUE_I(inode); |
269 | spin_lock(&info->lock); | |
270 | for (i = 0; i < info->attr.mq_curmsgs; i++) | |
271 | free_msg(info->messages[i]); | |
272 | kfree(info->messages); | |
273 | spin_unlock(&info->lock); | |
274 | ||
8834cf79 AGR |
275 | /* Total amount of bytes accounted for the mqueue */ |
276 | mq_bytes = info->attr.mq_maxmsg * (sizeof(struct msg_msg *) | |
277 | + info->attr.mq_msgsize); | |
1da177e4 LT |
278 | user = info->user; |
279 | if (user) { | |
280 | spin_lock(&mq_lock); | |
281 | user->mq_bytes -= mq_bytes; | |
7eafd7c7 SH |
282 | /* |
283 | * get_ns_from_inode() ensures that the | |
284 | * (ipc_ns = sb->s_fs_info) is either a valid ipc_ns | |
285 | * to which we now hold a reference, or it is NULL. | |
286 | * We can't put it here under mq_lock, though. | |
287 | */ | |
288 | if (ipc_ns) | |
289 | ipc_ns->mq_queues_count--; | |
1da177e4 LT |
290 | spin_unlock(&mq_lock); |
291 | free_uid(user); | |
292 | } | |
7eafd7c7 SH |
293 | if (ipc_ns) |
294 | put_ipc_ns(ipc_ns); | |
1da177e4 LT |
295 | } |
296 | ||
297 | static int mqueue_create(struct inode *dir, struct dentry *dentry, | |
4acdaf27 | 298 | umode_t mode, struct nameidata *nd) |
1da177e4 LT |
299 | { |
300 | struct inode *inode; | |
301 | struct mq_attr *attr = dentry->d_fsdata; | |
302 | int error; | |
7eafd7c7 | 303 | struct ipc_namespace *ipc_ns; |
1da177e4 LT |
304 | |
305 | spin_lock(&mq_lock); | |
7eafd7c7 SH |
306 | ipc_ns = __get_ns_from_inode(dir); |
307 | if (!ipc_ns) { | |
308 | error = -EACCES; | |
309 | goto out_unlock; | |
310 | } | |
614b84cf SH |
311 | if (ipc_ns->mq_queues_count >= ipc_ns->mq_queues_max && |
312 | !capable(CAP_SYS_RESOURCE)) { | |
1da177e4 | 313 | error = -ENOSPC; |
614b84cf | 314 | goto out_unlock; |
1da177e4 | 315 | } |
614b84cf | 316 | ipc_ns->mq_queues_count++; |
1da177e4 LT |
317 | spin_unlock(&mq_lock); |
318 | ||
7eafd7c7 | 319 | inode = mqueue_get_inode(dir->i_sb, ipc_ns, mode, attr); |
d40dcdb0 JS |
320 | if (IS_ERR(inode)) { |
321 | error = PTR_ERR(inode); | |
1da177e4 | 322 | spin_lock(&mq_lock); |
614b84cf SH |
323 | ipc_ns->mq_queues_count--; |
324 | goto out_unlock; | |
1da177e4 LT |
325 | } |
326 | ||
7eafd7c7 | 327 | put_ipc_ns(ipc_ns); |
1da177e4 LT |
328 | dir->i_size += DIRENT_SIZE; |
329 | dir->i_ctime = dir->i_mtime = dir->i_atime = CURRENT_TIME; | |
330 | ||
331 | d_instantiate(dentry, inode); | |
332 | dget(dentry); | |
333 | return 0; | |
614b84cf | 334 | out_unlock: |
1da177e4 | 335 | spin_unlock(&mq_lock); |
7eafd7c7 SH |
336 | if (ipc_ns) |
337 | put_ipc_ns(ipc_ns); | |
1da177e4 LT |
338 | return error; |
339 | } | |
340 | ||
341 | static int mqueue_unlink(struct inode *dir, struct dentry *dentry) | |
342 | { | |
343 | struct inode *inode = dentry->d_inode; | |
344 | ||
345 | dir->i_ctime = dir->i_mtime = dir->i_atime = CURRENT_TIME; | |
346 | dir->i_size -= DIRENT_SIZE; | |
9a53c3a7 | 347 | drop_nlink(inode); |
1da177e4 LT |
348 | dput(dentry); |
349 | return 0; | |
350 | } | |
351 | ||
352 | /* | |
353 | * This is routine for system read from queue file. | |
354 | * To avoid mess with doing here some sort of mq_receive we allow | |
355 | * to read only queue size & notification info (the only values | |
356 | * that are interesting from user point of view and aren't accessible | |
357 | * through std routines) | |
358 | */ | |
359 | static ssize_t mqueue_read_file(struct file *filp, char __user *u_data, | |
f1a43f93 | 360 | size_t count, loff_t *off) |
1da177e4 | 361 | { |
6d63079a | 362 | struct mqueue_inode_info *info = MQUEUE_I(filp->f_path.dentry->d_inode); |
1da177e4 | 363 | char buffer[FILENT_SIZE]; |
f1a43f93 | 364 | ssize_t ret; |
1da177e4 LT |
365 | |
366 | spin_lock(&info->lock); | |
367 | snprintf(buffer, sizeof(buffer), | |
368 | "QSIZE:%-10lu NOTIFY:%-5d SIGNO:%-5d NOTIFY_PID:%-6d\n", | |
369 | info->qsize, | |
370 | info->notify_owner ? info->notify.sigev_notify : 0, | |
371 | (info->notify_owner && | |
372 | info->notify.sigev_notify == SIGEV_SIGNAL) ? | |
373 | info->notify.sigev_signo : 0, | |
6c5f3e7b | 374 | pid_vnr(info->notify_owner)); |
1da177e4 LT |
375 | spin_unlock(&info->lock); |
376 | buffer[sizeof(buffer)-1] = '\0'; | |
1da177e4 | 377 | |
f1a43f93 AM |
378 | ret = simple_read_from_buffer(u_data, count, off, buffer, |
379 | strlen(buffer)); | |
380 | if (ret <= 0) | |
381 | return ret; | |
1da177e4 | 382 | |
6d63079a | 383 | filp->f_path.dentry->d_inode->i_atime = filp->f_path.dentry->d_inode->i_ctime = CURRENT_TIME; |
f1a43f93 | 384 | return ret; |
1da177e4 LT |
385 | } |
386 | ||
75e1fcc0 | 387 | static int mqueue_flush_file(struct file *filp, fl_owner_t id) |
1da177e4 | 388 | { |
6d63079a | 389 | struct mqueue_inode_info *info = MQUEUE_I(filp->f_path.dentry->d_inode); |
1da177e4 LT |
390 | |
391 | spin_lock(&info->lock); | |
a03fcb73 | 392 | if (task_tgid(current) == info->notify_owner) |
1da177e4 LT |
393 | remove_notification(info); |
394 | ||
395 | spin_unlock(&info->lock); | |
396 | return 0; | |
397 | } | |
398 | ||
399 | static unsigned int mqueue_poll_file(struct file *filp, struct poll_table_struct *poll_tab) | |
400 | { | |
6d63079a | 401 | struct mqueue_inode_info *info = MQUEUE_I(filp->f_path.dentry->d_inode); |
1da177e4 LT |
402 | int retval = 0; |
403 | ||
404 | poll_wait(filp, &info->wait_q, poll_tab); | |
405 | ||
406 | spin_lock(&info->lock); | |
407 | if (info->attr.mq_curmsgs) | |
408 | retval = POLLIN | POLLRDNORM; | |
409 | ||
410 | if (info->attr.mq_curmsgs < info->attr.mq_maxmsg) | |
411 | retval |= POLLOUT | POLLWRNORM; | |
412 | spin_unlock(&info->lock); | |
413 | ||
414 | return retval; | |
415 | } | |
416 | ||
417 | /* Adds current to info->e_wait_q[sr] before element with smaller prio */ | |
418 | static void wq_add(struct mqueue_inode_info *info, int sr, | |
419 | struct ext_wait_queue *ewp) | |
420 | { | |
421 | struct ext_wait_queue *walk; | |
422 | ||
423 | ewp->task = current; | |
424 | ||
425 | list_for_each_entry(walk, &info->e_wait_q[sr].list, list) { | |
426 | if (walk->task->static_prio <= current->static_prio) { | |
427 | list_add_tail(&ewp->list, &walk->list); | |
428 | return; | |
429 | } | |
430 | } | |
431 | list_add_tail(&ewp->list, &info->e_wait_q[sr].list); | |
432 | } | |
433 | ||
434 | /* | |
435 | * Puts current task to sleep. Caller must hold queue lock. After return | |
436 | * lock isn't held. | |
437 | * sr: SEND or RECV | |
438 | */ | |
439 | static int wq_sleep(struct mqueue_inode_info *info, int sr, | |
9ca7d8e6 | 440 | ktime_t *timeout, struct ext_wait_queue *ewp) |
1da177e4 LT |
441 | { |
442 | int retval; | |
443 | signed long time; | |
444 | ||
445 | wq_add(info, sr, ewp); | |
446 | ||
447 | for (;;) { | |
448 | set_current_state(TASK_INTERRUPTIBLE); | |
449 | ||
450 | spin_unlock(&info->lock); | |
32ea845d WG |
451 | time = schedule_hrtimeout_range_clock(timeout, 0, |
452 | HRTIMER_MODE_ABS, CLOCK_REALTIME); | |
1da177e4 LT |
453 | |
454 | while (ewp->state == STATE_PENDING) | |
455 | cpu_relax(); | |
456 | ||
457 | if (ewp->state == STATE_READY) { | |
458 | retval = 0; | |
459 | goto out; | |
460 | } | |
461 | spin_lock(&info->lock); | |
462 | if (ewp->state == STATE_READY) { | |
463 | retval = 0; | |
464 | goto out_unlock; | |
465 | } | |
466 | if (signal_pending(current)) { | |
467 | retval = -ERESTARTSYS; | |
468 | break; | |
469 | } | |
470 | if (time == 0) { | |
471 | retval = -ETIMEDOUT; | |
472 | break; | |
473 | } | |
474 | } | |
475 | list_del(&ewp->list); | |
476 | out_unlock: | |
477 | spin_unlock(&info->lock); | |
478 | out: | |
479 | return retval; | |
480 | } | |
481 | ||
482 | /* | |
483 | * Returns waiting task that should be serviced first or NULL if none exists | |
484 | */ | |
485 | static struct ext_wait_queue *wq_get_first_waiter( | |
486 | struct mqueue_inode_info *info, int sr) | |
487 | { | |
488 | struct list_head *ptr; | |
489 | ||
490 | ptr = info->e_wait_q[sr].list.prev; | |
491 | if (ptr == &info->e_wait_q[sr].list) | |
492 | return NULL; | |
493 | return list_entry(ptr, struct ext_wait_queue, list); | |
494 | } | |
495 | ||
496 | /* Auxiliary functions to manipulate messages' list */ | |
497 | static void msg_insert(struct msg_msg *ptr, struct mqueue_inode_info *info) | |
498 | { | |
499 | int k; | |
500 | ||
501 | k = info->attr.mq_curmsgs - 1; | |
502 | while (k >= 0 && info->messages[k]->m_type >= ptr->m_type) { | |
503 | info->messages[k + 1] = info->messages[k]; | |
504 | k--; | |
505 | } | |
506 | info->attr.mq_curmsgs++; | |
507 | info->qsize += ptr->m_ts; | |
508 | info->messages[k + 1] = ptr; | |
509 | } | |
510 | ||
511 | static inline struct msg_msg *msg_get(struct mqueue_inode_info *info) | |
512 | { | |
513 | info->qsize -= info->messages[--info->attr.mq_curmsgs]->m_ts; | |
514 | return info->messages[info->attr.mq_curmsgs]; | |
515 | } | |
516 | ||
517 | static inline void set_cookie(struct sk_buff *skb, char code) | |
518 | { | |
519 | ((char*)skb->data)[NOTIFY_COOKIE_LEN-1] = code; | |
520 | } | |
521 | ||
522 | /* | |
523 | * The next function is only to split too long sys_mq_timedsend | |
524 | */ | |
525 | static void __do_notify(struct mqueue_inode_info *info) | |
526 | { | |
527 | /* notification | |
528 | * invoked when there is registered process and there isn't process | |
529 | * waiting synchronously for message AND state of queue changed from | |
530 | * empty to not empty. Here we are sure that no one is waiting | |
531 | * synchronously. */ | |
532 | if (info->notify_owner && | |
533 | info->attr.mq_curmsgs == 1) { | |
534 | struct siginfo sig_i; | |
535 | switch (info->notify.sigev_notify) { | |
536 | case SIGEV_NONE: | |
537 | break; | |
538 | case SIGEV_SIGNAL: | |
539 | /* sends signal */ | |
540 | ||
541 | sig_i.si_signo = info->notify.sigev_signo; | |
542 | sig_i.si_errno = 0; | |
543 | sig_i.si_code = SI_MESGQ; | |
544 | sig_i.si_value = info->notify.sigev_value; | |
6b550f94 SH |
545 | /* map current pid/uid into info->owner's namespaces */ |
546 | rcu_read_lock(); | |
a6684999 SB |
547 | sig_i.si_pid = task_tgid_nr_ns(current, |
548 | ns_of_pid(info->notify_owner)); | |
6b550f94 SH |
549 | sig_i.si_uid = user_ns_map_uid(info->user->user_ns, |
550 | current_cred(), current_uid()); | |
551 | rcu_read_unlock(); | |
1da177e4 | 552 | |
a03fcb73 CLG |
553 | kill_pid_info(info->notify.sigev_signo, |
554 | &sig_i, info->notify_owner); | |
1da177e4 LT |
555 | break; |
556 | case SIGEV_THREAD: | |
557 | set_cookie(info->notify_cookie, NOTIFY_WOKENUP); | |
7ee015e0 | 558 | netlink_sendskb(info->notify_sock, info->notify_cookie); |
1da177e4 LT |
559 | break; |
560 | } | |
561 | /* after notification unregisters process */ | |
a03fcb73 CLG |
562 | put_pid(info->notify_owner); |
563 | info->notify_owner = NULL; | |
1da177e4 LT |
564 | } |
565 | wake_up(&info->wait_q); | |
566 | } | |
567 | ||
9ca7d8e6 CE |
568 | static int prepare_timeout(const struct timespec __user *u_abs_timeout, |
569 | ktime_t *expires, struct timespec *ts) | |
1da177e4 | 570 | { |
9ca7d8e6 CE |
571 | if (copy_from_user(ts, u_abs_timeout, sizeof(struct timespec))) |
572 | return -EFAULT; | |
573 | if (!timespec_valid(ts)) | |
574 | return -EINVAL; | |
1da177e4 | 575 | |
9ca7d8e6 CE |
576 | *expires = timespec_to_ktime(*ts); |
577 | return 0; | |
1da177e4 LT |
578 | } |
579 | ||
580 | static void remove_notification(struct mqueue_inode_info *info) | |
581 | { | |
a03fcb73 | 582 | if (info->notify_owner != NULL && |
1da177e4 LT |
583 | info->notify.sigev_notify == SIGEV_THREAD) { |
584 | set_cookie(info->notify_cookie, NOTIFY_REMOVED); | |
7ee015e0 | 585 | netlink_sendskb(info->notify_sock, info->notify_cookie); |
1da177e4 | 586 | } |
a03fcb73 CLG |
587 | put_pid(info->notify_owner); |
588 | info->notify_owner = NULL; | |
1da177e4 LT |
589 | } |
590 | ||
614b84cf | 591 | static int mq_attr_ok(struct ipc_namespace *ipc_ns, struct mq_attr *attr) |
1da177e4 LT |
592 | { |
593 | if (attr->mq_maxmsg <= 0 || attr->mq_msgsize <= 0) | |
594 | return 0; | |
595 | if (capable(CAP_SYS_RESOURCE)) { | |
596 | if (attr->mq_maxmsg > HARD_MSGMAX) | |
597 | return 0; | |
598 | } else { | |
614b84cf SH |
599 | if (attr->mq_maxmsg > ipc_ns->mq_msg_max || |
600 | attr->mq_msgsize > ipc_ns->mq_msgsize_max) | |
1da177e4 LT |
601 | return 0; |
602 | } | |
603 | /* check for overflow */ | |
604 | if (attr->mq_msgsize > ULONG_MAX/attr->mq_maxmsg) | |
605 | return 0; | |
8834cf79 AGR |
606 | if ((unsigned long)(attr->mq_maxmsg * (attr->mq_msgsize |
607 | + sizeof (struct msg_msg *))) < | |
1da177e4 LT |
608 | (unsigned long)(attr->mq_maxmsg * attr->mq_msgsize)) |
609 | return 0; | |
610 | return 1; | |
611 | } | |
612 | ||
613 | /* | |
614 | * Invoked when creating a new queue via sys_mq_open | |
615 | */ | |
614b84cf | 616 | static struct file *do_create(struct ipc_namespace *ipc_ns, struct dentry *dir, |
4acdaf27 | 617 | struct dentry *dentry, int oflag, umode_t mode, |
614b84cf | 618 | struct mq_attr *attr) |
1da177e4 | 619 | { |
745ca247 | 620 | const struct cred *cred = current_cred(); |
4a3fd211 | 621 | struct file *result; |
1da177e4 LT |
622 | int ret; |
623 | ||
564f6993 | 624 | if (attr) { |
8d8ffefa AGR |
625 | if (!mq_attr_ok(ipc_ns, attr)) { |
626 | ret = -EINVAL; | |
7c7dce92 | 627 | goto out; |
8d8ffefa | 628 | } |
1da177e4 | 629 | /* store for use during create */ |
564f6993 | 630 | dentry->d_fsdata = attr; |
1da177e4 LT |
631 | } |
632 | ||
ce3b0f8d | 633 | mode &= ~current_umask(); |
614b84cf | 634 | ret = mnt_want_write(ipc_ns->mq_mnt); |
4a3fd211 DH |
635 | if (ret) |
636 | goto out; | |
1da177e4 LT |
637 | ret = vfs_create(dir->d_inode, dentry, mode, NULL); |
638 | dentry->d_fsdata = NULL; | |
639 | if (ret) | |
4a3fd211 DH |
640 | goto out_drop_write; |
641 | ||
614b84cf | 642 | result = dentry_open(dentry, ipc_ns->mq_mnt, oflag, cred); |
4a3fd211 DH |
643 | /* |
644 | * dentry_open() took a persistent mnt_want_write(), | |
645 | * so we can now drop this one. | |
646 | */ | |
614b84cf | 647 | mnt_drop_write(ipc_ns->mq_mnt); |
4a3fd211 DH |
648 | return result; |
649 | ||
650 | out_drop_write: | |
614b84cf | 651 | mnt_drop_write(ipc_ns->mq_mnt); |
7c7dce92 AV |
652 | out: |
653 | dput(dentry); | |
614b84cf | 654 | mntput(ipc_ns->mq_mnt); |
7c7dce92 | 655 | return ERR_PTR(ret); |
1da177e4 LT |
656 | } |
657 | ||
658 | /* Opens existing queue */ | |
614b84cf SH |
659 | static struct file *do_open(struct ipc_namespace *ipc_ns, |
660 | struct dentry *dentry, int oflag) | |
1da177e4 | 661 | { |
04db0dde | 662 | int ret; |
745ca247 DH |
663 | const struct cred *cred = current_cred(); |
664 | ||
665 | static const int oflag2acc[O_ACCMODE] = { MAY_READ, MAY_WRITE, | |
666 | MAY_READ | MAY_WRITE }; | |
1da177e4 | 667 | |
7c7dce92 | 668 | if ((oflag & O_ACCMODE) == (O_RDWR | O_WRONLY)) { |
04db0dde AGR |
669 | ret = -EINVAL; |
670 | goto err; | |
7c7dce92 | 671 | } |
1da177e4 | 672 | |
f419a2e3 | 673 | if (inode_permission(dentry->d_inode, oflag2acc[oflag & O_ACCMODE])) { |
04db0dde AGR |
674 | ret = -EACCES; |
675 | goto err; | |
7c7dce92 | 676 | } |
1da177e4 | 677 | |
614b84cf | 678 | return dentry_open(dentry, ipc_ns->mq_mnt, oflag, cred); |
04db0dde AGR |
679 | |
680 | err: | |
681 | dput(dentry); | |
682 | mntput(ipc_ns->mq_mnt); | |
683 | return ERR_PTR(ret); | |
1da177e4 LT |
684 | } |
685 | ||
df0a4283 | 686 | SYSCALL_DEFINE4(mq_open, const char __user *, u_name, int, oflag, umode_t, mode, |
d5460c99 | 687 | struct mq_attr __user *, u_attr) |
1da177e4 LT |
688 | { |
689 | struct dentry *dentry; | |
690 | struct file *filp; | |
691 | char *name; | |
564f6993 | 692 | struct mq_attr attr; |
1da177e4 | 693 | int fd, error; |
7eafd7c7 | 694 | struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns; |
1da177e4 | 695 | |
564f6993 AV |
696 | if (u_attr && copy_from_user(&attr, u_attr, sizeof(struct mq_attr))) |
697 | return -EFAULT; | |
698 | ||
699 | audit_mq_open(oflag, mode, u_attr ? &attr : NULL); | |
20ca73bc | 700 | |
1da177e4 LT |
701 | if (IS_ERR(name = getname(u_name))) |
702 | return PTR_ERR(name); | |
703 | ||
269f2134 | 704 | fd = get_unused_fd_flags(O_CLOEXEC); |
1da177e4 LT |
705 | if (fd < 0) |
706 | goto out_putname; | |
707 | ||
614b84cf SH |
708 | mutex_lock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex); |
709 | dentry = lookup_one_len(name, ipc_ns->mq_mnt->mnt_root, strlen(name)); | |
1da177e4 LT |
710 | if (IS_ERR(dentry)) { |
711 | error = PTR_ERR(dentry); | |
4294a8ee | 712 | goto out_putfd; |
1da177e4 | 713 | } |
614b84cf | 714 | mntget(ipc_ns->mq_mnt); |
1da177e4 LT |
715 | |
716 | if (oflag & O_CREAT) { | |
717 | if (dentry->d_inode) { /* entry already exists */ | |
5a190ae6 | 718 | audit_inode(name, dentry); |
8d8ffefa AGR |
719 | if (oflag & O_EXCL) { |
720 | error = -EEXIST; | |
7c7dce92 | 721 | goto out; |
8d8ffefa | 722 | } |
614b84cf | 723 | filp = do_open(ipc_ns, dentry, oflag); |
1da177e4 | 724 | } else { |
614b84cf SH |
725 | filp = do_create(ipc_ns, ipc_ns->mq_mnt->mnt_root, |
726 | dentry, oflag, mode, | |
564f6993 | 727 | u_attr ? &attr : NULL); |
1da177e4 | 728 | } |
7c7dce92 | 729 | } else { |
8d8ffefa AGR |
730 | if (!dentry->d_inode) { |
731 | error = -ENOENT; | |
7c7dce92 | 732 | goto out; |
8d8ffefa | 733 | } |
5a190ae6 | 734 | audit_inode(name, dentry); |
614b84cf | 735 | filp = do_open(ipc_ns, dentry, oflag); |
7c7dce92 | 736 | } |
1da177e4 LT |
737 | |
738 | if (IS_ERR(filp)) { | |
739 | error = PTR_ERR(filp); | |
740 | goto out_putfd; | |
741 | } | |
742 | ||
1da177e4 LT |
743 | fd_install(fd, filp); |
744 | goto out_upsem; | |
745 | ||
7c7dce92 AV |
746 | out: |
747 | dput(dentry); | |
614b84cf | 748 | mntput(ipc_ns->mq_mnt); |
7c7dce92 | 749 | out_putfd: |
1da177e4 | 750 | put_unused_fd(fd); |
1da177e4 LT |
751 | fd = error; |
752 | out_upsem: | |
614b84cf | 753 | mutex_unlock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex); |
1da177e4 LT |
754 | out_putname: |
755 | putname(name); | |
756 | return fd; | |
757 | } | |
758 | ||
d5460c99 | 759 | SYSCALL_DEFINE1(mq_unlink, const char __user *, u_name) |
1da177e4 LT |
760 | { |
761 | int err; | |
762 | char *name; | |
763 | struct dentry *dentry; | |
764 | struct inode *inode = NULL; | |
7eafd7c7 | 765 | struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns; |
1da177e4 LT |
766 | |
767 | name = getname(u_name); | |
768 | if (IS_ERR(name)) | |
769 | return PTR_ERR(name); | |
770 | ||
614b84cf | 771 | mutex_lock_nested(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex, |
7a434814 | 772 | I_MUTEX_PARENT); |
614b84cf | 773 | dentry = lookup_one_len(name, ipc_ns->mq_mnt->mnt_root, strlen(name)); |
1da177e4 LT |
774 | if (IS_ERR(dentry)) { |
775 | err = PTR_ERR(dentry); | |
776 | goto out_unlock; | |
777 | } | |
778 | ||
779 | if (!dentry->d_inode) { | |
780 | err = -ENOENT; | |
781 | goto out_err; | |
782 | } | |
783 | ||
784 | inode = dentry->d_inode; | |
785 | if (inode) | |
7de9c6ee | 786 | ihold(inode); |
614b84cf | 787 | err = mnt_want_write(ipc_ns->mq_mnt); |
0622753b DH |
788 | if (err) |
789 | goto out_err; | |
1da177e4 | 790 | err = vfs_unlink(dentry->d_parent->d_inode, dentry); |
614b84cf | 791 | mnt_drop_write(ipc_ns->mq_mnt); |
1da177e4 LT |
792 | out_err: |
793 | dput(dentry); | |
794 | ||
795 | out_unlock: | |
614b84cf | 796 | mutex_unlock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex); |
1da177e4 LT |
797 | putname(name); |
798 | if (inode) | |
799 | iput(inode); | |
800 | ||
801 | return err; | |
802 | } | |
803 | ||
804 | /* Pipelined send and receive functions. | |
805 | * | |
806 | * If a receiver finds no waiting message, then it registers itself in the | |
807 | * list of waiting receivers. A sender checks that list before adding the new | |
808 | * message into the message array. If there is a waiting receiver, then it | |
809 | * bypasses the message array and directly hands the message over to the | |
810 | * receiver. | |
811 | * The receiver accepts the message and returns without grabbing the queue | |
812 | * spinlock. Therefore an intermediate STATE_PENDING state and memory barriers | |
813 | * are necessary. The same algorithm is used for sysv semaphores, see | |
7b7a317c | 814 | * ipc/sem.c for more details. |
1da177e4 LT |
815 | * |
816 | * The same algorithm is used for senders. | |
817 | */ | |
818 | ||
819 | /* pipelined_send() - send a message directly to the task waiting in | |
820 | * sys_mq_timedreceive() (without inserting message into a queue). | |
821 | */ | |
822 | static inline void pipelined_send(struct mqueue_inode_info *info, | |
823 | struct msg_msg *message, | |
824 | struct ext_wait_queue *receiver) | |
825 | { | |
826 | receiver->msg = message; | |
827 | list_del(&receiver->list); | |
828 | receiver->state = STATE_PENDING; | |
829 | wake_up_process(receiver->task); | |
d59dd462 | 830 | smp_wmb(); |
1da177e4 LT |
831 | receiver->state = STATE_READY; |
832 | } | |
833 | ||
834 | /* pipelined_receive() - if there is task waiting in sys_mq_timedsend() | |
835 | * gets its message and put to the queue (we have one free place for sure). */ | |
836 | static inline void pipelined_receive(struct mqueue_inode_info *info) | |
837 | { | |
838 | struct ext_wait_queue *sender = wq_get_first_waiter(info, SEND); | |
839 | ||
840 | if (!sender) { | |
841 | /* for poll */ | |
842 | wake_up_interruptible(&info->wait_q); | |
843 | return; | |
844 | } | |
845 | msg_insert(sender->msg, info); | |
846 | list_del(&sender->list); | |
847 | sender->state = STATE_PENDING; | |
848 | wake_up_process(sender->task); | |
d59dd462 | 849 | smp_wmb(); |
1da177e4 LT |
850 | sender->state = STATE_READY; |
851 | } | |
852 | ||
c4ea37c2 HC |
853 | SYSCALL_DEFINE5(mq_timedsend, mqd_t, mqdes, const char __user *, u_msg_ptr, |
854 | size_t, msg_len, unsigned int, msg_prio, | |
855 | const struct timespec __user *, u_abs_timeout) | |
1da177e4 LT |
856 | { |
857 | struct file *filp; | |
858 | struct inode *inode; | |
859 | struct ext_wait_queue wait; | |
860 | struct ext_wait_queue *receiver; | |
861 | struct msg_msg *msg_ptr; | |
862 | struct mqueue_inode_info *info; | |
9ca7d8e6 CE |
863 | ktime_t expires, *timeout = NULL; |
864 | struct timespec ts; | |
1da177e4 LT |
865 | int ret; |
866 | ||
c32c8af4 | 867 | if (u_abs_timeout) { |
9ca7d8e6 CE |
868 | int res = prepare_timeout(u_abs_timeout, &expires, &ts); |
869 | if (res) | |
870 | return res; | |
871 | timeout = &expires; | |
c32c8af4 | 872 | } |
20ca73bc | 873 | |
1da177e4 LT |
874 | if (unlikely(msg_prio >= (unsigned long) MQ_PRIO_MAX)) |
875 | return -EINVAL; | |
876 | ||
9ca7d8e6 | 877 | audit_mq_sendrecv(mqdes, msg_len, msg_prio, timeout ? &ts : NULL); |
1da177e4 | 878 | |
1da177e4 | 879 | filp = fget(mqdes); |
8d8ffefa AGR |
880 | if (unlikely(!filp)) { |
881 | ret = -EBADF; | |
1da177e4 | 882 | goto out; |
8d8ffefa | 883 | } |
1da177e4 | 884 | |
6d63079a | 885 | inode = filp->f_path.dentry->d_inode; |
8d8ffefa AGR |
886 | if (unlikely(filp->f_op != &mqueue_file_operations)) { |
887 | ret = -EBADF; | |
1da177e4 | 888 | goto out_fput; |
8d8ffefa | 889 | } |
1da177e4 | 890 | info = MQUEUE_I(inode); |
5a190ae6 | 891 | audit_inode(NULL, filp->f_path.dentry); |
1da177e4 | 892 | |
8d8ffefa AGR |
893 | if (unlikely(!(filp->f_mode & FMODE_WRITE))) { |
894 | ret = -EBADF; | |
1da177e4 | 895 | goto out_fput; |
8d8ffefa | 896 | } |
1da177e4 LT |
897 | |
898 | if (unlikely(msg_len > info->attr.mq_msgsize)) { | |
899 | ret = -EMSGSIZE; | |
900 | goto out_fput; | |
901 | } | |
902 | ||
903 | /* First try to allocate memory, before doing anything with | |
904 | * existing queues. */ | |
905 | msg_ptr = load_msg(u_msg_ptr, msg_len); | |
906 | if (IS_ERR(msg_ptr)) { | |
907 | ret = PTR_ERR(msg_ptr); | |
908 | goto out_fput; | |
909 | } | |
910 | msg_ptr->m_ts = msg_len; | |
911 | msg_ptr->m_type = msg_prio; | |
912 | ||
913 | spin_lock(&info->lock); | |
914 | ||
915 | if (info->attr.mq_curmsgs == info->attr.mq_maxmsg) { | |
916 | if (filp->f_flags & O_NONBLOCK) { | |
917 | spin_unlock(&info->lock); | |
918 | ret = -EAGAIN; | |
1da177e4 LT |
919 | } else { |
920 | wait.task = current; | |
921 | wait.msg = (void *) msg_ptr; | |
922 | wait.state = STATE_NONE; | |
923 | ret = wq_sleep(info, SEND, timeout, &wait); | |
924 | } | |
925 | if (ret < 0) | |
926 | free_msg(msg_ptr); | |
927 | } else { | |
928 | receiver = wq_get_first_waiter(info, RECV); | |
929 | if (receiver) { | |
930 | pipelined_send(info, msg_ptr, receiver); | |
931 | } else { | |
932 | /* adds message to the queue */ | |
933 | msg_insert(msg_ptr, info); | |
934 | __do_notify(info); | |
935 | } | |
936 | inode->i_atime = inode->i_mtime = inode->i_ctime = | |
937 | CURRENT_TIME; | |
938 | spin_unlock(&info->lock); | |
939 | ret = 0; | |
940 | } | |
941 | out_fput: | |
942 | fput(filp); | |
943 | out: | |
944 | return ret; | |
945 | } | |
946 | ||
c4ea37c2 HC |
947 | SYSCALL_DEFINE5(mq_timedreceive, mqd_t, mqdes, char __user *, u_msg_ptr, |
948 | size_t, msg_len, unsigned int __user *, u_msg_prio, | |
949 | const struct timespec __user *, u_abs_timeout) | |
1da177e4 | 950 | { |
1da177e4 LT |
951 | ssize_t ret; |
952 | struct msg_msg *msg_ptr; | |
953 | struct file *filp; | |
954 | struct inode *inode; | |
955 | struct mqueue_inode_info *info; | |
956 | struct ext_wait_queue wait; | |
9ca7d8e6 CE |
957 | ktime_t expires, *timeout = NULL; |
958 | struct timespec ts; | |
1da177e4 | 959 | |
c32c8af4 | 960 | if (u_abs_timeout) { |
9ca7d8e6 CE |
961 | int res = prepare_timeout(u_abs_timeout, &expires, &ts); |
962 | if (res) | |
963 | return res; | |
964 | timeout = &expires; | |
c32c8af4 | 965 | } |
20ca73bc | 966 | |
9ca7d8e6 | 967 | audit_mq_sendrecv(mqdes, msg_len, 0, timeout ? &ts : NULL); |
1da177e4 | 968 | |
1da177e4 | 969 | filp = fget(mqdes); |
8d8ffefa AGR |
970 | if (unlikely(!filp)) { |
971 | ret = -EBADF; | |
1da177e4 | 972 | goto out; |
8d8ffefa | 973 | } |
1da177e4 | 974 | |
6d63079a | 975 | inode = filp->f_path.dentry->d_inode; |
8d8ffefa AGR |
976 | if (unlikely(filp->f_op != &mqueue_file_operations)) { |
977 | ret = -EBADF; | |
1da177e4 | 978 | goto out_fput; |
8d8ffefa | 979 | } |
1da177e4 | 980 | info = MQUEUE_I(inode); |
5a190ae6 | 981 | audit_inode(NULL, filp->f_path.dentry); |
1da177e4 | 982 | |
8d8ffefa AGR |
983 | if (unlikely(!(filp->f_mode & FMODE_READ))) { |
984 | ret = -EBADF; | |
1da177e4 | 985 | goto out_fput; |
8d8ffefa | 986 | } |
1da177e4 LT |
987 | |
988 | /* checks if buffer is big enough */ | |
989 | if (unlikely(msg_len < info->attr.mq_msgsize)) { | |
990 | ret = -EMSGSIZE; | |
991 | goto out_fput; | |
992 | } | |
993 | ||
994 | spin_lock(&info->lock); | |
995 | if (info->attr.mq_curmsgs == 0) { | |
996 | if (filp->f_flags & O_NONBLOCK) { | |
997 | spin_unlock(&info->lock); | |
998 | ret = -EAGAIN; | |
1da177e4 LT |
999 | } else { |
1000 | wait.task = current; | |
1001 | wait.state = STATE_NONE; | |
1002 | ret = wq_sleep(info, RECV, timeout, &wait); | |
1003 | msg_ptr = wait.msg; | |
1004 | } | |
1005 | } else { | |
1006 | msg_ptr = msg_get(info); | |
1007 | ||
1008 | inode->i_atime = inode->i_mtime = inode->i_ctime = | |
1009 | CURRENT_TIME; | |
1010 | ||
1011 | /* There is now free space in queue. */ | |
1012 | pipelined_receive(info); | |
1013 | spin_unlock(&info->lock); | |
1014 | ret = 0; | |
1015 | } | |
1016 | if (ret == 0) { | |
1017 | ret = msg_ptr->m_ts; | |
1018 | ||
1019 | if ((u_msg_prio && put_user(msg_ptr->m_type, u_msg_prio)) || | |
1020 | store_msg(u_msg_ptr, msg_ptr, msg_ptr->m_ts)) { | |
1021 | ret = -EFAULT; | |
1022 | } | |
1023 | free_msg(msg_ptr); | |
1024 | } | |
1025 | out_fput: | |
1026 | fput(filp); | |
1027 | out: | |
1028 | return ret; | |
1029 | } | |
1030 | ||
1031 | /* | |
1032 | * Notes: the case when user wants us to deregister (with NULL as pointer) | |
1033 | * and he isn't currently owner of notification, will be silently discarded. | |
1034 | * It isn't explicitly defined in the POSIX. | |
1035 | */ | |
c4ea37c2 HC |
1036 | SYSCALL_DEFINE2(mq_notify, mqd_t, mqdes, |
1037 | const struct sigevent __user *, u_notification) | |
1da177e4 LT |
1038 | { |
1039 | int ret; | |
1040 | struct file *filp; | |
1041 | struct sock *sock; | |
1042 | struct inode *inode; | |
1043 | struct sigevent notification; | |
1044 | struct mqueue_inode_info *info; | |
1045 | struct sk_buff *nc; | |
1046 | ||
20114f71 | 1047 | if (u_notification) { |
1da177e4 LT |
1048 | if (copy_from_user(¬ification, u_notification, |
1049 | sizeof(struct sigevent))) | |
1050 | return -EFAULT; | |
20114f71 AV |
1051 | } |
1052 | ||
1053 | audit_mq_notify(mqdes, u_notification ? ¬ification : NULL); | |
1da177e4 | 1054 | |
20114f71 AV |
1055 | nc = NULL; |
1056 | sock = NULL; | |
1057 | if (u_notification != NULL) { | |
1da177e4 LT |
1058 | if (unlikely(notification.sigev_notify != SIGEV_NONE && |
1059 | notification.sigev_notify != SIGEV_SIGNAL && | |
1060 | notification.sigev_notify != SIGEV_THREAD)) | |
1061 | return -EINVAL; | |
1062 | if (notification.sigev_notify == SIGEV_SIGNAL && | |
7ed20e1a | 1063 | !valid_signal(notification.sigev_signo)) { |
1da177e4 LT |
1064 | return -EINVAL; |
1065 | } | |
1066 | if (notification.sigev_notify == SIGEV_THREAD) { | |
c3d8d1e3 PM |
1067 | long timeo; |
1068 | ||
1da177e4 LT |
1069 | /* create the notify skb */ |
1070 | nc = alloc_skb(NOTIFY_COOKIE_LEN, GFP_KERNEL); | |
8d8ffefa AGR |
1071 | if (!nc) { |
1072 | ret = -ENOMEM; | |
1da177e4 | 1073 | goto out; |
8d8ffefa | 1074 | } |
1da177e4 LT |
1075 | if (copy_from_user(nc->data, |
1076 | notification.sigev_value.sival_ptr, | |
1077 | NOTIFY_COOKIE_LEN)) { | |
8d8ffefa | 1078 | ret = -EFAULT; |
1da177e4 LT |
1079 | goto out; |
1080 | } | |
1081 | ||
1082 | /* TODO: add a header? */ | |
1083 | skb_put(nc, NOTIFY_COOKIE_LEN); | |
1084 | /* and attach it to the socket */ | |
1085 | retry: | |
1086 | filp = fget(notification.sigev_signo); | |
8d8ffefa AGR |
1087 | if (!filp) { |
1088 | ret = -EBADF; | |
1da177e4 | 1089 | goto out; |
8d8ffefa | 1090 | } |
1da177e4 LT |
1091 | sock = netlink_getsockbyfilp(filp); |
1092 | fput(filp); | |
1093 | if (IS_ERR(sock)) { | |
1094 | ret = PTR_ERR(sock); | |
1095 | sock = NULL; | |
1096 | goto out; | |
1097 | } | |
1098 | ||
c3d8d1e3 | 1099 | timeo = MAX_SCHEDULE_TIMEOUT; |
9457afee | 1100 | ret = netlink_attachskb(sock, nc, &timeo, NULL); |
1da177e4 | 1101 | if (ret == 1) |
8d8ffefa | 1102 | goto retry; |
1da177e4 LT |
1103 | if (ret) { |
1104 | sock = NULL; | |
1105 | nc = NULL; | |
1106 | goto out; | |
1107 | } | |
1108 | } | |
1109 | } | |
1110 | ||
1da177e4 | 1111 | filp = fget(mqdes); |
8d8ffefa AGR |
1112 | if (!filp) { |
1113 | ret = -EBADF; | |
1da177e4 | 1114 | goto out; |
8d8ffefa | 1115 | } |
1da177e4 | 1116 | |
6d63079a | 1117 | inode = filp->f_path.dentry->d_inode; |
8d8ffefa AGR |
1118 | if (unlikely(filp->f_op != &mqueue_file_operations)) { |
1119 | ret = -EBADF; | |
1da177e4 | 1120 | goto out_fput; |
8d8ffefa | 1121 | } |
1da177e4 LT |
1122 | info = MQUEUE_I(inode); |
1123 | ||
1124 | ret = 0; | |
1125 | spin_lock(&info->lock); | |
1126 | if (u_notification == NULL) { | |
a03fcb73 | 1127 | if (info->notify_owner == task_tgid(current)) { |
1da177e4 LT |
1128 | remove_notification(info); |
1129 | inode->i_atime = inode->i_ctime = CURRENT_TIME; | |
1130 | } | |
a03fcb73 | 1131 | } else if (info->notify_owner != NULL) { |
1da177e4 LT |
1132 | ret = -EBUSY; |
1133 | } else { | |
1134 | switch (notification.sigev_notify) { | |
1135 | case SIGEV_NONE: | |
1136 | info->notify.sigev_notify = SIGEV_NONE; | |
1137 | break; | |
1138 | case SIGEV_THREAD: | |
1139 | info->notify_sock = sock; | |
1140 | info->notify_cookie = nc; | |
1141 | sock = NULL; | |
1142 | nc = NULL; | |
1143 | info->notify.sigev_notify = SIGEV_THREAD; | |
1144 | break; | |
1145 | case SIGEV_SIGNAL: | |
1146 | info->notify.sigev_signo = notification.sigev_signo; | |
1147 | info->notify.sigev_value = notification.sigev_value; | |
1148 | info->notify.sigev_notify = SIGEV_SIGNAL; | |
1149 | break; | |
1150 | } | |
a03fcb73 CLG |
1151 | |
1152 | info->notify_owner = get_pid(task_tgid(current)); | |
1da177e4 LT |
1153 | inode->i_atime = inode->i_ctime = CURRENT_TIME; |
1154 | } | |
1155 | spin_unlock(&info->lock); | |
1156 | out_fput: | |
1157 | fput(filp); | |
1158 | out: | |
1159 | if (sock) { | |
1160 | netlink_detachskb(sock, nc); | |
1161 | } else if (nc) { | |
1162 | dev_kfree_skb(nc); | |
1163 | } | |
1164 | return ret; | |
1165 | } | |
1166 | ||
c4ea37c2 HC |
1167 | SYSCALL_DEFINE3(mq_getsetattr, mqd_t, mqdes, |
1168 | const struct mq_attr __user *, u_mqstat, | |
1169 | struct mq_attr __user *, u_omqstat) | |
1da177e4 LT |
1170 | { |
1171 | int ret; | |
1172 | struct mq_attr mqstat, omqstat; | |
1173 | struct file *filp; | |
1174 | struct inode *inode; | |
1175 | struct mqueue_inode_info *info; | |
1176 | ||
1177 | if (u_mqstat != NULL) { | |
1178 | if (copy_from_user(&mqstat, u_mqstat, sizeof(struct mq_attr))) | |
1179 | return -EFAULT; | |
1180 | if (mqstat.mq_flags & (~O_NONBLOCK)) | |
1181 | return -EINVAL; | |
1182 | } | |
1183 | ||
1da177e4 | 1184 | filp = fget(mqdes); |
8d8ffefa AGR |
1185 | if (!filp) { |
1186 | ret = -EBADF; | |
1da177e4 | 1187 | goto out; |
8d8ffefa | 1188 | } |
1da177e4 | 1189 | |
6d63079a | 1190 | inode = filp->f_path.dentry->d_inode; |
8d8ffefa AGR |
1191 | if (unlikely(filp->f_op != &mqueue_file_operations)) { |
1192 | ret = -EBADF; | |
1da177e4 | 1193 | goto out_fput; |
8d8ffefa | 1194 | } |
1da177e4 LT |
1195 | info = MQUEUE_I(inode); |
1196 | ||
1197 | spin_lock(&info->lock); | |
1198 | ||
1199 | omqstat = info->attr; | |
1200 | omqstat.mq_flags = filp->f_flags & O_NONBLOCK; | |
1201 | if (u_mqstat) { | |
7392906e | 1202 | audit_mq_getsetattr(mqdes, &mqstat); |
db1dd4d3 | 1203 | spin_lock(&filp->f_lock); |
1da177e4 LT |
1204 | if (mqstat.mq_flags & O_NONBLOCK) |
1205 | filp->f_flags |= O_NONBLOCK; | |
1206 | else | |
1207 | filp->f_flags &= ~O_NONBLOCK; | |
db1dd4d3 | 1208 | spin_unlock(&filp->f_lock); |
1da177e4 LT |
1209 | |
1210 | inode->i_atime = inode->i_ctime = CURRENT_TIME; | |
1211 | } | |
1212 | ||
1213 | spin_unlock(&info->lock); | |
1214 | ||
1215 | ret = 0; | |
1216 | if (u_omqstat != NULL && copy_to_user(u_omqstat, &omqstat, | |
1217 | sizeof(struct mq_attr))) | |
1218 | ret = -EFAULT; | |
1219 | ||
1220 | out_fput: | |
1221 | fput(filp); | |
1222 | out: | |
1223 | return ret; | |
1224 | } | |
1225 | ||
92e1d5be | 1226 | static const struct inode_operations mqueue_dir_inode_operations = { |
1da177e4 LT |
1227 | .lookup = simple_lookup, |
1228 | .create = mqueue_create, | |
1229 | .unlink = mqueue_unlink, | |
1230 | }; | |
1231 | ||
9a32144e | 1232 | static const struct file_operations mqueue_file_operations = { |
1da177e4 LT |
1233 | .flush = mqueue_flush_file, |
1234 | .poll = mqueue_poll_file, | |
1235 | .read = mqueue_read_file, | |
6038f373 | 1236 | .llseek = default_llseek, |
1da177e4 LT |
1237 | }; |
1238 | ||
b87221de | 1239 | static const struct super_operations mqueue_super_ops = { |
1da177e4 LT |
1240 | .alloc_inode = mqueue_alloc_inode, |
1241 | .destroy_inode = mqueue_destroy_inode, | |
6d8af64c | 1242 | .evict_inode = mqueue_evict_inode, |
1da177e4 | 1243 | .statfs = simple_statfs, |
1da177e4 LT |
1244 | }; |
1245 | ||
1246 | static struct file_system_type mqueue_fs_type = { | |
1247 | .name = "mqueue", | |
ceefda69 | 1248 | .mount = mqueue_mount, |
1da177e4 LT |
1249 | .kill_sb = kill_litter_super, |
1250 | }; | |
1251 | ||
7eafd7c7 SH |
1252 | int mq_init_ns(struct ipc_namespace *ns) |
1253 | { | |
1254 | ns->mq_queues_count = 0; | |
1255 | ns->mq_queues_max = DFLT_QUEUESMAX; | |
1256 | ns->mq_msg_max = DFLT_MSGMAX; | |
1257 | ns->mq_msgsize_max = DFLT_MSGSIZEMAX; | |
1258 | ||
1259 | ns->mq_mnt = kern_mount_data(&mqueue_fs_type, ns); | |
1260 | if (IS_ERR(ns->mq_mnt)) { | |
1261 | int err = PTR_ERR(ns->mq_mnt); | |
1262 | ns->mq_mnt = NULL; | |
1263 | return err; | |
1264 | } | |
1265 | return 0; | |
1266 | } | |
1267 | ||
1268 | void mq_clear_sbinfo(struct ipc_namespace *ns) | |
1269 | { | |
1270 | ns->mq_mnt->mnt_sb->s_fs_info = NULL; | |
1271 | } | |
1272 | ||
1273 | void mq_put_mnt(struct ipc_namespace *ns) | |
1274 | { | |
6f686574 | 1275 | kern_unmount(ns->mq_mnt); |
7eafd7c7 SH |
1276 | } |
1277 | ||
1da177e4 LT |
1278 | static int __init init_mqueue_fs(void) |
1279 | { | |
1280 | int error; | |
1281 | ||
1282 | mqueue_inode_cachep = kmem_cache_create("mqueue_inode_cache", | |
1283 | sizeof(struct mqueue_inode_info), 0, | |
20c2df83 | 1284 | SLAB_HWCACHE_ALIGN, init_once); |
1da177e4 LT |
1285 | if (mqueue_inode_cachep == NULL) |
1286 | return -ENOMEM; | |
1287 | ||
2329e392 | 1288 | /* ignore failures - they are not fatal */ |
bdc8e5f8 | 1289 | mq_sysctl_table = mq_register_sysctl_table(); |
1da177e4 LT |
1290 | |
1291 | error = register_filesystem(&mqueue_fs_type); | |
1292 | if (error) | |
1293 | goto out_sysctl; | |
1294 | ||
7eafd7c7 SH |
1295 | spin_lock_init(&mq_lock); |
1296 | ||
6f686574 AV |
1297 | error = mq_init_ns(&init_ipc_ns); |
1298 | if (error) | |
1da177e4 | 1299 | goto out_filesystem; |
1da177e4 | 1300 | |
1da177e4 LT |
1301 | return 0; |
1302 | ||
1303 | out_filesystem: | |
1304 | unregister_filesystem(&mqueue_fs_type); | |
1305 | out_sysctl: | |
1306 | if (mq_sysctl_table) | |
1307 | unregister_sysctl_table(mq_sysctl_table); | |
1a1d92c1 | 1308 | kmem_cache_destroy(mqueue_inode_cachep); |
1da177e4 LT |
1309 | return error; |
1310 | } | |
1311 | ||
1312 | __initcall(init_mqueue_fs); |