]>
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: | |
239521f3 | 9 | * Manfred Spraul (manfred@colorfullife.com) |
1da177e4 | 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> | |
5b5c4d1a | 27 | #include <linux/vmalloc.h> |
1da177e4 LT |
28 | #include <linux/netlink.h> |
29 | #include <linux/syscalls.h> | |
20ca73bc | 30 | #include <linux/audit.h> |
7ed20e1a | 31 | #include <linux/signal.h> |
5f921ae9 | 32 | #include <linux/mutex.h> |
b488893a PE |
33 | #include <linux/nsproxy.h> |
34 | #include <linux/pid.h> | |
614b84cf | 35 | #include <linux/ipc_namespace.h> |
6b550f94 | 36 | #include <linux/user_namespace.h> |
5a0e3ad6 | 37 | #include <linux/slab.h> |
84f001e1 | 38 | #include <linux/sched/wake_q.h> |
3f07c014 | 39 | #include <linux/sched/signal.h> |
8703e8a4 | 40 | #include <linux/sched/user.h> |
5f921ae9 | 41 | |
1da177e4 LT |
42 | #include <net/sock.h> |
43 | #include "util.h" | |
44 | ||
45 | #define MQUEUE_MAGIC 0x19800202 | |
46 | #define DIRENT_SIZE 20 | |
47 | #define FILENT_SIZE 80 | |
48 | ||
49 | #define SEND 0 | |
50 | #define RECV 1 | |
51 | ||
52 | #define STATE_NONE 0 | |
fa6004ad | 53 | #define STATE_READY 1 |
1da177e4 | 54 | |
d6629859 DL |
55 | struct posix_msg_tree_node { |
56 | struct rb_node rb_node; | |
57 | struct list_head msg_list; | |
58 | int priority; | |
59 | }; | |
60 | ||
1da177e4 LT |
61 | struct ext_wait_queue { /* queue of sleeping tasks */ |
62 | struct task_struct *task; | |
63 | struct list_head list; | |
64 | struct msg_msg *msg; /* ptr of loaded message */ | |
65 | int state; /* one of STATE_* values */ | |
66 | }; | |
67 | ||
68 | struct mqueue_inode_info { | |
69 | spinlock_t lock; | |
70 | struct inode vfs_inode; | |
71 | wait_queue_head_t wait_q; | |
72 | ||
d6629859 | 73 | struct rb_root msg_tree; |
ce2d52cc | 74 | struct posix_msg_tree_node *node_cache; |
1da177e4 LT |
75 | struct mq_attr attr; |
76 | ||
77 | struct sigevent notify; | |
239521f3 | 78 | struct pid *notify_owner; |
6f9ac6d9 | 79 | struct user_namespace *notify_user_ns; |
338cec32 | 80 | struct user_struct *user; /* user who created, for accounting */ |
1da177e4 LT |
81 | struct sock *notify_sock; |
82 | struct sk_buff *notify_cookie; | |
83 | ||
84 | /* for tasks waiting for free space and messages, respectively */ | |
85 | struct ext_wait_queue e_wait_q[2]; | |
86 | ||
87 | unsigned long qsize; /* size of queue in memory (sum of all msgs) */ | |
88 | }; | |
89 | ||
92e1d5be | 90 | static const struct inode_operations mqueue_dir_inode_operations; |
9a32144e | 91 | static const struct file_operations mqueue_file_operations; |
b87221de | 92 | static const struct super_operations mqueue_super_ops; |
1da177e4 LT |
93 | static void remove_notification(struct mqueue_inode_info *info); |
94 | ||
e18b890b | 95 | static struct kmem_cache *mqueue_inode_cachep; |
1da177e4 | 96 | |
239521f3 | 97 | static struct ctl_table_header *mq_sysctl_table; |
1da177e4 LT |
98 | |
99 | static inline struct mqueue_inode_info *MQUEUE_I(struct inode *inode) | |
100 | { | |
101 | return container_of(inode, struct mqueue_inode_info, vfs_inode); | |
102 | } | |
103 | ||
7eafd7c7 SH |
104 | /* |
105 | * This routine should be called with the mq_lock held. | |
106 | */ | |
107 | static inline struct ipc_namespace *__get_ns_from_inode(struct inode *inode) | |
614b84cf | 108 | { |
7eafd7c7 | 109 | return get_ipc_ns(inode->i_sb->s_fs_info); |
614b84cf SH |
110 | } |
111 | ||
7eafd7c7 | 112 | static struct ipc_namespace *get_ns_from_inode(struct inode *inode) |
614b84cf | 113 | { |
7eafd7c7 SH |
114 | struct ipc_namespace *ns; |
115 | ||
116 | spin_lock(&mq_lock); | |
117 | ns = __get_ns_from_inode(inode); | |
118 | spin_unlock(&mq_lock); | |
119 | return ns; | |
614b84cf SH |
120 | } |
121 | ||
d6629859 DL |
122 | /* Auxiliary functions to manipulate messages' list */ |
123 | static int msg_insert(struct msg_msg *msg, struct mqueue_inode_info *info) | |
124 | { | |
125 | struct rb_node **p, *parent = NULL; | |
126 | struct posix_msg_tree_node *leaf; | |
127 | ||
128 | p = &info->msg_tree.rb_node; | |
129 | while (*p) { | |
130 | parent = *p; | |
131 | leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node); | |
132 | ||
133 | if (likely(leaf->priority == msg->m_type)) | |
134 | goto insert_msg; | |
135 | else if (msg->m_type < leaf->priority) | |
136 | p = &(*p)->rb_left; | |
137 | else | |
138 | p = &(*p)->rb_right; | |
139 | } | |
ce2d52cc DL |
140 | if (info->node_cache) { |
141 | leaf = info->node_cache; | |
142 | info->node_cache = NULL; | |
143 | } else { | |
144 | leaf = kmalloc(sizeof(*leaf), GFP_ATOMIC); | |
145 | if (!leaf) | |
146 | return -ENOMEM; | |
ce2d52cc | 147 | INIT_LIST_HEAD(&leaf->msg_list); |
ce2d52cc | 148 | } |
d6629859 DL |
149 | leaf->priority = msg->m_type; |
150 | rb_link_node(&leaf->rb_node, parent, p); | |
151 | rb_insert_color(&leaf->rb_node, &info->msg_tree); | |
d6629859 DL |
152 | insert_msg: |
153 | info->attr.mq_curmsgs++; | |
154 | info->qsize += msg->m_ts; | |
155 | list_add_tail(&msg->m_list, &leaf->msg_list); | |
156 | return 0; | |
157 | } | |
158 | ||
159 | static inline struct msg_msg *msg_get(struct mqueue_inode_info *info) | |
160 | { | |
161 | struct rb_node **p, *parent = NULL; | |
162 | struct posix_msg_tree_node *leaf; | |
163 | struct msg_msg *msg; | |
164 | ||
165 | try_again: | |
166 | p = &info->msg_tree.rb_node; | |
167 | while (*p) { | |
168 | parent = *p; | |
169 | /* | |
170 | * During insert, low priorities go to the left and high to the | |
171 | * right. On receive, we want the highest priorities first, so | |
172 | * walk all the way to the right. | |
173 | */ | |
174 | p = &(*p)->rb_right; | |
175 | } | |
176 | if (!parent) { | |
177 | if (info->attr.mq_curmsgs) { | |
178 | pr_warn_once("Inconsistency in POSIX message queue, " | |
179 | "no tree element, but supposedly messages " | |
180 | "should exist!\n"); | |
181 | info->attr.mq_curmsgs = 0; | |
182 | } | |
183 | return NULL; | |
184 | } | |
185 | leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node); | |
ce2d52cc | 186 | if (unlikely(list_empty(&leaf->msg_list))) { |
d6629859 DL |
187 | pr_warn_once("Inconsistency in POSIX message queue, " |
188 | "empty leaf node but we haven't implemented " | |
189 | "lazy leaf delete!\n"); | |
190 | rb_erase(&leaf->rb_node, &info->msg_tree); | |
ce2d52cc | 191 | if (info->node_cache) { |
ce2d52cc DL |
192 | kfree(leaf); |
193 | } else { | |
194 | info->node_cache = leaf; | |
195 | } | |
d6629859 DL |
196 | goto try_again; |
197 | } else { | |
198 | msg = list_first_entry(&leaf->msg_list, | |
199 | struct msg_msg, m_list); | |
200 | list_del(&msg->m_list); | |
201 | if (list_empty(&leaf->msg_list)) { | |
202 | rb_erase(&leaf->rb_node, &info->msg_tree); | |
ce2d52cc | 203 | if (info->node_cache) { |
ce2d52cc DL |
204 | kfree(leaf); |
205 | } else { | |
206 | info->node_cache = leaf; | |
207 | } | |
d6629859 DL |
208 | } |
209 | } | |
210 | info->attr.mq_curmsgs--; | |
211 | info->qsize -= msg->m_ts; | |
212 | return msg; | |
213 | } | |
214 | ||
7eafd7c7 | 215 | static struct inode *mqueue_get_inode(struct super_block *sb, |
1b9d5ff7 | 216 | struct ipc_namespace *ipc_ns, umode_t mode, |
7eafd7c7 | 217 | struct mq_attr *attr) |
1da177e4 | 218 | { |
86a264ab | 219 | struct user_struct *u = current_user(); |
1da177e4 | 220 | struct inode *inode; |
d40dcdb0 | 221 | int ret = -ENOMEM; |
1da177e4 LT |
222 | |
223 | inode = new_inode(sb); | |
04715206 JS |
224 | if (!inode) |
225 | goto err; | |
226 | ||
227 | inode->i_ino = get_next_ino(); | |
228 | inode->i_mode = mode; | |
229 | inode->i_uid = current_fsuid(); | |
230 | inode->i_gid = current_fsgid(); | |
078cd827 | 231 | inode->i_mtime = inode->i_ctime = inode->i_atime = current_time(inode); |
04715206 JS |
232 | |
233 | if (S_ISREG(mode)) { | |
234 | struct mqueue_inode_info *info; | |
d6629859 | 235 | unsigned long mq_bytes, mq_treesize; |
04715206 JS |
236 | |
237 | inode->i_fop = &mqueue_file_operations; | |
238 | inode->i_size = FILENT_SIZE; | |
239 | /* mqueue specific info */ | |
240 | info = MQUEUE_I(inode); | |
241 | spin_lock_init(&info->lock); | |
242 | init_waitqueue_head(&info->wait_q); | |
243 | INIT_LIST_HEAD(&info->e_wait_q[0].list); | |
244 | INIT_LIST_HEAD(&info->e_wait_q[1].list); | |
245 | info->notify_owner = NULL; | |
6f9ac6d9 | 246 | info->notify_user_ns = NULL; |
04715206 JS |
247 | info->qsize = 0; |
248 | info->user = NULL; /* set when all is ok */ | |
d6629859 | 249 | info->msg_tree = RB_ROOT; |
ce2d52cc | 250 | info->node_cache = NULL; |
04715206 | 251 | memset(&info->attr, 0, sizeof(info->attr)); |
cef0184c KM |
252 | info->attr.mq_maxmsg = min(ipc_ns->mq_msg_max, |
253 | ipc_ns->mq_msg_default); | |
254 | info->attr.mq_msgsize = min(ipc_ns->mq_msgsize_max, | |
255 | ipc_ns->mq_msgsize_default); | |
04715206 JS |
256 | if (attr) { |
257 | info->attr.mq_maxmsg = attr->mq_maxmsg; | |
258 | info->attr.mq_msgsize = attr->mq_msgsize; | |
259 | } | |
d6629859 DL |
260 | /* |
261 | * We used to allocate a static array of pointers and account | |
262 | * the size of that array as well as one msg_msg struct per | |
263 | * possible message into the queue size. That's no longer | |
264 | * accurate as the queue is now an rbtree and will grow and | |
265 | * shrink depending on usage patterns. We can, however, still | |
266 | * account one msg_msg struct per message, but the nodes are | |
267 | * allocated depending on priority usage, and most programs | |
268 | * only use one, or a handful, of priorities. However, since | |
269 | * this is pinned memory, we need to assume worst case, so | |
270 | * that means the min(mq_maxmsg, max_priorities) * struct | |
271 | * posix_msg_tree_node. | |
272 | */ | |
273 | mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) + | |
274 | min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) * | |
275 | sizeof(struct posix_msg_tree_node); | |
04715206 | 276 | |
d6629859 DL |
277 | mq_bytes = mq_treesize + (info->attr.mq_maxmsg * |
278 | info->attr.mq_msgsize); | |
1da177e4 | 279 | |
04715206 JS |
280 | spin_lock(&mq_lock); |
281 | if (u->mq_bytes + mq_bytes < u->mq_bytes || | |
2a4e64b8 | 282 | u->mq_bytes + mq_bytes > rlimit(RLIMIT_MSGQUEUE)) { |
04715206 JS |
283 | spin_unlock(&mq_lock); |
284 | /* mqueue_evict_inode() releases info->messages */ | |
d40dcdb0 | 285 | ret = -EMFILE; |
04715206 | 286 | goto out_inode; |
1da177e4 | 287 | } |
04715206 JS |
288 | u->mq_bytes += mq_bytes; |
289 | spin_unlock(&mq_lock); | |
290 | ||
291 | /* all is ok */ | |
292 | info->user = get_uid(u); | |
293 | } else if (S_ISDIR(mode)) { | |
294 | inc_nlink(inode); | |
295 | /* Some things misbehave if size == 0 on a directory */ | |
296 | inode->i_size = 2 * DIRENT_SIZE; | |
297 | inode->i_op = &mqueue_dir_inode_operations; | |
298 | inode->i_fop = &simple_dir_operations; | |
1da177e4 | 299 | } |
04715206 | 300 | |
1da177e4 LT |
301 | return inode; |
302 | out_inode: | |
1da177e4 | 303 | iput(inode); |
04715206 | 304 | err: |
d40dcdb0 | 305 | return ERR_PTR(ret); |
1da177e4 LT |
306 | } |
307 | ||
308 | static int mqueue_fill_super(struct super_block *sb, void *data, int silent) | |
309 | { | |
310 | struct inode *inode; | |
d91ee87d | 311 | struct ipc_namespace *ns = sb->s_fs_info; |
1da177e4 | 312 | |
a2982cc9 | 313 | sb->s_iflags |= SB_I_NOEXEC | SB_I_NODEV; |
09cbfeaf KS |
314 | sb->s_blocksize = PAGE_SIZE; |
315 | sb->s_blocksize_bits = PAGE_SHIFT; | |
1da177e4 LT |
316 | sb->s_magic = MQUEUE_MAGIC; |
317 | sb->s_op = &mqueue_super_ops; | |
318 | ||
48fde701 AV |
319 | inode = mqueue_get_inode(sb, ns, S_IFDIR | S_ISVTX | S_IRWXUGO, NULL); |
320 | if (IS_ERR(inode)) | |
321 | return PTR_ERR(inode); | |
1da177e4 | 322 | |
48fde701 AV |
323 | sb->s_root = d_make_root(inode); |
324 | if (!sb->s_root) | |
325 | return -ENOMEM; | |
326 | return 0; | |
1da177e4 LT |
327 | } |
328 | ||
ceefda69 | 329 | static struct dentry *mqueue_mount(struct file_system_type *fs_type, |
454e2398 | 330 | int flags, const char *dev_name, |
ceefda69 | 331 | void *data) |
1da177e4 | 332 | { |
d91ee87d | 333 | struct ipc_namespace *ns; |
1751e8a6 | 334 | if (flags & SB_KERNMOUNT) { |
d91ee87d EB |
335 | ns = data; |
336 | data = NULL; | |
337 | } else { | |
338 | ns = current->nsproxy->ipc_ns; | |
a636b702 | 339 | } |
d91ee87d | 340 | return mount_ns(fs_type, flags, data, ns, ns->user_ns, mqueue_fill_super); |
1da177e4 LT |
341 | } |
342 | ||
51cc5068 | 343 | static void init_once(void *foo) |
1da177e4 LT |
344 | { |
345 | struct mqueue_inode_info *p = (struct mqueue_inode_info *) foo; | |
346 | ||
a35afb83 | 347 | inode_init_once(&p->vfs_inode); |
1da177e4 LT |
348 | } |
349 | ||
350 | static struct inode *mqueue_alloc_inode(struct super_block *sb) | |
351 | { | |
352 | struct mqueue_inode_info *ei; | |
353 | ||
e94b1766 | 354 | ei = kmem_cache_alloc(mqueue_inode_cachep, GFP_KERNEL); |
1da177e4 LT |
355 | if (!ei) |
356 | return NULL; | |
357 | return &ei->vfs_inode; | |
358 | } | |
359 | ||
fa0d7e3d | 360 | static void mqueue_i_callback(struct rcu_head *head) |
1da177e4 | 361 | { |
fa0d7e3d | 362 | struct inode *inode = container_of(head, struct inode, i_rcu); |
1da177e4 LT |
363 | kmem_cache_free(mqueue_inode_cachep, MQUEUE_I(inode)); |
364 | } | |
365 | ||
fa0d7e3d NP |
366 | static void mqueue_destroy_inode(struct inode *inode) |
367 | { | |
368 | call_rcu(&inode->i_rcu, mqueue_i_callback); | |
369 | } | |
370 | ||
6d8af64c | 371 | static void mqueue_evict_inode(struct inode *inode) |
1da177e4 LT |
372 | { |
373 | struct mqueue_inode_info *info; | |
374 | struct user_struct *user; | |
7eafd7c7 | 375 | struct ipc_namespace *ipc_ns; |
52457b64 LR |
376 | struct msg_msg *msg, *nmsg; |
377 | LIST_HEAD(tmp_msg); | |
1da177e4 | 378 | |
dbd5768f | 379 | clear_inode(inode); |
6d8af64c AV |
380 | |
381 | if (S_ISDIR(inode->i_mode)) | |
1da177e4 | 382 | return; |
6d8af64c | 383 | |
7eafd7c7 | 384 | ipc_ns = get_ns_from_inode(inode); |
1da177e4 LT |
385 | info = MQUEUE_I(inode); |
386 | spin_lock(&info->lock); | |
d6629859 | 387 | while ((msg = msg_get(info)) != NULL) |
52457b64 | 388 | list_add_tail(&msg->m_list, &tmp_msg); |
ce2d52cc | 389 | kfree(info->node_cache); |
1da177e4 LT |
390 | spin_unlock(&info->lock); |
391 | ||
52457b64 LR |
392 | list_for_each_entry_safe(msg, nmsg, &tmp_msg, m_list) { |
393 | list_del(&msg->m_list); | |
394 | free_msg(msg); | |
395 | } | |
396 | ||
1da177e4 LT |
397 | user = info->user; |
398 | if (user) { | |
71379582 KC |
399 | unsigned long mq_bytes, mq_treesize; |
400 | ||
401 | /* Total amount of bytes accounted for the mqueue */ | |
402 | mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) + | |
403 | min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) * | |
404 | sizeof(struct posix_msg_tree_node); | |
405 | ||
406 | mq_bytes = mq_treesize + (info->attr.mq_maxmsg * | |
407 | info->attr.mq_msgsize); | |
408 | ||
1da177e4 LT |
409 | spin_lock(&mq_lock); |
410 | user->mq_bytes -= mq_bytes; | |
7eafd7c7 SH |
411 | /* |
412 | * get_ns_from_inode() ensures that the | |
413 | * (ipc_ns = sb->s_fs_info) is either a valid ipc_ns | |
414 | * to which we now hold a reference, or it is NULL. | |
415 | * We can't put it here under mq_lock, though. | |
416 | */ | |
417 | if (ipc_ns) | |
418 | ipc_ns->mq_queues_count--; | |
1da177e4 LT |
419 | spin_unlock(&mq_lock); |
420 | free_uid(user); | |
421 | } | |
7eafd7c7 SH |
422 | if (ipc_ns) |
423 | put_ipc_ns(ipc_ns); | |
1da177e4 LT |
424 | } |
425 | ||
426 | static int mqueue_create(struct inode *dir, struct dentry *dentry, | |
ebfc3b49 | 427 | umode_t mode, bool excl) |
1da177e4 LT |
428 | { |
429 | struct inode *inode; | |
430 | struct mq_attr *attr = dentry->d_fsdata; | |
431 | int error; | |
7eafd7c7 | 432 | struct ipc_namespace *ipc_ns; |
1da177e4 LT |
433 | |
434 | spin_lock(&mq_lock); | |
7eafd7c7 SH |
435 | ipc_ns = __get_ns_from_inode(dir); |
436 | if (!ipc_ns) { | |
437 | error = -EACCES; | |
438 | goto out_unlock; | |
439 | } | |
f3713fd9 DB |
440 | |
441 | if (ipc_ns->mq_queues_count >= ipc_ns->mq_queues_max && | |
442 | !capable(CAP_SYS_RESOURCE)) { | |
1da177e4 | 443 | error = -ENOSPC; |
614b84cf | 444 | goto out_unlock; |
1da177e4 | 445 | } |
614b84cf | 446 | ipc_ns->mq_queues_count++; |
1da177e4 LT |
447 | spin_unlock(&mq_lock); |
448 | ||
7eafd7c7 | 449 | inode = mqueue_get_inode(dir->i_sb, ipc_ns, mode, attr); |
d40dcdb0 JS |
450 | if (IS_ERR(inode)) { |
451 | error = PTR_ERR(inode); | |
1da177e4 | 452 | spin_lock(&mq_lock); |
614b84cf SH |
453 | ipc_ns->mq_queues_count--; |
454 | goto out_unlock; | |
1da177e4 LT |
455 | } |
456 | ||
7eafd7c7 | 457 | put_ipc_ns(ipc_ns); |
1da177e4 | 458 | dir->i_size += DIRENT_SIZE; |
078cd827 | 459 | dir->i_ctime = dir->i_mtime = dir->i_atime = current_time(dir); |
1da177e4 LT |
460 | |
461 | d_instantiate(dentry, inode); | |
462 | dget(dentry); | |
463 | return 0; | |
614b84cf | 464 | out_unlock: |
1da177e4 | 465 | spin_unlock(&mq_lock); |
7eafd7c7 SH |
466 | if (ipc_ns) |
467 | put_ipc_ns(ipc_ns); | |
1da177e4 LT |
468 | return error; |
469 | } | |
470 | ||
471 | static int mqueue_unlink(struct inode *dir, struct dentry *dentry) | |
472 | { | |
75c3cfa8 | 473 | struct inode *inode = d_inode(dentry); |
1da177e4 | 474 | |
078cd827 | 475 | dir->i_ctime = dir->i_mtime = dir->i_atime = current_time(dir); |
1da177e4 | 476 | dir->i_size -= DIRENT_SIZE; |
239521f3 MS |
477 | drop_nlink(inode); |
478 | dput(dentry); | |
479 | return 0; | |
1da177e4 LT |
480 | } |
481 | ||
482 | /* | |
483 | * This is routine for system read from queue file. | |
484 | * To avoid mess with doing here some sort of mq_receive we allow | |
485 | * to read only queue size & notification info (the only values | |
486 | * that are interesting from user point of view and aren't accessible | |
487 | * through std routines) | |
488 | */ | |
489 | static ssize_t mqueue_read_file(struct file *filp, char __user *u_data, | |
f1a43f93 | 490 | size_t count, loff_t *off) |
1da177e4 | 491 | { |
496ad9aa | 492 | struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp)); |
1da177e4 | 493 | char buffer[FILENT_SIZE]; |
f1a43f93 | 494 | ssize_t ret; |
1da177e4 LT |
495 | |
496 | spin_lock(&info->lock); | |
497 | snprintf(buffer, sizeof(buffer), | |
498 | "QSIZE:%-10lu NOTIFY:%-5d SIGNO:%-5d NOTIFY_PID:%-6d\n", | |
499 | info->qsize, | |
500 | info->notify_owner ? info->notify.sigev_notify : 0, | |
501 | (info->notify_owner && | |
502 | info->notify.sigev_notify == SIGEV_SIGNAL) ? | |
503 | info->notify.sigev_signo : 0, | |
6c5f3e7b | 504 | pid_vnr(info->notify_owner)); |
1da177e4 LT |
505 | spin_unlock(&info->lock); |
506 | buffer[sizeof(buffer)-1] = '\0'; | |
1da177e4 | 507 | |
f1a43f93 AM |
508 | ret = simple_read_from_buffer(u_data, count, off, buffer, |
509 | strlen(buffer)); | |
510 | if (ret <= 0) | |
511 | return ret; | |
1da177e4 | 512 | |
078cd827 | 513 | file_inode(filp)->i_atime = file_inode(filp)->i_ctime = current_time(file_inode(filp)); |
f1a43f93 | 514 | return ret; |
1da177e4 LT |
515 | } |
516 | ||
75e1fcc0 | 517 | static int mqueue_flush_file(struct file *filp, fl_owner_t id) |
1da177e4 | 518 | { |
496ad9aa | 519 | struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp)); |
1da177e4 LT |
520 | |
521 | spin_lock(&info->lock); | |
a03fcb73 | 522 | if (task_tgid(current) == info->notify_owner) |
1da177e4 LT |
523 | remove_notification(info); |
524 | ||
525 | spin_unlock(&info->lock); | |
526 | return 0; | |
527 | } | |
528 | ||
529 | static unsigned int mqueue_poll_file(struct file *filp, struct poll_table_struct *poll_tab) | |
530 | { | |
496ad9aa | 531 | struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp)); |
1da177e4 LT |
532 | int retval = 0; |
533 | ||
534 | poll_wait(filp, &info->wait_q, poll_tab); | |
535 | ||
536 | spin_lock(&info->lock); | |
537 | if (info->attr.mq_curmsgs) | |
538 | retval = POLLIN | POLLRDNORM; | |
539 | ||
540 | if (info->attr.mq_curmsgs < info->attr.mq_maxmsg) | |
541 | retval |= POLLOUT | POLLWRNORM; | |
542 | spin_unlock(&info->lock); | |
543 | ||
544 | return retval; | |
545 | } | |
546 | ||
547 | /* Adds current to info->e_wait_q[sr] before element with smaller prio */ | |
548 | static void wq_add(struct mqueue_inode_info *info, int sr, | |
549 | struct ext_wait_queue *ewp) | |
550 | { | |
551 | struct ext_wait_queue *walk; | |
552 | ||
553 | ewp->task = current; | |
554 | ||
555 | list_for_each_entry(walk, &info->e_wait_q[sr].list, list) { | |
556 | if (walk->task->static_prio <= current->static_prio) { | |
557 | list_add_tail(&ewp->list, &walk->list); | |
558 | return; | |
559 | } | |
560 | } | |
561 | list_add_tail(&ewp->list, &info->e_wait_q[sr].list); | |
562 | } | |
563 | ||
564 | /* | |
565 | * Puts current task to sleep. Caller must hold queue lock. After return | |
566 | * lock isn't held. | |
567 | * sr: SEND or RECV | |
568 | */ | |
569 | static int wq_sleep(struct mqueue_inode_info *info, int sr, | |
9ca7d8e6 | 570 | ktime_t *timeout, struct ext_wait_queue *ewp) |
eac0b1c3 | 571 | __releases(&info->lock) |
1da177e4 LT |
572 | { |
573 | int retval; | |
574 | signed long time; | |
575 | ||
576 | wq_add(info, sr, ewp); | |
577 | ||
578 | for (;;) { | |
fa6004ad | 579 | __set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 LT |
580 | |
581 | spin_unlock(&info->lock); | |
32ea845d WG |
582 | time = schedule_hrtimeout_range_clock(timeout, 0, |
583 | HRTIMER_MODE_ABS, CLOCK_REALTIME); | |
1da177e4 | 584 | |
1da177e4 LT |
585 | if (ewp->state == STATE_READY) { |
586 | retval = 0; | |
587 | goto out; | |
588 | } | |
589 | spin_lock(&info->lock); | |
590 | if (ewp->state == STATE_READY) { | |
591 | retval = 0; | |
592 | goto out_unlock; | |
593 | } | |
594 | if (signal_pending(current)) { | |
595 | retval = -ERESTARTSYS; | |
596 | break; | |
597 | } | |
598 | if (time == 0) { | |
599 | retval = -ETIMEDOUT; | |
600 | break; | |
601 | } | |
602 | } | |
603 | list_del(&ewp->list); | |
604 | out_unlock: | |
605 | spin_unlock(&info->lock); | |
606 | out: | |
607 | return retval; | |
608 | } | |
609 | ||
610 | /* | |
611 | * Returns waiting task that should be serviced first or NULL if none exists | |
612 | */ | |
613 | static struct ext_wait_queue *wq_get_first_waiter( | |
614 | struct mqueue_inode_info *info, int sr) | |
615 | { | |
616 | struct list_head *ptr; | |
617 | ||
618 | ptr = info->e_wait_q[sr].list.prev; | |
619 | if (ptr == &info->e_wait_q[sr].list) | |
620 | return NULL; | |
621 | return list_entry(ptr, struct ext_wait_queue, list); | |
622 | } | |
623 | ||
1da177e4 LT |
624 | |
625 | static inline void set_cookie(struct sk_buff *skb, char code) | |
626 | { | |
239521f3 | 627 | ((char *)skb->data)[NOTIFY_COOKIE_LEN-1] = code; |
1da177e4 LT |
628 | } |
629 | ||
630 | /* | |
631 | * The next function is only to split too long sys_mq_timedsend | |
632 | */ | |
633 | static void __do_notify(struct mqueue_inode_info *info) | |
634 | { | |
635 | /* notification | |
636 | * invoked when there is registered process and there isn't process | |
637 | * waiting synchronously for message AND state of queue changed from | |
638 | * empty to not empty. Here we are sure that no one is waiting | |
639 | * synchronously. */ | |
640 | if (info->notify_owner && | |
641 | info->attr.mq_curmsgs == 1) { | |
642 | struct siginfo sig_i; | |
643 | switch (info->notify.sigev_notify) { | |
644 | case SIGEV_NONE: | |
645 | break; | |
646 | case SIGEV_SIGNAL: | |
647 | /* sends signal */ | |
648 | ||
649 | sig_i.si_signo = info->notify.sigev_signo; | |
650 | sig_i.si_errno = 0; | |
651 | sig_i.si_code = SI_MESGQ; | |
652 | sig_i.si_value = info->notify.sigev_value; | |
6b550f94 SH |
653 | /* map current pid/uid into info->owner's namespaces */ |
654 | rcu_read_lock(); | |
a6684999 SB |
655 | sig_i.si_pid = task_tgid_nr_ns(current, |
656 | ns_of_pid(info->notify_owner)); | |
76b6db01 | 657 | sig_i.si_uid = from_kuid_munged(info->notify_user_ns, current_uid()); |
6b550f94 | 658 | rcu_read_unlock(); |
1da177e4 | 659 | |
a03fcb73 CLG |
660 | kill_pid_info(info->notify.sigev_signo, |
661 | &sig_i, info->notify_owner); | |
1da177e4 LT |
662 | break; |
663 | case SIGEV_THREAD: | |
664 | set_cookie(info->notify_cookie, NOTIFY_WOKENUP); | |
7ee015e0 | 665 | netlink_sendskb(info->notify_sock, info->notify_cookie); |
1da177e4 LT |
666 | break; |
667 | } | |
668 | /* after notification unregisters process */ | |
a03fcb73 | 669 | put_pid(info->notify_owner); |
6f9ac6d9 | 670 | put_user_ns(info->notify_user_ns); |
a03fcb73 | 671 | info->notify_owner = NULL; |
6f9ac6d9 | 672 | info->notify_user_ns = NULL; |
1da177e4 LT |
673 | } |
674 | wake_up(&info->wait_q); | |
675 | } | |
676 | ||
9ca7d8e6 | 677 | static int prepare_timeout(const struct timespec __user *u_abs_timeout, |
b9047726 | 678 | struct timespec64 *ts) |
1da177e4 | 679 | { |
b9047726 | 680 | if (get_timespec64(ts, u_abs_timeout)) |
9ca7d8e6 | 681 | return -EFAULT; |
b9047726 | 682 | if (!timespec64_valid(ts)) |
9ca7d8e6 | 683 | return -EINVAL; |
9ca7d8e6 | 684 | return 0; |
1da177e4 LT |
685 | } |
686 | ||
687 | static void remove_notification(struct mqueue_inode_info *info) | |
688 | { | |
a03fcb73 | 689 | if (info->notify_owner != NULL && |
1da177e4 LT |
690 | info->notify.sigev_notify == SIGEV_THREAD) { |
691 | set_cookie(info->notify_cookie, NOTIFY_REMOVED); | |
7ee015e0 | 692 | netlink_sendskb(info->notify_sock, info->notify_cookie); |
1da177e4 | 693 | } |
a03fcb73 | 694 | put_pid(info->notify_owner); |
6f9ac6d9 | 695 | put_user_ns(info->notify_user_ns); |
a03fcb73 | 696 | info->notify_owner = NULL; |
6f9ac6d9 | 697 | info->notify_user_ns = NULL; |
1da177e4 LT |
698 | } |
699 | ||
614b84cf | 700 | static int mq_attr_ok(struct ipc_namespace *ipc_ns, struct mq_attr *attr) |
1da177e4 | 701 | { |
2c12ea49 DL |
702 | int mq_treesize; |
703 | unsigned long total_size; | |
704 | ||
1da177e4 | 705 | if (attr->mq_maxmsg <= 0 || attr->mq_msgsize <= 0) |
113289cc | 706 | return -EINVAL; |
1da177e4 | 707 | if (capable(CAP_SYS_RESOURCE)) { |
02967ea0 DL |
708 | if (attr->mq_maxmsg > HARD_MSGMAX || |
709 | attr->mq_msgsize > HARD_MSGSIZEMAX) | |
113289cc | 710 | return -EINVAL; |
1da177e4 | 711 | } else { |
614b84cf SH |
712 | if (attr->mq_maxmsg > ipc_ns->mq_msg_max || |
713 | attr->mq_msgsize > ipc_ns->mq_msgsize_max) | |
113289cc | 714 | return -EINVAL; |
1da177e4 LT |
715 | } |
716 | /* check for overflow */ | |
717 | if (attr->mq_msgsize > ULONG_MAX/attr->mq_maxmsg) | |
113289cc | 718 | return -EOVERFLOW; |
2c12ea49 DL |
719 | mq_treesize = attr->mq_maxmsg * sizeof(struct msg_msg) + |
720 | min_t(unsigned int, attr->mq_maxmsg, MQ_PRIO_MAX) * | |
721 | sizeof(struct posix_msg_tree_node); | |
722 | total_size = attr->mq_maxmsg * attr->mq_msgsize; | |
723 | if (total_size + mq_treesize < total_size) | |
113289cc DL |
724 | return -EOVERFLOW; |
725 | return 0; | |
1da177e4 LT |
726 | } |
727 | ||
728 | /* | |
729 | * Invoked when creating a new queue via sys_mq_open | |
730 | */ | |
765927b2 AV |
731 | static struct file *do_create(struct ipc_namespace *ipc_ns, struct inode *dir, |
732 | struct path *path, int oflag, umode_t mode, | |
614b84cf | 733 | struct mq_attr *attr) |
1da177e4 | 734 | { |
745ca247 | 735 | const struct cred *cred = current_cred(); |
1da177e4 LT |
736 | int ret; |
737 | ||
564f6993 | 738 | if (attr) { |
113289cc DL |
739 | ret = mq_attr_ok(ipc_ns, attr); |
740 | if (ret) | |
765927b2 | 741 | return ERR_PTR(ret); |
1da177e4 | 742 | /* store for use during create */ |
765927b2 | 743 | path->dentry->d_fsdata = attr; |
113289cc DL |
744 | } else { |
745 | struct mq_attr def_attr; | |
746 | ||
747 | def_attr.mq_maxmsg = min(ipc_ns->mq_msg_max, | |
748 | ipc_ns->mq_msg_default); | |
749 | def_attr.mq_msgsize = min(ipc_ns->mq_msgsize_max, | |
750 | ipc_ns->mq_msgsize_default); | |
751 | ret = mq_attr_ok(ipc_ns, &def_attr); | |
752 | if (ret) | |
765927b2 | 753 | return ERR_PTR(ret); |
1da177e4 LT |
754 | } |
755 | ||
ce3b0f8d | 756 | mode &= ~current_umask(); |
765927b2 AV |
757 | ret = vfs_create(dir, path->dentry, mode, true); |
758 | path->dentry->d_fsdata = NULL; | |
312b90fb AV |
759 | if (ret) |
760 | return ERR_PTR(ret); | |
761 | return dentry_open(path, oflag, cred); | |
1da177e4 LT |
762 | } |
763 | ||
764 | /* Opens existing queue */ | |
765927b2 | 765 | static struct file *do_open(struct path *path, int oflag) |
1da177e4 | 766 | { |
745ca247 DH |
767 | static const int oflag2acc[O_ACCMODE] = { MAY_READ, MAY_WRITE, |
768 | MAY_READ | MAY_WRITE }; | |
765927b2 AV |
769 | int acc; |
770 | if ((oflag & O_ACCMODE) == (O_RDWR | O_WRONLY)) | |
771 | return ERR_PTR(-EINVAL); | |
772 | acc = oflag2acc[oflag & O_ACCMODE]; | |
75c3cfa8 | 773 | if (inode_permission(d_inode(path->dentry), acc)) |
765927b2 AV |
774 | return ERR_PTR(-EACCES); |
775 | return dentry_open(path, oflag, current_cred()); | |
1da177e4 LT |
776 | } |
777 | ||
0d060606 AV |
778 | static int do_mq_open(const char __user *u_name, int oflag, umode_t mode, |
779 | struct mq_attr *attr) | |
1da177e4 | 780 | { |
765927b2 | 781 | struct path path; |
1da177e4 | 782 | struct file *filp; |
91a27b2a | 783 | struct filename *name; |
1da177e4 | 784 | int fd, error; |
7eafd7c7 | 785 | struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns; |
312b90fb AV |
786 | struct vfsmount *mnt = ipc_ns->mq_mnt; |
787 | struct dentry *root = mnt->mnt_root; | |
788 | int ro; | |
1da177e4 | 789 | |
0d060606 | 790 | audit_mq_open(oflag, mode, attr); |
20ca73bc | 791 | |
1da177e4 LT |
792 | if (IS_ERR(name = getname(u_name))) |
793 | return PTR_ERR(name); | |
794 | ||
269f2134 | 795 | fd = get_unused_fd_flags(O_CLOEXEC); |
1da177e4 LT |
796 | if (fd < 0) |
797 | goto out_putname; | |
798 | ||
312b90fb | 799 | ro = mnt_want_write(mnt); /* we'll drop it in any case */ |
765927b2 | 800 | error = 0; |
5955102c | 801 | inode_lock(d_inode(root)); |
91a27b2a | 802 | path.dentry = lookup_one_len(name->name, root, strlen(name->name)); |
765927b2 AV |
803 | if (IS_ERR(path.dentry)) { |
804 | error = PTR_ERR(path.dentry); | |
4294a8ee | 805 | goto out_putfd; |
1da177e4 | 806 | } |
312b90fb | 807 | path.mnt = mntget(mnt); |
1da177e4 LT |
808 | |
809 | if (oflag & O_CREAT) { | |
75c3cfa8 | 810 | if (d_really_is_positive(path.dentry)) { /* entry already exists */ |
adb5c247 | 811 | audit_inode(name, path.dentry, 0); |
8d8ffefa AGR |
812 | if (oflag & O_EXCL) { |
813 | error = -EEXIST; | |
7c7dce92 | 814 | goto out; |
8d8ffefa | 815 | } |
765927b2 | 816 | filp = do_open(&path, oflag); |
1da177e4 | 817 | } else { |
312b90fb AV |
818 | if (ro) { |
819 | error = ro; | |
820 | goto out; | |
821 | } | |
79f6530c | 822 | audit_inode_parent_hidden(name, root); |
0d060606 AV |
823 | filp = do_create(ipc_ns, d_inode(root), &path, |
824 | oflag, mode, attr); | |
1da177e4 | 825 | } |
7c7dce92 | 826 | } else { |
75c3cfa8 | 827 | if (d_really_is_negative(path.dentry)) { |
8d8ffefa | 828 | error = -ENOENT; |
7c7dce92 | 829 | goto out; |
8d8ffefa | 830 | } |
adb5c247 | 831 | audit_inode(name, path.dentry, 0); |
765927b2 | 832 | filp = do_open(&path, oflag); |
7c7dce92 | 833 | } |
1da177e4 | 834 | |
765927b2 AV |
835 | if (!IS_ERR(filp)) |
836 | fd_install(fd, filp); | |
837 | else | |
1da177e4 | 838 | error = PTR_ERR(filp); |
7c7dce92 | 839 | out: |
765927b2 | 840 | path_put(&path); |
7c7dce92 | 841 | out_putfd: |
765927b2 AV |
842 | if (error) { |
843 | put_unused_fd(fd); | |
844 | fd = error; | |
845 | } | |
5955102c | 846 | inode_unlock(d_inode(root)); |
38d78e58 VD |
847 | if (!ro) |
848 | mnt_drop_write(mnt); | |
1da177e4 LT |
849 | out_putname: |
850 | putname(name); | |
851 | return fd; | |
852 | } | |
853 | ||
0d060606 AV |
854 | SYSCALL_DEFINE4(mq_open, const char __user *, u_name, int, oflag, umode_t, mode, |
855 | struct mq_attr __user *, u_attr) | |
856 | { | |
857 | struct mq_attr attr; | |
858 | if (u_attr && copy_from_user(&attr, u_attr, sizeof(struct mq_attr))) | |
859 | return -EFAULT; | |
860 | ||
861 | return do_mq_open(u_name, oflag, mode, u_attr ? &attr : NULL); | |
862 | } | |
863 | ||
d5460c99 | 864 | SYSCALL_DEFINE1(mq_unlink, const char __user *, u_name) |
1da177e4 LT |
865 | { |
866 | int err; | |
91a27b2a | 867 | struct filename *name; |
1da177e4 LT |
868 | struct dentry *dentry; |
869 | struct inode *inode = NULL; | |
7eafd7c7 | 870 | struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns; |
312b90fb | 871 | struct vfsmount *mnt = ipc_ns->mq_mnt; |
1da177e4 LT |
872 | |
873 | name = getname(u_name); | |
874 | if (IS_ERR(name)) | |
875 | return PTR_ERR(name); | |
876 | ||
79f6530c | 877 | audit_inode_parent_hidden(name, mnt->mnt_root); |
312b90fb AV |
878 | err = mnt_want_write(mnt); |
879 | if (err) | |
880 | goto out_name; | |
5955102c | 881 | inode_lock_nested(d_inode(mnt->mnt_root), I_MUTEX_PARENT); |
91a27b2a JL |
882 | dentry = lookup_one_len(name->name, mnt->mnt_root, |
883 | strlen(name->name)); | |
1da177e4 LT |
884 | if (IS_ERR(dentry)) { |
885 | err = PTR_ERR(dentry); | |
886 | goto out_unlock; | |
887 | } | |
888 | ||
75c3cfa8 | 889 | inode = d_inode(dentry); |
312b90fb AV |
890 | if (!inode) { |
891 | err = -ENOENT; | |
892 | } else { | |
7de9c6ee | 893 | ihold(inode); |
75c3cfa8 | 894 | err = vfs_unlink(d_inode(dentry->d_parent), dentry, NULL); |
312b90fb | 895 | } |
1da177e4 LT |
896 | dput(dentry); |
897 | ||
898 | out_unlock: | |
5955102c | 899 | inode_unlock(d_inode(mnt->mnt_root)); |
1da177e4 LT |
900 | if (inode) |
901 | iput(inode); | |
312b90fb AV |
902 | mnt_drop_write(mnt); |
903 | out_name: | |
904 | putname(name); | |
1da177e4 LT |
905 | |
906 | return err; | |
907 | } | |
908 | ||
909 | /* Pipelined send and receive functions. | |
910 | * | |
911 | * If a receiver finds no waiting message, then it registers itself in the | |
912 | * list of waiting receivers. A sender checks that list before adding the new | |
913 | * message into the message array. If there is a waiting receiver, then it | |
914 | * bypasses the message array and directly hands the message over to the | |
fa6004ad DB |
915 | * receiver. The receiver accepts the message and returns without grabbing the |
916 | * queue spinlock: | |
917 | * | |
918 | * - Set pointer to message. | |
919 | * - Queue the receiver task for later wakeup (without the info->lock). | |
920 | * - Update its state to STATE_READY. Now the receiver can continue. | |
921 | * - Wake up the process after the lock is dropped. Should the process wake up | |
922 | * before this wakeup (due to a timeout or a signal) it will either see | |
923 | * STATE_READY and continue or acquire the lock to check the state again. | |
1da177e4 LT |
924 | * |
925 | * The same algorithm is used for senders. | |
926 | */ | |
927 | ||
928 | /* pipelined_send() - send a message directly to the task waiting in | |
929 | * sys_mq_timedreceive() (without inserting message into a queue). | |
930 | */ | |
fa6004ad DB |
931 | static inline void pipelined_send(struct wake_q_head *wake_q, |
932 | struct mqueue_inode_info *info, | |
1da177e4 LT |
933 | struct msg_msg *message, |
934 | struct ext_wait_queue *receiver) | |
935 | { | |
936 | receiver->msg = message; | |
937 | list_del(&receiver->list); | |
fa6004ad DB |
938 | wake_q_add(wake_q, receiver->task); |
939 | /* | |
940 | * Rely on the implicit cmpxchg barrier from wake_q_add such | |
941 | * that we can ensure that updating receiver->state is the last | |
942 | * write operation: As once set, the receiver can continue, | |
943 | * and if we don't have the reference count from the wake_q, | |
944 | * yet, at that point we can later have a use-after-free | |
945 | * condition and bogus wakeup. | |
946 | */ | |
1da177e4 LT |
947 | receiver->state = STATE_READY; |
948 | } | |
949 | ||
950 | /* pipelined_receive() - if there is task waiting in sys_mq_timedsend() | |
951 | * gets its message and put to the queue (we have one free place for sure). */ | |
fa6004ad DB |
952 | static inline void pipelined_receive(struct wake_q_head *wake_q, |
953 | struct mqueue_inode_info *info) | |
1da177e4 LT |
954 | { |
955 | struct ext_wait_queue *sender = wq_get_first_waiter(info, SEND); | |
956 | ||
957 | if (!sender) { | |
958 | /* for poll */ | |
959 | wake_up_interruptible(&info->wait_q); | |
960 | return; | |
961 | } | |
d6629859 DL |
962 | if (msg_insert(sender->msg, info)) |
963 | return; | |
fa6004ad | 964 | |
1da177e4 | 965 | list_del(&sender->list); |
fa6004ad | 966 | wake_q_add(wake_q, sender->task); |
1da177e4 LT |
967 | sender->state = STATE_READY; |
968 | } | |
969 | ||
0d060606 AV |
970 | static int do_mq_timedsend(mqd_t mqdes, const char __user *u_msg_ptr, |
971 | size_t msg_len, unsigned int msg_prio, | |
b9047726 | 972 | struct timespec64 *ts) |
1da177e4 | 973 | { |
2903ff01 | 974 | struct fd f; |
1da177e4 LT |
975 | struct inode *inode; |
976 | struct ext_wait_queue wait; | |
977 | struct ext_wait_queue *receiver; | |
978 | struct msg_msg *msg_ptr; | |
979 | struct mqueue_inode_info *info; | |
9ca7d8e6 | 980 | ktime_t expires, *timeout = NULL; |
ce2d52cc | 981 | struct posix_msg_tree_node *new_leaf = NULL; |
2903ff01 | 982 | int ret = 0; |
194a6b5b | 983 | DEFINE_WAKE_Q(wake_q); |
1da177e4 LT |
984 | |
985 | if (unlikely(msg_prio >= (unsigned long) MQ_PRIO_MAX)) | |
986 | return -EINVAL; | |
987 | ||
0d060606 | 988 | if (ts) { |
b9047726 | 989 | expires = timespec64_to_ktime(*ts); |
0d060606 AV |
990 | timeout = &expires; |
991 | } | |
992 | ||
993 | audit_mq_sendrecv(mqdes, msg_len, msg_prio, ts); | |
1da177e4 | 994 | |
2903ff01 AV |
995 | f = fdget(mqdes); |
996 | if (unlikely(!f.file)) { | |
8d8ffefa | 997 | ret = -EBADF; |
1da177e4 | 998 | goto out; |
8d8ffefa | 999 | } |
1da177e4 | 1000 | |
496ad9aa | 1001 | inode = file_inode(f.file); |
2903ff01 | 1002 | if (unlikely(f.file->f_op != &mqueue_file_operations)) { |
8d8ffefa | 1003 | ret = -EBADF; |
1da177e4 | 1004 | goto out_fput; |
8d8ffefa | 1005 | } |
1da177e4 | 1006 | info = MQUEUE_I(inode); |
9f45f5bf | 1007 | audit_file(f.file); |
1da177e4 | 1008 | |
2903ff01 | 1009 | if (unlikely(!(f.file->f_mode & FMODE_WRITE))) { |
8d8ffefa | 1010 | ret = -EBADF; |
1da177e4 | 1011 | goto out_fput; |
8d8ffefa | 1012 | } |
1da177e4 LT |
1013 | |
1014 | if (unlikely(msg_len > info->attr.mq_msgsize)) { | |
1015 | ret = -EMSGSIZE; | |
1016 | goto out_fput; | |
1017 | } | |
1018 | ||
1019 | /* First try to allocate memory, before doing anything with | |
1020 | * existing queues. */ | |
1021 | msg_ptr = load_msg(u_msg_ptr, msg_len); | |
1022 | if (IS_ERR(msg_ptr)) { | |
1023 | ret = PTR_ERR(msg_ptr); | |
1024 | goto out_fput; | |
1025 | } | |
1026 | msg_ptr->m_ts = msg_len; | |
1027 | msg_ptr->m_type = msg_prio; | |
1028 | ||
ce2d52cc DL |
1029 | /* |
1030 | * msg_insert really wants us to have a valid, spare node struct so | |
1031 | * it doesn't have to kmalloc a GFP_ATOMIC allocation, but it will | |
1032 | * fall back to that if necessary. | |
1033 | */ | |
1034 | if (!info->node_cache) | |
1035 | new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL); | |
1036 | ||
1da177e4 LT |
1037 | spin_lock(&info->lock); |
1038 | ||
ce2d52cc DL |
1039 | if (!info->node_cache && new_leaf) { |
1040 | /* Save our speculative allocation into the cache */ | |
ce2d52cc DL |
1041 | INIT_LIST_HEAD(&new_leaf->msg_list); |
1042 | info->node_cache = new_leaf; | |
ce2d52cc DL |
1043 | new_leaf = NULL; |
1044 | } else { | |
1045 | kfree(new_leaf); | |
1046 | } | |
1047 | ||
1da177e4 | 1048 | if (info->attr.mq_curmsgs == info->attr.mq_maxmsg) { |
2903ff01 | 1049 | if (f.file->f_flags & O_NONBLOCK) { |
1da177e4 | 1050 | ret = -EAGAIN; |
1da177e4 LT |
1051 | } else { |
1052 | wait.task = current; | |
1053 | wait.msg = (void *) msg_ptr; | |
1054 | wait.state = STATE_NONE; | |
1055 | ret = wq_sleep(info, SEND, timeout, &wait); | |
ce2d52cc DL |
1056 | /* |
1057 | * wq_sleep must be called with info->lock held, and | |
1058 | * returns with the lock released | |
1059 | */ | |
1060 | goto out_free; | |
1da177e4 | 1061 | } |
1da177e4 LT |
1062 | } else { |
1063 | receiver = wq_get_first_waiter(info, RECV); | |
1064 | if (receiver) { | |
fa6004ad | 1065 | pipelined_send(&wake_q, info, msg_ptr, receiver); |
1da177e4 LT |
1066 | } else { |
1067 | /* adds message to the queue */ | |
ce2d52cc DL |
1068 | ret = msg_insert(msg_ptr, info); |
1069 | if (ret) | |
1070 | goto out_unlock; | |
1da177e4 LT |
1071 | __do_notify(info); |
1072 | } | |
1073 | inode->i_atime = inode->i_mtime = inode->i_ctime = | |
078cd827 | 1074 | current_time(inode); |
1da177e4 | 1075 | } |
ce2d52cc DL |
1076 | out_unlock: |
1077 | spin_unlock(&info->lock); | |
fa6004ad | 1078 | wake_up_q(&wake_q); |
ce2d52cc DL |
1079 | out_free: |
1080 | if (ret) | |
1081 | free_msg(msg_ptr); | |
1da177e4 | 1082 | out_fput: |
2903ff01 | 1083 | fdput(f); |
1da177e4 LT |
1084 | out: |
1085 | return ret; | |
1086 | } | |
1087 | ||
0d060606 AV |
1088 | static int do_mq_timedreceive(mqd_t mqdes, char __user *u_msg_ptr, |
1089 | size_t msg_len, unsigned int __user *u_msg_prio, | |
b9047726 | 1090 | struct timespec64 *ts) |
1da177e4 | 1091 | { |
1da177e4 LT |
1092 | ssize_t ret; |
1093 | struct msg_msg *msg_ptr; | |
2903ff01 | 1094 | struct fd f; |
1da177e4 LT |
1095 | struct inode *inode; |
1096 | struct mqueue_inode_info *info; | |
1097 | struct ext_wait_queue wait; | |
9ca7d8e6 | 1098 | ktime_t expires, *timeout = NULL; |
ce2d52cc | 1099 | struct posix_msg_tree_node *new_leaf = NULL; |
1da177e4 | 1100 | |
0d060606 | 1101 | if (ts) { |
b9047726 | 1102 | expires = timespec64_to_ktime(*ts); |
9ca7d8e6 | 1103 | timeout = &expires; |
c32c8af4 | 1104 | } |
20ca73bc | 1105 | |
0d060606 | 1106 | audit_mq_sendrecv(mqdes, msg_len, 0, ts); |
1da177e4 | 1107 | |
2903ff01 AV |
1108 | f = fdget(mqdes); |
1109 | if (unlikely(!f.file)) { | |
8d8ffefa | 1110 | ret = -EBADF; |
1da177e4 | 1111 | goto out; |
8d8ffefa | 1112 | } |
1da177e4 | 1113 | |
496ad9aa | 1114 | inode = file_inode(f.file); |
2903ff01 | 1115 | if (unlikely(f.file->f_op != &mqueue_file_operations)) { |
8d8ffefa | 1116 | ret = -EBADF; |
1da177e4 | 1117 | goto out_fput; |
8d8ffefa | 1118 | } |
1da177e4 | 1119 | info = MQUEUE_I(inode); |
9f45f5bf | 1120 | audit_file(f.file); |
1da177e4 | 1121 | |
2903ff01 | 1122 | if (unlikely(!(f.file->f_mode & FMODE_READ))) { |
8d8ffefa | 1123 | ret = -EBADF; |
1da177e4 | 1124 | goto out_fput; |
8d8ffefa | 1125 | } |
1da177e4 LT |
1126 | |
1127 | /* checks if buffer is big enough */ | |
1128 | if (unlikely(msg_len < info->attr.mq_msgsize)) { | |
1129 | ret = -EMSGSIZE; | |
1130 | goto out_fput; | |
1131 | } | |
1132 | ||
ce2d52cc DL |
1133 | /* |
1134 | * msg_insert really wants us to have a valid, spare node struct so | |
1135 | * it doesn't have to kmalloc a GFP_ATOMIC allocation, but it will | |
1136 | * fall back to that if necessary. | |
1137 | */ | |
1138 | if (!info->node_cache) | |
1139 | new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL); | |
1140 | ||
1da177e4 | 1141 | spin_lock(&info->lock); |
ce2d52cc DL |
1142 | |
1143 | if (!info->node_cache && new_leaf) { | |
1144 | /* Save our speculative allocation into the cache */ | |
ce2d52cc DL |
1145 | INIT_LIST_HEAD(&new_leaf->msg_list); |
1146 | info->node_cache = new_leaf; | |
ce2d52cc DL |
1147 | } else { |
1148 | kfree(new_leaf); | |
1149 | } | |
1150 | ||
1da177e4 | 1151 | if (info->attr.mq_curmsgs == 0) { |
2903ff01 | 1152 | if (f.file->f_flags & O_NONBLOCK) { |
1da177e4 LT |
1153 | spin_unlock(&info->lock); |
1154 | ret = -EAGAIN; | |
1da177e4 LT |
1155 | } else { |
1156 | wait.task = current; | |
1157 | wait.state = STATE_NONE; | |
1158 | ret = wq_sleep(info, RECV, timeout, &wait); | |
1159 | msg_ptr = wait.msg; | |
1160 | } | |
1161 | } else { | |
194a6b5b | 1162 | DEFINE_WAKE_Q(wake_q); |
fa6004ad | 1163 | |
1da177e4 LT |
1164 | msg_ptr = msg_get(info); |
1165 | ||
1166 | inode->i_atime = inode->i_mtime = inode->i_ctime = | |
078cd827 | 1167 | current_time(inode); |
1da177e4 LT |
1168 | |
1169 | /* There is now free space in queue. */ | |
fa6004ad | 1170 | pipelined_receive(&wake_q, info); |
1da177e4 | 1171 | spin_unlock(&info->lock); |
fa6004ad | 1172 | wake_up_q(&wake_q); |
1da177e4 LT |
1173 | ret = 0; |
1174 | } | |
1175 | if (ret == 0) { | |
1176 | ret = msg_ptr->m_ts; | |
1177 | ||
1178 | if ((u_msg_prio && put_user(msg_ptr->m_type, u_msg_prio)) || | |
1179 | store_msg(u_msg_ptr, msg_ptr, msg_ptr->m_ts)) { | |
1180 | ret = -EFAULT; | |
1181 | } | |
1182 | free_msg(msg_ptr); | |
1183 | } | |
1184 | out_fput: | |
2903ff01 | 1185 | fdput(f); |
1da177e4 LT |
1186 | out: |
1187 | return ret; | |
1188 | } | |
1189 | ||
0d060606 AV |
1190 | SYSCALL_DEFINE5(mq_timedsend, mqd_t, mqdes, const char __user *, u_msg_ptr, |
1191 | size_t, msg_len, unsigned int, msg_prio, | |
1192 | const struct timespec __user *, u_abs_timeout) | |
1193 | { | |
b9047726 | 1194 | struct timespec64 ts, *p = NULL; |
0d060606 AV |
1195 | if (u_abs_timeout) { |
1196 | int res = prepare_timeout(u_abs_timeout, &ts); | |
1197 | if (res) | |
1198 | return res; | |
1199 | p = &ts; | |
1200 | } | |
1201 | return do_mq_timedsend(mqdes, u_msg_ptr, msg_len, msg_prio, p); | |
1202 | } | |
1203 | ||
1204 | SYSCALL_DEFINE5(mq_timedreceive, mqd_t, mqdes, char __user *, u_msg_ptr, | |
1205 | size_t, msg_len, unsigned int __user *, u_msg_prio, | |
1206 | const struct timespec __user *, u_abs_timeout) | |
1207 | { | |
b9047726 | 1208 | struct timespec64 ts, *p = NULL; |
0d060606 AV |
1209 | if (u_abs_timeout) { |
1210 | int res = prepare_timeout(u_abs_timeout, &ts); | |
1211 | if (res) | |
1212 | return res; | |
1213 | p = &ts; | |
1214 | } | |
1215 | return do_mq_timedreceive(mqdes, u_msg_ptr, msg_len, u_msg_prio, p); | |
1216 | } | |
1217 | ||
1da177e4 LT |
1218 | /* |
1219 | * Notes: the case when user wants us to deregister (with NULL as pointer) | |
1220 | * and he isn't currently owner of notification, will be silently discarded. | |
1221 | * It isn't explicitly defined in the POSIX. | |
1222 | */ | |
0d060606 | 1223 | static int do_mq_notify(mqd_t mqdes, const struct sigevent *notification) |
1da177e4 | 1224 | { |
2903ff01 AV |
1225 | int ret; |
1226 | struct fd f; | |
1da177e4 LT |
1227 | struct sock *sock; |
1228 | struct inode *inode; | |
1da177e4 LT |
1229 | struct mqueue_inode_info *info; |
1230 | struct sk_buff *nc; | |
1231 | ||
0d060606 | 1232 | audit_mq_notify(mqdes, notification); |
1da177e4 | 1233 | |
20114f71 AV |
1234 | nc = NULL; |
1235 | sock = NULL; | |
0d060606 AV |
1236 | if (notification != NULL) { |
1237 | if (unlikely(notification->sigev_notify != SIGEV_NONE && | |
1238 | notification->sigev_notify != SIGEV_SIGNAL && | |
1239 | notification->sigev_notify != SIGEV_THREAD)) | |
1da177e4 | 1240 | return -EINVAL; |
0d060606 AV |
1241 | if (notification->sigev_notify == SIGEV_SIGNAL && |
1242 | !valid_signal(notification->sigev_signo)) { | |
1da177e4 LT |
1243 | return -EINVAL; |
1244 | } | |
0d060606 | 1245 | if (notification->sigev_notify == SIGEV_THREAD) { |
c3d8d1e3 PM |
1246 | long timeo; |
1247 | ||
1da177e4 LT |
1248 | /* create the notify skb */ |
1249 | nc = alloc_skb(NOTIFY_COOKIE_LEN, GFP_KERNEL); | |
8d8ffefa AGR |
1250 | if (!nc) { |
1251 | ret = -ENOMEM; | |
1da177e4 | 1252 | goto out; |
8d8ffefa | 1253 | } |
1da177e4 | 1254 | if (copy_from_user(nc->data, |
0d060606 | 1255 | notification->sigev_value.sival_ptr, |
1da177e4 | 1256 | NOTIFY_COOKIE_LEN)) { |
8d8ffefa | 1257 | ret = -EFAULT; |
1da177e4 LT |
1258 | goto out; |
1259 | } | |
1260 | ||
1261 | /* TODO: add a header? */ | |
1262 | skb_put(nc, NOTIFY_COOKIE_LEN); | |
1263 | /* and attach it to the socket */ | |
1264 | retry: | |
0d060606 | 1265 | f = fdget(notification->sigev_signo); |
2903ff01 | 1266 | if (!f.file) { |
8d8ffefa | 1267 | ret = -EBADF; |
1da177e4 | 1268 | goto out; |
8d8ffefa | 1269 | } |
2903ff01 AV |
1270 | sock = netlink_getsockbyfilp(f.file); |
1271 | fdput(f); | |
1da177e4 LT |
1272 | if (IS_ERR(sock)) { |
1273 | ret = PTR_ERR(sock); | |
1274 | sock = NULL; | |
1275 | goto out; | |
1276 | } | |
1277 | ||
c3d8d1e3 | 1278 | timeo = MAX_SCHEDULE_TIMEOUT; |
9457afee | 1279 | ret = netlink_attachskb(sock, nc, &timeo, NULL); |
f991af3d CW |
1280 | if (ret == 1) { |
1281 | sock = NULL; | |
8d8ffefa | 1282 | goto retry; |
f991af3d | 1283 | } |
1da177e4 LT |
1284 | if (ret) { |
1285 | sock = NULL; | |
1286 | nc = NULL; | |
1287 | goto out; | |
1288 | } | |
1289 | } | |
1290 | } | |
1291 | ||
2903ff01 AV |
1292 | f = fdget(mqdes); |
1293 | if (!f.file) { | |
8d8ffefa | 1294 | ret = -EBADF; |
1da177e4 | 1295 | goto out; |
8d8ffefa | 1296 | } |
1da177e4 | 1297 | |
496ad9aa | 1298 | inode = file_inode(f.file); |
2903ff01 | 1299 | if (unlikely(f.file->f_op != &mqueue_file_operations)) { |
8d8ffefa | 1300 | ret = -EBADF; |
1da177e4 | 1301 | goto out_fput; |
8d8ffefa | 1302 | } |
1da177e4 LT |
1303 | info = MQUEUE_I(inode); |
1304 | ||
1305 | ret = 0; | |
1306 | spin_lock(&info->lock); | |
0d060606 | 1307 | if (notification == NULL) { |
a03fcb73 | 1308 | if (info->notify_owner == task_tgid(current)) { |
1da177e4 | 1309 | remove_notification(info); |
078cd827 | 1310 | inode->i_atime = inode->i_ctime = current_time(inode); |
1da177e4 | 1311 | } |
a03fcb73 | 1312 | } else if (info->notify_owner != NULL) { |
1da177e4 LT |
1313 | ret = -EBUSY; |
1314 | } else { | |
0d060606 | 1315 | switch (notification->sigev_notify) { |
1da177e4 LT |
1316 | case SIGEV_NONE: |
1317 | info->notify.sigev_notify = SIGEV_NONE; | |
1318 | break; | |
1319 | case SIGEV_THREAD: | |
1320 | info->notify_sock = sock; | |
1321 | info->notify_cookie = nc; | |
1322 | sock = NULL; | |
1323 | nc = NULL; | |
1324 | info->notify.sigev_notify = SIGEV_THREAD; | |
1325 | break; | |
1326 | case SIGEV_SIGNAL: | |
0d060606 AV |
1327 | info->notify.sigev_signo = notification->sigev_signo; |
1328 | info->notify.sigev_value = notification->sigev_value; | |
1da177e4 LT |
1329 | info->notify.sigev_notify = SIGEV_SIGNAL; |
1330 | break; | |
1331 | } | |
a03fcb73 CLG |
1332 | |
1333 | info->notify_owner = get_pid(task_tgid(current)); | |
6f9ac6d9 | 1334 | info->notify_user_ns = get_user_ns(current_user_ns()); |
078cd827 | 1335 | inode->i_atime = inode->i_ctime = current_time(inode); |
1da177e4 LT |
1336 | } |
1337 | spin_unlock(&info->lock); | |
1338 | out_fput: | |
2903ff01 | 1339 | fdput(f); |
1da177e4 | 1340 | out: |
3ab08fe2 | 1341 | if (sock) |
1da177e4 | 1342 | netlink_detachskb(sock, nc); |
3ab08fe2 | 1343 | else if (nc) |
1da177e4 | 1344 | dev_kfree_skb(nc); |
3ab08fe2 | 1345 | |
1da177e4 LT |
1346 | return ret; |
1347 | } | |
1348 | ||
0d060606 AV |
1349 | SYSCALL_DEFINE2(mq_notify, mqd_t, mqdes, |
1350 | const struct sigevent __user *, u_notification) | |
1351 | { | |
1352 | struct sigevent n, *p = NULL; | |
1353 | if (u_notification) { | |
1354 | if (copy_from_user(&n, u_notification, sizeof(struct sigevent))) | |
1355 | return -EFAULT; | |
1356 | p = &n; | |
1357 | } | |
1358 | return do_mq_notify(mqdes, p); | |
1359 | } | |
1360 | ||
1361 | static int do_mq_getsetattr(int mqdes, struct mq_attr *new, struct mq_attr *old) | |
1da177e4 | 1362 | { |
2903ff01 | 1363 | struct fd f; |
1da177e4 LT |
1364 | struct inode *inode; |
1365 | struct mqueue_inode_info *info; | |
1366 | ||
0d060606 AV |
1367 | if (new && (new->mq_flags & (~O_NONBLOCK))) |
1368 | return -EINVAL; | |
1da177e4 | 1369 | |
2903ff01 | 1370 | f = fdget(mqdes); |
0d060606 AV |
1371 | if (!f.file) |
1372 | return -EBADF; | |
1da177e4 | 1373 | |
2903ff01 | 1374 | if (unlikely(f.file->f_op != &mqueue_file_operations)) { |
0d060606 AV |
1375 | fdput(f); |
1376 | return -EBADF; | |
8d8ffefa | 1377 | } |
0d060606 AV |
1378 | |
1379 | inode = file_inode(f.file); | |
1da177e4 LT |
1380 | info = MQUEUE_I(inode); |
1381 | ||
1382 | spin_lock(&info->lock); | |
1383 | ||
0d060606 AV |
1384 | if (old) { |
1385 | *old = info->attr; | |
1386 | old->mq_flags = f.file->f_flags & O_NONBLOCK; | |
1387 | } | |
1388 | if (new) { | |
1389 | audit_mq_getsetattr(mqdes, new); | |
2903ff01 | 1390 | spin_lock(&f.file->f_lock); |
0d060606 | 1391 | if (new->mq_flags & O_NONBLOCK) |
2903ff01 | 1392 | f.file->f_flags |= O_NONBLOCK; |
1da177e4 | 1393 | else |
2903ff01 AV |
1394 | f.file->f_flags &= ~O_NONBLOCK; |
1395 | spin_unlock(&f.file->f_lock); | |
1da177e4 | 1396 | |
078cd827 | 1397 | inode->i_atime = inode->i_ctime = current_time(inode); |
1da177e4 LT |
1398 | } |
1399 | ||
1400 | spin_unlock(&info->lock); | |
0d060606 AV |
1401 | fdput(f); |
1402 | return 0; | |
1403 | } | |
1da177e4 | 1404 | |
0d060606 AV |
1405 | SYSCALL_DEFINE3(mq_getsetattr, mqd_t, mqdes, |
1406 | const struct mq_attr __user *, u_mqstat, | |
1407 | struct mq_attr __user *, u_omqstat) | |
1408 | { | |
1409 | int ret; | |
1410 | struct mq_attr mqstat, omqstat; | |
1411 | struct mq_attr *new = NULL, *old = NULL; | |
1da177e4 | 1412 | |
0d060606 AV |
1413 | if (u_mqstat) { |
1414 | new = &mqstat; | |
1415 | if (copy_from_user(new, u_mqstat, sizeof(struct mq_attr))) | |
1416 | return -EFAULT; | |
1417 | } | |
1418 | if (u_omqstat) | |
1419 | old = &omqstat; | |
1420 | ||
1421 | ret = do_mq_getsetattr(mqdes, new, old); | |
1422 | if (ret || !old) | |
1423 | return ret; | |
1424 | ||
1425 | if (copy_to_user(u_omqstat, old, sizeof(struct mq_attr))) | |
1426 | return -EFAULT; | |
1427 | return 0; | |
1428 | } | |
1429 | ||
1430 | #ifdef CONFIG_COMPAT | |
1431 | ||
1432 | struct compat_mq_attr { | |
1433 | compat_long_t mq_flags; /* message queue flags */ | |
1434 | compat_long_t mq_maxmsg; /* maximum number of messages */ | |
1435 | compat_long_t mq_msgsize; /* maximum message size */ | |
1436 | compat_long_t mq_curmsgs; /* number of messages currently queued */ | |
1437 | compat_long_t __reserved[4]; /* ignored for input, zeroed for output */ | |
1438 | }; | |
1439 | ||
1440 | static inline int get_compat_mq_attr(struct mq_attr *attr, | |
1441 | const struct compat_mq_attr __user *uattr) | |
1442 | { | |
1443 | struct compat_mq_attr v; | |
1444 | ||
1445 | if (copy_from_user(&v, uattr, sizeof(*uattr))) | |
1446 | return -EFAULT; | |
1447 | ||
1448 | memset(attr, 0, sizeof(*attr)); | |
1449 | attr->mq_flags = v.mq_flags; | |
1450 | attr->mq_maxmsg = v.mq_maxmsg; | |
1451 | attr->mq_msgsize = v.mq_msgsize; | |
1452 | attr->mq_curmsgs = v.mq_curmsgs; | |
1453 | return 0; | |
1454 | } | |
1455 | ||
1456 | static inline int put_compat_mq_attr(const struct mq_attr *attr, | |
1457 | struct compat_mq_attr __user *uattr) | |
1458 | { | |
1459 | struct compat_mq_attr v; | |
1460 | ||
1461 | memset(&v, 0, sizeof(v)); | |
1462 | v.mq_flags = attr->mq_flags; | |
1463 | v.mq_maxmsg = attr->mq_maxmsg; | |
1464 | v.mq_msgsize = attr->mq_msgsize; | |
1465 | v.mq_curmsgs = attr->mq_curmsgs; | |
1466 | if (copy_to_user(uattr, &v, sizeof(*uattr))) | |
1467 | return -EFAULT; | |
1468 | return 0; | |
1469 | } | |
1470 | ||
1471 | COMPAT_SYSCALL_DEFINE4(mq_open, const char __user *, u_name, | |
1472 | int, oflag, compat_mode_t, mode, | |
1473 | struct compat_mq_attr __user *, u_attr) | |
1474 | { | |
1475 | struct mq_attr attr, *p = NULL; | |
1476 | if (u_attr && oflag & O_CREAT) { | |
1477 | p = &attr; | |
1478 | if (get_compat_mq_attr(&attr, u_attr)) | |
1479 | return -EFAULT; | |
1480 | } | |
1481 | return do_mq_open(u_name, oflag, mode, p); | |
1482 | } | |
1483 | ||
1484 | static int compat_prepare_timeout(const struct compat_timespec __user *p, | |
b9047726 | 1485 | struct timespec64 *ts) |
0d060606 | 1486 | { |
b9047726 | 1487 | if (compat_get_timespec64(ts, p)) |
0d060606 | 1488 | return -EFAULT; |
b9047726 | 1489 | if (!timespec64_valid(ts)) |
0d060606 AV |
1490 | return -EINVAL; |
1491 | return 0; | |
1492 | } | |
1493 | ||
1494 | COMPAT_SYSCALL_DEFINE5(mq_timedsend, mqd_t, mqdes, | |
1495 | const char __user *, u_msg_ptr, | |
1496 | compat_size_t, msg_len, unsigned int, msg_prio, | |
1497 | const struct compat_timespec __user *, u_abs_timeout) | |
1498 | { | |
b9047726 | 1499 | struct timespec64 ts, *p = NULL; |
0d060606 AV |
1500 | if (u_abs_timeout) { |
1501 | int res = compat_prepare_timeout(u_abs_timeout, &ts); | |
1502 | if (res) | |
1503 | return res; | |
1504 | p = &ts; | |
1505 | } | |
1506 | return do_mq_timedsend(mqdes, u_msg_ptr, msg_len, msg_prio, p); | |
1507 | } | |
1508 | ||
1509 | COMPAT_SYSCALL_DEFINE5(mq_timedreceive, mqd_t, mqdes, | |
1510 | char __user *, u_msg_ptr, | |
1511 | compat_size_t, msg_len, unsigned int __user *, u_msg_prio, | |
1512 | const struct compat_timespec __user *, u_abs_timeout) | |
1513 | { | |
b9047726 | 1514 | struct timespec64 ts, *p = NULL; |
0d060606 AV |
1515 | if (u_abs_timeout) { |
1516 | int res = compat_prepare_timeout(u_abs_timeout, &ts); | |
1517 | if (res) | |
1518 | return res; | |
1519 | p = &ts; | |
1520 | } | |
1521 | return do_mq_timedreceive(mqdes, u_msg_ptr, msg_len, u_msg_prio, p); | |
1522 | } | |
1523 | ||
1524 | COMPAT_SYSCALL_DEFINE2(mq_notify, mqd_t, mqdes, | |
1525 | const struct compat_sigevent __user *, u_notification) | |
1526 | { | |
1527 | struct sigevent n, *p = NULL; | |
1528 | if (u_notification) { | |
1529 | if (get_compat_sigevent(&n, u_notification)) | |
1530 | return -EFAULT; | |
1531 | if (n.sigev_notify == SIGEV_THREAD) | |
1532 | n.sigev_value.sival_ptr = compat_ptr(n.sigev_value.sival_int); | |
1533 | p = &n; | |
1534 | } | |
1535 | return do_mq_notify(mqdes, p); | |
1536 | } | |
1537 | ||
1538 | COMPAT_SYSCALL_DEFINE3(mq_getsetattr, mqd_t, mqdes, | |
1539 | const struct compat_mq_attr __user *, u_mqstat, | |
1540 | struct compat_mq_attr __user *, u_omqstat) | |
1541 | { | |
1542 | int ret; | |
1543 | struct mq_attr mqstat, omqstat; | |
1544 | struct mq_attr *new = NULL, *old = NULL; | |
1545 | ||
1546 | if (u_mqstat) { | |
1547 | new = &mqstat; | |
1548 | if (get_compat_mq_attr(new, u_mqstat)) | |
1549 | return -EFAULT; | |
1550 | } | |
1551 | if (u_omqstat) | |
1552 | old = &omqstat; | |
1553 | ||
1554 | ret = do_mq_getsetattr(mqdes, new, old); | |
1555 | if (ret || !old) | |
1556 | return ret; | |
1557 | ||
1558 | if (put_compat_mq_attr(old, u_omqstat)) | |
1559 | return -EFAULT; | |
1560 | return 0; | |
1da177e4 | 1561 | } |
0d060606 | 1562 | #endif |
1da177e4 | 1563 | |
92e1d5be | 1564 | static const struct inode_operations mqueue_dir_inode_operations = { |
1da177e4 LT |
1565 | .lookup = simple_lookup, |
1566 | .create = mqueue_create, | |
1567 | .unlink = mqueue_unlink, | |
1568 | }; | |
1569 | ||
9a32144e | 1570 | static const struct file_operations mqueue_file_operations = { |
1da177e4 LT |
1571 | .flush = mqueue_flush_file, |
1572 | .poll = mqueue_poll_file, | |
1573 | .read = mqueue_read_file, | |
6038f373 | 1574 | .llseek = default_llseek, |
1da177e4 LT |
1575 | }; |
1576 | ||
b87221de | 1577 | static const struct super_operations mqueue_super_ops = { |
1da177e4 LT |
1578 | .alloc_inode = mqueue_alloc_inode, |
1579 | .destroy_inode = mqueue_destroy_inode, | |
6d8af64c | 1580 | .evict_inode = mqueue_evict_inode, |
1da177e4 | 1581 | .statfs = simple_statfs, |
1da177e4 LT |
1582 | }; |
1583 | ||
1584 | static struct file_system_type mqueue_fs_type = { | |
1585 | .name = "mqueue", | |
ceefda69 | 1586 | .mount = mqueue_mount, |
1da177e4 | 1587 | .kill_sb = kill_litter_super, |
bc1b69ed | 1588 | .fs_flags = FS_USERNS_MOUNT, |
1da177e4 LT |
1589 | }; |
1590 | ||
7eafd7c7 SH |
1591 | int mq_init_ns(struct ipc_namespace *ns) |
1592 | { | |
1593 | ns->mq_queues_count = 0; | |
1594 | ns->mq_queues_max = DFLT_QUEUESMAX; | |
1595 | ns->mq_msg_max = DFLT_MSGMAX; | |
1596 | ns->mq_msgsize_max = DFLT_MSGSIZEMAX; | |
cef0184c KM |
1597 | ns->mq_msg_default = DFLT_MSG; |
1598 | ns->mq_msgsize_default = DFLT_MSGSIZE; | |
7eafd7c7 SH |
1599 | |
1600 | ns->mq_mnt = kern_mount_data(&mqueue_fs_type, ns); | |
1601 | if (IS_ERR(ns->mq_mnt)) { | |
1602 | int err = PTR_ERR(ns->mq_mnt); | |
1603 | ns->mq_mnt = NULL; | |
1604 | return err; | |
1605 | } | |
1606 | return 0; | |
1607 | } | |
1608 | ||
1609 | void mq_clear_sbinfo(struct ipc_namespace *ns) | |
1610 | { | |
1611 | ns->mq_mnt->mnt_sb->s_fs_info = NULL; | |
1612 | } | |
1613 | ||
1614 | void mq_put_mnt(struct ipc_namespace *ns) | |
1615 | { | |
6f686574 | 1616 | kern_unmount(ns->mq_mnt); |
7eafd7c7 SH |
1617 | } |
1618 | ||
1da177e4 LT |
1619 | static int __init init_mqueue_fs(void) |
1620 | { | |
1621 | int error; | |
1622 | ||
1623 | mqueue_inode_cachep = kmem_cache_create("mqueue_inode_cache", | |
1624 | sizeof(struct mqueue_inode_info), 0, | |
5d097056 | 1625 | SLAB_HWCACHE_ALIGN|SLAB_ACCOUNT, init_once); |
1da177e4 LT |
1626 | if (mqueue_inode_cachep == NULL) |
1627 | return -ENOMEM; | |
1628 | ||
2329e392 | 1629 | /* ignore failures - they are not fatal */ |
bdc8e5f8 | 1630 | mq_sysctl_table = mq_register_sysctl_table(); |
1da177e4 LT |
1631 | |
1632 | error = register_filesystem(&mqueue_fs_type); | |
1633 | if (error) | |
1634 | goto out_sysctl; | |
1635 | ||
7eafd7c7 SH |
1636 | spin_lock_init(&mq_lock); |
1637 | ||
6f686574 AV |
1638 | error = mq_init_ns(&init_ipc_ns); |
1639 | if (error) | |
1da177e4 | 1640 | goto out_filesystem; |
1da177e4 | 1641 | |
1da177e4 LT |
1642 | return 0; |
1643 | ||
1644 | out_filesystem: | |
1645 | unregister_filesystem(&mqueue_fs_type); | |
1646 | out_sysctl: | |
1647 | if (mq_sysctl_table) | |
1648 | unregister_sysctl_table(mq_sysctl_table); | |
1a1d92c1 | 1649 | kmem_cache_destroy(mqueue_inode_cachep); |
1da177e4 LT |
1650 | return error; |
1651 | } | |
1652 | ||
6d08a256 | 1653 | device_initcall(init_mqueue_fs); |