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> | |
935c6912 | 21 | #include <linux/fs_context.h> |
1da177e4 LT |
22 | #include <linux/namei.h> |
23 | #include <linux/sysctl.h> | |
24 | #include <linux/poll.h> | |
25 | #include <linux/mqueue.h> | |
26 | #include <linux/msg.h> | |
27 | #include <linux/skbuff.h> | |
5b5c4d1a | 28 | #include <linux/vmalloc.h> |
1da177e4 LT |
29 | #include <linux/netlink.h> |
30 | #include <linux/syscalls.h> | |
20ca73bc | 31 | #include <linux/audit.h> |
7ed20e1a | 32 | #include <linux/signal.h> |
5f921ae9 | 33 | #include <linux/mutex.h> |
b488893a PE |
34 | #include <linux/nsproxy.h> |
35 | #include <linux/pid.h> | |
614b84cf | 36 | #include <linux/ipc_namespace.h> |
6b550f94 | 37 | #include <linux/user_namespace.h> |
5a0e3ad6 | 38 | #include <linux/slab.h> |
84f001e1 | 39 | #include <linux/sched/wake_q.h> |
3f07c014 | 40 | #include <linux/sched/signal.h> |
8703e8a4 | 41 | #include <linux/sched/user.h> |
5f921ae9 | 42 | |
1da177e4 LT |
43 | #include <net/sock.h> |
44 | #include "util.h" | |
45 | ||
935c6912 DH |
46 | struct mqueue_fs_context { |
47 | struct ipc_namespace *ipc_ns; | |
48 | }; | |
49 | ||
1da177e4 LT |
50 | #define MQUEUE_MAGIC 0x19800202 |
51 | #define DIRENT_SIZE 20 | |
52 | #define FILENT_SIZE 80 | |
53 | ||
54 | #define SEND 0 | |
55 | #define RECV 1 | |
56 | ||
57 | #define STATE_NONE 0 | |
fa6004ad | 58 | #define STATE_READY 1 |
1da177e4 | 59 | |
d6629859 DL |
60 | struct posix_msg_tree_node { |
61 | struct rb_node rb_node; | |
62 | struct list_head msg_list; | |
63 | int priority; | |
64 | }; | |
65 | ||
1da177e4 LT |
66 | struct ext_wait_queue { /* queue of sleeping tasks */ |
67 | struct task_struct *task; | |
68 | struct list_head list; | |
69 | struct msg_msg *msg; /* ptr of loaded message */ | |
70 | int state; /* one of STATE_* values */ | |
71 | }; | |
72 | ||
73 | struct mqueue_inode_info { | |
74 | spinlock_t lock; | |
75 | struct inode vfs_inode; | |
76 | wait_queue_head_t wait_q; | |
77 | ||
d6629859 | 78 | struct rb_root msg_tree; |
a5091fda | 79 | struct rb_node *msg_tree_rightmost; |
ce2d52cc | 80 | struct posix_msg_tree_node *node_cache; |
1da177e4 LT |
81 | struct mq_attr attr; |
82 | ||
83 | struct sigevent notify; | |
239521f3 | 84 | struct pid *notify_owner; |
6f9ac6d9 | 85 | struct user_namespace *notify_user_ns; |
338cec32 | 86 | struct user_struct *user; /* user who created, for accounting */ |
1da177e4 LT |
87 | struct sock *notify_sock; |
88 | struct sk_buff *notify_cookie; | |
89 | ||
90 | /* for tasks waiting for free space and messages, respectively */ | |
91 | struct ext_wait_queue e_wait_q[2]; | |
92 | ||
93 | unsigned long qsize; /* size of queue in memory (sum of all msgs) */ | |
94 | }; | |
95 | ||
935c6912 | 96 | static struct file_system_type mqueue_fs_type; |
92e1d5be | 97 | static const struct inode_operations mqueue_dir_inode_operations; |
9a32144e | 98 | static const struct file_operations mqueue_file_operations; |
b87221de | 99 | static const struct super_operations mqueue_super_ops; |
935c6912 | 100 | static const struct fs_context_operations mqueue_fs_context_ops; |
1da177e4 LT |
101 | static void remove_notification(struct mqueue_inode_info *info); |
102 | ||
e18b890b | 103 | static struct kmem_cache *mqueue_inode_cachep; |
1da177e4 | 104 | |
239521f3 | 105 | static struct ctl_table_header *mq_sysctl_table; |
1da177e4 LT |
106 | |
107 | static inline struct mqueue_inode_info *MQUEUE_I(struct inode *inode) | |
108 | { | |
109 | return container_of(inode, struct mqueue_inode_info, vfs_inode); | |
110 | } | |
111 | ||
7eafd7c7 SH |
112 | /* |
113 | * This routine should be called with the mq_lock held. | |
114 | */ | |
115 | static inline struct ipc_namespace *__get_ns_from_inode(struct inode *inode) | |
614b84cf | 116 | { |
7eafd7c7 | 117 | return get_ipc_ns(inode->i_sb->s_fs_info); |
614b84cf SH |
118 | } |
119 | ||
7eafd7c7 | 120 | static struct ipc_namespace *get_ns_from_inode(struct inode *inode) |
614b84cf | 121 | { |
7eafd7c7 SH |
122 | struct ipc_namespace *ns; |
123 | ||
124 | spin_lock(&mq_lock); | |
125 | ns = __get_ns_from_inode(inode); | |
126 | spin_unlock(&mq_lock); | |
127 | return ns; | |
614b84cf SH |
128 | } |
129 | ||
d6629859 DL |
130 | /* Auxiliary functions to manipulate messages' list */ |
131 | static int msg_insert(struct msg_msg *msg, struct mqueue_inode_info *info) | |
132 | { | |
133 | struct rb_node **p, *parent = NULL; | |
134 | struct posix_msg_tree_node *leaf; | |
a5091fda | 135 | bool rightmost = true; |
d6629859 DL |
136 | |
137 | p = &info->msg_tree.rb_node; | |
138 | while (*p) { | |
139 | parent = *p; | |
140 | leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node); | |
141 | ||
142 | if (likely(leaf->priority == msg->m_type)) | |
143 | goto insert_msg; | |
a5091fda | 144 | else if (msg->m_type < leaf->priority) { |
d6629859 | 145 | p = &(*p)->rb_left; |
a5091fda DB |
146 | rightmost = false; |
147 | } else | |
d6629859 DL |
148 | p = &(*p)->rb_right; |
149 | } | |
ce2d52cc DL |
150 | if (info->node_cache) { |
151 | leaf = info->node_cache; | |
152 | info->node_cache = NULL; | |
153 | } else { | |
154 | leaf = kmalloc(sizeof(*leaf), GFP_ATOMIC); | |
155 | if (!leaf) | |
156 | return -ENOMEM; | |
ce2d52cc | 157 | INIT_LIST_HEAD(&leaf->msg_list); |
ce2d52cc | 158 | } |
d6629859 | 159 | leaf->priority = msg->m_type; |
a5091fda DB |
160 | |
161 | if (rightmost) | |
162 | info->msg_tree_rightmost = &leaf->rb_node; | |
163 | ||
d6629859 DL |
164 | rb_link_node(&leaf->rb_node, parent, p); |
165 | rb_insert_color(&leaf->rb_node, &info->msg_tree); | |
d6629859 DL |
166 | insert_msg: |
167 | info->attr.mq_curmsgs++; | |
168 | info->qsize += msg->m_ts; | |
169 | list_add_tail(&msg->m_list, &leaf->msg_list); | |
170 | return 0; | |
171 | } | |
172 | ||
a5091fda DB |
173 | static inline void msg_tree_erase(struct posix_msg_tree_node *leaf, |
174 | struct mqueue_inode_info *info) | |
175 | { | |
176 | struct rb_node *node = &leaf->rb_node; | |
177 | ||
178 | if (info->msg_tree_rightmost == node) | |
179 | info->msg_tree_rightmost = rb_prev(node); | |
180 | ||
181 | rb_erase(node, &info->msg_tree); | |
182 | if (info->node_cache) { | |
183 | kfree(leaf); | |
184 | } else { | |
185 | info->node_cache = leaf; | |
186 | } | |
187 | } | |
188 | ||
d6629859 DL |
189 | static inline struct msg_msg *msg_get(struct mqueue_inode_info *info) |
190 | { | |
a5091fda | 191 | struct rb_node *parent = NULL; |
d6629859 DL |
192 | struct posix_msg_tree_node *leaf; |
193 | struct msg_msg *msg; | |
194 | ||
195 | try_again: | |
a5091fda DB |
196 | /* |
197 | * During insert, low priorities go to the left and high to the | |
198 | * right. On receive, we want the highest priorities first, so | |
199 | * walk all the way to the right. | |
200 | */ | |
201 | parent = info->msg_tree_rightmost; | |
d6629859 DL |
202 | if (!parent) { |
203 | if (info->attr.mq_curmsgs) { | |
204 | pr_warn_once("Inconsistency in POSIX message queue, " | |
205 | "no tree element, but supposedly messages " | |
206 | "should exist!\n"); | |
207 | info->attr.mq_curmsgs = 0; | |
208 | } | |
209 | return NULL; | |
210 | } | |
211 | leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node); | |
ce2d52cc | 212 | if (unlikely(list_empty(&leaf->msg_list))) { |
d6629859 DL |
213 | pr_warn_once("Inconsistency in POSIX message queue, " |
214 | "empty leaf node but we haven't implemented " | |
215 | "lazy leaf delete!\n"); | |
a5091fda | 216 | msg_tree_erase(leaf, info); |
d6629859 DL |
217 | goto try_again; |
218 | } else { | |
219 | msg = list_first_entry(&leaf->msg_list, | |
220 | struct msg_msg, m_list); | |
221 | list_del(&msg->m_list); | |
222 | if (list_empty(&leaf->msg_list)) { | |
a5091fda | 223 | msg_tree_erase(leaf, info); |
d6629859 DL |
224 | } |
225 | } | |
226 | info->attr.mq_curmsgs--; | |
227 | info->qsize -= msg->m_ts; | |
228 | return msg; | |
229 | } | |
230 | ||
7eafd7c7 | 231 | static struct inode *mqueue_get_inode(struct super_block *sb, |
1b9d5ff7 | 232 | struct ipc_namespace *ipc_ns, umode_t mode, |
7eafd7c7 | 233 | struct mq_attr *attr) |
1da177e4 | 234 | { |
86a264ab | 235 | struct user_struct *u = current_user(); |
1da177e4 | 236 | struct inode *inode; |
d40dcdb0 | 237 | int ret = -ENOMEM; |
1da177e4 LT |
238 | |
239 | inode = new_inode(sb); | |
04715206 JS |
240 | if (!inode) |
241 | goto err; | |
242 | ||
243 | inode->i_ino = get_next_ino(); | |
244 | inode->i_mode = mode; | |
245 | inode->i_uid = current_fsuid(); | |
246 | inode->i_gid = current_fsgid(); | |
078cd827 | 247 | inode->i_mtime = inode->i_ctime = inode->i_atime = current_time(inode); |
04715206 JS |
248 | |
249 | if (S_ISREG(mode)) { | |
250 | struct mqueue_inode_info *info; | |
d6629859 | 251 | unsigned long mq_bytes, mq_treesize; |
04715206 JS |
252 | |
253 | inode->i_fop = &mqueue_file_operations; | |
254 | inode->i_size = FILENT_SIZE; | |
255 | /* mqueue specific info */ | |
256 | info = MQUEUE_I(inode); | |
257 | spin_lock_init(&info->lock); | |
258 | init_waitqueue_head(&info->wait_q); | |
259 | INIT_LIST_HEAD(&info->e_wait_q[0].list); | |
260 | INIT_LIST_HEAD(&info->e_wait_q[1].list); | |
261 | info->notify_owner = NULL; | |
6f9ac6d9 | 262 | info->notify_user_ns = NULL; |
04715206 JS |
263 | info->qsize = 0; |
264 | info->user = NULL; /* set when all is ok */ | |
d6629859 | 265 | info->msg_tree = RB_ROOT; |
a5091fda | 266 | info->msg_tree_rightmost = NULL; |
ce2d52cc | 267 | info->node_cache = NULL; |
04715206 | 268 | memset(&info->attr, 0, sizeof(info->attr)); |
cef0184c KM |
269 | info->attr.mq_maxmsg = min(ipc_ns->mq_msg_max, |
270 | ipc_ns->mq_msg_default); | |
271 | info->attr.mq_msgsize = min(ipc_ns->mq_msgsize_max, | |
272 | ipc_ns->mq_msgsize_default); | |
04715206 JS |
273 | if (attr) { |
274 | info->attr.mq_maxmsg = attr->mq_maxmsg; | |
275 | info->attr.mq_msgsize = attr->mq_msgsize; | |
276 | } | |
d6629859 DL |
277 | /* |
278 | * We used to allocate a static array of pointers and account | |
279 | * the size of that array as well as one msg_msg struct per | |
280 | * possible message into the queue size. That's no longer | |
281 | * accurate as the queue is now an rbtree and will grow and | |
282 | * shrink depending on usage patterns. We can, however, still | |
283 | * account one msg_msg struct per message, but the nodes are | |
284 | * allocated depending on priority usage, and most programs | |
285 | * only use one, or a handful, of priorities. However, since | |
286 | * this is pinned memory, we need to assume worst case, so | |
287 | * that means the min(mq_maxmsg, max_priorities) * struct | |
288 | * posix_msg_tree_node. | |
289 | */ | |
05c1b290 AV |
290 | |
291 | ret = -EINVAL; | |
292 | if (info->attr.mq_maxmsg <= 0 || info->attr.mq_msgsize <= 0) | |
293 | goto out_inode; | |
294 | if (capable(CAP_SYS_RESOURCE)) { | |
295 | if (info->attr.mq_maxmsg > HARD_MSGMAX || | |
296 | info->attr.mq_msgsize > HARD_MSGSIZEMAX) | |
297 | goto out_inode; | |
298 | } else { | |
299 | if (info->attr.mq_maxmsg > ipc_ns->mq_msg_max || | |
300 | info->attr.mq_msgsize > ipc_ns->mq_msgsize_max) | |
301 | goto out_inode; | |
302 | } | |
303 | ret = -EOVERFLOW; | |
304 | /* check for overflow */ | |
305 | if (info->attr.mq_msgsize > ULONG_MAX/info->attr.mq_maxmsg) | |
306 | goto out_inode; | |
d6629859 DL |
307 | mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) + |
308 | min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) * | |
309 | sizeof(struct posix_msg_tree_node); | |
05c1b290 AV |
310 | mq_bytes = info->attr.mq_maxmsg * info->attr.mq_msgsize; |
311 | if (mq_bytes + mq_treesize < mq_bytes) | |
312 | goto out_inode; | |
313 | mq_bytes += mq_treesize; | |
04715206 JS |
314 | spin_lock(&mq_lock); |
315 | if (u->mq_bytes + mq_bytes < u->mq_bytes || | |
2a4e64b8 | 316 | u->mq_bytes + mq_bytes > rlimit(RLIMIT_MSGQUEUE)) { |
04715206 JS |
317 | spin_unlock(&mq_lock); |
318 | /* mqueue_evict_inode() releases info->messages */ | |
d40dcdb0 | 319 | ret = -EMFILE; |
04715206 | 320 | goto out_inode; |
1da177e4 | 321 | } |
04715206 JS |
322 | u->mq_bytes += mq_bytes; |
323 | spin_unlock(&mq_lock); | |
324 | ||
325 | /* all is ok */ | |
326 | info->user = get_uid(u); | |
327 | } else if (S_ISDIR(mode)) { | |
328 | inc_nlink(inode); | |
329 | /* Some things misbehave if size == 0 on a directory */ | |
330 | inode->i_size = 2 * DIRENT_SIZE; | |
331 | inode->i_op = &mqueue_dir_inode_operations; | |
332 | inode->i_fop = &simple_dir_operations; | |
1da177e4 | 333 | } |
04715206 | 334 | |
1da177e4 LT |
335 | return inode; |
336 | out_inode: | |
1da177e4 | 337 | iput(inode); |
04715206 | 338 | err: |
d40dcdb0 | 339 | return ERR_PTR(ret); |
1da177e4 LT |
340 | } |
341 | ||
935c6912 | 342 | static int mqueue_fill_super(struct super_block *sb, struct fs_context *fc) |
1da177e4 LT |
343 | { |
344 | struct inode *inode; | |
cfb2f6f6 | 345 | struct ipc_namespace *ns = sb->s_fs_info; |
1da177e4 | 346 | |
a2982cc9 | 347 | sb->s_iflags |= SB_I_NOEXEC | SB_I_NODEV; |
09cbfeaf KS |
348 | sb->s_blocksize = PAGE_SIZE; |
349 | sb->s_blocksize_bits = PAGE_SHIFT; | |
1da177e4 LT |
350 | sb->s_magic = MQUEUE_MAGIC; |
351 | sb->s_op = &mqueue_super_ops; | |
352 | ||
48fde701 AV |
353 | inode = mqueue_get_inode(sb, ns, S_IFDIR | S_ISVTX | S_IRWXUGO, NULL); |
354 | if (IS_ERR(inode)) | |
355 | return PTR_ERR(inode); | |
1da177e4 | 356 | |
48fde701 AV |
357 | sb->s_root = d_make_root(inode); |
358 | if (!sb->s_root) | |
359 | return -ENOMEM; | |
360 | return 0; | |
1da177e4 LT |
361 | } |
362 | ||
935c6912 | 363 | static int mqueue_get_tree(struct fs_context *fc) |
1da177e4 | 364 | { |
935c6912 DH |
365 | struct mqueue_fs_context *ctx = fc->fs_private; |
366 | ||
367 | put_user_ns(fc->user_ns); | |
368 | fc->user_ns = get_user_ns(ctx->ipc_ns->user_ns); | |
369 | fc->s_fs_info = ctx->ipc_ns; | |
370 | return vfs_get_super(fc, vfs_get_keyed_super, mqueue_fill_super); | |
371 | } | |
372 | ||
373 | static void mqueue_fs_context_free(struct fs_context *fc) | |
374 | { | |
375 | struct mqueue_fs_context *ctx = fc->fs_private; | |
376 | ||
377 | if (ctx->ipc_ns) | |
378 | put_ipc_ns(ctx->ipc_ns); | |
379 | kfree(ctx); | |
380 | } | |
381 | ||
382 | static int mqueue_init_fs_context(struct fs_context *fc) | |
383 | { | |
384 | struct mqueue_fs_context *ctx; | |
385 | ||
386 | ctx = kzalloc(sizeof(struct mqueue_fs_context), GFP_KERNEL); | |
387 | if (!ctx) | |
388 | return -ENOMEM; | |
389 | ||
390 | ctx->ipc_ns = get_ipc_ns(current->nsproxy->ipc_ns); | |
391 | fc->fs_private = ctx; | |
392 | fc->ops = &mqueue_fs_context_ops; | |
393 | return 0; | |
394 | } | |
395 | ||
396 | static struct vfsmount *mq_create_mount(struct ipc_namespace *ns) | |
397 | { | |
398 | struct mqueue_fs_context *ctx; | |
399 | struct fs_context *fc; | |
400 | struct vfsmount *mnt; | |
401 | ||
402 | fc = fs_context_for_mount(&mqueue_fs_type, SB_KERNMOUNT); | |
403 | if (IS_ERR(fc)) | |
404 | return ERR_CAST(fc); | |
405 | ||
406 | ctx = fc->fs_private; | |
407 | put_ipc_ns(ctx->ipc_ns); | |
408 | ctx->ipc_ns = get_ipc_ns(ns); | |
409 | ||
410 | mnt = fc_mount(fc); | |
411 | put_fs_context(fc); | |
412 | return mnt; | |
1da177e4 LT |
413 | } |
414 | ||
51cc5068 | 415 | static void init_once(void *foo) |
1da177e4 LT |
416 | { |
417 | struct mqueue_inode_info *p = (struct mqueue_inode_info *) foo; | |
418 | ||
a35afb83 | 419 | inode_init_once(&p->vfs_inode); |
1da177e4 LT |
420 | } |
421 | ||
422 | static struct inode *mqueue_alloc_inode(struct super_block *sb) | |
423 | { | |
424 | struct mqueue_inode_info *ei; | |
425 | ||
e94b1766 | 426 | ei = kmem_cache_alloc(mqueue_inode_cachep, GFP_KERNEL); |
1da177e4 LT |
427 | if (!ei) |
428 | return NULL; | |
429 | return &ei->vfs_inode; | |
430 | } | |
431 | ||
015d7956 | 432 | static void mqueue_free_inode(struct inode *inode) |
1da177e4 LT |
433 | { |
434 | kmem_cache_free(mqueue_inode_cachep, MQUEUE_I(inode)); | |
435 | } | |
436 | ||
6d8af64c | 437 | static void mqueue_evict_inode(struct inode *inode) |
1da177e4 LT |
438 | { |
439 | struct mqueue_inode_info *info; | |
440 | struct user_struct *user; | |
d6629859 | 441 | unsigned long mq_bytes, mq_treesize; |
7eafd7c7 | 442 | struct ipc_namespace *ipc_ns; |
d6a2946a LR |
443 | struct msg_msg *msg, *nmsg; |
444 | LIST_HEAD(tmp_msg); | |
1da177e4 | 445 | |
dbd5768f | 446 | clear_inode(inode); |
6d8af64c AV |
447 | |
448 | if (S_ISDIR(inode->i_mode)) | |
1da177e4 | 449 | return; |
6d8af64c | 450 | |
7eafd7c7 | 451 | ipc_ns = get_ns_from_inode(inode); |
1da177e4 LT |
452 | info = MQUEUE_I(inode); |
453 | spin_lock(&info->lock); | |
d6629859 | 454 | while ((msg = msg_get(info)) != NULL) |
d6a2946a | 455 | list_add_tail(&msg->m_list, &tmp_msg); |
ce2d52cc | 456 | kfree(info->node_cache); |
1da177e4 LT |
457 | spin_unlock(&info->lock); |
458 | ||
d6a2946a LR |
459 | list_for_each_entry_safe(msg, nmsg, &tmp_msg, m_list) { |
460 | list_del(&msg->m_list); | |
461 | free_msg(msg); | |
462 | } | |
463 | ||
8834cf79 | 464 | /* Total amount of bytes accounted for the mqueue */ |
d6629859 DL |
465 | mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) + |
466 | min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) * | |
467 | sizeof(struct posix_msg_tree_node); | |
468 | ||
469 | mq_bytes = mq_treesize + (info->attr.mq_maxmsg * | |
470 | info->attr.mq_msgsize); | |
471 | ||
1da177e4 LT |
472 | user = info->user; |
473 | if (user) { | |
474 | spin_lock(&mq_lock); | |
475 | user->mq_bytes -= mq_bytes; | |
7eafd7c7 SH |
476 | /* |
477 | * get_ns_from_inode() ensures that the | |
478 | * (ipc_ns = sb->s_fs_info) is either a valid ipc_ns | |
479 | * to which we now hold a reference, or it is NULL. | |
480 | * We can't put it here under mq_lock, though. | |
481 | */ | |
482 | if (ipc_ns) | |
483 | ipc_ns->mq_queues_count--; | |
1da177e4 LT |
484 | spin_unlock(&mq_lock); |
485 | free_uid(user); | |
486 | } | |
7eafd7c7 SH |
487 | if (ipc_ns) |
488 | put_ipc_ns(ipc_ns); | |
1da177e4 LT |
489 | } |
490 | ||
eecec19d | 491 | static int mqueue_create_attr(struct dentry *dentry, umode_t mode, void *arg) |
1da177e4 | 492 | { |
eecec19d | 493 | struct inode *dir = dentry->d_parent->d_inode; |
1da177e4 | 494 | struct inode *inode; |
eecec19d | 495 | struct mq_attr *attr = arg; |
1da177e4 | 496 | int error; |
7eafd7c7 | 497 | struct ipc_namespace *ipc_ns; |
1da177e4 LT |
498 | |
499 | spin_lock(&mq_lock); | |
7eafd7c7 SH |
500 | ipc_ns = __get_ns_from_inode(dir); |
501 | if (!ipc_ns) { | |
502 | error = -EACCES; | |
503 | goto out_unlock; | |
504 | } | |
f3713fd9 DB |
505 | |
506 | if (ipc_ns->mq_queues_count >= ipc_ns->mq_queues_max && | |
507 | !capable(CAP_SYS_RESOURCE)) { | |
1da177e4 | 508 | error = -ENOSPC; |
614b84cf | 509 | goto out_unlock; |
1da177e4 | 510 | } |
614b84cf | 511 | ipc_ns->mq_queues_count++; |
1da177e4 LT |
512 | spin_unlock(&mq_lock); |
513 | ||
7eafd7c7 | 514 | inode = mqueue_get_inode(dir->i_sb, ipc_ns, mode, attr); |
d40dcdb0 JS |
515 | if (IS_ERR(inode)) { |
516 | error = PTR_ERR(inode); | |
1da177e4 | 517 | spin_lock(&mq_lock); |
614b84cf SH |
518 | ipc_ns->mq_queues_count--; |
519 | goto out_unlock; | |
1da177e4 LT |
520 | } |
521 | ||
7eafd7c7 | 522 | put_ipc_ns(ipc_ns); |
1da177e4 | 523 | dir->i_size += DIRENT_SIZE; |
078cd827 | 524 | dir->i_ctime = dir->i_mtime = dir->i_atime = current_time(dir); |
1da177e4 LT |
525 | |
526 | d_instantiate(dentry, inode); | |
527 | dget(dentry); | |
528 | return 0; | |
614b84cf | 529 | out_unlock: |
1da177e4 | 530 | spin_unlock(&mq_lock); |
7eafd7c7 SH |
531 | if (ipc_ns) |
532 | put_ipc_ns(ipc_ns); | |
1da177e4 LT |
533 | return error; |
534 | } | |
535 | ||
eecec19d AV |
536 | static int mqueue_create(struct inode *dir, struct dentry *dentry, |
537 | umode_t mode, bool excl) | |
538 | { | |
539 | return mqueue_create_attr(dentry, mode, NULL); | |
540 | } | |
541 | ||
1da177e4 LT |
542 | static int mqueue_unlink(struct inode *dir, struct dentry *dentry) |
543 | { | |
75c3cfa8 | 544 | struct inode *inode = d_inode(dentry); |
1da177e4 | 545 | |
078cd827 | 546 | dir->i_ctime = dir->i_mtime = dir->i_atime = current_time(dir); |
1da177e4 | 547 | dir->i_size -= DIRENT_SIZE; |
239521f3 MS |
548 | drop_nlink(inode); |
549 | dput(dentry); | |
550 | return 0; | |
1da177e4 LT |
551 | } |
552 | ||
553 | /* | |
554 | * This is routine for system read from queue file. | |
555 | * To avoid mess with doing here some sort of mq_receive we allow | |
556 | * to read only queue size & notification info (the only values | |
557 | * that are interesting from user point of view and aren't accessible | |
558 | * through std routines) | |
559 | */ | |
560 | static ssize_t mqueue_read_file(struct file *filp, char __user *u_data, | |
f1a43f93 | 561 | size_t count, loff_t *off) |
1da177e4 | 562 | { |
496ad9aa | 563 | struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp)); |
1da177e4 | 564 | char buffer[FILENT_SIZE]; |
f1a43f93 | 565 | ssize_t ret; |
1da177e4 LT |
566 | |
567 | spin_lock(&info->lock); | |
568 | snprintf(buffer, sizeof(buffer), | |
569 | "QSIZE:%-10lu NOTIFY:%-5d SIGNO:%-5d NOTIFY_PID:%-6d\n", | |
570 | info->qsize, | |
571 | info->notify_owner ? info->notify.sigev_notify : 0, | |
572 | (info->notify_owner && | |
573 | info->notify.sigev_notify == SIGEV_SIGNAL) ? | |
574 | info->notify.sigev_signo : 0, | |
6c5f3e7b | 575 | pid_vnr(info->notify_owner)); |
1da177e4 LT |
576 | spin_unlock(&info->lock); |
577 | buffer[sizeof(buffer)-1] = '\0'; | |
1da177e4 | 578 | |
f1a43f93 AM |
579 | ret = simple_read_from_buffer(u_data, count, off, buffer, |
580 | strlen(buffer)); | |
581 | if (ret <= 0) | |
582 | return ret; | |
1da177e4 | 583 | |
078cd827 | 584 | file_inode(filp)->i_atime = file_inode(filp)->i_ctime = current_time(file_inode(filp)); |
f1a43f93 | 585 | return ret; |
1da177e4 LT |
586 | } |
587 | ||
75e1fcc0 | 588 | static int mqueue_flush_file(struct file *filp, fl_owner_t id) |
1da177e4 | 589 | { |
496ad9aa | 590 | struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp)); |
1da177e4 LT |
591 | |
592 | spin_lock(&info->lock); | |
a03fcb73 | 593 | if (task_tgid(current) == info->notify_owner) |
1da177e4 LT |
594 | remove_notification(info); |
595 | ||
596 | spin_unlock(&info->lock); | |
597 | return 0; | |
598 | } | |
599 | ||
9dd95748 | 600 | static __poll_t mqueue_poll_file(struct file *filp, struct poll_table_struct *poll_tab) |
1da177e4 | 601 | { |
496ad9aa | 602 | struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp)); |
9dd95748 | 603 | __poll_t retval = 0; |
1da177e4 LT |
604 | |
605 | poll_wait(filp, &info->wait_q, poll_tab); | |
606 | ||
607 | spin_lock(&info->lock); | |
608 | if (info->attr.mq_curmsgs) | |
a9a08845 | 609 | retval = EPOLLIN | EPOLLRDNORM; |
1da177e4 LT |
610 | |
611 | if (info->attr.mq_curmsgs < info->attr.mq_maxmsg) | |
a9a08845 | 612 | retval |= EPOLLOUT | EPOLLWRNORM; |
1da177e4 LT |
613 | spin_unlock(&info->lock); |
614 | ||
615 | return retval; | |
616 | } | |
617 | ||
618 | /* Adds current to info->e_wait_q[sr] before element with smaller prio */ | |
619 | static void wq_add(struct mqueue_inode_info *info, int sr, | |
620 | struct ext_wait_queue *ewp) | |
621 | { | |
622 | struct ext_wait_queue *walk; | |
623 | ||
1da177e4 | 624 | list_for_each_entry(walk, &info->e_wait_q[sr].list, list) { |
68e34f4e | 625 | if (walk->task->prio <= current->prio) { |
1da177e4 LT |
626 | list_add_tail(&ewp->list, &walk->list); |
627 | return; | |
628 | } | |
629 | } | |
630 | list_add_tail(&ewp->list, &info->e_wait_q[sr].list); | |
631 | } | |
632 | ||
633 | /* | |
634 | * Puts current task to sleep. Caller must hold queue lock. After return | |
635 | * lock isn't held. | |
636 | * sr: SEND or RECV | |
637 | */ | |
638 | static int wq_sleep(struct mqueue_inode_info *info, int sr, | |
9ca7d8e6 | 639 | ktime_t *timeout, struct ext_wait_queue *ewp) |
eac0b1c3 | 640 | __releases(&info->lock) |
1da177e4 LT |
641 | { |
642 | int retval; | |
643 | signed long time; | |
644 | ||
645 | wq_add(info, sr, ewp); | |
646 | ||
647 | for (;;) { | |
fa6004ad | 648 | __set_current_state(TASK_INTERRUPTIBLE); |
1da177e4 LT |
649 | |
650 | spin_unlock(&info->lock); | |
32ea845d WG |
651 | time = schedule_hrtimeout_range_clock(timeout, 0, |
652 | HRTIMER_MODE_ABS, CLOCK_REALTIME); | |
1da177e4 | 653 | |
1da177e4 LT |
654 | if (ewp->state == STATE_READY) { |
655 | retval = 0; | |
656 | goto out; | |
657 | } | |
658 | spin_lock(&info->lock); | |
659 | if (ewp->state == STATE_READY) { | |
660 | retval = 0; | |
661 | goto out_unlock; | |
662 | } | |
663 | if (signal_pending(current)) { | |
664 | retval = -ERESTARTSYS; | |
665 | break; | |
666 | } | |
667 | if (time == 0) { | |
668 | retval = -ETIMEDOUT; | |
669 | break; | |
670 | } | |
671 | } | |
672 | list_del(&ewp->list); | |
673 | out_unlock: | |
674 | spin_unlock(&info->lock); | |
675 | out: | |
676 | return retval; | |
677 | } | |
678 | ||
679 | /* | |
680 | * Returns waiting task that should be serviced first or NULL if none exists | |
681 | */ | |
682 | static struct ext_wait_queue *wq_get_first_waiter( | |
683 | struct mqueue_inode_info *info, int sr) | |
684 | { | |
685 | struct list_head *ptr; | |
686 | ||
687 | ptr = info->e_wait_q[sr].list.prev; | |
688 | if (ptr == &info->e_wait_q[sr].list) | |
689 | return NULL; | |
690 | return list_entry(ptr, struct ext_wait_queue, list); | |
691 | } | |
692 | ||
1da177e4 LT |
693 | |
694 | static inline void set_cookie(struct sk_buff *skb, char code) | |
695 | { | |
239521f3 | 696 | ((char *)skb->data)[NOTIFY_COOKIE_LEN-1] = code; |
1da177e4 LT |
697 | } |
698 | ||
699 | /* | |
700 | * The next function is only to split too long sys_mq_timedsend | |
701 | */ | |
702 | static void __do_notify(struct mqueue_inode_info *info) | |
703 | { | |
704 | /* notification | |
705 | * invoked when there is registered process and there isn't process | |
706 | * waiting synchronously for message AND state of queue changed from | |
707 | * empty to not empty. Here we are sure that no one is waiting | |
708 | * synchronously. */ | |
709 | if (info->notify_owner && | |
710 | info->attr.mq_curmsgs == 1) { | |
ae7795bc | 711 | struct kernel_siginfo sig_i; |
1da177e4 LT |
712 | switch (info->notify.sigev_notify) { |
713 | case SIGEV_NONE: | |
714 | break; | |
715 | case SIGEV_SIGNAL: | |
716 | /* sends signal */ | |
717 | ||
faf1f22b | 718 | clear_siginfo(&sig_i); |
1da177e4 LT |
719 | sig_i.si_signo = info->notify.sigev_signo; |
720 | sig_i.si_errno = 0; | |
721 | sig_i.si_code = SI_MESGQ; | |
722 | sig_i.si_value = info->notify.sigev_value; | |
6b550f94 SH |
723 | /* map current pid/uid into info->owner's namespaces */ |
724 | rcu_read_lock(); | |
a6684999 SB |
725 | sig_i.si_pid = task_tgid_nr_ns(current, |
726 | ns_of_pid(info->notify_owner)); | |
76b6db01 | 727 | sig_i.si_uid = from_kuid_munged(info->notify_user_ns, current_uid()); |
6b550f94 | 728 | rcu_read_unlock(); |
1da177e4 | 729 | |
a03fcb73 CLG |
730 | kill_pid_info(info->notify.sigev_signo, |
731 | &sig_i, info->notify_owner); | |
1da177e4 LT |
732 | break; |
733 | case SIGEV_THREAD: | |
734 | set_cookie(info->notify_cookie, NOTIFY_WOKENUP); | |
7ee015e0 | 735 | netlink_sendskb(info->notify_sock, info->notify_cookie); |
1da177e4 LT |
736 | break; |
737 | } | |
738 | /* after notification unregisters process */ | |
a03fcb73 | 739 | put_pid(info->notify_owner); |
6f9ac6d9 | 740 | put_user_ns(info->notify_user_ns); |
a03fcb73 | 741 | info->notify_owner = NULL; |
6f9ac6d9 | 742 | info->notify_user_ns = NULL; |
1da177e4 LT |
743 | } |
744 | wake_up(&info->wait_q); | |
745 | } | |
746 | ||
21fc538d | 747 | static int prepare_timeout(const struct __kernel_timespec __user *u_abs_timeout, |
b9047726 | 748 | struct timespec64 *ts) |
1da177e4 | 749 | { |
b9047726 | 750 | if (get_timespec64(ts, u_abs_timeout)) |
9ca7d8e6 | 751 | return -EFAULT; |
b9047726 | 752 | if (!timespec64_valid(ts)) |
9ca7d8e6 | 753 | return -EINVAL; |
9ca7d8e6 | 754 | return 0; |
1da177e4 LT |
755 | } |
756 | ||
757 | static void remove_notification(struct mqueue_inode_info *info) | |
758 | { | |
a03fcb73 | 759 | if (info->notify_owner != NULL && |
1da177e4 LT |
760 | info->notify.sigev_notify == SIGEV_THREAD) { |
761 | set_cookie(info->notify_cookie, NOTIFY_REMOVED); | |
7ee015e0 | 762 | netlink_sendskb(info->notify_sock, info->notify_cookie); |
1da177e4 | 763 | } |
a03fcb73 | 764 | put_pid(info->notify_owner); |
6f9ac6d9 | 765 | put_user_ns(info->notify_user_ns); |
a03fcb73 | 766 | info->notify_owner = NULL; |
6f9ac6d9 | 767 | info->notify_user_ns = NULL; |
1da177e4 LT |
768 | } |
769 | ||
066cc813 AV |
770 | static int prepare_open(struct dentry *dentry, int oflag, int ro, |
771 | umode_t mode, struct filename *name, | |
614b84cf | 772 | struct mq_attr *attr) |
1da177e4 | 773 | { |
745ca247 DH |
774 | static const int oflag2acc[O_ACCMODE] = { MAY_READ, MAY_WRITE, |
775 | MAY_READ | MAY_WRITE }; | |
765927b2 | 776 | int acc; |
066cc813 | 777 | |
9b20d7fc AV |
778 | if (d_really_is_negative(dentry)) { |
779 | if (!(oflag & O_CREAT)) | |
066cc813 | 780 | return -ENOENT; |
9b20d7fc AV |
781 | if (ro) |
782 | return ro; | |
783 | audit_inode_parent_hidden(name, dentry->d_parent); | |
784 | return vfs_mkobj(dentry, mode & ~current_umask(), | |
785 | mqueue_create_attr, attr); | |
066cc813 | 786 | } |
9b20d7fc AV |
787 | /* it already existed */ |
788 | audit_inode(name, dentry, 0); | |
789 | if ((oflag & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL)) | |
790 | return -EEXIST; | |
765927b2 | 791 | if ((oflag & O_ACCMODE) == (O_RDWR | O_WRONLY)) |
af4a5372 | 792 | return -EINVAL; |
765927b2 | 793 | acc = oflag2acc[oflag & O_ACCMODE]; |
066cc813 | 794 | return inode_permission(d_inode(dentry), acc); |
1da177e4 LT |
795 | } |
796 | ||
0d060606 AV |
797 | static int do_mq_open(const char __user *u_name, int oflag, umode_t mode, |
798 | struct mq_attr *attr) | |
1da177e4 | 799 | { |
cfb2f6f6 EB |
800 | struct vfsmount *mnt = current->nsproxy->ipc_ns->mq_mnt; |
801 | struct dentry *root = mnt->mnt_root; | |
91a27b2a | 802 | struct filename *name; |
a713fd7f | 803 | struct path path; |
1da177e4 | 804 | int fd, error; |
312b90fb | 805 | int ro; |
1da177e4 | 806 | |
0d060606 | 807 | audit_mq_open(oflag, mode, attr); |
20ca73bc | 808 | |
1da177e4 LT |
809 | if (IS_ERR(name = getname(u_name))) |
810 | return PTR_ERR(name); | |
811 | ||
269f2134 | 812 | fd = get_unused_fd_flags(O_CLOEXEC); |
1da177e4 LT |
813 | if (fd < 0) |
814 | goto out_putname; | |
815 | ||
312b90fb | 816 | ro = mnt_want_write(mnt); /* we'll drop it in any case */ |
5955102c | 817 | inode_lock(d_inode(root)); |
91a27b2a | 818 | path.dentry = lookup_one_len(name->name, root, strlen(name->name)); |
765927b2 AV |
819 | if (IS_ERR(path.dentry)) { |
820 | error = PTR_ERR(path.dentry); | |
4294a8ee | 821 | goto out_putfd; |
1da177e4 | 822 | } |
312b90fb | 823 | path.mnt = mntget(mnt); |
066cc813 AV |
824 | error = prepare_open(path.dentry, oflag, ro, mode, name, attr); |
825 | if (!error) { | |
826 | struct file *file = dentry_open(&path, oflag, current_cred()); | |
827 | if (!IS_ERR(file)) | |
828 | fd_install(fd, file); | |
829 | else | |
830 | error = PTR_ERR(file); | |
7c7dce92 | 831 | } |
765927b2 | 832 | path_put(&path); |
7c7dce92 | 833 | out_putfd: |
765927b2 AV |
834 | if (error) { |
835 | put_unused_fd(fd); | |
836 | fd = error; | |
837 | } | |
5955102c | 838 | inode_unlock(d_inode(root)); |
38d78e58 VD |
839 | if (!ro) |
840 | mnt_drop_write(mnt); | |
1da177e4 LT |
841 | out_putname: |
842 | putname(name); | |
843 | return fd; | |
844 | } | |
845 | ||
0d060606 AV |
846 | SYSCALL_DEFINE4(mq_open, const char __user *, u_name, int, oflag, umode_t, mode, |
847 | struct mq_attr __user *, u_attr) | |
848 | { | |
849 | struct mq_attr attr; | |
850 | if (u_attr && copy_from_user(&attr, u_attr, sizeof(struct mq_attr))) | |
851 | return -EFAULT; | |
852 | ||
853 | return do_mq_open(u_name, oflag, mode, u_attr ? &attr : NULL); | |
854 | } | |
855 | ||
d5460c99 | 856 | SYSCALL_DEFINE1(mq_unlink, const char __user *, u_name) |
1da177e4 LT |
857 | { |
858 | int err; | |
91a27b2a | 859 | struct filename *name; |
1da177e4 LT |
860 | struct dentry *dentry; |
861 | struct inode *inode = NULL; | |
7eafd7c7 | 862 | struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns; |
312b90fb | 863 | struct vfsmount *mnt = ipc_ns->mq_mnt; |
1da177e4 LT |
864 | |
865 | name = getname(u_name); | |
866 | if (IS_ERR(name)) | |
867 | return PTR_ERR(name); | |
868 | ||
79f6530c | 869 | audit_inode_parent_hidden(name, mnt->mnt_root); |
312b90fb AV |
870 | err = mnt_want_write(mnt); |
871 | if (err) | |
872 | goto out_name; | |
5955102c | 873 | inode_lock_nested(d_inode(mnt->mnt_root), I_MUTEX_PARENT); |
91a27b2a JL |
874 | dentry = lookup_one_len(name->name, mnt->mnt_root, |
875 | strlen(name->name)); | |
1da177e4 LT |
876 | if (IS_ERR(dentry)) { |
877 | err = PTR_ERR(dentry); | |
878 | goto out_unlock; | |
879 | } | |
880 | ||
75c3cfa8 | 881 | inode = d_inode(dentry); |
312b90fb AV |
882 | if (!inode) { |
883 | err = -ENOENT; | |
884 | } else { | |
7de9c6ee | 885 | ihold(inode); |
75c3cfa8 | 886 | err = vfs_unlink(d_inode(dentry->d_parent), dentry, NULL); |
312b90fb | 887 | } |
1da177e4 LT |
888 | dput(dentry); |
889 | ||
890 | out_unlock: | |
5955102c | 891 | inode_unlock(d_inode(mnt->mnt_root)); |
1da177e4 LT |
892 | if (inode) |
893 | iput(inode); | |
312b90fb AV |
894 | mnt_drop_write(mnt); |
895 | out_name: | |
896 | putname(name); | |
1da177e4 LT |
897 | |
898 | return err; | |
899 | } | |
900 | ||
901 | /* Pipelined send and receive functions. | |
902 | * | |
903 | * If a receiver finds no waiting message, then it registers itself in the | |
904 | * list of waiting receivers. A sender checks that list before adding the new | |
905 | * message into the message array. If there is a waiting receiver, then it | |
906 | * bypasses the message array and directly hands the message over to the | |
fa6004ad DB |
907 | * receiver. The receiver accepts the message and returns without grabbing the |
908 | * queue spinlock: | |
909 | * | |
910 | * - Set pointer to message. | |
911 | * - Queue the receiver task for later wakeup (without the info->lock). | |
912 | * - Update its state to STATE_READY. Now the receiver can continue. | |
913 | * - Wake up the process after the lock is dropped. Should the process wake up | |
914 | * before this wakeup (due to a timeout or a signal) it will either see | |
915 | * STATE_READY and continue or acquire the lock to check the state again. | |
1da177e4 LT |
916 | * |
917 | * The same algorithm is used for senders. | |
918 | */ | |
919 | ||
920 | /* pipelined_send() - send a message directly to the task waiting in | |
921 | * sys_mq_timedreceive() (without inserting message into a queue). | |
922 | */ | |
fa6004ad DB |
923 | static inline void pipelined_send(struct wake_q_head *wake_q, |
924 | struct mqueue_inode_info *info, | |
1da177e4 LT |
925 | struct msg_msg *message, |
926 | struct ext_wait_queue *receiver) | |
927 | { | |
928 | receiver->msg = message; | |
929 | list_del(&receiver->list); | |
fa6004ad DB |
930 | wake_q_add(wake_q, receiver->task); |
931 | /* | |
932 | * Rely on the implicit cmpxchg barrier from wake_q_add such | |
933 | * that we can ensure that updating receiver->state is the last | |
934 | * write operation: As once set, the receiver can continue, | |
935 | * and if we don't have the reference count from the wake_q, | |
936 | * yet, at that point we can later have a use-after-free | |
937 | * condition and bogus wakeup. | |
938 | */ | |
1da177e4 LT |
939 | receiver->state = STATE_READY; |
940 | } | |
941 | ||
942 | /* pipelined_receive() - if there is task waiting in sys_mq_timedsend() | |
943 | * gets its message and put to the queue (we have one free place for sure). */ | |
fa6004ad DB |
944 | static inline void pipelined_receive(struct wake_q_head *wake_q, |
945 | struct mqueue_inode_info *info) | |
1da177e4 LT |
946 | { |
947 | struct ext_wait_queue *sender = wq_get_first_waiter(info, SEND); | |
948 | ||
949 | if (!sender) { | |
950 | /* for poll */ | |
951 | wake_up_interruptible(&info->wait_q); | |
952 | return; | |
953 | } | |
d6629859 DL |
954 | if (msg_insert(sender->msg, info)) |
955 | return; | |
fa6004ad | 956 | |
1da177e4 | 957 | list_del(&sender->list); |
fa6004ad | 958 | wake_q_add(wake_q, sender->task); |
1da177e4 LT |
959 | sender->state = STATE_READY; |
960 | } | |
961 | ||
0d060606 AV |
962 | static int do_mq_timedsend(mqd_t mqdes, const char __user *u_msg_ptr, |
963 | size_t msg_len, unsigned int msg_prio, | |
b9047726 | 964 | struct timespec64 *ts) |
1da177e4 | 965 | { |
2903ff01 | 966 | struct fd f; |
1da177e4 LT |
967 | struct inode *inode; |
968 | struct ext_wait_queue wait; | |
969 | struct ext_wait_queue *receiver; | |
970 | struct msg_msg *msg_ptr; | |
971 | struct mqueue_inode_info *info; | |
9ca7d8e6 | 972 | ktime_t expires, *timeout = NULL; |
ce2d52cc | 973 | struct posix_msg_tree_node *new_leaf = NULL; |
2903ff01 | 974 | int ret = 0; |
194a6b5b | 975 | DEFINE_WAKE_Q(wake_q); |
1da177e4 LT |
976 | |
977 | if (unlikely(msg_prio >= (unsigned long) MQ_PRIO_MAX)) | |
978 | return -EINVAL; | |
979 | ||
0d060606 | 980 | if (ts) { |
b9047726 | 981 | expires = timespec64_to_ktime(*ts); |
0d060606 AV |
982 | timeout = &expires; |
983 | } | |
984 | ||
985 | audit_mq_sendrecv(mqdes, msg_len, msg_prio, ts); | |
1da177e4 | 986 | |
2903ff01 AV |
987 | f = fdget(mqdes); |
988 | if (unlikely(!f.file)) { | |
8d8ffefa | 989 | ret = -EBADF; |
1da177e4 | 990 | goto out; |
8d8ffefa | 991 | } |
1da177e4 | 992 | |
496ad9aa | 993 | inode = file_inode(f.file); |
2903ff01 | 994 | if (unlikely(f.file->f_op != &mqueue_file_operations)) { |
8d8ffefa | 995 | ret = -EBADF; |
1da177e4 | 996 | goto out_fput; |
8d8ffefa | 997 | } |
1da177e4 | 998 | info = MQUEUE_I(inode); |
9f45f5bf | 999 | audit_file(f.file); |
1da177e4 | 1000 | |
2903ff01 | 1001 | if (unlikely(!(f.file->f_mode & FMODE_WRITE))) { |
8d8ffefa | 1002 | ret = -EBADF; |
1da177e4 | 1003 | goto out_fput; |
8d8ffefa | 1004 | } |
1da177e4 LT |
1005 | |
1006 | if (unlikely(msg_len > info->attr.mq_msgsize)) { | |
1007 | ret = -EMSGSIZE; | |
1008 | goto out_fput; | |
1009 | } | |
1010 | ||
1011 | /* First try to allocate memory, before doing anything with | |
1012 | * existing queues. */ | |
1013 | msg_ptr = load_msg(u_msg_ptr, msg_len); | |
1014 | if (IS_ERR(msg_ptr)) { | |
1015 | ret = PTR_ERR(msg_ptr); | |
1016 | goto out_fput; | |
1017 | } | |
1018 | msg_ptr->m_ts = msg_len; | |
1019 | msg_ptr->m_type = msg_prio; | |
1020 | ||
ce2d52cc DL |
1021 | /* |
1022 | * msg_insert really wants us to have a valid, spare node struct so | |
1023 | * it doesn't have to kmalloc a GFP_ATOMIC allocation, but it will | |
1024 | * fall back to that if necessary. | |
1025 | */ | |
1026 | if (!info->node_cache) | |
1027 | new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL); | |
1028 | ||
1da177e4 LT |
1029 | spin_lock(&info->lock); |
1030 | ||
ce2d52cc DL |
1031 | if (!info->node_cache && new_leaf) { |
1032 | /* Save our speculative allocation into the cache */ | |
ce2d52cc DL |
1033 | INIT_LIST_HEAD(&new_leaf->msg_list); |
1034 | info->node_cache = new_leaf; | |
ce2d52cc DL |
1035 | new_leaf = NULL; |
1036 | } else { | |
1037 | kfree(new_leaf); | |
1038 | } | |
1039 | ||
1da177e4 | 1040 | if (info->attr.mq_curmsgs == info->attr.mq_maxmsg) { |
2903ff01 | 1041 | if (f.file->f_flags & O_NONBLOCK) { |
1da177e4 | 1042 | ret = -EAGAIN; |
1da177e4 LT |
1043 | } else { |
1044 | wait.task = current; | |
1045 | wait.msg = (void *) msg_ptr; | |
1046 | wait.state = STATE_NONE; | |
1047 | ret = wq_sleep(info, SEND, timeout, &wait); | |
ce2d52cc DL |
1048 | /* |
1049 | * wq_sleep must be called with info->lock held, and | |
1050 | * returns with the lock released | |
1051 | */ | |
1052 | goto out_free; | |
1da177e4 | 1053 | } |
1da177e4 LT |
1054 | } else { |
1055 | receiver = wq_get_first_waiter(info, RECV); | |
1056 | if (receiver) { | |
fa6004ad | 1057 | pipelined_send(&wake_q, info, msg_ptr, receiver); |
1da177e4 LT |
1058 | } else { |
1059 | /* adds message to the queue */ | |
ce2d52cc DL |
1060 | ret = msg_insert(msg_ptr, info); |
1061 | if (ret) | |
1062 | goto out_unlock; | |
1da177e4 LT |
1063 | __do_notify(info); |
1064 | } | |
1065 | inode->i_atime = inode->i_mtime = inode->i_ctime = | |
078cd827 | 1066 | current_time(inode); |
1da177e4 | 1067 | } |
ce2d52cc DL |
1068 | out_unlock: |
1069 | spin_unlock(&info->lock); | |
fa6004ad | 1070 | wake_up_q(&wake_q); |
ce2d52cc DL |
1071 | out_free: |
1072 | if (ret) | |
1073 | free_msg(msg_ptr); | |
1da177e4 | 1074 | out_fput: |
2903ff01 | 1075 | fdput(f); |
1da177e4 LT |
1076 | out: |
1077 | return ret; | |
1078 | } | |
1079 | ||
0d060606 AV |
1080 | static int do_mq_timedreceive(mqd_t mqdes, char __user *u_msg_ptr, |
1081 | size_t msg_len, unsigned int __user *u_msg_prio, | |
b9047726 | 1082 | struct timespec64 *ts) |
1da177e4 | 1083 | { |
1da177e4 LT |
1084 | ssize_t ret; |
1085 | struct msg_msg *msg_ptr; | |
2903ff01 | 1086 | struct fd f; |
1da177e4 LT |
1087 | struct inode *inode; |
1088 | struct mqueue_inode_info *info; | |
1089 | struct ext_wait_queue wait; | |
9ca7d8e6 | 1090 | ktime_t expires, *timeout = NULL; |
ce2d52cc | 1091 | struct posix_msg_tree_node *new_leaf = NULL; |
1da177e4 | 1092 | |
0d060606 | 1093 | if (ts) { |
b9047726 | 1094 | expires = timespec64_to_ktime(*ts); |
9ca7d8e6 | 1095 | timeout = &expires; |
c32c8af4 | 1096 | } |
20ca73bc | 1097 | |
0d060606 | 1098 | audit_mq_sendrecv(mqdes, msg_len, 0, ts); |
1da177e4 | 1099 | |
2903ff01 AV |
1100 | f = fdget(mqdes); |
1101 | if (unlikely(!f.file)) { | |
8d8ffefa | 1102 | ret = -EBADF; |
1da177e4 | 1103 | goto out; |
8d8ffefa | 1104 | } |
1da177e4 | 1105 | |
496ad9aa | 1106 | inode = file_inode(f.file); |
2903ff01 | 1107 | if (unlikely(f.file->f_op != &mqueue_file_operations)) { |
8d8ffefa | 1108 | ret = -EBADF; |
1da177e4 | 1109 | goto out_fput; |
8d8ffefa | 1110 | } |
1da177e4 | 1111 | info = MQUEUE_I(inode); |
9f45f5bf | 1112 | audit_file(f.file); |
1da177e4 | 1113 | |
2903ff01 | 1114 | if (unlikely(!(f.file->f_mode & FMODE_READ))) { |
8d8ffefa | 1115 | ret = -EBADF; |
1da177e4 | 1116 | goto out_fput; |
8d8ffefa | 1117 | } |
1da177e4 LT |
1118 | |
1119 | /* checks if buffer is big enough */ | |
1120 | if (unlikely(msg_len < info->attr.mq_msgsize)) { | |
1121 | ret = -EMSGSIZE; | |
1122 | goto out_fput; | |
1123 | } | |
1124 | ||
ce2d52cc DL |
1125 | /* |
1126 | * msg_insert really wants us to have a valid, spare node struct so | |
1127 | * it doesn't have to kmalloc a GFP_ATOMIC allocation, but it will | |
1128 | * fall back to that if necessary. | |
1129 | */ | |
1130 | if (!info->node_cache) | |
1131 | new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL); | |
1132 | ||
1da177e4 | 1133 | spin_lock(&info->lock); |
ce2d52cc DL |
1134 | |
1135 | if (!info->node_cache && new_leaf) { | |
1136 | /* Save our speculative allocation into the cache */ | |
ce2d52cc DL |
1137 | INIT_LIST_HEAD(&new_leaf->msg_list); |
1138 | info->node_cache = new_leaf; | |
ce2d52cc DL |
1139 | } else { |
1140 | kfree(new_leaf); | |
1141 | } | |
1142 | ||
1da177e4 | 1143 | if (info->attr.mq_curmsgs == 0) { |
2903ff01 | 1144 | if (f.file->f_flags & O_NONBLOCK) { |
1da177e4 LT |
1145 | spin_unlock(&info->lock); |
1146 | ret = -EAGAIN; | |
1da177e4 LT |
1147 | } else { |
1148 | wait.task = current; | |
1149 | wait.state = STATE_NONE; | |
1150 | ret = wq_sleep(info, RECV, timeout, &wait); | |
1151 | msg_ptr = wait.msg; | |
1152 | } | |
1153 | } else { | |
194a6b5b | 1154 | DEFINE_WAKE_Q(wake_q); |
fa6004ad | 1155 | |
1da177e4 LT |
1156 | msg_ptr = msg_get(info); |
1157 | ||
1158 | inode->i_atime = inode->i_mtime = inode->i_ctime = | |
078cd827 | 1159 | current_time(inode); |
1da177e4 LT |
1160 | |
1161 | /* There is now free space in queue. */ | |
fa6004ad | 1162 | pipelined_receive(&wake_q, info); |
1da177e4 | 1163 | spin_unlock(&info->lock); |
fa6004ad | 1164 | wake_up_q(&wake_q); |
1da177e4 LT |
1165 | ret = 0; |
1166 | } | |
1167 | if (ret == 0) { | |
1168 | ret = msg_ptr->m_ts; | |
1169 | ||
1170 | if ((u_msg_prio && put_user(msg_ptr->m_type, u_msg_prio)) || | |
1171 | store_msg(u_msg_ptr, msg_ptr, msg_ptr->m_ts)) { | |
1172 | ret = -EFAULT; | |
1173 | } | |
1174 | free_msg(msg_ptr); | |
1175 | } | |
1176 | out_fput: | |
2903ff01 | 1177 | fdput(f); |
1da177e4 LT |
1178 | out: |
1179 | return ret; | |
1180 | } | |
1181 | ||
0d060606 AV |
1182 | SYSCALL_DEFINE5(mq_timedsend, mqd_t, mqdes, const char __user *, u_msg_ptr, |
1183 | size_t, msg_len, unsigned int, msg_prio, | |
21fc538d | 1184 | const struct __kernel_timespec __user *, u_abs_timeout) |
0d060606 | 1185 | { |
b9047726 | 1186 | struct timespec64 ts, *p = NULL; |
0d060606 AV |
1187 | if (u_abs_timeout) { |
1188 | int res = prepare_timeout(u_abs_timeout, &ts); | |
1189 | if (res) | |
1190 | return res; | |
1191 | p = &ts; | |
1192 | } | |
1193 | return do_mq_timedsend(mqdes, u_msg_ptr, msg_len, msg_prio, p); | |
1194 | } | |
1195 | ||
1196 | SYSCALL_DEFINE5(mq_timedreceive, mqd_t, mqdes, char __user *, u_msg_ptr, | |
1197 | size_t, msg_len, unsigned int __user *, u_msg_prio, | |
21fc538d | 1198 | const struct __kernel_timespec __user *, u_abs_timeout) |
0d060606 | 1199 | { |
b9047726 | 1200 | struct timespec64 ts, *p = NULL; |
0d060606 AV |
1201 | if (u_abs_timeout) { |
1202 | int res = prepare_timeout(u_abs_timeout, &ts); | |
1203 | if (res) | |
1204 | return res; | |
1205 | p = &ts; | |
1206 | } | |
1207 | return do_mq_timedreceive(mqdes, u_msg_ptr, msg_len, u_msg_prio, p); | |
1208 | } | |
1209 | ||
1da177e4 LT |
1210 | /* |
1211 | * Notes: the case when user wants us to deregister (with NULL as pointer) | |
1212 | * and he isn't currently owner of notification, will be silently discarded. | |
1213 | * It isn't explicitly defined in the POSIX. | |
1214 | */ | |
0d060606 | 1215 | static int do_mq_notify(mqd_t mqdes, const struct sigevent *notification) |
1da177e4 | 1216 | { |
2903ff01 AV |
1217 | int ret; |
1218 | struct fd f; | |
1da177e4 LT |
1219 | struct sock *sock; |
1220 | struct inode *inode; | |
1da177e4 LT |
1221 | struct mqueue_inode_info *info; |
1222 | struct sk_buff *nc; | |
1223 | ||
0d060606 | 1224 | audit_mq_notify(mqdes, notification); |
1da177e4 | 1225 | |
20114f71 AV |
1226 | nc = NULL; |
1227 | sock = NULL; | |
0d060606 AV |
1228 | if (notification != NULL) { |
1229 | if (unlikely(notification->sigev_notify != SIGEV_NONE && | |
1230 | notification->sigev_notify != SIGEV_SIGNAL && | |
1231 | notification->sigev_notify != SIGEV_THREAD)) | |
1da177e4 | 1232 | return -EINVAL; |
0d060606 AV |
1233 | if (notification->sigev_notify == SIGEV_SIGNAL && |
1234 | !valid_signal(notification->sigev_signo)) { | |
1da177e4 LT |
1235 | return -EINVAL; |
1236 | } | |
0d060606 | 1237 | if (notification->sigev_notify == SIGEV_THREAD) { |
c3d8d1e3 PM |
1238 | long timeo; |
1239 | ||
1da177e4 LT |
1240 | /* create the notify skb */ |
1241 | nc = alloc_skb(NOTIFY_COOKIE_LEN, GFP_KERNEL); | |
8d8ffefa AGR |
1242 | if (!nc) { |
1243 | ret = -ENOMEM; | |
1da177e4 | 1244 | goto out; |
8d8ffefa | 1245 | } |
1da177e4 | 1246 | if (copy_from_user(nc->data, |
0d060606 | 1247 | notification->sigev_value.sival_ptr, |
1da177e4 | 1248 | NOTIFY_COOKIE_LEN)) { |
8d8ffefa | 1249 | ret = -EFAULT; |
1da177e4 LT |
1250 | goto out; |
1251 | } | |
1252 | ||
1253 | /* TODO: add a header? */ | |
1254 | skb_put(nc, NOTIFY_COOKIE_LEN); | |
1255 | /* and attach it to the socket */ | |
1256 | retry: | |
0d060606 | 1257 | f = fdget(notification->sigev_signo); |
2903ff01 | 1258 | if (!f.file) { |
8d8ffefa | 1259 | ret = -EBADF; |
1da177e4 | 1260 | goto out; |
8d8ffefa | 1261 | } |
2903ff01 AV |
1262 | sock = netlink_getsockbyfilp(f.file); |
1263 | fdput(f); | |
1da177e4 LT |
1264 | if (IS_ERR(sock)) { |
1265 | ret = PTR_ERR(sock); | |
1266 | sock = NULL; | |
1267 | goto out; | |
1268 | } | |
1269 | ||
c3d8d1e3 | 1270 | timeo = MAX_SCHEDULE_TIMEOUT; |
9457afee | 1271 | ret = netlink_attachskb(sock, nc, &timeo, NULL); |
f991af3d CW |
1272 | if (ret == 1) { |
1273 | sock = NULL; | |
8d8ffefa | 1274 | goto retry; |
f991af3d | 1275 | } |
1da177e4 LT |
1276 | if (ret) { |
1277 | sock = NULL; | |
1278 | nc = NULL; | |
1279 | goto out; | |
1280 | } | |
1281 | } | |
1282 | } | |
1283 | ||
2903ff01 AV |
1284 | f = fdget(mqdes); |
1285 | if (!f.file) { | |
8d8ffefa | 1286 | ret = -EBADF; |
1da177e4 | 1287 | goto out; |
8d8ffefa | 1288 | } |
1da177e4 | 1289 | |
496ad9aa | 1290 | inode = file_inode(f.file); |
2903ff01 | 1291 | if (unlikely(f.file->f_op != &mqueue_file_operations)) { |
8d8ffefa | 1292 | ret = -EBADF; |
1da177e4 | 1293 | goto out_fput; |
8d8ffefa | 1294 | } |
1da177e4 LT |
1295 | info = MQUEUE_I(inode); |
1296 | ||
1297 | ret = 0; | |
1298 | spin_lock(&info->lock); | |
0d060606 | 1299 | if (notification == NULL) { |
a03fcb73 | 1300 | if (info->notify_owner == task_tgid(current)) { |
1da177e4 | 1301 | remove_notification(info); |
078cd827 | 1302 | inode->i_atime = inode->i_ctime = current_time(inode); |
1da177e4 | 1303 | } |
a03fcb73 | 1304 | } else if (info->notify_owner != NULL) { |
1da177e4 LT |
1305 | ret = -EBUSY; |
1306 | } else { | |
0d060606 | 1307 | switch (notification->sigev_notify) { |
1da177e4 LT |
1308 | case SIGEV_NONE: |
1309 | info->notify.sigev_notify = SIGEV_NONE; | |
1310 | break; | |
1311 | case SIGEV_THREAD: | |
1312 | info->notify_sock = sock; | |
1313 | info->notify_cookie = nc; | |
1314 | sock = NULL; | |
1315 | nc = NULL; | |
1316 | info->notify.sigev_notify = SIGEV_THREAD; | |
1317 | break; | |
1318 | case SIGEV_SIGNAL: | |
0d060606 AV |
1319 | info->notify.sigev_signo = notification->sigev_signo; |
1320 | info->notify.sigev_value = notification->sigev_value; | |
1da177e4 LT |
1321 | info->notify.sigev_notify = SIGEV_SIGNAL; |
1322 | break; | |
1323 | } | |
a03fcb73 CLG |
1324 | |
1325 | info->notify_owner = get_pid(task_tgid(current)); | |
6f9ac6d9 | 1326 | info->notify_user_ns = get_user_ns(current_user_ns()); |
078cd827 | 1327 | inode->i_atime = inode->i_ctime = current_time(inode); |
1da177e4 LT |
1328 | } |
1329 | spin_unlock(&info->lock); | |
1330 | out_fput: | |
2903ff01 | 1331 | fdput(f); |
1da177e4 | 1332 | out: |
3ab08fe2 | 1333 | if (sock) |
1da177e4 | 1334 | netlink_detachskb(sock, nc); |
3ab08fe2 | 1335 | else if (nc) |
1da177e4 | 1336 | dev_kfree_skb(nc); |
3ab08fe2 | 1337 | |
1da177e4 LT |
1338 | return ret; |
1339 | } | |
1340 | ||
0d060606 AV |
1341 | SYSCALL_DEFINE2(mq_notify, mqd_t, mqdes, |
1342 | const struct sigevent __user *, u_notification) | |
1343 | { | |
1344 | struct sigevent n, *p = NULL; | |
1345 | if (u_notification) { | |
1346 | if (copy_from_user(&n, u_notification, sizeof(struct sigevent))) | |
1347 | return -EFAULT; | |
1348 | p = &n; | |
1349 | } | |
1350 | return do_mq_notify(mqdes, p); | |
1351 | } | |
1352 | ||
1353 | static int do_mq_getsetattr(int mqdes, struct mq_attr *new, struct mq_attr *old) | |
1da177e4 | 1354 | { |
2903ff01 | 1355 | struct fd f; |
1da177e4 LT |
1356 | struct inode *inode; |
1357 | struct mqueue_inode_info *info; | |
1358 | ||
0d060606 AV |
1359 | if (new && (new->mq_flags & (~O_NONBLOCK))) |
1360 | return -EINVAL; | |
1da177e4 | 1361 | |
2903ff01 | 1362 | f = fdget(mqdes); |
0d060606 AV |
1363 | if (!f.file) |
1364 | return -EBADF; | |
1da177e4 | 1365 | |
2903ff01 | 1366 | if (unlikely(f.file->f_op != &mqueue_file_operations)) { |
0d060606 AV |
1367 | fdput(f); |
1368 | return -EBADF; | |
8d8ffefa | 1369 | } |
0d060606 AV |
1370 | |
1371 | inode = file_inode(f.file); | |
1da177e4 LT |
1372 | info = MQUEUE_I(inode); |
1373 | ||
1374 | spin_lock(&info->lock); | |
1375 | ||
0d060606 AV |
1376 | if (old) { |
1377 | *old = info->attr; | |
1378 | old->mq_flags = f.file->f_flags & O_NONBLOCK; | |
1379 | } | |
1380 | if (new) { | |
1381 | audit_mq_getsetattr(mqdes, new); | |
2903ff01 | 1382 | spin_lock(&f.file->f_lock); |
0d060606 | 1383 | if (new->mq_flags & O_NONBLOCK) |
2903ff01 | 1384 | f.file->f_flags |= O_NONBLOCK; |
1da177e4 | 1385 | else |
2903ff01 AV |
1386 | f.file->f_flags &= ~O_NONBLOCK; |
1387 | spin_unlock(&f.file->f_lock); | |
1da177e4 | 1388 | |
078cd827 | 1389 | inode->i_atime = inode->i_ctime = current_time(inode); |
1da177e4 LT |
1390 | } |
1391 | ||
1392 | spin_unlock(&info->lock); | |
0d060606 AV |
1393 | fdput(f); |
1394 | return 0; | |
1395 | } | |
1da177e4 | 1396 | |
0d060606 AV |
1397 | SYSCALL_DEFINE3(mq_getsetattr, mqd_t, mqdes, |
1398 | const struct mq_attr __user *, u_mqstat, | |
1399 | struct mq_attr __user *, u_omqstat) | |
1400 | { | |
1401 | int ret; | |
1402 | struct mq_attr mqstat, omqstat; | |
1403 | struct mq_attr *new = NULL, *old = NULL; | |
1da177e4 | 1404 | |
0d060606 AV |
1405 | if (u_mqstat) { |
1406 | new = &mqstat; | |
1407 | if (copy_from_user(new, u_mqstat, sizeof(struct mq_attr))) | |
1408 | return -EFAULT; | |
1409 | } | |
1410 | if (u_omqstat) | |
1411 | old = &omqstat; | |
1412 | ||
1413 | ret = do_mq_getsetattr(mqdes, new, old); | |
1414 | if (ret || !old) | |
1415 | return ret; | |
1416 | ||
1417 | if (copy_to_user(u_omqstat, old, sizeof(struct mq_attr))) | |
1418 | return -EFAULT; | |
1419 | return 0; | |
1420 | } | |
1421 | ||
1422 | #ifdef CONFIG_COMPAT | |
1423 | ||
1424 | struct compat_mq_attr { | |
1425 | compat_long_t mq_flags; /* message queue flags */ | |
1426 | compat_long_t mq_maxmsg; /* maximum number of messages */ | |
1427 | compat_long_t mq_msgsize; /* maximum message size */ | |
1428 | compat_long_t mq_curmsgs; /* number of messages currently queued */ | |
1429 | compat_long_t __reserved[4]; /* ignored for input, zeroed for output */ | |
1430 | }; | |
1431 | ||
1432 | static inline int get_compat_mq_attr(struct mq_attr *attr, | |
1433 | const struct compat_mq_attr __user *uattr) | |
1434 | { | |
1435 | struct compat_mq_attr v; | |
1436 | ||
1437 | if (copy_from_user(&v, uattr, sizeof(*uattr))) | |
1438 | return -EFAULT; | |
1439 | ||
1440 | memset(attr, 0, sizeof(*attr)); | |
1441 | attr->mq_flags = v.mq_flags; | |
1442 | attr->mq_maxmsg = v.mq_maxmsg; | |
1443 | attr->mq_msgsize = v.mq_msgsize; | |
1444 | attr->mq_curmsgs = v.mq_curmsgs; | |
1445 | return 0; | |
1446 | } | |
1447 | ||
1448 | static inline int put_compat_mq_attr(const struct mq_attr *attr, | |
1449 | struct compat_mq_attr __user *uattr) | |
1450 | { | |
1451 | struct compat_mq_attr v; | |
1452 | ||
1453 | memset(&v, 0, sizeof(v)); | |
1454 | v.mq_flags = attr->mq_flags; | |
1455 | v.mq_maxmsg = attr->mq_maxmsg; | |
1456 | v.mq_msgsize = attr->mq_msgsize; | |
1457 | v.mq_curmsgs = attr->mq_curmsgs; | |
1458 | if (copy_to_user(uattr, &v, sizeof(*uattr))) | |
1459 | return -EFAULT; | |
1460 | return 0; | |
1461 | } | |
1462 | ||
1463 | COMPAT_SYSCALL_DEFINE4(mq_open, const char __user *, u_name, | |
1464 | int, oflag, compat_mode_t, mode, | |
1465 | struct compat_mq_attr __user *, u_attr) | |
1466 | { | |
1467 | struct mq_attr attr, *p = NULL; | |
1468 | if (u_attr && oflag & O_CREAT) { | |
1469 | p = &attr; | |
1470 | if (get_compat_mq_attr(&attr, u_attr)) | |
1471 | return -EFAULT; | |
1472 | } | |
1473 | return do_mq_open(u_name, oflag, mode, p); | |
1474 | } | |
1475 | ||
b0d17578 AB |
1476 | COMPAT_SYSCALL_DEFINE2(mq_notify, mqd_t, mqdes, |
1477 | const struct compat_sigevent __user *, u_notification) | |
1478 | { | |
1479 | struct sigevent n, *p = NULL; | |
1480 | if (u_notification) { | |
1481 | if (get_compat_sigevent(&n, u_notification)) | |
1482 | return -EFAULT; | |
1483 | if (n.sigev_notify == SIGEV_THREAD) | |
1484 | n.sigev_value.sival_ptr = compat_ptr(n.sigev_value.sival_int); | |
1485 | p = &n; | |
1486 | } | |
1487 | return do_mq_notify(mqdes, p); | |
1488 | } | |
1489 | ||
1490 | COMPAT_SYSCALL_DEFINE3(mq_getsetattr, mqd_t, mqdes, | |
1491 | const struct compat_mq_attr __user *, u_mqstat, | |
1492 | struct compat_mq_attr __user *, u_omqstat) | |
1493 | { | |
1494 | int ret; | |
1495 | struct mq_attr mqstat, omqstat; | |
1496 | struct mq_attr *new = NULL, *old = NULL; | |
1497 | ||
1498 | if (u_mqstat) { | |
1499 | new = &mqstat; | |
1500 | if (get_compat_mq_attr(new, u_mqstat)) | |
1501 | return -EFAULT; | |
1502 | } | |
1503 | if (u_omqstat) | |
1504 | old = &omqstat; | |
1505 | ||
1506 | ret = do_mq_getsetattr(mqdes, new, old); | |
1507 | if (ret || !old) | |
1508 | return ret; | |
1509 | ||
1510 | if (put_compat_mq_attr(old, u_omqstat)) | |
1511 | return -EFAULT; | |
1512 | return 0; | |
1513 | } | |
1514 | #endif | |
1515 | ||
1516 | #ifdef CONFIG_COMPAT_32BIT_TIME | |
9afc5eee | 1517 | static int compat_prepare_timeout(const struct old_timespec32 __user *p, |
b9047726 | 1518 | struct timespec64 *ts) |
0d060606 | 1519 | { |
9afc5eee | 1520 | if (get_old_timespec32(ts, p)) |
0d060606 | 1521 | return -EFAULT; |
b9047726 | 1522 | if (!timespec64_valid(ts)) |
0d060606 AV |
1523 | return -EINVAL; |
1524 | return 0; | |
1525 | } | |
1526 | ||
8dabe724 AB |
1527 | SYSCALL_DEFINE5(mq_timedsend_time32, mqd_t, mqdes, |
1528 | const char __user *, u_msg_ptr, | |
1529 | unsigned int, msg_len, unsigned int, msg_prio, | |
1530 | const struct old_timespec32 __user *, u_abs_timeout) | |
0d060606 | 1531 | { |
b9047726 | 1532 | struct timespec64 ts, *p = NULL; |
0d060606 AV |
1533 | if (u_abs_timeout) { |
1534 | int res = compat_prepare_timeout(u_abs_timeout, &ts); | |
1535 | if (res) | |
1536 | return res; | |
1537 | p = &ts; | |
1538 | } | |
1539 | return do_mq_timedsend(mqdes, u_msg_ptr, msg_len, msg_prio, p); | |
1540 | } | |
1541 | ||
8dabe724 AB |
1542 | SYSCALL_DEFINE5(mq_timedreceive_time32, mqd_t, mqdes, |
1543 | char __user *, u_msg_ptr, | |
1544 | unsigned int, msg_len, unsigned int __user *, u_msg_prio, | |
1545 | const struct old_timespec32 __user *, u_abs_timeout) | |
0d060606 | 1546 | { |
b9047726 | 1547 | struct timespec64 ts, *p = NULL; |
0d060606 AV |
1548 | if (u_abs_timeout) { |
1549 | int res = compat_prepare_timeout(u_abs_timeout, &ts); | |
1550 | if (res) | |
1551 | return res; | |
1552 | p = &ts; | |
1553 | } | |
1554 | return do_mq_timedreceive(mqdes, u_msg_ptr, msg_len, u_msg_prio, p); | |
1555 | } | |
0d060606 | 1556 | #endif |
1da177e4 | 1557 | |
92e1d5be | 1558 | static const struct inode_operations mqueue_dir_inode_operations = { |
1da177e4 LT |
1559 | .lookup = simple_lookup, |
1560 | .create = mqueue_create, | |
1561 | .unlink = mqueue_unlink, | |
1562 | }; | |
1563 | ||
9a32144e | 1564 | static const struct file_operations mqueue_file_operations = { |
1da177e4 LT |
1565 | .flush = mqueue_flush_file, |
1566 | .poll = mqueue_poll_file, | |
1567 | .read = mqueue_read_file, | |
6038f373 | 1568 | .llseek = default_llseek, |
1da177e4 LT |
1569 | }; |
1570 | ||
b87221de | 1571 | static const struct super_operations mqueue_super_ops = { |
1da177e4 | 1572 | .alloc_inode = mqueue_alloc_inode, |
015d7956 | 1573 | .free_inode = mqueue_free_inode, |
6d8af64c | 1574 | .evict_inode = mqueue_evict_inode, |
1da177e4 | 1575 | .statfs = simple_statfs, |
1da177e4 LT |
1576 | }; |
1577 | ||
935c6912 DH |
1578 | static const struct fs_context_operations mqueue_fs_context_ops = { |
1579 | .free = mqueue_fs_context_free, | |
1580 | .get_tree = mqueue_get_tree, | |
1581 | }; | |
1582 | ||
1da177e4 | 1583 | static struct file_system_type mqueue_fs_type = { |
935c6912 DH |
1584 | .name = "mqueue", |
1585 | .init_fs_context = mqueue_init_fs_context, | |
1586 | .kill_sb = kill_litter_super, | |
1587 | .fs_flags = FS_USERNS_MOUNT, | |
1da177e4 LT |
1588 | }; |
1589 | ||
7eafd7c7 SH |
1590 | int mq_init_ns(struct ipc_namespace *ns) |
1591 | { | |
935c6912 DH |
1592 | struct vfsmount *m; |
1593 | ||
7eafd7c7 SH |
1594 | ns->mq_queues_count = 0; |
1595 | ns->mq_queues_max = DFLT_QUEUESMAX; | |
1596 | ns->mq_msg_max = DFLT_MSGMAX; | |
1597 | ns->mq_msgsize_max = DFLT_MSGSIZEMAX; | |
cef0184c KM |
1598 | ns->mq_msg_default = DFLT_MSG; |
1599 | ns->mq_msgsize_default = DFLT_MSGSIZE; | |
7eafd7c7 | 1600 | |
935c6912 DH |
1601 | m = mq_create_mount(ns); |
1602 | if (IS_ERR(m)) | |
1603 | return PTR_ERR(m); | |
1604 | ns->mq_mnt = m; | |
7eafd7c7 SH |
1605 | return 0; |
1606 | } | |
1607 | ||
1608 | void mq_clear_sbinfo(struct ipc_namespace *ns) | |
1609 | { | |
cfb2f6f6 | 1610 | ns->mq_mnt->mnt_sb->s_fs_info = NULL; |
7eafd7c7 SH |
1611 | } |
1612 | ||
1613 | void mq_put_mnt(struct ipc_namespace *ns) | |
1614 | { | |
cfb2f6f6 | 1615 | kern_unmount(ns->mq_mnt); |
7eafd7c7 SH |
1616 | } |
1617 | ||
1da177e4 LT |
1618 | static int __init init_mqueue_fs(void) |
1619 | { | |
1620 | int error; | |
1621 | ||
1622 | mqueue_inode_cachep = kmem_cache_create("mqueue_inode_cache", | |
1623 | sizeof(struct mqueue_inode_info), 0, | |
5d097056 | 1624 | SLAB_HWCACHE_ALIGN|SLAB_ACCOUNT, init_once); |
1da177e4 LT |
1625 | if (mqueue_inode_cachep == NULL) |
1626 | return -ENOMEM; | |
1627 | ||
2329e392 | 1628 | /* ignore failures - they are not fatal */ |
bdc8e5f8 | 1629 | mq_sysctl_table = mq_register_sysctl_table(); |
1da177e4 LT |
1630 | |
1631 | error = register_filesystem(&mqueue_fs_type); | |
1632 | if (error) | |
1633 | goto out_sysctl; | |
1634 | ||
7eafd7c7 SH |
1635 | spin_lock_init(&mq_lock); |
1636 | ||
6f686574 AV |
1637 | error = mq_init_ns(&init_ipc_ns); |
1638 | if (error) | |
1da177e4 | 1639 | goto out_filesystem; |
1da177e4 | 1640 | |
1da177e4 LT |
1641 | return 0; |
1642 | ||
1643 | out_filesystem: | |
1644 | unregister_filesystem(&mqueue_fs_type); | |
1645 | out_sysctl: | |
1646 | if (mq_sysctl_table) | |
1647 | unregister_sysctl_table(mq_sysctl_table); | |
1a1d92c1 | 1648 | kmem_cache_destroy(mqueue_inode_cachep); |
1da177e4 LT |
1649 | return error; |
1650 | } | |
1651 | ||
6d08a256 | 1652 | device_initcall(init_mqueue_fs); |