Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/ipc/msg.c | |
5a06a363 | 3 | * Copyright (C) 1992 Krishna Balasubramanian |
1da177e4 LT |
4 | * |
5 | * Removed all the remaining kerneld mess | |
6 | * Catch the -EFAULT stuff properly | |
7 | * Use GFP_KERNEL for messages as in 1.2 | |
8 | * Fixed up the unchecked user space derefs | |
9 | * Copyright (C) 1998 Alan Cox & Andi Kleen | |
10 | * | |
11 | * /proc/sysvipc/msg support (c) 1999 Dragos Acostachioaie <dragos@iname.com> | |
12 | * | |
13 | * mostly rewritten, threaded and wake-one semantics added | |
14 | * MSGMAX limit removed, sysctl's added | |
624dffcb | 15 | * (c) 1999 Manfred Spraul <manfred@colorfullife.com> |
073115d6 SG |
16 | * |
17 | * support for audit of ipc object properties and permission changes | |
18 | * Dustin Kirkland <dustin.kirkland@us.ibm.com> | |
1e786937 KK |
19 | * |
20 | * namespaces support | |
21 | * OpenVZ, SWsoft Inc. | |
22 | * Pavel Emelianov <xemul@openvz.org> | |
1da177e4 LT |
23 | */ |
24 | ||
c59ede7b | 25 | #include <linux/capability.h> |
1da177e4 LT |
26 | #include <linux/msg.h> |
27 | #include <linux/spinlock.h> | |
28 | #include <linux/init.h> | |
f7bf3df8 | 29 | #include <linux/mm.h> |
1da177e4 LT |
30 | #include <linux/proc_fs.h> |
31 | #include <linux/list.h> | |
32 | #include <linux/security.h> | |
33 | #include <linux/sched.h> | |
34 | #include <linux/syscalls.h> | |
35 | #include <linux/audit.h> | |
19b4946c | 36 | #include <linux/seq_file.h> |
3e148c79 | 37 | #include <linux/rwsem.h> |
1e786937 | 38 | #include <linux/nsproxy.h> |
ae5e1b22 | 39 | #include <linux/ipc_namespace.h> |
5f921ae9 | 40 | |
1da177e4 | 41 | #include <asm/current.h> |
7153e402 | 42 | #include <linux/uaccess.h> |
1da177e4 LT |
43 | #include "util.h" |
44 | ||
5a06a363 IM |
45 | /* |
46 | * one msg_receiver structure for each sleeping receiver: | |
47 | */ | |
1da177e4 | 48 | struct msg_receiver { |
5a06a363 IM |
49 | struct list_head r_list; |
50 | struct task_struct *r_tsk; | |
1da177e4 | 51 | |
5a06a363 IM |
52 | int r_mode; |
53 | long r_msgtype; | |
54 | long r_maxsize; | |
1da177e4 | 55 | |
80491eb9 | 56 | struct msg_msg *volatile r_msg; |
1da177e4 LT |
57 | }; |
58 | ||
59 | /* one msg_sender for each sleeping sender */ | |
60 | struct msg_sender { | |
5a06a363 IM |
61 | struct list_head list; |
62 | struct task_struct *tsk; | |
1da177e4 LT |
63 | }; |
64 | ||
65 | #define SEARCH_ANY 1 | |
66 | #define SEARCH_EQUAL 2 | |
67 | #define SEARCH_NOTEQUAL 3 | |
68 | #define SEARCH_LESSEQUAL 4 | |
8ac6ed58 | 69 | #define SEARCH_NUMBER 5 |
1da177e4 | 70 | |
ed2ddbf8 | 71 | #define msg_ids(ns) ((ns)->ids[IPC_MSG_IDS]) |
1da177e4 | 72 | |
01b8b07a | 73 | static void freeque(struct ipc_namespace *, struct kern_ipc_perm *); |
7748dbfa | 74 | static int newque(struct ipc_namespace *, struct ipc_params *); |
1da177e4 | 75 | #ifdef CONFIG_PROC_FS |
19b4946c | 76 | static int sysvipc_msg_proc_show(struct seq_file *s, void *it); |
1da177e4 LT |
77 | #endif |
78 | ||
f7bf3df8 ND |
79 | /* |
80 | * Scale msgmni with the available lowmem size: the memory dedicated to msg | |
81 | * queues should occupy at most 1/MSG_MEM_SCALE of lowmem. | |
4d89dc6a ND |
82 | * Also take into account the number of nsproxies created so far. |
83 | * This should be done staying within the (MSGMNI , IPCMNI/nr_ipc_ns) range. | |
f7bf3df8 | 84 | */ |
b6b337ad | 85 | void recompute_msgmni(struct ipc_namespace *ns) |
f7bf3df8 ND |
86 | { |
87 | struct sysinfo i; | |
88 | unsigned long allowed; | |
4d89dc6a | 89 | int nb_ns; |
f7bf3df8 ND |
90 | |
91 | si_meminfo(&i); | |
92 | allowed = (((i.totalram - i.totalhigh) / MSG_MEM_SCALE) * i.mem_unit) | |
93 | / MSGMNB; | |
4d89dc6a ND |
94 | nb_ns = atomic_read(&nr_ipc_ns); |
95 | allowed /= nb_ns; | |
f7bf3df8 ND |
96 | |
97 | if (allowed < MSGMNI) { | |
98 | ns->msg_ctlmni = MSGMNI; | |
dfcceb26 | 99 | return; |
f7bf3df8 ND |
100 | } |
101 | ||
4d89dc6a ND |
102 | if (allowed > IPCMNI / nb_ns) { |
103 | ns->msg_ctlmni = IPCMNI / nb_ns; | |
dfcceb26 | 104 | return; |
f7bf3df8 ND |
105 | } |
106 | ||
107 | ns->msg_ctlmni = allowed; | |
f7bf3df8 ND |
108 | } |
109 | ||
ed2ddbf8 | 110 | void msg_init_ns(struct ipc_namespace *ns) |
1e786937 | 111 | { |
1e786937 KK |
112 | ns->msg_ctlmax = MSGMAX; |
113 | ns->msg_ctlmnb = MSGMNB; | |
f7bf3df8 ND |
114 | |
115 | recompute_msgmni(ns); | |
116 | ||
3ac88a41 KK |
117 | atomic_set(&ns->msg_bytes, 0); |
118 | atomic_set(&ns->msg_hdrs, 0); | |
ed2ddbf8 | 119 | ipc_init_ids(&ns->ids[IPC_MSG_IDS]); |
1e786937 KK |
120 | } |
121 | ||
ae5e1b22 | 122 | #ifdef CONFIG_IPC_NS |
1e786937 KK |
123 | void msg_exit_ns(struct ipc_namespace *ns) |
124 | { | |
01b8b07a | 125 | free_ipcs(ns, &msg_ids(ns), freeque); |
7d6feeb2 | 126 | idr_destroy(&ns->ids[IPC_MSG_IDS].ipcs_idr); |
1e786937 | 127 | } |
ae5e1b22 | 128 | #endif |
1e786937 | 129 | |
5a06a363 | 130 | void __init msg_init(void) |
1da177e4 | 131 | { |
ed2ddbf8 | 132 | msg_init_ns(&init_ipc_ns); |
dfcceb26 ND |
133 | |
134 | printk(KERN_INFO "msgmni has been set to %d\n", | |
135 | init_ipc_ns.msg_ctlmni); | |
136 | ||
19b4946c MW |
137 | ipc_init_proc_interface("sysvipc/msg", |
138 | " key msqid perms cbytes qnum lspid lrpid uid gid cuid cgid stime rtime ctime\n", | |
1e786937 | 139 | IPC_MSG_IDS, sysvipc_msg_proc_show); |
1da177e4 LT |
140 | } |
141 | ||
a5001a0d DB |
142 | static inline struct msg_queue *msq_obtain_object(struct ipc_namespace *ns, int id) |
143 | { | |
144 | struct kern_ipc_perm *ipcp = ipc_obtain_object(&msg_ids(ns), id); | |
145 | ||
146 | if (IS_ERR(ipcp)) | |
147 | return ERR_CAST(ipcp); | |
148 | ||
149 | return container_of(ipcp, struct msg_queue, q_perm); | |
150 | } | |
151 | ||
152 | static inline struct msg_queue *msq_obtain_object_check(struct ipc_namespace *ns, | |
153 | int id) | |
154 | { | |
155 | struct kern_ipc_perm *ipcp = ipc_obtain_object_check(&msg_ids(ns), id); | |
156 | ||
157 | if (IS_ERR(ipcp)) | |
158 | return ERR_CAST(ipcp); | |
159 | ||
160 | return container_of(ipcp, struct msg_queue, q_perm); | |
161 | } | |
162 | ||
7ca7e564 ND |
163 | static inline void msg_rmid(struct ipc_namespace *ns, struct msg_queue *s) |
164 | { | |
165 | ipc_rmid(&msg_ids(ns), &s->q_perm); | |
166 | } | |
167 | ||
53dad6d3 DB |
168 | static void msg_rcu_free(struct rcu_head *head) |
169 | { | |
170 | struct ipc_rcu *p = container_of(head, struct ipc_rcu, rcu); | |
171 | struct msg_queue *msq = ipc_rcu_to_struct(p); | |
172 | ||
173 | security_msg_queue_free(msq); | |
174 | ipc_rcu_free(head); | |
175 | } | |
176 | ||
f4566f04 ND |
177 | /** |
178 | * newque - Create a new msg queue | |
179 | * @ns: namespace | |
180 | * @params: ptr to the structure that contains the key and msgflg | |
181 | * | |
d9a605e4 | 182 | * Called with msg_ids.rwsem held (writer) |
f4566f04 | 183 | */ |
7748dbfa | 184 | static int newque(struct ipc_namespace *ns, struct ipc_params *params) |
1da177e4 | 185 | { |
1da177e4 | 186 | struct msg_queue *msq; |
5a06a363 | 187 | int id, retval; |
7748dbfa ND |
188 | key_t key = params->key; |
189 | int msgflg = params->flg; | |
1da177e4 | 190 | |
5a06a363 IM |
191 | msq = ipc_rcu_alloc(sizeof(*msq)); |
192 | if (!msq) | |
1da177e4 LT |
193 | return -ENOMEM; |
194 | ||
5a06a363 | 195 | msq->q_perm.mode = msgflg & S_IRWXUGO; |
1da177e4 LT |
196 | msq->q_perm.key = key; |
197 | ||
198 | msq->q_perm.security = NULL; | |
199 | retval = security_msg_queue_alloc(msq); | |
200 | if (retval) { | |
53dad6d3 | 201 | ipc_rcu_putref(msq, ipc_rcu_free); |
1da177e4 LT |
202 | return retval; |
203 | } | |
204 | ||
dbfcd91f | 205 | /* ipc_addid() locks msq upon success. */ |
1e786937 | 206 | id = ipc_addid(&msg_ids(ns), &msq->q_perm, ns->msg_ctlmni); |
283bb7fa | 207 | if (id < 0) { |
53dad6d3 | 208 | ipc_rcu_putref(msq, msg_rcu_free); |
283bb7fa | 209 | return id; |
1da177e4 LT |
210 | } |
211 | ||
212 | msq->q_stime = msq->q_rtime = 0; | |
213 | msq->q_ctime = get_seconds(); | |
214 | msq->q_cbytes = msq->q_qnum = 0; | |
1e786937 | 215 | msq->q_qbytes = ns->msg_ctlmnb; |
1da177e4 LT |
216 | msq->q_lspid = msq->q_lrpid = 0; |
217 | INIT_LIST_HEAD(&msq->q_messages); | |
218 | INIT_LIST_HEAD(&msq->q_receivers); | |
219 | INIT_LIST_HEAD(&msq->q_senders); | |
7ca7e564 | 220 | |
cf9d5d78 | 221 | ipc_unlock_object(&msq->q_perm); |
dbfcd91f | 222 | rcu_read_unlock(); |
1da177e4 | 223 | |
7ca7e564 | 224 | return msq->q_perm.id; |
1da177e4 LT |
225 | } |
226 | ||
5a06a363 | 227 | static inline void ss_add(struct msg_queue *msq, struct msg_sender *mss) |
1da177e4 | 228 | { |
5a06a363 IM |
229 | mss->tsk = current; |
230 | current->state = TASK_INTERRUPTIBLE; | |
231 | list_add_tail(&mss->list, &msq->q_senders); | |
1da177e4 LT |
232 | } |
233 | ||
5a06a363 | 234 | static inline void ss_del(struct msg_sender *mss) |
1da177e4 | 235 | { |
5a06a363 | 236 | if (mss->list.next != NULL) |
1da177e4 LT |
237 | list_del(&mss->list); |
238 | } | |
239 | ||
5a06a363 | 240 | static void ss_wakeup(struct list_head *h, int kill) |
1da177e4 | 241 | { |
41239fe8 | 242 | struct msg_sender *mss, *t; |
1da177e4 | 243 | |
41239fe8 | 244 | list_for_each_entry_safe(mss, t, h, list) { |
5a06a363 IM |
245 | if (kill) |
246 | mss->list.next = NULL; | |
1da177e4 LT |
247 | wake_up_process(mss->tsk); |
248 | } | |
249 | } | |
250 | ||
5a06a363 | 251 | static void expunge_all(struct msg_queue *msq, int res) |
1da177e4 | 252 | { |
41239fe8 | 253 | struct msg_receiver *msr, *t; |
5a06a363 | 254 | |
41239fe8 | 255 | list_for_each_entry_safe(msr, t, &msq->q_receivers, r_list) { |
ffa571da | 256 | msr->r_msg = NULL; /* initialize expunge ordering */ |
1da177e4 | 257 | wake_up_process(msr->r_tsk); |
ffa571da DB |
258 | /* |
259 | * Ensure that the wakeup is visible before setting r_msg as | |
260 | * the receiving end depends on it: either spinning on a nil, | |
261 | * or dealing with -EAGAIN cases. See lockless receive part 1 | |
262 | * and 2 in do_msgrcv(). | |
263 | */ | |
1da177e4 LT |
264 | smp_mb(); |
265 | msr->r_msg = ERR_PTR(res); | |
266 | } | |
267 | } | |
5a06a363 IM |
268 | |
269 | /* | |
270 | * freeque() wakes up waiters on the sender and receiver waiting queue, | |
f4566f04 ND |
271 | * removes the message queue from message queue ID IDR, and cleans up all the |
272 | * messages associated with this queue. | |
1da177e4 | 273 | * |
d9a605e4 DB |
274 | * msg_ids.rwsem (writer) and the spinlock for this message queue are held |
275 | * before freeque() is called. msg_ids.rwsem remains locked on exit. | |
1da177e4 | 276 | */ |
01b8b07a | 277 | static void freeque(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp) |
1da177e4 | 278 | { |
41239fe8 | 279 | struct msg_msg *msg, *t; |
01b8b07a | 280 | struct msg_queue *msq = container_of(ipcp, struct msg_queue, q_perm); |
1da177e4 | 281 | |
5a06a363 IM |
282 | expunge_all(msq, -EIDRM); |
283 | ss_wakeup(&msq->q_senders, 1); | |
7ca7e564 | 284 | msg_rmid(ns, msq); |
4718787d DB |
285 | ipc_unlock_object(&msq->q_perm); |
286 | rcu_read_unlock(); | |
5a06a363 | 287 | |
41239fe8 | 288 | list_for_each_entry_safe(msg, t, &msq->q_messages, m_list) { |
3ac88a41 | 289 | atomic_dec(&ns->msg_hdrs); |
1da177e4 LT |
290 | free_msg(msg); |
291 | } | |
3ac88a41 | 292 | atomic_sub(msq->q_cbytes, &ns->msg_bytes); |
53dad6d3 | 293 | ipc_rcu_putref(msq, msg_rcu_free); |
1da177e4 LT |
294 | } |
295 | ||
f4566f04 | 296 | /* |
d9a605e4 | 297 | * Called with msg_ids.rwsem and ipcp locked. |
f4566f04 | 298 | */ |
03f02c76 | 299 | static inline int msg_security(struct kern_ipc_perm *ipcp, int msgflg) |
7748dbfa | 300 | { |
03f02c76 ND |
301 | struct msg_queue *msq = container_of(ipcp, struct msg_queue, q_perm); |
302 | ||
303 | return security_msg_queue_associate(msq, msgflg); | |
7748dbfa ND |
304 | } |
305 | ||
e48fbb69 | 306 | SYSCALL_DEFINE2(msgget, key_t, key, int, msgflg) |
1da177e4 | 307 | { |
1e786937 | 308 | struct ipc_namespace *ns; |
eb66ec44 MK |
309 | static const struct ipc_ops msg_ops = { |
310 | .getnew = newque, | |
311 | .associate = msg_security, | |
312 | }; | |
7748dbfa | 313 | struct ipc_params msg_params; |
1e786937 KK |
314 | |
315 | ns = current->nsproxy->ipc_ns; | |
7ca7e564 | 316 | |
7748dbfa ND |
317 | msg_params.key = key; |
318 | msg_params.flg = msgflg; | |
5a06a363 | 319 | |
7748dbfa | 320 | return ipcget(ns, &msg_ids(ns), &msg_ops, &msg_params); |
1da177e4 LT |
321 | } |
322 | ||
5a06a363 IM |
323 | static inline unsigned long |
324 | copy_msqid_to_user(void __user *buf, struct msqid64_ds *in, int version) | |
1da177e4 | 325 | { |
239521f3 | 326 | switch (version) { |
1da177e4 | 327 | case IPC_64: |
5a06a363 | 328 | return copy_to_user(buf, in, sizeof(*in)); |
1da177e4 | 329 | case IPC_OLD: |
5a06a363 | 330 | { |
1da177e4 LT |
331 | struct msqid_ds out; |
332 | ||
5a06a363 | 333 | memset(&out, 0, sizeof(out)); |
1da177e4 LT |
334 | |
335 | ipc64_perm_to_ipc_perm(&in->msg_perm, &out.msg_perm); | |
336 | ||
337 | out.msg_stime = in->msg_stime; | |
338 | out.msg_rtime = in->msg_rtime; | |
339 | out.msg_ctime = in->msg_ctime; | |
340 | ||
4be929be AD |
341 | if (in->msg_cbytes > USHRT_MAX) |
342 | out.msg_cbytes = USHRT_MAX; | |
1da177e4 LT |
343 | else |
344 | out.msg_cbytes = in->msg_cbytes; | |
345 | out.msg_lcbytes = in->msg_cbytes; | |
346 | ||
4be929be AD |
347 | if (in->msg_qnum > USHRT_MAX) |
348 | out.msg_qnum = USHRT_MAX; | |
1da177e4 LT |
349 | else |
350 | out.msg_qnum = in->msg_qnum; | |
351 | ||
4be929be AD |
352 | if (in->msg_qbytes > USHRT_MAX) |
353 | out.msg_qbytes = USHRT_MAX; | |
1da177e4 LT |
354 | else |
355 | out.msg_qbytes = in->msg_qbytes; | |
356 | out.msg_lqbytes = in->msg_qbytes; | |
357 | ||
358 | out.msg_lspid = in->msg_lspid; | |
359 | out.msg_lrpid = in->msg_lrpid; | |
360 | ||
5a06a363 IM |
361 | return copy_to_user(buf, &out, sizeof(out)); |
362 | } | |
1da177e4 LT |
363 | default: |
364 | return -EINVAL; | |
365 | } | |
366 | } | |
367 | ||
5a06a363 | 368 | static inline unsigned long |
016d7132 | 369 | copy_msqid_from_user(struct msqid64_ds *out, void __user *buf, int version) |
1da177e4 | 370 | { |
239521f3 | 371 | switch (version) { |
1da177e4 | 372 | case IPC_64: |
016d7132 | 373 | if (copy_from_user(out, buf, sizeof(*out))) |
1da177e4 | 374 | return -EFAULT; |
1da177e4 | 375 | return 0; |
1da177e4 | 376 | case IPC_OLD: |
5a06a363 | 377 | { |
1da177e4 LT |
378 | struct msqid_ds tbuf_old; |
379 | ||
5a06a363 | 380 | if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old))) |
1da177e4 LT |
381 | return -EFAULT; |
382 | ||
239521f3 MS |
383 | out->msg_perm.uid = tbuf_old.msg_perm.uid; |
384 | out->msg_perm.gid = tbuf_old.msg_perm.gid; | |
385 | out->msg_perm.mode = tbuf_old.msg_perm.mode; | |
1da177e4 | 386 | |
5a06a363 | 387 | if (tbuf_old.msg_qbytes == 0) |
016d7132 | 388 | out->msg_qbytes = tbuf_old.msg_lqbytes; |
1da177e4 | 389 | else |
016d7132 | 390 | out->msg_qbytes = tbuf_old.msg_qbytes; |
1da177e4 LT |
391 | |
392 | return 0; | |
5a06a363 | 393 | } |
1da177e4 LT |
394 | default: |
395 | return -EINVAL; | |
396 | } | |
397 | } | |
398 | ||
a0d092fc | 399 | /* |
d9a605e4 | 400 | * This function handles some msgctl commands which require the rwsem |
a0d092fc | 401 | * to be held in write mode. |
d9a605e4 | 402 | * NOTE: no locks must be held, the rwsem is taken inside this function. |
a0d092fc PP |
403 | */ |
404 | static int msgctl_down(struct ipc_namespace *ns, int msqid, int cmd, | |
405 | struct msqid_ds __user *buf, int version) | |
1da177e4 | 406 | { |
1da177e4 | 407 | struct kern_ipc_perm *ipcp; |
f1970c48 | 408 | struct msqid64_ds uninitialized_var(msqid64); |
a0d092fc PP |
409 | struct msg_queue *msq; |
410 | int err; | |
411 | ||
412 | if (cmd == IPC_SET) { | |
016d7132 | 413 | if (copy_msqid_from_user(&msqid64, buf, version)) |
a0d092fc PP |
414 | return -EFAULT; |
415 | } | |
416 | ||
d9a605e4 | 417 | down_write(&msg_ids(ns).rwsem); |
7b4cc5d8 DB |
418 | rcu_read_lock(); |
419 | ||
15724ecb DB |
420 | ipcp = ipcctl_pre_down_nolock(ns, &msg_ids(ns), msqid, cmd, |
421 | &msqid64.msg_perm, msqid64.msg_qbytes); | |
7b4cc5d8 DB |
422 | if (IS_ERR(ipcp)) { |
423 | err = PTR_ERR(ipcp); | |
7b4cc5d8 DB |
424 | goto out_unlock1; |
425 | } | |
a0d092fc | 426 | |
a5f75e7f | 427 | msq = container_of(ipcp, struct msg_queue, q_perm); |
a0d092fc PP |
428 | |
429 | err = security_msg_queue_msgctl(msq, cmd); | |
430 | if (err) | |
15724ecb | 431 | goto out_unlock1; |
a0d092fc PP |
432 | |
433 | switch (cmd) { | |
434 | case IPC_RMID: | |
15724ecb | 435 | ipc_lock_object(&msq->q_perm); |
7b4cc5d8 | 436 | /* freeque unlocks the ipc object and rcu */ |
a0d092fc PP |
437 | freeque(ns, ipcp); |
438 | goto out_up; | |
439 | case IPC_SET: | |
016d7132 | 440 | if (msqid64.msg_qbytes > ns->msg_ctlmnb && |
a0d092fc PP |
441 | !capable(CAP_SYS_RESOURCE)) { |
442 | err = -EPERM; | |
15724ecb | 443 | goto out_unlock1; |
a0d092fc PP |
444 | } |
445 | ||
15724ecb | 446 | ipc_lock_object(&msq->q_perm); |
1efdb69b EB |
447 | err = ipc_update_perm(&msqid64.msg_perm, ipcp); |
448 | if (err) | |
7b4cc5d8 | 449 | goto out_unlock0; |
1efdb69b | 450 | |
016d7132 | 451 | msq->q_qbytes = msqid64.msg_qbytes; |
a0d092fc | 452 | |
a0d092fc PP |
453 | msq->q_ctime = get_seconds(); |
454 | /* sleeping receivers might be excluded by | |
455 | * stricter permissions. | |
456 | */ | |
457 | expunge_all(msq, -EAGAIN); | |
458 | /* sleeping senders might be able to send | |
459 | * due to a larger queue size. | |
460 | */ | |
461 | ss_wakeup(&msq->q_senders, 0); | |
462 | break; | |
463 | default: | |
464 | err = -EINVAL; | |
15724ecb | 465 | goto out_unlock1; |
a0d092fc | 466 | } |
7b4cc5d8 DB |
467 | |
468 | out_unlock0: | |
469 | ipc_unlock_object(&msq->q_perm); | |
470 | out_unlock1: | |
471 | rcu_read_unlock(); | |
a0d092fc | 472 | out_up: |
d9a605e4 | 473 | up_write(&msg_ids(ns).rwsem); |
a0d092fc PP |
474 | return err; |
475 | } | |
476 | ||
2cafed30 DB |
477 | static int msgctl_nolock(struct ipc_namespace *ns, int msqid, |
478 | int cmd, int version, void __user *buf) | |
a0d092fc | 479 | { |
2cafed30 | 480 | int err; |
5a06a363 | 481 | struct msg_queue *msq; |
1da177e4 LT |
482 | |
483 | switch (cmd) { | |
5a06a363 IM |
484 | case IPC_INFO: |
485 | case MSG_INFO: | |
486 | { | |
1da177e4 LT |
487 | struct msginfo msginfo; |
488 | int max_id; | |
5a06a363 | 489 | |
1da177e4 LT |
490 | if (!buf) |
491 | return -EFAULT; | |
2cafed30 | 492 | |
5a06a363 IM |
493 | /* |
494 | * We must not return kernel stack data. | |
1da177e4 LT |
495 | * due to padding, it's not enough |
496 | * to set all member fields. | |
497 | */ | |
1da177e4 LT |
498 | err = security_msg_queue_msgctl(NULL, cmd); |
499 | if (err) | |
500 | return err; | |
501 | ||
5a06a363 | 502 | memset(&msginfo, 0, sizeof(msginfo)); |
1e786937 KK |
503 | msginfo.msgmni = ns->msg_ctlmni; |
504 | msginfo.msgmax = ns->msg_ctlmax; | |
505 | msginfo.msgmnb = ns->msg_ctlmnb; | |
1da177e4 LT |
506 | msginfo.msgssz = MSGSSZ; |
507 | msginfo.msgseg = MSGSEG; | |
d9a605e4 | 508 | down_read(&msg_ids(ns).rwsem); |
1da177e4 | 509 | if (cmd == MSG_INFO) { |
1e786937 | 510 | msginfo.msgpool = msg_ids(ns).in_use; |
3ac88a41 KK |
511 | msginfo.msgmap = atomic_read(&ns->msg_hdrs); |
512 | msginfo.msgtql = atomic_read(&ns->msg_bytes); | |
1da177e4 LT |
513 | } else { |
514 | msginfo.msgmap = MSGMAP; | |
515 | msginfo.msgpool = MSGPOOL; | |
516 | msginfo.msgtql = MSGTQL; | |
517 | } | |
7ca7e564 | 518 | max_id = ipc_get_maxid(&msg_ids(ns)); |
d9a605e4 | 519 | up_read(&msg_ids(ns).rwsem); |
5a06a363 | 520 | if (copy_to_user(buf, &msginfo, sizeof(struct msginfo))) |
1da177e4 | 521 | return -EFAULT; |
5a06a363 | 522 | return (max_id < 0) ? 0 : max_id; |
1da177e4 | 523 | } |
2cafed30 DB |
524 | |
525 | case MSG_STAT: | |
1da177e4 LT |
526 | case IPC_STAT: |
527 | { | |
528 | struct msqid64_ds tbuf; | |
529 | int success_return; | |
5a06a363 | 530 | |
1da177e4 LT |
531 | if (!buf) |
532 | return -EFAULT; | |
1da177e4 | 533 | |
ac0ba20e DB |
534 | memset(&tbuf, 0, sizeof(tbuf)); |
535 | ||
536 | rcu_read_lock(); | |
5a06a363 | 537 | if (cmd == MSG_STAT) { |
ac0ba20e DB |
538 | msq = msq_obtain_object(ns, msqid); |
539 | if (IS_ERR(msq)) { | |
540 | err = PTR_ERR(msq); | |
541 | goto out_unlock; | |
542 | } | |
7ca7e564 | 543 | success_return = msq->q_perm.id; |
1da177e4 | 544 | } else { |
ac0ba20e DB |
545 | msq = msq_obtain_object_check(ns, msqid); |
546 | if (IS_ERR(msq)) { | |
547 | err = PTR_ERR(msq); | |
548 | goto out_unlock; | |
549 | } | |
1da177e4 LT |
550 | success_return = 0; |
551 | } | |
ac0ba20e | 552 | |
1da177e4 | 553 | err = -EACCES; |
b0e77598 | 554 | if (ipcperms(ns, &msq->q_perm, S_IRUGO)) |
1da177e4 LT |
555 | goto out_unlock; |
556 | ||
557 | err = security_msg_queue_msgctl(msq, cmd); | |
558 | if (err) | |
559 | goto out_unlock; | |
560 | ||
561 | kernel_to_ipc64_perm(&msq->q_perm, &tbuf.msg_perm); | |
562 | tbuf.msg_stime = msq->q_stime; | |
563 | tbuf.msg_rtime = msq->q_rtime; | |
564 | tbuf.msg_ctime = msq->q_ctime; | |
565 | tbuf.msg_cbytes = msq->q_cbytes; | |
566 | tbuf.msg_qnum = msq->q_qnum; | |
567 | tbuf.msg_qbytes = msq->q_qbytes; | |
568 | tbuf.msg_lspid = msq->q_lspid; | |
569 | tbuf.msg_lrpid = msq->q_lrpid; | |
ac0ba20e DB |
570 | rcu_read_unlock(); |
571 | ||
1da177e4 LT |
572 | if (copy_msqid_to_user(buf, &tbuf, version)) |
573 | return -EFAULT; | |
574 | return success_return; | |
575 | } | |
2cafed30 | 576 | |
1da177e4 | 577 | default: |
2cafed30 | 578 | return -EINVAL; |
1da177e4 LT |
579 | } |
580 | ||
2cafed30 | 581 | return err; |
1da177e4 | 582 | out_unlock: |
ac0ba20e | 583 | rcu_read_unlock(); |
1da177e4 LT |
584 | return err; |
585 | } | |
586 | ||
2cafed30 DB |
587 | SYSCALL_DEFINE3(msgctl, int, msqid, int, cmd, struct msqid_ds __user *, buf) |
588 | { | |
589 | int version; | |
590 | struct ipc_namespace *ns; | |
591 | ||
592 | if (msqid < 0 || cmd < 0) | |
593 | return -EINVAL; | |
594 | ||
595 | version = ipc_parse_version(&cmd); | |
596 | ns = current->nsproxy->ipc_ns; | |
597 | ||
598 | switch (cmd) { | |
599 | case IPC_INFO: | |
600 | case MSG_INFO: | |
601 | case MSG_STAT: /* msqid is an index rather than a msg queue id */ | |
602 | case IPC_STAT: | |
603 | return msgctl_nolock(ns, msqid, cmd, version, buf); | |
604 | case IPC_SET: | |
605 | case IPC_RMID: | |
606 | return msgctl_down(ns, msqid, cmd, buf, version); | |
607 | default: | |
608 | return -EINVAL; | |
609 | } | |
610 | } | |
611 | ||
5a06a363 | 612 | static int testmsg(struct msg_msg *msg, long type, int mode) |
1da177e4 | 613 | { |
239521f3 | 614 | switch (mode) |
1da177e4 LT |
615 | { |
616 | case SEARCH_ANY: | |
8ac6ed58 | 617 | case SEARCH_NUMBER: |
1da177e4 LT |
618 | return 1; |
619 | case SEARCH_LESSEQUAL: | |
239521f3 | 620 | if (msg->m_type <= type) |
1da177e4 LT |
621 | return 1; |
622 | break; | |
623 | case SEARCH_EQUAL: | |
5a06a363 | 624 | if (msg->m_type == type) |
1da177e4 LT |
625 | return 1; |
626 | break; | |
627 | case SEARCH_NOTEQUAL: | |
5a06a363 | 628 | if (msg->m_type != type) |
1da177e4 LT |
629 | return 1; |
630 | break; | |
631 | } | |
632 | return 0; | |
633 | } | |
634 | ||
5a06a363 | 635 | static inline int pipelined_send(struct msg_queue *msq, struct msg_msg *msg) |
1da177e4 | 636 | { |
41239fe8 | 637 | struct msg_receiver *msr, *t; |
5a06a363 | 638 | |
41239fe8 | 639 | list_for_each_entry_safe(msr, t, &msq->q_receivers, r_list) { |
5a06a363 IM |
640 | if (testmsg(msg, msr->r_msgtype, msr->r_mode) && |
641 | !security_msg_queue_msgrcv(msq, msg, msr->r_tsk, | |
642 | msr->r_msgtype, msr->r_mode)) { | |
643 | ||
1da177e4 | 644 | list_del(&msr->r_list); |
5a06a363 | 645 | if (msr->r_maxsize < msg->m_ts) { |
ffa571da | 646 | /* initialize pipelined send ordering */ |
1da177e4 LT |
647 | msr->r_msg = NULL; |
648 | wake_up_process(msr->r_tsk); | |
ffa571da | 649 | smp_mb(); /* see barrier comment below */ |
1da177e4 LT |
650 | msr->r_msg = ERR_PTR(-E2BIG); |
651 | } else { | |
652 | msr->r_msg = NULL; | |
b488893a | 653 | msq->q_lrpid = task_pid_vnr(msr->r_tsk); |
1da177e4 LT |
654 | msq->q_rtime = get_seconds(); |
655 | wake_up_process(msr->r_tsk); | |
ffa571da DB |
656 | /* |
657 | * Ensure that the wakeup is visible before | |
658 | * setting r_msg, as the receiving end depends | |
659 | * on it. See lockless receive part 1 and 2 in | |
660 | * do_msgrcv(). | |
661 | */ | |
1da177e4 LT |
662 | smp_mb(); |
663 | msr->r_msg = msg; | |
5a06a363 | 664 | |
1da177e4 LT |
665 | return 1; |
666 | } | |
667 | } | |
668 | } | |
ffa571da | 669 | |
1da177e4 LT |
670 | return 0; |
671 | } | |
672 | ||
651971cb | 673 | long do_msgsnd(int msqid, long mtype, void __user *mtext, |
674 | size_t msgsz, int msgflg) | |
1da177e4 LT |
675 | { |
676 | struct msg_queue *msq; | |
677 | struct msg_msg *msg; | |
1da177e4 | 678 | int err; |
1e786937 KK |
679 | struct ipc_namespace *ns; |
680 | ||
681 | ns = current->nsproxy->ipc_ns; | |
5a06a363 | 682 | |
1e786937 | 683 | if (msgsz > ns->msg_ctlmax || (long) msgsz < 0 || msqid < 0) |
1da177e4 | 684 | return -EINVAL; |
1da177e4 LT |
685 | if (mtype < 1) |
686 | return -EINVAL; | |
687 | ||
651971cb | 688 | msg = load_msg(mtext, msgsz); |
5a06a363 | 689 | if (IS_ERR(msg)) |
1da177e4 LT |
690 | return PTR_ERR(msg); |
691 | ||
692 | msg->m_type = mtype; | |
693 | msg->m_ts = msgsz; | |
694 | ||
3dd1f784 DB |
695 | rcu_read_lock(); |
696 | msq = msq_obtain_object_check(ns, msqid); | |
023a5355 ND |
697 | if (IS_ERR(msq)) { |
698 | err = PTR_ERR(msq); | |
3dd1f784 | 699 | goto out_unlock1; |
023a5355 | 700 | } |
1da177e4 | 701 | |
bebcb928 MS |
702 | ipc_lock_object(&msq->q_perm); |
703 | ||
1da177e4 LT |
704 | for (;;) { |
705 | struct msg_sender s; | |
706 | ||
5a06a363 | 707 | err = -EACCES; |
b0e77598 | 708 | if (ipcperms(ns, &msq->q_perm, S_IWUGO)) |
bebcb928 | 709 | goto out_unlock0; |
1da177e4 | 710 | |
4271b05a | 711 | /* raced with RMID? */ |
0f3d2b01 | 712 | if (!ipc_valid_object(&msq->q_perm)) { |
4271b05a DB |
713 | err = -EIDRM; |
714 | goto out_unlock0; | |
715 | } | |
716 | ||
1da177e4 LT |
717 | err = security_msg_queue_msgsnd(msq, msg, msgflg); |
718 | if (err) | |
bebcb928 | 719 | goto out_unlock0; |
1da177e4 | 720 | |
5a06a363 | 721 | if (msgsz + msq->q_cbytes <= msq->q_qbytes && |
1da177e4 LT |
722 | 1 + msq->q_qnum <= msq->q_qbytes) { |
723 | break; | |
724 | } | |
725 | ||
726 | /* queue full, wait: */ | |
5a06a363 IM |
727 | if (msgflg & IPC_NOWAIT) { |
728 | err = -EAGAIN; | |
bebcb928 | 729 | goto out_unlock0; |
1da177e4 | 730 | } |
3dd1f784 | 731 | |
ffa571da | 732 | /* enqueue the sender and prepare to block */ |
1da177e4 | 733 | ss_add(msq, &s); |
6062a8dc RR |
734 | |
735 | if (!ipc_rcu_getref(msq)) { | |
736 | err = -EIDRM; | |
3dd1f784 | 737 | goto out_unlock0; |
6062a8dc RR |
738 | } |
739 | ||
3dd1f784 DB |
740 | ipc_unlock_object(&msq->q_perm); |
741 | rcu_read_unlock(); | |
1da177e4 LT |
742 | schedule(); |
743 | ||
3dd1f784 DB |
744 | rcu_read_lock(); |
745 | ipc_lock_object(&msq->q_perm); | |
746 | ||
53dad6d3 | 747 | ipc_rcu_putref(msq, ipc_rcu_free); |
0f3d2b01 RA |
748 | /* raced with RMID? */ |
749 | if (!ipc_valid_object(&msq->q_perm)) { | |
1da177e4 | 750 | err = -EIDRM; |
3dd1f784 | 751 | goto out_unlock0; |
1da177e4 | 752 | } |
3dd1f784 | 753 | |
1da177e4 | 754 | ss_del(&s); |
5a06a363 | 755 | |
1da177e4 | 756 | if (signal_pending(current)) { |
5a06a363 | 757 | err = -ERESTARTNOHAND; |
3dd1f784 | 758 | goto out_unlock0; |
1da177e4 | 759 | } |
3dd1f784 | 760 | |
1da177e4 | 761 | } |
b488893a | 762 | msq->q_lspid = task_tgid_vnr(current); |
1da177e4 LT |
763 | msq->q_stime = get_seconds(); |
764 | ||
5a06a363 | 765 | if (!pipelined_send(msq, msg)) { |
25985edc | 766 | /* no one is waiting for this message, enqueue it */ |
5a06a363 | 767 | list_add_tail(&msg->m_list, &msq->q_messages); |
1da177e4 LT |
768 | msq->q_cbytes += msgsz; |
769 | msq->q_qnum++; | |
3ac88a41 KK |
770 | atomic_add(msgsz, &ns->msg_bytes); |
771 | atomic_inc(&ns->msg_hdrs); | |
1da177e4 | 772 | } |
5a06a363 | 773 | |
1da177e4 LT |
774 | err = 0; |
775 | msg = NULL; | |
776 | ||
3dd1f784 DB |
777 | out_unlock0: |
778 | ipc_unlock_object(&msq->q_perm); | |
779 | out_unlock1: | |
780 | rcu_read_unlock(); | |
5a06a363 | 781 | if (msg != NULL) |
1da177e4 LT |
782 | free_msg(msg); |
783 | return err; | |
784 | } | |
785 | ||
e48fbb69 HC |
786 | SYSCALL_DEFINE4(msgsnd, int, msqid, struct msgbuf __user *, msgp, size_t, msgsz, |
787 | int, msgflg) | |
651971cb | 788 | { |
789 | long mtype; | |
790 | ||
791 | if (get_user(mtype, &msgp->mtype)) | |
792 | return -EFAULT; | |
793 | return do_msgsnd(msqid, mtype, msgp->mtext, msgsz, msgflg); | |
794 | } | |
795 | ||
5a06a363 | 796 | static inline int convert_mode(long *msgtyp, int msgflg) |
1da177e4 | 797 | { |
8ac6ed58 PH |
798 | if (msgflg & MSG_COPY) |
799 | return SEARCH_NUMBER; | |
5a06a363 | 800 | /* |
1da177e4 LT |
801 | * find message of correct type. |
802 | * msgtyp = 0 => get first. | |
803 | * msgtyp > 0 => get first message of matching type. | |
5a06a363 | 804 | * msgtyp < 0 => get message with least type must be < abs(msgtype). |
1da177e4 | 805 | */ |
5a06a363 | 806 | if (*msgtyp == 0) |
1da177e4 | 807 | return SEARCH_ANY; |
5a06a363 IM |
808 | if (*msgtyp < 0) { |
809 | *msgtyp = -*msgtyp; | |
1da177e4 LT |
810 | return SEARCH_LESSEQUAL; |
811 | } | |
5a06a363 | 812 | if (msgflg & MSG_EXCEPT) |
1da177e4 LT |
813 | return SEARCH_NOTEQUAL; |
814 | return SEARCH_EQUAL; | |
815 | } | |
816 | ||
f9dd87f4 SK |
817 | static long do_msg_fill(void __user *dest, struct msg_msg *msg, size_t bufsz) |
818 | { | |
819 | struct msgbuf __user *msgp = dest; | |
820 | size_t msgsz; | |
821 | ||
822 | if (put_user(msg->m_type, &msgp->mtype)) | |
823 | return -EFAULT; | |
824 | ||
825 | msgsz = (bufsz > msg->m_ts) ? msg->m_ts : bufsz; | |
826 | if (store_msg(msgp->mtext, msg, msgsz)) | |
827 | return -EFAULT; | |
828 | return msgsz; | |
829 | } | |
830 | ||
4a674f34 | 831 | #ifdef CONFIG_CHECKPOINT_RESTORE |
3fcfe786 SK |
832 | /* |
833 | * This function creates new kernel message structure, large enough to store | |
834 | * bufsz message bytes. | |
835 | */ | |
8ac6ed58 | 836 | static inline struct msg_msg *prepare_copy(void __user *buf, size_t bufsz) |
4a674f34 SK |
837 | { |
838 | struct msg_msg *copy; | |
839 | ||
4a674f34 SK |
840 | /* |
841 | * Create dummy message to copy real message to. | |
842 | */ | |
843 | copy = load_msg(buf, bufsz); | |
844 | if (!IS_ERR(copy)) | |
845 | copy->m_ts = bufsz; | |
846 | return copy; | |
847 | } | |
848 | ||
85398aa8 | 849 | static inline void free_copy(struct msg_msg *copy) |
4a674f34 | 850 | { |
85398aa8 | 851 | if (copy) |
4a674f34 SK |
852 | free_msg(copy); |
853 | } | |
854 | #else | |
8ac6ed58 | 855 | static inline struct msg_msg *prepare_copy(void __user *buf, size_t bufsz) |
b30efe27 SK |
856 | { |
857 | return ERR_PTR(-ENOSYS); | |
858 | } | |
859 | ||
85398aa8 SK |
860 | static inline void free_copy(struct msg_msg *copy) |
861 | { | |
862 | } | |
4a674f34 SK |
863 | #endif |
864 | ||
daaf74cf PH |
865 | static struct msg_msg *find_msg(struct msg_queue *msq, long *msgtyp, int mode) |
866 | { | |
368ae537 | 867 | struct msg_msg *msg, *found = NULL; |
daaf74cf PH |
868 | long count = 0; |
869 | ||
870 | list_for_each_entry(msg, &msq->q_messages, m_list) { | |
871 | if (testmsg(msg, *msgtyp, mode) && | |
872 | !security_msg_queue_msgrcv(msq, msg, current, | |
873 | *msgtyp, mode)) { | |
874 | if (mode == SEARCH_LESSEQUAL && msg->m_type != 1) { | |
875 | *msgtyp = msg->m_type - 1; | |
368ae537 | 876 | found = msg; |
daaf74cf PH |
877 | } else if (mode == SEARCH_NUMBER) { |
878 | if (*msgtyp == count) | |
879 | return msg; | |
880 | } else | |
881 | return msg; | |
882 | count++; | |
883 | } | |
884 | } | |
885 | ||
368ae537 | 886 | return found ?: ERR_PTR(-EAGAIN); |
daaf74cf PH |
887 | } |
888 | ||
41a0d523 | 889 | long do_msgrcv(int msqid, void __user *buf, size_t bufsz, long msgtyp, int msgflg, |
f9dd87f4 | 890 | long (*msg_handler)(void __user *, struct msg_msg *, size_t)) |
1da177e4 | 891 | { |
1da177e4 | 892 | int mode; |
41a0d523 | 893 | struct msg_queue *msq; |
1e786937 | 894 | struct ipc_namespace *ns; |
41a0d523 | 895 | struct msg_msg *msg, *copy = NULL; |
1da177e4 | 896 | |
88b9e456 PH |
897 | ns = current->nsproxy->ipc_ns; |
898 | ||
f9dd87f4 | 899 | if (msqid < 0 || (long) bufsz < 0) |
1da177e4 | 900 | return -EINVAL; |
41a0d523 | 901 | |
4a674f34 | 902 | if (msgflg & MSG_COPY) { |
4f87dac3 MK |
903 | if ((msgflg & MSG_EXCEPT) || !(msgflg & IPC_NOWAIT)) |
904 | return -EINVAL; | |
8ac6ed58 | 905 | copy = prepare_copy(buf, min_t(size_t, bufsz, ns->msg_ctlmax)); |
4a674f34 SK |
906 | if (IS_ERR(copy)) |
907 | return PTR_ERR(copy); | |
908 | } | |
5a06a363 | 909 | mode = convert_mode(&msgtyp, msgflg); |
1da177e4 | 910 | |
41a0d523 DB |
911 | rcu_read_lock(); |
912 | msq = msq_obtain_object_check(ns, msqid); | |
4a674f34 | 913 | if (IS_ERR(msq)) { |
41a0d523 | 914 | rcu_read_unlock(); |
85398aa8 | 915 | free_copy(copy); |
023a5355 | 916 | return PTR_ERR(msq); |
4a674f34 | 917 | } |
1da177e4 LT |
918 | |
919 | for (;;) { | |
920 | struct msg_receiver msr_d; | |
1da177e4 LT |
921 | |
922 | msg = ERR_PTR(-EACCES); | |
b0e77598 | 923 | if (ipcperms(ns, &msq->q_perm, S_IRUGO)) |
41a0d523 | 924 | goto out_unlock1; |
1da177e4 | 925 | |
41a0d523 | 926 | ipc_lock_object(&msq->q_perm); |
4271b05a DB |
927 | |
928 | /* raced with RMID? */ | |
0f3d2b01 | 929 | if (!ipc_valid_object(&msq->q_perm)) { |
4271b05a DB |
930 | msg = ERR_PTR(-EIDRM); |
931 | goto out_unlock0; | |
932 | } | |
933 | ||
daaf74cf | 934 | msg = find_msg(msq, &msgtyp, mode); |
5a06a363 IM |
935 | if (!IS_ERR(msg)) { |
936 | /* | |
937 | * Found a suitable message. | |
938 | * Unlink it from the queue. | |
939 | */ | |
f9dd87f4 | 940 | if ((bufsz < msg->m_ts) && !(msgflg & MSG_NOERROR)) { |
1da177e4 | 941 | msg = ERR_PTR(-E2BIG); |
41a0d523 | 942 | goto out_unlock0; |
1da177e4 | 943 | } |
3fcfe786 SK |
944 | /* |
945 | * If we are copying, then do not unlink message and do | |
946 | * not update queue parameters. | |
947 | */ | |
852028af PH |
948 | if (msgflg & MSG_COPY) { |
949 | msg = copy_msg(msg, copy); | |
41a0d523 | 950 | goto out_unlock0; |
852028af | 951 | } |
41a0d523 | 952 | |
1da177e4 LT |
953 | list_del(&msg->m_list); |
954 | msq->q_qnum--; | |
955 | msq->q_rtime = get_seconds(); | |
b488893a | 956 | msq->q_lrpid = task_tgid_vnr(current); |
1da177e4 | 957 | msq->q_cbytes -= msg->m_ts; |
3ac88a41 KK |
958 | atomic_sub(msg->m_ts, &ns->msg_bytes); |
959 | atomic_dec(&ns->msg_hdrs); | |
5a06a363 | 960 | ss_wakeup(&msq->q_senders, 0); |
41a0d523 DB |
961 | |
962 | goto out_unlock0; | |
1da177e4 | 963 | } |
41a0d523 | 964 | |
1da177e4 LT |
965 | /* No message waiting. Wait for a message */ |
966 | if (msgflg & IPC_NOWAIT) { | |
967 | msg = ERR_PTR(-ENOMSG); | |
41a0d523 | 968 | goto out_unlock0; |
1da177e4 | 969 | } |
41a0d523 | 970 | |
5a06a363 | 971 | list_add_tail(&msr_d.r_list, &msq->q_receivers); |
1da177e4 LT |
972 | msr_d.r_tsk = current; |
973 | msr_d.r_msgtype = msgtyp; | |
974 | msr_d.r_mode = mode; | |
5a06a363 | 975 | if (msgflg & MSG_NOERROR) |
1da177e4 | 976 | msr_d.r_maxsize = INT_MAX; |
5a06a363 | 977 | else |
f9dd87f4 | 978 | msr_d.r_maxsize = bufsz; |
1da177e4 LT |
979 | msr_d.r_msg = ERR_PTR(-EAGAIN); |
980 | current->state = TASK_INTERRUPTIBLE; | |
1da177e4 | 981 | |
41a0d523 DB |
982 | ipc_unlock_object(&msq->q_perm); |
983 | rcu_read_unlock(); | |
1da177e4 LT |
984 | schedule(); |
985 | ||
986 | /* Lockless receive, part 1: | |
987 | * Disable preemption. We don't hold a reference to the queue | |
988 | * and getting a reference would defeat the idea of a lockless | |
989 | * operation, thus the code relies on rcu to guarantee the | |
25985edc | 990 | * existence of msq: |
1da177e4 LT |
991 | * Prior to destruction, expunge_all(-EIRDM) changes r_msg. |
992 | * Thus if r_msg is -EAGAIN, then the queue not yet destroyed. | |
993 | * rcu_read_lock() prevents preemption between reading r_msg | |
41a0d523 | 994 | * and acquiring the q_perm.lock in ipc_lock_object(). |
1da177e4 LT |
995 | */ |
996 | rcu_read_lock(); | |
997 | ||
998 | /* Lockless receive, part 2: | |
999 | * Wait until pipelined_send or expunge_all are outside of | |
1000 | * wake_up_process(). There is a race with exit(), see | |
1001 | * ipc/mqueue.c for the details. | |
1002 | */ | |
239521f3 | 1003 | msg = (struct msg_msg *)msr_d.r_msg; |
1da177e4 LT |
1004 | while (msg == NULL) { |
1005 | cpu_relax(); | |
5a06a363 | 1006 | msg = (struct msg_msg *)msr_d.r_msg; |
1da177e4 LT |
1007 | } |
1008 | ||
1009 | /* Lockless receive, part 3: | |
1010 | * If there is a message or an error then accept it without | |
1011 | * locking. | |
1012 | */ | |
41a0d523 DB |
1013 | if (msg != ERR_PTR(-EAGAIN)) |
1014 | goto out_unlock1; | |
1da177e4 LT |
1015 | |
1016 | /* Lockless receive, part 3: | |
1017 | * Acquire the queue spinlock. | |
1018 | */ | |
41a0d523 | 1019 | ipc_lock_object(&msq->q_perm); |
1da177e4 LT |
1020 | |
1021 | /* Lockless receive, part 4: | |
1022 | * Repeat test after acquiring the spinlock. | |
1023 | */ | |
239521f3 | 1024 | msg = (struct msg_msg *)msr_d.r_msg; |
5a06a363 | 1025 | if (msg != ERR_PTR(-EAGAIN)) |
41a0d523 | 1026 | goto out_unlock0; |
1da177e4 LT |
1027 | |
1028 | list_del(&msr_d.r_list); | |
1029 | if (signal_pending(current)) { | |
1030 | msg = ERR_PTR(-ERESTARTNOHAND); | |
41a0d523 | 1031 | goto out_unlock0; |
1da177e4 | 1032 | } |
41a0d523 DB |
1033 | |
1034 | ipc_unlock_object(&msq->q_perm); | |
1da177e4 | 1035 | } |
41a0d523 DB |
1036 | |
1037 | out_unlock0: | |
1038 | ipc_unlock_object(&msq->q_perm); | |
1039 | out_unlock1: | |
1040 | rcu_read_unlock(); | |
4a674f34 | 1041 | if (IS_ERR(msg)) { |
85398aa8 | 1042 | free_copy(copy); |
5a06a363 | 1043 | return PTR_ERR(msg); |
4a674f34 | 1044 | } |
1da177e4 | 1045 | |
f9dd87f4 | 1046 | bufsz = msg_handler(buf, msg, bufsz); |
1da177e4 | 1047 | free_msg(msg); |
5a06a363 | 1048 | |
f9dd87f4 | 1049 | return bufsz; |
1da177e4 LT |
1050 | } |
1051 | ||
e48fbb69 HC |
1052 | SYSCALL_DEFINE5(msgrcv, int, msqid, struct msgbuf __user *, msgp, size_t, msgsz, |
1053 | long, msgtyp, int, msgflg) | |
651971cb | 1054 | { |
f9dd87f4 | 1055 | return do_msgrcv(msqid, msgp, msgsz, msgtyp, msgflg, do_msg_fill); |
651971cb | 1056 | } |
1057 | ||
1da177e4 | 1058 | #ifdef CONFIG_PROC_FS |
19b4946c | 1059 | static int sysvipc_msg_proc_show(struct seq_file *s, void *it) |
1da177e4 | 1060 | { |
1efdb69b | 1061 | struct user_namespace *user_ns = seq_user_ns(s); |
19b4946c MW |
1062 | struct msg_queue *msq = it; |
1063 | ||
1064 | return seq_printf(s, | |
5a06a363 IM |
1065 | "%10d %10d %4o %10lu %10lu %5u %5u %5u %5u %5u %5u %10lu %10lu %10lu\n", |
1066 | msq->q_perm.key, | |
7ca7e564 | 1067 | msq->q_perm.id, |
5a06a363 IM |
1068 | msq->q_perm.mode, |
1069 | msq->q_cbytes, | |
1070 | msq->q_qnum, | |
1071 | msq->q_lspid, | |
1072 | msq->q_lrpid, | |
1efdb69b EB |
1073 | from_kuid_munged(user_ns, msq->q_perm.uid), |
1074 | from_kgid_munged(user_ns, msq->q_perm.gid), | |
1075 | from_kuid_munged(user_ns, msq->q_perm.cuid), | |
1076 | from_kgid_munged(user_ns, msq->q_perm.cgid), | |
5a06a363 IM |
1077 | msq->q_stime, |
1078 | msq->q_rtime, | |
1079 | msq->q_ctime); | |
1da177e4 LT |
1080 | } |
1081 | #endif |