blob: 32494e8cc7a5cd5fa72b874da80c149f007e84cb [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/ipc/msg.c
Ingo Molnar5a06a362006-07-30 03:04:11 -07003 * Copyright (C) 1992 Krishna Balasubramanian
Linus Torvalds1da177e2005-04-16 15:20:36 -07004 *
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
Christian Kujau624dffc2006-01-15 02:43:54 +010015 * (c) 1999 Manfred Spraul <manfred@colorfullife.com>
Steve Grubb073115d2006-04-02 17:07:33 -040016 *
17 * support for audit of ipc object properties and permission changes
18 * Dustin Kirkland <dustin.kirkland@us.ibm.com>
Kirill Korotaev1e786932006-10-02 02:18:21 -070019 *
20 * namespaces support
21 * OpenVZ, SWsoft Inc.
22 * Pavel Emelianov <xemul@openvz.org>
Linus Torvalds1da177e2005-04-16 15:20:36 -070023 */
24
Randy.Dunlapc59ede72006-01-11 12:17:46 -080025#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include <linux/slab.h>
27#include <linux/msg.h>
28#include <linux/spinlock.h>
29#include <linux/init.h>
Nadia Derbeyf7bf3df2008-04-29 01:00:39 -070030#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <linux/proc_fs.h>
32#include <linux/list.h>
33#include <linux/security.h>
34#include <linux/sched.h>
35#include <linux/syscalls.h>
36#include <linux/audit.h>
Mike Waychison19b49462005-09-06 15:17:10 -070037#include <linux/seq_file.h>
Nadia Derbey3e148c72007-10-18 23:40:54 -070038#include <linux/rwsem.h>
Kirill Korotaev1e786932006-10-02 02:18:21 -070039#include <linux/nsproxy.h>
Pavel Emelyanovae5e1b22008-02-08 04:18:22 -080040#include <linux/ipc_namespace.h>
Ingo Molnar5f921ae2006-03-26 01:37:17 -080041
Linus Torvalds1da177e2005-04-16 15:20:36 -070042#include <asm/current.h>
43#include <asm/uaccess.h>
44#include "util.h"
45
Ingo Molnar5a06a362006-07-30 03:04:11 -070046/*
47 * one msg_receiver structure for each sleeping receiver:
48 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070049struct msg_receiver {
Ingo Molnar5a06a362006-07-30 03:04:11 -070050 struct list_head r_list;
51 struct task_struct *r_tsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -070052
Ingo Molnar5a06a362006-07-30 03:04:11 -070053 int r_mode;
54 long r_msgtype;
55 long r_maxsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -070056
Linus Torvalds80491eb2006-11-04 09:55:00 -080057 struct msg_msg *volatile r_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -070058};
59
60/* one msg_sender for each sleeping sender */
61struct msg_sender {
Ingo Molnar5a06a362006-07-30 03:04:11 -070062 struct list_head list;
63 struct task_struct *tsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -070064};
65
66#define SEARCH_ANY 1
67#define SEARCH_EQUAL 2
68#define SEARCH_NOTEQUAL 3
69#define SEARCH_LESSEQUAL 4
70
Pierre Peiffered2ddbf2008-02-08 04:18:57 -080071#define msg_ids(ns) ((ns)->ids[IPC_MSG_IDS])
Kirill Korotaev1e786932006-10-02 02:18:21 -070072
Ingo Molnar5a06a362006-07-30 03:04:11 -070073#define msg_unlock(msq) ipc_unlock(&(msq)->q_perm)
Linus Torvalds1da177e2005-04-16 15:20:36 -070074
Pierre Peiffer01b8b072008-02-08 04:18:57 -080075static void freeque(struct ipc_namespace *, struct kern_ipc_perm *);
Nadia Derbey7748dbf2007-10-18 23:40:49 -070076static int newque(struct ipc_namespace *, struct ipc_params *);
Linus Torvalds1da177e2005-04-16 15:20:36 -070077#ifdef CONFIG_PROC_FS
Mike Waychison19b49462005-09-06 15:17:10 -070078static int sysvipc_msg_proc_show(struct seq_file *s, void *it);
Linus Torvalds1da177e2005-04-16 15:20:36 -070079#endif
80
Nadia Derbeyf7bf3df2008-04-29 01:00:39 -070081/*
82 * Scale msgmni with the available lowmem size: the memory dedicated to msg
83 * queues should occupy at most 1/MSG_MEM_SCALE of lowmem.
Nadia Derbey4d89dc62008-04-29 01:00:40 -070084 * Also take into account the number of nsproxies created so far.
85 * This should be done staying within the (MSGMNI , IPCMNI/nr_ipc_ns) range.
Nadia Derbeyf7bf3df2008-04-29 01:00:39 -070086 */
Nadia Derbeyb6b337a2008-04-29 01:00:42 -070087void recompute_msgmni(struct ipc_namespace *ns)
Nadia Derbeyf7bf3df2008-04-29 01:00:39 -070088{
89 struct sysinfo i;
90 unsigned long allowed;
Nadia Derbey4d89dc62008-04-29 01:00:40 -070091 int nb_ns;
Nadia Derbeyf7bf3df2008-04-29 01:00:39 -070092
93 si_meminfo(&i);
94 allowed = (((i.totalram - i.totalhigh) / MSG_MEM_SCALE) * i.mem_unit)
95 / MSGMNB;
Nadia Derbey4d89dc62008-04-29 01:00:40 -070096 nb_ns = atomic_read(&nr_ipc_ns);
97 allowed /= nb_ns;
Nadia Derbeyf7bf3df2008-04-29 01:00:39 -070098
99 if (allowed < MSGMNI) {
100 ns->msg_ctlmni = MSGMNI;
101 goto out_callback;
102 }
103
Nadia Derbey4d89dc62008-04-29 01:00:40 -0700104 if (allowed > IPCMNI / nb_ns) {
105 ns->msg_ctlmni = IPCMNI / nb_ns;
Nadia Derbeyf7bf3df2008-04-29 01:00:39 -0700106 goto out_callback;
107 }
108
109 ns->msg_ctlmni = allowed;
110
111out_callback:
112
113 printk(KERN_INFO "msgmni has been set to %d for ipc namespace %p\n",
114 ns->msg_ctlmni, ns);
115}
116
Pierre Peiffered2ddbf2008-02-08 04:18:57 -0800117void msg_init_ns(struct ipc_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118{
Kirill Korotaev1e786932006-10-02 02:18:21 -0700119 ns->msg_ctlmax = MSGMAX;
120 ns->msg_ctlmnb = MSGMNB;
Nadia Derbeyf7bf3df2008-04-29 01:00:39 -0700121
122 recompute_msgmni(ns);
123
Kirill Korotaev3ac88a42007-10-18 23:40:56 -0700124 atomic_set(&ns->msg_bytes, 0);
125 atomic_set(&ns->msg_hdrs, 0);
Pierre Peiffered2ddbf2008-02-08 04:18:57 -0800126 ipc_init_ids(&ns->ids[IPC_MSG_IDS]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127}
128
Pavel Emelyanovae5e1b22008-02-08 04:18:22 -0800129#ifdef CONFIG_IPC_NS
Kirill Korotaev1e786932006-10-02 02:18:21 -0700130void msg_exit_ns(struct ipc_namespace *ns)
131{
Pierre Peiffer01b8b072008-02-08 04:18:57 -0800132 free_ipcs(ns, &msg_ids(ns), freeque);
Kirill Korotaev1e786932006-10-02 02:18:21 -0700133}
Pavel Emelyanovae5e1b22008-02-08 04:18:22 -0800134#endif
Kirill Korotaev1e786932006-10-02 02:18:21 -0700135
136void __init msg_init(void)
137{
Pierre Peiffered2ddbf2008-02-08 04:18:57 -0800138 msg_init_ns(&init_ipc_ns);
Kirill Korotaev1e786932006-10-02 02:18:21 -0700139 ipc_init_proc_interface("sysvipc/msg",
140 " key msqid perms cbytes qnum lspid lrpid uid gid cuid cgid stime rtime ctime\n",
141 IPC_MSG_IDS, sysvipc_msg_proc_show);
142}
143
Nadia Derbey3e148c72007-10-18 23:40:54 -0700144/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700145 * msg_lock_(check_) routines are called in the paths where the rw_mutex
146 * is not held.
147 */
Nadia Derbey023a5352007-10-18 23:40:51 -0700148static inline struct msg_queue *msg_lock(struct ipc_namespace *ns, int id)
149{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700150 struct kern_ipc_perm *ipcp = ipc_lock(&msg_ids(ns), id);
151
Pierre Peifferb1ed88b2008-02-06 01:36:23 -0800152 if (IS_ERR(ipcp))
153 return (struct msg_queue *)ipcp;
154
Nadia Derbey03f02c72007-10-18 23:40:51 -0700155 return container_of(ipcp, struct msg_queue, q_perm);
Nadia Derbey023a5352007-10-18 23:40:51 -0700156}
157
158static inline struct msg_queue *msg_lock_check(struct ipc_namespace *ns,
159 int id)
160{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700161 struct kern_ipc_perm *ipcp = ipc_lock_check(&msg_ids(ns), id);
162
Pierre Peifferb1ed88b2008-02-06 01:36:23 -0800163 if (IS_ERR(ipcp))
164 return (struct msg_queue *)ipcp;
165
Nadia Derbey03f02c72007-10-18 23:40:51 -0700166 return container_of(ipcp, struct msg_queue, q_perm);
Nadia Derbey023a5352007-10-18 23:40:51 -0700167}
168
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700169static inline void msg_rmid(struct ipc_namespace *ns, struct msg_queue *s)
170{
171 ipc_rmid(&msg_ids(ns), &s->q_perm);
172}
173
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700174/**
175 * newque - Create a new msg queue
176 * @ns: namespace
177 * @params: ptr to the structure that contains the key and msgflg
178 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700179 * Called with msg_ids.rw_mutex held (writer)
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700180 */
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700181static int newque(struct ipc_namespace *ns, struct ipc_params *params)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 struct msg_queue *msq;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700184 int id, retval;
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700185 key_t key = params->key;
186 int msgflg = params->flg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187
Ingo Molnar5a06a362006-07-30 03:04:11 -0700188 msq = ipc_rcu_alloc(sizeof(*msq));
189 if (!msq)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 return -ENOMEM;
191
Ingo Molnar5a06a362006-07-30 03:04:11 -0700192 msq->q_perm.mode = msgflg & S_IRWXUGO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 msq->q_perm.key = key;
194
195 msq->q_perm.security = NULL;
196 retval = security_msg_queue_alloc(msq);
197 if (retval) {
198 ipc_rcu_putref(msq);
199 return retval;
200 }
201
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700202 /*
203 * ipc_addid() locks msq
204 */
Kirill Korotaev1e786932006-10-02 02:18:21 -0700205 id = ipc_addid(&msg_ids(ns), &msq->q_perm, ns->msg_ctlmni);
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700206 if (id < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 security_msg_queue_free(msq);
208 ipc_rcu_putref(msq);
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700209 return id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210 }
211
212 msq->q_stime = msq->q_rtime = 0;
213 msq->q_ctime = get_seconds();
214 msq->q_cbytes = msq->q_qnum = 0;
Kirill Korotaev1e786932006-10-02 02:18:21 -0700215 msq->q_qbytes = ns->msg_ctlmnb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 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);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700220
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 msg_unlock(msq);
222
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700223 return msq->q_perm.id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224}
225
Ingo Molnar5a06a362006-07-30 03:04:11 -0700226static inline void ss_add(struct msg_queue *msq, struct msg_sender *mss)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227{
Ingo Molnar5a06a362006-07-30 03:04:11 -0700228 mss->tsk = current;
229 current->state = TASK_INTERRUPTIBLE;
230 list_add_tail(&mss->list, &msq->q_senders);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231}
232
Ingo Molnar5a06a362006-07-30 03:04:11 -0700233static inline void ss_del(struct msg_sender *mss)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234{
Ingo Molnar5a06a362006-07-30 03:04:11 -0700235 if (mss->list.next != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 list_del(&mss->list);
237}
238
Ingo Molnar5a06a362006-07-30 03:04:11 -0700239static void ss_wakeup(struct list_head *h, int kill)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240{
241 struct list_head *tmp;
242
243 tmp = h->next;
244 while (tmp != h) {
Ingo Molnar5a06a362006-07-30 03:04:11 -0700245 struct msg_sender *mss;
246
247 mss = list_entry(tmp, struct msg_sender, list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 tmp = tmp->next;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700249 if (kill)
250 mss->list.next = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 wake_up_process(mss->tsk);
252 }
253}
254
Ingo Molnar5a06a362006-07-30 03:04:11 -0700255static void expunge_all(struct msg_queue *msq, int res)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256{
257 struct list_head *tmp;
258
259 tmp = msq->q_receivers.next;
260 while (tmp != &msq->q_receivers) {
Ingo Molnar5a06a362006-07-30 03:04:11 -0700261 struct msg_receiver *msr;
262
263 msr = list_entry(tmp, struct msg_receiver, r_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 tmp = tmp->next;
265 msr->r_msg = NULL;
266 wake_up_process(msr->r_tsk);
267 smp_mb();
268 msr->r_msg = ERR_PTR(res);
269 }
270}
Ingo Molnar5a06a362006-07-30 03:04:11 -0700271
272/*
273 * freeque() wakes up waiters on the sender and receiver waiting queue,
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700274 * removes the message queue from message queue ID IDR, and cleans up all the
275 * messages associated with this queue.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700277 * msg_ids.rw_mutex (writer) and the spinlock for this message queue are held
278 * before freeque() is called. msg_ids.rw_mutex remains locked on exit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 */
Pierre Peiffer01b8b072008-02-08 04:18:57 -0800280static void freeque(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281{
282 struct list_head *tmp;
Pierre Peiffer01b8b072008-02-08 04:18:57 -0800283 struct msg_queue *msq = container_of(ipcp, struct msg_queue, q_perm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284
Ingo Molnar5a06a362006-07-30 03:04:11 -0700285 expunge_all(msq, -EIDRM);
286 ss_wakeup(&msq->q_senders, 1);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700287 msg_rmid(ns, msq);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 msg_unlock(msq);
Ingo Molnar5a06a362006-07-30 03:04:11 -0700289
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 tmp = msq->q_messages.next;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700291 while (tmp != &msq->q_messages) {
292 struct msg_msg *msg = list_entry(tmp, struct msg_msg, m_list);
293
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 tmp = tmp->next;
Kirill Korotaev3ac88a42007-10-18 23:40:56 -0700295 atomic_dec(&ns->msg_hdrs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296 free_msg(msg);
297 }
Kirill Korotaev3ac88a42007-10-18 23:40:56 -0700298 atomic_sub(msq->q_cbytes, &ns->msg_bytes);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 security_msg_queue_free(msq);
300 ipc_rcu_putref(msq);
301}
302
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700303/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700304 * Called with msg_ids.rw_mutex and ipcp locked.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700305 */
Nadia Derbey03f02c72007-10-18 23:40:51 -0700306static inline int msg_security(struct kern_ipc_perm *ipcp, int msgflg)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700307{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700308 struct msg_queue *msq = container_of(ipcp, struct msg_queue, q_perm);
309
310 return security_msg_queue_associate(msq, msgflg);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700311}
312
Ingo Molnar5a06a362006-07-30 03:04:11 -0700313asmlinkage long sys_msgget(key_t key, int msgflg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314{
Kirill Korotaev1e786932006-10-02 02:18:21 -0700315 struct ipc_namespace *ns;
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700316 struct ipc_ops msg_ops;
317 struct ipc_params msg_params;
Kirill Korotaev1e786932006-10-02 02:18:21 -0700318
319 ns = current->nsproxy->ipc_ns;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700320
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700321 msg_ops.getnew = newque;
322 msg_ops.associate = msg_security;
323 msg_ops.more_checks = NULL;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700324
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700325 msg_params.key = key;
326 msg_params.flg = msgflg;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700327
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700328 return ipcget(ns, &msg_ids(ns), &msg_ops, &msg_params);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329}
330
Ingo Molnar5a06a362006-07-30 03:04:11 -0700331static inline unsigned long
332copy_msqid_to_user(void __user *buf, struct msqid64_ds *in, int version)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333{
334 switch(version) {
335 case IPC_64:
Ingo Molnar5a06a362006-07-30 03:04:11 -0700336 return copy_to_user(buf, in, sizeof(*in));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 case IPC_OLD:
Ingo Molnar5a06a362006-07-30 03:04:11 -0700338 {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339 struct msqid_ds out;
340
Ingo Molnar5a06a362006-07-30 03:04:11 -0700341 memset(&out, 0, sizeof(out));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342
343 ipc64_perm_to_ipc_perm(&in->msg_perm, &out.msg_perm);
344
345 out.msg_stime = in->msg_stime;
346 out.msg_rtime = in->msg_rtime;
347 out.msg_ctime = in->msg_ctime;
348
Zhang, Yanmin44f564a2008-04-29 01:00:55 -0700349 if (in->msg_cbytes > USHORT_MAX)
350 out.msg_cbytes = USHORT_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 else
352 out.msg_cbytes = in->msg_cbytes;
353 out.msg_lcbytes = in->msg_cbytes;
354
Zhang, Yanmin44f564a2008-04-29 01:00:55 -0700355 if (in->msg_qnum > USHORT_MAX)
356 out.msg_qnum = USHORT_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 else
358 out.msg_qnum = in->msg_qnum;
359
Zhang, Yanmin44f564a2008-04-29 01:00:55 -0700360 if (in->msg_qbytes > USHORT_MAX)
361 out.msg_qbytes = USHORT_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 else
363 out.msg_qbytes = in->msg_qbytes;
364 out.msg_lqbytes = in->msg_qbytes;
365
366 out.msg_lspid = in->msg_lspid;
367 out.msg_lrpid = in->msg_lrpid;
368
Ingo Molnar5a06a362006-07-30 03:04:11 -0700369 return copy_to_user(buf, &out, sizeof(out));
370 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 default:
372 return -EINVAL;
373 }
374}
375
Ingo Molnar5a06a362006-07-30 03:04:11 -0700376static inline unsigned long
Pierre Peiffer016d7132008-04-29 01:00:50 -0700377copy_msqid_from_user(struct msqid64_ds *out, void __user *buf, int version)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378{
379 switch(version) {
380 case IPC_64:
Pierre Peiffer016d7132008-04-29 01:00:50 -0700381 if (copy_from_user(out, buf, sizeof(*out)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 case IPC_OLD:
Ingo Molnar5a06a362006-07-30 03:04:11 -0700385 {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 struct msqid_ds tbuf_old;
387
Ingo Molnar5a06a362006-07-30 03:04:11 -0700388 if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 return -EFAULT;
390
Pierre Peiffer016d7132008-04-29 01:00:50 -0700391 out->msg_perm.uid = tbuf_old.msg_perm.uid;
392 out->msg_perm.gid = tbuf_old.msg_perm.gid;
393 out->msg_perm.mode = tbuf_old.msg_perm.mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394
Ingo Molnar5a06a362006-07-30 03:04:11 -0700395 if (tbuf_old.msg_qbytes == 0)
Pierre Peiffer016d7132008-04-29 01:00:50 -0700396 out->msg_qbytes = tbuf_old.msg_lqbytes;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 else
Pierre Peiffer016d7132008-04-29 01:00:50 -0700398 out->msg_qbytes = tbuf_old.msg_qbytes;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399
400 return 0;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700401 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 default:
403 return -EINVAL;
404 }
405}
406
Pierre Peiffera0d092f2008-04-29 01:00:48 -0700407/*
408 * This function handles some msgctl commands which require the rw_mutex
409 * to be held in write mode.
410 * NOTE: no locks must be held, the rw_mutex is taken inside this function.
411 */
412static int msgctl_down(struct ipc_namespace *ns, int msqid, int cmd,
413 struct msqid_ds __user *buf, int version)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 struct kern_ipc_perm *ipcp;
Pierre Peiffer016d7132008-04-29 01:00:50 -0700416 struct msqid64_ds msqid64;
Pierre Peiffera0d092f2008-04-29 01:00:48 -0700417 struct msg_queue *msq;
418 int err;
419
420 if (cmd == IPC_SET) {
Pierre Peiffer016d7132008-04-29 01:00:50 -0700421 if (copy_msqid_from_user(&msqid64, buf, version))
Pierre Peiffera0d092f2008-04-29 01:00:48 -0700422 return -EFAULT;
423 }
424
Pierre Peiffera5f75e72008-04-29 01:00:54 -0700425 ipcp = ipcctl_pre_down(&msg_ids(ns), msqid, cmd,
426 &msqid64.msg_perm, msqid64.msg_qbytes);
427 if (IS_ERR(ipcp))
428 return PTR_ERR(ipcp);
Pierre Peiffera0d092f2008-04-29 01:00:48 -0700429
Pierre Peiffera5f75e72008-04-29 01:00:54 -0700430 msq = container_of(ipcp, struct msg_queue, q_perm);
Pierre Peiffera0d092f2008-04-29 01:00:48 -0700431
432 err = security_msg_queue_msgctl(msq, cmd);
433 if (err)
434 goto out_unlock;
435
436 switch (cmd) {
437 case IPC_RMID:
438 freeque(ns, ipcp);
439 goto out_up;
440 case IPC_SET:
Pierre Peiffer016d7132008-04-29 01:00:50 -0700441 if (msqid64.msg_qbytes > ns->msg_ctlmnb &&
Pierre Peiffera0d092f2008-04-29 01:00:48 -0700442 !capable(CAP_SYS_RESOURCE)) {
443 err = -EPERM;
444 goto out_unlock;
445 }
446
Pierre Peiffer016d7132008-04-29 01:00:50 -0700447 msq->q_qbytes = msqid64.msg_qbytes;
Pierre Peiffera0d092f2008-04-29 01:00:48 -0700448
Pierre Peiffer8f4a3802008-04-29 01:00:51 -0700449 ipc_update_perm(&msqid64.msg_perm, ipcp);
Pierre Peiffera0d092f2008-04-29 01:00:48 -0700450 msq->q_ctime = get_seconds();
451 /* sleeping receivers might be excluded by
452 * stricter permissions.
453 */
454 expunge_all(msq, -EAGAIN);
455 /* sleeping senders might be able to send
456 * due to a larger queue size.
457 */
458 ss_wakeup(&msq->q_senders, 0);
459 break;
460 default:
461 err = -EINVAL;
462 }
463out_unlock:
464 msg_unlock(msq);
465out_up:
466 up_write(&msg_ids(ns).rw_mutex);
467 return err;
468}
469
470asmlinkage long sys_msgctl(int msqid, int cmd, struct msqid_ds __user *buf)
471{
Ingo Molnar5a06a362006-07-30 03:04:11 -0700472 struct msg_queue *msq;
473 int err, version;
Kirill Korotaev1e786932006-10-02 02:18:21 -0700474 struct ipc_namespace *ns;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700475
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 if (msqid < 0 || cmd < 0)
477 return -EINVAL;
478
479 version = ipc_parse_version(&cmd);
Kirill Korotaev1e786932006-10-02 02:18:21 -0700480 ns = current->nsproxy->ipc_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481
482 switch (cmd) {
Ingo Molnar5a06a362006-07-30 03:04:11 -0700483 case IPC_INFO:
484 case MSG_INFO:
485 {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 struct msginfo msginfo;
487 int max_id;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700488
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 if (!buf)
490 return -EFAULT;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700491 /*
492 * We must not return kernel stack data.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 * due to padding, it's not enough
494 * to set all member fields.
495 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 err = security_msg_queue_msgctl(NULL, cmd);
497 if (err)
498 return err;
499
Ingo Molnar5a06a362006-07-30 03:04:11 -0700500 memset(&msginfo, 0, sizeof(msginfo));
Kirill Korotaev1e786932006-10-02 02:18:21 -0700501 msginfo.msgmni = ns->msg_ctlmni;
502 msginfo.msgmax = ns->msg_ctlmax;
503 msginfo.msgmnb = ns->msg_ctlmnb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 msginfo.msgssz = MSGSSZ;
505 msginfo.msgseg = MSGSEG;
Nadia Derbey3e148c72007-10-18 23:40:54 -0700506 down_read(&msg_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 if (cmd == MSG_INFO) {
Kirill Korotaev1e786932006-10-02 02:18:21 -0700508 msginfo.msgpool = msg_ids(ns).in_use;
Kirill Korotaev3ac88a42007-10-18 23:40:56 -0700509 msginfo.msgmap = atomic_read(&ns->msg_hdrs);
510 msginfo.msgtql = atomic_read(&ns->msg_bytes);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 } else {
512 msginfo.msgmap = MSGMAP;
513 msginfo.msgpool = MSGPOOL;
514 msginfo.msgtql = MSGTQL;
515 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700516 max_id = ipc_get_maxid(&msg_ids(ns));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700517 up_read(&msg_ids(ns).rw_mutex);
Ingo Molnar5a06a362006-07-30 03:04:11 -0700518 if (copy_to_user(buf, &msginfo, sizeof(struct msginfo)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 return -EFAULT;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700520 return (max_id < 0) ? 0 : max_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700522 case MSG_STAT: /* msqid is an index rather than a msg queue id */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 case IPC_STAT:
524 {
525 struct msqid64_ds tbuf;
526 int success_return;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700527
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 if (!buf)
529 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530
Ingo Molnar5a06a362006-07-30 03:04:11 -0700531 if (cmd == MSG_STAT) {
Nadia Derbey023a5352007-10-18 23:40:51 -0700532 msq = msg_lock(ns, msqid);
533 if (IS_ERR(msq))
534 return PTR_ERR(msq);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700535 success_return = msq->q_perm.id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 } else {
Nadia Derbey023a5352007-10-18 23:40:51 -0700537 msq = msg_lock_check(ns, msqid);
538 if (IS_ERR(msq))
539 return PTR_ERR(msq);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 success_return = 0;
541 }
542 err = -EACCES;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700543 if (ipcperms(&msq->q_perm, S_IRUGO))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 goto out_unlock;
545
546 err = security_msg_queue_msgctl(msq, cmd);
547 if (err)
548 goto out_unlock;
549
Nadia Derbey023a5352007-10-18 23:40:51 -0700550 memset(&tbuf, 0, sizeof(tbuf));
551
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 kernel_to_ipc64_perm(&msq->q_perm, &tbuf.msg_perm);
553 tbuf.msg_stime = msq->q_stime;
554 tbuf.msg_rtime = msq->q_rtime;
555 tbuf.msg_ctime = msq->q_ctime;
556 tbuf.msg_cbytes = msq->q_cbytes;
557 tbuf.msg_qnum = msq->q_qnum;
558 tbuf.msg_qbytes = msq->q_qbytes;
559 tbuf.msg_lspid = msq->q_lspid;
560 tbuf.msg_lrpid = msq->q_lrpid;
561 msg_unlock(msq);
562 if (copy_msqid_to_user(buf, &tbuf, version))
563 return -EFAULT;
564 return success_return;
565 }
566 case IPC_SET:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 case IPC_RMID:
Pierre Peiffera0d092f2008-04-29 01:00:48 -0700568 err = msgctl_down(ns, msqid, cmd, buf, version);
569 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 default:
571 return -EINVAL;
572 }
573
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574out_unlock:
575 msg_unlock(msq);
576 return err;
577}
578
Ingo Molnar5a06a362006-07-30 03:04:11 -0700579static int testmsg(struct msg_msg *msg, long type, int mode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580{
581 switch(mode)
582 {
583 case SEARCH_ANY:
584 return 1;
585 case SEARCH_LESSEQUAL:
Ingo Molnar5a06a362006-07-30 03:04:11 -0700586 if (msg->m_type <=type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 return 1;
588 break;
589 case SEARCH_EQUAL:
Ingo Molnar5a06a362006-07-30 03:04:11 -0700590 if (msg->m_type == type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 return 1;
592 break;
593 case SEARCH_NOTEQUAL:
Ingo Molnar5a06a362006-07-30 03:04:11 -0700594 if (msg->m_type != type)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 return 1;
596 break;
597 }
598 return 0;
599}
600
Ingo Molnar5a06a362006-07-30 03:04:11 -0700601static inline int pipelined_send(struct msg_queue *msq, struct msg_msg *msg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602{
Ingo Molnar5a06a362006-07-30 03:04:11 -0700603 struct list_head *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604
605 tmp = msq->q_receivers.next;
606 while (tmp != &msq->q_receivers) {
Ingo Molnar5a06a362006-07-30 03:04:11 -0700607 struct msg_receiver *msr;
608
609 msr = list_entry(tmp, struct msg_receiver, r_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610 tmp = tmp->next;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700611 if (testmsg(msg, msr->r_msgtype, msr->r_mode) &&
612 !security_msg_queue_msgrcv(msq, msg, msr->r_tsk,
613 msr->r_msgtype, msr->r_mode)) {
614
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 list_del(&msr->r_list);
Ingo Molnar5a06a362006-07-30 03:04:11 -0700616 if (msr->r_maxsize < msg->m_ts) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 msr->r_msg = NULL;
618 wake_up_process(msr->r_tsk);
619 smp_mb();
620 msr->r_msg = ERR_PTR(-E2BIG);
621 } else {
622 msr->r_msg = NULL;
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700623 msq->q_lrpid = task_pid_vnr(msr->r_tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 msq->q_rtime = get_seconds();
625 wake_up_process(msr->r_tsk);
626 smp_mb();
627 msr->r_msg = msg;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700628
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 return 1;
630 }
631 }
632 }
633 return 0;
634}
635
suzuki651971c2006-12-06 20:37:48 -0800636long do_msgsnd(int msqid, long mtype, void __user *mtext,
637 size_t msgsz, int msgflg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638{
639 struct msg_queue *msq;
640 struct msg_msg *msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 int err;
Kirill Korotaev1e786932006-10-02 02:18:21 -0700642 struct ipc_namespace *ns;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700643
Kirill Korotaev1e786932006-10-02 02:18:21 -0700644 ns = current->nsproxy->ipc_ns;
645
646 if (msgsz > ns->msg_ctlmax || (long) msgsz < 0 || msqid < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648 if (mtype < 1)
649 return -EINVAL;
650
suzuki651971c2006-12-06 20:37:48 -0800651 msg = load_msg(mtext, msgsz);
Ingo Molnar5a06a362006-07-30 03:04:11 -0700652 if (IS_ERR(msg))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 return PTR_ERR(msg);
654
655 msg->m_type = mtype;
656 msg->m_ts = msgsz;
657
Nadia Derbey023a5352007-10-18 23:40:51 -0700658 msq = msg_lock_check(ns, msqid);
659 if (IS_ERR(msq)) {
660 err = PTR_ERR(msq);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661 goto out_free;
Nadia Derbey023a5352007-10-18 23:40:51 -0700662 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663
664 for (;;) {
665 struct msg_sender s;
666
Ingo Molnar5a06a362006-07-30 03:04:11 -0700667 err = -EACCES;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 if (ipcperms(&msq->q_perm, S_IWUGO))
669 goto out_unlock_free;
670
671 err = security_msg_queue_msgsnd(msq, msg, msgflg);
672 if (err)
673 goto out_unlock_free;
674
Ingo Molnar5a06a362006-07-30 03:04:11 -0700675 if (msgsz + msq->q_cbytes <= msq->q_qbytes &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 1 + msq->q_qnum <= msq->q_qbytes) {
677 break;
678 }
679
680 /* queue full, wait: */
Ingo Molnar5a06a362006-07-30 03:04:11 -0700681 if (msgflg & IPC_NOWAIT) {
682 err = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 goto out_unlock_free;
684 }
685 ss_add(msq, &s);
686 ipc_rcu_getref(msq);
687 msg_unlock(msq);
688 schedule();
689
690 ipc_lock_by_ptr(&msq->q_perm);
691 ipc_rcu_putref(msq);
692 if (msq->q_perm.deleted) {
693 err = -EIDRM;
694 goto out_unlock_free;
695 }
696 ss_del(&s);
Ingo Molnar5a06a362006-07-30 03:04:11 -0700697
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 if (signal_pending(current)) {
Ingo Molnar5a06a362006-07-30 03:04:11 -0700699 err = -ERESTARTNOHAND;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 goto out_unlock_free;
701 }
702 }
703
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700704 msq->q_lspid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 msq->q_stime = get_seconds();
706
Ingo Molnar5a06a362006-07-30 03:04:11 -0700707 if (!pipelined_send(msq, msg)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 /* noone is waiting for this message, enqueue it */
Ingo Molnar5a06a362006-07-30 03:04:11 -0700709 list_add_tail(&msg->m_list, &msq->q_messages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 msq->q_cbytes += msgsz;
711 msq->q_qnum++;
Kirill Korotaev3ac88a42007-10-18 23:40:56 -0700712 atomic_add(msgsz, &ns->msg_bytes);
713 atomic_inc(&ns->msg_hdrs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 }
Ingo Molnar5a06a362006-07-30 03:04:11 -0700715
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 err = 0;
717 msg = NULL;
718
719out_unlock_free:
720 msg_unlock(msq);
721out_free:
Ingo Molnar5a06a362006-07-30 03:04:11 -0700722 if (msg != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 free_msg(msg);
724 return err;
725}
726
suzuki651971c2006-12-06 20:37:48 -0800727asmlinkage long
728sys_msgsnd(int msqid, struct msgbuf __user *msgp, size_t msgsz, int msgflg)
729{
730 long mtype;
731
732 if (get_user(mtype, &msgp->mtype))
733 return -EFAULT;
734 return do_msgsnd(msqid, mtype, msgp->mtext, msgsz, msgflg);
735}
736
Ingo Molnar5a06a362006-07-30 03:04:11 -0700737static inline int convert_mode(long *msgtyp, int msgflg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738{
Ingo Molnar5a06a362006-07-30 03:04:11 -0700739 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 * find message of correct type.
741 * msgtyp = 0 => get first.
742 * msgtyp > 0 => get first message of matching type.
Ingo Molnar5a06a362006-07-30 03:04:11 -0700743 * msgtyp < 0 => get message with least type must be < abs(msgtype).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 */
Ingo Molnar5a06a362006-07-30 03:04:11 -0700745 if (*msgtyp == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 return SEARCH_ANY;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700747 if (*msgtyp < 0) {
748 *msgtyp = -*msgtyp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 return SEARCH_LESSEQUAL;
750 }
Ingo Molnar5a06a362006-07-30 03:04:11 -0700751 if (msgflg & MSG_EXCEPT)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 return SEARCH_NOTEQUAL;
753 return SEARCH_EQUAL;
754}
755
suzuki651971c2006-12-06 20:37:48 -0800756long do_msgrcv(int msqid, long *pmtype, void __user *mtext,
757 size_t msgsz, long msgtyp, int msgflg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758{
759 struct msg_queue *msq;
760 struct msg_msg *msg;
761 int mode;
Kirill Korotaev1e786932006-10-02 02:18:21 -0700762 struct ipc_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763
764 if (msqid < 0 || (long) msgsz < 0)
765 return -EINVAL;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700766 mode = convert_mode(&msgtyp, msgflg);
Kirill Korotaev1e786932006-10-02 02:18:21 -0700767 ns = current->nsproxy->ipc_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768
Nadia Derbey023a5352007-10-18 23:40:51 -0700769 msq = msg_lock_check(ns, msqid);
770 if (IS_ERR(msq))
771 return PTR_ERR(msq);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772
773 for (;;) {
774 struct msg_receiver msr_d;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700775 struct list_head *tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776
777 msg = ERR_PTR(-EACCES);
Ingo Molnar5a06a362006-07-30 03:04:11 -0700778 if (ipcperms(&msq->q_perm, S_IRUGO))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 goto out_unlock;
780
781 msg = ERR_PTR(-EAGAIN);
782 tmp = msq->q_messages.next;
783 while (tmp != &msq->q_messages) {
784 struct msg_msg *walk_msg;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700785
786 walk_msg = list_entry(tmp, struct msg_msg, m_list);
787 if (testmsg(walk_msg, msgtyp, mode) &&
788 !security_msg_queue_msgrcv(msq, walk_msg, current,
789 msgtyp, mode)) {
790
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 msg = walk_msg;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700792 if (mode == SEARCH_LESSEQUAL &&
793 walk_msg->m_type != 1) {
794 msg = walk_msg;
795 msgtyp = walk_msg->m_type - 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 } else {
Ingo Molnar5a06a362006-07-30 03:04:11 -0700797 msg = walk_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 break;
799 }
800 }
801 tmp = tmp->next;
802 }
Ingo Molnar5a06a362006-07-30 03:04:11 -0700803 if (!IS_ERR(msg)) {
804 /*
805 * Found a suitable message.
806 * Unlink it from the queue.
807 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 if ((msgsz < msg->m_ts) && !(msgflg & MSG_NOERROR)) {
809 msg = ERR_PTR(-E2BIG);
810 goto out_unlock;
811 }
812 list_del(&msg->m_list);
813 msq->q_qnum--;
814 msq->q_rtime = get_seconds();
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700815 msq->q_lrpid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 msq->q_cbytes -= msg->m_ts;
Kirill Korotaev3ac88a42007-10-18 23:40:56 -0700817 atomic_sub(msg->m_ts, &ns->msg_bytes);
818 atomic_dec(&ns->msg_hdrs);
Ingo Molnar5a06a362006-07-30 03:04:11 -0700819 ss_wakeup(&msq->q_senders, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 msg_unlock(msq);
821 break;
822 }
823 /* No message waiting. Wait for a message */
824 if (msgflg & IPC_NOWAIT) {
825 msg = ERR_PTR(-ENOMSG);
826 goto out_unlock;
827 }
Ingo Molnar5a06a362006-07-30 03:04:11 -0700828 list_add_tail(&msr_d.r_list, &msq->q_receivers);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 msr_d.r_tsk = current;
830 msr_d.r_msgtype = msgtyp;
831 msr_d.r_mode = mode;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700832 if (msgflg & MSG_NOERROR)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 msr_d.r_maxsize = INT_MAX;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700834 else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 msr_d.r_maxsize = msgsz;
836 msr_d.r_msg = ERR_PTR(-EAGAIN);
837 current->state = TASK_INTERRUPTIBLE;
838 msg_unlock(msq);
839
840 schedule();
841
842 /* Lockless receive, part 1:
843 * Disable preemption. We don't hold a reference to the queue
844 * and getting a reference would defeat the idea of a lockless
845 * operation, thus the code relies on rcu to guarantee the
846 * existance of msq:
847 * Prior to destruction, expunge_all(-EIRDM) changes r_msg.
848 * Thus if r_msg is -EAGAIN, then the queue not yet destroyed.
849 * rcu_read_lock() prevents preemption between reading r_msg
850 * and the spin_lock() inside ipc_lock_by_ptr().
851 */
852 rcu_read_lock();
853
854 /* Lockless receive, part 2:
855 * Wait until pipelined_send or expunge_all are outside of
856 * wake_up_process(). There is a race with exit(), see
857 * ipc/mqueue.c for the details.
858 */
Ingo Molnar5a06a362006-07-30 03:04:11 -0700859 msg = (struct msg_msg*)msr_d.r_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 while (msg == NULL) {
861 cpu_relax();
Ingo Molnar5a06a362006-07-30 03:04:11 -0700862 msg = (struct msg_msg *)msr_d.r_msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 }
864
865 /* Lockless receive, part 3:
866 * If there is a message or an error then accept it without
867 * locking.
868 */
Ingo Molnar5a06a362006-07-30 03:04:11 -0700869 if (msg != ERR_PTR(-EAGAIN)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870 rcu_read_unlock();
871 break;
872 }
873
874 /* Lockless receive, part 3:
875 * Acquire the queue spinlock.
876 */
877 ipc_lock_by_ptr(&msq->q_perm);
878 rcu_read_unlock();
879
880 /* Lockless receive, part 4:
881 * Repeat test after acquiring the spinlock.
882 */
883 msg = (struct msg_msg*)msr_d.r_msg;
Ingo Molnar5a06a362006-07-30 03:04:11 -0700884 if (msg != ERR_PTR(-EAGAIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885 goto out_unlock;
886
887 list_del(&msr_d.r_list);
888 if (signal_pending(current)) {
889 msg = ERR_PTR(-ERESTARTNOHAND);
890out_unlock:
891 msg_unlock(msq);
892 break;
893 }
894 }
895 if (IS_ERR(msg))
Ingo Molnar5a06a362006-07-30 03:04:11 -0700896 return PTR_ERR(msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897
898 msgsz = (msgsz > msg->m_ts) ? msg->m_ts : msgsz;
suzuki651971c2006-12-06 20:37:48 -0800899 *pmtype = msg->m_type;
900 if (store_msg(mtext, msg, msgsz))
Ingo Molnar5a06a362006-07-30 03:04:11 -0700901 msgsz = -EFAULT;
suzuki651971c2006-12-06 20:37:48 -0800902
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 free_msg(msg);
Ingo Molnar5a06a362006-07-30 03:04:11 -0700904
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 return msgsz;
906}
907
suzuki651971c2006-12-06 20:37:48 -0800908asmlinkage long sys_msgrcv(int msqid, struct msgbuf __user *msgp, size_t msgsz,
909 long msgtyp, int msgflg)
910{
911 long err, mtype;
912
913 err = do_msgrcv(msqid, &mtype, msgp->mtext, msgsz, msgtyp, msgflg);
914 if (err < 0)
915 goto out;
916
917 if (put_user(mtype, &msgp->mtype))
918 err = -EFAULT;
919out:
920 return err;
921}
922
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923#ifdef CONFIG_PROC_FS
Mike Waychison19b49462005-09-06 15:17:10 -0700924static int sysvipc_msg_proc_show(struct seq_file *s, void *it)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925{
Mike Waychison19b49462005-09-06 15:17:10 -0700926 struct msg_queue *msq = it;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927
Mike Waychison19b49462005-09-06 15:17:10 -0700928 return seq_printf(s,
Ingo Molnar5a06a362006-07-30 03:04:11 -0700929 "%10d %10d %4o %10lu %10lu %5u %5u %5u %5u %5u %5u %10lu %10lu %10lu\n",
930 msq->q_perm.key,
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700931 msq->q_perm.id,
Ingo Molnar5a06a362006-07-30 03:04:11 -0700932 msq->q_perm.mode,
933 msq->q_cbytes,
934 msq->q_qnum,
935 msq->q_lspid,
936 msq->q_lrpid,
937 msq->q_perm.uid,
938 msq->q_perm.gid,
939 msq->q_perm.cuid,
940 msq->q_perm.cgid,
941 msq->q_stime,
942 msq->q_rtime,
943 msq->q_ctime);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944}
945#endif