blob: 8ce57691e7b60994d9cc97620b7550c603df5391 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * POSIX message queues filesystem for Linux.
3 *
4 * Copyright (C) 2003,2004 Krzysztof Benedyczak (golbi@mat.uni.torun.pl)
Michal Wronskif66e9282006-10-03 23:23:27 +02005 * Michal Wronski (michal.wronski@gmail.com)
Linus Torvalds1da177e2005-04-16 15:20:36 -07006 *
7 * Spinlocks: Mohamed Abbas (abbas.mohamed@intel.com)
8 * Lockless receive & send, fd based notify:
9 * Manfred Spraul (manfred@colorfullife.com)
10 *
George C. Wilson20ca73b2006-05-24 16:09:55 -050011 * Audit: George Wilson (ltcgcw@us.ibm.com)
12 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070013 * This file is released under the GPL.
14 */
15
Randy.Dunlapc59ede72006-01-11 12:17:46 -080016#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/init.h>
18#include <linux/pagemap.h>
19#include <linux/file.h>
20#include <linux/mount.h>
21#include <linux/namei.h>
22#include <linux/sysctl.h>
23#include <linux/poll.h>
24#include <linux/mqueue.h>
25#include <linux/msg.h>
26#include <linux/skbuff.h>
Doug Ledford5b5c4d12012-05-31 16:26:30 -070027#include <linux/vmalloc.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070028#include <linux/netlink.h>
29#include <linux/syscalls.h>
George C. Wilson20ca73b2006-05-24 16:09:55 -050030#include <linux/audit.h>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070031#include <linux/signal.h>
Ingo Molnar5f921ae2006-03-26 01:37:17 -080032#include <linux/mutex.h>
Pavel Emelyanovb4888932007-10-18 23:40:14 -070033#include <linux/nsproxy.h>
34#include <linux/pid.h>
Serge E. Hallyn614b84c2009-04-06 19:01:08 -070035#include <linux/ipc_namespace.h>
Serge E. Hallyn6b550f92012-01-10 15:11:37 -080036#include <linux/user_namespace.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090037#include <linux/slab.h>
Ingo Molnar5f921ae2006-03-26 01:37:17 -080038
Linus Torvalds1da177e2005-04-16 15:20:36 -070039#include <net/sock.h>
40#include "util.h"
41
42#define MQUEUE_MAGIC 0x19800202
43#define DIRENT_SIZE 20
44#define FILENT_SIZE 80
45
46#define SEND 0
47#define RECV 1
48
49#define STATE_NONE 0
50#define STATE_PENDING 1
51#define STATE_READY 2
52
Doug Ledfordd6629852012-05-31 16:26:35 -070053struct posix_msg_tree_node {
54 struct rb_node rb_node;
55 struct list_head msg_list;
56 int priority;
57};
58
Linus Torvalds1da177e2005-04-16 15:20:36 -070059struct ext_wait_queue { /* queue of sleeping tasks */
60 struct task_struct *task;
61 struct list_head list;
62 struct msg_msg *msg; /* ptr of loaded message */
63 int state; /* one of STATE_* values */
64};
65
66struct mqueue_inode_info {
67 spinlock_t lock;
68 struct inode vfs_inode;
69 wait_queue_head_t wait_q;
70
Doug Ledfordd6629852012-05-31 16:26:35 -070071 struct rb_root msg_tree;
Doug Ledfordce2d52c2012-05-31 16:26:38 -070072 struct posix_msg_tree_node *node_cache;
Linus Torvalds1da177e2005-04-16 15:20:36 -070073 struct mq_attr attr;
74
75 struct sigevent notify;
Cedric Le Goatera03fcb72006-10-02 02:17:26 -070076 struct pid* notify_owner;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -080077 struct user_namespace *notify_user_ns;
Adrian Bunk338cec32005-09-10 00:26:54 -070078 struct user_struct *user; /* user who created, for accounting */
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 struct sock *notify_sock;
80 struct sk_buff *notify_cookie;
81
82 /* for tasks waiting for free space and messages, respectively */
83 struct ext_wait_queue e_wait_q[2];
84
85 unsigned long qsize; /* size of queue in memory (sum of all msgs) */
86};
87
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -080088static const struct inode_operations mqueue_dir_inode_operations;
Arjan van de Ven9a321442007-02-12 00:55:35 -080089static const struct file_operations mqueue_file_operations;
Alexey Dobriyanb87221d2009-09-21 17:01:09 -070090static const struct super_operations mqueue_super_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -070091static void remove_notification(struct mqueue_inode_info *info);
92
Christoph Lametere18b8902006-12-06 20:33:20 -080093static struct kmem_cache *mqueue_inode_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -070094
95static struct ctl_table_header * mq_sysctl_table;
96
97static inline struct mqueue_inode_info *MQUEUE_I(struct inode *inode)
98{
99 return container_of(inode, struct mqueue_inode_info, vfs_inode);
100}
101
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700102/*
103 * This routine should be called with the mq_lock held.
104 */
105static inline struct ipc_namespace *__get_ns_from_inode(struct inode *inode)
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700106{
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700107 return get_ipc_ns(inode->i_sb->s_fs_info);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700108}
109
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700110static struct ipc_namespace *get_ns_from_inode(struct inode *inode)
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700111{
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700112 struct ipc_namespace *ns;
113
114 spin_lock(&mq_lock);
115 ns = __get_ns_from_inode(inode);
116 spin_unlock(&mq_lock);
117 return ns;
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700118}
119
Doug Ledfordd6629852012-05-31 16:26:35 -0700120/* Auxiliary functions to manipulate messages' list */
121static int msg_insert(struct msg_msg *msg, struct mqueue_inode_info *info)
122{
123 struct rb_node **p, *parent = NULL;
124 struct posix_msg_tree_node *leaf;
125
126 p = &info->msg_tree.rb_node;
127 while (*p) {
128 parent = *p;
129 leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node);
130
131 if (likely(leaf->priority == msg->m_type))
132 goto insert_msg;
133 else if (msg->m_type < leaf->priority)
134 p = &(*p)->rb_left;
135 else
136 p = &(*p)->rb_right;
137 }
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700138 if (info->node_cache) {
139 leaf = info->node_cache;
140 info->node_cache = NULL;
141 } else {
142 leaf = kmalloc(sizeof(*leaf), GFP_ATOMIC);
143 if (!leaf)
144 return -ENOMEM;
145 rb_init_node(&leaf->rb_node);
146 INIT_LIST_HEAD(&leaf->msg_list);
147 info->qsize += sizeof(*leaf);
148 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700149 leaf->priority = msg->m_type;
150 rb_link_node(&leaf->rb_node, parent, p);
151 rb_insert_color(&leaf->rb_node, &info->msg_tree);
Doug Ledfordd6629852012-05-31 16:26:35 -0700152insert_msg:
153 info->attr.mq_curmsgs++;
154 info->qsize += msg->m_ts;
155 list_add_tail(&msg->m_list, &leaf->msg_list);
156 return 0;
157}
158
159static inline struct msg_msg *msg_get(struct mqueue_inode_info *info)
160{
161 struct rb_node **p, *parent = NULL;
162 struct posix_msg_tree_node *leaf;
163 struct msg_msg *msg;
164
165try_again:
166 p = &info->msg_tree.rb_node;
167 while (*p) {
168 parent = *p;
169 /*
170 * During insert, low priorities go to the left and high to the
171 * right. On receive, we want the highest priorities first, so
172 * walk all the way to the right.
173 */
174 p = &(*p)->rb_right;
175 }
176 if (!parent) {
177 if (info->attr.mq_curmsgs) {
178 pr_warn_once("Inconsistency in POSIX message queue, "
179 "no tree element, but supposedly messages "
180 "should exist!\n");
181 info->attr.mq_curmsgs = 0;
182 }
183 return NULL;
184 }
185 leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700186 if (unlikely(list_empty(&leaf->msg_list))) {
Doug Ledfordd6629852012-05-31 16:26:35 -0700187 pr_warn_once("Inconsistency in POSIX message queue, "
188 "empty leaf node but we haven't implemented "
189 "lazy leaf delete!\n");
190 rb_erase(&leaf->rb_node, &info->msg_tree);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700191 if (info->node_cache) {
192 info->qsize -= sizeof(*leaf);
193 kfree(leaf);
194 } else {
195 info->node_cache = leaf;
196 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700197 goto try_again;
198 } else {
199 msg = list_first_entry(&leaf->msg_list,
200 struct msg_msg, m_list);
201 list_del(&msg->m_list);
202 if (list_empty(&leaf->msg_list)) {
203 rb_erase(&leaf->rb_node, &info->msg_tree);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700204 if (info->node_cache) {
205 info->qsize -= sizeof(*leaf);
206 kfree(leaf);
207 } else {
208 info->node_cache = leaf;
209 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700210 }
211 }
212 info->attr.mq_curmsgs--;
213 info->qsize -= msg->m_ts;
214 return msg;
215}
216
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700217static struct inode *mqueue_get_inode(struct super_block *sb,
Al Viro1b9d5ff72011-07-24 14:18:20 -0400218 struct ipc_namespace *ipc_ns, umode_t mode,
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700219 struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220{
David Howells86a264a2008-11-14 10:39:18 +1100221 struct user_struct *u = current_user();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 struct inode *inode;
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700223 int ret = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224
225 inode = new_inode(sb);
Jiri Slaby04715202011-07-26 16:08:46 -0700226 if (!inode)
227 goto err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228
Jiri Slaby04715202011-07-26 16:08:46 -0700229 inode->i_ino = get_next_ino();
230 inode->i_mode = mode;
231 inode->i_uid = current_fsuid();
232 inode->i_gid = current_fsgid();
233 inode->i_mtime = inode->i_ctime = inode->i_atime = CURRENT_TIME;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234
Jiri Slaby04715202011-07-26 16:08:46 -0700235 if (S_ISREG(mode)) {
236 struct mqueue_inode_info *info;
Doug Ledfordd6629852012-05-31 16:26:35 -0700237 unsigned long mq_bytes, mq_treesize;
André Goddard Rosac8308b12010-02-23 04:04:23 -0300238
Jiri Slaby04715202011-07-26 16:08:46 -0700239 inode->i_fop = &mqueue_file_operations;
240 inode->i_size = FILENT_SIZE;
241 /* mqueue specific info */
242 info = MQUEUE_I(inode);
243 spin_lock_init(&info->lock);
244 init_waitqueue_head(&info->wait_q);
245 INIT_LIST_HEAD(&info->e_wait_q[0].list);
246 INIT_LIST_HEAD(&info->e_wait_q[1].list);
247 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800248 info->notify_user_ns = NULL;
Jiri Slaby04715202011-07-26 16:08:46 -0700249 info->qsize = 0;
250 info->user = NULL; /* set when all is ok */
Doug Ledfordd6629852012-05-31 16:26:35 -0700251 info->msg_tree = RB_ROOT;
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700252 info->node_cache = NULL;
Jiri Slaby04715202011-07-26 16:08:46 -0700253 memset(&info->attr, 0, sizeof(info->attr));
KOSAKI Motohirocef01842012-05-31 16:26:33 -0700254 info->attr.mq_maxmsg = min(ipc_ns->mq_msg_max,
255 ipc_ns->mq_msg_default);
256 info->attr.mq_msgsize = min(ipc_ns->mq_msgsize_max,
257 ipc_ns->mq_msgsize_default);
Jiri Slaby04715202011-07-26 16:08:46 -0700258 if (attr) {
259 info->attr.mq_maxmsg = attr->mq_maxmsg;
260 info->attr.mq_msgsize = attr->mq_msgsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700262 /*
263 * We used to allocate a static array of pointers and account
264 * the size of that array as well as one msg_msg struct per
265 * possible message into the queue size. That's no longer
266 * accurate as the queue is now an rbtree and will grow and
267 * shrink depending on usage patterns. We can, however, still
268 * account one msg_msg struct per message, but the nodes are
269 * allocated depending on priority usage, and most programs
270 * only use one, or a handful, of priorities. However, since
271 * this is pinned memory, we need to assume worst case, so
272 * that means the min(mq_maxmsg, max_priorities) * struct
273 * posix_msg_tree_node.
274 */
275 mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) +
276 min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) *
277 sizeof(struct posix_msg_tree_node);
Jiri Slaby04715202011-07-26 16:08:46 -0700278
Doug Ledfordd6629852012-05-31 16:26:35 -0700279 mq_bytes = mq_treesize + (info->attr.mq_maxmsg *
280 info->attr.mq_msgsize);
Jiri Slaby04715202011-07-26 16:08:46 -0700281
282 spin_lock(&mq_lock);
283 if (u->mq_bytes + mq_bytes < u->mq_bytes ||
Davidlohr Bueso2a4e64b2012-01-20 14:34:01 -0800284 u->mq_bytes + mq_bytes > rlimit(RLIMIT_MSGQUEUE)) {
Jiri Slaby04715202011-07-26 16:08:46 -0700285 spin_unlock(&mq_lock);
286 /* mqueue_evict_inode() releases info->messages */
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700287 ret = -EMFILE;
Jiri Slaby04715202011-07-26 16:08:46 -0700288 goto out_inode;
289 }
290 u->mq_bytes += mq_bytes;
291 spin_unlock(&mq_lock);
292
293 /* all is ok */
294 info->user = get_uid(u);
295 } else if (S_ISDIR(mode)) {
296 inc_nlink(inode);
297 /* Some things misbehave if size == 0 on a directory */
298 inode->i_size = 2 * DIRENT_SIZE;
299 inode->i_op = &mqueue_dir_inode_operations;
300 inode->i_fop = &simple_dir_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301 }
Jiri Slaby04715202011-07-26 16:08:46 -0700302
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 return inode;
304out_inode:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 iput(inode);
Jiri Slaby04715202011-07-26 16:08:46 -0700306err:
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700307 return ERR_PTR(ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308}
309
310static int mqueue_fill_super(struct super_block *sb, void *data, int silent)
311{
312 struct inode *inode;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700313 struct ipc_namespace *ns = data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314
315 sb->s_blocksize = PAGE_CACHE_SIZE;
316 sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
317 sb->s_magic = MQUEUE_MAGIC;
318 sb->s_op = &mqueue_super_ops;
319
Al Viro48fde702012-01-08 22:15:13 -0500320 inode = mqueue_get_inode(sb, ns, S_IFDIR | S_ISVTX | S_IRWXUGO, NULL);
321 if (IS_ERR(inode))
322 return PTR_ERR(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323
Al Viro48fde702012-01-08 22:15:13 -0500324 sb->s_root = d_make_root(inode);
325 if (!sb->s_root)
326 return -ENOMEM;
327 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328}
329
Al Viroceefda62010-07-26 13:16:50 +0400330static struct dentry *mqueue_mount(struct file_system_type *fs_type,
David Howells454e2392006-06-23 02:02:57 -0700331 int flags, const char *dev_name,
Al Viroceefda62010-07-26 13:16:50 +0400332 void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333{
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700334 if (!(flags & MS_KERNMOUNT))
335 data = current->nsproxy->ipc_ns;
Al Viroceefda62010-07-26 13:16:50 +0400336 return mount_ns(fs_type, flags, data, mqueue_fill_super);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337}
338
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700339static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340{
341 struct mqueue_inode_info *p = (struct mqueue_inode_info *) foo;
342
Christoph Lametera35afb82007-05-16 22:10:57 -0700343 inode_init_once(&p->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344}
345
346static struct inode *mqueue_alloc_inode(struct super_block *sb)
347{
348 struct mqueue_inode_info *ei;
349
Christoph Lametere94b1762006-12-06 20:33:17 -0800350 ei = kmem_cache_alloc(mqueue_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 if (!ei)
352 return NULL;
353 return &ei->vfs_inode;
354}
355
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100356static void mqueue_i_callback(struct rcu_head *head)
357{
358 struct inode *inode = container_of(head, struct inode, i_rcu);
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100359 kmem_cache_free(mqueue_inode_cachep, MQUEUE_I(inode));
360}
361
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362static void mqueue_destroy_inode(struct inode *inode)
363{
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100364 call_rcu(&inode->i_rcu, mqueue_i_callback);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365}
366
Al Viro6d8af642010-06-05 16:29:45 -0400367static void mqueue_evict_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368{
369 struct mqueue_inode_info *info;
370 struct user_struct *user;
Doug Ledfordd6629852012-05-31 16:26:35 -0700371 unsigned long mq_bytes, mq_treesize;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700372 struct ipc_namespace *ipc_ns;
Doug Ledfordd6629852012-05-31 16:26:35 -0700373 struct msg_msg *msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374
Jan Karadbd57682012-05-03 14:48:02 +0200375 clear_inode(inode);
Al Viro6d8af642010-06-05 16:29:45 -0400376
377 if (S_ISDIR(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 return;
Al Viro6d8af642010-06-05 16:29:45 -0400379
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700380 ipc_ns = get_ns_from_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 info = MQUEUE_I(inode);
382 spin_lock(&info->lock);
Doug Ledfordd6629852012-05-31 16:26:35 -0700383 while ((msg = msg_get(info)) != NULL)
384 free_msg(msg);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700385 kfree(info->node_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 spin_unlock(&info->lock);
387
André Goddard Rosa8834cf72010-02-23 04:04:24 -0300388 /* Total amount of bytes accounted for the mqueue */
Doug Ledfordd6629852012-05-31 16:26:35 -0700389 mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) +
390 min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) *
391 sizeof(struct posix_msg_tree_node);
392
393 mq_bytes = mq_treesize + (info->attr.mq_maxmsg *
394 info->attr.mq_msgsize);
395
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 user = info->user;
397 if (user) {
398 spin_lock(&mq_lock);
399 user->mq_bytes -= mq_bytes;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700400 /*
401 * get_ns_from_inode() ensures that the
402 * (ipc_ns = sb->s_fs_info) is either a valid ipc_ns
403 * to which we now hold a reference, or it is NULL.
404 * We can't put it here under mq_lock, though.
405 */
406 if (ipc_ns)
407 ipc_ns->mq_queues_count--;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 spin_unlock(&mq_lock);
409 free_uid(user);
410 }
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700411 if (ipc_ns)
412 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413}
414
415static int mqueue_create(struct inode *dir, struct dentry *dentry,
Al Viro4acdaf22011-07-26 01:42:34 -0400416 umode_t mode, struct nameidata *nd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417{
418 struct inode *inode;
419 struct mq_attr *attr = dentry->d_fsdata;
420 int error;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700421 struct ipc_namespace *ipc_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422
423 spin_lock(&mq_lock);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700424 ipc_ns = __get_ns_from_inode(dir);
425 if (!ipc_ns) {
426 error = -EACCES;
427 goto out_unlock;
428 }
Doug Ledford02967ea2012-05-31 16:26:29 -0700429 if (ipc_ns->mq_queues_count >= HARD_QUEUESMAX ||
430 (ipc_ns->mq_queues_count >= ipc_ns->mq_queues_max &&
431 !capable(CAP_SYS_RESOURCE))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 error = -ENOSPC;
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700433 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 }
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700435 ipc_ns->mq_queues_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 spin_unlock(&mq_lock);
437
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700438 inode = mqueue_get_inode(dir->i_sb, ipc_ns, mode, attr);
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700439 if (IS_ERR(inode)) {
440 error = PTR_ERR(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 spin_lock(&mq_lock);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700442 ipc_ns->mq_queues_count--;
443 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 }
445
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700446 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 dir->i_size += DIRENT_SIZE;
448 dir->i_ctime = dir->i_mtime = dir->i_atime = CURRENT_TIME;
449
450 d_instantiate(dentry, inode);
451 dget(dentry);
452 return 0;
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700453out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 spin_unlock(&mq_lock);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700455 if (ipc_ns)
456 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 return error;
458}
459
460static int mqueue_unlink(struct inode *dir, struct dentry *dentry)
461{
462 struct inode *inode = dentry->d_inode;
463
464 dir->i_ctime = dir->i_mtime = dir->i_atime = CURRENT_TIME;
465 dir->i_size -= DIRENT_SIZE;
Dave Hansen9a53c3a2006-09-30 23:29:03 -0700466 drop_nlink(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 dput(dentry);
468 return 0;
469}
470
471/*
472* This is routine for system read from queue file.
473* To avoid mess with doing here some sort of mq_receive we allow
474* to read only queue size & notification info (the only values
475* that are interesting from user point of view and aren't accessible
476* through std routines)
477*/
478static ssize_t mqueue_read_file(struct file *filp, char __user *u_data,
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700479 size_t count, loff_t *off)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480{
Josef Sipek6d630792006-12-08 02:37:11 -0800481 struct mqueue_inode_info *info = MQUEUE_I(filp->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 char buffer[FILENT_SIZE];
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700483 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
485 spin_lock(&info->lock);
486 snprintf(buffer, sizeof(buffer),
487 "QSIZE:%-10lu NOTIFY:%-5d SIGNO:%-5d NOTIFY_PID:%-6d\n",
488 info->qsize,
489 info->notify_owner ? info->notify.sigev_notify : 0,
490 (info->notify_owner &&
491 info->notify.sigev_notify == SIGEV_SIGNAL) ?
492 info->notify.sigev_signo : 0,
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800493 pid_vnr(info->notify_owner));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 spin_unlock(&info->lock);
495 buffer[sizeof(buffer)-1] = '\0';
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700497 ret = simple_read_from_buffer(u_data, count, off, buffer,
498 strlen(buffer));
499 if (ret <= 0)
500 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501
Josef Sipek6d630792006-12-08 02:37:11 -0800502 filp->f_path.dentry->d_inode->i_atime = filp->f_path.dentry->d_inode->i_ctime = CURRENT_TIME;
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700503 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504}
505
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -0700506static int mqueue_flush_file(struct file *filp, fl_owner_t id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507{
Josef Sipek6d630792006-12-08 02:37:11 -0800508 struct mqueue_inode_info *info = MQUEUE_I(filp->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509
510 spin_lock(&info->lock);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700511 if (task_tgid(current) == info->notify_owner)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 remove_notification(info);
513
514 spin_unlock(&info->lock);
515 return 0;
516}
517
518static unsigned int mqueue_poll_file(struct file *filp, struct poll_table_struct *poll_tab)
519{
Josef Sipek6d630792006-12-08 02:37:11 -0800520 struct mqueue_inode_info *info = MQUEUE_I(filp->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 int retval = 0;
522
523 poll_wait(filp, &info->wait_q, poll_tab);
524
525 spin_lock(&info->lock);
526 if (info->attr.mq_curmsgs)
527 retval = POLLIN | POLLRDNORM;
528
529 if (info->attr.mq_curmsgs < info->attr.mq_maxmsg)
530 retval |= POLLOUT | POLLWRNORM;
531 spin_unlock(&info->lock);
532
533 return retval;
534}
535
536/* Adds current to info->e_wait_q[sr] before element with smaller prio */
537static void wq_add(struct mqueue_inode_info *info, int sr,
538 struct ext_wait_queue *ewp)
539{
540 struct ext_wait_queue *walk;
541
542 ewp->task = current;
543
544 list_for_each_entry(walk, &info->e_wait_q[sr].list, list) {
545 if (walk->task->static_prio <= current->static_prio) {
546 list_add_tail(&ewp->list, &walk->list);
547 return;
548 }
549 }
550 list_add_tail(&ewp->list, &info->e_wait_q[sr].list);
551}
552
553/*
554 * Puts current task to sleep. Caller must hold queue lock. After return
555 * lock isn't held.
556 * sr: SEND or RECV
557 */
558static int wq_sleep(struct mqueue_inode_info *info, int sr,
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200559 ktime_t *timeout, struct ext_wait_queue *ewp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560{
561 int retval;
562 signed long time;
563
564 wq_add(info, sr, ewp);
565
566 for (;;) {
567 set_current_state(TASK_INTERRUPTIBLE);
568
569 spin_unlock(&info->lock);
Wanlong Gao32ea8452011-10-31 17:06:35 -0700570 time = schedule_hrtimeout_range_clock(timeout, 0,
571 HRTIMER_MODE_ABS, CLOCK_REALTIME);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572
573 while (ewp->state == STATE_PENDING)
574 cpu_relax();
575
576 if (ewp->state == STATE_READY) {
577 retval = 0;
578 goto out;
579 }
580 spin_lock(&info->lock);
581 if (ewp->state == STATE_READY) {
582 retval = 0;
583 goto out_unlock;
584 }
585 if (signal_pending(current)) {
586 retval = -ERESTARTSYS;
587 break;
588 }
589 if (time == 0) {
590 retval = -ETIMEDOUT;
591 break;
592 }
593 }
594 list_del(&ewp->list);
595out_unlock:
596 spin_unlock(&info->lock);
597out:
598 return retval;
599}
600
601/*
602 * Returns waiting task that should be serviced first or NULL if none exists
603 */
604static struct ext_wait_queue *wq_get_first_waiter(
605 struct mqueue_inode_info *info, int sr)
606{
607 struct list_head *ptr;
608
609 ptr = info->e_wait_q[sr].list.prev;
610 if (ptr == &info->e_wait_q[sr].list)
611 return NULL;
612 return list_entry(ptr, struct ext_wait_queue, list);
613}
614
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615
616static inline void set_cookie(struct sk_buff *skb, char code)
617{
618 ((char*)skb->data)[NOTIFY_COOKIE_LEN-1] = code;
619}
620
621/*
622 * The next function is only to split too long sys_mq_timedsend
623 */
624static void __do_notify(struct mqueue_inode_info *info)
625{
626 /* notification
627 * invoked when there is registered process and there isn't process
628 * waiting synchronously for message AND state of queue changed from
629 * empty to not empty. Here we are sure that no one is waiting
630 * synchronously. */
631 if (info->notify_owner &&
632 info->attr.mq_curmsgs == 1) {
633 struct siginfo sig_i;
634 switch (info->notify.sigev_notify) {
635 case SIGEV_NONE:
636 break;
637 case SIGEV_SIGNAL:
638 /* sends signal */
639
640 sig_i.si_signo = info->notify.sigev_signo;
641 sig_i.si_errno = 0;
642 sig_i.si_code = SI_MESGQ;
643 sig_i.si_value = info->notify.sigev_value;
Serge E. Hallyn6b550f92012-01-10 15:11:37 -0800644 /* map current pid/uid into info->owner's namespaces */
645 rcu_read_lock();
Sukadev Bhattiprolua6684992009-01-07 18:08:50 -0800646 sig_i.si_pid = task_tgid_nr_ns(current,
647 ns_of_pid(info->notify_owner));
Eric W. Biederman76b6db02012-03-14 15:24:19 -0700648 sig_i.si_uid = from_kuid_munged(info->notify_user_ns, current_uid());
Serge E. Hallyn6b550f92012-01-10 15:11:37 -0800649 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700651 kill_pid_info(info->notify.sigev_signo,
652 &sig_i, info->notify_owner);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 break;
654 case SIGEV_THREAD:
655 set_cookie(info->notify_cookie, NOTIFY_WOKENUP);
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700656 netlink_sendskb(info->notify_sock, info->notify_cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 break;
658 }
659 /* after notification unregisters process */
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700660 put_pid(info->notify_owner);
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800661 put_user_ns(info->notify_user_ns);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700662 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800663 info->notify_user_ns = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 }
665 wake_up(&info->wait_q);
666}
667
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200668static int prepare_timeout(const struct timespec __user *u_abs_timeout,
669 ktime_t *expires, struct timespec *ts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670{
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200671 if (copy_from_user(ts, u_abs_timeout, sizeof(struct timespec)))
672 return -EFAULT;
673 if (!timespec_valid(ts))
674 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200676 *expires = timespec_to_ktime(*ts);
677 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678}
679
680static void remove_notification(struct mqueue_inode_info *info)
681{
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700682 if (info->notify_owner != NULL &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 info->notify.sigev_notify == SIGEV_THREAD) {
684 set_cookie(info->notify_cookie, NOTIFY_REMOVED);
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700685 netlink_sendskb(info->notify_sock, info->notify_cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700687 put_pid(info->notify_owner);
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800688 put_user_ns(info->notify_user_ns);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700689 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800690 info->notify_user_ns = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691}
692
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700693static int mq_attr_ok(struct ipc_namespace *ipc_ns, struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694{
Doug Ledford2c12ea42012-05-31 16:26:36 -0700695 int mq_treesize;
696 unsigned long total_size;
697
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 if (attr->mq_maxmsg <= 0 || attr->mq_msgsize <= 0)
Doug Ledford113289c2012-05-31 16:26:36 -0700699 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 if (capable(CAP_SYS_RESOURCE)) {
Doug Ledford02967ea2012-05-31 16:26:29 -0700701 if (attr->mq_maxmsg > HARD_MSGMAX ||
702 attr->mq_msgsize > HARD_MSGSIZEMAX)
Doug Ledford113289c2012-05-31 16:26:36 -0700703 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 } else {
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700705 if (attr->mq_maxmsg > ipc_ns->mq_msg_max ||
706 attr->mq_msgsize > ipc_ns->mq_msgsize_max)
Doug Ledford113289c2012-05-31 16:26:36 -0700707 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 }
709 /* check for overflow */
710 if (attr->mq_msgsize > ULONG_MAX/attr->mq_maxmsg)
Doug Ledford113289c2012-05-31 16:26:36 -0700711 return -EOVERFLOW;
Doug Ledford2c12ea42012-05-31 16:26:36 -0700712 mq_treesize = attr->mq_maxmsg * sizeof(struct msg_msg) +
713 min_t(unsigned int, attr->mq_maxmsg, MQ_PRIO_MAX) *
714 sizeof(struct posix_msg_tree_node);
715 total_size = attr->mq_maxmsg * attr->mq_msgsize;
716 if (total_size + mq_treesize < total_size)
Doug Ledford113289c2012-05-31 16:26:36 -0700717 return -EOVERFLOW;
718 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719}
720
721/*
722 * Invoked when creating a new queue via sys_mq_open
723 */
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700724static struct file *do_create(struct ipc_namespace *ipc_ns, struct dentry *dir,
Al Viro4acdaf22011-07-26 01:42:34 -0400725 struct dentry *dentry, int oflag, umode_t mode,
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700726 struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727{
David Howells745ca242008-11-14 10:39:22 +1100728 const struct cred *cred = current_cred();
Dave Hansen4a3fd212008-02-15 14:37:48 -0800729 struct file *result;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 int ret;
731
Al Viro564f6992008-12-14 04:02:26 -0500732 if (attr) {
Doug Ledford113289c2012-05-31 16:26:36 -0700733 ret = mq_attr_ok(ipc_ns, attr);
734 if (ret)
Alexander Viro7c7dce92006-01-14 15:29:55 -0500735 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 /* store for use during create */
Al Viro564f6992008-12-14 04:02:26 -0500737 dentry->d_fsdata = attr;
Doug Ledford113289c2012-05-31 16:26:36 -0700738 } else {
739 struct mq_attr def_attr;
740
741 def_attr.mq_maxmsg = min(ipc_ns->mq_msg_max,
742 ipc_ns->mq_msg_default);
743 def_attr.mq_msgsize = min(ipc_ns->mq_msgsize_max,
744 ipc_ns->mq_msgsize_default);
745 ret = mq_attr_ok(ipc_ns, &def_attr);
746 if (ret)
747 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 }
749
Al Viroce3b0f82009-03-29 19:08:22 -0400750 mode &= ~current_umask();
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700751 ret = mnt_want_write(ipc_ns->mq_mnt);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800752 if (ret)
753 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 ret = vfs_create(dir->d_inode, dentry, mode, NULL);
755 dentry->d_fsdata = NULL;
756 if (ret)
Dave Hansen4a3fd212008-02-15 14:37:48 -0800757 goto out_drop_write;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700759 result = dentry_open(dentry, ipc_ns->mq_mnt, oflag, cred);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800760 /*
761 * dentry_open() took a persistent mnt_want_write(),
762 * so we can now drop this one.
763 */
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700764 mnt_drop_write(ipc_ns->mq_mnt);
Dave Hansen4a3fd212008-02-15 14:37:48 -0800765 return result;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766
Dave Hansen4a3fd212008-02-15 14:37:48 -0800767out_drop_write:
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700768 mnt_drop_write(ipc_ns->mq_mnt);
Alexander Viro7c7dce92006-01-14 15:29:55 -0500769out:
770 dput(dentry);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700771 mntput(ipc_ns->mq_mnt);
Alexander Viro7c7dce92006-01-14 15:29:55 -0500772 return ERR_PTR(ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773}
774
775/* Opens existing queue */
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700776static struct file *do_open(struct ipc_namespace *ipc_ns,
777 struct dentry *dentry, int oflag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778{
André Goddard Rosa04db0dd2010-02-23 04:04:25 -0300779 int ret;
David Howells745ca242008-11-14 10:39:22 +1100780 const struct cred *cred = current_cred();
781
782 static const int oflag2acc[O_ACCMODE] = { MAY_READ, MAY_WRITE,
783 MAY_READ | MAY_WRITE };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784
Alexander Viro7c7dce92006-01-14 15:29:55 -0500785 if ((oflag & O_ACCMODE) == (O_RDWR | O_WRONLY)) {
André Goddard Rosa04db0dd2010-02-23 04:04:25 -0300786 ret = -EINVAL;
787 goto err;
Alexander Viro7c7dce92006-01-14 15:29:55 -0500788 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789
Al Virof419a2e2008-07-22 00:07:17 -0400790 if (inode_permission(dentry->d_inode, oflag2acc[oflag & O_ACCMODE])) {
André Goddard Rosa04db0dd2010-02-23 04:04:25 -0300791 ret = -EACCES;
792 goto err;
Alexander Viro7c7dce92006-01-14 15:29:55 -0500793 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700795 return dentry_open(dentry, ipc_ns->mq_mnt, oflag, cred);
André Goddard Rosa04db0dd2010-02-23 04:04:25 -0300796
797err:
798 dput(dentry);
799 mntput(ipc_ns->mq_mnt);
800 return ERR_PTR(ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801}
802
Al Virodf0a4282011-07-26 05:26:10 -0400803SYSCALL_DEFINE4(mq_open, const char __user *, u_name, int, oflag, umode_t, mode,
Heiko Carstensd5460c92009-01-14 14:14:27 +0100804 struct mq_attr __user *, u_attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805{
806 struct dentry *dentry;
807 struct file *filp;
808 char *name;
Al Viro564f6992008-12-14 04:02:26 -0500809 struct mq_attr attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 int fd, error;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700811 struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812
Al Viro564f6992008-12-14 04:02:26 -0500813 if (u_attr && copy_from_user(&attr, u_attr, sizeof(struct mq_attr)))
814 return -EFAULT;
815
816 audit_mq_open(oflag, mode, u_attr ? &attr : NULL);
George C. Wilson20ca73b2006-05-24 16:09:55 -0500817
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 if (IS_ERR(name = getname(u_name)))
819 return PTR_ERR(name);
820
Ulrich Drepper269f2132008-05-03 15:28:45 -0400821 fd = get_unused_fd_flags(O_CLOEXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 if (fd < 0)
823 goto out_putname;
824
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700825 mutex_lock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex);
826 dentry = lookup_one_len(name, ipc_ns->mq_mnt->mnt_root, strlen(name));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 if (IS_ERR(dentry)) {
828 error = PTR_ERR(dentry);
André Goddard Rosa4294a8e2010-02-23 04:04:28 -0300829 goto out_putfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 }
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700831 mntget(ipc_ns->mq_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832
833 if (oflag & O_CREAT) {
834 if (dentry->d_inode) { /* entry already exists */
Al Viro5a190ae2007-06-07 12:19:32 -0400835 audit_inode(name, dentry);
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300836 if (oflag & O_EXCL) {
837 error = -EEXIST;
Alexander Viro7c7dce92006-01-14 15:29:55 -0500838 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300839 }
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700840 filp = do_open(ipc_ns, dentry, oflag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 } else {
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700842 filp = do_create(ipc_ns, ipc_ns->mq_mnt->mnt_root,
843 dentry, oflag, mode,
Al Viro564f6992008-12-14 04:02:26 -0500844 u_attr ? &attr : NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 }
Alexander Viro7c7dce92006-01-14 15:29:55 -0500846 } else {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300847 if (!dentry->d_inode) {
848 error = -ENOENT;
Alexander Viro7c7dce92006-01-14 15:29:55 -0500849 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300850 }
Al Viro5a190ae2007-06-07 12:19:32 -0400851 audit_inode(name, dentry);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700852 filp = do_open(ipc_ns, dentry, oflag);
Alexander Viro7c7dce92006-01-14 15:29:55 -0500853 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854
855 if (IS_ERR(filp)) {
856 error = PTR_ERR(filp);
857 goto out_putfd;
858 }
859
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 fd_install(fd, filp);
861 goto out_upsem;
862
Alexander Viro7c7dce92006-01-14 15:29:55 -0500863out:
864 dput(dentry);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700865 mntput(ipc_ns->mq_mnt);
Alexander Viro7c7dce92006-01-14 15:29:55 -0500866out_putfd:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867 put_unused_fd(fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 fd = error;
869out_upsem:
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700870 mutex_unlock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871out_putname:
872 putname(name);
873 return fd;
874}
875
Heiko Carstensd5460c92009-01-14 14:14:27 +0100876SYSCALL_DEFINE1(mq_unlink, const char __user *, u_name)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877{
878 int err;
879 char *name;
880 struct dentry *dentry;
881 struct inode *inode = NULL;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700882 struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883
884 name = getname(u_name);
885 if (IS_ERR(name))
886 return PTR_ERR(name);
887
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700888 mutex_lock_nested(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex,
Peter Zijlstra7a434812007-03-06 01:42:09 -0800889 I_MUTEX_PARENT);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700890 dentry = lookup_one_len(name, ipc_ns->mq_mnt->mnt_root, strlen(name));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 if (IS_ERR(dentry)) {
892 err = PTR_ERR(dentry);
893 goto out_unlock;
894 }
895
896 if (!dentry->d_inode) {
897 err = -ENOENT;
898 goto out_err;
899 }
900
901 inode = dentry->d_inode;
902 if (inode)
Al Viro7de9c6ee2010-10-23 11:11:40 -0400903 ihold(inode);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700904 err = mnt_want_write(ipc_ns->mq_mnt);
Dave Hansen06227532008-02-15 14:37:34 -0800905 if (err)
906 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 err = vfs_unlink(dentry->d_parent->d_inode, dentry);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700908 mnt_drop_write(ipc_ns->mq_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909out_err:
910 dput(dentry);
911
912out_unlock:
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700913 mutex_unlock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 putname(name);
915 if (inode)
916 iput(inode);
917
918 return err;
919}
920
921/* Pipelined send and receive functions.
922 *
923 * If a receiver finds no waiting message, then it registers itself in the
924 * list of waiting receivers. A sender checks that list before adding the new
925 * message into the message array. If there is a waiting receiver, then it
926 * bypasses the message array and directly hands the message over to the
927 * receiver.
928 * The receiver accepts the message and returns without grabbing the queue
929 * spinlock. Therefore an intermediate STATE_PENDING state and memory barriers
930 * are necessary. The same algorithm is used for sysv semaphores, see
Serge E. Hallyn7b7a3172006-03-28 01:56:23 -0800931 * ipc/sem.c for more details.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 *
933 * The same algorithm is used for senders.
934 */
935
936/* pipelined_send() - send a message directly to the task waiting in
937 * sys_mq_timedreceive() (without inserting message into a queue).
938 */
939static inline void pipelined_send(struct mqueue_inode_info *info,
940 struct msg_msg *message,
941 struct ext_wait_queue *receiver)
942{
943 receiver->msg = message;
944 list_del(&receiver->list);
945 receiver->state = STATE_PENDING;
946 wake_up_process(receiver->task);
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700947 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 receiver->state = STATE_READY;
949}
950
951/* pipelined_receive() - if there is task waiting in sys_mq_timedsend()
952 * gets its message and put to the queue (we have one free place for sure). */
953static inline void pipelined_receive(struct mqueue_inode_info *info)
954{
955 struct ext_wait_queue *sender = wq_get_first_waiter(info, SEND);
956
957 if (!sender) {
958 /* for poll */
959 wake_up_interruptible(&info->wait_q);
960 return;
961 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700962 if (msg_insert(sender->msg, info))
963 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 list_del(&sender->list);
965 sender->state = STATE_PENDING;
966 wake_up_process(sender->task);
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700967 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 sender->state = STATE_READY;
969}
970
Heiko Carstensc4ea37c2009-01-14 14:14:28 +0100971SYSCALL_DEFINE5(mq_timedsend, mqd_t, mqdes, const char __user *, u_msg_ptr,
972 size_t, msg_len, unsigned int, msg_prio,
973 const struct timespec __user *, u_abs_timeout)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974{
975 struct file *filp;
976 struct inode *inode;
977 struct ext_wait_queue wait;
978 struct ext_wait_queue *receiver;
979 struct msg_msg *msg_ptr;
980 struct mqueue_inode_info *info;
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200981 ktime_t expires, *timeout = NULL;
982 struct timespec ts;
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700983 struct posix_msg_tree_node *new_leaf = NULL;
984 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985
Al Viroc32c8af2008-12-14 03:46:48 -0500986 if (u_abs_timeout) {
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200987 int res = prepare_timeout(u_abs_timeout, &expires, &ts);
988 if (res)
989 return res;
990 timeout = &expires;
Al Viroc32c8af2008-12-14 03:46:48 -0500991 }
George C. Wilson20ca73b2006-05-24 16:09:55 -0500992
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 if (unlikely(msg_prio >= (unsigned long) MQ_PRIO_MAX))
994 return -EINVAL;
995
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200996 audit_mq_sendrecv(mqdes, msg_len, msg_prio, timeout ? &ts : NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 filp = fget(mqdes);
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300999 if (unlikely(!filp)) {
1000 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001002 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003
Josef Sipek6d630792006-12-08 02:37:11 -08001004 inode = filp->f_path.dentry->d_inode;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001005 if (unlikely(filp->f_op != &mqueue_file_operations)) {
1006 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001008 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 info = MQUEUE_I(inode);
Al Viro5a190ae2007-06-07 12:19:32 -04001010 audit_inode(NULL, filp->f_path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001012 if (unlikely(!(filp->f_mode & FMODE_WRITE))) {
1013 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001015 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016
1017 if (unlikely(msg_len > info->attr.mq_msgsize)) {
1018 ret = -EMSGSIZE;
1019 goto out_fput;
1020 }
1021
1022 /* First try to allocate memory, before doing anything with
1023 * existing queues. */
1024 msg_ptr = load_msg(u_msg_ptr, msg_len);
1025 if (IS_ERR(msg_ptr)) {
1026 ret = PTR_ERR(msg_ptr);
1027 goto out_fput;
1028 }
1029 msg_ptr->m_ts = msg_len;
1030 msg_ptr->m_type = msg_prio;
1031
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001032 /*
1033 * msg_insert really wants us to have a valid, spare node struct so
1034 * it doesn't have to kmalloc a GFP_ATOMIC allocation, but it will
1035 * fall back to that if necessary.
1036 */
1037 if (!info->node_cache)
1038 new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL);
1039
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 spin_lock(&info->lock);
1041
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001042 if (!info->node_cache && new_leaf) {
1043 /* Save our speculative allocation into the cache */
1044 rb_init_node(&new_leaf->rb_node);
1045 INIT_LIST_HEAD(&new_leaf->msg_list);
1046 info->node_cache = new_leaf;
1047 info->qsize += sizeof(*new_leaf);
1048 new_leaf = NULL;
1049 } else {
1050 kfree(new_leaf);
1051 }
1052
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 if (info->attr.mq_curmsgs == info->attr.mq_maxmsg) {
1054 if (filp->f_flags & O_NONBLOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 ret = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 } else {
1057 wait.task = current;
1058 wait.msg = (void *) msg_ptr;
1059 wait.state = STATE_NONE;
1060 ret = wq_sleep(info, SEND, timeout, &wait);
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001061 /*
1062 * wq_sleep must be called with info->lock held, and
1063 * returns with the lock released
1064 */
1065 goto out_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 } else {
1068 receiver = wq_get_first_waiter(info, RECV);
1069 if (receiver) {
1070 pipelined_send(info, msg_ptr, receiver);
1071 } else {
1072 /* adds message to the queue */
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001073 ret = msg_insert(msg_ptr, info);
1074 if (ret)
1075 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 __do_notify(info);
1077 }
1078 inode->i_atime = inode->i_mtime = inode->i_ctime =
1079 CURRENT_TIME;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 }
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001081out_unlock:
1082 spin_unlock(&info->lock);
1083out_free:
1084 if (ret)
1085 free_msg(msg_ptr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086out_fput:
1087 fput(filp);
1088out:
1089 return ret;
1090}
1091
Heiko Carstensc4ea37c2009-01-14 14:14:28 +01001092SYSCALL_DEFINE5(mq_timedreceive, mqd_t, mqdes, char __user *, u_msg_ptr,
1093 size_t, msg_len, unsigned int __user *, u_msg_prio,
1094 const struct timespec __user *, u_abs_timeout)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 ssize_t ret;
1097 struct msg_msg *msg_ptr;
1098 struct file *filp;
1099 struct inode *inode;
1100 struct mqueue_inode_info *info;
1101 struct ext_wait_queue wait;
Carsten Emde9ca7d8e2010-04-02 22:40:20 +02001102 ktime_t expires, *timeout = NULL;
1103 struct timespec ts;
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001104 struct posix_msg_tree_node *new_leaf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105
Al Viroc32c8af2008-12-14 03:46:48 -05001106 if (u_abs_timeout) {
Carsten Emde9ca7d8e2010-04-02 22:40:20 +02001107 int res = prepare_timeout(u_abs_timeout, &expires, &ts);
1108 if (res)
1109 return res;
1110 timeout = &expires;
Al Viroc32c8af2008-12-14 03:46:48 -05001111 }
George C. Wilson20ca73b2006-05-24 16:09:55 -05001112
Carsten Emde9ca7d8e2010-04-02 22:40:20 +02001113 audit_mq_sendrecv(mqdes, msg_len, 0, timeout ? &ts : NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 filp = fget(mqdes);
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001116 if (unlikely(!filp)) {
1117 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001119 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120
Josef Sipek6d630792006-12-08 02:37:11 -08001121 inode = filp->f_path.dentry->d_inode;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001122 if (unlikely(filp->f_op != &mqueue_file_operations)) {
1123 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001125 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 info = MQUEUE_I(inode);
Al Viro5a190ae2007-06-07 12:19:32 -04001127 audit_inode(NULL, filp->f_path.dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001129 if (unlikely(!(filp->f_mode & FMODE_READ))) {
1130 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001132 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133
1134 /* checks if buffer is big enough */
1135 if (unlikely(msg_len < info->attr.mq_msgsize)) {
1136 ret = -EMSGSIZE;
1137 goto out_fput;
1138 }
1139
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001140 /*
1141 * msg_insert really wants us to have a valid, spare node struct so
1142 * it doesn't have to kmalloc a GFP_ATOMIC allocation, but it will
1143 * fall back to that if necessary.
1144 */
1145 if (!info->node_cache)
1146 new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL);
1147
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 spin_lock(&info->lock);
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001149
1150 if (!info->node_cache && new_leaf) {
1151 /* Save our speculative allocation into the cache */
1152 rb_init_node(&new_leaf->rb_node);
1153 INIT_LIST_HEAD(&new_leaf->msg_list);
1154 info->node_cache = new_leaf;
1155 info->qsize += sizeof(*new_leaf);
1156 } else {
1157 kfree(new_leaf);
1158 }
1159
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160 if (info->attr.mq_curmsgs == 0) {
1161 if (filp->f_flags & O_NONBLOCK) {
1162 spin_unlock(&info->lock);
1163 ret = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 } else {
1165 wait.task = current;
1166 wait.state = STATE_NONE;
1167 ret = wq_sleep(info, RECV, timeout, &wait);
1168 msg_ptr = wait.msg;
1169 }
1170 } else {
1171 msg_ptr = msg_get(info);
1172
1173 inode->i_atime = inode->i_mtime = inode->i_ctime =
1174 CURRENT_TIME;
1175
1176 /* There is now free space in queue. */
1177 pipelined_receive(info);
1178 spin_unlock(&info->lock);
1179 ret = 0;
1180 }
1181 if (ret == 0) {
1182 ret = msg_ptr->m_ts;
1183
1184 if ((u_msg_prio && put_user(msg_ptr->m_type, u_msg_prio)) ||
1185 store_msg(u_msg_ptr, msg_ptr, msg_ptr->m_ts)) {
1186 ret = -EFAULT;
1187 }
1188 free_msg(msg_ptr);
1189 }
1190out_fput:
1191 fput(filp);
1192out:
1193 return ret;
1194}
1195
1196/*
1197 * Notes: the case when user wants us to deregister (with NULL as pointer)
1198 * and he isn't currently owner of notification, will be silently discarded.
1199 * It isn't explicitly defined in the POSIX.
1200 */
Heiko Carstensc4ea37c2009-01-14 14:14:28 +01001201SYSCALL_DEFINE2(mq_notify, mqd_t, mqdes,
1202 const struct sigevent __user *, u_notification)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203{
1204 int ret;
1205 struct file *filp;
1206 struct sock *sock;
1207 struct inode *inode;
1208 struct sigevent notification;
1209 struct mqueue_inode_info *info;
1210 struct sk_buff *nc;
1211
Al Viro20114f72008-12-10 07:16:12 -05001212 if (u_notification) {
1213 if (copy_from_user(&notification, u_notification,
1214 sizeof(struct sigevent)))
1215 return -EFAULT;
1216 }
1217
1218 audit_mq_notify(mqdes, u_notification ? &notification : NULL);
George C. Wilson20ca73b2006-05-24 16:09:55 -05001219
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 nc = NULL;
1221 sock = NULL;
1222 if (u_notification != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223 if (unlikely(notification.sigev_notify != SIGEV_NONE &&
1224 notification.sigev_notify != SIGEV_SIGNAL &&
1225 notification.sigev_notify != SIGEV_THREAD))
1226 return -EINVAL;
1227 if (notification.sigev_notify == SIGEV_SIGNAL &&
Jesper Juhl7ed20e12005-05-01 08:59:14 -07001228 !valid_signal(notification.sigev_signo)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 return -EINVAL;
1230 }
1231 if (notification.sigev_notify == SIGEV_THREAD) {
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001232 long timeo;
1233
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234 /* create the notify skb */
1235 nc = alloc_skb(NOTIFY_COOKIE_LEN, GFP_KERNEL);
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001236 if (!nc) {
1237 ret = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001239 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240 if (copy_from_user(nc->data,
1241 notification.sigev_value.sival_ptr,
1242 NOTIFY_COOKIE_LEN)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001243 ret = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244 goto out;
1245 }
1246
1247 /* TODO: add a header? */
1248 skb_put(nc, NOTIFY_COOKIE_LEN);
1249 /* and attach it to the socket */
1250retry:
1251 filp = fget(notification.sigev_signo);
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001252 if (!filp) {
1253 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001255 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 sock = netlink_getsockbyfilp(filp);
1257 fput(filp);
1258 if (IS_ERR(sock)) {
1259 ret = PTR_ERR(sock);
1260 sock = NULL;
1261 goto out;
1262 }
1263
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001264 timeo = MAX_SCHEDULE_TIMEOUT;
Denis V. Lunev9457afe2008-06-05 11:23:39 -07001265 ret = netlink_attachskb(sock, nc, &timeo, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 if (ret == 1)
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001267 goto retry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 if (ret) {
1269 sock = NULL;
1270 nc = NULL;
1271 goto out;
1272 }
1273 }
1274 }
1275
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 filp = fget(mqdes);
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001277 if (!filp) {
1278 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001280 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281
Josef Sipek6d630792006-12-08 02:37:11 -08001282 inode = filp->f_path.dentry->d_inode;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001283 if (unlikely(filp->f_op != &mqueue_file_operations)) {
1284 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001286 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 info = MQUEUE_I(inode);
1288
1289 ret = 0;
1290 spin_lock(&info->lock);
1291 if (u_notification == NULL) {
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001292 if (info->notify_owner == task_tgid(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293 remove_notification(info);
1294 inode->i_atime = inode->i_ctime = CURRENT_TIME;
1295 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001296 } else if (info->notify_owner != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297 ret = -EBUSY;
1298 } else {
1299 switch (notification.sigev_notify) {
1300 case SIGEV_NONE:
1301 info->notify.sigev_notify = SIGEV_NONE;
1302 break;
1303 case SIGEV_THREAD:
1304 info->notify_sock = sock;
1305 info->notify_cookie = nc;
1306 sock = NULL;
1307 nc = NULL;
1308 info->notify.sigev_notify = SIGEV_THREAD;
1309 break;
1310 case SIGEV_SIGNAL:
1311 info->notify.sigev_signo = notification.sigev_signo;
1312 info->notify.sigev_value = notification.sigev_value;
1313 info->notify.sigev_notify = SIGEV_SIGNAL;
1314 break;
1315 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001316
1317 info->notify_owner = get_pid(task_tgid(current));
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -08001318 info->notify_user_ns = get_user_ns(current_user_ns());
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 inode->i_atime = inode->i_ctime = CURRENT_TIME;
1320 }
1321 spin_unlock(&info->lock);
1322out_fput:
1323 fput(filp);
1324out:
1325 if (sock) {
1326 netlink_detachskb(sock, nc);
1327 } else if (nc) {
1328 dev_kfree_skb(nc);
1329 }
1330 return ret;
1331}
1332
Heiko Carstensc4ea37c2009-01-14 14:14:28 +01001333SYSCALL_DEFINE3(mq_getsetattr, mqd_t, mqdes,
1334 const struct mq_attr __user *, u_mqstat,
1335 struct mq_attr __user *, u_omqstat)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336{
1337 int ret;
1338 struct mq_attr mqstat, omqstat;
1339 struct file *filp;
1340 struct inode *inode;
1341 struct mqueue_inode_info *info;
1342
1343 if (u_mqstat != NULL) {
1344 if (copy_from_user(&mqstat, u_mqstat, sizeof(struct mq_attr)))
1345 return -EFAULT;
1346 if (mqstat.mq_flags & (~O_NONBLOCK))
1347 return -EINVAL;
1348 }
1349
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 filp = fget(mqdes);
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001351 if (!filp) {
1352 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001354 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355
Josef Sipek6d630792006-12-08 02:37:11 -08001356 inode = filp->f_path.dentry->d_inode;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001357 if (unlikely(filp->f_op != &mqueue_file_operations)) {
1358 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001360 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361 info = MQUEUE_I(inode);
1362
1363 spin_lock(&info->lock);
1364
1365 omqstat = info->attr;
1366 omqstat.mq_flags = filp->f_flags & O_NONBLOCK;
1367 if (u_mqstat) {
Al Viro73929062008-12-10 06:58:59 -05001368 audit_mq_getsetattr(mqdes, &mqstat);
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -07001369 spin_lock(&filp->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 if (mqstat.mq_flags & O_NONBLOCK)
1371 filp->f_flags |= O_NONBLOCK;
1372 else
1373 filp->f_flags &= ~O_NONBLOCK;
Jonathan Corbetdb1dd4d2009-02-06 15:25:24 -07001374 spin_unlock(&filp->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375
1376 inode->i_atime = inode->i_ctime = CURRENT_TIME;
1377 }
1378
1379 spin_unlock(&info->lock);
1380
1381 ret = 0;
1382 if (u_omqstat != NULL && copy_to_user(u_omqstat, &omqstat,
1383 sizeof(struct mq_attr)))
1384 ret = -EFAULT;
1385
1386out_fput:
1387 fput(filp);
1388out:
1389 return ret;
1390}
1391
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -08001392static const struct inode_operations mqueue_dir_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393 .lookup = simple_lookup,
1394 .create = mqueue_create,
1395 .unlink = mqueue_unlink,
1396};
1397
Arjan van de Ven9a321442007-02-12 00:55:35 -08001398static const struct file_operations mqueue_file_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 .flush = mqueue_flush_file,
1400 .poll = mqueue_poll_file,
1401 .read = mqueue_read_file,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001402 .llseek = default_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403};
1404
Alexey Dobriyanb87221d2009-09-21 17:01:09 -07001405static const struct super_operations mqueue_super_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406 .alloc_inode = mqueue_alloc_inode,
1407 .destroy_inode = mqueue_destroy_inode,
Al Viro6d8af642010-06-05 16:29:45 -04001408 .evict_inode = mqueue_evict_inode,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410};
1411
1412static struct file_system_type mqueue_fs_type = {
1413 .name = "mqueue",
Al Viroceefda62010-07-26 13:16:50 +04001414 .mount = mqueue_mount,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415 .kill_sb = kill_litter_super,
1416};
1417
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001418int mq_init_ns(struct ipc_namespace *ns)
1419{
1420 ns->mq_queues_count = 0;
1421 ns->mq_queues_max = DFLT_QUEUESMAX;
1422 ns->mq_msg_max = DFLT_MSGMAX;
1423 ns->mq_msgsize_max = DFLT_MSGSIZEMAX;
KOSAKI Motohirocef01842012-05-31 16:26:33 -07001424 ns->mq_msg_default = DFLT_MSG;
1425 ns->mq_msgsize_default = DFLT_MSGSIZE;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001426
1427 ns->mq_mnt = kern_mount_data(&mqueue_fs_type, ns);
1428 if (IS_ERR(ns->mq_mnt)) {
1429 int err = PTR_ERR(ns->mq_mnt);
1430 ns->mq_mnt = NULL;
1431 return err;
1432 }
1433 return 0;
1434}
1435
1436void mq_clear_sbinfo(struct ipc_namespace *ns)
1437{
1438 ns->mq_mnt->mnt_sb->s_fs_info = NULL;
1439}
1440
1441void mq_put_mnt(struct ipc_namespace *ns)
1442{
Al Viro6f686572011-12-09 00:38:50 -05001443 kern_unmount(ns->mq_mnt);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001444}
1445
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446static int __init init_mqueue_fs(void)
1447{
1448 int error;
1449
1450 mqueue_inode_cachep = kmem_cache_create("mqueue_inode_cache",
1451 sizeof(struct mqueue_inode_info), 0,
Paul Mundt20c2df82007-07-20 10:11:58 +09001452 SLAB_HWCACHE_ALIGN, init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 if (mqueue_inode_cachep == NULL)
1454 return -ENOMEM;
1455
André Goddard Rosa2329e3922010-02-23 04:04:27 -03001456 /* ignore failures - they are not fatal */
Serge E. Hallynbdc8e5f2009-04-06 19:01:11 -07001457 mq_sysctl_table = mq_register_sysctl_table();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458
1459 error = register_filesystem(&mqueue_fs_type);
1460 if (error)
1461 goto out_sysctl;
1462
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001463 spin_lock_init(&mq_lock);
1464
Al Viro6f686572011-12-09 00:38:50 -05001465 error = mq_init_ns(&init_ipc_ns);
1466 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467 goto out_filesystem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 return 0;
1470
1471out_filesystem:
1472 unregister_filesystem(&mqueue_fs_type);
1473out_sysctl:
1474 if (mq_sysctl_table)
1475 unregister_sysctl_table(mq_sysctl_table);
Alexey Dobriyan1a1d92c2006-09-27 01:49:40 -07001476 kmem_cache_destroy(mqueue_inode_cachep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477 return error;
1478}
1479
1480__initcall(init_mqueue_fs);