blob: 3aaea7ffd077c874cb1b420013458020057f45a6 [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:
Manfred Spraul239521f2014-01-27 17:07:04 -08009 * Manfred Spraul (manfred@colorfullife.com)
Linus Torvalds1da177e2005-04-16 15:20:36 -070010 *
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;
Manfred Spraul239521f2014-01-27 17:07:04 -080076 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
Manfred Spraul239521f2014-01-27 17:07:04 -080095static struct ctl_table_header *mq_sysctl_table;
Linus Torvalds1da177e2005-04-16 15:20:36 -070096
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;
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700145 INIT_LIST_HEAD(&leaf->msg_list);
146 info->qsize += sizeof(*leaf);
147 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700148 leaf->priority = msg->m_type;
149 rb_link_node(&leaf->rb_node, parent, p);
150 rb_insert_color(&leaf->rb_node, &info->msg_tree);
Doug Ledfordd6629852012-05-31 16:26:35 -0700151insert_msg:
152 info->attr.mq_curmsgs++;
153 info->qsize += msg->m_ts;
154 list_add_tail(&msg->m_list, &leaf->msg_list);
155 return 0;
156}
157
158static inline struct msg_msg *msg_get(struct mqueue_inode_info *info)
159{
160 struct rb_node **p, *parent = NULL;
161 struct posix_msg_tree_node *leaf;
162 struct msg_msg *msg;
163
164try_again:
165 p = &info->msg_tree.rb_node;
166 while (*p) {
167 parent = *p;
168 /*
169 * During insert, low priorities go to the left and high to the
170 * right. On receive, we want the highest priorities first, so
171 * walk all the way to the right.
172 */
173 p = &(*p)->rb_right;
174 }
175 if (!parent) {
176 if (info->attr.mq_curmsgs) {
177 pr_warn_once("Inconsistency in POSIX message queue, "
178 "no tree element, but supposedly messages "
179 "should exist!\n");
180 info->attr.mq_curmsgs = 0;
181 }
182 return NULL;
183 }
184 leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700185 if (unlikely(list_empty(&leaf->msg_list))) {
Doug Ledfordd6629852012-05-31 16:26:35 -0700186 pr_warn_once("Inconsistency in POSIX message queue, "
187 "empty leaf node but we haven't implemented "
188 "lazy leaf delete!\n");
189 rb_erase(&leaf->rb_node, &info->msg_tree);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700190 if (info->node_cache) {
191 info->qsize -= sizeof(*leaf);
192 kfree(leaf);
193 } else {
194 info->node_cache = leaf;
195 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700196 goto try_again;
197 } else {
198 msg = list_first_entry(&leaf->msg_list,
199 struct msg_msg, m_list);
200 list_del(&msg->m_list);
201 if (list_empty(&leaf->msg_list)) {
202 rb_erase(&leaf->rb_node, &info->msg_tree);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700203 if (info->node_cache) {
204 info->qsize -= sizeof(*leaf);
205 kfree(leaf);
206 } else {
207 info->node_cache = leaf;
208 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700209 }
210 }
211 info->attr.mq_curmsgs--;
212 info->qsize -= msg->m_ts;
213 return msg;
214}
215
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700216static struct inode *mqueue_get_inode(struct super_block *sb,
Al Viro1b9d5ff72011-07-24 14:18:20 -0400217 struct ipc_namespace *ipc_ns, umode_t mode,
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700218 struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219{
David Howells86a264a2008-11-14 10:39:18 +1100220 struct user_struct *u = current_user();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 struct inode *inode;
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700222 int ret = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223
224 inode = new_inode(sb);
Jiri Slaby04715202011-07-26 16:08:46 -0700225 if (!inode)
226 goto err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227
Jiri Slaby04715202011-07-26 16:08:46 -0700228 inode->i_ino = get_next_ino();
229 inode->i_mode = mode;
230 inode->i_uid = current_fsuid();
231 inode->i_gid = current_fsgid();
232 inode->i_mtime = inode->i_ctime = inode->i_atime = CURRENT_TIME;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233
Jiri Slaby04715202011-07-26 16:08:46 -0700234 if (S_ISREG(mode)) {
235 struct mqueue_inode_info *info;
Doug Ledfordd6629852012-05-31 16:26:35 -0700236 unsigned long mq_bytes, mq_treesize;
André Goddard Rosac8308b12010-02-23 04:04:23 -0300237
Jiri Slaby04715202011-07-26 16:08:46 -0700238 inode->i_fop = &mqueue_file_operations;
239 inode->i_size = FILENT_SIZE;
240 /* mqueue specific info */
241 info = MQUEUE_I(inode);
242 spin_lock_init(&info->lock);
243 init_waitqueue_head(&info->wait_q);
244 INIT_LIST_HEAD(&info->e_wait_q[0].list);
245 INIT_LIST_HEAD(&info->e_wait_q[1].list);
246 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800247 info->notify_user_ns = NULL;
Jiri Slaby04715202011-07-26 16:08:46 -0700248 info->qsize = 0;
249 info->user = NULL; /* set when all is ok */
Doug Ledfordd6629852012-05-31 16:26:35 -0700250 info->msg_tree = RB_ROOT;
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700251 info->node_cache = NULL;
Jiri Slaby04715202011-07-26 16:08:46 -0700252 memset(&info->attr, 0, sizeof(info->attr));
KOSAKI Motohirocef01842012-05-31 16:26:33 -0700253 info->attr.mq_maxmsg = min(ipc_ns->mq_msg_max,
254 ipc_ns->mq_msg_default);
255 info->attr.mq_msgsize = min(ipc_ns->mq_msgsize_max,
256 ipc_ns->mq_msgsize_default);
Jiri Slaby04715202011-07-26 16:08:46 -0700257 if (attr) {
258 info->attr.mq_maxmsg = attr->mq_maxmsg;
259 info->attr.mq_msgsize = attr->mq_msgsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700261 /*
262 * We used to allocate a static array of pointers and account
263 * the size of that array as well as one msg_msg struct per
264 * possible message into the queue size. That's no longer
265 * accurate as the queue is now an rbtree and will grow and
266 * shrink depending on usage patterns. We can, however, still
267 * account one msg_msg struct per message, but the nodes are
268 * allocated depending on priority usage, and most programs
269 * only use one, or a handful, of priorities. However, since
270 * this is pinned memory, we need to assume worst case, so
271 * that means the min(mq_maxmsg, max_priorities) * struct
272 * posix_msg_tree_node.
273 */
274 mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) +
275 min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) *
276 sizeof(struct posix_msg_tree_node);
Jiri Slaby04715202011-07-26 16:08:46 -0700277
Doug Ledfordd6629852012-05-31 16:26:35 -0700278 mq_bytes = mq_treesize + (info->attr.mq_maxmsg *
279 info->attr.mq_msgsize);
Jiri Slaby04715202011-07-26 16:08:46 -0700280
281 spin_lock(&mq_lock);
282 if (u->mq_bytes + mq_bytes < u->mq_bytes ||
Davidlohr Bueso2a4e64b2012-01-20 14:34:01 -0800283 u->mq_bytes + mq_bytes > rlimit(RLIMIT_MSGQUEUE)) {
Jiri Slaby04715202011-07-26 16:08:46 -0700284 spin_unlock(&mq_lock);
285 /* mqueue_evict_inode() releases info->messages */
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700286 ret = -EMFILE;
Jiri Slaby04715202011-07-26 16:08:46 -0700287 goto out_inode;
288 }
289 u->mq_bytes += mq_bytes;
290 spin_unlock(&mq_lock);
291
292 /* all is ok */
293 info->user = get_uid(u);
294 } else if (S_ISDIR(mode)) {
295 inc_nlink(inode);
296 /* Some things misbehave if size == 0 on a directory */
297 inode->i_size = 2 * DIRENT_SIZE;
298 inode->i_op = &mqueue_dir_inode_operations;
299 inode->i_fop = &simple_dir_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 }
Jiri Slaby04715202011-07-26 16:08:46 -0700301
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 return inode;
303out_inode:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 iput(inode);
Jiri Slaby04715202011-07-26 16:08:46 -0700305err:
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700306 return ERR_PTR(ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307}
308
309static int mqueue_fill_super(struct super_block *sb, void *data, int silent)
310{
311 struct inode *inode;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700312 struct ipc_namespace *ns = data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313
314 sb->s_blocksize = PAGE_CACHE_SIZE;
315 sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
316 sb->s_magic = MQUEUE_MAGIC;
317 sb->s_op = &mqueue_super_ops;
318
Al Viro48fde702012-01-08 22:15:13 -0500319 inode = mqueue_get_inode(sb, ns, S_IFDIR | S_ISVTX | S_IRWXUGO, NULL);
320 if (IS_ERR(inode))
321 return PTR_ERR(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322
Al Viro48fde702012-01-08 22:15:13 -0500323 sb->s_root = d_make_root(inode);
324 if (!sb->s_root)
325 return -ENOMEM;
326 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327}
328
Al Viroceefda62010-07-26 13:16:50 +0400329static struct dentry *mqueue_mount(struct file_system_type *fs_type,
David Howells454e2392006-06-23 02:02:57 -0700330 int flags, const char *dev_name,
Al Viroceefda62010-07-26 13:16:50 +0400331 void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332{
Eric W. Biedermana636b702013-03-21 18:13:15 -0700333 if (!(flags & MS_KERNMOUNT)) {
334 struct ipc_namespace *ns = current->nsproxy->ipc_ns;
335 /* Don't allow mounting unless the caller has CAP_SYS_ADMIN
336 * over the ipc namespace.
337 */
338 if (!ns_capable(ns->user_ns, CAP_SYS_ADMIN))
339 return ERR_PTR(-EPERM);
340
341 data = ns;
342 }
Al Viroceefda62010-07-26 13:16:50 +0400343 return mount_ns(fs_type, flags, data, mqueue_fill_super);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344}
345
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700346static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347{
348 struct mqueue_inode_info *p = (struct mqueue_inode_info *) foo;
349
Christoph Lametera35afb82007-05-16 22:10:57 -0700350 inode_init_once(&p->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351}
352
353static struct inode *mqueue_alloc_inode(struct super_block *sb)
354{
355 struct mqueue_inode_info *ei;
356
Christoph Lametere94b1762006-12-06 20:33:17 -0800357 ei = kmem_cache_alloc(mqueue_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 if (!ei)
359 return NULL;
360 return &ei->vfs_inode;
361}
362
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100363static void mqueue_i_callback(struct rcu_head *head)
364{
365 struct inode *inode = container_of(head, struct inode, i_rcu);
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100366 kmem_cache_free(mqueue_inode_cachep, MQUEUE_I(inode));
367}
368
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369static void mqueue_destroy_inode(struct inode *inode)
370{
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100371 call_rcu(&inode->i_rcu, mqueue_i_callback);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372}
373
Al Viro6d8af642010-06-05 16:29:45 -0400374static void mqueue_evict_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375{
376 struct mqueue_inode_info *info;
377 struct user_struct *user;
Doug Ledfordd6629852012-05-31 16:26:35 -0700378 unsigned long mq_bytes, mq_treesize;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700379 struct ipc_namespace *ipc_ns;
Doug Ledfordd6629852012-05-31 16:26:35 -0700380 struct msg_msg *msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381
Jan Karadbd57682012-05-03 14:48:02 +0200382 clear_inode(inode);
Al Viro6d8af642010-06-05 16:29:45 -0400383
384 if (S_ISDIR(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 return;
Al Viro6d8af642010-06-05 16:29:45 -0400386
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700387 ipc_ns = get_ns_from_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 info = MQUEUE_I(inode);
389 spin_lock(&info->lock);
Doug Ledfordd6629852012-05-31 16:26:35 -0700390 while ((msg = msg_get(info)) != NULL)
391 free_msg(msg);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700392 kfree(info->node_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 spin_unlock(&info->lock);
394
André Goddard Rosa8834cf72010-02-23 04:04:24 -0300395 /* Total amount of bytes accounted for the mqueue */
Doug Ledfordd6629852012-05-31 16:26:35 -0700396 mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) +
397 min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) *
398 sizeof(struct posix_msg_tree_node);
399
400 mq_bytes = mq_treesize + (info->attr.mq_maxmsg *
401 info->attr.mq_msgsize);
402
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 user = info->user;
404 if (user) {
405 spin_lock(&mq_lock);
406 user->mq_bytes -= mq_bytes;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700407 /*
408 * get_ns_from_inode() ensures that the
409 * (ipc_ns = sb->s_fs_info) is either a valid ipc_ns
410 * to which we now hold a reference, or it is NULL.
411 * We can't put it here under mq_lock, though.
412 */
413 if (ipc_ns)
414 ipc_ns->mq_queues_count--;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 spin_unlock(&mq_lock);
416 free_uid(user);
417 }
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700418 if (ipc_ns)
419 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420}
421
422static int mqueue_create(struct inode *dir, struct dentry *dentry,
Al Viroebfc3b42012-06-10 18:05:36 -0400423 umode_t mode, bool excl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424{
425 struct inode *inode;
426 struct mq_attr *attr = dentry->d_fsdata;
427 int error;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700428 struct ipc_namespace *ipc_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429
430 spin_lock(&mq_lock);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700431 ipc_ns = __get_ns_from_inode(dir);
432 if (!ipc_ns) {
433 error = -EACCES;
434 goto out_unlock;
435 }
Davidlohr Buesof3713fd2014-02-25 15:01:45 -0800436
437 if (ipc_ns->mq_queues_count >= ipc_ns->mq_queues_max &&
438 !capable(CAP_SYS_RESOURCE)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 error = -ENOSPC;
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700440 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 }
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700442 ipc_ns->mq_queues_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 spin_unlock(&mq_lock);
444
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700445 inode = mqueue_get_inode(dir->i_sb, ipc_ns, mode, attr);
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700446 if (IS_ERR(inode)) {
447 error = PTR_ERR(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 spin_lock(&mq_lock);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700449 ipc_ns->mq_queues_count--;
450 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 }
452
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700453 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 dir->i_size += DIRENT_SIZE;
455 dir->i_ctime = dir->i_mtime = dir->i_atime = CURRENT_TIME;
456
457 d_instantiate(dentry, inode);
458 dget(dentry);
459 return 0;
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700460out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 spin_unlock(&mq_lock);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700462 if (ipc_ns)
463 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 return error;
465}
466
467static int mqueue_unlink(struct inode *dir, struct dentry *dentry)
468{
David Howells75c3cfa2015-03-17 22:26:12 +0000469 struct inode *inode = d_inode(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470
471 dir->i_ctime = dir->i_mtime = dir->i_atime = CURRENT_TIME;
472 dir->i_size -= DIRENT_SIZE;
Manfred Spraul239521f2014-01-27 17:07:04 -0800473 drop_nlink(inode);
474 dput(dentry);
475 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476}
477
478/*
479* This is routine for system read from queue file.
480* To avoid mess with doing here some sort of mq_receive we allow
481* to read only queue size & notification info (the only values
482* that are interesting from user point of view and aren't accessible
483* through std routines)
484*/
485static ssize_t mqueue_read_file(struct file *filp, char __user *u_data,
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700486 size_t count, loff_t *off)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487{
Al Viro496ad9a2013-01-23 17:07:38 -0500488 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 char buffer[FILENT_SIZE];
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700490 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491
492 spin_lock(&info->lock);
493 snprintf(buffer, sizeof(buffer),
494 "QSIZE:%-10lu NOTIFY:%-5d SIGNO:%-5d NOTIFY_PID:%-6d\n",
495 info->qsize,
496 info->notify_owner ? info->notify.sigev_notify : 0,
497 (info->notify_owner &&
498 info->notify.sigev_notify == SIGEV_SIGNAL) ?
499 info->notify.sigev_signo : 0,
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800500 pid_vnr(info->notify_owner));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 spin_unlock(&info->lock);
502 buffer[sizeof(buffer)-1] = '\0';
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700504 ret = simple_read_from_buffer(u_data, count, off, buffer,
505 strlen(buffer));
506 if (ret <= 0)
507 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508
Al Viro496ad9a2013-01-23 17:07:38 -0500509 file_inode(filp)->i_atime = file_inode(filp)->i_ctime = CURRENT_TIME;
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700510 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511}
512
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -0700513static int mqueue_flush_file(struct file *filp, fl_owner_t id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514{
Al Viro496ad9a2013-01-23 17:07:38 -0500515 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516
517 spin_lock(&info->lock);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700518 if (task_tgid(current) == info->notify_owner)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 remove_notification(info);
520
521 spin_unlock(&info->lock);
522 return 0;
523}
524
525static unsigned int mqueue_poll_file(struct file *filp, struct poll_table_struct *poll_tab)
526{
Al Viro496ad9a2013-01-23 17:07:38 -0500527 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 int retval = 0;
529
530 poll_wait(filp, &info->wait_q, poll_tab);
531
532 spin_lock(&info->lock);
533 if (info->attr.mq_curmsgs)
534 retval = POLLIN | POLLRDNORM;
535
536 if (info->attr.mq_curmsgs < info->attr.mq_maxmsg)
537 retval |= POLLOUT | POLLWRNORM;
538 spin_unlock(&info->lock);
539
540 return retval;
541}
542
543/* Adds current to info->e_wait_q[sr] before element with smaller prio */
544static void wq_add(struct mqueue_inode_info *info, int sr,
545 struct ext_wait_queue *ewp)
546{
547 struct ext_wait_queue *walk;
548
549 ewp->task = current;
550
551 list_for_each_entry(walk, &info->e_wait_q[sr].list, list) {
552 if (walk->task->static_prio <= current->static_prio) {
553 list_add_tail(&ewp->list, &walk->list);
554 return;
555 }
556 }
557 list_add_tail(&ewp->list, &info->e_wait_q[sr].list);
558}
559
560/*
561 * Puts current task to sleep. Caller must hold queue lock. After return
562 * lock isn't held.
563 * sr: SEND or RECV
564 */
565static int wq_sleep(struct mqueue_inode_info *info, int sr,
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200566 ktime_t *timeout, struct ext_wait_queue *ewp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567{
568 int retval;
569 signed long time;
570
571 wq_add(info, sr, ewp);
572
573 for (;;) {
574 set_current_state(TASK_INTERRUPTIBLE);
575
576 spin_unlock(&info->lock);
Wanlong Gao32ea8452011-10-31 17:06:35 -0700577 time = schedule_hrtimeout_range_clock(timeout, 0,
578 HRTIMER_MODE_ABS, CLOCK_REALTIME);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579
580 while (ewp->state == STATE_PENDING)
581 cpu_relax();
582
583 if (ewp->state == STATE_READY) {
584 retval = 0;
585 goto out;
586 }
587 spin_lock(&info->lock);
588 if (ewp->state == STATE_READY) {
589 retval = 0;
590 goto out_unlock;
591 }
592 if (signal_pending(current)) {
593 retval = -ERESTARTSYS;
594 break;
595 }
596 if (time == 0) {
597 retval = -ETIMEDOUT;
598 break;
599 }
600 }
601 list_del(&ewp->list);
602out_unlock:
603 spin_unlock(&info->lock);
604out:
605 return retval;
606}
607
608/*
609 * Returns waiting task that should be serviced first or NULL if none exists
610 */
611static struct ext_wait_queue *wq_get_first_waiter(
612 struct mqueue_inode_info *info, int sr)
613{
614 struct list_head *ptr;
615
616 ptr = info->e_wait_q[sr].list.prev;
617 if (ptr == &info->e_wait_q[sr].list)
618 return NULL;
619 return list_entry(ptr, struct ext_wait_queue, list);
620}
621
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622
623static inline void set_cookie(struct sk_buff *skb, char code)
624{
Manfred Spraul239521f2014-01-27 17:07:04 -0800625 ((char *)skb->data)[NOTIFY_COOKIE_LEN-1] = code;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626}
627
628/*
629 * The next function is only to split too long sys_mq_timedsend
630 */
631static void __do_notify(struct mqueue_inode_info *info)
632{
633 /* notification
634 * invoked when there is registered process and there isn't process
635 * waiting synchronously for message AND state of queue changed from
636 * empty to not empty. Here we are sure that no one is waiting
637 * synchronously. */
638 if (info->notify_owner &&
639 info->attr.mq_curmsgs == 1) {
640 struct siginfo sig_i;
641 switch (info->notify.sigev_notify) {
642 case SIGEV_NONE:
643 break;
644 case SIGEV_SIGNAL:
645 /* sends signal */
646
647 sig_i.si_signo = info->notify.sigev_signo;
648 sig_i.si_errno = 0;
649 sig_i.si_code = SI_MESGQ;
650 sig_i.si_value = info->notify.sigev_value;
Serge E. Hallyn6b550f92012-01-10 15:11:37 -0800651 /* map current pid/uid into info->owner's namespaces */
652 rcu_read_lock();
Sukadev Bhattiprolua6684992009-01-07 18:08:50 -0800653 sig_i.si_pid = task_tgid_nr_ns(current,
654 ns_of_pid(info->notify_owner));
Eric W. Biederman76b6db02012-03-14 15:24:19 -0700655 sig_i.si_uid = from_kuid_munged(info->notify_user_ns, current_uid());
Serge E. Hallyn6b550f92012-01-10 15:11:37 -0800656 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700658 kill_pid_info(info->notify.sigev_signo,
659 &sig_i, info->notify_owner);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 break;
661 case SIGEV_THREAD:
662 set_cookie(info->notify_cookie, NOTIFY_WOKENUP);
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700663 netlink_sendskb(info->notify_sock, info->notify_cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 break;
665 }
666 /* after notification unregisters process */
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700667 put_pid(info->notify_owner);
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800668 put_user_ns(info->notify_user_ns);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700669 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800670 info->notify_user_ns = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 }
672 wake_up(&info->wait_q);
673}
674
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200675static int prepare_timeout(const struct timespec __user *u_abs_timeout,
676 ktime_t *expires, struct timespec *ts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677{
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200678 if (copy_from_user(ts, u_abs_timeout, sizeof(struct timespec)))
679 return -EFAULT;
680 if (!timespec_valid(ts))
681 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200683 *expires = timespec_to_ktime(*ts);
684 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685}
686
687static void remove_notification(struct mqueue_inode_info *info)
688{
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700689 if (info->notify_owner != NULL &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690 info->notify.sigev_notify == SIGEV_THREAD) {
691 set_cookie(info->notify_cookie, NOTIFY_REMOVED);
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700692 netlink_sendskb(info->notify_sock, info->notify_cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700694 put_pid(info->notify_owner);
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800695 put_user_ns(info->notify_user_ns);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700696 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800697 info->notify_user_ns = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698}
699
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700700static int mq_attr_ok(struct ipc_namespace *ipc_ns, struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701{
Doug Ledford2c12ea42012-05-31 16:26:36 -0700702 int mq_treesize;
703 unsigned long total_size;
704
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 if (attr->mq_maxmsg <= 0 || attr->mq_msgsize <= 0)
Doug Ledford113289c2012-05-31 16:26:36 -0700706 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707 if (capable(CAP_SYS_RESOURCE)) {
Doug Ledford02967ea2012-05-31 16:26:29 -0700708 if (attr->mq_maxmsg > HARD_MSGMAX ||
709 attr->mq_msgsize > HARD_MSGSIZEMAX)
Doug Ledford113289c2012-05-31 16:26:36 -0700710 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 } else {
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700712 if (attr->mq_maxmsg > ipc_ns->mq_msg_max ||
713 attr->mq_msgsize > ipc_ns->mq_msgsize_max)
Doug Ledford113289c2012-05-31 16:26:36 -0700714 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715 }
716 /* check for overflow */
717 if (attr->mq_msgsize > ULONG_MAX/attr->mq_maxmsg)
Doug Ledford113289c2012-05-31 16:26:36 -0700718 return -EOVERFLOW;
Doug Ledford2c12ea42012-05-31 16:26:36 -0700719 mq_treesize = attr->mq_maxmsg * sizeof(struct msg_msg) +
720 min_t(unsigned int, attr->mq_maxmsg, MQ_PRIO_MAX) *
721 sizeof(struct posix_msg_tree_node);
722 total_size = attr->mq_maxmsg * attr->mq_msgsize;
723 if (total_size + mq_treesize < total_size)
Doug Ledford113289c2012-05-31 16:26:36 -0700724 return -EOVERFLOW;
725 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726}
727
728/*
729 * Invoked when creating a new queue via sys_mq_open
730 */
Al Viro765927b2012-06-26 21:58:53 +0400731static struct file *do_create(struct ipc_namespace *ipc_ns, struct inode *dir,
732 struct path *path, int oflag, umode_t mode,
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700733 struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734{
David Howells745ca242008-11-14 10:39:22 +1100735 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 int ret;
737
Al Viro564f6992008-12-14 04:02:26 -0500738 if (attr) {
Doug Ledford113289c2012-05-31 16:26:36 -0700739 ret = mq_attr_ok(ipc_ns, attr);
740 if (ret)
Al Viro765927b2012-06-26 21:58:53 +0400741 return ERR_PTR(ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742 /* store for use during create */
Al Viro765927b2012-06-26 21:58:53 +0400743 path->dentry->d_fsdata = attr;
Doug Ledford113289c2012-05-31 16:26:36 -0700744 } else {
745 struct mq_attr def_attr;
746
747 def_attr.mq_maxmsg = min(ipc_ns->mq_msg_max,
748 ipc_ns->mq_msg_default);
749 def_attr.mq_msgsize = min(ipc_ns->mq_msgsize_max,
750 ipc_ns->mq_msgsize_default);
751 ret = mq_attr_ok(ipc_ns, &def_attr);
752 if (ret)
Al Viro765927b2012-06-26 21:58:53 +0400753 return ERR_PTR(ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 }
755
Al Viroce3b0f82009-03-29 19:08:22 -0400756 mode &= ~current_umask();
Al Viro765927b2012-06-26 21:58:53 +0400757 ret = vfs_create(dir, path->dentry, mode, true);
758 path->dentry->d_fsdata = NULL;
Al Viro312b90f2012-08-06 10:18:17 +0400759 if (ret)
760 return ERR_PTR(ret);
761 return dentry_open(path, oflag, cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762}
763
764/* Opens existing queue */
Al Viro765927b2012-06-26 21:58:53 +0400765static struct file *do_open(struct path *path, int oflag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766{
David Howells745ca242008-11-14 10:39:22 +1100767 static const int oflag2acc[O_ACCMODE] = { MAY_READ, MAY_WRITE,
768 MAY_READ | MAY_WRITE };
Al Viro765927b2012-06-26 21:58:53 +0400769 int acc;
770 if ((oflag & O_ACCMODE) == (O_RDWR | O_WRONLY))
771 return ERR_PTR(-EINVAL);
772 acc = oflag2acc[oflag & O_ACCMODE];
David Howells75c3cfa2015-03-17 22:26:12 +0000773 if (inode_permission(d_inode(path->dentry), acc))
Al Viro765927b2012-06-26 21:58:53 +0400774 return ERR_PTR(-EACCES);
775 return dentry_open(path, oflag, current_cred());
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776}
777
Al Virodf0a4282011-07-26 05:26:10 -0400778SYSCALL_DEFINE4(mq_open, const char __user *, u_name, int, oflag, umode_t, mode,
Heiko Carstensd5460c92009-01-14 14:14:27 +0100779 struct mq_attr __user *, u_attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780{
Al Viro765927b2012-06-26 21:58:53 +0400781 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 struct file *filp;
Jeff Layton91a27b22012-10-10 15:25:28 -0400783 struct filename *name;
Al Viro564f6992008-12-14 04:02:26 -0500784 struct mq_attr attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 int fd, error;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700786 struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns;
Al Viro312b90f2012-08-06 10:18:17 +0400787 struct vfsmount *mnt = ipc_ns->mq_mnt;
788 struct dentry *root = mnt->mnt_root;
789 int ro;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790
Al Viro564f6992008-12-14 04:02:26 -0500791 if (u_attr && copy_from_user(&attr, u_attr, sizeof(struct mq_attr)))
792 return -EFAULT;
793
794 audit_mq_open(oflag, mode, u_attr ? &attr : NULL);
George C. Wilson20ca73b2006-05-24 16:09:55 -0500795
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 if (IS_ERR(name = getname(u_name)))
797 return PTR_ERR(name);
798
Ulrich Drepper269f2132008-05-03 15:28:45 -0400799 fd = get_unused_fd_flags(O_CLOEXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 if (fd < 0)
801 goto out_putname;
802
Al Viro312b90f2012-08-06 10:18:17 +0400803 ro = mnt_want_write(mnt); /* we'll drop it in any case */
Al Viro765927b2012-06-26 21:58:53 +0400804 error = 0;
David Howells75c3cfa2015-03-17 22:26:12 +0000805 mutex_lock(&d_inode(root)->i_mutex);
Jeff Layton91a27b22012-10-10 15:25:28 -0400806 path.dentry = lookup_one_len(name->name, root, strlen(name->name));
Al Viro765927b2012-06-26 21:58:53 +0400807 if (IS_ERR(path.dentry)) {
808 error = PTR_ERR(path.dentry);
André Goddard Rosa4294a8e2010-02-23 04:04:28 -0300809 goto out_putfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 }
Al Viro312b90f2012-08-06 10:18:17 +0400811 path.mnt = mntget(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812
813 if (oflag & O_CREAT) {
David Howells75c3cfa2015-03-17 22:26:12 +0000814 if (d_really_is_positive(path.dentry)) { /* entry already exists */
Jeff Laytonadb5c242012-10-10 16:43:13 -0400815 audit_inode(name, path.dentry, 0);
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300816 if (oflag & O_EXCL) {
817 error = -EEXIST;
Alexander Viro7c7dce92006-01-14 15:29:55 -0500818 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300819 }
Al Viro765927b2012-06-26 21:58:53 +0400820 filp = do_open(&path, oflag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 } else {
Al Viro312b90f2012-08-06 10:18:17 +0400822 if (ro) {
823 error = ro;
824 goto out;
825 }
Jeff Layton79f65302013-07-08 15:59:36 -0700826 audit_inode_parent_hidden(name, root);
David Howells75c3cfa2015-03-17 22:26:12 +0000827 filp = do_create(ipc_ns, d_inode(root),
Al Viro765927b2012-06-26 21:58:53 +0400828 &path, oflag, mode,
Al Viro564f6992008-12-14 04:02:26 -0500829 u_attr ? &attr : NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 }
Alexander Viro7c7dce92006-01-14 15:29:55 -0500831 } else {
David Howells75c3cfa2015-03-17 22:26:12 +0000832 if (d_really_is_negative(path.dentry)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300833 error = -ENOENT;
Alexander Viro7c7dce92006-01-14 15:29:55 -0500834 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300835 }
Jeff Laytonadb5c242012-10-10 16:43:13 -0400836 audit_inode(name, path.dentry, 0);
Al Viro765927b2012-06-26 21:58:53 +0400837 filp = do_open(&path, oflag);
Alexander Viro7c7dce92006-01-14 15:29:55 -0500838 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839
Al Viro765927b2012-06-26 21:58:53 +0400840 if (!IS_ERR(filp))
841 fd_install(fd, filp);
842 else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 error = PTR_ERR(filp);
Alexander Viro7c7dce92006-01-14 15:29:55 -0500844out:
Al Viro765927b2012-06-26 21:58:53 +0400845 path_put(&path);
Alexander Viro7c7dce92006-01-14 15:29:55 -0500846out_putfd:
Al Viro765927b2012-06-26 21:58:53 +0400847 if (error) {
848 put_unused_fd(fd);
849 fd = error;
850 }
David Howells75c3cfa2015-03-17 22:26:12 +0000851 mutex_unlock(&d_inode(root)->i_mutex);
Vladimir Davydov38d78e52013-03-22 15:04:51 -0700852 if (!ro)
853 mnt_drop_write(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854out_putname:
855 putname(name);
856 return fd;
857}
858
Heiko Carstensd5460c92009-01-14 14:14:27 +0100859SYSCALL_DEFINE1(mq_unlink, const char __user *, u_name)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860{
861 int err;
Jeff Layton91a27b22012-10-10 15:25:28 -0400862 struct filename *name;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 struct dentry *dentry;
864 struct inode *inode = NULL;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700865 struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns;
Al Viro312b90f2012-08-06 10:18:17 +0400866 struct vfsmount *mnt = ipc_ns->mq_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867
868 name = getname(u_name);
869 if (IS_ERR(name))
870 return PTR_ERR(name);
871
Jeff Layton79f65302013-07-08 15:59:36 -0700872 audit_inode_parent_hidden(name, mnt->mnt_root);
Al Viro312b90f2012-08-06 10:18:17 +0400873 err = mnt_want_write(mnt);
874 if (err)
875 goto out_name;
David Howells75c3cfa2015-03-17 22:26:12 +0000876 mutex_lock_nested(&d_inode(mnt->mnt_root)->i_mutex, I_MUTEX_PARENT);
Jeff Layton91a27b22012-10-10 15:25:28 -0400877 dentry = lookup_one_len(name->name, mnt->mnt_root,
878 strlen(name->name));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 if (IS_ERR(dentry)) {
880 err = PTR_ERR(dentry);
881 goto out_unlock;
882 }
883
David Howells75c3cfa2015-03-17 22:26:12 +0000884 inode = d_inode(dentry);
Al Viro312b90f2012-08-06 10:18:17 +0400885 if (!inode) {
886 err = -ENOENT;
887 } else {
Al Viro7de9c6ee2010-10-23 11:11:40 -0400888 ihold(inode);
David Howells75c3cfa2015-03-17 22:26:12 +0000889 err = vfs_unlink(d_inode(dentry->d_parent), dentry, NULL);
Al Viro312b90f2012-08-06 10:18:17 +0400890 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 dput(dentry);
892
893out_unlock:
David Howells75c3cfa2015-03-17 22:26:12 +0000894 mutex_unlock(&d_inode(mnt->mnt_root)->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 if (inode)
896 iput(inode);
Al Viro312b90f2012-08-06 10:18:17 +0400897 mnt_drop_write(mnt);
898out_name:
899 putname(name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900
901 return err;
902}
903
904/* Pipelined send and receive functions.
905 *
906 * If a receiver finds no waiting message, then it registers itself in the
907 * list of waiting receivers. A sender checks that list before adding the new
908 * message into the message array. If there is a waiting receiver, then it
909 * bypasses the message array and directly hands the message over to the
910 * receiver.
911 * The receiver accepts the message and returns without grabbing the queue
912 * spinlock. Therefore an intermediate STATE_PENDING state and memory barriers
913 * are necessary. The same algorithm is used for sysv semaphores, see
Serge E. Hallyn7b7a3172006-03-28 01:56:23 -0800914 * ipc/sem.c for more details.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 *
916 * The same algorithm is used for senders.
917 */
918
919/* pipelined_send() - send a message directly to the task waiting in
920 * sys_mq_timedreceive() (without inserting message into a queue).
921 */
922static inline void pipelined_send(struct mqueue_inode_info *info,
923 struct msg_msg *message,
924 struct ext_wait_queue *receiver)
925{
926 receiver->msg = message;
927 list_del(&receiver->list);
928 receiver->state = STATE_PENDING;
929 wake_up_process(receiver->task);
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700930 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 receiver->state = STATE_READY;
932}
933
934/* pipelined_receive() - if there is task waiting in sys_mq_timedsend()
935 * gets its message and put to the queue (we have one free place for sure). */
936static inline void pipelined_receive(struct mqueue_inode_info *info)
937{
938 struct ext_wait_queue *sender = wq_get_first_waiter(info, SEND);
939
940 if (!sender) {
941 /* for poll */
942 wake_up_interruptible(&info->wait_q);
943 return;
944 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700945 if (msg_insert(sender->msg, info))
946 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947 list_del(&sender->list);
948 sender->state = STATE_PENDING;
949 wake_up_process(sender->task);
akpm@osdl.orgd59dd462005-05-01 08:58:47 -0700950 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951 sender->state = STATE_READY;
952}
953
Heiko Carstensc4ea37c2009-01-14 14:14:28 +0100954SYSCALL_DEFINE5(mq_timedsend, mqd_t, mqdes, const char __user *, u_msg_ptr,
955 size_t, msg_len, unsigned int, msg_prio,
956 const struct timespec __user *, u_abs_timeout)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957{
Al Viro2903ff02012-08-28 12:52:22 -0400958 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 struct inode *inode;
960 struct ext_wait_queue wait;
961 struct ext_wait_queue *receiver;
962 struct msg_msg *msg_ptr;
963 struct mqueue_inode_info *info;
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200964 ktime_t expires, *timeout = NULL;
965 struct timespec ts;
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700966 struct posix_msg_tree_node *new_leaf = NULL;
Al Viro2903ff02012-08-28 12:52:22 -0400967 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968
Al Viroc32c8af2008-12-14 03:46:48 -0500969 if (u_abs_timeout) {
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200970 int res = prepare_timeout(u_abs_timeout, &expires, &ts);
971 if (res)
972 return res;
973 timeout = &expires;
Al Viroc32c8af2008-12-14 03:46:48 -0500974 }
George C. Wilson20ca73b2006-05-24 16:09:55 -0500975
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976 if (unlikely(msg_prio >= (unsigned long) MQ_PRIO_MAX))
977 return -EINVAL;
978
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200979 audit_mq_sendrecv(mqdes, msg_len, msg_prio, timeout ? &ts : NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980
Al Viro2903ff02012-08-28 12:52:22 -0400981 f = fdget(mqdes);
982 if (unlikely(!f.file)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300983 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300985 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986
Al Viro496ad9a2013-01-23 17:07:38 -0500987 inode = file_inode(f.file);
Al Viro2903ff02012-08-28 12:52:22 -0400988 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300989 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300991 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 info = MQUEUE_I(inode);
Al Viro9f45f5b2014-10-31 17:44:57 -0400993 audit_file(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994
Al Viro2903ff02012-08-28 12:52:22 -0400995 if (unlikely(!(f.file->f_mode & FMODE_WRITE))) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300996 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300998 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999
1000 if (unlikely(msg_len > info->attr.mq_msgsize)) {
1001 ret = -EMSGSIZE;
1002 goto out_fput;
1003 }
1004
1005 /* First try to allocate memory, before doing anything with
1006 * existing queues. */
1007 msg_ptr = load_msg(u_msg_ptr, msg_len);
1008 if (IS_ERR(msg_ptr)) {
1009 ret = PTR_ERR(msg_ptr);
1010 goto out_fput;
1011 }
1012 msg_ptr->m_ts = msg_len;
1013 msg_ptr->m_type = msg_prio;
1014
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001015 /*
1016 * msg_insert really wants us to have a valid, spare node struct so
1017 * it doesn't have to kmalloc a GFP_ATOMIC allocation, but it will
1018 * fall back to that if necessary.
1019 */
1020 if (!info->node_cache)
1021 new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL);
1022
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 spin_lock(&info->lock);
1024
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001025 if (!info->node_cache && new_leaf) {
1026 /* Save our speculative allocation into the cache */
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001027 INIT_LIST_HEAD(&new_leaf->msg_list);
1028 info->node_cache = new_leaf;
1029 info->qsize += sizeof(*new_leaf);
1030 new_leaf = NULL;
1031 } else {
1032 kfree(new_leaf);
1033 }
1034
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 if (info->attr.mq_curmsgs == info->attr.mq_maxmsg) {
Al Viro2903ff02012-08-28 12:52:22 -04001036 if (f.file->f_flags & O_NONBLOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 ret = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038 } else {
1039 wait.task = current;
1040 wait.msg = (void *) msg_ptr;
1041 wait.state = STATE_NONE;
1042 ret = wq_sleep(info, SEND, timeout, &wait);
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001043 /*
1044 * wq_sleep must be called with info->lock held, and
1045 * returns with the lock released
1046 */
1047 goto out_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 } else {
1050 receiver = wq_get_first_waiter(info, RECV);
1051 if (receiver) {
1052 pipelined_send(info, msg_ptr, receiver);
1053 } else {
1054 /* adds message to the queue */
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001055 ret = msg_insert(msg_ptr, info);
1056 if (ret)
1057 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 __do_notify(info);
1059 }
1060 inode->i_atime = inode->i_mtime = inode->i_ctime =
1061 CURRENT_TIME;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062 }
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001063out_unlock:
1064 spin_unlock(&info->lock);
1065out_free:
1066 if (ret)
1067 free_msg(msg_ptr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04001069 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070out:
1071 return ret;
1072}
1073
Heiko Carstensc4ea37c2009-01-14 14:14:28 +01001074SYSCALL_DEFINE5(mq_timedreceive, mqd_t, mqdes, char __user *, u_msg_ptr,
1075 size_t, msg_len, unsigned int __user *, u_msg_prio,
1076 const struct timespec __user *, u_abs_timeout)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 ssize_t ret;
1079 struct msg_msg *msg_ptr;
Al Viro2903ff02012-08-28 12:52:22 -04001080 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 struct inode *inode;
1082 struct mqueue_inode_info *info;
1083 struct ext_wait_queue wait;
Carsten Emde9ca7d8e2010-04-02 22:40:20 +02001084 ktime_t expires, *timeout = NULL;
1085 struct timespec ts;
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001086 struct posix_msg_tree_node *new_leaf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087
Al Viroc32c8af2008-12-14 03:46:48 -05001088 if (u_abs_timeout) {
Carsten Emde9ca7d8e2010-04-02 22:40:20 +02001089 int res = prepare_timeout(u_abs_timeout, &expires, &ts);
1090 if (res)
1091 return res;
1092 timeout = &expires;
Al Viroc32c8af2008-12-14 03:46:48 -05001093 }
George C. Wilson20ca73b2006-05-24 16:09:55 -05001094
Carsten Emde9ca7d8e2010-04-02 22:40:20 +02001095 audit_mq_sendrecv(mqdes, msg_len, 0, timeout ? &ts : NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096
Al Viro2903ff02012-08-28 12:52:22 -04001097 f = fdget(mqdes);
1098 if (unlikely(!f.file)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001099 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001101 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102
Al Viro496ad9a2013-01-23 17:07:38 -05001103 inode = file_inode(f.file);
Al Viro2903ff02012-08-28 12:52:22 -04001104 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001105 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001107 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 info = MQUEUE_I(inode);
Al Viro9f45f5b2014-10-31 17:44:57 -04001109 audit_file(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110
Al Viro2903ff02012-08-28 12:52:22 -04001111 if (unlikely(!(f.file->f_mode & FMODE_READ))) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001112 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001114 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115
1116 /* checks if buffer is big enough */
1117 if (unlikely(msg_len < info->attr.mq_msgsize)) {
1118 ret = -EMSGSIZE;
1119 goto out_fput;
1120 }
1121
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001122 /*
1123 * msg_insert really wants us to have a valid, spare node struct so
1124 * it doesn't have to kmalloc a GFP_ATOMIC allocation, but it will
1125 * fall back to that if necessary.
1126 */
1127 if (!info->node_cache)
1128 new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL);
1129
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 spin_lock(&info->lock);
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001131
1132 if (!info->node_cache && new_leaf) {
1133 /* Save our speculative allocation into the cache */
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001134 INIT_LIST_HEAD(&new_leaf->msg_list);
1135 info->node_cache = new_leaf;
1136 info->qsize += sizeof(*new_leaf);
1137 } else {
1138 kfree(new_leaf);
1139 }
1140
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 if (info->attr.mq_curmsgs == 0) {
Al Viro2903ff02012-08-28 12:52:22 -04001142 if (f.file->f_flags & O_NONBLOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 spin_unlock(&info->lock);
1144 ret = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145 } else {
1146 wait.task = current;
1147 wait.state = STATE_NONE;
1148 ret = wq_sleep(info, RECV, timeout, &wait);
1149 msg_ptr = wait.msg;
1150 }
1151 } else {
1152 msg_ptr = msg_get(info);
1153
1154 inode->i_atime = inode->i_mtime = inode->i_ctime =
1155 CURRENT_TIME;
1156
1157 /* There is now free space in queue. */
1158 pipelined_receive(info);
1159 spin_unlock(&info->lock);
1160 ret = 0;
1161 }
1162 if (ret == 0) {
1163 ret = msg_ptr->m_ts;
1164
1165 if ((u_msg_prio && put_user(msg_ptr->m_type, u_msg_prio)) ||
1166 store_msg(u_msg_ptr, msg_ptr, msg_ptr->m_ts)) {
1167 ret = -EFAULT;
1168 }
1169 free_msg(msg_ptr);
1170 }
1171out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04001172 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173out:
1174 return ret;
1175}
1176
1177/*
1178 * Notes: the case when user wants us to deregister (with NULL as pointer)
1179 * and he isn't currently owner of notification, will be silently discarded.
1180 * It isn't explicitly defined in the POSIX.
1181 */
Heiko Carstensc4ea37c2009-01-14 14:14:28 +01001182SYSCALL_DEFINE2(mq_notify, mqd_t, mqdes,
1183 const struct sigevent __user *, u_notification)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184{
Al Viro2903ff02012-08-28 12:52:22 -04001185 int ret;
1186 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 struct sock *sock;
1188 struct inode *inode;
1189 struct sigevent notification;
1190 struct mqueue_inode_info *info;
1191 struct sk_buff *nc;
1192
Al Viro20114f72008-12-10 07:16:12 -05001193 if (u_notification) {
1194 if (copy_from_user(&notification, u_notification,
1195 sizeof(struct sigevent)))
1196 return -EFAULT;
1197 }
1198
1199 audit_mq_notify(mqdes, u_notification ? &notification : NULL);
George C. Wilson20ca73b2006-05-24 16:09:55 -05001200
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 nc = NULL;
1202 sock = NULL;
1203 if (u_notification != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204 if (unlikely(notification.sigev_notify != SIGEV_NONE &&
1205 notification.sigev_notify != SIGEV_SIGNAL &&
1206 notification.sigev_notify != SIGEV_THREAD))
1207 return -EINVAL;
1208 if (notification.sigev_notify == SIGEV_SIGNAL &&
Jesper Juhl7ed20e12005-05-01 08:59:14 -07001209 !valid_signal(notification.sigev_signo)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210 return -EINVAL;
1211 }
1212 if (notification.sigev_notify == SIGEV_THREAD) {
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001213 long timeo;
1214
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215 /* create the notify skb */
1216 nc = alloc_skb(NOTIFY_COOKIE_LEN, GFP_KERNEL);
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001217 if (!nc) {
1218 ret = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001220 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221 if (copy_from_user(nc->data,
1222 notification.sigev_value.sival_ptr,
1223 NOTIFY_COOKIE_LEN)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001224 ret = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225 goto out;
1226 }
1227
1228 /* TODO: add a header? */
1229 skb_put(nc, NOTIFY_COOKIE_LEN);
1230 /* and attach it to the socket */
1231retry:
Al Viro2903ff02012-08-28 12:52:22 -04001232 f = fdget(notification.sigev_signo);
1233 if (!f.file) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001234 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001236 }
Al Viro2903ff02012-08-28 12:52:22 -04001237 sock = netlink_getsockbyfilp(f.file);
1238 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 if (IS_ERR(sock)) {
1240 ret = PTR_ERR(sock);
1241 sock = NULL;
1242 goto out;
1243 }
1244
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001245 timeo = MAX_SCHEDULE_TIMEOUT;
Denis V. Lunev9457afe2008-06-05 11:23:39 -07001246 ret = netlink_attachskb(sock, nc, &timeo, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247 if (ret == 1)
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001248 goto retry;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249 if (ret) {
1250 sock = NULL;
1251 nc = NULL;
1252 goto out;
1253 }
1254 }
1255 }
1256
Al Viro2903ff02012-08-28 12:52:22 -04001257 f = fdget(mqdes);
1258 if (!f.file) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001259 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001261 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262
Al Viro496ad9a2013-01-23 17:07:38 -05001263 inode = file_inode(f.file);
Al Viro2903ff02012-08-28 12:52:22 -04001264 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001265 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001267 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 info = MQUEUE_I(inode);
1269
1270 ret = 0;
1271 spin_lock(&info->lock);
1272 if (u_notification == NULL) {
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001273 if (info->notify_owner == task_tgid(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 remove_notification(info);
1275 inode->i_atime = inode->i_ctime = CURRENT_TIME;
1276 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001277 } else if (info->notify_owner != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278 ret = -EBUSY;
1279 } else {
1280 switch (notification.sigev_notify) {
1281 case SIGEV_NONE:
1282 info->notify.sigev_notify = SIGEV_NONE;
1283 break;
1284 case SIGEV_THREAD:
1285 info->notify_sock = sock;
1286 info->notify_cookie = nc;
1287 sock = NULL;
1288 nc = NULL;
1289 info->notify.sigev_notify = SIGEV_THREAD;
1290 break;
1291 case SIGEV_SIGNAL:
1292 info->notify.sigev_signo = notification.sigev_signo;
1293 info->notify.sigev_value = notification.sigev_value;
1294 info->notify.sigev_notify = SIGEV_SIGNAL;
1295 break;
1296 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001297
1298 info->notify_owner = get_pid(task_tgid(current));
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -08001299 info->notify_user_ns = get_user_ns(current_user_ns());
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300 inode->i_atime = inode->i_ctime = CURRENT_TIME;
1301 }
1302 spin_unlock(&info->lock);
1303out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04001304 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305out:
Davidlohr Bueso3ab08fe2014-01-27 17:07:06 -08001306 if (sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 netlink_detachskb(sock, nc);
Davidlohr Bueso3ab08fe2014-01-27 17:07:06 -08001308 else if (nc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309 dev_kfree_skb(nc);
Davidlohr Bueso3ab08fe2014-01-27 17:07:06 -08001310
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 return ret;
1312}
1313
Heiko Carstensc4ea37c2009-01-14 14:14:28 +01001314SYSCALL_DEFINE3(mq_getsetattr, mqd_t, mqdes,
1315 const struct mq_attr __user *, u_mqstat,
1316 struct mq_attr __user *, u_omqstat)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317{
1318 int ret;
1319 struct mq_attr mqstat, omqstat;
Al Viro2903ff02012-08-28 12:52:22 -04001320 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 struct inode *inode;
1322 struct mqueue_inode_info *info;
1323
1324 if (u_mqstat != NULL) {
1325 if (copy_from_user(&mqstat, u_mqstat, sizeof(struct mq_attr)))
1326 return -EFAULT;
1327 if (mqstat.mq_flags & (~O_NONBLOCK))
1328 return -EINVAL;
1329 }
1330
Al Viro2903ff02012-08-28 12:52:22 -04001331 f = fdget(mqdes);
1332 if (!f.file) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001333 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001335 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336
Al Viro496ad9a2013-01-23 17:07:38 -05001337 inode = file_inode(f.file);
Al Viro2903ff02012-08-28 12:52:22 -04001338 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001339 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001341 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 info = MQUEUE_I(inode);
1343
1344 spin_lock(&info->lock);
1345
1346 omqstat = info->attr;
Al Viro2903ff02012-08-28 12:52:22 -04001347 omqstat.mq_flags = f.file->f_flags & O_NONBLOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348 if (u_mqstat) {
Al Viro73929062008-12-10 06:58:59 -05001349 audit_mq_getsetattr(mqdes, &mqstat);
Al Viro2903ff02012-08-28 12:52:22 -04001350 spin_lock(&f.file->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351 if (mqstat.mq_flags & O_NONBLOCK)
Al Viro2903ff02012-08-28 12:52:22 -04001352 f.file->f_flags |= O_NONBLOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353 else
Al Viro2903ff02012-08-28 12:52:22 -04001354 f.file->f_flags &= ~O_NONBLOCK;
1355 spin_unlock(&f.file->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356
1357 inode->i_atime = inode->i_ctime = CURRENT_TIME;
1358 }
1359
1360 spin_unlock(&info->lock);
1361
1362 ret = 0;
1363 if (u_omqstat != NULL && copy_to_user(u_omqstat, &omqstat,
1364 sizeof(struct mq_attr)))
1365 ret = -EFAULT;
1366
1367out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04001368 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369out:
1370 return ret;
1371}
1372
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -08001373static const struct inode_operations mqueue_dir_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 .lookup = simple_lookup,
1375 .create = mqueue_create,
1376 .unlink = mqueue_unlink,
1377};
1378
Arjan van de Ven9a321442007-02-12 00:55:35 -08001379static const struct file_operations mqueue_file_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380 .flush = mqueue_flush_file,
1381 .poll = mqueue_poll_file,
1382 .read = mqueue_read_file,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001383 .llseek = default_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384};
1385
Alexey Dobriyanb87221d2009-09-21 17:01:09 -07001386static const struct super_operations mqueue_super_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 .alloc_inode = mqueue_alloc_inode,
1388 .destroy_inode = mqueue_destroy_inode,
Al Viro6d8af642010-06-05 16:29:45 -04001389 .evict_inode = mqueue_evict_inode,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391};
1392
1393static struct file_system_type mqueue_fs_type = {
1394 .name = "mqueue",
Al Viroceefda62010-07-26 13:16:50 +04001395 .mount = mqueue_mount,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396 .kill_sb = kill_litter_super,
Gao fengbc1b69e2013-01-28 11:09:01 +08001397 .fs_flags = FS_USERNS_MOUNT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398};
1399
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001400int mq_init_ns(struct ipc_namespace *ns)
1401{
1402 ns->mq_queues_count = 0;
1403 ns->mq_queues_max = DFLT_QUEUESMAX;
1404 ns->mq_msg_max = DFLT_MSGMAX;
1405 ns->mq_msgsize_max = DFLT_MSGSIZEMAX;
KOSAKI Motohirocef01842012-05-31 16:26:33 -07001406 ns->mq_msg_default = DFLT_MSG;
1407 ns->mq_msgsize_default = DFLT_MSGSIZE;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001408
1409 ns->mq_mnt = kern_mount_data(&mqueue_fs_type, ns);
1410 if (IS_ERR(ns->mq_mnt)) {
1411 int err = PTR_ERR(ns->mq_mnt);
1412 ns->mq_mnt = NULL;
1413 return err;
1414 }
1415 return 0;
1416}
1417
1418void mq_clear_sbinfo(struct ipc_namespace *ns)
1419{
1420 ns->mq_mnt->mnt_sb->s_fs_info = NULL;
1421}
1422
1423void mq_put_mnt(struct ipc_namespace *ns)
1424{
Al Viro6f686572011-12-09 00:38:50 -05001425 kern_unmount(ns->mq_mnt);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001426}
1427
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428static int __init init_mqueue_fs(void)
1429{
1430 int error;
1431
1432 mqueue_inode_cachep = kmem_cache_create("mqueue_inode_cache",
1433 sizeof(struct mqueue_inode_info), 0,
Paul Mundt20c2df82007-07-20 10:11:58 +09001434 SLAB_HWCACHE_ALIGN, init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 if (mqueue_inode_cachep == NULL)
1436 return -ENOMEM;
1437
André Goddard Rosa2329e3922010-02-23 04:04:27 -03001438 /* ignore failures - they are not fatal */
Serge E. Hallynbdc8e5f2009-04-06 19:01:11 -07001439 mq_sysctl_table = mq_register_sysctl_table();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440
1441 error = register_filesystem(&mqueue_fs_type);
1442 if (error)
1443 goto out_sysctl;
1444
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001445 spin_lock_init(&mq_lock);
1446
Al Viro6f686572011-12-09 00:38:50 -05001447 error = mq_init_ns(&init_ipc_ns);
1448 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 goto out_filesystem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451 return 0;
1452
1453out_filesystem:
1454 unregister_filesystem(&mqueue_fs_type);
1455out_sysctl:
1456 if (mq_sysctl_table)
1457 unregister_sysctl_table(mq_sysctl_table);
Alexey Dobriyan1a1d92c2006-09-27 01:49:40 -07001458 kmem_cache_destroy(mqueue_inode_cachep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459 return error;
1460}
1461
Davidlohr Bueso6d08a252014-04-07 15:39:18 -07001462device_initcall(init_mqueue_fs);