blob: b3f22e610c7e9e28679321f12d136d3e60de216a [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
Davidlohr Buesofa6004a2015-05-04 07:02:46 -070050#define STATE_READY 1
Linus Torvalds1da177e2005-04-16 15:20:36 -070051
Doug Ledfordd6629852012-05-31 16:26:35 -070052struct posix_msg_tree_node {
53 struct rb_node rb_node;
54 struct list_head msg_list;
55 int priority;
56};
57
Linus Torvalds1da177e2005-04-16 15:20:36 -070058struct ext_wait_queue { /* queue of sleeping tasks */
59 struct task_struct *task;
60 struct list_head list;
61 struct msg_msg *msg; /* ptr of loaded message */
62 int state; /* one of STATE_* values */
63};
64
65struct mqueue_inode_info {
66 spinlock_t lock;
67 struct inode vfs_inode;
68 wait_queue_head_t wait_q;
69
Doug Ledfordd6629852012-05-31 16:26:35 -070070 struct rb_root msg_tree;
Doug Ledfordce2d52c2012-05-31 16:26:38 -070071 struct posix_msg_tree_node *node_cache;
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 struct mq_attr attr;
73
74 struct sigevent notify;
Manfred Spraul239521f2014-01-27 17:07:04 -080075 struct pid *notify_owner;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -080076 struct user_namespace *notify_user_ns;
Adrian Bunk338cec32005-09-10 00:26:54 -070077 struct user_struct *user; /* user who created, for accounting */
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 struct sock *notify_sock;
79 struct sk_buff *notify_cookie;
80
81 /* for tasks waiting for free space and messages, respectively */
82 struct ext_wait_queue e_wait_q[2];
83
84 unsigned long qsize; /* size of queue in memory (sum of all msgs) */
85};
86
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -080087static const struct inode_operations mqueue_dir_inode_operations;
Arjan van de Ven9a321442007-02-12 00:55:35 -080088static const struct file_operations mqueue_file_operations;
Alexey Dobriyanb87221d2009-09-21 17:01:09 -070089static const struct super_operations mqueue_super_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -070090static void remove_notification(struct mqueue_inode_info *info);
91
Christoph Lametere18b8902006-12-06 20:33:20 -080092static struct kmem_cache *mqueue_inode_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -070093
Manfred Spraul239521f2014-01-27 17:07:04 -080094static struct ctl_table_header *mq_sysctl_table;
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
96static inline struct mqueue_inode_info *MQUEUE_I(struct inode *inode)
97{
98 return container_of(inode, struct mqueue_inode_info, vfs_inode);
99}
100
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700101/*
102 * This routine should be called with the mq_lock held.
103 */
104static inline struct ipc_namespace *__get_ns_from_inode(struct inode *inode)
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700105{
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700106 return get_ipc_ns(inode->i_sb->s_fs_info);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700107}
108
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700109static struct ipc_namespace *get_ns_from_inode(struct inode *inode)
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700110{
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700111 struct ipc_namespace *ns;
112
113 spin_lock(&mq_lock);
114 ns = __get_ns_from_inode(inode);
115 spin_unlock(&mq_lock);
116 return ns;
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700117}
118
Doug Ledfordd6629852012-05-31 16:26:35 -0700119/* Auxiliary functions to manipulate messages' list */
120static int msg_insert(struct msg_msg *msg, struct mqueue_inode_info *info)
121{
122 struct rb_node **p, *parent = NULL;
123 struct posix_msg_tree_node *leaf;
124
125 p = &info->msg_tree.rb_node;
126 while (*p) {
127 parent = *p;
128 leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node);
129
130 if (likely(leaf->priority == msg->m_type))
131 goto insert_msg;
132 else if (msg->m_type < leaf->priority)
133 p = &(*p)->rb_left;
134 else
135 p = &(*p)->rb_right;
136 }
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700137 if (info->node_cache) {
138 leaf = info->node_cache;
139 info->node_cache = NULL;
140 } else {
141 leaf = kmalloc(sizeof(*leaf), GFP_ATOMIC);
142 if (!leaf)
143 return -ENOMEM;
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700144 INIT_LIST_HEAD(&leaf->msg_list);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700145 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700146 leaf->priority = msg->m_type;
147 rb_link_node(&leaf->rb_node, parent, p);
148 rb_insert_color(&leaf->rb_node, &info->msg_tree);
Doug Ledfordd6629852012-05-31 16:26:35 -0700149insert_msg:
150 info->attr.mq_curmsgs++;
151 info->qsize += msg->m_ts;
152 list_add_tail(&msg->m_list, &leaf->msg_list);
153 return 0;
154}
155
156static inline struct msg_msg *msg_get(struct mqueue_inode_info *info)
157{
158 struct rb_node **p, *parent = NULL;
159 struct posix_msg_tree_node *leaf;
160 struct msg_msg *msg;
161
162try_again:
163 p = &info->msg_tree.rb_node;
164 while (*p) {
165 parent = *p;
166 /*
167 * During insert, low priorities go to the left and high to the
168 * right. On receive, we want the highest priorities first, so
169 * walk all the way to the right.
170 */
171 p = &(*p)->rb_right;
172 }
173 if (!parent) {
174 if (info->attr.mq_curmsgs) {
175 pr_warn_once("Inconsistency in POSIX message queue, "
176 "no tree element, but supposedly messages "
177 "should exist!\n");
178 info->attr.mq_curmsgs = 0;
179 }
180 return NULL;
181 }
182 leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700183 if (unlikely(list_empty(&leaf->msg_list))) {
Doug Ledfordd6629852012-05-31 16:26:35 -0700184 pr_warn_once("Inconsistency in POSIX message queue, "
185 "empty leaf node but we haven't implemented "
186 "lazy leaf delete!\n");
187 rb_erase(&leaf->rb_node, &info->msg_tree);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700188 if (info->node_cache) {
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700189 kfree(leaf);
190 } else {
191 info->node_cache = leaf;
192 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700193 goto try_again;
194 } else {
195 msg = list_first_entry(&leaf->msg_list,
196 struct msg_msg, m_list);
197 list_del(&msg->m_list);
198 if (list_empty(&leaf->msg_list)) {
199 rb_erase(&leaf->rb_node, &info->msg_tree);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700200 if (info->node_cache) {
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700201 kfree(leaf);
202 } else {
203 info->node_cache = leaf;
204 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700205 }
206 }
207 info->attr.mq_curmsgs--;
208 info->qsize -= msg->m_ts;
209 return msg;
210}
211
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700212static struct inode *mqueue_get_inode(struct super_block *sb,
Al Viro1b9d5ff72011-07-24 14:18:20 -0400213 struct ipc_namespace *ipc_ns, umode_t mode,
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700214 struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215{
David Howells86a264a2008-11-14 10:39:18 +1100216 struct user_struct *u = current_user();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 struct inode *inode;
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700218 int ret = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219
220 inode = new_inode(sb);
Jiri Slaby04715202011-07-26 16:08:46 -0700221 if (!inode)
222 goto err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223
Jiri Slaby04715202011-07-26 16:08:46 -0700224 inode->i_ino = get_next_ino();
225 inode->i_mode = mode;
226 inode->i_uid = current_fsuid();
227 inode->i_gid = current_fsgid();
Deepa Dinamani078cd822016-09-14 07:48:04 -0700228 inode->i_mtime = inode->i_ctime = inode->i_atime = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229
Jiri Slaby04715202011-07-26 16:08:46 -0700230 if (S_ISREG(mode)) {
231 struct mqueue_inode_info *info;
Doug Ledfordd6629852012-05-31 16:26:35 -0700232 unsigned long mq_bytes, mq_treesize;
André Goddard Rosac8308b12010-02-23 04:04:23 -0300233
Jiri Slaby04715202011-07-26 16:08:46 -0700234 inode->i_fop = &mqueue_file_operations;
235 inode->i_size = FILENT_SIZE;
236 /* mqueue specific info */
237 info = MQUEUE_I(inode);
238 spin_lock_init(&info->lock);
239 init_waitqueue_head(&info->wait_q);
240 INIT_LIST_HEAD(&info->e_wait_q[0].list);
241 INIT_LIST_HEAD(&info->e_wait_q[1].list);
242 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800243 info->notify_user_ns = NULL;
Jiri Slaby04715202011-07-26 16:08:46 -0700244 info->qsize = 0;
245 info->user = NULL; /* set when all is ok */
Doug Ledfordd6629852012-05-31 16:26:35 -0700246 info->msg_tree = RB_ROOT;
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700247 info->node_cache = NULL;
Jiri Slaby04715202011-07-26 16:08:46 -0700248 memset(&info->attr, 0, sizeof(info->attr));
KOSAKI Motohirocef01842012-05-31 16:26:33 -0700249 info->attr.mq_maxmsg = min(ipc_ns->mq_msg_max,
250 ipc_ns->mq_msg_default);
251 info->attr.mq_msgsize = min(ipc_ns->mq_msgsize_max,
252 ipc_ns->mq_msgsize_default);
Jiri Slaby04715202011-07-26 16:08:46 -0700253 if (attr) {
254 info->attr.mq_maxmsg = attr->mq_maxmsg;
255 info->attr.mq_msgsize = attr->mq_msgsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700257 /*
258 * We used to allocate a static array of pointers and account
259 * the size of that array as well as one msg_msg struct per
260 * possible message into the queue size. That's no longer
261 * accurate as the queue is now an rbtree and will grow and
262 * shrink depending on usage patterns. We can, however, still
263 * account one msg_msg struct per message, but the nodes are
264 * allocated depending on priority usage, and most programs
265 * only use one, or a handful, of priorities. However, since
266 * this is pinned memory, we need to assume worst case, so
267 * that means the min(mq_maxmsg, max_priorities) * struct
268 * posix_msg_tree_node.
269 */
270 mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) +
271 min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) *
272 sizeof(struct posix_msg_tree_node);
Jiri Slaby04715202011-07-26 16:08:46 -0700273
Doug Ledfordd6629852012-05-31 16:26:35 -0700274 mq_bytes = mq_treesize + (info->attr.mq_maxmsg *
275 info->attr.mq_msgsize);
Jiri Slaby04715202011-07-26 16:08:46 -0700276
277 spin_lock(&mq_lock);
278 if (u->mq_bytes + mq_bytes < u->mq_bytes ||
Davidlohr Bueso2a4e64b2012-01-20 14:34:01 -0800279 u->mq_bytes + mq_bytes > rlimit(RLIMIT_MSGQUEUE)) {
Jiri Slaby04715202011-07-26 16:08:46 -0700280 spin_unlock(&mq_lock);
281 /* mqueue_evict_inode() releases info->messages */
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700282 ret = -EMFILE;
Jiri Slaby04715202011-07-26 16:08:46 -0700283 goto out_inode;
284 }
285 u->mq_bytes += mq_bytes;
286 spin_unlock(&mq_lock);
287
288 /* all is ok */
289 info->user = get_uid(u);
290 } else if (S_ISDIR(mode)) {
291 inc_nlink(inode);
292 /* Some things misbehave if size == 0 on a directory */
293 inode->i_size = 2 * DIRENT_SIZE;
294 inode->i_op = &mqueue_dir_inode_operations;
295 inode->i_fop = &simple_dir_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296 }
Jiri Slaby04715202011-07-26 16:08:46 -0700297
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 return inode;
299out_inode:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 iput(inode);
Jiri Slaby04715202011-07-26 16:08:46 -0700301err:
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700302 return ERR_PTR(ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303}
304
305static int mqueue_fill_super(struct super_block *sb, void *data, int silent)
306{
307 struct inode *inode;
Eric W. Biedermand91ee872016-05-23 14:51:59 -0500308 struct ipc_namespace *ns = sb->s_fs_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309
Eric W. Biedermana2982cc2016-06-09 15:34:02 -0500310 sb->s_iflags |= SB_I_NOEXEC | SB_I_NODEV;
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300311 sb->s_blocksize = PAGE_SIZE;
312 sb->s_blocksize_bits = PAGE_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313 sb->s_magic = MQUEUE_MAGIC;
314 sb->s_op = &mqueue_super_ops;
315
Al Viro48fde702012-01-08 22:15:13 -0500316 inode = mqueue_get_inode(sb, ns, S_IFDIR | S_ISVTX | S_IRWXUGO, NULL);
317 if (IS_ERR(inode))
318 return PTR_ERR(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319
Al Viro48fde702012-01-08 22:15:13 -0500320 sb->s_root = d_make_root(inode);
321 if (!sb->s_root)
322 return -ENOMEM;
323 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324}
325
Al Viroceefda62010-07-26 13:16:50 +0400326static struct dentry *mqueue_mount(struct file_system_type *fs_type,
David Howells454e2392006-06-23 02:02:57 -0700327 int flags, const char *dev_name,
Al Viroceefda62010-07-26 13:16:50 +0400328 void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329{
Eric W. Biedermand91ee872016-05-23 14:51:59 -0500330 struct ipc_namespace *ns;
331 if (flags & MS_KERNMOUNT) {
332 ns = data;
333 data = NULL;
334 } else {
335 ns = current->nsproxy->ipc_ns;
Eric W. Biedermana636b702013-03-21 18:13:15 -0700336 }
Eric W. Biedermand91ee872016-05-23 14:51:59 -0500337 return mount_ns(fs_type, flags, data, ns, ns->user_ns, mqueue_fill_super);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338}
339
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700340static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341{
342 struct mqueue_inode_info *p = (struct mqueue_inode_info *) foo;
343
Christoph Lametera35afb82007-05-16 22:10:57 -0700344 inode_init_once(&p->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345}
346
347static struct inode *mqueue_alloc_inode(struct super_block *sb)
348{
349 struct mqueue_inode_info *ei;
350
Christoph Lametere94b1762006-12-06 20:33:17 -0800351 ei = kmem_cache_alloc(mqueue_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 if (!ei)
353 return NULL;
354 return &ei->vfs_inode;
355}
356
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100357static void mqueue_i_callback(struct rcu_head *head)
358{
359 struct inode *inode = container_of(head, struct inode, i_rcu);
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100360 kmem_cache_free(mqueue_inode_cachep, MQUEUE_I(inode));
361}
362
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363static void mqueue_destroy_inode(struct inode *inode)
364{
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100365 call_rcu(&inode->i_rcu, mqueue_i_callback);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366}
367
Al Viro6d8af642010-06-05 16:29:45 -0400368static void mqueue_evict_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369{
370 struct mqueue_inode_info *info;
371 struct user_struct *user;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700372 struct ipc_namespace *ipc_ns;
Li Rongqing7b5598c2019-05-14 15:46:20 -0700373 struct msg_msg *msg, *nmsg;
374 LIST_HEAD(tmp_msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375
Jan Karadbd57682012-05-03 14:48:02 +0200376 clear_inode(inode);
Al Viro6d8af642010-06-05 16:29:45 -0400377
378 if (S_ISDIR(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 return;
Al Viro6d8af642010-06-05 16:29:45 -0400380
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700381 ipc_ns = get_ns_from_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 info = MQUEUE_I(inode);
383 spin_lock(&info->lock);
Doug Ledfordd6629852012-05-31 16:26:35 -0700384 while ((msg = msg_get(info)) != NULL)
Li Rongqing7b5598c2019-05-14 15:46:20 -0700385 list_add_tail(&msg->m_list, &tmp_msg);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700386 kfree(info->node_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387 spin_unlock(&info->lock);
388
Li Rongqing7b5598c2019-05-14 15:46:20 -0700389 list_for_each_entry_safe(msg, nmsg, &tmp_msg, m_list) {
390 list_del(&msg->m_list);
391 free_msg(msg);
392 }
393
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 user = info->user;
395 if (user) {
Kees Cook62369d52019-07-16 16:30:21 -0700396 unsigned long mq_bytes, mq_treesize;
397
398 /* Total amount of bytes accounted for the mqueue */
399 mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) +
400 min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) *
401 sizeof(struct posix_msg_tree_node);
402
403 mq_bytes = mq_treesize + (info->attr.mq_maxmsg *
404 info->attr.mq_msgsize);
405
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 spin_lock(&mq_lock);
407 user->mq_bytes -= mq_bytes;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700408 /*
409 * get_ns_from_inode() ensures that the
410 * (ipc_ns = sb->s_fs_info) is either a valid ipc_ns
411 * to which we now hold a reference, or it is NULL.
412 * We can't put it here under mq_lock, though.
413 */
414 if (ipc_ns)
415 ipc_ns->mq_queues_count--;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 spin_unlock(&mq_lock);
417 free_uid(user);
418 }
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700419 if (ipc_ns)
420 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421}
422
423static int mqueue_create(struct inode *dir, struct dentry *dentry,
Al Viroebfc3b42012-06-10 18:05:36 -0400424 umode_t mode, bool excl)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425{
426 struct inode *inode;
427 struct mq_attr *attr = dentry->d_fsdata;
428 int error;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700429 struct ipc_namespace *ipc_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430
431 spin_lock(&mq_lock);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700432 ipc_ns = __get_ns_from_inode(dir);
433 if (!ipc_ns) {
434 error = -EACCES;
435 goto out_unlock;
436 }
Davidlohr Buesof3713fd2014-02-25 15:01:45 -0800437
438 if (ipc_ns->mq_queues_count >= ipc_ns->mq_queues_max &&
439 !capable(CAP_SYS_RESOURCE)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 error = -ENOSPC;
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700441 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 }
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700443 ipc_ns->mq_queues_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 spin_unlock(&mq_lock);
445
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700446 inode = mqueue_get_inode(dir->i_sb, ipc_ns, mode, attr);
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700447 if (IS_ERR(inode)) {
448 error = PTR_ERR(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 spin_lock(&mq_lock);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700450 ipc_ns->mq_queues_count--;
451 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 }
453
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700454 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 dir->i_size += DIRENT_SIZE;
Deepa Dinamani078cd822016-09-14 07:48:04 -0700456 dir->i_ctime = dir->i_mtime = dir->i_atime = current_time(dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457
458 d_instantiate(dentry, inode);
459 dget(dentry);
460 return 0;
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700461out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 spin_unlock(&mq_lock);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700463 if (ipc_ns)
464 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 return error;
466}
467
468static int mqueue_unlink(struct inode *dir, struct dentry *dentry)
469{
David Howells75c3cfa2015-03-17 22:26:12 +0000470 struct inode *inode = d_inode(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471
Deepa Dinamani078cd822016-09-14 07:48:04 -0700472 dir->i_ctime = dir->i_mtime = dir->i_atime = current_time(dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 dir->i_size -= DIRENT_SIZE;
Manfred Spraul239521f2014-01-27 17:07:04 -0800474 drop_nlink(inode);
475 dput(dentry);
476 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477}
478
479/*
480* This is routine for system read from queue file.
481* To avoid mess with doing here some sort of mq_receive we allow
482* to read only queue size & notification info (the only values
483* that are interesting from user point of view and aren't accessible
484* through std routines)
485*/
486static ssize_t mqueue_read_file(struct file *filp, char __user *u_data,
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700487 size_t count, loff_t *off)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488{
Al Viro496ad9a2013-01-23 17:07:38 -0500489 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 char buffer[FILENT_SIZE];
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700491 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492
493 spin_lock(&info->lock);
494 snprintf(buffer, sizeof(buffer),
495 "QSIZE:%-10lu NOTIFY:%-5d SIGNO:%-5d NOTIFY_PID:%-6d\n",
496 info->qsize,
497 info->notify_owner ? info->notify.sigev_notify : 0,
498 (info->notify_owner &&
499 info->notify.sigev_notify == SIGEV_SIGNAL) ?
500 info->notify.sigev_signo : 0,
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800501 pid_vnr(info->notify_owner));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 spin_unlock(&info->lock);
503 buffer[sizeof(buffer)-1] = '\0';
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700505 ret = simple_read_from_buffer(u_data, count, off, buffer,
506 strlen(buffer));
507 if (ret <= 0)
508 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509
Deepa Dinamani078cd822016-09-14 07:48:04 -0700510 file_inode(filp)->i_atime = file_inode(filp)->i_ctime = current_time(file_inode(filp));
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700511 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512}
513
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -0700514static int mqueue_flush_file(struct file *filp, fl_owner_t id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515{
Al Viro496ad9a2013-01-23 17:07:38 -0500516 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517
518 spin_lock(&info->lock);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700519 if (task_tgid(current) == info->notify_owner)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 remove_notification(info);
521
522 spin_unlock(&info->lock);
523 return 0;
524}
525
526static unsigned int mqueue_poll_file(struct file *filp, struct poll_table_struct *poll_tab)
527{
Al Viro496ad9a2013-01-23 17:07:38 -0500528 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 int retval = 0;
530
531 poll_wait(filp, &info->wait_q, poll_tab);
532
533 spin_lock(&info->lock);
534 if (info->attr.mq_curmsgs)
535 retval = POLLIN | POLLRDNORM;
536
537 if (info->attr.mq_curmsgs < info->attr.mq_maxmsg)
538 retval |= POLLOUT | POLLWRNORM;
539 spin_unlock(&info->lock);
540
541 return retval;
542}
543
544/* Adds current to info->e_wait_q[sr] before element with smaller prio */
545static void wq_add(struct mqueue_inode_info *info, int sr,
546 struct ext_wait_queue *ewp)
547{
548 struct ext_wait_queue *walk;
549
550 ewp->task = current;
551
552 list_for_each_entry(walk, &info->e_wait_q[sr].list, list) {
553 if (walk->task->static_prio <= current->static_prio) {
554 list_add_tail(&ewp->list, &walk->list);
555 return;
556 }
557 }
558 list_add_tail(&ewp->list, &info->e_wait_q[sr].list);
559}
560
561/*
562 * Puts current task to sleep. Caller must hold queue lock. After return
563 * lock isn't held.
564 * sr: SEND or RECV
565 */
566static int wq_sleep(struct mqueue_inode_info *info, int sr,
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200567 ktime_t *timeout, struct ext_wait_queue *ewp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568{
569 int retval;
570 signed long time;
571
572 wq_add(info, sr, ewp);
573
574 for (;;) {
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700575 __set_current_state(TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576
577 spin_unlock(&info->lock);
Wanlong Gao32ea8452011-10-31 17:06:35 -0700578 time = schedule_hrtimeout_range_clock(timeout, 0,
579 HRTIMER_MODE_ABS, CLOCK_REALTIME);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 if (ewp->state == STATE_READY) {
582 retval = 0;
583 goto out;
584 }
585 spin_lock(&info->lock);
586 if (ewp->state == STATE_READY) {
587 retval = 0;
588 goto out_unlock;
589 }
590 if (signal_pending(current)) {
591 retval = -ERESTARTSYS;
592 break;
593 }
594 if (time == 0) {
595 retval = -ETIMEDOUT;
596 break;
597 }
598 }
599 list_del(&ewp->list);
600out_unlock:
601 spin_unlock(&info->lock);
602out:
603 return retval;
604}
605
606/*
607 * Returns waiting task that should be serviced first or NULL if none exists
608 */
609static struct ext_wait_queue *wq_get_first_waiter(
610 struct mqueue_inode_info *info, int sr)
611{
612 struct list_head *ptr;
613
614 ptr = info->e_wait_q[sr].list.prev;
615 if (ptr == &info->e_wait_q[sr].list)
616 return NULL;
617 return list_entry(ptr, struct ext_wait_queue, list);
618}
619
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620
621static inline void set_cookie(struct sk_buff *skb, char code)
622{
Manfred Spraul239521f2014-01-27 17:07:04 -0800623 ((char *)skb->data)[NOTIFY_COOKIE_LEN-1] = code;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624}
625
626/*
627 * The next function is only to split too long sys_mq_timedsend
628 */
629static void __do_notify(struct mqueue_inode_info *info)
630{
631 /* notification
632 * invoked when there is registered process and there isn't process
633 * waiting synchronously for message AND state of queue changed from
634 * empty to not empty. Here we are sure that no one is waiting
635 * synchronously. */
636 if (info->notify_owner &&
637 info->attr.mq_curmsgs == 1) {
638 struct siginfo sig_i;
639 switch (info->notify.sigev_notify) {
640 case SIGEV_NONE:
641 break;
642 case SIGEV_SIGNAL:
643 /* sends signal */
644
645 sig_i.si_signo = info->notify.sigev_signo;
646 sig_i.si_errno = 0;
647 sig_i.si_code = SI_MESGQ;
648 sig_i.si_value = info->notify.sigev_value;
Serge E. Hallyn6b550f92012-01-10 15:11:37 -0800649 /* map current pid/uid into info->owner's namespaces */
650 rcu_read_lock();
Sukadev Bhattiprolua6684992009-01-07 18:08:50 -0800651 sig_i.si_pid = task_tgid_nr_ns(current,
652 ns_of_pid(info->notify_owner));
Eric W. Biederman76b6db02012-03-14 15:24:19 -0700653 sig_i.si_uid = from_kuid_munged(info->notify_user_ns, current_uid());
Serge E. Hallyn6b550f92012-01-10 15:11:37 -0800654 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700656 kill_pid_info(info->notify.sigev_signo,
657 &sig_i, info->notify_owner);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 break;
659 case SIGEV_THREAD:
660 set_cookie(info->notify_cookie, NOTIFY_WOKENUP);
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700661 netlink_sendskb(info->notify_sock, info->notify_cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 break;
663 }
664 /* after notification unregisters process */
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700665 put_pid(info->notify_owner);
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800666 put_user_ns(info->notify_user_ns);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700667 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800668 info->notify_user_ns = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 }
670 wake_up(&info->wait_q);
671}
672
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200673static int prepare_timeout(const struct timespec __user *u_abs_timeout,
674 ktime_t *expires, struct timespec *ts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675{
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200676 if (copy_from_user(ts, u_abs_timeout, sizeof(struct timespec)))
677 return -EFAULT;
678 if (!timespec_valid(ts))
679 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200681 *expires = timespec_to_ktime(*ts);
682 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683}
684
685static void remove_notification(struct mqueue_inode_info *info)
686{
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700687 if (info->notify_owner != NULL &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 info->notify.sigev_notify == SIGEV_THREAD) {
689 set_cookie(info->notify_cookie, NOTIFY_REMOVED);
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700690 netlink_sendskb(info->notify_sock, info->notify_cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700692 put_pid(info->notify_owner);
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800693 put_user_ns(info->notify_user_ns);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700694 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800695 info->notify_user_ns = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696}
697
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700698static int mq_attr_ok(struct ipc_namespace *ipc_ns, struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699{
Doug Ledford2c12ea42012-05-31 16:26:36 -0700700 int mq_treesize;
701 unsigned long total_size;
702
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 if (attr->mq_maxmsg <= 0 || attr->mq_msgsize <= 0)
Doug Ledford113289c2012-05-31 16:26:36 -0700704 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 if (capable(CAP_SYS_RESOURCE)) {
Doug Ledford02967ea2012-05-31 16:26:29 -0700706 if (attr->mq_maxmsg > HARD_MSGMAX ||
707 attr->mq_msgsize > HARD_MSGSIZEMAX)
Doug Ledford113289c2012-05-31 16:26:36 -0700708 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 } else {
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700710 if (attr->mq_maxmsg > ipc_ns->mq_msg_max ||
711 attr->mq_msgsize > ipc_ns->mq_msgsize_max)
Doug Ledford113289c2012-05-31 16:26:36 -0700712 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713 }
714 /* check for overflow */
715 if (attr->mq_msgsize > ULONG_MAX/attr->mq_maxmsg)
Doug Ledford113289c2012-05-31 16:26:36 -0700716 return -EOVERFLOW;
Doug Ledford2c12ea42012-05-31 16:26:36 -0700717 mq_treesize = attr->mq_maxmsg * sizeof(struct msg_msg) +
718 min_t(unsigned int, attr->mq_maxmsg, MQ_PRIO_MAX) *
719 sizeof(struct posix_msg_tree_node);
720 total_size = attr->mq_maxmsg * attr->mq_msgsize;
721 if (total_size + mq_treesize < total_size)
Doug Ledford113289c2012-05-31 16:26:36 -0700722 return -EOVERFLOW;
723 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724}
725
726/*
727 * Invoked when creating a new queue via sys_mq_open
728 */
Al Viro765927b2012-06-26 21:58:53 +0400729static struct file *do_create(struct ipc_namespace *ipc_ns, struct inode *dir,
730 struct path *path, int oflag, umode_t mode,
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700731 struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732{
David Howells745ca242008-11-14 10:39:22 +1100733 const struct cred *cred = current_cred();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 int ret;
735
Al Viro564f6992008-12-14 04:02:26 -0500736 if (attr) {
Doug Ledford113289c2012-05-31 16:26:36 -0700737 ret = mq_attr_ok(ipc_ns, attr);
738 if (ret)
Al Viro765927b2012-06-26 21:58:53 +0400739 return ERR_PTR(ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 /* store for use during create */
Al Viro765927b2012-06-26 21:58:53 +0400741 path->dentry->d_fsdata = attr;
Doug Ledford113289c2012-05-31 16:26:36 -0700742 } else {
743 struct mq_attr def_attr;
744
745 def_attr.mq_maxmsg = min(ipc_ns->mq_msg_max,
746 ipc_ns->mq_msg_default);
747 def_attr.mq_msgsize = min(ipc_ns->mq_msgsize_max,
748 ipc_ns->mq_msgsize_default);
749 ret = mq_attr_ok(ipc_ns, &def_attr);
750 if (ret)
Al Viro765927b2012-06-26 21:58:53 +0400751 return ERR_PTR(ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 }
753
Al Viroce3b0f82009-03-29 19:08:22 -0400754 mode &= ~current_umask();
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -0700755 ret = vfs_create2(path->mnt, dir, path->dentry, mode, true);
Al Viro765927b2012-06-26 21:58:53 +0400756 path->dentry->d_fsdata = NULL;
Al Viro312b90f2012-08-06 10:18:17 +0400757 if (ret)
758 return ERR_PTR(ret);
759 return dentry_open(path, oflag, cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760}
761
762/* Opens existing queue */
Al Viro765927b2012-06-26 21:58:53 +0400763static struct file *do_open(struct path *path, int oflag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764{
David Howells745ca242008-11-14 10:39:22 +1100765 static const int oflag2acc[O_ACCMODE] = { MAY_READ, MAY_WRITE,
766 MAY_READ | MAY_WRITE };
Al Viro765927b2012-06-26 21:58:53 +0400767 int acc;
768 if ((oflag & O_ACCMODE) == (O_RDWR | O_WRONLY))
769 return ERR_PTR(-EINVAL);
770 acc = oflag2acc[oflag & O_ACCMODE];
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -0700771 if (inode_permission2(path->mnt, d_inode(path->dentry), acc))
Al Viro765927b2012-06-26 21:58:53 +0400772 return ERR_PTR(-EACCES);
773 return dentry_open(path, oflag, current_cred());
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774}
775
Al Virodf0a4282011-07-26 05:26:10 -0400776SYSCALL_DEFINE4(mq_open, const char __user *, u_name, int, oflag, umode_t, mode,
Heiko Carstensd5460c92009-01-14 14:14:27 +0100777 struct mq_attr __user *, u_attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778{
Al Viro765927b2012-06-26 21:58:53 +0400779 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 struct file *filp;
Jeff Layton91a27b22012-10-10 15:25:28 -0400781 struct filename *name;
Al Viro564f6992008-12-14 04:02:26 -0500782 struct mq_attr attr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 int fd, error;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700784 struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns;
Al Viro312b90f2012-08-06 10:18:17 +0400785 struct vfsmount *mnt = ipc_ns->mq_mnt;
786 struct dentry *root = mnt->mnt_root;
787 int ro;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788
Al Viro564f6992008-12-14 04:02:26 -0500789 if (u_attr && copy_from_user(&attr, u_attr, sizeof(struct mq_attr)))
790 return -EFAULT;
791
792 audit_mq_open(oflag, mode, u_attr ? &attr : NULL);
George C. Wilson20ca73b2006-05-24 16:09:55 -0500793
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 if (IS_ERR(name = getname(u_name)))
795 return PTR_ERR(name);
796
Ulrich Drepper269f2132008-05-03 15:28:45 -0400797 fd = get_unused_fd_flags(O_CLOEXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 if (fd < 0)
799 goto out_putname;
800
Al Viro312b90f2012-08-06 10:18:17 +0400801 ro = mnt_want_write(mnt); /* we'll drop it in any case */
Al Viro765927b2012-06-26 21:58:53 +0400802 error = 0;
Al Viro59551022016-01-22 15:40:57 -0500803 inode_lock(d_inode(root));
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -0700804 path.dentry = lookup_one_len2(name->name, mnt, root, strlen(name->name));
Al Viro765927b2012-06-26 21:58:53 +0400805 if (IS_ERR(path.dentry)) {
806 error = PTR_ERR(path.dentry);
André Goddard Rosa4294a8e2010-02-23 04:04:28 -0300807 goto out_putfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 }
Al Viro312b90f2012-08-06 10:18:17 +0400809 path.mnt = mntget(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810
811 if (oflag & O_CREAT) {
David Howells75c3cfa2015-03-17 22:26:12 +0000812 if (d_really_is_positive(path.dentry)) { /* entry already exists */
Jeff Laytonadb5c242012-10-10 16:43:13 -0400813 audit_inode(name, path.dentry, 0);
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300814 if (oflag & O_EXCL) {
815 error = -EEXIST;
Alexander Viro7c7dce92006-01-14 15:29:55 -0500816 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300817 }
Al Viro765927b2012-06-26 21:58:53 +0400818 filp = do_open(&path, oflag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 } else {
Al Viro312b90f2012-08-06 10:18:17 +0400820 if (ro) {
821 error = ro;
822 goto out;
823 }
Jeff Layton79f65302013-07-08 15:59:36 -0700824 audit_inode_parent_hidden(name, root);
David Howells75c3cfa2015-03-17 22:26:12 +0000825 filp = do_create(ipc_ns, d_inode(root),
Al Viro765927b2012-06-26 21:58:53 +0400826 &path, oflag, mode,
Al Viro564f6992008-12-14 04:02:26 -0500827 u_attr ? &attr : NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 }
Alexander Viro7c7dce92006-01-14 15:29:55 -0500829 } else {
David Howells75c3cfa2015-03-17 22:26:12 +0000830 if (d_really_is_negative(path.dentry)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300831 error = -ENOENT;
Alexander Viro7c7dce92006-01-14 15:29:55 -0500832 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300833 }
Jeff Laytonadb5c242012-10-10 16:43:13 -0400834 audit_inode(name, path.dentry, 0);
Al Viro765927b2012-06-26 21:58:53 +0400835 filp = do_open(&path, oflag);
Alexander Viro7c7dce92006-01-14 15:29:55 -0500836 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837
Al Viro765927b2012-06-26 21:58:53 +0400838 if (!IS_ERR(filp))
839 fd_install(fd, filp);
840 else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 error = PTR_ERR(filp);
Alexander Viro7c7dce92006-01-14 15:29:55 -0500842out:
Al Viro765927b2012-06-26 21:58:53 +0400843 path_put(&path);
Alexander Viro7c7dce92006-01-14 15:29:55 -0500844out_putfd:
Al Viro765927b2012-06-26 21:58:53 +0400845 if (error) {
846 put_unused_fd(fd);
847 fd = error;
848 }
Al Viro59551022016-01-22 15:40:57 -0500849 inode_unlock(d_inode(root));
Vladimir Davydov38d78e52013-03-22 15:04:51 -0700850 if (!ro)
851 mnt_drop_write(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852out_putname:
853 putname(name);
854 return fd;
855}
856
Heiko Carstensd5460c92009-01-14 14:14:27 +0100857SYSCALL_DEFINE1(mq_unlink, const char __user *, u_name)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858{
859 int err;
Jeff Layton91a27b22012-10-10 15:25:28 -0400860 struct filename *name;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 struct dentry *dentry;
862 struct inode *inode = NULL;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700863 struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns;
Al Viro312b90f2012-08-06 10:18:17 +0400864 struct vfsmount *mnt = ipc_ns->mq_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865
866 name = getname(u_name);
867 if (IS_ERR(name))
868 return PTR_ERR(name);
869
Jeff Layton79f65302013-07-08 15:59:36 -0700870 audit_inode_parent_hidden(name, mnt->mnt_root);
Al Viro312b90f2012-08-06 10:18:17 +0400871 err = mnt_want_write(mnt);
872 if (err)
873 goto out_name;
Al Viro59551022016-01-22 15:40:57 -0500874 inode_lock_nested(d_inode(mnt->mnt_root), I_MUTEX_PARENT);
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -0700875 dentry = lookup_one_len2(name->name, mnt, mnt->mnt_root,
Jeff Layton91a27b22012-10-10 15:25:28 -0400876 strlen(name->name));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 if (IS_ERR(dentry)) {
878 err = PTR_ERR(dentry);
879 goto out_unlock;
880 }
881
David Howells75c3cfa2015-03-17 22:26:12 +0000882 inode = d_inode(dentry);
Al Viro312b90f2012-08-06 10:18:17 +0400883 if (!inode) {
884 err = -ENOENT;
885 } else {
Al Viro7de9c6ee2010-10-23 11:11:40 -0400886 ihold(inode);
Daniel Rosenbergbbcd0ff2016-10-26 16:27:45 -0700887 err = vfs_unlink2(mnt, d_inode(dentry->d_parent), dentry, NULL);
Al Viro312b90f2012-08-06 10:18:17 +0400888 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 dput(dentry);
890
891out_unlock:
Al Viro59551022016-01-22 15:40:57 -0500892 inode_unlock(d_inode(mnt->mnt_root));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 if (inode)
894 iput(inode);
Al Viro312b90f2012-08-06 10:18:17 +0400895 mnt_drop_write(mnt);
896out_name:
897 putname(name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898
899 return err;
900}
901
902/* Pipelined send and receive functions.
903 *
904 * If a receiver finds no waiting message, then it registers itself in the
905 * list of waiting receivers. A sender checks that list before adding the new
906 * message into the message array. If there is a waiting receiver, then it
907 * bypasses the message array and directly hands the message over to the
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700908 * receiver. The receiver accepts the message and returns without grabbing the
909 * queue spinlock:
910 *
911 * - Set pointer to message.
912 * - Queue the receiver task for later wakeup (without the info->lock).
913 * - Update its state to STATE_READY. Now the receiver can continue.
914 * - Wake up the process after the lock is dropped. Should the process wake up
915 * before this wakeup (due to a timeout or a signal) it will either see
916 * STATE_READY and continue or acquire the lock to check the state again.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917 *
918 * The same algorithm is used for senders.
919 */
920
921/* pipelined_send() - send a message directly to the task waiting in
922 * sys_mq_timedreceive() (without inserting message into a queue).
923 */
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700924static inline void pipelined_send(struct wake_q_head *wake_q,
925 struct mqueue_inode_info *info,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 struct msg_msg *message,
927 struct ext_wait_queue *receiver)
928{
929 receiver->msg = message;
930 list_del(&receiver->list);
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700931 wake_q_add(wake_q, receiver->task);
932 /*
933 * Rely on the implicit cmpxchg barrier from wake_q_add such
934 * that we can ensure that updating receiver->state is the last
935 * write operation: As once set, the receiver can continue,
936 * and if we don't have the reference count from the wake_q,
937 * yet, at that point we can later have a use-after-free
938 * condition and bogus wakeup.
939 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 receiver->state = STATE_READY;
941}
942
943/* pipelined_receive() - if there is task waiting in sys_mq_timedsend()
944 * gets its message and put to the queue (we have one free place for sure). */
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700945static inline void pipelined_receive(struct wake_q_head *wake_q,
946 struct mqueue_inode_info *info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947{
948 struct ext_wait_queue *sender = wq_get_first_waiter(info, SEND);
949
950 if (!sender) {
951 /* for poll */
952 wake_up_interruptible(&info->wait_q);
953 return;
954 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700955 if (msg_insert(sender->msg, info))
956 return;
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700957
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 list_del(&sender->list);
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700959 wake_q_add(wake_q, sender->task);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960 sender->state = STATE_READY;
961}
962
Heiko Carstensc4ea37c2009-01-14 14:14:28 +0100963SYSCALL_DEFINE5(mq_timedsend, mqd_t, mqdes, const char __user *, u_msg_ptr,
964 size_t, msg_len, unsigned int, msg_prio,
965 const struct timespec __user *, u_abs_timeout)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966{
Al Viro2903ff02012-08-28 12:52:22 -0400967 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 struct inode *inode;
969 struct ext_wait_queue wait;
970 struct ext_wait_queue *receiver;
971 struct msg_msg *msg_ptr;
972 struct mqueue_inode_info *info;
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200973 ktime_t expires, *timeout = NULL;
974 struct timespec ts;
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700975 struct posix_msg_tree_node *new_leaf = NULL;
Al Viro2903ff02012-08-28 12:52:22 -0400976 int ret = 0;
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700977 WAKE_Q(wake_q);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978
Al Viroc32c8af2008-12-14 03:46:48 -0500979 if (u_abs_timeout) {
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200980 int res = prepare_timeout(u_abs_timeout, &expires, &ts);
981 if (res)
982 return res;
983 timeout = &expires;
Al Viroc32c8af2008-12-14 03:46:48 -0500984 }
George C. Wilson20ca73b2006-05-24 16:09:55 -0500985
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 if (unlikely(msg_prio >= (unsigned long) MQ_PRIO_MAX))
987 return -EINVAL;
988
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200989 audit_mq_sendrecv(mqdes, msg_len, msg_prio, timeout ? &ts : NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990
Al Viro2903ff02012-08-28 12:52:22 -0400991 f = fdget(mqdes);
992 if (unlikely(!f.file)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300993 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300995 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996
Al Viro496ad9a2013-01-23 17:07:38 -0500997 inode = file_inode(f.file);
Al Viro2903ff02012-08-28 12:52:22 -0400998 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300999 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001001 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 info = MQUEUE_I(inode);
Al Viro9f45f5b2014-10-31 17:44:57 -04001003 audit_file(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004
Al Viro2903ff02012-08-28 12:52:22 -04001005 if (unlikely(!(f.file->f_mode & FMODE_WRITE))) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001006 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
1010 if (unlikely(msg_len > info->attr.mq_msgsize)) {
1011 ret = -EMSGSIZE;
1012 goto out_fput;
1013 }
1014
1015 /* First try to allocate memory, before doing anything with
1016 * existing queues. */
1017 msg_ptr = load_msg(u_msg_ptr, msg_len);
1018 if (IS_ERR(msg_ptr)) {
1019 ret = PTR_ERR(msg_ptr);
1020 goto out_fput;
1021 }
1022 msg_ptr->m_ts = msg_len;
1023 msg_ptr->m_type = msg_prio;
1024
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001025 /*
1026 * msg_insert really wants us to have a valid, spare node struct so
1027 * it doesn't have to kmalloc a GFP_ATOMIC allocation, but it will
1028 * fall back to that if necessary.
1029 */
1030 if (!info->node_cache)
1031 new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL);
1032
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 spin_lock(&info->lock);
1034
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001035 if (!info->node_cache && new_leaf) {
1036 /* Save our speculative allocation into the cache */
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001037 INIT_LIST_HEAD(&new_leaf->msg_list);
1038 info->node_cache = new_leaf;
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001039 new_leaf = NULL;
1040 } else {
1041 kfree(new_leaf);
1042 }
1043
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 if (info->attr.mq_curmsgs == info->attr.mq_maxmsg) {
Al Viro2903ff02012-08-28 12:52:22 -04001045 if (f.file->f_flags & O_NONBLOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046 ret = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 } else {
1048 wait.task = current;
1049 wait.msg = (void *) msg_ptr;
1050 wait.state = STATE_NONE;
1051 ret = wq_sleep(info, SEND, timeout, &wait);
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001052 /*
1053 * wq_sleep must be called with info->lock held, and
1054 * returns with the lock released
1055 */
1056 goto out_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 } else {
1059 receiver = wq_get_first_waiter(info, RECV);
1060 if (receiver) {
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001061 pipelined_send(&wake_q, info, msg_ptr, receiver);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062 } else {
1063 /* adds message to the queue */
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001064 ret = msg_insert(msg_ptr, info);
1065 if (ret)
1066 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 __do_notify(info);
1068 }
1069 inode->i_atime = inode->i_mtime = inode->i_ctime =
Deepa Dinamani078cd822016-09-14 07:48:04 -07001070 current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 }
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001072out_unlock:
1073 spin_unlock(&info->lock);
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001074 wake_up_q(&wake_q);
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001075out_free:
1076 if (ret)
1077 free_msg(msg_ptr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04001079 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080out:
1081 return ret;
1082}
1083
Heiko Carstensc4ea37c2009-01-14 14:14:28 +01001084SYSCALL_DEFINE5(mq_timedreceive, mqd_t, mqdes, char __user *, u_msg_ptr,
1085 size_t, msg_len, unsigned int __user *, u_msg_prio,
1086 const struct timespec __user *, u_abs_timeout)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 ssize_t ret;
1089 struct msg_msg *msg_ptr;
Al Viro2903ff02012-08-28 12:52:22 -04001090 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 struct inode *inode;
1092 struct mqueue_inode_info *info;
1093 struct ext_wait_queue wait;
Carsten Emde9ca7d8e2010-04-02 22:40:20 +02001094 ktime_t expires, *timeout = NULL;
1095 struct timespec ts;
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001096 struct posix_msg_tree_node *new_leaf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097
Al Viroc32c8af2008-12-14 03:46:48 -05001098 if (u_abs_timeout) {
Carsten Emde9ca7d8e2010-04-02 22:40:20 +02001099 int res = prepare_timeout(u_abs_timeout, &expires, &ts);
1100 if (res)
1101 return res;
1102 timeout = &expires;
Al Viroc32c8af2008-12-14 03:46:48 -05001103 }
George C. Wilson20ca73b2006-05-24 16:09:55 -05001104
Carsten Emde9ca7d8e2010-04-02 22:40:20 +02001105 audit_mq_sendrecv(mqdes, msg_len, 0, timeout ? &ts : NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106
Al Viro2903ff02012-08-28 12:52:22 -04001107 f = fdget(mqdes);
1108 if (unlikely(!f.file)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001109 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001111 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112
Al Viro496ad9a2013-01-23 17:07:38 -05001113 inode = file_inode(f.file);
Al Viro2903ff02012-08-28 12:52:22 -04001114 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001115 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001117 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 info = MQUEUE_I(inode);
Al Viro9f45f5b2014-10-31 17:44:57 -04001119 audit_file(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120
Al Viro2903ff02012-08-28 12:52:22 -04001121 if (unlikely(!(f.file->f_mode & FMODE_READ))) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001122 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001124 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125
1126 /* checks if buffer is big enough */
1127 if (unlikely(msg_len < info->attr.mq_msgsize)) {
1128 ret = -EMSGSIZE;
1129 goto out_fput;
1130 }
1131
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001132 /*
1133 * msg_insert really wants us to have a valid, spare node struct so
1134 * it doesn't have to kmalloc a GFP_ATOMIC allocation, but it will
1135 * fall back to that if necessary.
1136 */
1137 if (!info->node_cache)
1138 new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL);
1139
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 spin_lock(&info->lock);
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001141
1142 if (!info->node_cache && new_leaf) {
1143 /* Save our speculative allocation into the cache */
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001144 INIT_LIST_HEAD(&new_leaf->msg_list);
1145 info->node_cache = new_leaf;
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001146 } else {
1147 kfree(new_leaf);
1148 }
1149
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150 if (info->attr.mq_curmsgs == 0) {
Al Viro2903ff02012-08-28 12:52:22 -04001151 if (f.file->f_flags & O_NONBLOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 spin_unlock(&info->lock);
1153 ret = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154 } else {
1155 wait.task = current;
1156 wait.state = STATE_NONE;
1157 ret = wq_sleep(info, RECV, timeout, &wait);
1158 msg_ptr = wait.msg;
1159 }
1160 } else {
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001161 WAKE_Q(wake_q);
1162
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163 msg_ptr = msg_get(info);
1164
1165 inode->i_atime = inode->i_mtime = inode->i_ctime =
Deepa Dinamani078cd822016-09-14 07:48:04 -07001166 current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167
1168 /* There is now free space in queue. */
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001169 pipelined_receive(&wake_q, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170 spin_unlock(&info->lock);
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001171 wake_up_q(&wake_q);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 ret = 0;
1173 }
1174 if (ret == 0) {
1175 ret = msg_ptr->m_ts;
1176
1177 if ((u_msg_prio && put_user(msg_ptr->m_type, u_msg_prio)) ||
1178 store_msg(u_msg_ptr, msg_ptr, msg_ptr->m_ts)) {
1179 ret = -EFAULT;
1180 }
1181 free_msg(msg_ptr);
1182 }
1183out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04001184 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185out:
1186 return ret;
1187}
1188
1189/*
1190 * Notes: the case when user wants us to deregister (with NULL as pointer)
1191 * and he isn't currently owner of notification, will be silently discarded.
1192 * It isn't explicitly defined in the POSIX.
1193 */
Heiko Carstensc4ea37c2009-01-14 14:14:28 +01001194SYSCALL_DEFINE2(mq_notify, mqd_t, mqdes,
1195 const struct sigevent __user *, u_notification)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196{
Al Viro2903ff02012-08-28 12:52:22 -04001197 int ret;
1198 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 struct sock *sock;
1200 struct inode *inode;
1201 struct sigevent notification;
1202 struct mqueue_inode_info *info;
1203 struct sk_buff *nc;
1204
Al Viro20114f72008-12-10 07:16:12 -05001205 if (u_notification) {
1206 if (copy_from_user(&notification, u_notification,
1207 sizeof(struct sigevent)))
1208 return -EFAULT;
1209 }
1210
1211 audit_mq_notify(mqdes, u_notification ? &notification : NULL);
George C. Wilson20ca73b2006-05-24 16:09:55 -05001212
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 nc = NULL;
1214 sock = NULL;
1215 if (u_notification != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 if (unlikely(notification.sigev_notify != SIGEV_NONE &&
1217 notification.sigev_notify != SIGEV_SIGNAL &&
1218 notification.sigev_notify != SIGEV_THREAD))
1219 return -EINVAL;
1220 if (notification.sigev_notify == SIGEV_SIGNAL &&
Jesper Juhl7ed20e12005-05-01 08:59:14 -07001221 !valid_signal(notification.sigev_signo)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222 return -EINVAL;
1223 }
1224 if (notification.sigev_notify == SIGEV_THREAD) {
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001225 long timeo;
1226
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 /* create the notify skb */
1228 nc = alloc_skb(NOTIFY_COOKIE_LEN, GFP_KERNEL);
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001229 if (!nc) {
1230 ret = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001232 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 if (copy_from_user(nc->data,
1234 notification.sigev_value.sival_ptr,
1235 NOTIFY_COOKIE_LEN)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001236 ret = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 goto out;
1238 }
1239
1240 /* TODO: add a header? */
1241 skb_put(nc, NOTIFY_COOKIE_LEN);
1242 /* and attach it to the socket */
1243retry:
Al Viro2903ff02012-08-28 12:52:22 -04001244 f = fdget(notification.sigev_signo);
1245 if (!f.file) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001246 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001248 }
Al Viro2903ff02012-08-28 12:52:22 -04001249 sock = netlink_getsockbyfilp(f.file);
1250 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 if (IS_ERR(sock)) {
1252 ret = PTR_ERR(sock);
1253 sock = NULL;
1254 goto out;
1255 }
1256
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001257 timeo = MAX_SCHEDULE_TIMEOUT;
Denis V. Lunev9457afe2008-06-05 11:23:39 -07001258 ret = netlink_attachskb(sock, nc, &timeo, NULL);
Cong Wange6952842017-07-09 13:19:55 -07001259 if (ret == 1) {
1260 sock = NULL;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001261 goto retry;
Cong Wange6952842017-07-09 13:19:55 -07001262 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 if (ret) {
1264 sock = NULL;
1265 nc = NULL;
1266 goto out;
1267 }
1268 }
1269 }
1270
Al Viro2903ff02012-08-28 12:52:22 -04001271 f = fdget(mqdes);
1272 if (!f.file) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001273 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001275 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276
Al Viro496ad9a2013-01-23 17:07:38 -05001277 inode = file_inode(f.file);
Al Viro2903ff02012-08-28 12:52:22 -04001278 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001279 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001281 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282 info = MQUEUE_I(inode);
1283
1284 ret = 0;
1285 spin_lock(&info->lock);
1286 if (u_notification == NULL) {
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001287 if (info->notify_owner == task_tgid(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288 remove_notification(info);
Deepa Dinamani078cd822016-09-14 07:48:04 -07001289 inode->i_atime = inode->i_ctime = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001291 } else if (info->notify_owner != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 ret = -EBUSY;
1293 } else {
1294 switch (notification.sigev_notify) {
1295 case SIGEV_NONE:
1296 info->notify.sigev_notify = SIGEV_NONE;
1297 break;
1298 case SIGEV_THREAD:
1299 info->notify_sock = sock;
1300 info->notify_cookie = nc;
1301 sock = NULL;
1302 nc = NULL;
1303 info->notify.sigev_notify = SIGEV_THREAD;
1304 break;
1305 case SIGEV_SIGNAL:
1306 info->notify.sigev_signo = notification.sigev_signo;
1307 info->notify.sigev_value = notification.sigev_value;
1308 info->notify.sigev_notify = SIGEV_SIGNAL;
1309 break;
1310 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001311
1312 info->notify_owner = get_pid(task_tgid(current));
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -08001313 info->notify_user_ns = get_user_ns(current_user_ns());
Deepa Dinamani078cd822016-09-14 07:48:04 -07001314 inode->i_atime = inode->i_ctime = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315 }
1316 spin_unlock(&info->lock);
1317out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04001318 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319out:
Davidlohr Bueso3ab08fe2014-01-27 17:07:06 -08001320 if (sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 netlink_detachskb(sock, nc);
Davidlohr Bueso3ab08fe2014-01-27 17:07:06 -08001322 else if (nc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 dev_kfree_skb(nc);
Davidlohr Bueso3ab08fe2014-01-27 17:07:06 -08001324
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 return ret;
1326}
1327
Heiko Carstensc4ea37c2009-01-14 14:14:28 +01001328SYSCALL_DEFINE3(mq_getsetattr, mqd_t, mqdes,
1329 const struct mq_attr __user *, u_mqstat,
1330 struct mq_attr __user *, u_omqstat)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331{
1332 int ret;
1333 struct mq_attr mqstat, omqstat;
Al Viro2903ff02012-08-28 12:52:22 -04001334 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 struct inode *inode;
1336 struct mqueue_inode_info *info;
1337
1338 if (u_mqstat != NULL) {
1339 if (copy_from_user(&mqstat, u_mqstat, sizeof(struct mq_attr)))
1340 return -EFAULT;
1341 if (mqstat.mq_flags & (~O_NONBLOCK))
1342 return -EINVAL;
1343 }
1344
Al Viro2903ff02012-08-28 12:52:22 -04001345 f = fdget(mqdes);
1346 if (!f.file) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001347 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001349 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350
Al Viro496ad9a2013-01-23 17:07:38 -05001351 inode = file_inode(f.file);
Al Viro2903ff02012-08-28 12:52:22 -04001352 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001353 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001355 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 info = MQUEUE_I(inode);
1357
1358 spin_lock(&info->lock);
1359
1360 omqstat = info->attr;
Al Viro2903ff02012-08-28 12:52:22 -04001361 omqstat.mq_flags = f.file->f_flags & O_NONBLOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 if (u_mqstat) {
Al Viro73929062008-12-10 06:58:59 -05001363 audit_mq_getsetattr(mqdes, &mqstat);
Al Viro2903ff02012-08-28 12:52:22 -04001364 spin_lock(&f.file->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 if (mqstat.mq_flags & O_NONBLOCK)
Al Viro2903ff02012-08-28 12:52:22 -04001366 f.file->f_flags |= O_NONBLOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367 else
Al Viro2903ff02012-08-28 12:52:22 -04001368 f.file->f_flags &= ~O_NONBLOCK;
1369 spin_unlock(&f.file->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370
Deepa Dinamani078cd822016-09-14 07:48:04 -07001371 inode->i_atime = inode->i_ctime = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 }
1373
1374 spin_unlock(&info->lock);
1375
1376 ret = 0;
1377 if (u_omqstat != NULL && copy_to_user(u_omqstat, &omqstat,
1378 sizeof(struct mq_attr)))
1379 ret = -EFAULT;
1380
1381out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04001382 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383out:
1384 return ret;
1385}
1386
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -08001387static const struct inode_operations mqueue_dir_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 .lookup = simple_lookup,
1389 .create = mqueue_create,
1390 .unlink = mqueue_unlink,
1391};
1392
Arjan van de Ven9a321442007-02-12 00:55:35 -08001393static const struct file_operations mqueue_file_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 .flush = mqueue_flush_file,
1395 .poll = mqueue_poll_file,
1396 .read = mqueue_read_file,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001397 .llseek = default_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398};
1399
Alexey Dobriyanb87221d2009-09-21 17:01:09 -07001400static const struct super_operations mqueue_super_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 .alloc_inode = mqueue_alloc_inode,
1402 .destroy_inode = mqueue_destroy_inode,
Al Viro6d8af642010-06-05 16:29:45 -04001403 .evict_inode = mqueue_evict_inode,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405};
1406
1407static struct file_system_type mqueue_fs_type = {
1408 .name = "mqueue",
Al Viroceefda62010-07-26 13:16:50 +04001409 .mount = mqueue_mount,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410 .kill_sb = kill_litter_super,
Gao fengbc1b69e2013-01-28 11:09:01 +08001411 .fs_flags = FS_USERNS_MOUNT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412};
1413
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001414int mq_init_ns(struct ipc_namespace *ns)
1415{
1416 ns->mq_queues_count = 0;
1417 ns->mq_queues_max = DFLT_QUEUESMAX;
1418 ns->mq_msg_max = DFLT_MSGMAX;
1419 ns->mq_msgsize_max = DFLT_MSGSIZEMAX;
KOSAKI Motohirocef01842012-05-31 16:26:33 -07001420 ns->mq_msg_default = DFLT_MSG;
1421 ns->mq_msgsize_default = DFLT_MSGSIZE;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001422
1423 ns->mq_mnt = kern_mount_data(&mqueue_fs_type, ns);
1424 if (IS_ERR(ns->mq_mnt)) {
1425 int err = PTR_ERR(ns->mq_mnt);
1426 ns->mq_mnt = NULL;
1427 return err;
1428 }
1429 return 0;
1430}
1431
1432void mq_clear_sbinfo(struct ipc_namespace *ns)
1433{
1434 ns->mq_mnt->mnt_sb->s_fs_info = NULL;
1435}
1436
1437void mq_put_mnt(struct ipc_namespace *ns)
1438{
Al Viro6f686572011-12-09 00:38:50 -05001439 kern_unmount(ns->mq_mnt);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001440}
1441
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442static int __init init_mqueue_fs(void)
1443{
1444 int error;
1445
1446 mqueue_inode_cachep = kmem_cache_create("mqueue_inode_cache",
1447 sizeof(struct mqueue_inode_info), 0,
Vladimir Davydov5d097052016-01-14 15:18:21 -08001448 SLAB_HWCACHE_ALIGN|SLAB_ACCOUNT, init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 if (mqueue_inode_cachep == NULL)
1450 return -ENOMEM;
1451
André Goddard Rosa2329e392010-02-23 04:04:27 -03001452 /* ignore failures - they are not fatal */
Serge E. Hallynbdc8e5f2009-04-06 19:01:11 -07001453 mq_sysctl_table = mq_register_sysctl_table();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454
1455 error = register_filesystem(&mqueue_fs_type);
1456 if (error)
1457 goto out_sysctl;
1458
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001459 spin_lock_init(&mq_lock);
1460
Al Viro6f686572011-12-09 00:38:50 -05001461 error = mq_init_ns(&init_ipc_ns);
1462 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463 goto out_filesystem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465 return 0;
1466
1467out_filesystem:
1468 unregister_filesystem(&mqueue_fs_type);
1469out_sysctl:
1470 if (mq_sysctl_table)
1471 unregister_sysctl_table(mq_sysctl_table);
Alexey Dobriyan1a1d92c2006-09-27 01:49:40 -07001472 kmem_cache_destroy(mqueue_inode_cachep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473 return error;
1474}
1475
Davidlohr Bueso6d08a252014-04-07 15:39:18 -07001476device_initcall(init_mqueue_fs);