blob: 8031464ed4ae24e2611876a520d8232d0c0eacd8 [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>
David Howells935c69122018-11-01 23:07:25 +000021#include <linux/fs_context.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/namei.h>
23#include <linux/sysctl.h>
24#include <linux/poll.h>
25#include <linux/mqueue.h>
26#include <linux/msg.h>
27#include <linux/skbuff.h>
Doug Ledford5b5c4d12012-05-31 16:26:30 -070028#include <linux/vmalloc.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/netlink.h>
30#include <linux/syscalls.h>
George C. Wilson20ca73b2006-05-24 16:09:55 -050031#include <linux/audit.h>
Jesper Juhl7ed20e12005-05-01 08:59:14 -070032#include <linux/signal.h>
Ingo Molnar5f921ae2006-03-26 01:37:17 -080033#include <linux/mutex.h>
Pavel Emelyanovb4888932007-10-18 23:40:14 -070034#include <linux/nsproxy.h>
35#include <linux/pid.h>
Serge E. Hallyn614b84c2009-04-06 19:01:08 -070036#include <linux/ipc_namespace.h>
Serge E. Hallyn6b550f92012-01-10 15:11:37 -080037#include <linux/user_namespace.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090038#include <linux/slab.h>
Ingo Molnar84f001e2017-02-01 16:36:40 +010039#include <linux/sched/wake_q.h>
Ingo Molnar3f07c012017-02-08 18:51:30 +010040#include <linux/sched/signal.h>
Ingo Molnar8703e8a2017-02-08 18:51:30 +010041#include <linux/sched/user.h>
Ingo Molnar5f921ae2006-03-26 01:37:17 -080042
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#include <net/sock.h>
44#include "util.h"
45
David Howells935c69122018-11-01 23:07:25 +000046struct mqueue_fs_context {
47 struct ipc_namespace *ipc_ns;
48};
49
Linus Torvalds1da177e2005-04-16 15:20:36 -070050#define MQUEUE_MAGIC 0x19800202
51#define DIRENT_SIZE 20
52#define FILENT_SIZE 80
53
54#define SEND 0
55#define RECV 1
56
57#define STATE_NONE 0
Davidlohr Buesofa6004a2015-05-04 07:02:46 -070058#define STATE_READY 1
Linus Torvalds1da177e2005-04-16 15:20:36 -070059
Doug Ledfordd6629852012-05-31 16:26:35 -070060struct posix_msg_tree_node {
61 struct rb_node rb_node;
62 struct list_head msg_list;
63 int priority;
64};
65
Manfred Spraulc5b2cbd2020-02-03 17:34:36 -080066/*
67 * Locking:
68 *
69 * Accesses to a message queue are synchronized by acquiring info->lock.
70 *
71 * There are two notable exceptions:
72 * - The actual wakeup of a sleeping task is performed using the wake_q
73 * framework. info->lock is already released when wake_up_q is called.
74 * - The exit codepaths after sleeping check ext_wait_queue->state without
75 * any locks. If it is STATE_READY, then the syscall is completed without
76 * acquiring info->lock.
77 *
78 * MQ_BARRIER:
79 * To achieve proper release/acquire memory barrier pairing, the state is set to
80 * STATE_READY with smp_store_release(), and it is read with READ_ONCE followed
81 * by smp_acquire__after_ctrl_dep(). In addition, wake_q_add_safe() is used.
82 *
83 * This prevents the following races:
84 *
85 * 1) With the simple wake_q_add(), the task could be gone already before
86 * the increase of the reference happens
87 * Thread A
88 * Thread B
89 * WRITE_ONCE(wait.state, STATE_NONE);
90 * schedule_hrtimeout()
91 * wake_q_add(A)
92 * if (cmpxchg()) // success
93 * ->state = STATE_READY (reordered)
94 * <timeout returns>
95 * if (wait.state == STATE_READY) return;
96 * sysret to user space
97 * sys_exit()
98 * get_task_struct() // UaF
99 *
100 * Solution: Use wake_q_add_safe() and perform the get_task_struct() before
101 * the smp_store_release() that does ->state = STATE_READY.
102 *
103 * 2) Without proper _release/_acquire barriers, the woken up task
104 * could read stale data
105 *
106 * Thread A
107 * Thread B
108 * do_mq_timedreceive
109 * WRITE_ONCE(wait.state, STATE_NONE);
110 * schedule_hrtimeout()
111 * state = STATE_READY;
112 * <timeout returns>
113 * if (wait.state == STATE_READY) return;
114 * msg_ptr = wait.msg; // Access to stale data!
115 * receiver->msg = message; (reordered)
116 *
117 * Solution: use _release and _acquire barriers.
118 *
119 * 3) There is intentionally no barrier when setting current->state
120 * to TASK_INTERRUPTIBLE: spin_unlock(&info->lock) provides the
121 * release memory barrier, and the wakeup is triggered when holding
122 * info->lock, i.e. spin_lock(&info->lock) provided a pairing
123 * acquire memory barrier.
124 */
125
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126struct ext_wait_queue { /* queue of sleeping tasks */
127 struct task_struct *task;
128 struct list_head list;
129 struct msg_msg *msg; /* ptr of loaded message */
130 int state; /* one of STATE_* values */
131};
132
133struct mqueue_inode_info {
134 spinlock_t lock;
135 struct inode vfs_inode;
136 wait_queue_head_t wait_q;
137
Doug Ledfordd6629852012-05-31 16:26:35 -0700138 struct rb_root msg_tree;
Davidlohr Buesoa5091fd2019-05-14 15:46:26 -0700139 struct rb_node *msg_tree_rightmost;
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700140 struct posix_msg_tree_node *node_cache;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141 struct mq_attr attr;
142
143 struct sigevent notify;
Manfred Spraul239521f2014-01-27 17:07:04 -0800144 struct pid *notify_owner;
Oleg Nesterovb5f20062020-05-07 18:35:39 -0700145 u32 notify_self_exec_id;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800146 struct user_namespace *notify_user_ns;
Adrian Bunk338cec32005-09-10 00:26:54 -0700147 struct user_struct *user; /* user who created, for accounting */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148 struct sock *notify_sock;
149 struct sk_buff *notify_cookie;
150
151 /* for tasks waiting for free space and messages, respectively */
152 struct ext_wait_queue e_wait_q[2];
153
154 unsigned long qsize; /* size of queue in memory (sum of all msgs) */
155};
156
David Howells935c69122018-11-01 23:07:25 +0000157static struct file_system_type mqueue_fs_type;
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -0800158static const struct inode_operations mqueue_dir_inode_operations;
Arjan van de Ven9a321442007-02-12 00:55:35 -0800159static const struct file_operations mqueue_file_operations;
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700160static const struct super_operations mqueue_super_ops;
David Howells935c69122018-11-01 23:07:25 +0000161static const struct fs_context_operations mqueue_fs_context_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162static void remove_notification(struct mqueue_inode_info *info);
163
Christoph Lametere18b8902006-12-06 20:33:20 -0800164static struct kmem_cache *mqueue_inode_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165
Manfred Spraul239521f2014-01-27 17:07:04 -0800166static struct ctl_table_header *mq_sysctl_table;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167
168static inline struct mqueue_inode_info *MQUEUE_I(struct inode *inode)
169{
170 return container_of(inode, struct mqueue_inode_info, vfs_inode);
171}
172
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700173/*
174 * This routine should be called with the mq_lock held.
175 */
176static inline struct ipc_namespace *__get_ns_from_inode(struct inode *inode)
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700177{
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700178 return get_ipc_ns(inode->i_sb->s_fs_info);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700179}
180
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700181static struct ipc_namespace *get_ns_from_inode(struct inode *inode)
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700182{
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700183 struct ipc_namespace *ns;
184
185 spin_lock(&mq_lock);
186 ns = __get_ns_from_inode(inode);
187 spin_unlock(&mq_lock);
188 return ns;
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700189}
190
Doug Ledfordd6629852012-05-31 16:26:35 -0700191/* Auxiliary functions to manipulate messages' list */
192static int msg_insert(struct msg_msg *msg, struct mqueue_inode_info *info)
193{
194 struct rb_node **p, *parent = NULL;
195 struct posix_msg_tree_node *leaf;
Davidlohr Buesoa5091fd2019-05-14 15:46:26 -0700196 bool rightmost = true;
Doug Ledfordd6629852012-05-31 16:26:35 -0700197
198 p = &info->msg_tree.rb_node;
199 while (*p) {
200 parent = *p;
201 leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node);
202
203 if (likely(leaf->priority == msg->m_type))
204 goto insert_msg;
Davidlohr Buesoa5091fd2019-05-14 15:46:26 -0700205 else if (msg->m_type < leaf->priority) {
Doug Ledfordd6629852012-05-31 16:26:35 -0700206 p = &(*p)->rb_left;
Davidlohr Buesoa5091fd2019-05-14 15:46:26 -0700207 rightmost = false;
208 } else
Doug Ledfordd6629852012-05-31 16:26:35 -0700209 p = &(*p)->rb_right;
210 }
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700211 if (info->node_cache) {
212 leaf = info->node_cache;
213 info->node_cache = NULL;
214 } else {
215 leaf = kmalloc(sizeof(*leaf), GFP_ATOMIC);
216 if (!leaf)
217 return -ENOMEM;
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700218 INIT_LIST_HEAD(&leaf->msg_list);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700219 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700220 leaf->priority = msg->m_type;
Davidlohr Buesoa5091fd2019-05-14 15:46:26 -0700221
222 if (rightmost)
223 info->msg_tree_rightmost = &leaf->rb_node;
224
Doug Ledfordd6629852012-05-31 16:26:35 -0700225 rb_link_node(&leaf->rb_node, parent, p);
226 rb_insert_color(&leaf->rb_node, &info->msg_tree);
Doug Ledfordd6629852012-05-31 16:26:35 -0700227insert_msg:
228 info->attr.mq_curmsgs++;
229 info->qsize += msg->m_ts;
230 list_add_tail(&msg->m_list, &leaf->msg_list);
231 return 0;
232}
233
Davidlohr Buesoa5091fd2019-05-14 15:46:26 -0700234static inline void msg_tree_erase(struct posix_msg_tree_node *leaf,
235 struct mqueue_inode_info *info)
236{
237 struct rb_node *node = &leaf->rb_node;
238
239 if (info->msg_tree_rightmost == node)
240 info->msg_tree_rightmost = rb_prev(node);
241
242 rb_erase(node, &info->msg_tree);
Somala Swaraj43afe4d2020-04-06 20:12:53 -0700243 if (info->node_cache)
Davidlohr Buesoa5091fd2019-05-14 15:46:26 -0700244 kfree(leaf);
Somala Swaraj43afe4d2020-04-06 20:12:53 -0700245 else
Davidlohr Buesoa5091fd2019-05-14 15:46:26 -0700246 info->node_cache = leaf;
Davidlohr Buesoa5091fd2019-05-14 15:46:26 -0700247}
248
Doug Ledfordd6629852012-05-31 16:26:35 -0700249static inline struct msg_msg *msg_get(struct mqueue_inode_info *info)
250{
Davidlohr Buesoa5091fd2019-05-14 15:46:26 -0700251 struct rb_node *parent = NULL;
Doug Ledfordd6629852012-05-31 16:26:35 -0700252 struct posix_msg_tree_node *leaf;
253 struct msg_msg *msg;
254
255try_again:
Davidlohr Buesoa5091fd2019-05-14 15:46:26 -0700256 /*
257 * During insert, low priorities go to the left and high to the
258 * right. On receive, we want the highest priorities first, so
259 * walk all the way to the right.
260 */
261 parent = info->msg_tree_rightmost;
Doug Ledfordd6629852012-05-31 16:26:35 -0700262 if (!parent) {
263 if (info->attr.mq_curmsgs) {
264 pr_warn_once("Inconsistency in POSIX message queue, "
265 "no tree element, but supposedly messages "
266 "should exist!\n");
267 info->attr.mq_curmsgs = 0;
268 }
269 return NULL;
270 }
271 leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700272 if (unlikely(list_empty(&leaf->msg_list))) {
Doug Ledfordd6629852012-05-31 16:26:35 -0700273 pr_warn_once("Inconsistency in POSIX message queue, "
274 "empty leaf node but we haven't implemented "
275 "lazy leaf delete!\n");
Davidlohr Buesoa5091fd2019-05-14 15:46:26 -0700276 msg_tree_erase(leaf, info);
Doug Ledfordd6629852012-05-31 16:26:35 -0700277 goto try_again;
278 } else {
279 msg = list_first_entry(&leaf->msg_list,
280 struct msg_msg, m_list);
281 list_del(&msg->m_list);
282 if (list_empty(&leaf->msg_list)) {
Davidlohr Buesoa5091fd2019-05-14 15:46:26 -0700283 msg_tree_erase(leaf, info);
Doug Ledfordd6629852012-05-31 16:26:35 -0700284 }
285 }
286 info->attr.mq_curmsgs--;
287 info->qsize -= msg->m_ts;
288 return msg;
289}
290
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700291static struct inode *mqueue_get_inode(struct super_block *sb,
Al Viro1b9d5ff72011-07-24 14:18:20 -0400292 struct ipc_namespace *ipc_ns, umode_t mode,
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700293 struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294{
David Howells86a264a2008-11-14 10:39:18 +1100295 struct user_struct *u = current_user();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296 struct inode *inode;
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700297 int ret = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298
299 inode = new_inode(sb);
Jiri Slaby04715202011-07-26 16:08:46 -0700300 if (!inode)
301 goto err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302
Jiri Slaby04715202011-07-26 16:08:46 -0700303 inode->i_ino = get_next_ino();
304 inode->i_mode = mode;
305 inode->i_uid = current_fsuid();
306 inode->i_gid = current_fsgid();
Deepa Dinamani078cd822016-09-14 07:48:04 -0700307 inode->i_mtime = inode->i_ctime = inode->i_atime = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
Jiri Slaby04715202011-07-26 16:08:46 -0700309 if (S_ISREG(mode)) {
310 struct mqueue_inode_info *info;
Doug Ledfordd6629852012-05-31 16:26:35 -0700311 unsigned long mq_bytes, mq_treesize;
André Goddard Rosac8308b12010-02-23 04:04:23 -0300312
Jiri Slaby04715202011-07-26 16:08:46 -0700313 inode->i_fop = &mqueue_file_operations;
314 inode->i_size = FILENT_SIZE;
315 /* mqueue specific info */
316 info = MQUEUE_I(inode);
317 spin_lock_init(&info->lock);
318 init_waitqueue_head(&info->wait_q);
319 INIT_LIST_HEAD(&info->e_wait_q[0].list);
320 INIT_LIST_HEAD(&info->e_wait_q[1].list);
321 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800322 info->notify_user_ns = NULL;
Jiri Slaby04715202011-07-26 16:08:46 -0700323 info->qsize = 0;
324 info->user = NULL; /* set when all is ok */
Doug Ledfordd6629852012-05-31 16:26:35 -0700325 info->msg_tree = RB_ROOT;
Davidlohr Buesoa5091fd2019-05-14 15:46:26 -0700326 info->msg_tree_rightmost = NULL;
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700327 info->node_cache = NULL;
Jiri Slaby04715202011-07-26 16:08:46 -0700328 memset(&info->attr, 0, sizeof(info->attr));
KOSAKI Motohirocef01842012-05-31 16:26:33 -0700329 info->attr.mq_maxmsg = min(ipc_ns->mq_msg_max,
330 ipc_ns->mq_msg_default);
331 info->attr.mq_msgsize = min(ipc_ns->mq_msgsize_max,
332 ipc_ns->mq_msgsize_default);
Jiri Slaby04715202011-07-26 16:08:46 -0700333 if (attr) {
334 info->attr.mq_maxmsg = attr->mq_maxmsg;
335 info->attr.mq_msgsize = attr->mq_msgsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700337 /*
338 * We used to allocate a static array of pointers and account
339 * the size of that array as well as one msg_msg struct per
340 * possible message into the queue size. That's no longer
341 * accurate as the queue is now an rbtree and will grow and
342 * shrink depending on usage patterns. We can, however, still
343 * account one msg_msg struct per message, but the nodes are
344 * allocated depending on priority usage, and most programs
345 * only use one, or a handful, of priorities. However, since
346 * this is pinned memory, we need to assume worst case, so
347 * that means the min(mq_maxmsg, max_priorities) * struct
348 * posix_msg_tree_node.
349 */
Al Viro05c1b292017-12-01 17:43:43 -0500350
351 ret = -EINVAL;
352 if (info->attr.mq_maxmsg <= 0 || info->attr.mq_msgsize <= 0)
353 goto out_inode;
354 if (capable(CAP_SYS_RESOURCE)) {
355 if (info->attr.mq_maxmsg > HARD_MSGMAX ||
356 info->attr.mq_msgsize > HARD_MSGSIZEMAX)
357 goto out_inode;
358 } else {
359 if (info->attr.mq_maxmsg > ipc_ns->mq_msg_max ||
360 info->attr.mq_msgsize > ipc_ns->mq_msgsize_max)
361 goto out_inode;
362 }
363 ret = -EOVERFLOW;
364 /* check for overflow */
365 if (info->attr.mq_msgsize > ULONG_MAX/info->attr.mq_maxmsg)
366 goto out_inode;
Doug Ledfordd6629852012-05-31 16:26:35 -0700367 mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) +
368 min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) *
369 sizeof(struct posix_msg_tree_node);
Al Viro05c1b292017-12-01 17:43:43 -0500370 mq_bytes = info->attr.mq_maxmsg * info->attr.mq_msgsize;
371 if (mq_bytes + mq_treesize < mq_bytes)
372 goto out_inode;
373 mq_bytes += mq_treesize;
Jiri Slaby04715202011-07-26 16:08:46 -0700374 spin_lock(&mq_lock);
375 if (u->mq_bytes + mq_bytes < u->mq_bytes ||
Davidlohr Bueso2a4e64b2012-01-20 14:34:01 -0800376 u->mq_bytes + mq_bytes > rlimit(RLIMIT_MSGQUEUE)) {
Jiri Slaby04715202011-07-26 16:08:46 -0700377 spin_unlock(&mq_lock);
378 /* mqueue_evict_inode() releases info->messages */
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700379 ret = -EMFILE;
Jiri Slaby04715202011-07-26 16:08:46 -0700380 goto out_inode;
381 }
382 u->mq_bytes += mq_bytes;
383 spin_unlock(&mq_lock);
384
385 /* all is ok */
386 info->user = get_uid(u);
387 } else if (S_ISDIR(mode)) {
388 inc_nlink(inode);
389 /* Some things misbehave if size == 0 on a directory */
390 inode->i_size = 2 * DIRENT_SIZE;
391 inode->i_op = &mqueue_dir_inode_operations;
392 inode->i_fop = &simple_dir_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 }
Jiri Slaby04715202011-07-26 16:08:46 -0700394
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 return inode;
396out_inode:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 iput(inode);
Jiri Slaby04715202011-07-26 16:08:46 -0700398err:
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700399 return ERR_PTR(ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400}
401
David Howells935c69122018-11-01 23:07:25 +0000402static int mqueue_fill_super(struct super_block *sb, struct fs_context *fc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403{
404 struct inode *inode;
Eric W. Biedermancfb2f6f2018-03-24 11:28:14 -0500405 struct ipc_namespace *ns = sb->s_fs_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406
Eric W. Biedermana2982cc2016-06-09 15:34:02 -0500407 sb->s_iflags |= SB_I_NOEXEC | SB_I_NODEV;
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300408 sb->s_blocksize = PAGE_SIZE;
409 sb->s_blocksize_bits = PAGE_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 sb->s_magic = MQUEUE_MAGIC;
411 sb->s_op = &mqueue_super_ops;
412
Al Viro48fde702012-01-08 22:15:13 -0500413 inode = mqueue_get_inode(sb, ns, S_IFDIR | S_ISVTX | S_IRWXUGO, NULL);
414 if (IS_ERR(inode))
415 return PTR_ERR(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416
Al Viro48fde702012-01-08 22:15:13 -0500417 sb->s_root = d_make_root(inode);
418 if (!sb->s_root)
419 return -ENOMEM;
420 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421}
422
David Howells935c69122018-11-01 23:07:25 +0000423static int mqueue_get_tree(struct fs_context *fc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424{
David Howells935c69122018-11-01 23:07:25 +0000425 struct mqueue_fs_context *ctx = fc->fs_private;
426
Al Viro533770c2019-09-03 19:05:48 -0400427 return get_tree_keyed(fc, mqueue_fill_super, ctx->ipc_ns);
David Howells935c69122018-11-01 23:07:25 +0000428}
429
430static void mqueue_fs_context_free(struct fs_context *fc)
431{
432 struct mqueue_fs_context *ctx = fc->fs_private;
433
Al Viro709a643d2019-05-12 17:46:05 -0400434 put_ipc_ns(ctx->ipc_ns);
David Howells935c69122018-11-01 23:07:25 +0000435 kfree(ctx);
436}
437
438static int mqueue_init_fs_context(struct fs_context *fc)
439{
440 struct mqueue_fs_context *ctx;
441
442 ctx = kzalloc(sizeof(struct mqueue_fs_context), GFP_KERNEL);
443 if (!ctx)
444 return -ENOMEM;
445
446 ctx->ipc_ns = get_ipc_ns(current->nsproxy->ipc_ns);
Al Viro709a643d2019-05-12 17:46:05 -0400447 put_user_ns(fc->user_ns);
448 fc->user_ns = get_user_ns(ctx->ipc_ns->user_ns);
David Howells935c69122018-11-01 23:07:25 +0000449 fc->fs_private = ctx;
450 fc->ops = &mqueue_fs_context_ops;
451 return 0;
452}
453
454static struct vfsmount *mq_create_mount(struct ipc_namespace *ns)
455{
456 struct mqueue_fs_context *ctx;
457 struct fs_context *fc;
458 struct vfsmount *mnt;
459
460 fc = fs_context_for_mount(&mqueue_fs_type, SB_KERNMOUNT);
461 if (IS_ERR(fc))
462 return ERR_CAST(fc);
463
464 ctx = fc->fs_private;
465 put_ipc_ns(ctx->ipc_ns);
466 ctx->ipc_ns = get_ipc_ns(ns);
Al Viro709a643d2019-05-12 17:46:05 -0400467 put_user_ns(fc->user_ns);
468 fc->user_ns = get_user_ns(ctx->ipc_ns->user_ns);
David Howells935c69122018-11-01 23:07:25 +0000469
470 mnt = fc_mount(fc);
471 put_fs_context(fc);
472 return mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473}
474
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700475static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476{
477 struct mqueue_inode_info *p = (struct mqueue_inode_info *) foo;
478
Christoph Lametera35afb82007-05-16 22:10:57 -0700479 inode_init_once(&p->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480}
481
482static struct inode *mqueue_alloc_inode(struct super_block *sb)
483{
484 struct mqueue_inode_info *ei;
485
Christoph Lametere94b1762006-12-06 20:33:17 -0800486 ei = kmem_cache_alloc(mqueue_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 if (!ei)
488 return NULL;
489 return &ei->vfs_inode;
490}
491
Al Viro015d7952019-04-15 22:30:30 -0400492static void mqueue_free_inode(struct inode *inode)
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100493{
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100494 kmem_cache_free(mqueue_inode_cachep, MQUEUE_I(inode));
495}
496
Al Viro6d8af642010-06-05 16:29:45 -0400497static void mqueue_evict_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498{
499 struct mqueue_inode_info *info;
500 struct user_struct *user;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700501 struct ipc_namespace *ipc_ns;
Li Rongqingd6a29462019-05-14 15:46:20 -0700502 struct msg_msg *msg, *nmsg;
503 LIST_HEAD(tmp_msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504
Jan Karadbd57682012-05-03 14:48:02 +0200505 clear_inode(inode);
Al Viro6d8af642010-06-05 16:29:45 -0400506
507 if (S_ISDIR(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 return;
Al Viro6d8af642010-06-05 16:29:45 -0400509
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700510 ipc_ns = get_ns_from_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 info = MQUEUE_I(inode);
512 spin_lock(&info->lock);
Doug Ledfordd6629852012-05-31 16:26:35 -0700513 while ((msg = msg_get(info)) != NULL)
Li Rongqingd6a29462019-05-14 15:46:20 -0700514 list_add_tail(&msg->m_list, &tmp_msg);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700515 kfree(info->node_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 spin_unlock(&info->lock);
517
Li Rongqingd6a29462019-05-14 15:46:20 -0700518 list_for_each_entry_safe(msg, nmsg, &tmp_msg, m_list) {
519 list_del(&msg->m_list);
520 free_msg(msg);
521 }
522
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 user = info->user;
524 if (user) {
Kees Cooka318f122019-07-16 16:30:21 -0700525 unsigned long mq_bytes, mq_treesize;
526
527 /* Total amount of bytes accounted for the mqueue */
528 mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) +
529 min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) *
530 sizeof(struct posix_msg_tree_node);
531
532 mq_bytes = mq_treesize + (info->attr.mq_maxmsg *
533 info->attr.mq_msgsize);
534
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 spin_lock(&mq_lock);
536 user->mq_bytes -= mq_bytes;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700537 /*
538 * get_ns_from_inode() ensures that the
539 * (ipc_ns = sb->s_fs_info) is either a valid ipc_ns
540 * to which we now hold a reference, or it is NULL.
541 * We can't put it here under mq_lock, though.
542 */
543 if (ipc_ns)
544 ipc_ns->mq_queues_count--;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 spin_unlock(&mq_lock);
546 free_uid(user);
547 }
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700548 if (ipc_ns)
549 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550}
551
Al Viroeecec192017-12-01 17:26:05 -0500552static int mqueue_create_attr(struct dentry *dentry, umode_t mode, void *arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553{
Al Viroeecec192017-12-01 17:26:05 -0500554 struct inode *dir = dentry->d_parent->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 struct inode *inode;
Al Viroeecec192017-12-01 17:26:05 -0500556 struct mq_attr *attr = arg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 int error;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700558 struct ipc_namespace *ipc_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559
560 spin_lock(&mq_lock);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700561 ipc_ns = __get_ns_from_inode(dir);
562 if (!ipc_ns) {
563 error = -EACCES;
564 goto out_unlock;
565 }
Davidlohr Buesof3713fd2014-02-25 15:01:45 -0800566
567 if (ipc_ns->mq_queues_count >= ipc_ns->mq_queues_max &&
568 !capable(CAP_SYS_RESOURCE)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569 error = -ENOSPC;
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700570 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 }
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700572 ipc_ns->mq_queues_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 spin_unlock(&mq_lock);
574
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700575 inode = mqueue_get_inode(dir->i_sb, ipc_ns, mode, attr);
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700576 if (IS_ERR(inode)) {
577 error = PTR_ERR(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 spin_lock(&mq_lock);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700579 ipc_ns->mq_queues_count--;
580 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 }
582
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700583 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 dir->i_size += DIRENT_SIZE;
Deepa Dinamani078cd822016-09-14 07:48:04 -0700585 dir->i_ctime = dir->i_mtime = dir->i_atime = current_time(dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586
587 d_instantiate(dentry, inode);
588 dget(dentry);
589 return 0;
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700590out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 spin_unlock(&mq_lock);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700592 if (ipc_ns)
593 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594 return error;
595}
596
Christian Brauner549c7292021-01-21 14:19:43 +0100597static int mqueue_create(struct user_namespace *mnt_userns, struct inode *dir,
598 struct dentry *dentry, umode_t mode, bool excl)
Al Viroeecec192017-12-01 17:26:05 -0500599{
600 return mqueue_create_attr(dentry, mode, NULL);
601}
602
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603static int mqueue_unlink(struct inode *dir, struct dentry *dentry)
604{
David Howells75c3cfa2015-03-17 22:26:12 +0000605 struct inode *inode = d_inode(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606
Deepa Dinamani078cd822016-09-14 07:48:04 -0700607 dir->i_ctime = dir->i_mtime = dir->i_atime = current_time(dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 dir->i_size -= DIRENT_SIZE;
Manfred Spraul239521f2014-01-27 17:07:04 -0800609 drop_nlink(inode);
610 dput(dentry);
611 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612}
613
614/*
615* This is routine for system read from queue file.
616* To avoid mess with doing here some sort of mq_receive we allow
617* to read only queue size & notification info (the only values
618* that are interesting from user point of view and aren't accessible
619* through std routines)
620*/
621static ssize_t mqueue_read_file(struct file *filp, char __user *u_data,
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700622 size_t count, loff_t *off)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623{
Al Viro496ad9a2013-01-23 17:07:38 -0500624 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 char buffer[FILENT_SIZE];
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700626 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627
628 spin_lock(&info->lock);
629 snprintf(buffer, sizeof(buffer),
630 "QSIZE:%-10lu NOTIFY:%-5d SIGNO:%-5d NOTIFY_PID:%-6d\n",
631 info->qsize,
632 info->notify_owner ? info->notify.sigev_notify : 0,
633 (info->notify_owner &&
634 info->notify.sigev_notify == SIGEV_SIGNAL) ?
635 info->notify.sigev_signo : 0,
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800636 pid_vnr(info->notify_owner));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 spin_unlock(&info->lock);
638 buffer[sizeof(buffer)-1] = '\0';
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700640 ret = simple_read_from_buffer(u_data, count, off, buffer,
641 strlen(buffer));
642 if (ret <= 0)
643 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644
Deepa Dinamani078cd822016-09-14 07:48:04 -0700645 file_inode(filp)->i_atime = file_inode(filp)->i_ctime = current_time(file_inode(filp));
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700646 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647}
648
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -0700649static int mqueue_flush_file(struct file *filp, fl_owner_t id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650{
Al Viro496ad9a2013-01-23 17:07:38 -0500651 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652
653 spin_lock(&info->lock);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700654 if (task_tgid(current) == info->notify_owner)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 remove_notification(info);
656
657 spin_unlock(&info->lock);
658 return 0;
659}
660
Al Viro9dd95742017-07-03 00:42:43 -0400661static __poll_t mqueue_poll_file(struct file *filp, struct poll_table_struct *poll_tab)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662{
Al Viro496ad9a2013-01-23 17:07:38 -0500663 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
Al Viro9dd95742017-07-03 00:42:43 -0400664 __poll_t retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665
666 poll_wait(filp, &info->wait_q, poll_tab);
667
668 spin_lock(&info->lock);
669 if (info->attr.mq_curmsgs)
Linus Torvaldsa9a08842018-02-11 14:34:03 -0800670 retval = EPOLLIN | EPOLLRDNORM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
672 if (info->attr.mq_curmsgs < info->attr.mq_maxmsg)
Linus Torvaldsa9a08842018-02-11 14:34:03 -0800673 retval |= EPOLLOUT | EPOLLWRNORM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 spin_unlock(&info->lock);
675
676 return retval;
677}
678
679/* Adds current to info->e_wait_q[sr] before element with smaller prio */
680static void wq_add(struct mqueue_inode_info *info, int sr,
681 struct ext_wait_queue *ewp)
682{
683 struct ext_wait_queue *walk;
684
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 list_for_each_entry(walk, &info->e_wait_q[sr].list, list) {
Jonathan Haws68e34f42018-02-06 15:40:52 -0800686 if (walk->task->prio <= current->prio) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 list_add_tail(&ewp->list, &walk->list);
688 return;
689 }
690 }
691 list_add_tail(&ewp->list, &info->e_wait_q[sr].list);
692}
693
694/*
695 * Puts current task to sleep. Caller must hold queue lock. After return
696 * lock isn't held.
697 * sr: SEND or RECV
698 */
699static int wq_sleep(struct mqueue_inode_info *info, int sr,
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200700 ktime_t *timeout, struct ext_wait_queue *ewp)
Luc Van Oostenryckeac0b1c2017-02-27 14:28:21 -0800701 __releases(&info->lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702{
703 int retval;
704 signed long time;
705
706 wq_add(info, sr, ewp);
707
708 for (;;) {
Manfred Spraulc5b2cbd2020-02-03 17:34:36 -0800709 /* memory barrier not required, we hold info->lock */
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700710 __set_current_state(TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711
712 spin_unlock(&info->lock);
Wanlong Gao32ea8452011-10-31 17:06:35 -0700713 time = schedule_hrtimeout_range_clock(timeout, 0,
714 HRTIMER_MODE_ABS, CLOCK_REALTIME);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715
Manfred Spraulc5b2cbd2020-02-03 17:34:36 -0800716 if (READ_ONCE(ewp->state) == STATE_READY) {
717 /* see MQ_BARRIER for purpose/pairing */
718 smp_acquire__after_ctrl_dep();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 retval = 0;
720 goto out;
721 }
722 spin_lock(&info->lock);
Manfred Spraulc5b2cbd2020-02-03 17:34:36 -0800723
724 /* we hold info->lock, so no memory barrier required */
725 if (READ_ONCE(ewp->state) == STATE_READY) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 retval = 0;
727 goto out_unlock;
728 }
729 if (signal_pending(current)) {
730 retval = -ERESTARTSYS;
731 break;
732 }
733 if (time == 0) {
734 retval = -ETIMEDOUT;
735 break;
736 }
737 }
738 list_del(&ewp->list);
739out_unlock:
740 spin_unlock(&info->lock);
741out:
742 return retval;
743}
744
745/*
746 * Returns waiting task that should be serviced first or NULL if none exists
747 */
748static struct ext_wait_queue *wq_get_first_waiter(
749 struct mqueue_inode_info *info, int sr)
750{
751 struct list_head *ptr;
752
753 ptr = info->e_wait_q[sr].list.prev;
754 if (ptr == &info->e_wait_q[sr].list)
755 return NULL;
756 return list_entry(ptr, struct ext_wait_queue, list);
757}
758
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759
760static inline void set_cookie(struct sk_buff *skb, char code)
761{
Manfred Spraul239521f2014-01-27 17:07:04 -0800762 ((char *)skb->data)[NOTIFY_COOKIE_LEN-1] = code;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763}
764
765/*
766 * The next function is only to split too long sys_mq_timedsend
767 */
768static void __do_notify(struct mqueue_inode_info *info)
769{
770 /* notification
771 * invoked when there is registered process and there isn't process
772 * waiting synchronously for message AND state of queue changed from
773 * empty to not empty. Here we are sure that no one is waiting
774 * synchronously. */
775 if (info->notify_owner &&
776 info->attr.mq_curmsgs == 1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 switch (info->notify.sigev_notify) {
778 case SIGEV_NONE:
779 break;
Oleg Nesterovb5f20062020-05-07 18:35:39 -0700780 case SIGEV_SIGNAL: {
781 struct kernel_siginfo sig_i;
782 struct task_struct *task;
783
784 /* do_mq_notify() accepts sigev_signo == 0, why?? */
785 if (!info->notify.sigev_signo)
786 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
Eric W. Biedermanfaf1f222018-01-05 17:27:42 -0600788 clear_siginfo(&sig_i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 sig_i.si_signo = info->notify.sigev_signo;
790 sig_i.si_errno = 0;
791 sig_i.si_code = SI_MESGQ;
792 sig_i.si_value = info->notify.sigev_value;
Serge E. Hallyn6b550f92012-01-10 15:11:37 -0800793 rcu_read_lock();
Oleg Nesterovb5f20062020-05-07 18:35:39 -0700794 /* map current pid/uid into info->owner's namespaces */
Sukadev Bhattiprolua6684992009-01-07 18:08:50 -0800795 sig_i.si_pid = task_tgid_nr_ns(current,
796 ns_of_pid(info->notify_owner));
Oleg Nesterovb5f20062020-05-07 18:35:39 -0700797 sig_i.si_uid = from_kuid_munged(info->notify_user_ns,
798 current_uid());
799 /*
800 * We can't use kill_pid_info(), this signal should
801 * bypass check_kill_permission(). It is from kernel
802 * but si_fromuser() can't know this.
803 * We do check the self_exec_id, to avoid sending
804 * signals to programs that don't expect them.
805 */
806 task = pid_task(info->notify_owner, PIDTYPE_TGID);
807 if (task && task->self_exec_id ==
808 info->notify_self_exec_id) {
809 do_send_sig_info(info->notify.sigev_signo,
810 &sig_i, task, PIDTYPE_TGID);
811 }
Serge E. Hallyn6b550f92012-01-10 15:11:37 -0800812 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 break;
Oleg Nesterovb5f20062020-05-07 18:35:39 -0700814 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 case SIGEV_THREAD:
816 set_cookie(info->notify_cookie, NOTIFY_WOKENUP);
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700817 netlink_sendskb(info->notify_sock, info->notify_cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 break;
819 }
820 /* after notification unregisters process */
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700821 put_pid(info->notify_owner);
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800822 put_user_ns(info->notify_user_ns);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700823 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800824 info->notify_user_ns = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 }
826 wake_up(&info->wait_q);
827}
828
Arnd Bergmann21fc5382018-04-13 13:58:00 +0200829static int prepare_timeout(const struct __kernel_timespec __user *u_abs_timeout,
Deepa Dinamanib9047722017-08-02 19:51:11 -0700830 struct timespec64 *ts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831{
Deepa Dinamanib9047722017-08-02 19:51:11 -0700832 if (get_timespec64(ts, u_abs_timeout))
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200833 return -EFAULT;
Deepa Dinamanib9047722017-08-02 19:51:11 -0700834 if (!timespec64_valid(ts))
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200835 return -EINVAL;
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200836 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837}
838
839static void remove_notification(struct mqueue_inode_info *info)
840{
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700841 if (info->notify_owner != NULL &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 info->notify.sigev_notify == SIGEV_THREAD) {
843 set_cookie(info->notify_cookie, NOTIFY_REMOVED);
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700844 netlink_sendskb(info->notify_sock, info->notify_cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700846 put_pid(info->notify_owner);
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800847 put_user_ns(info->notify_user_ns);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700848 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800849 info->notify_user_ns = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850}
851
Al Viro066cc812017-12-01 17:51:39 -0500852static int prepare_open(struct dentry *dentry, int oflag, int ro,
853 umode_t mode, struct filename *name,
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700854 struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855{
David Howells745ca242008-11-14 10:39:22 +1100856 static const int oflag2acc[O_ACCMODE] = { MAY_READ, MAY_WRITE,
857 MAY_READ | MAY_WRITE };
Al Viro765927b2012-06-26 21:58:53 +0400858 int acc;
Al Viro066cc812017-12-01 17:51:39 -0500859
Al Viro9b20d7f2017-12-01 17:57:02 -0500860 if (d_really_is_negative(dentry)) {
861 if (!(oflag & O_CREAT))
Al Viro066cc812017-12-01 17:51:39 -0500862 return -ENOENT;
Al Viro9b20d7f2017-12-01 17:57:02 -0500863 if (ro)
864 return ro;
865 audit_inode_parent_hidden(name, dentry->d_parent);
866 return vfs_mkobj(dentry, mode & ~current_umask(),
867 mqueue_create_attr, attr);
Al Viro066cc812017-12-01 17:51:39 -0500868 }
Al Viro9b20d7f2017-12-01 17:57:02 -0500869 /* it already existed */
870 audit_inode(name, dentry, 0);
871 if ((oflag & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL))
872 return -EEXIST;
Al Viro765927b2012-06-26 21:58:53 +0400873 if ((oflag & O_ACCMODE) == (O_RDWR | O_WRONLY))
Al Viroaf4a5372017-12-01 17:34:22 -0500874 return -EINVAL;
Al Viro765927b2012-06-26 21:58:53 +0400875 acc = oflag2acc[oflag & O_ACCMODE];
Christian Brauner47291ba2021-01-21 14:19:24 +0100876 return inode_permission(&init_user_ns, d_inode(dentry), acc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877}
878
Al Viro0d060602017-06-27 21:32:36 -0400879static int do_mq_open(const char __user *u_name, int oflag, umode_t mode,
880 struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881{
Eric W. Biedermancfb2f6f2018-03-24 11:28:14 -0500882 struct vfsmount *mnt = current->nsproxy->ipc_ns->mq_mnt;
883 struct dentry *root = mnt->mnt_root;
Jeff Layton91a27b22012-10-10 15:25:28 -0400884 struct filename *name;
Al Viroa713fd72017-12-01 18:01:09 -0500885 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 int fd, error;
Al Viro312b90f2012-08-06 10:18:17 +0400887 int ro;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888
Al Viro0d060602017-06-27 21:32:36 -0400889 audit_mq_open(oflag, mode, attr);
George C. Wilson20ca73b2006-05-24 16:09:55 -0500890
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 if (IS_ERR(name = getname(u_name)))
892 return PTR_ERR(name);
893
Ulrich Drepper269f2132008-05-03 15:28:45 -0400894 fd = get_unused_fd_flags(O_CLOEXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 if (fd < 0)
896 goto out_putname;
897
Al Viro312b90f2012-08-06 10:18:17 +0400898 ro = mnt_want_write(mnt); /* we'll drop it in any case */
Al Viro59551022016-01-22 15:40:57 -0500899 inode_lock(d_inode(root));
Jeff Layton91a27b22012-10-10 15:25:28 -0400900 path.dentry = lookup_one_len(name->name, root, strlen(name->name));
Al Viro765927b2012-06-26 21:58:53 +0400901 if (IS_ERR(path.dentry)) {
902 error = PTR_ERR(path.dentry);
André Goddard Rosa4294a8e2010-02-23 04:04:28 -0300903 goto out_putfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 }
Al Viro312b90f2012-08-06 10:18:17 +0400905 path.mnt = mntget(mnt);
Al Viro066cc812017-12-01 17:51:39 -0500906 error = prepare_open(path.dentry, oflag, ro, mode, name, attr);
907 if (!error) {
908 struct file *file = dentry_open(&path, oflag, current_cred());
909 if (!IS_ERR(file))
910 fd_install(fd, file);
911 else
912 error = PTR_ERR(file);
Alexander Viro7c7dce92006-01-14 15:29:55 -0500913 }
Al Viro765927b2012-06-26 21:58:53 +0400914 path_put(&path);
Alexander Viro7c7dce92006-01-14 15:29:55 -0500915out_putfd:
Al Viro765927b2012-06-26 21:58:53 +0400916 if (error) {
917 put_unused_fd(fd);
918 fd = error;
919 }
Al Viro59551022016-01-22 15:40:57 -0500920 inode_unlock(d_inode(root));
Vladimir Davydov38d78e52013-03-22 15:04:51 -0700921 if (!ro)
922 mnt_drop_write(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923out_putname:
924 putname(name);
925 return fd;
926}
927
Al Viro0d060602017-06-27 21:32:36 -0400928SYSCALL_DEFINE4(mq_open, const char __user *, u_name, int, oflag, umode_t, mode,
929 struct mq_attr __user *, u_attr)
930{
931 struct mq_attr attr;
932 if (u_attr && copy_from_user(&attr, u_attr, sizeof(struct mq_attr)))
933 return -EFAULT;
934
935 return do_mq_open(u_name, oflag, mode, u_attr ? &attr : NULL);
936}
937
Heiko Carstensd5460c92009-01-14 14:14:27 +0100938SYSCALL_DEFINE1(mq_unlink, const char __user *, u_name)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939{
940 int err;
Jeff Layton91a27b22012-10-10 15:25:28 -0400941 struct filename *name;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 struct dentry *dentry;
943 struct inode *inode = NULL;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700944 struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns;
Al Viro312b90f2012-08-06 10:18:17 +0400945 struct vfsmount *mnt = ipc_ns->mq_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946
947 name = getname(u_name);
948 if (IS_ERR(name))
949 return PTR_ERR(name);
950
Jeff Layton79f65302013-07-08 15:59:36 -0700951 audit_inode_parent_hidden(name, mnt->mnt_root);
Al Viro312b90f2012-08-06 10:18:17 +0400952 err = mnt_want_write(mnt);
953 if (err)
954 goto out_name;
Al Viro59551022016-01-22 15:40:57 -0500955 inode_lock_nested(d_inode(mnt->mnt_root), I_MUTEX_PARENT);
Jeff Layton91a27b22012-10-10 15:25:28 -0400956 dentry = lookup_one_len(name->name, mnt->mnt_root,
957 strlen(name->name));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 if (IS_ERR(dentry)) {
959 err = PTR_ERR(dentry);
960 goto out_unlock;
961 }
962
David Howells75c3cfa2015-03-17 22:26:12 +0000963 inode = d_inode(dentry);
Al Viro312b90f2012-08-06 10:18:17 +0400964 if (!inode) {
965 err = -ENOENT;
966 } else {
Al Viro7de9c6ee2010-10-23 11:11:40 -0400967 ihold(inode);
Christian Brauner6521f892021-01-21 14:19:33 +0100968 err = vfs_unlink(&init_user_ns, d_inode(dentry->d_parent),
969 dentry, NULL);
Al Viro312b90f2012-08-06 10:18:17 +0400970 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 dput(dentry);
972
973out_unlock:
Al Viro59551022016-01-22 15:40:57 -0500974 inode_unlock(d_inode(mnt->mnt_root));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 if (inode)
976 iput(inode);
Al Viro312b90f2012-08-06 10:18:17 +0400977 mnt_drop_write(mnt);
978out_name:
979 putname(name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980
981 return err;
982}
983
984/* Pipelined send and receive functions.
985 *
986 * If a receiver finds no waiting message, then it registers itself in the
987 * list of waiting receivers. A sender checks that list before adding the new
988 * message into the message array. If there is a waiting receiver, then it
989 * bypasses the message array and directly hands the message over to the
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700990 * receiver. The receiver accepts the message and returns without grabbing the
991 * queue spinlock:
992 *
993 * - Set pointer to message.
994 * - Queue the receiver task for later wakeup (without the info->lock).
995 * - Update its state to STATE_READY. Now the receiver can continue.
996 * - Wake up the process after the lock is dropped. Should the process wake up
997 * before this wakeup (due to a timeout or a signal) it will either see
998 * STATE_READY and continue or acquire the lock to check the state again.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 *
1000 * The same algorithm is used for senders.
1001 */
1002
Davidlohr Buesoed29f172020-02-03 17:34:32 -08001003static inline void __pipelined_op(struct wake_q_head *wake_q,
1004 struct mqueue_inode_info *info,
1005 struct ext_wait_queue *this)
1006{
1007 list_del(&this->list);
Manfred Spraulc5b2cbd2020-02-03 17:34:36 -08001008 get_task_struct(this->task);
1009
1010 /* see MQ_BARRIER for purpose/pairing */
1011 smp_store_release(&this->state, STATE_READY);
1012 wake_q_add_safe(wake_q, this->task);
Davidlohr Buesoed29f172020-02-03 17:34:32 -08001013}
1014
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015/* pipelined_send() - send a message directly to the task waiting in
1016 * sys_mq_timedreceive() (without inserting message into a queue).
1017 */
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001018static inline void pipelined_send(struct wake_q_head *wake_q,
1019 struct mqueue_inode_info *info,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 struct msg_msg *message,
1021 struct ext_wait_queue *receiver)
1022{
1023 receiver->msg = message;
Davidlohr Buesoed29f172020-02-03 17:34:32 -08001024 __pipelined_op(wake_q, info, receiver);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025}
1026
1027/* pipelined_receive() - if there is task waiting in sys_mq_timedsend()
1028 * gets its message and put to the queue (we have one free place for sure). */
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001029static inline void pipelined_receive(struct wake_q_head *wake_q,
1030 struct mqueue_inode_info *info)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031{
1032 struct ext_wait_queue *sender = wq_get_first_waiter(info, SEND);
1033
1034 if (!sender) {
1035 /* for poll */
1036 wake_up_interruptible(&info->wait_q);
1037 return;
1038 }
Doug Ledfordd6629852012-05-31 16:26:35 -07001039 if (msg_insert(sender->msg, info))
1040 return;
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001041
Davidlohr Buesoed29f172020-02-03 17:34:32 -08001042 __pipelined_op(wake_q, info, sender);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043}
1044
Al Viro0d060602017-06-27 21:32:36 -04001045static int do_mq_timedsend(mqd_t mqdes, const char __user *u_msg_ptr,
1046 size_t msg_len, unsigned int msg_prio,
Deepa Dinamanib9047722017-08-02 19:51:11 -07001047 struct timespec64 *ts)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048{
Al Viro2903ff02012-08-28 12:52:22 -04001049 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 struct inode *inode;
1051 struct ext_wait_queue wait;
1052 struct ext_wait_queue *receiver;
1053 struct msg_msg *msg_ptr;
1054 struct mqueue_inode_info *info;
Carsten Emde9ca7d8e2010-04-02 22:40:20 +02001055 ktime_t expires, *timeout = NULL;
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001056 struct posix_msg_tree_node *new_leaf = NULL;
Al Viro2903ff02012-08-28 12:52:22 -04001057 int ret = 0;
Waiman Long194a6b52016-11-17 11:46:38 -05001058 DEFINE_WAKE_Q(wake_q);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059
1060 if (unlikely(msg_prio >= (unsigned long) MQ_PRIO_MAX))
1061 return -EINVAL;
1062
Al Viro0d060602017-06-27 21:32:36 -04001063 if (ts) {
Deepa Dinamanib9047722017-08-02 19:51:11 -07001064 expires = timespec64_to_ktime(*ts);
Al Viro0d060602017-06-27 21:32:36 -04001065 timeout = &expires;
1066 }
1067
1068 audit_mq_sendrecv(mqdes, msg_len, msg_prio, ts);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069
Al Viro2903ff02012-08-28 12:52:22 -04001070 f = fdget(mqdes);
1071 if (unlikely(!f.file)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001072 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001074 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075
Al Viro496ad9a2013-01-23 17:07:38 -05001076 inode = file_inode(f.file);
Al Viro2903ff02012-08-28 12:52:22 -04001077 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001078 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001080 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 info = MQUEUE_I(inode);
Al Viro9f45f5b2014-10-31 17:44:57 -04001082 audit_file(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083
Al Viro2903ff02012-08-28 12:52:22 -04001084 if (unlikely(!(f.file->f_mode & FMODE_WRITE))) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001085 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001087 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088
1089 if (unlikely(msg_len > info->attr.mq_msgsize)) {
1090 ret = -EMSGSIZE;
1091 goto out_fput;
1092 }
1093
1094 /* First try to allocate memory, before doing anything with
1095 * existing queues. */
1096 msg_ptr = load_msg(u_msg_ptr, msg_len);
1097 if (IS_ERR(msg_ptr)) {
1098 ret = PTR_ERR(msg_ptr);
1099 goto out_fput;
1100 }
1101 msg_ptr->m_ts = msg_len;
1102 msg_ptr->m_type = msg_prio;
1103
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001104 /*
1105 * msg_insert really wants us to have a valid, spare node struct so
1106 * it doesn't have to kmalloc a GFP_ATOMIC allocation, but it will
1107 * fall back to that if necessary.
1108 */
1109 if (!info->node_cache)
1110 new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL);
1111
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 spin_lock(&info->lock);
1113
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001114 if (!info->node_cache && new_leaf) {
1115 /* Save our speculative allocation into the cache */
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001116 INIT_LIST_HEAD(&new_leaf->msg_list);
1117 info->node_cache = new_leaf;
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001118 new_leaf = NULL;
1119 } else {
1120 kfree(new_leaf);
1121 }
1122
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 if (info->attr.mq_curmsgs == info->attr.mq_maxmsg) {
Al Viro2903ff02012-08-28 12:52:22 -04001124 if (f.file->f_flags & O_NONBLOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125 ret = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 } else {
1127 wait.task = current;
1128 wait.msg = (void *) msg_ptr;
Manfred Spraulc5b2cbd2020-02-03 17:34:36 -08001129
1130 /* memory barrier not required, we hold info->lock */
1131 WRITE_ONCE(wait.state, STATE_NONE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 ret = wq_sleep(info, SEND, timeout, &wait);
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001133 /*
1134 * wq_sleep must be called with info->lock held, and
1135 * returns with the lock released
1136 */
1137 goto out_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 } else {
1140 receiver = wq_get_first_waiter(info, RECV);
1141 if (receiver) {
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001142 pipelined_send(&wake_q, info, msg_ptr, receiver);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 } else {
1144 /* adds message to the queue */
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001145 ret = msg_insert(msg_ptr, info);
1146 if (ret)
1147 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 __do_notify(info);
1149 }
1150 inode->i_atime = inode->i_mtime = inode->i_ctime =
Deepa Dinamani078cd822016-09-14 07:48:04 -07001151 current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 }
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001153out_unlock:
1154 spin_unlock(&info->lock);
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001155 wake_up_q(&wake_q);
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001156out_free:
1157 if (ret)
1158 free_msg(msg_ptr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04001160 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161out:
1162 return ret;
1163}
1164
Al Viro0d060602017-06-27 21:32:36 -04001165static int do_mq_timedreceive(mqd_t mqdes, char __user *u_msg_ptr,
1166 size_t msg_len, unsigned int __user *u_msg_prio,
Deepa Dinamanib9047722017-08-02 19:51:11 -07001167 struct timespec64 *ts)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169 ssize_t ret;
1170 struct msg_msg *msg_ptr;
Al Viro2903ff02012-08-28 12:52:22 -04001171 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 struct inode *inode;
1173 struct mqueue_inode_info *info;
1174 struct ext_wait_queue wait;
Carsten Emde9ca7d8e2010-04-02 22:40:20 +02001175 ktime_t expires, *timeout = NULL;
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001176 struct posix_msg_tree_node *new_leaf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177
Al Viro0d060602017-06-27 21:32:36 -04001178 if (ts) {
Deepa Dinamanib9047722017-08-02 19:51:11 -07001179 expires = timespec64_to_ktime(*ts);
Carsten Emde9ca7d8e2010-04-02 22:40:20 +02001180 timeout = &expires;
Al Viroc32c8af2008-12-14 03:46:48 -05001181 }
George C. Wilson20ca73b2006-05-24 16:09:55 -05001182
Al Viro0d060602017-06-27 21:32:36 -04001183 audit_mq_sendrecv(mqdes, msg_len, 0, ts);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184
Al Viro2903ff02012-08-28 12:52:22 -04001185 f = fdget(mqdes);
1186 if (unlikely(!f.file)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001187 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001189 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190
Al Viro496ad9a2013-01-23 17:07:38 -05001191 inode = file_inode(f.file);
Al Viro2903ff02012-08-28 12:52:22 -04001192 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001193 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001195 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 info = MQUEUE_I(inode);
Al Viro9f45f5b2014-10-31 17:44:57 -04001197 audit_file(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198
Al Viro2903ff02012-08-28 12:52:22 -04001199 if (unlikely(!(f.file->f_mode & FMODE_READ))) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001200 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001202 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203
1204 /* checks if buffer is big enough */
1205 if (unlikely(msg_len < info->attr.mq_msgsize)) {
1206 ret = -EMSGSIZE;
1207 goto out_fput;
1208 }
1209
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001210 /*
1211 * msg_insert really wants us to have a valid, spare node struct so
1212 * it doesn't have to kmalloc a GFP_ATOMIC allocation, but it will
1213 * fall back to that if necessary.
1214 */
1215 if (!info->node_cache)
1216 new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL);
1217
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 spin_lock(&info->lock);
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001219
1220 if (!info->node_cache && new_leaf) {
1221 /* Save our speculative allocation into the cache */
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001222 INIT_LIST_HEAD(&new_leaf->msg_list);
1223 info->node_cache = new_leaf;
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001224 } else {
1225 kfree(new_leaf);
1226 }
1227
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228 if (info->attr.mq_curmsgs == 0) {
Al Viro2903ff02012-08-28 12:52:22 -04001229 if (f.file->f_flags & O_NONBLOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 spin_unlock(&info->lock);
1231 ret = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 } else {
1233 wait.task = current;
Manfred Spraulc5b2cbd2020-02-03 17:34:36 -08001234
1235 /* memory barrier not required, we hold info->lock */
1236 WRITE_ONCE(wait.state, STATE_NONE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 ret = wq_sleep(info, RECV, timeout, &wait);
1238 msg_ptr = wait.msg;
1239 }
1240 } else {
Waiman Long194a6b52016-11-17 11:46:38 -05001241 DEFINE_WAKE_Q(wake_q);
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001242
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243 msg_ptr = msg_get(info);
1244
1245 inode->i_atime = inode->i_mtime = inode->i_ctime =
Deepa Dinamani078cd822016-09-14 07:48:04 -07001246 current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247
1248 /* There is now free space in queue. */
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001249 pipelined_receive(&wake_q, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 spin_unlock(&info->lock);
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001251 wake_up_q(&wake_q);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252 ret = 0;
1253 }
1254 if (ret == 0) {
1255 ret = msg_ptr->m_ts;
1256
1257 if ((u_msg_prio && put_user(msg_ptr->m_type, u_msg_prio)) ||
1258 store_msg(u_msg_ptr, msg_ptr, msg_ptr->m_ts)) {
1259 ret = -EFAULT;
1260 }
1261 free_msg(msg_ptr);
1262 }
1263out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04001264 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265out:
1266 return ret;
1267}
1268
Al Viro0d060602017-06-27 21:32:36 -04001269SYSCALL_DEFINE5(mq_timedsend, mqd_t, mqdes, const char __user *, u_msg_ptr,
1270 size_t, msg_len, unsigned int, msg_prio,
Arnd Bergmann21fc5382018-04-13 13:58:00 +02001271 const struct __kernel_timespec __user *, u_abs_timeout)
Al Viro0d060602017-06-27 21:32:36 -04001272{
Deepa Dinamanib9047722017-08-02 19:51:11 -07001273 struct timespec64 ts, *p = NULL;
Al Viro0d060602017-06-27 21:32:36 -04001274 if (u_abs_timeout) {
1275 int res = prepare_timeout(u_abs_timeout, &ts);
1276 if (res)
1277 return res;
1278 p = &ts;
1279 }
1280 return do_mq_timedsend(mqdes, u_msg_ptr, msg_len, msg_prio, p);
1281}
1282
1283SYSCALL_DEFINE5(mq_timedreceive, mqd_t, mqdes, char __user *, u_msg_ptr,
1284 size_t, msg_len, unsigned int __user *, u_msg_prio,
Arnd Bergmann21fc5382018-04-13 13:58:00 +02001285 const struct __kernel_timespec __user *, u_abs_timeout)
Al Viro0d060602017-06-27 21:32:36 -04001286{
Deepa Dinamanib9047722017-08-02 19:51:11 -07001287 struct timespec64 ts, *p = NULL;
Al Viro0d060602017-06-27 21:32:36 -04001288 if (u_abs_timeout) {
1289 int res = prepare_timeout(u_abs_timeout, &ts);
1290 if (res)
1291 return res;
1292 p = &ts;
1293 }
1294 return do_mq_timedreceive(mqdes, u_msg_ptr, msg_len, u_msg_prio, p);
1295}
1296
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297/*
1298 * Notes: the case when user wants us to deregister (with NULL as pointer)
1299 * and he isn't currently owner of notification, will be silently discarded.
1300 * It isn't explicitly defined in the POSIX.
1301 */
Al Viro0d060602017-06-27 21:32:36 -04001302static int do_mq_notify(mqd_t mqdes, const struct sigevent *notification)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303{
Al Viro2903ff02012-08-28 12:52:22 -04001304 int ret;
1305 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306 struct sock *sock;
1307 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308 struct mqueue_inode_info *info;
1309 struct sk_buff *nc;
1310
Al Viro0d060602017-06-27 21:32:36 -04001311 audit_mq_notify(mqdes, notification);
George C. Wilson20ca73b2006-05-24 16:09:55 -05001312
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313 nc = NULL;
1314 sock = NULL;
Al Viro0d060602017-06-27 21:32:36 -04001315 if (notification != NULL) {
1316 if (unlikely(notification->sigev_notify != SIGEV_NONE &&
1317 notification->sigev_notify != SIGEV_SIGNAL &&
1318 notification->sigev_notify != SIGEV_THREAD))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 return -EINVAL;
Al Viro0d060602017-06-27 21:32:36 -04001320 if (notification->sigev_notify == SIGEV_SIGNAL &&
1321 !valid_signal(notification->sigev_signo)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322 return -EINVAL;
1323 }
Al Viro0d060602017-06-27 21:32:36 -04001324 if (notification->sigev_notify == SIGEV_THREAD) {
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001325 long timeo;
1326
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 /* create the notify skb */
1328 nc = alloc_skb(NOTIFY_COOKIE_LEN, GFP_KERNEL);
Markus Elfringc2317402019-09-25 16:48:17 -07001329 if (!nc)
1330 return -ENOMEM;
1331
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 if (copy_from_user(nc->data,
Al Viro0d060602017-06-27 21:32:36 -04001333 notification->sigev_value.sival_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334 NOTIFY_COOKIE_LEN)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001335 ret = -EFAULT;
Markus Elfringc2317402019-09-25 16:48:17 -07001336 goto free_skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 }
1338
1339 /* TODO: add a header? */
1340 skb_put(nc, NOTIFY_COOKIE_LEN);
1341 /* and attach it to the socket */
1342retry:
Al Viro0d060602017-06-27 21:32:36 -04001343 f = fdget(notification->sigev_signo);
Al Viro2903ff02012-08-28 12:52:22 -04001344 if (!f.file) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001345 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001347 }
Al Viro2903ff02012-08-28 12:52:22 -04001348 sock = netlink_getsockbyfilp(f.file);
1349 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 if (IS_ERR(sock)) {
1351 ret = PTR_ERR(sock);
Markus Elfringc2317402019-09-25 16:48:17 -07001352 goto free_skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353 }
1354
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001355 timeo = MAX_SCHEDULE_TIMEOUT;
Denis V. Lunev9457afe2008-06-05 11:23:39 -07001356 ret = netlink_attachskb(sock, nc, &timeo, NULL);
Cong Wangf991af32017-07-09 13:19:55 -07001357 if (ret == 1) {
1358 sock = NULL;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001359 goto retry;
Cong Wangf991af32017-07-09 13:19:55 -07001360 }
Markus Elfringc2317402019-09-25 16:48:17 -07001361 if (ret)
1362 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 }
1364 }
1365
Al Viro2903ff02012-08-28 12:52:22 -04001366 f = fdget(mqdes);
1367 if (!f.file) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001368 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001370 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371
Al Viro496ad9a2013-01-23 17:07:38 -05001372 inode = file_inode(f.file);
Al Viro2903ff02012-08-28 12:52:22 -04001373 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001374 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001376 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377 info = MQUEUE_I(inode);
1378
1379 ret = 0;
1380 spin_lock(&info->lock);
Al Viro0d060602017-06-27 21:32:36 -04001381 if (notification == NULL) {
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001382 if (info->notify_owner == task_tgid(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 remove_notification(info);
Deepa Dinamani078cd822016-09-14 07:48:04 -07001384 inode->i_atime = inode->i_ctime = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001386 } else if (info->notify_owner != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 ret = -EBUSY;
1388 } else {
Al Viro0d060602017-06-27 21:32:36 -04001389 switch (notification->sigev_notify) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 case SIGEV_NONE:
1391 info->notify.sigev_notify = SIGEV_NONE;
1392 break;
1393 case SIGEV_THREAD:
1394 info->notify_sock = sock;
1395 info->notify_cookie = nc;
1396 sock = NULL;
1397 nc = NULL;
1398 info->notify.sigev_notify = SIGEV_THREAD;
1399 break;
1400 case SIGEV_SIGNAL:
Al Viro0d060602017-06-27 21:32:36 -04001401 info->notify.sigev_signo = notification->sigev_signo;
1402 info->notify.sigev_value = notification->sigev_value;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403 info->notify.sigev_notify = SIGEV_SIGNAL;
Oleg Nesterovb5f20062020-05-07 18:35:39 -07001404 info->notify_self_exec_id = current->self_exec_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405 break;
1406 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001407
1408 info->notify_owner = get_pid(task_tgid(current));
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -08001409 info->notify_user_ns = get_user_ns(current_user_ns());
Deepa Dinamani078cd822016-09-14 07:48:04 -07001410 inode->i_atime = inode->i_ctime = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411 }
1412 spin_unlock(&info->lock);
1413out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04001414 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415out:
Davidlohr Bueso3ab08fe2014-01-27 17:07:06 -08001416 if (sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417 netlink_detachskb(sock, nc);
Markus Elfring97b0b1a2019-09-25 16:48:14 -07001418 else
Markus Elfringc2317402019-09-25 16:48:17 -07001419free_skb:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 dev_kfree_skb(nc);
Davidlohr Bueso3ab08fe2014-01-27 17:07:06 -08001421
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 return ret;
1423}
1424
Al Viro0d060602017-06-27 21:32:36 -04001425SYSCALL_DEFINE2(mq_notify, mqd_t, mqdes,
1426 const struct sigevent __user *, u_notification)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427{
Al Viro0d060602017-06-27 21:32:36 -04001428 struct sigevent n, *p = NULL;
1429 if (u_notification) {
1430 if (copy_from_user(&n, u_notification, sizeof(struct sigevent)))
1431 return -EFAULT;
1432 p = &n;
1433 }
1434 return do_mq_notify(mqdes, p);
1435}
1436
1437static int do_mq_getsetattr(int mqdes, struct mq_attr *new, struct mq_attr *old)
1438{
Al Viro2903ff02012-08-28 12:52:22 -04001439 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440 struct inode *inode;
1441 struct mqueue_inode_info *info;
1442
Al Viro0d060602017-06-27 21:32:36 -04001443 if (new && (new->mq_flags & (~O_NONBLOCK)))
1444 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445
Al Viro2903ff02012-08-28 12:52:22 -04001446 f = fdget(mqdes);
Al Viro0d060602017-06-27 21:32:36 -04001447 if (!f.file)
1448 return -EBADF;
1449
1450 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
1451 fdput(f);
1452 return -EBADF;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001453 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454
Al Viro496ad9a2013-01-23 17:07:38 -05001455 inode = file_inode(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456 info = MQUEUE_I(inode);
1457
1458 spin_lock(&info->lock);
1459
Al Viro0d060602017-06-27 21:32:36 -04001460 if (old) {
1461 *old = info->attr;
1462 old->mq_flags = f.file->f_flags & O_NONBLOCK;
1463 }
1464 if (new) {
1465 audit_mq_getsetattr(mqdes, new);
Al Viro2903ff02012-08-28 12:52:22 -04001466 spin_lock(&f.file->f_lock);
Al Viro0d060602017-06-27 21:32:36 -04001467 if (new->mq_flags & O_NONBLOCK)
Al Viro2903ff02012-08-28 12:52:22 -04001468 f.file->f_flags |= O_NONBLOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 else
Al Viro2903ff02012-08-28 12:52:22 -04001470 f.file->f_flags &= ~O_NONBLOCK;
1471 spin_unlock(&f.file->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472
Deepa Dinamani078cd822016-09-14 07:48:04 -07001473 inode->i_atime = inode->i_ctime = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 }
1475
1476 spin_unlock(&info->lock);
Al Viro2903ff02012-08-28 12:52:22 -04001477 fdput(f);
Al Viro0d060602017-06-27 21:32:36 -04001478 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479}
1480
Al Viro0d060602017-06-27 21:32:36 -04001481SYSCALL_DEFINE3(mq_getsetattr, mqd_t, mqdes,
1482 const struct mq_attr __user *, u_mqstat,
1483 struct mq_attr __user *, u_omqstat)
1484{
1485 int ret;
1486 struct mq_attr mqstat, omqstat;
1487 struct mq_attr *new = NULL, *old = NULL;
1488
1489 if (u_mqstat) {
1490 new = &mqstat;
1491 if (copy_from_user(new, u_mqstat, sizeof(struct mq_attr)))
1492 return -EFAULT;
1493 }
1494 if (u_omqstat)
1495 old = &omqstat;
1496
1497 ret = do_mq_getsetattr(mqdes, new, old);
1498 if (ret || !old)
1499 return ret;
1500
1501 if (copy_to_user(u_omqstat, old, sizeof(struct mq_attr)))
1502 return -EFAULT;
1503 return 0;
1504}
1505
1506#ifdef CONFIG_COMPAT
1507
1508struct compat_mq_attr {
1509 compat_long_t mq_flags; /* message queue flags */
1510 compat_long_t mq_maxmsg; /* maximum number of messages */
1511 compat_long_t mq_msgsize; /* maximum message size */
1512 compat_long_t mq_curmsgs; /* number of messages currently queued */
1513 compat_long_t __reserved[4]; /* ignored for input, zeroed for output */
1514};
1515
1516static inline int get_compat_mq_attr(struct mq_attr *attr,
1517 const struct compat_mq_attr __user *uattr)
1518{
1519 struct compat_mq_attr v;
1520
1521 if (copy_from_user(&v, uattr, sizeof(*uattr)))
1522 return -EFAULT;
1523
1524 memset(attr, 0, sizeof(*attr));
1525 attr->mq_flags = v.mq_flags;
1526 attr->mq_maxmsg = v.mq_maxmsg;
1527 attr->mq_msgsize = v.mq_msgsize;
1528 attr->mq_curmsgs = v.mq_curmsgs;
1529 return 0;
1530}
1531
1532static inline int put_compat_mq_attr(const struct mq_attr *attr,
1533 struct compat_mq_attr __user *uattr)
1534{
1535 struct compat_mq_attr v;
1536
1537 memset(&v, 0, sizeof(v));
1538 v.mq_flags = attr->mq_flags;
1539 v.mq_maxmsg = attr->mq_maxmsg;
1540 v.mq_msgsize = attr->mq_msgsize;
1541 v.mq_curmsgs = attr->mq_curmsgs;
1542 if (copy_to_user(uattr, &v, sizeof(*uattr)))
1543 return -EFAULT;
1544 return 0;
1545}
1546
1547COMPAT_SYSCALL_DEFINE4(mq_open, const char __user *, u_name,
1548 int, oflag, compat_mode_t, mode,
1549 struct compat_mq_attr __user *, u_attr)
1550{
1551 struct mq_attr attr, *p = NULL;
1552 if (u_attr && oflag & O_CREAT) {
1553 p = &attr;
1554 if (get_compat_mq_attr(&attr, u_attr))
1555 return -EFAULT;
1556 }
1557 return do_mq_open(u_name, oflag, mode, p);
1558}
1559
Arnd Bergmannb0d17572018-04-13 13:58:23 +02001560COMPAT_SYSCALL_DEFINE2(mq_notify, mqd_t, mqdes,
1561 const struct compat_sigevent __user *, u_notification)
1562{
1563 struct sigevent n, *p = NULL;
1564 if (u_notification) {
1565 if (get_compat_sigevent(&n, u_notification))
1566 return -EFAULT;
1567 if (n.sigev_notify == SIGEV_THREAD)
1568 n.sigev_value.sival_ptr = compat_ptr(n.sigev_value.sival_int);
1569 p = &n;
1570 }
1571 return do_mq_notify(mqdes, p);
1572}
1573
1574COMPAT_SYSCALL_DEFINE3(mq_getsetattr, mqd_t, mqdes,
1575 const struct compat_mq_attr __user *, u_mqstat,
1576 struct compat_mq_attr __user *, u_omqstat)
1577{
1578 int ret;
1579 struct mq_attr mqstat, omqstat;
1580 struct mq_attr *new = NULL, *old = NULL;
1581
1582 if (u_mqstat) {
1583 new = &mqstat;
1584 if (get_compat_mq_attr(new, u_mqstat))
1585 return -EFAULT;
1586 }
1587 if (u_omqstat)
1588 old = &omqstat;
1589
1590 ret = do_mq_getsetattr(mqdes, new, old);
1591 if (ret || !old)
1592 return ret;
1593
1594 if (put_compat_mq_attr(old, u_omqstat))
1595 return -EFAULT;
1596 return 0;
1597}
1598#endif
1599
1600#ifdef CONFIG_COMPAT_32BIT_TIME
Arnd Bergmann9afc5ee2018-07-13 12:52:28 +02001601static int compat_prepare_timeout(const struct old_timespec32 __user *p,
Deepa Dinamanib9047722017-08-02 19:51:11 -07001602 struct timespec64 *ts)
Al Viro0d060602017-06-27 21:32:36 -04001603{
Arnd Bergmann9afc5ee2018-07-13 12:52:28 +02001604 if (get_old_timespec32(ts, p))
Al Viro0d060602017-06-27 21:32:36 -04001605 return -EFAULT;
Deepa Dinamanib9047722017-08-02 19:51:11 -07001606 if (!timespec64_valid(ts))
Al Viro0d060602017-06-27 21:32:36 -04001607 return -EINVAL;
1608 return 0;
1609}
1610
Arnd Bergmann8dabe722019-01-07 00:33:08 +01001611SYSCALL_DEFINE5(mq_timedsend_time32, mqd_t, mqdes,
1612 const char __user *, u_msg_ptr,
1613 unsigned int, msg_len, unsigned int, msg_prio,
1614 const struct old_timespec32 __user *, u_abs_timeout)
Al Viro0d060602017-06-27 21:32:36 -04001615{
Deepa Dinamanib9047722017-08-02 19:51:11 -07001616 struct timespec64 ts, *p = NULL;
Al Viro0d060602017-06-27 21:32:36 -04001617 if (u_abs_timeout) {
1618 int res = compat_prepare_timeout(u_abs_timeout, &ts);
1619 if (res)
1620 return res;
1621 p = &ts;
1622 }
1623 return do_mq_timedsend(mqdes, u_msg_ptr, msg_len, msg_prio, p);
1624}
1625
Arnd Bergmann8dabe722019-01-07 00:33:08 +01001626SYSCALL_DEFINE5(mq_timedreceive_time32, mqd_t, mqdes,
1627 char __user *, u_msg_ptr,
1628 unsigned int, msg_len, unsigned int __user *, u_msg_prio,
1629 const struct old_timespec32 __user *, u_abs_timeout)
Al Viro0d060602017-06-27 21:32:36 -04001630{
Deepa Dinamanib9047722017-08-02 19:51:11 -07001631 struct timespec64 ts, *p = NULL;
Al Viro0d060602017-06-27 21:32:36 -04001632 if (u_abs_timeout) {
1633 int res = compat_prepare_timeout(u_abs_timeout, &ts);
1634 if (res)
1635 return res;
1636 p = &ts;
1637 }
1638 return do_mq_timedreceive(mqdes, u_msg_ptr, msg_len, u_msg_prio, p);
1639}
Al Viro0d060602017-06-27 21:32:36 -04001640#endif
1641
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -08001642static const struct inode_operations mqueue_dir_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 .lookup = simple_lookup,
1644 .create = mqueue_create,
1645 .unlink = mqueue_unlink,
1646};
1647
Arjan van de Ven9a321442007-02-12 00:55:35 -08001648static const struct file_operations mqueue_file_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649 .flush = mqueue_flush_file,
1650 .poll = mqueue_poll_file,
1651 .read = mqueue_read_file,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001652 .llseek = default_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653};
1654
Alexey Dobriyanb87221d2009-09-21 17:01:09 -07001655static const struct super_operations mqueue_super_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656 .alloc_inode = mqueue_alloc_inode,
Al Viro015d7952019-04-15 22:30:30 -04001657 .free_inode = mqueue_free_inode,
Al Viro6d8af642010-06-05 16:29:45 -04001658 .evict_inode = mqueue_evict_inode,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660};
1661
David Howells935c69122018-11-01 23:07:25 +00001662static const struct fs_context_operations mqueue_fs_context_ops = {
1663 .free = mqueue_fs_context_free,
1664 .get_tree = mqueue_get_tree,
1665};
1666
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667static struct file_system_type mqueue_fs_type = {
David Howells935c69122018-11-01 23:07:25 +00001668 .name = "mqueue",
1669 .init_fs_context = mqueue_init_fs_context,
1670 .kill_sb = kill_litter_super,
1671 .fs_flags = FS_USERNS_MOUNT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672};
1673
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001674int mq_init_ns(struct ipc_namespace *ns)
1675{
David Howells935c69122018-11-01 23:07:25 +00001676 struct vfsmount *m;
1677
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001678 ns->mq_queues_count = 0;
1679 ns->mq_queues_max = DFLT_QUEUESMAX;
1680 ns->mq_msg_max = DFLT_MSGMAX;
1681 ns->mq_msgsize_max = DFLT_MSGSIZEMAX;
KOSAKI Motohirocef01842012-05-31 16:26:33 -07001682 ns->mq_msg_default = DFLT_MSG;
1683 ns->mq_msgsize_default = DFLT_MSGSIZE;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001684
David Howells935c69122018-11-01 23:07:25 +00001685 m = mq_create_mount(ns);
1686 if (IS_ERR(m))
1687 return PTR_ERR(m);
1688 ns->mq_mnt = m;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001689 return 0;
1690}
1691
1692void mq_clear_sbinfo(struct ipc_namespace *ns)
1693{
Eric W. Biedermancfb2f6f2018-03-24 11:28:14 -05001694 ns->mq_mnt->mnt_sb->s_fs_info = NULL;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001695}
1696
1697void mq_put_mnt(struct ipc_namespace *ns)
1698{
Eric W. Biedermancfb2f6f2018-03-24 11:28:14 -05001699 kern_unmount(ns->mq_mnt);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001700}
1701
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702static int __init init_mqueue_fs(void)
1703{
1704 int error;
1705
1706 mqueue_inode_cachep = kmem_cache_create("mqueue_inode_cache",
1707 sizeof(struct mqueue_inode_info), 0,
Vladimir Davydov5d097052016-01-14 15:18:21 -08001708 SLAB_HWCACHE_ALIGN|SLAB_ACCOUNT, init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709 if (mqueue_inode_cachep == NULL)
1710 return -ENOMEM;
1711
André Goddard Rosa2329e3922010-02-23 04:04:27 -03001712 /* ignore failures - they are not fatal */
Serge E. Hallynbdc8e5f2009-04-06 19:01:11 -07001713 mq_sysctl_table = mq_register_sysctl_table();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714
1715 error = register_filesystem(&mqueue_fs_type);
1716 if (error)
1717 goto out_sysctl;
1718
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001719 spin_lock_init(&mq_lock);
1720
Al Viro6f686572011-12-09 00:38:50 -05001721 error = mq_init_ns(&init_ipc_ns);
1722 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723 goto out_filesystem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725 return 0;
1726
1727out_filesystem:
1728 unregister_filesystem(&mqueue_fs_type);
1729out_sysctl:
1730 if (mq_sysctl_table)
1731 unregister_sysctl_table(mq_sysctl_table);
Alexey Dobriyan1a1d92c2006-09-27 01:49:40 -07001732 kmem_cache_destroy(mqueue_inode_cachep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 return error;
1734}
1735
Davidlohr Bueso6d08a252014-04-07 15:39:18 -07001736device_initcall(init_mqueue_fs);