blob: 8c6a04111fded6932717336fc9062cc161199009 [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
Linus Torvalds1da177e2005-04-16 15:20:36 -070066struct ext_wait_queue { /* queue of sleeping tasks */
67 struct task_struct *task;
68 struct list_head list;
69 struct msg_msg *msg; /* ptr of loaded message */
70 int state; /* one of STATE_* values */
71};
72
73struct mqueue_inode_info {
74 spinlock_t lock;
75 struct inode vfs_inode;
76 wait_queue_head_t wait_q;
77
Doug Ledfordd6629852012-05-31 16:26:35 -070078 struct rb_root msg_tree;
Doug Ledfordce2d52c2012-05-31 16:26:38 -070079 struct posix_msg_tree_node *node_cache;
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 struct mq_attr attr;
81
82 struct sigevent notify;
Manfred Spraul239521f2014-01-27 17:07:04 -080083 struct pid *notify_owner;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -080084 struct user_namespace *notify_user_ns;
Adrian Bunk338cec32005-09-10 00:26:54 -070085 struct user_struct *user; /* user who created, for accounting */
Linus Torvalds1da177e2005-04-16 15:20:36 -070086 struct sock *notify_sock;
87 struct sk_buff *notify_cookie;
88
89 /* for tasks waiting for free space and messages, respectively */
90 struct ext_wait_queue e_wait_q[2];
91
92 unsigned long qsize; /* size of queue in memory (sum of all msgs) */
93};
94
David Howells935c69122018-11-01 23:07:25 +000095static struct file_system_type mqueue_fs_type;
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -080096static const struct inode_operations mqueue_dir_inode_operations;
Arjan van de Ven9a321442007-02-12 00:55:35 -080097static const struct file_operations mqueue_file_operations;
Alexey Dobriyanb87221d2009-09-21 17:01:09 -070098static const struct super_operations mqueue_super_ops;
David Howells935c69122018-11-01 23:07:25 +000099static const struct fs_context_operations mqueue_fs_context_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100static void remove_notification(struct mqueue_inode_info *info);
101
Christoph Lametere18b8902006-12-06 20:33:20 -0800102static struct kmem_cache *mqueue_inode_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103
Manfred Spraul239521f2014-01-27 17:07:04 -0800104static struct ctl_table_header *mq_sysctl_table;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105
106static inline struct mqueue_inode_info *MQUEUE_I(struct inode *inode)
107{
108 return container_of(inode, struct mqueue_inode_info, vfs_inode);
109}
110
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700111/*
112 * This routine should be called with the mq_lock held.
113 */
114static inline struct ipc_namespace *__get_ns_from_inode(struct inode *inode)
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700115{
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700116 return get_ipc_ns(inode->i_sb->s_fs_info);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700117}
118
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700119static struct ipc_namespace *get_ns_from_inode(struct inode *inode)
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700120{
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700121 struct ipc_namespace *ns;
122
123 spin_lock(&mq_lock);
124 ns = __get_ns_from_inode(inode);
125 spin_unlock(&mq_lock);
126 return ns;
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700127}
128
Doug Ledfordd6629852012-05-31 16:26:35 -0700129/* Auxiliary functions to manipulate messages' list */
130static int msg_insert(struct msg_msg *msg, struct mqueue_inode_info *info)
131{
132 struct rb_node **p, *parent = NULL;
133 struct posix_msg_tree_node *leaf;
134
135 p = &info->msg_tree.rb_node;
136 while (*p) {
137 parent = *p;
138 leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node);
139
140 if (likely(leaf->priority == msg->m_type))
141 goto insert_msg;
142 else if (msg->m_type < leaf->priority)
143 p = &(*p)->rb_left;
144 else
145 p = &(*p)->rb_right;
146 }
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700147 if (info->node_cache) {
148 leaf = info->node_cache;
149 info->node_cache = NULL;
150 } else {
151 leaf = kmalloc(sizeof(*leaf), GFP_ATOMIC);
152 if (!leaf)
153 return -ENOMEM;
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700154 INIT_LIST_HEAD(&leaf->msg_list);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700155 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700156 leaf->priority = msg->m_type;
157 rb_link_node(&leaf->rb_node, parent, p);
158 rb_insert_color(&leaf->rb_node, &info->msg_tree);
Doug Ledfordd6629852012-05-31 16:26:35 -0700159insert_msg:
160 info->attr.mq_curmsgs++;
161 info->qsize += msg->m_ts;
162 list_add_tail(&msg->m_list, &leaf->msg_list);
163 return 0;
164}
165
166static inline struct msg_msg *msg_get(struct mqueue_inode_info *info)
167{
168 struct rb_node **p, *parent = NULL;
169 struct posix_msg_tree_node *leaf;
170 struct msg_msg *msg;
171
172try_again:
173 p = &info->msg_tree.rb_node;
174 while (*p) {
175 parent = *p;
176 /*
177 * During insert, low priorities go to the left and high to the
178 * right. On receive, we want the highest priorities first, so
179 * walk all the way to the right.
180 */
181 p = &(*p)->rb_right;
182 }
183 if (!parent) {
184 if (info->attr.mq_curmsgs) {
185 pr_warn_once("Inconsistency in POSIX message queue, "
186 "no tree element, but supposedly messages "
187 "should exist!\n");
188 info->attr.mq_curmsgs = 0;
189 }
190 return NULL;
191 }
192 leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700193 if (unlikely(list_empty(&leaf->msg_list))) {
Doug Ledfordd6629852012-05-31 16:26:35 -0700194 pr_warn_once("Inconsistency in POSIX message queue, "
195 "empty leaf node but we haven't implemented "
196 "lazy leaf delete!\n");
197 rb_erase(&leaf->rb_node, &info->msg_tree);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700198 if (info->node_cache) {
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700199 kfree(leaf);
200 } else {
201 info->node_cache = leaf;
202 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700203 goto try_again;
204 } else {
205 msg = list_first_entry(&leaf->msg_list,
206 struct msg_msg, m_list);
207 list_del(&msg->m_list);
208 if (list_empty(&leaf->msg_list)) {
209 rb_erase(&leaf->rb_node, &info->msg_tree);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700210 if (info->node_cache) {
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700211 kfree(leaf);
212 } else {
213 info->node_cache = leaf;
214 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700215 }
216 }
217 info->attr.mq_curmsgs--;
218 info->qsize -= msg->m_ts;
219 return msg;
220}
221
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700222static struct inode *mqueue_get_inode(struct super_block *sb,
Al Viro1b9d5ff72011-07-24 14:18:20 -0400223 struct ipc_namespace *ipc_ns, umode_t mode,
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700224 struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225{
David Howells86a264a2008-11-14 10:39:18 +1100226 struct user_struct *u = current_user();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 struct inode *inode;
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700228 int ret = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229
230 inode = new_inode(sb);
Jiri Slaby04715202011-07-26 16:08:46 -0700231 if (!inode)
232 goto err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233
Jiri Slaby04715202011-07-26 16:08:46 -0700234 inode->i_ino = get_next_ino();
235 inode->i_mode = mode;
236 inode->i_uid = current_fsuid();
237 inode->i_gid = current_fsgid();
Deepa Dinamani078cd822016-09-14 07:48:04 -0700238 inode->i_mtime = inode->i_ctime = inode->i_atime = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
Jiri Slaby04715202011-07-26 16:08:46 -0700240 if (S_ISREG(mode)) {
241 struct mqueue_inode_info *info;
Doug Ledfordd6629852012-05-31 16:26:35 -0700242 unsigned long mq_bytes, mq_treesize;
André Goddard Rosac8308b12010-02-23 04:04:23 -0300243
Jiri Slaby04715202011-07-26 16:08:46 -0700244 inode->i_fop = &mqueue_file_operations;
245 inode->i_size = FILENT_SIZE;
246 /* mqueue specific info */
247 info = MQUEUE_I(inode);
248 spin_lock_init(&info->lock);
249 init_waitqueue_head(&info->wait_q);
250 INIT_LIST_HEAD(&info->e_wait_q[0].list);
251 INIT_LIST_HEAD(&info->e_wait_q[1].list);
252 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800253 info->notify_user_ns = NULL;
Jiri Slaby04715202011-07-26 16:08:46 -0700254 info->qsize = 0;
255 info->user = NULL; /* set when all is ok */
Doug Ledfordd6629852012-05-31 16:26:35 -0700256 info->msg_tree = RB_ROOT;
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700257 info->node_cache = NULL;
Jiri Slaby04715202011-07-26 16:08:46 -0700258 memset(&info->attr, 0, sizeof(info->attr));
KOSAKI Motohirocef01842012-05-31 16:26:33 -0700259 info->attr.mq_maxmsg = min(ipc_ns->mq_msg_max,
260 ipc_ns->mq_msg_default);
261 info->attr.mq_msgsize = min(ipc_ns->mq_msgsize_max,
262 ipc_ns->mq_msgsize_default);
Jiri Slaby04715202011-07-26 16:08:46 -0700263 if (attr) {
264 info->attr.mq_maxmsg = attr->mq_maxmsg;
265 info->attr.mq_msgsize = attr->mq_msgsize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700267 /*
268 * We used to allocate a static array of pointers and account
269 * the size of that array as well as one msg_msg struct per
270 * possible message into the queue size. That's no longer
271 * accurate as the queue is now an rbtree and will grow and
272 * shrink depending on usage patterns. We can, however, still
273 * account one msg_msg struct per message, but the nodes are
274 * allocated depending on priority usage, and most programs
275 * only use one, or a handful, of priorities. However, since
276 * this is pinned memory, we need to assume worst case, so
277 * that means the min(mq_maxmsg, max_priorities) * struct
278 * posix_msg_tree_node.
279 */
Al Viro05c1b292017-12-01 17:43:43 -0500280
281 ret = -EINVAL;
282 if (info->attr.mq_maxmsg <= 0 || info->attr.mq_msgsize <= 0)
283 goto out_inode;
284 if (capable(CAP_SYS_RESOURCE)) {
285 if (info->attr.mq_maxmsg > HARD_MSGMAX ||
286 info->attr.mq_msgsize > HARD_MSGSIZEMAX)
287 goto out_inode;
288 } else {
289 if (info->attr.mq_maxmsg > ipc_ns->mq_msg_max ||
290 info->attr.mq_msgsize > ipc_ns->mq_msgsize_max)
291 goto out_inode;
292 }
293 ret = -EOVERFLOW;
294 /* check for overflow */
295 if (info->attr.mq_msgsize > ULONG_MAX/info->attr.mq_maxmsg)
296 goto out_inode;
Doug Ledfordd6629852012-05-31 16:26:35 -0700297 mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) +
298 min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) *
299 sizeof(struct posix_msg_tree_node);
Al Viro05c1b292017-12-01 17:43:43 -0500300 mq_bytes = info->attr.mq_maxmsg * info->attr.mq_msgsize;
301 if (mq_bytes + mq_treesize < mq_bytes)
302 goto out_inode;
303 mq_bytes += mq_treesize;
Jiri Slaby04715202011-07-26 16:08:46 -0700304 spin_lock(&mq_lock);
305 if (u->mq_bytes + mq_bytes < u->mq_bytes ||
Davidlohr Bueso2a4e64b2012-01-20 14:34:01 -0800306 u->mq_bytes + mq_bytes > rlimit(RLIMIT_MSGQUEUE)) {
Jiri Slaby04715202011-07-26 16:08:46 -0700307 spin_unlock(&mq_lock);
308 /* mqueue_evict_inode() releases info->messages */
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700309 ret = -EMFILE;
Jiri Slaby04715202011-07-26 16:08:46 -0700310 goto out_inode;
311 }
312 u->mq_bytes += mq_bytes;
313 spin_unlock(&mq_lock);
314
315 /* all is ok */
316 info->user = get_uid(u);
317 } else if (S_ISDIR(mode)) {
318 inc_nlink(inode);
319 /* Some things misbehave if size == 0 on a directory */
320 inode->i_size = 2 * DIRENT_SIZE;
321 inode->i_op = &mqueue_dir_inode_operations;
322 inode->i_fop = &simple_dir_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 }
Jiri Slaby04715202011-07-26 16:08:46 -0700324
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325 return inode;
326out_inode:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 iput(inode);
Jiri Slaby04715202011-07-26 16:08:46 -0700328err:
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700329 return ERR_PTR(ret);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330}
331
David Howells935c69122018-11-01 23:07:25 +0000332static int mqueue_fill_super(struct super_block *sb, struct fs_context *fc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333{
334 struct inode *inode;
Eric W. Biedermancfb2f6f2018-03-24 11:28:14 -0500335 struct ipc_namespace *ns = sb->s_fs_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336
Eric W. Biedermana2982cc2016-06-09 15:34:02 -0500337 sb->s_iflags |= SB_I_NOEXEC | SB_I_NODEV;
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300338 sb->s_blocksize = PAGE_SIZE;
339 sb->s_blocksize_bits = PAGE_SHIFT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 sb->s_magic = MQUEUE_MAGIC;
341 sb->s_op = &mqueue_super_ops;
342
Al Viro48fde702012-01-08 22:15:13 -0500343 inode = mqueue_get_inode(sb, ns, S_IFDIR | S_ISVTX | S_IRWXUGO, NULL);
344 if (IS_ERR(inode))
345 return PTR_ERR(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346
Al Viro48fde702012-01-08 22:15:13 -0500347 sb->s_root = d_make_root(inode);
348 if (!sb->s_root)
349 return -ENOMEM;
350 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351}
352
David Howells935c69122018-11-01 23:07:25 +0000353static int mqueue_get_tree(struct fs_context *fc)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354{
David Howells935c69122018-11-01 23:07:25 +0000355 struct mqueue_fs_context *ctx = fc->fs_private;
356
357 put_user_ns(fc->user_ns);
358 fc->user_ns = get_user_ns(ctx->ipc_ns->user_ns);
359 fc->s_fs_info = ctx->ipc_ns;
360 return vfs_get_super(fc, vfs_get_keyed_super, mqueue_fill_super);
361}
362
363static void mqueue_fs_context_free(struct fs_context *fc)
364{
365 struct mqueue_fs_context *ctx = fc->fs_private;
366
367 if (ctx->ipc_ns)
368 put_ipc_ns(ctx->ipc_ns);
369 kfree(ctx);
370}
371
372static int mqueue_init_fs_context(struct fs_context *fc)
373{
374 struct mqueue_fs_context *ctx;
375
376 ctx = kzalloc(sizeof(struct mqueue_fs_context), GFP_KERNEL);
377 if (!ctx)
378 return -ENOMEM;
379
380 ctx->ipc_ns = get_ipc_ns(current->nsproxy->ipc_ns);
381 fc->fs_private = ctx;
382 fc->ops = &mqueue_fs_context_ops;
383 return 0;
384}
385
386static struct vfsmount *mq_create_mount(struct ipc_namespace *ns)
387{
388 struct mqueue_fs_context *ctx;
389 struct fs_context *fc;
390 struct vfsmount *mnt;
391
392 fc = fs_context_for_mount(&mqueue_fs_type, SB_KERNMOUNT);
393 if (IS_ERR(fc))
394 return ERR_CAST(fc);
395
396 ctx = fc->fs_private;
397 put_ipc_ns(ctx->ipc_ns);
398 ctx->ipc_ns = get_ipc_ns(ns);
399
400 mnt = fc_mount(fc);
401 put_fs_context(fc);
402 return mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403}
404
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700405static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406{
407 struct mqueue_inode_info *p = (struct mqueue_inode_info *) foo;
408
Christoph Lametera35afb82007-05-16 22:10:57 -0700409 inode_init_once(&p->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410}
411
412static struct inode *mqueue_alloc_inode(struct super_block *sb)
413{
414 struct mqueue_inode_info *ei;
415
Christoph Lametere94b1762006-12-06 20:33:17 -0800416 ei = kmem_cache_alloc(mqueue_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 if (!ei)
418 return NULL;
419 return &ei->vfs_inode;
420}
421
Al Viro015d7952019-04-15 22:30:30 -0400422static void mqueue_free_inode(struct inode *inode)
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100423{
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100424 kmem_cache_free(mqueue_inode_cachep, MQUEUE_I(inode));
425}
426
Al Viro6d8af642010-06-05 16:29:45 -0400427static void mqueue_evict_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428{
429 struct mqueue_inode_info *info;
430 struct user_struct *user;
Doug Ledfordd6629852012-05-31 16:26:35 -0700431 unsigned long mq_bytes, mq_treesize;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700432 struct ipc_namespace *ipc_ns;
Li Rongqingd6a29462019-05-14 15:46:20 -0700433 struct msg_msg *msg, *nmsg;
434 LIST_HEAD(tmp_msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435
Jan Karadbd57682012-05-03 14:48:02 +0200436 clear_inode(inode);
Al Viro6d8af642010-06-05 16:29:45 -0400437
438 if (S_ISDIR(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 return;
Al Viro6d8af642010-06-05 16:29:45 -0400440
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700441 ipc_ns = get_ns_from_inode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 info = MQUEUE_I(inode);
443 spin_lock(&info->lock);
Doug Ledfordd6629852012-05-31 16:26:35 -0700444 while ((msg = msg_get(info)) != NULL)
Li Rongqingd6a29462019-05-14 15:46:20 -0700445 list_add_tail(&msg->m_list, &tmp_msg);
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700446 kfree(info->node_cache);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 spin_unlock(&info->lock);
448
Li Rongqingd6a29462019-05-14 15:46:20 -0700449 list_for_each_entry_safe(msg, nmsg, &tmp_msg, m_list) {
450 list_del(&msg->m_list);
451 free_msg(msg);
452 }
453
André Goddard Rosa8834cf72010-02-23 04:04:24 -0300454 /* Total amount of bytes accounted for the mqueue */
Doug Ledfordd6629852012-05-31 16:26:35 -0700455 mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) +
456 min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) *
457 sizeof(struct posix_msg_tree_node);
458
459 mq_bytes = mq_treesize + (info->attr.mq_maxmsg *
460 info->attr.mq_msgsize);
461
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 user = info->user;
463 if (user) {
464 spin_lock(&mq_lock);
465 user->mq_bytes -= mq_bytes;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700466 /*
467 * get_ns_from_inode() ensures that the
468 * (ipc_ns = sb->s_fs_info) is either a valid ipc_ns
469 * to which we now hold a reference, or it is NULL.
470 * We can't put it here under mq_lock, though.
471 */
472 if (ipc_ns)
473 ipc_ns->mq_queues_count--;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 spin_unlock(&mq_lock);
475 free_uid(user);
476 }
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700477 if (ipc_ns)
478 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479}
480
Al Viroeecec192017-12-01 17:26:05 -0500481static int mqueue_create_attr(struct dentry *dentry, umode_t mode, void *arg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482{
Al Viroeecec192017-12-01 17:26:05 -0500483 struct inode *dir = dentry->d_parent->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 struct inode *inode;
Al Viroeecec192017-12-01 17:26:05 -0500485 struct mq_attr *attr = arg;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 int error;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700487 struct ipc_namespace *ipc_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488
489 spin_lock(&mq_lock);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700490 ipc_ns = __get_ns_from_inode(dir);
491 if (!ipc_ns) {
492 error = -EACCES;
493 goto out_unlock;
494 }
Davidlohr Buesof3713fd2014-02-25 15:01:45 -0800495
496 if (ipc_ns->mq_queues_count >= ipc_ns->mq_queues_max &&
497 !capable(CAP_SYS_RESOURCE)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 error = -ENOSPC;
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700499 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 }
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700501 ipc_ns->mq_queues_count++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 spin_unlock(&mq_lock);
503
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700504 inode = mqueue_get_inode(dir->i_sb, ipc_ns, mode, attr);
Jiri Slabyd40dcdb2011-07-26 16:08:47 -0700505 if (IS_ERR(inode)) {
506 error = PTR_ERR(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 spin_lock(&mq_lock);
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700508 ipc_ns->mq_queues_count--;
509 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 }
511
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700512 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 dir->i_size += DIRENT_SIZE;
Deepa Dinamani078cd822016-09-14 07:48:04 -0700514 dir->i_ctime = dir->i_mtime = dir->i_atime = current_time(dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
516 d_instantiate(dentry, inode);
517 dget(dentry);
518 return 0;
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700519out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 spin_unlock(&mq_lock);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700521 if (ipc_ns)
522 put_ipc_ns(ipc_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 return error;
524}
525
Al Viroeecec192017-12-01 17:26:05 -0500526static int mqueue_create(struct inode *dir, struct dentry *dentry,
527 umode_t mode, bool excl)
528{
529 return mqueue_create_attr(dentry, mode, NULL);
530}
531
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532static int mqueue_unlink(struct inode *dir, struct dentry *dentry)
533{
David Howells75c3cfa2015-03-17 22:26:12 +0000534 struct inode *inode = d_inode(dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535
Deepa Dinamani078cd822016-09-14 07:48:04 -0700536 dir->i_ctime = dir->i_mtime = dir->i_atime = current_time(dir);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 dir->i_size -= DIRENT_SIZE;
Manfred Spraul239521f2014-01-27 17:07:04 -0800538 drop_nlink(inode);
539 dput(dentry);
540 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541}
542
543/*
544* This is routine for system read from queue file.
545* To avoid mess with doing here some sort of mq_receive we allow
546* to read only queue size & notification info (the only values
547* that are interesting from user point of view and aren't accessible
548* through std routines)
549*/
550static ssize_t mqueue_read_file(struct file *filp, char __user *u_data,
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700551 size_t count, loff_t *off)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552{
Al Viro496ad9a2013-01-23 17:07:38 -0500553 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 char buffer[FILENT_SIZE];
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700555 ssize_t ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556
557 spin_lock(&info->lock);
558 snprintf(buffer, sizeof(buffer),
559 "QSIZE:%-10lu NOTIFY:%-5d SIGNO:%-5d NOTIFY_PID:%-6d\n",
560 info->qsize,
561 info->notify_owner ? info->notify.sigev_notify : 0,
562 (info->notify_owner &&
563 info->notify.sigev_notify == SIGEV_SIGNAL) ?
564 info->notify.sigev_signo : 0,
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -0800565 pid_vnr(info->notify_owner));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 spin_unlock(&info->lock);
567 buffer[sizeof(buffer)-1] = '\0';
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700569 ret = simple_read_from_buffer(u_data, count, off, buffer,
570 strlen(buffer));
571 if (ret <= 0)
572 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573
Deepa Dinamani078cd822016-09-14 07:48:04 -0700574 file_inode(filp)->i_atime = file_inode(filp)->i_ctime = current_time(file_inode(filp));
Akinobu Mitaf1a43f92008-07-25 01:48:07 -0700575 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576}
577
Miklos Szeredi75e1fcc2006-06-23 02:05:12 -0700578static int mqueue_flush_file(struct file *filp, fl_owner_t id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579{
Al Viro496ad9a2013-01-23 17:07:38 -0500580 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581
582 spin_lock(&info->lock);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700583 if (task_tgid(current) == info->notify_owner)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 remove_notification(info);
585
586 spin_unlock(&info->lock);
587 return 0;
588}
589
Al Viro9dd95742017-07-03 00:42:43 -0400590static __poll_t mqueue_poll_file(struct file *filp, struct poll_table_struct *poll_tab)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591{
Al Viro496ad9a2013-01-23 17:07:38 -0500592 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
Al Viro9dd95742017-07-03 00:42:43 -0400593 __poll_t retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594
595 poll_wait(filp, &info->wait_q, poll_tab);
596
597 spin_lock(&info->lock);
598 if (info->attr.mq_curmsgs)
Linus Torvaldsa9a08842018-02-11 14:34:03 -0800599 retval = EPOLLIN | EPOLLRDNORM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
601 if (info->attr.mq_curmsgs < info->attr.mq_maxmsg)
Linus Torvaldsa9a08842018-02-11 14:34:03 -0800602 retval |= EPOLLOUT | EPOLLWRNORM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 spin_unlock(&info->lock);
604
605 return retval;
606}
607
608/* Adds current to info->e_wait_q[sr] before element with smaller prio */
609static void wq_add(struct mqueue_inode_info *info, int sr,
610 struct ext_wait_queue *ewp)
611{
612 struct ext_wait_queue *walk;
613
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 list_for_each_entry(walk, &info->e_wait_q[sr].list, list) {
Jonathan Haws68e34f42018-02-06 15:40:52 -0800615 if (walk->task->prio <= current->prio) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 list_add_tail(&ewp->list, &walk->list);
617 return;
618 }
619 }
620 list_add_tail(&ewp->list, &info->e_wait_q[sr].list);
621}
622
623/*
624 * Puts current task to sleep. Caller must hold queue lock. After return
625 * lock isn't held.
626 * sr: SEND or RECV
627 */
628static int wq_sleep(struct mqueue_inode_info *info, int sr,
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200629 ktime_t *timeout, struct ext_wait_queue *ewp)
Luc Van Oostenryckeac0b1c2017-02-27 14:28:21 -0800630 __releases(&info->lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631{
632 int retval;
633 signed long time;
634
635 wq_add(info, sr, ewp);
636
637 for (;;) {
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700638 __set_current_state(TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639
640 spin_unlock(&info->lock);
Wanlong Gao32ea8452011-10-31 17:06:35 -0700641 time = schedule_hrtimeout_range_clock(timeout, 0,
642 HRTIMER_MODE_ABS, CLOCK_REALTIME);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 if (ewp->state == STATE_READY) {
645 retval = 0;
646 goto out;
647 }
648 spin_lock(&info->lock);
649 if (ewp->state == STATE_READY) {
650 retval = 0;
651 goto out_unlock;
652 }
653 if (signal_pending(current)) {
654 retval = -ERESTARTSYS;
655 break;
656 }
657 if (time == 0) {
658 retval = -ETIMEDOUT;
659 break;
660 }
661 }
662 list_del(&ewp->list);
663out_unlock:
664 spin_unlock(&info->lock);
665out:
666 return retval;
667}
668
669/*
670 * Returns waiting task that should be serviced first or NULL if none exists
671 */
672static struct ext_wait_queue *wq_get_first_waiter(
673 struct mqueue_inode_info *info, int sr)
674{
675 struct list_head *ptr;
676
677 ptr = info->e_wait_q[sr].list.prev;
678 if (ptr == &info->e_wait_q[sr].list)
679 return NULL;
680 return list_entry(ptr, struct ext_wait_queue, list);
681}
682
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683
684static inline void set_cookie(struct sk_buff *skb, char code)
685{
Manfred Spraul239521f2014-01-27 17:07:04 -0800686 ((char *)skb->data)[NOTIFY_COOKIE_LEN-1] = code;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687}
688
689/*
690 * The next function is only to split too long sys_mq_timedsend
691 */
692static void __do_notify(struct mqueue_inode_info *info)
693{
694 /* notification
695 * invoked when there is registered process and there isn't process
696 * waiting synchronously for message AND state of queue changed from
697 * empty to not empty. Here we are sure that no one is waiting
698 * synchronously. */
699 if (info->notify_owner &&
700 info->attr.mq_curmsgs == 1) {
Eric W. Biedermanae7795b2018-09-25 11:27:20 +0200701 struct kernel_siginfo sig_i;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 switch (info->notify.sigev_notify) {
703 case SIGEV_NONE:
704 break;
705 case SIGEV_SIGNAL:
706 /* sends signal */
707
Eric W. Biedermanfaf1f222018-01-05 17:27:42 -0600708 clear_siginfo(&sig_i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 sig_i.si_signo = info->notify.sigev_signo;
710 sig_i.si_errno = 0;
711 sig_i.si_code = SI_MESGQ;
712 sig_i.si_value = info->notify.sigev_value;
Serge E. Hallyn6b550f92012-01-10 15:11:37 -0800713 /* map current pid/uid into info->owner's namespaces */
714 rcu_read_lock();
Sukadev Bhattiprolua6684992009-01-07 18:08:50 -0800715 sig_i.si_pid = task_tgid_nr_ns(current,
716 ns_of_pid(info->notify_owner));
Eric W. Biederman76b6db02012-03-14 15:24:19 -0700717 sig_i.si_uid = from_kuid_munged(info->notify_user_ns, current_uid());
Serge E. Hallyn6b550f92012-01-10 15:11:37 -0800718 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700720 kill_pid_info(info->notify.sigev_signo,
721 &sig_i, info->notify_owner);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722 break;
723 case SIGEV_THREAD:
724 set_cookie(info->notify_cookie, NOTIFY_WOKENUP);
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700725 netlink_sendskb(info->notify_sock, info->notify_cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 break;
727 }
728 /* after notification unregisters process */
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700729 put_pid(info->notify_owner);
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800730 put_user_ns(info->notify_user_ns);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700731 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800732 info->notify_user_ns = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 }
734 wake_up(&info->wait_q);
735}
736
Arnd Bergmann21fc5382018-04-13 13:58:00 +0200737static int prepare_timeout(const struct __kernel_timespec __user *u_abs_timeout,
Deepa Dinamanib9047722017-08-02 19:51:11 -0700738 struct timespec64 *ts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739{
Deepa Dinamanib9047722017-08-02 19:51:11 -0700740 if (get_timespec64(ts, u_abs_timeout))
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200741 return -EFAULT;
Deepa Dinamanib9047722017-08-02 19:51:11 -0700742 if (!timespec64_valid(ts))
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200743 return -EINVAL;
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200744 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745}
746
747static void remove_notification(struct mqueue_inode_info *info)
748{
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700749 if (info->notify_owner != NULL &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 info->notify.sigev_notify == SIGEV_THREAD) {
751 set_cookie(info->notify_cookie, NOTIFY_REMOVED);
Denis V. Lunev7ee015e2007-10-10 21:14:03 -0700752 netlink_sendskb(info->notify_sock, info->notify_cookie);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700754 put_pid(info->notify_owner);
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800755 put_user_ns(info->notify_user_ns);
Cedric Le Goatera03fcb72006-10-02 02:17:26 -0700756 info->notify_owner = NULL;
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -0800757 info->notify_user_ns = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758}
759
Al Viro066cc812017-12-01 17:51:39 -0500760static int prepare_open(struct dentry *dentry, int oflag, int ro,
761 umode_t mode, struct filename *name,
Serge E. Hallyn614b84c2009-04-06 19:01:08 -0700762 struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763{
David Howells745ca242008-11-14 10:39:22 +1100764 static const int oflag2acc[O_ACCMODE] = { MAY_READ, MAY_WRITE,
765 MAY_READ | MAY_WRITE };
Al Viro765927b2012-06-26 21:58:53 +0400766 int acc;
Al Viro066cc812017-12-01 17:51:39 -0500767
Al Viro9b20d7f2017-12-01 17:57:02 -0500768 if (d_really_is_negative(dentry)) {
769 if (!(oflag & O_CREAT))
Al Viro066cc812017-12-01 17:51:39 -0500770 return -ENOENT;
Al Viro9b20d7f2017-12-01 17:57:02 -0500771 if (ro)
772 return ro;
773 audit_inode_parent_hidden(name, dentry->d_parent);
774 return vfs_mkobj(dentry, mode & ~current_umask(),
775 mqueue_create_attr, attr);
Al Viro066cc812017-12-01 17:51:39 -0500776 }
Al Viro9b20d7f2017-12-01 17:57:02 -0500777 /* it already existed */
778 audit_inode(name, dentry, 0);
779 if ((oflag & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL))
780 return -EEXIST;
Al Viro765927b2012-06-26 21:58:53 +0400781 if ((oflag & O_ACCMODE) == (O_RDWR | O_WRONLY))
Al Viroaf4a5372017-12-01 17:34:22 -0500782 return -EINVAL;
Al Viro765927b2012-06-26 21:58:53 +0400783 acc = oflag2acc[oflag & O_ACCMODE];
Al Viro066cc812017-12-01 17:51:39 -0500784 return inode_permission(d_inode(dentry), acc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785}
786
Al Viro0d060602017-06-27 21:32:36 -0400787static int do_mq_open(const char __user *u_name, int oflag, umode_t mode,
788 struct mq_attr *attr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789{
Eric W. Biedermancfb2f6f2018-03-24 11:28:14 -0500790 struct vfsmount *mnt = current->nsproxy->ipc_ns->mq_mnt;
791 struct dentry *root = mnt->mnt_root;
Jeff Layton91a27b22012-10-10 15:25:28 -0400792 struct filename *name;
Al Viroa713fd72017-12-01 18:01:09 -0500793 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 int fd, error;
Al Viro312b90f2012-08-06 10:18:17 +0400795 int ro;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796
Al Viro0d060602017-06-27 21:32:36 -0400797 audit_mq_open(oflag, mode, attr);
George C. Wilson20ca73b2006-05-24 16:09:55 -0500798
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 if (IS_ERR(name = getname(u_name)))
800 return PTR_ERR(name);
801
Ulrich Drepper269f2132008-05-03 15:28:45 -0400802 fd = get_unused_fd_flags(O_CLOEXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 if (fd < 0)
804 goto out_putname;
805
Al Viro312b90f2012-08-06 10:18:17 +0400806 ro = mnt_want_write(mnt); /* we'll drop it in any case */
Al Viro59551022016-01-22 15:40:57 -0500807 inode_lock(d_inode(root));
Jeff Layton91a27b22012-10-10 15:25:28 -0400808 path.dentry = lookup_one_len(name->name, root, strlen(name->name));
Al Viro765927b2012-06-26 21:58:53 +0400809 if (IS_ERR(path.dentry)) {
810 error = PTR_ERR(path.dentry);
André Goddard Rosa4294a8e2010-02-23 04:04:28 -0300811 goto out_putfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 }
Al Viro312b90f2012-08-06 10:18:17 +0400813 path.mnt = mntget(mnt);
Al Viro066cc812017-12-01 17:51:39 -0500814 error = prepare_open(path.dentry, oflag, ro, mode, name, attr);
815 if (!error) {
816 struct file *file = dentry_open(&path, oflag, current_cred());
817 if (!IS_ERR(file))
818 fd_install(fd, file);
819 else
820 error = PTR_ERR(file);
Alexander Viro7c7dce92006-01-14 15:29:55 -0500821 }
Al Viro765927b2012-06-26 21:58:53 +0400822 path_put(&path);
Alexander Viro7c7dce92006-01-14 15:29:55 -0500823out_putfd:
Al Viro765927b2012-06-26 21:58:53 +0400824 if (error) {
825 put_unused_fd(fd);
826 fd = error;
827 }
Al Viro59551022016-01-22 15:40:57 -0500828 inode_unlock(d_inode(root));
Vladimir Davydov38d78e52013-03-22 15:04:51 -0700829 if (!ro)
830 mnt_drop_write(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831out_putname:
832 putname(name);
833 return fd;
834}
835
Al Viro0d060602017-06-27 21:32:36 -0400836SYSCALL_DEFINE4(mq_open, const char __user *, u_name, int, oflag, umode_t, mode,
837 struct mq_attr __user *, u_attr)
838{
839 struct mq_attr attr;
840 if (u_attr && copy_from_user(&attr, u_attr, sizeof(struct mq_attr)))
841 return -EFAULT;
842
843 return do_mq_open(u_name, oflag, mode, u_attr ? &attr : NULL);
844}
845
Heiko Carstensd5460c92009-01-14 14:14:27 +0100846SYSCALL_DEFINE1(mq_unlink, const char __user *, u_name)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847{
848 int err;
Jeff Layton91a27b22012-10-10 15:25:28 -0400849 struct filename *name;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 struct dentry *dentry;
851 struct inode *inode = NULL;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -0700852 struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns;
Al Viro312b90f2012-08-06 10:18:17 +0400853 struct vfsmount *mnt = ipc_ns->mq_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854
855 name = getname(u_name);
856 if (IS_ERR(name))
857 return PTR_ERR(name);
858
Jeff Layton79f65302013-07-08 15:59:36 -0700859 audit_inode_parent_hidden(name, mnt->mnt_root);
Al Viro312b90f2012-08-06 10:18:17 +0400860 err = mnt_want_write(mnt);
861 if (err)
862 goto out_name;
Al Viro59551022016-01-22 15:40:57 -0500863 inode_lock_nested(d_inode(mnt->mnt_root), I_MUTEX_PARENT);
Jeff Layton91a27b22012-10-10 15:25:28 -0400864 dentry = lookup_one_len(name->name, mnt->mnt_root,
865 strlen(name->name));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 if (IS_ERR(dentry)) {
867 err = PTR_ERR(dentry);
868 goto out_unlock;
869 }
870
David Howells75c3cfa2015-03-17 22:26:12 +0000871 inode = d_inode(dentry);
Al Viro312b90f2012-08-06 10:18:17 +0400872 if (!inode) {
873 err = -ENOENT;
874 } else {
Al Viro7de9c6ee2010-10-23 11:11:40 -0400875 ihold(inode);
David Howells75c3cfa2015-03-17 22:26:12 +0000876 err = vfs_unlink(d_inode(dentry->d_parent), dentry, NULL);
Al Viro312b90f2012-08-06 10:18:17 +0400877 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 dput(dentry);
879
880out_unlock:
Al Viro59551022016-01-22 15:40:57 -0500881 inode_unlock(d_inode(mnt->mnt_root));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882 if (inode)
883 iput(inode);
Al Viro312b90f2012-08-06 10:18:17 +0400884 mnt_drop_write(mnt);
885out_name:
886 putname(name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887
888 return err;
889}
890
891/* Pipelined send and receive functions.
892 *
893 * If a receiver finds no waiting message, then it registers itself in the
894 * list of waiting receivers. A sender checks that list before adding the new
895 * message into the message array. If there is a waiting receiver, then it
896 * bypasses the message array and directly hands the message over to the
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700897 * receiver. The receiver accepts the message and returns without grabbing the
898 * queue spinlock:
899 *
900 * - Set pointer to message.
901 * - Queue the receiver task for later wakeup (without the info->lock).
902 * - Update its state to STATE_READY. Now the receiver can continue.
903 * - Wake up the process after the lock is dropped. Should the process wake up
904 * before this wakeup (due to a timeout or a signal) it will either see
905 * STATE_READY and continue or acquire the lock to check the state again.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 *
907 * The same algorithm is used for senders.
908 */
909
910/* pipelined_send() - send a message directly to the task waiting in
911 * sys_mq_timedreceive() (without inserting message into a queue).
912 */
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700913static inline void pipelined_send(struct wake_q_head *wake_q,
914 struct mqueue_inode_info *info,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 struct msg_msg *message,
916 struct ext_wait_queue *receiver)
917{
918 receiver->msg = message;
919 list_del(&receiver->list);
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700920 wake_q_add(wake_q, receiver->task);
921 /*
922 * Rely on the implicit cmpxchg barrier from wake_q_add such
923 * that we can ensure that updating receiver->state is the last
924 * write operation: As once set, the receiver can continue,
925 * and if we don't have the reference count from the wake_q,
926 * yet, at that point we can later have a use-after-free
927 * condition and bogus wakeup.
928 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 receiver->state = STATE_READY;
930}
931
932/* pipelined_receive() - if there is task waiting in sys_mq_timedsend()
933 * gets its message and put to the queue (we have one free place for sure). */
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700934static inline void pipelined_receive(struct wake_q_head *wake_q,
935 struct mqueue_inode_info *info)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936{
937 struct ext_wait_queue *sender = wq_get_first_waiter(info, SEND);
938
939 if (!sender) {
940 /* for poll */
941 wake_up_interruptible(&info->wait_q);
942 return;
943 }
Doug Ledfordd6629852012-05-31 16:26:35 -0700944 if (msg_insert(sender->msg, info))
945 return;
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700946
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947 list_del(&sender->list);
Davidlohr Buesofa6004a2015-05-04 07:02:46 -0700948 wake_q_add(wake_q, sender->task);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 sender->state = STATE_READY;
950}
951
Al Viro0d060602017-06-27 21:32:36 -0400952static int do_mq_timedsend(mqd_t mqdes, const char __user *u_msg_ptr,
953 size_t msg_len, unsigned int msg_prio,
Deepa Dinamanib9047722017-08-02 19:51:11 -0700954 struct timespec64 *ts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955{
Al Viro2903ff02012-08-28 12:52:22 -0400956 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 struct inode *inode;
958 struct ext_wait_queue wait;
959 struct ext_wait_queue *receiver;
960 struct msg_msg *msg_ptr;
961 struct mqueue_inode_info *info;
Carsten Emde9ca7d8e2010-04-02 22:40:20 +0200962 ktime_t expires, *timeout = NULL;
Doug Ledfordce2d52c2012-05-31 16:26:38 -0700963 struct posix_msg_tree_node *new_leaf = NULL;
Al Viro2903ff02012-08-28 12:52:22 -0400964 int ret = 0;
Waiman Long194a6b52016-11-17 11:46:38 -0500965 DEFINE_WAKE_Q(wake_q);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966
967 if (unlikely(msg_prio >= (unsigned long) MQ_PRIO_MAX))
968 return -EINVAL;
969
Al Viro0d060602017-06-27 21:32:36 -0400970 if (ts) {
Deepa Dinamanib9047722017-08-02 19:51:11 -0700971 expires = timespec64_to_ktime(*ts);
Al Viro0d060602017-06-27 21:32:36 -0400972 timeout = &expires;
973 }
974
975 audit_mq_sendrecv(mqdes, msg_len, msg_prio, ts);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976
Al Viro2903ff02012-08-28 12:52:22 -0400977 f = fdget(mqdes);
978 if (unlikely(!f.file)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300979 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300981 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982
Al Viro496ad9a2013-01-23 17:07:38 -0500983 inode = file_inode(f.file);
Al Viro2903ff02012-08-28 12:52:22 -0400984 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300985 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300987 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988 info = MQUEUE_I(inode);
Al Viro9f45f5b2014-10-31 17:44:57 -0400989 audit_file(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990
Al Viro2903ff02012-08-28 12:52:22 -0400991 if (unlikely(!(f.file->f_mode & FMODE_WRITE))) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300992 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -0300994 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995
996 if (unlikely(msg_len > info->attr.mq_msgsize)) {
997 ret = -EMSGSIZE;
998 goto out_fput;
999 }
1000
1001 /* First try to allocate memory, before doing anything with
1002 * existing queues. */
1003 msg_ptr = load_msg(u_msg_ptr, msg_len);
1004 if (IS_ERR(msg_ptr)) {
1005 ret = PTR_ERR(msg_ptr);
1006 goto out_fput;
1007 }
1008 msg_ptr->m_ts = msg_len;
1009 msg_ptr->m_type = msg_prio;
1010
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001011 /*
1012 * msg_insert really wants us to have a valid, spare node struct so
1013 * it doesn't have to kmalloc a GFP_ATOMIC allocation, but it will
1014 * fall back to that if necessary.
1015 */
1016 if (!info->node_cache)
1017 new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL);
1018
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 spin_lock(&info->lock);
1020
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001021 if (!info->node_cache && new_leaf) {
1022 /* Save our speculative allocation into the cache */
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001023 INIT_LIST_HEAD(&new_leaf->msg_list);
1024 info->node_cache = new_leaf;
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001025 new_leaf = NULL;
1026 } else {
1027 kfree(new_leaf);
1028 }
1029
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030 if (info->attr.mq_curmsgs == info->attr.mq_maxmsg) {
Al Viro2903ff02012-08-28 12:52:22 -04001031 if (f.file->f_flags & O_NONBLOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 ret = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 } else {
1034 wait.task = current;
1035 wait.msg = (void *) msg_ptr;
1036 wait.state = STATE_NONE;
1037 ret = wq_sleep(info, SEND, timeout, &wait);
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001038 /*
1039 * wq_sleep must be called with info->lock held, and
1040 * returns with the lock released
1041 */
1042 goto out_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 } else {
1045 receiver = wq_get_first_waiter(info, RECV);
1046 if (receiver) {
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001047 pipelined_send(&wake_q, info, msg_ptr, receiver);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 } else {
1049 /* adds message to the queue */
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001050 ret = msg_insert(msg_ptr, info);
1051 if (ret)
1052 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 __do_notify(info);
1054 }
1055 inode->i_atime = inode->i_mtime = inode->i_ctime =
Deepa Dinamani078cd822016-09-14 07:48:04 -07001056 current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 }
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001058out_unlock:
1059 spin_unlock(&info->lock);
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001060 wake_up_q(&wake_q);
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001061out_free:
1062 if (ret)
1063 free_msg(msg_ptr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04001065 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066out:
1067 return ret;
1068}
1069
Al Viro0d060602017-06-27 21:32:36 -04001070static int do_mq_timedreceive(mqd_t mqdes, char __user *u_msg_ptr,
1071 size_t msg_len, unsigned int __user *u_msg_prio,
Deepa Dinamanib9047722017-08-02 19:51:11 -07001072 struct timespec64 *ts)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074 ssize_t ret;
1075 struct msg_msg *msg_ptr;
Al Viro2903ff02012-08-28 12:52:22 -04001076 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077 struct inode *inode;
1078 struct mqueue_inode_info *info;
1079 struct ext_wait_queue wait;
Carsten Emde9ca7d8e2010-04-02 22:40:20 +02001080 ktime_t expires, *timeout = NULL;
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001081 struct posix_msg_tree_node *new_leaf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082
Al Viro0d060602017-06-27 21:32:36 -04001083 if (ts) {
Deepa Dinamanib9047722017-08-02 19:51:11 -07001084 expires = timespec64_to_ktime(*ts);
Carsten Emde9ca7d8e2010-04-02 22:40:20 +02001085 timeout = &expires;
Al Viroc32c8af2008-12-14 03:46:48 -05001086 }
George C. Wilson20ca73b2006-05-24 16:09:55 -05001087
Al Viro0d060602017-06-27 21:32:36 -04001088 audit_mq_sendrecv(mqdes, msg_len, 0, ts);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089
Al Viro2903ff02012-08-28 12:52:22 -04001090 f = fdget(mqdes);
1091 if (unlikely(!f.file)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001092 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001094 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095
Al Viro496ad9a2013-01-23 17:07:38 -05001096 inode = file_inode(f.file);
Al Viro2903ff02012-08-28 12:52:22 -04001097 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001098 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001100 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 info = MQUEUE_I(inode);
Al Viro9f45f5b2014-10-31 17:44:57 -04001102 audit_file(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103
Al Viro2903ff02012-08-28 12:52:22 -04001104 if (unlikely(!(f.file->f_mode & FMODE_READ))) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001105 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001107 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108
1109 /* checks if buffer is big enough */
1110 if (unlikely(msg_len < info->attr.mq_msgsize)) {
1111 ret = -EMSGSIZE;
1112 goto out_fput;
1113 }
1114
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001115 /*
1116 * msg_insert really wants us to have a valid, spare node struct so
1117 * it doesn't have to kmalloc a GFP_ATOMIC allocation, but it will
1118 * fall back to that if necessary.
1119 */
1120 if (!info->node_cache)
1121 new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL);
1122
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 spin_lock(&info->lock);
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001124
1125 if (!info->node_cache && new_leaf) {
1126 /* Save our speculative allocation into the cache */
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001127 INIT_LIST_HEAD(&new_leaf->msg_list);
1128 info->node_cache = new_leaf;
Doug Ledfordce2d52c2012-05-31 16:26:38 -07001129 } else {
1130 kfree(new_leaf);
1131 }
1132
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 if (info->attr.mq_curmsgs == 0) {
Al Viro2903ff02012-08-28 12:52:22 -04001134 if (f.file->f_flags & O_NONBLOCK) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135 spin_unlock(&info->lock);
1136 ret = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 } else {
1138 wait.task = current;
1139 wait.state = STATE_NONE;
1140 ret = wq_sleep(info, RECV, timeout, &wait);
1141 msg_ptr = wait.msg;
1142 }
1143 } else {
Waiman Long194a6b52016-11-17 11:46:38 -05001144 DEFINE_WAKE_Q(wake_q);
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001145
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 msg_ptr = msg_get(info);
1147
1148 inode->i_atime = inode->i_mtime = inode->i_ctime =
Deepa Dinamani078cd822016-09-14 07:48:04 -07001149 current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150
1151 /* There is now free space in queue. */
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001152 pipelined_receive(&wake_q, info);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 spin_unlock(&info->lock);
Davidlohr Buesofa6004a2015-05-04 07:02:46 -07001154 wake_up_q(&wake_q);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155 ret = 0;
1156 }
1157 if (ret == 0) {
1158 ret = msg_ptr->m_ts;
1159
1160 if ((u_msg_prio && put_user(msg_ptr->m_type, u_msg_prio)) ||
1161 store_msg(u_msg_ptr, msg_ptr, msg_ptr->m_ts)) {
1162 ret = -EFAULT;
1163 }
1164 free_msg(msg_ptr);
1165 }
1166out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04001167 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168out:
1169 return ret;
1170}
1171
Al Viro0d060602017-06-27 21:32:36 -04001172SYSCALL_DEFINE5(mq_timedsend, mqd_t, mqdes, const char __user *, u_msg_ptr,
1173 size_t, msg_len, unsigned int, msg_prio,
Arnd Bergmann21fc5382018-04-13 13:58:00 +02001174 const struct __kernel_timespec __user *, u_abs_timeout)
Al Viro0d060602017-06-27 21:32:36 -04001175{
Deepa Dinamanib9047722017-08-02 19:51:11 -07001176 struct timespec64 ts, *p = NULL;
Al Viro0d060602017-06-27 21:32:36 -04001177 if (u_abs_timeout) {
1178 int res = prepare_timeout(u_abs_timeout, &ts);
1179 if (res)
1180 return res;
1181 p = &ts;
1182 }
1183 return do_mq_timedsend(mqdes, u_msg_ptr, msg_len, msg_prio, p);
1184}
1185
1186SYSCALL_DEFINE5(mq_timedreceive, mqd_t, mqdes, char __user *, u_msg_ptr,
1187 size_t, msg_len, unsigned int __user *, u_msg_prio,
Arnd Bergmann21fc5382018-04-13 13:58:00 +02001188 const struct __kernel_timespec __user *, u_abs_timeout)
Al Viro0d060602017-06-27 21:32:36 -04001189{
Deepa Dinamanib9047722017-08-02 19:51:11 -07001190 struct timespec64 ts, *p = NULL;
Al Viro0d060602017-06-27 21:32:36 -04001191 if (u_abs_timeout) {
1192 int res = prepare_timeout(u_abs_timeout, &ts);
1193 if (res)
1194 return res;
1195 p = &ts;
1196 }
1197 return do_mq_timedreceive(mqdes, u_msg_ptr, msg_len, u_msg_prio, p);
1198}
1199
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200/*
1201 * Notes: the case when user wants us to deregister (with NULL as pointer)
1202 * and he isn't currently owner of notification, will be silently discarded.
1203 * It isn't explicitly defined in the POSIX.
1204 */
Al Viro0d060602017-06-27 21:32:36 -04001205static int do_mq_notify(mqd_t mqdes, const struct sigevent *notification)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206{
Al Viro2903ff02012-08-28 12:52:22 -04001207 int ret;
1208 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209 struct sock *sock;
1210 struct inode *inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 struct mqueue_inode_info *info;
1212 struct sk_buff *nc;
1213
Al Viro0d060602017-06-27 21:32:36 -04001214 audit_mq_notify(mqdes, notification);
George C. Wilson20ca73b2006-05-24 16:09:55 -05001215
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 nc = NULL;
1217 sock = NULL;
Al Viro0d060602017-06-27 21:32:36 -04001218 if (notification != NULL) {
1219 if (unlikely(notification->sigev_notify != SIGEV_NONE &&
1220 notification->sigev_notify != SIGEV_SIGNAL &&
1221 notification->sigev_notify != SIGEV_THREAD))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222 return -EINVAL;
Al Viro0d060602017-06-27 21:32:36 -04001223 if (notification->sigev_notify == SIGEV_SIGNAL &&
1224 !valid_signal(notification->sigev_signo)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225 return -EINVAL;
1226 }
Al Viro0d060602017-06-27 21:32:36 -04001227 if (notification->sigev_notify == SIGEV_THREAD) {
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001228 long timeo;
1229
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 /* create the notify skb */
1231 nc = alloc_skb(NOTIFY_COOKIE_LEN, GFP_KERNEL);
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001232 if (!nc) {
1233 ret = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001235 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 if (copy_from_user(nc->data,
Al Viro0d060602017-06-27 21:32:36 -04001237 notification->sigev_value.sival_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238 NOTIFY_COOKIE_LEN)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001239 ret = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240 goto out;
1241 }
1242
1243 /* TODO: add a header? */
1244 skb_put(nc, NOTIFY_COOKIE_LEN);
1245 /* and attach it to the socket */
1246retry:
Al Viro0d060602017-06-27 21:32:36 -04001247 f = fdget(notification->sigev_signo);
Al Viro2903ff02012-08-28 12:52:22 -04001248 if (!f.file) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001249 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001251 }
Al Viro2903ff02012-08-28 12:52:22 -04001252 sock = netlink_getsockbyfilp(f.file);
1253 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254 if (IS_ERR(sock)) {
1255 ret = PTR_ERR(sock);
1256 sock = NULL;
1257 goto out;
1258 }
1259
Patrick McHardyc3d8d1e2007-11-07 02:42:09 -08001260 timeo = MAX_SCHEDULE_TIMEOUT;
Denis V. Lunev9457afe2008-06-05 11:23:39 -07001261 ret = netlink_attachskb(sock, nc, &timeo, NULL);
Cong Wangf991af32017-07-09 13:19:55 -07001262 if (ret == 1) {
1263 sock = NULL;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001264 goto retry;
Cong Wangf991af32017-07-09 13:19:55 -07001265 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 if (ret) {
1267 sock = NULL;
1268 nc = NULL;
1269 goto out;
1270 }
1271 }
1272 }
1273
Al Viro2903ff02012-08-28 12:52:22 -04001274 f = fdget(mqdes);
1275 if (!f.file) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001276 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277 goto out;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001278 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279
Al Viro496ad9a2013-01-23 17:07:38 -05001280 inode = file_inode(f.file);
Al Viro2903ff02012-08-28 12:52:22 -04001281 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001282 ret = -EBADF;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 goto out_fput;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001284 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 info = MQUEUE_I(inode);
1286
1287 ret = 0;
1288 spin_lock(&info->lock);
Al Viro0d060602017-06-27 21:32:36 -04001289 if (notification == NULL) {
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001290 if (info->notify_owner == task_tgid(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 remove_notification(info);
Deepa Dinamani078cd822016-09-14 07:48:04 -07001292 inode->i_atime = inode->i_ctime = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001294 } else if (info->notify_owner != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295 ret = -EBUSY;
1296 } else {
Al Viro0d060602017-06-27 21:32:36 -04001297 switch (notification->sigev_notify) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 case SIGEV_NONE:
1299 info->notify.sigev_notify = SIGEV_NONE;
1300 break;
1301 case SIGEV_THREAD:
1302 info->notify_sock = sock;
1303 info->notify_cookie = nc;
1304 sock = NULL;
1305 nc = NULL;
1306 info->notify.sigev_notify = SIGEV_THREAD;
1307 break;
1308 case SIGEV_SIGNAL:
Al Viro0d060602017-06-27 21:32:36 -04001309 info->notify.sigev_signo = notification->sigev_signo;
1310 info->notify.sigev_value = notification->sigev_value;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 info->notify.sigev_notify = SIGEV_SIGNAL;
1312 break;
1313 }
Cedric Le Goatera03fcb72006-10-02 02:17:26 -07001314
1315 info->notify_owner = get_pid(task_tgid(current));
Eric W. Biederman6f9ac6d2011-11-16 22:57:55 -08001316 info->notify_user_ns = get_user_ns(current_user_ns());
Deepa Dinamani078cd822016-09-14 07:48:04 -07001317 inode->i_atime = inode->i_ctime = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318 }
1319 spin_unlock(&info->lock);
1320out_fput:
Al Viro2903ff02012-08-28 12:52:22 -04001321 fdput(f);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322out:
Davidlohr Bueso3ab08fe2014-01-27 17:07:06 -08001323 if (sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324 netlink_detachskb(sock, nc);
Davidlohr Bueso3ab08fe2014-01-27 17:07:06 -08001325 else if (nc)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326 dev_kfree_skb(nc);
Davidlohr Bueso3ab08fe2014-01-27 17:07:06 -08001327
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 return ret;
1329}
1330
Al Viro0d060602017-06-27 21:32:36 -04001331SYSCALL_DEFINE2(mq_notify, mqd_t, mqdes,
1332 const struct sigevent __user *, u_notification)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333{
Al Viro0d060602017-06-27 21:32:36 -04001334 struct sigevent n, *p = NULL;
1335 if (u_notification) {
1336 if (copy_from_user(&n, u_notification, sizeof(struct sigevent)))
1337 return -EFAULT;
1338 p = &n;
1339 }
1340 return do_mq_notify(mqdes, p);
1341}
1342
1343static int do_mq_getsetattr(int mqdes, struct mq_attr *new, struct mq_attr *old)
1344{
Al Viro2903ff02012-08-28 12:52:22 -04001345 struct fd f;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346 struct inode *inode;
1347 struct mqueue_inode_info *info;
1348
Al Viro0d060602017-06-27 21:32:36 -04001349 if (new && (new->mq_flags & (~O_NONBLOCK)))
1350 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351
Al Viro2903ff02012-08-28 12:52:22 -04001352 f = fdget(mqdes);
Al Viro0d060602017-06-27 21:32:36 -04001353 if (!f.file)
1354 return -EBADF;
1355
1356 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
1357 fdput(f);
1358 return -EBADF;
André Goddard Rosa8d8ffef2010-02-23 04:04:26 -03001359 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360
Al Viro496ad9a2013-01-23 17:07:38 -05001361 inode = file_inode(f.file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 info = MQUEUE_I(inode);
1363
1364 spin_lock(&info->lock);
1365
Al Viro0d060602017-06-27 21:32:36 -04001366 if (old) {
1367 *old = info->attr;
1368 old->mq_flags = f.file->f_flags & O_NONBLOCK;
1369 }
1370 if (new) {
1371 audit_mq_getsetattr(mqdes, new);
Al Viro2903ff02012-08-28 12:52:22 -04001372 spin_lock(&f.file->f_lock);
Al Viro0d060602017-06-27 21:32:36 -04001373 if (new->mq_flags & O_NONBLOCK)
Al Viro2903ff02012-08-28 12:52:22 -04001374 f.file->f_flags |= O_NONBLOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 else
Al Viro2903ff02012-08-28 12:52:22 -04001376 f.file->f_flags &= ~O_NONBLOCK;
1377 spin_unlock(&f.file->f_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378
Deepa Dinamani078cd822016-09-14 07:48:04 -07001379 inode->i_atime = inode->i_ctime = current_time(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380 }
1381
1382 spin_unlock(&info->lock);
Al Viro2903ff02012-08-28 12:52:22 -04001383 fdput(f);
Al Viro0d060602017-06-27 21:32:36 -04001384 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385}
1386
Al Viro0d060602017-06-27 21:32:36 -04001387SYSCALL_DEFINE3(mq_getsetattr, mqd_t, mqdes,
1388 const struct mq_attr __user *, u_mqstat,
1389 struct mq_attr __user *, u_omqstat)
1390{
1391 int ret;
1392 struct mq_attr mqstat, omqstat;
1393 struct mq_attr *new = NULL, *old = NULL;
1394
1395 if (u_mqstat) {
1396 new = &mqstat;
1397 if (copy_from_user(new, u_mqstat, sizeof(struct mq_attr)))
1398 return -EFAULT;
1399 }
1400 if (u_omqstat)
1401 old = &omqstat;
1402
1403 ret = do_mq_getsetattr(mqdes, new, old);
1404 if (ret || !old)
1405 return ret;
1406
1407 if (copy_to_user(u_omqstat, old, sizeof(struct mq_attr)))
1408 return -EFAULT;
1409 return 0;
1410}
1411
1412#ifdef CONFIG_COMPAT
1413
1414struct compat_mq_attr {
1415 compat_long_t mq_flags; /* message queue flags */
1416 compat_long_t mq_maxmsg; /* maximum number of messages */
1417 compat_long_t mq_msgsize; /* maximum message size */
1418 compat_long_t mq_curmsgs; /* number of messages currently queued */
1419 compat_long_t __reserved[4]; /* ignored for input, zeroed for output */
1420};
1421
1422static inline int get_compat_mq_attr(struct mq_attr *attr,
1423 const struct compat_mq_attr __user *uattr)
1424{
1425 struct compat_mq_attr v;
1426
1427 if (copy_from_user(&v, uattr, sizeof(*uattr)))
1428 return -EFAULT;
1429
1430 memset(attr, 0, sizeof(*attr));
1431 attr->mq_flags = v.mq_flags;
1432 attr->mq_maxmsg = v.mq_maxmsg;
1433 attr->mq_msgsize = v.mq_msgsize;
1434 attr->mq_curmsgs = v.mq_curmsgs;
1435 return 0;
1436}
1437
1438static inline int put_compat_mq_attr(const struct mq_attr *attr,
1439 struct compat_mq_attr __user *uattr)
1440{
1441 struct compat_mq_attr v;
1442
1443 memset(&v, 0, sizeof(v));
1444 v.mq_flags = attr->mq_flags;
1445 v.mq_maxmsg = attr->mq_maxmsg;
1446 v.mq_msgsize = attr->mq_msgsize;
1447 v.mq_curmsgs = attr->mq_curmsgs;
1448 if (copy_to_user(uattr, &v, sizeof(*uattr)))
1449 return -EFAULT;
1450 return 0;
1451}
1452
1453COMPAT_SYSCALL_DEFINE4(mq_open, const char __user *, u_name,
1454 int, oflag, compat_mode_t, mode,
1455 struct compat_mq_attr __user *, u_attr)
1456{
1457 struct mq_attr attr, *p = NULL;
1458 if (u_attr && oflag & O_CREAT) {
1459 p = &attr;
1460 if (get_compat_mq_attr(&attr, u_attr))
1461 return -EFAULT;
1462 }
1463 return do_mq_open(u_name, oflag, mode, p);
1464}
1465
Arnd Bergmannb0d17572018-04-13 13:58:23 +02001466COMPAT_SYSCALL_DEFINE2(mq_notify, mqd_t, mqdes,
1467 const struct compat_sigevent __user *, u_notification)
1468{
1469 struct sigevent n, *p = NULL;
1470 if (u_notification) {
1471 if (get_compat_sigevent(&n, u_notification))
1472 return -EFAULT;
1473 if (n.sigev_notify == SIGEV_THREAD)
1474 n.sigev_value.sival_ptr = compat_ptr(n.sigev_value.sival_int);
1475 p = &n;
1476 }
1477 return do_mq_notify(mqdes, p);
1478}
1479
1480COMPAT_SYSCALL_DEFINE3(mq_getsetattr, mqd_t, mqdes,
1481 const struct compat_mq_attr __user *, u_mqstat,
1482 struct compat_mq_attr __user *, u_omqstat)
1483{
1484 int ret;
1485 struct mq_attr mqstat, omqstat;
1486 struct mq_attr *new = NULL, *old = NULL;
1487
1488 if (u_mqstat) {
1489 new = &mqstat;
1490 if (get_compat_mq_attr(new, u_mqstat))
1491 return -EFAULT;
1492 }
1493 if (u_omqstat)
1494 old = &omqstat;
1495
1496 ret = do_mq_getsetattr(mqdes, new, old);
1497 if (ret || !old)
1498 return ret;
1499
1500 if (put_compat_mq_attr(old, u_omqstat))
1501 return -EFAULT;
1502 return 0;
1503}
1504#endif
1505
1506#ifdef CONFIG_COMPAT_32BIT_TIME
Arnd Bergmann9afc5ee2018-07-13 12:52:28 +02001507static int compat_prepare_timeout(const struct old_timespec32 __user *p,
Deepa Dinamanib9047722017-08-02 19:51:11 -07001508 struct timespec64 *ts)
Al Viro0d060602017-06-27 21:32:36 -04001509{
Arnd Bergmann9afc5ee2018-07-13 12:52:28 +02001510 if (get_old_timespec32(ts, p))
Al Viro0d060602017-06-27 21:32:36 -04001511 return -EFAULT;
Deepa Dinamanib9047722017-08-02 19:51:11 -07001512 if (!timespec64_valid(ts))
Al Viro0d060602017-06-27 21:32:36 -04001513 return -EINVAL;
1514 return 0;
1515}
1516
Arnd Bergmann8dabe722019-01-07 00:33:08 +01001517SYSCALL_DEFINE5(mq_timedsend_time32, mqd_t, mqdes,
1518 const char __user *, u_msg_ptr,
1519 unsigned int, msg_len, unsigned int, msg_prio,
1520 const struct old_timespec32 __user *, u_abs_timeout)
Al Viro0d060602017-06-27 21:32:36 -04001521{
Deepa Dinamanib9047722017-08-02 19:51:11 -07001522 struct timespec64 ts, *p = NULL;
Al Viro0d060602017-06-27 21:32:36 -04001523 if (u_abs_timeout) {
1524 int res = compat_prepare_timeout(u_abs_timeout, &ts);
1525 if (res)
1526 return res;
1527 p = &ts;
1528 }
1529 return do_mq_timedsend(mqdes, u_msg_ptr, msg_len, msg_prio, p);
1530}
1531
Arnd Bergmann8dabe722019-01-07 00:33:08 +01001532SYSCALL_DEFINE5(mq_timedreceive_time32, mqd_t, mqdes,
1533 char __user *, u_msg_ptr,
1534 unsigned int, msg_len, unsigned int __user *, u_msg_prio,
1535 const struct old_timespec32 __user *, u_abs_timeout)
Al Viro0d060602017-06-27 21:32:36 -04001536{
Deepa Dinamanib9047722017-08-02 19:51:11 -07001537 struct timespec64 ts, *p = NULL;
Al Viro0d060602017-06-27 21:32:36 -04001538 if (u_abs_timeout) {
1539 int res = compat_prepare_timeout(u_abs_timeout, &ts);
1540 if (res)
1541 return res;
1542 p = &ts;
1543 }
1544 return do_mq_timedreceive(mqdes, u_msg_ptr, msg_len, u_msg_prio, p);
1545}
Al Viro0d060602017-06-27 21:32:36 -04001546#endif
1547
Arjan van de Ven92e1d5b2007-02-12 00:55:39 -08001548static const struct inode_operations mqueue_dir_inode_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 .lookup = simple_lookup,
1550 .create = mqueue_create,
1551 .unlink = mqueue_unlink,
1552};
1553
Arjan van de Ven9a321442007-02-12 00:55:35 -08001554static const struct file_operations mqueue_file_operations = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 .flush = mqueue_flush_file,
1556 .poll = mqueue_poll_file,
1557 .read = mqueue_read_file,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001558 .llseek = default_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559};
1560
Alexey Dobriyanb87221d2009-09-21 17:01:09 -07001561static const struct super_operations mqueue_super_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562 .alloc_inode = mqueue_alloc_inode,
Al Viro015d7952019-04-15 22:30:30 -04001563 .free_inode = mqueue_free_inode,
Al Viro6d8af642010-06-05 16:29:45 -04001564 .evict_inode = mqueue_evict_inode,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566};
1567
David Howells935c69122018-11-01 23:07:25 +00001568static const struct fs_context_operations mqueue_fs_context_ops = {
1569 .free = mqueue_fs_context_free,
1570 .get_tree = mqueue_get_tree,
1571};
1572
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573static struct file_system_type mqueue_fs_type = {
David Howells935c69122018-11-01 23:07:25 +00001574 .name = "mqueue",
1575 .init_fs_context = mqueue_init_fs_context,
1576 .kill_sb = kill_litter_super,
1577 .fs_flags = FS_USERNS_MOUNT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578};
1579
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001580int mq_init_ns(struct ipc_namespace *ns)
1581{
David Howells935c69122018-11-01 23:07:25 +00001582 struct vfsmount *m;
1583
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001584 ns->mq_queues_count = 0;
1585 ns->mq_queues_max = DFLT_QUEUESMAX;
1586 ns->mq_msg_max = DFLT_MSGMAX;
1587 ns->mq_msgsize_max = DFLT_MSGSIZEMAX;
KOSAKI Motohirocef01842012-05-31 16:26:33 -07001588 ns->mq_msg_default = DFLT_MSG;
1589 ns->mq_msgsize_default = DFLT_MSGSIZE;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001590
David Howells935c69122018-11-01 23:07:25 +00001591 m = mq_create_mount(ns);
1592 if (IS_ERR(m))
1593 return PTR_ERR(m);
1594 ns->mq_mnt = m;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001595 return 0;
1596}
1597
1598void mq_clear_sbinfo(struct ipc_namespace *ns)
1599{
Eric W. Biedermancfb2f6f2018-03-24 11:28:14 -05001600 ns->mq_mnt->mnt_sb->s_fs_info = NULL;
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001601}
1602
1603void mq_put_mnt(struct ipc_namespace *ns)
1604{
Eric W. Biedermancfb2f6f2018-03-24 11:28:14 -05001605 kern_unmount(ns->mq_mnt);
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001606}
1607
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608static int __init init_mqueue_fs(void)
1609{
1610 int error;
1611
1612 mqueue_inode_cachep = kmem_cache_create("mqueue_inode_cache",
1613 sizeof(struct mqueue_inode_info), 0,
Vladimir Davydov5d097052016-01-14 15:18:21 -08001614 SLAB_HWCACHE_ALIGN|SLAB_ACCOUNT, init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615 if (mqueue_inode_cachep == NULL)
1616 return -ENOMEM;
1617
André Goddard Rosa2329e3922010-02-23 04:04:27 -03001618 /* ignore failures - they are not fatal */
Serge E. Hallynbdc8e5f2009-04-06 19:01:11 -07001619 mq_sysctl_table = mq_register_sysctl_table();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620
1621 error = register_filesystem(&mqueue_fs_type);
1622 if (error)
1623 goto out_sysctl;
1624
Serge E. Hallyn7eafd7c2009-04-06 19:01:10 -07001625 spin_lock_init(&mq_lock);
1626
Al Viro6f686572011-12-09 00:38:50 -05001627 error = mq_init_ns(&init_ipc_ns);
1628 if (error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629 goto out_filesystem;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631 return 0;
1632
1633out_filesystem:
1634 unregister_filesystem(&mqueue_fs_type);
1635out_sysctl:
1636 if (mq_sysctl_table)
1637 unregister_sysctl_table(mq_sysctl_table);
Alexey Dobriyan1a1d92c2006-09-27 01:49:40 -07001638 kmem_cache_destroy(mqueue_inode_cachep);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639 return error;
1640}
1641
Davidlohr Bueso6d08a252014-04-07 15:39:18 -07001642device_initcall(init_mqueue_fs);