1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16#include <linux/capability.h>
17#include <linux/init.h>
18#include <linux/pagemap.h>
19#include <linux/file.h>
20#include <linux/mount.h>
21#include <linux/fs_context.h>
22#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>
28#include <linux/vmalloc.h>
29#include <linux/netlink.h>
30#include <linux/syscalls.h>
31#include <linux/audit.h>
32#include <linux/signal.h>
33#include <linux/mutex.h>
34#include <linux/nsproxy.h>
35#include <linux/pid.h>
36#include <linux/ipc_namespace.h>
37#include <linux/user_namespace.h>
38#include <linux/slab.h>
39#include <linux/sched/wake_q.h>
40#include <linux/sched/signal.h>
41#include <linux/sched/user.h>
42
43#include <net/sock.h>
44#include "util.h"
45
46struct mqueue_fs_context {
47 struct ipc_namespace *ipc_ns;
48};
49
50#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
58#define STATE_READY 1
59
60struct posix_msg_tree_node {
61 struct rb_node rb_node;
62 struct list_head msg_list;
63 int priority;
64};
65
66struct ext_wait_queue {
67 struct task_struct *task;
68 struct list_head list;
69 struct msg_msg *msg;
70 int state;
71};
72
73struct mqueue_inode_info {
74 spinlock_t lock;
75 struct inode vfs_inode;
76 wait_queue_head_t wait_q;
77
78 struct rb_root msg_tree;
79 struct rb_node *msg_tree_rightmost;
80 struct posix_msg_tree_node *node_cache;
81 struct mq_attr attr;
82
83 struct sigevent notify;
84 struct pid *notify_owner;
85 struct user_namespace *notify_user_ns;
86 struct user_struct *user;
87 struct sock *notify_sock;
88 struct sk_buff *notify_cookie;
89
90
91 struct ext_wait_queue e_wait_q[2];
92
93 unsigned long qsize;
94};
95
96static struct file_system_type mqueue_fs_type;
97static const struct inode_operations mqueue_dir_inode_operations;
98static const struct file_operations mqueue_file_operations;
99static const struct super_operations mqueue_super_ops;
100static const struct fs_context_operations mqueue_fs_context_ops;
101static void remove_notification(struct mqueue_inode_info *info);
102
103static struct kmem_cache *mqueue_inode_cachep;
104
105static struct ctl_table_header *mq_sysctl_table;
106
107static inline struct mqueue_inode_info *MQUEUE_I(struct inode *inode)
108{
109 return container_of(inode, struct mqueue_inode_info, vfs_inode);
110}
111
112
113
114
115static inline struct ipc_namespace *__get_ns_from_inode(struct inode *inode)
116{
117 return get_ipc_ns(inode->i_sb->s_fs_info);
118}
119
120static struct ipc_namespace *get_ns_from_inode(struct inode *inode)
121{
122 struct ipc_namespace *ns;
123
124 spin_lock(&mq_lock);
125 ns = __get_ns_from_inode(inode);
126 spin_unlock(&mq_lock);
127 return ns;
128}
129
130
131static int msg_insert(struct msg_msg *msg, struct mqueue_inode_info *info)
132{
133 struct rb_node **p, *parent = NULL;
134 struct posix_msg_tree_node *leaf;
135 bool rightmost = true;
136
137 p = &info->msg_tree.rb_node;
138 while (*p) {
139 parent = *p;
140 leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node);
141
142 if (likely(leaf->priority == msg->m_type))
143 goto insert_msg;
144 else if (msg->m_type < leaf->priority) {
145 p = &(*p)->rb_left;
146 rightmost = false;
147 } else
148 p = &(*p)->rb_right;
149 }
150 if (info->node_cache) {
151 leaf = info->node_cache;
152 info->node_cache = NULL;
153 } else {
154 leaf = kmalloc(sizeof(*leaf), GFP_ATOMIC);
155 if (!leaf)
156 return -ENOMEM;
157 INIT_LIST_HEAD(&leaf->msg_list);
158 }
159 leaf->priority = msg->m_type;
160
161 if (rightmost)
162 info->msg_tree_rightmost = &leaf->rb_node;
163
164 rb_link_node(&leaf->rb_node, parent, p);
165 rb_insert_color(&leaf->rb_node, &info->msg_tree);
166insert_msg:
167 info->attr.mq_curmsgs++;
168 info->qsize += msg->m_ts;
169 list_add_tail(&msg->m_list, &leaf->msg_list);
170 return 0;
171}
172
173static inline void msg_tree_erase(struct posix_msg_tree_node *leaf,
174 struct mqueue_inode_info *info)
175{
176 struct rb_node *node = &leaf->rb_node;
177
178 if (info->msg_tree_rightmost == node)
179 info->msg_tree_rightmost = rb_prev(node);
180
181 rb_erase(node, &info->msg_tree);
182 if (info->node_cache) {
183 kfree(leaf);
184 } else {
185 info->node_cache = leaf;
186 }
187}
188
189static inline struct msg_msg *msg_get(struct mqueue_inode_info *info)
190{
191 struct rb_node *parent = NULL;
192 struct posix_msg_tree_node *leaf;
193 struct msg_msg *msg;
194
195try_again:
196
197
198
199
200
201 parent = info->msg_tree_rightmost;
202 if (!parent) {
203 if (info->attr.mq_curmsgs) {
204 pr_warn_once("Inconsistency in POSIX message queue, "
205 "no tree element, but supposedly messages "
206 "should exist!\n");
207 info->attr.mq_curmsgs = 0;
208 }
209 return NULL;
210 }
211 leaf = rb_entry(parent, struct posix_msg_tree_node, rb_node);
212 if (unlikely(list_empty(&leaf->msg_list))) {
213 pr_warn_once("Inconsistency in POSIX message queue, "
214 "empty leaf node but we haven't implemented "
215 "lazy leaf delete!\n");
216 msg_tree_erase(leaf, info);
217 goto try_again;
218 } else {
219 msg = list_first_entry(&leaf->msg_list,
220 struct msg_msg, m_list);
221 list_del(&msg->m_list);
222 if (list_empty(&leaf->msg_list)) {
223 msg_tree_erase(leaf, info);
224 }
225 }
226 info->attr.mq_curmsgs--;
227 info->qsize -= msg->m_ts;
228 return msg;
229}
230
231static struct inode *mqueue_get_inode(struct super_block *sb,
232 struct ipc_namespace *ipc_ns, umode_t mode,
233 struct mq_attr *attr)
234{
235 struct user_struct *u = current_user();
236 struct inode *inode;
237 int ret = -ENOMEM;
238
239 inode = new_inode(sb);
240 if (!inode)
241 goto err;
242
243 inode->i_ino = get_next_ino();
244 inode->i_mode = mode;
245 inode->i_uid = current_fsuid();
246 inode->i_gid = current_fsgid();
247 inode->i_mtime = inode->i_ctime = inode->i_atime = current_time(inode);
248
249 if (S_ISREG(mode)) {
250 struct mqueue_inode_info *info;
251 unsigned long mq_bytes, mq_treesize;
252
253 inode->i_fop = &mqueue_file_operations;
254 inode->i_size = FILENT_SIZE;
255
256 info = MQUEUE_I(inode);
257 spin_lock_init(&info->lock);
258 init_waitqueue_head(&info->wait_q);
259 INIT_LIST_HEAD(&info->e_wait_q[0].list);
260 INIT_LIST_HEAD(&info->e_wait_q[1].list);
261 info->notify_owner = NULL;
262 info->notify_user_ns = NULL;
263 info->qsize = 0;
264 info->user = NULL;
265 info->msg_tree = RB_ROOT;
266 info->msg_tree_rightmost = NULL;
267 info->node_cache = NULL;
268 memset(&info->attr, 0, sizeof(info->attr));
269 info->attr.mq_maxmsg = min(ipc_ns->mq_msg_max,
270 ipc_ns->mq_msg_default);
271 info->attr.mq_msgsize = min(ipc_ns->mq_msgsize_max,
272 ipc_ns->mq_msgsize_default);
273 if (attr) {
274 info->attr.mq_maxmsg = attr->mq_maxmsg;
275 info->attr.mq_msgsize = attr->mq_msgsize;
276 }
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291 ret = -EINVAL;
292 if (info->attr.mq_maxmsg <= 0 || info->attr.mq_msgsize <= 0)
293 goto out_inode;
294 if (capable(CAP_SYS_RESOURCE)) {
295 if (info->attr.mq_maxmsg > HARD_MSGMAX ||
296 info->attr.mq_msgsize > HARD_MSGSIZEMAX)
297 goto out_inode;
298 } else {
299 if (info->attr.mq_maxmsg > ipc_ns->mq_msg_max ||
300 info->attr.mq_msgsize > ipc_ns->mq_msgsize_max)
301 goto out_inode;
302 }
303 ret = -EOVERFLOW;
304
305 if (info->attr.mq_msgsize > ULONG_MAX/info->attr.mq_maxmsg)
306 goto out_inode;
307 mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) +
308 min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) *
309 sizeof(struct posix_msg_tree_node);
310 mq_bytes = info->attr.mq_maxmsg * info->attr.mq_msgsize;
311 if (mq_bytes + mq_treesize < mq_bytes)
312 goto out_inode;
313 mq_bytes += mq_treesize;
314 spin_lock(&mq_lock);
315 if (u->mq_bytes + mq_bytes < u->mq_bytes ||
316 u->mq_bytes + mq_bytes > rlimit(RLIMIT_MSGQUEUE)) {
317 spin_unlock(&mq_lock);
318
319 ret = -EMFILE;
320 goto out_inode;
321 }
322 u->mq_bytes += mq_bytes;
323 spin_unlock(&mq_lock);
324
325
326 info->user = get_uid(u);
327 } else if (S_ISDIR(mode)) {
328 inc_nlink(inode);
329
330 inode->i_size = 2 * DIRENT_SIZE;
331 inode->i_op = &mqueue_dir_inode_operations;
332 inode->i_fop = &simple_dir_operations;
333 }
334
335 return inode;
336out_inode:
337 iput(inode);
338err:
339 return ERR_PTR(ret);
340}
341
342static int mqueue_fill_super(struct super_block *sb, struct fs_context *fc)
343{
344 struct inode *inode;
345 struct ipc_namespace *ns = sb->s_fs_info;
346
347 sb->s_iflags |= SB_I_NOEXEC | SB_I_NODEV;
348 sb->s_blocksize = PAGE_SIZE;
349 sb->s_blocksize_bits = PAGE_SHIFT;
350 sb->s_magic = MQUEUE_MAGIC;
351 sb->s_op = &mqueue_super_ops;
352
353 inode = mqueue_get_inode(sb, ns, S_IFDIR | S_ISVTX | S_IRWXUGO, NULL);
354 if (IS_ERR(inode))
355 return PTR_ERR(inode);
356
357 sb->s_root = d_make_root(inode);
358 if (!sb->s_root)
359 return -ENOMEM;
360 return 0;
361}
362
363static int mqueue_get_tree(struct fs_context *fc)
364{
365 struct mqueue_fs_context *ctx = fc->fs_private;
366
367 fc->s_fs_info = ctx->ipc_ns;
368 return vfs_get_super(fc, vfs_get_keyed_super, mqueue_fill_super);
369}
370
371static void mqueue_fs_context_free(struct fs_context *fc)
372{
373 struct mqueue_fs_context *ctx = fc->fs_private;
374
375 put_ipc_ns(ctx->ipc_ns);
376 kfree(ctx);
377}
378
379static int mqueue_init_fs_context(struct fs_context *fc)
380{
381 struct mqueue_fs_context *ctx;
382
383 ctx = kzalloc(sizeof(struct mqueue_fs_context), GFP_KERNEL);
384 if (!ctx)
385 return -ENOMEM;
386
387 ctx->ipc_ns = get_ipc_ns(current->nsproxy->ipc_ns);
388 put_user_ns(fc->user_ns);
389 fc->user_ns = get_user_ns(ctx->ipc_ns->user_ns);
390 fc->fs_private = ctx;
391 fc->ops = &mqueue_fs_context_ops;
392 return 0;
393}
394
395static struct vfsmount *mq_create_mount(struct ipc_namespace *ns)
396{
397 struct mqueue_fs_context *ctx;
398 struct fs_context *fc;
399 struct vfsmount *mnt;
400
401 fc = fs_context_for_mount(&mqueue_fs_type, SB_KERNMOUNT);
402 if (IS_ERR(fc))
403 return ERR_CAST(fc);
404
405 ctx = fc->fs_private;
406 put_ipc_ns(ctx->ipc_ns);
407 ctx->ipc_ns = get_ipc_ns(ns);
408 put_user_ns(fc->user_ns);
409 fc->user_ns = get_user_ns(ctx->ipc_ns->user_ns);
410
411 mnt = fc_mount(fc);
412 put_fs_context(fc);
413 return mnt;
414}
415
416static void init_once(void *foo)
417{
418 struct mqueue_inode_info *p = (struct mqueue_inode_info *) foo;
419
420 inode_init_once(&p->vfs_inode);
421}
422
423static struct inode *mqueue_alloc_inode(struct super_block *sb)
424{
425 struct mqueue_inode_info *ei;
426
427 ei = kmem_cache_alloc(mqueue_inode_cachep, GFP_KERNEL);
428 if (!ei)
429 return NULL;
430 return &ei->vfs_inode;
431}
432
433static void mqueue_free_inode(struct inode *inode)
434{
435 kmem_cache_free(mqueue_inode_cachep, MQUEUE_I(inode));
436}
437
438static void mqueue_evict_inode(struct inode *inode)
439{
440 struct mqueue_inode_info *info;
441 struct user_struct *user;
442 struct ipc_namespace *ipc_ns;
443 struct msg_msg *msg, *nmsg;
444 LIST_HEAD(tmp_msg);
445
446 clear_inode(inode);
447
448 if (S_ISDIR(inode->i_mode))
449 return;
450
451 ipc_ns = get_ns_from_inode(inode);
452 info = MQUEUE_I(inode);
453 spin_lock(&info->lock);
454 while ((msg = msg_get(info)) != NULL)
455 list_add_tail(&msg->m_list, &tmp_msg);
456 kfree(info->node_cache);
457 spin_unlock(&info->lock);
458
459 list_for_each_entry_safe(msg, nmsg, &tmp_msg, m_list) {
460 list_del(&msg->m_list);
461 free_msg(msg);
462 }
463
464 user = info->user;
465 if (user) {
466 unsigned long mq_bytes, mq_treesize;
467
468
469 mq_treesize = info->attr.mq_maxmsg * sizeof(struct msg_msg) +
470 min_t(unsigned int, info->attr.mq_maxmsg, MQ_PRIO_MAX) *
471 sizeof(struct posix_msg_tree_node);
472
473 mq_bytes = mq_treesize + (info->attr.mq_maxmsg *
474 info->attr.mq_msgsize);
475
476 spin_lock(&mq_lock);
477 user->mq_bytes -= mq_bytes;
478
479
480
481
482
483
484 if (ipc_ns)
485 ipc_ns->mq_queues_count--;
486 spin_unlock(&mq_lock);
487 free_uid(user);
488 }
489 if (ipc_ns)
490 put_ipc_ns(ipc_ns);
491}
492
493static int mqueue_create_attr(struct dentry *dentry, umode_t mode, void *arg)
494{
495 struct inode *dir = dentry->d_parent->d_inode;
496 struct inode *inode;
497 struct mq_attr *attr = arg;
498 int error;
499 struct ipc_namespace *ipc_ns;
500
501 spin_lock(&mq_lock);
502 ipc_ns = __get_ns_from_inode(dir);
503 if (!ipc_ns) {
504 error = -EACCES;
505 goto out_unlock;
506 }
507
508 if (ipc_ns->mq_queues_count >= ipc_ns->mq_queues_max &&
509 !capable(CAP_SYS_RESOURCE)) {
510 error = -ENOSPC;
511 goto out_unlock;
512 }
513 ipc_ns->mq_queues_count++;
514 spin_unlock(&mq_lock);
515
516 inode = mqueue_get_inode(dir->i_sb, ipc_ns, mode, attr);
517 if (IS_ERR(inode)) {
518 error = PTR_ERR(inode);
519 spin_lock(&mq_lock);
520 ipc_ns->mq_queues_count--;
521 goto out_unlock;
522 }
523
524 put_ipc_ns(ipc_ns);
525 dir->i_size += DIRENT_SIZE;
526 dir->i_ctime = dir->i_mtime = dir->i_atime = current_time(dir);
527
528 d_instantiate(dentry, inode);
529 dget(dentry);
530 return 0;
531out_unlock:
532 spin_unlock(&mq_lock);
533 if (ipc_ns)
534 put_ipc_ns(ipc_ns);
535 return error;
536}
537
538static int mqueue_create(struct inode *dir, struct dentry *dentry,
539 umode_t mode, bool excl)
540{
541 return mqueue_create_attr(dentry, mode, NULL);
542}
543
544static int mqueue_unlink(struct inode *dir, struct dentry *dentry)
545{
546 struct inode *inode = d_inode(dentry);
547
548 dir->i_ctime = dir->i_mtime = dir->i_atime = current_time(dir);
549 dir->i_size -= DIRENT_SIZE;
550 drop_nlink(inode);
551 dput(dentry);
552 return 0;
553}
554
555
556
557
558
559
560
561
562static ssize_t mqueue_read_file(struct file *filp, char __user *u_data,
563 size_t count, loff_t *off)
564{
565 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
566 char buffer[FILENT_SIZE];
567 ssize_t ret;
568
569 spin_lock(&info->lock);
570 snprintf(buffer, sizeof(buffer),
571 "QSIZE:%-10lu NOTIFY:%-5d SIGNO:%-5d NOTIFY_PID:%-6d\n",
572 info->qsize,
573 info->notify_owner ? info->notify.sigev_notify : 0,
574 (info->notify_owner &&
575 info->notify.sigev_notify == SIGEV_SIGNAL) ?
576 info->notify.sigev_signo : 0,
577 pid_vnr(info->notify_owner));
578 spin_unlock(&info->lock);
579 buffer[sizeof(buffer)-1] = '\0';
580
581 ret = simple_read_from_buffer(u_data, count, off, buffer,
582 strlen(buffer));
583 if (ret <= 0)
584 return ret;
585
586 file_inode(filp)->i_atime = file_inode(filp)->i_ctime = current_time(file_inode(filp));
587 return ret;
588}
589
590static int mqueue_flush_file(struct file *filp, fl_owner_t id)
591{
592 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
593
594 spin_lock(&info->lock);
595 if (task_tgid(current) == info->notify_owner)
596 remove_notification(info);
597
598 spin_unlock(&info->lock);
599 return 0;
600}
601
602static __poll_t mqueue_poll_file(struct file *filp, struct poll_table_struct *poll_tab)
603{
604 struct mqueue_inode_info *info = MQUEUE_I(file_inode(filp));
605 __poll_t retval = 0;
606
607 poll_wait(filp, &info->wait_q, poll_tab);
608
609 spin_lock(&info->lock);
610 if (info->attr.mq_curmsgs)
611 retval = EPOLLIN | EPOLLRDNORM;
612
613 if (info->attr.mq_curmsgs < info->attr.mq_maxmsg)
614 retval |= EPOLLOUT | EPOLLWRNORM;
615 spin_unlock(&info->lock);
616
617 return retval;
618}
619
620
621static void wq_add(struct mqueue_inode_info *info, int sr,
622 struct ext_wait_queue *ewp)
623{
624 struct ext_wait_queue *walk;
625
626 list_for_each_entry(walk, &info->e_wait_q[sr].list, list) {
627 if (walk->task->prio <= current->prio) {
628 list_add_tail(&ewp->list, &walk->list);
629 return;
630 }
631 }
632 list_add_tail(&ewp->list, &info->e_wait_q[sr].list);
633}
634
635
636
637
638
639
640static int wq_sleep(struct mqueue_inode_info *info, int sr,
641 ktime_t *timeout, struct ext_wait_queue *ewp)
642 __releases(&info->lock)
643{
644 int retval;
645 signed long time;
646
647 wq_add(info, sr, ewp);
648
649 for (;;) {
650 __set_current_state(TASK_INTERRUPTIBLE);
651
652 spin_unlock(&info->lock);
653 time = schedule_hrtimeout_range_clock(timeout, 0,
654 HRTIMER_MODE_ABS, CLOCK_REALTIME);
655
656 if (ewp->state == STATE_READY) {
657 retval = 0;
658 goto out;
659 }
660 spin_lock(&info->lock);
661 if (ewp->state == STATE_READY) {
662 retval = 0;
663 goto out_unlock;
664 }
665 if (signal_pending(current)) {
666 retval = -ERESTARTSYS;
667 break;
668 }
669 if (time == 0) {
670 retval = -ETIMEDOUT;
671 break;
672 }
673 }
674 list_del(&ewp->list);
675out_unlock:
676 spin_unlock(&info->lock);
677out:
678 return retval;
679}
680
681
682
683
684static struct ext_wait_queue *wq_get_first_waiter(
685 struct mqueue_inode_info *info, int sr)
686{
687 struct list_head *ptr;
688
689 ptr = info->e_wait_q[sr].list.prev;
690 if (ptr == &info->e_wait_q[sr].list)
691 return NULL;
692 return list_entry(ptr, struct ext_wait_queue, list);
693}
694
695
696static inline void set_cookie(struct sk_buff *skb, char code)
697{
698 ((char *)skb->data)[NOTIFY_COOKIE_LEN-1] = code;
699}
700
701
702
703
704static void __do_notify(struct mqueue_inode_info *info)
705{
706
707
708
709
710
711 if (info->notify_owner &&
712 info->attr.mq_curmsgs == 1) {
713 struct kernel_siginfo sig_i;
714 switch (info->notify.sigev_notify) {
715 case SIGEV_NONE:
716 break;
717 case SIGEV_SIGNAL:
718
719
720 clear_siginfo(&sig_i);
721 sig_i.si_signo = info->notify.sigev_signo;
722 sig_i.si_errno = 0;
723 sig_i.si_code = SI_MESGQ;
724 sig_i.si_value = info->notify.sigev_value;
725
726 rcu_read_lock();
727 sig_i.si_pid = task_tgid_nr_ns(current,
728 ns_of_pid(info->notify_owner));
729 sig_i.si_uid = from_kuid_munged(info->notify_user_ns, current_uid());
730 rcu_read_unlock();
731
732 kill_pid_info(info->notify.sigev_signo,
733 &sig_i, info->notify_owner);
734 break;
735 case SIGEV_THREAD:
736 set_cookie(info->notify_cookie, NOTIFY_WOKENUP);
737 netlink_sendskb(info->notify_sock, info->notify_cookie);
738 break;
739 }
740
741 put_pid(info->notify_owner);
742 put_user_ns(info->notify_user_ns);
743 info->notify_owner = NULL;
744 info->notify_user_ns = NULL;
745 }
746 wake_up(&info->wait_q);
747}
748
749static int prepare_timeout(const struct __kernel_timespec __user *u_abs_timeout,
750 struct timespec64 *ts)
751{
752 if (get_timespec64(ts, u_abs_timeout))
753 return -EFAULT;
754 if (!timespec64_valid(ts))
755 return -EINVAL;
756 return 0;
757}
758
759static void remove_notification(struct mqueue_inode_info *info)
760{
761 if (info->notify_owner != NULL &&
762 info->notify.sigev_notify == SIGEV_THREAD) {
763 set_cookie(info->notify_cookie, NOTIFY_REMOVED);
764 netlink_sendskb(info->notify_sock, info->notify_cookie);
765 }
766 put_pid(info->notify_owner);
767 put_user_ns(info->notify_user_ns);
768 info->notify_owner = NULL;
769 info->notify_user_ns = NULL;
770}
771
772static int prepare_open(struct dentry *dentry, int oflag, int ro,
773 umode_t mode, struct filename *name,
774 struct mq_attr *attr)
775{
776 static const int oflag2acc[O_ACCMODE] = { MAY_READ, MAY_WRITE,
777 MAY_READ | MAY_WRITE };
778 int acc;
779
780 if (d_really_is_negative(dentry)) {
781 if (!(oflag & O_CREAT))
782 return -ENOENT;
783 if (ro)
784 return ro;
785 audit_inode_parent_hidden(name, dentry->d_parent);
786 return vfs_mkobj(dentry, mode & ~current_umask(),
787 mqueue_create_attr, attr);
788 }
789
790 audit_inode(name, dentry, 0);
791 if ((oflag & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL))
792 return -EEXIST;
793 if ((oflag & O_ACCMODE) == (O_RDWR | O_WRONLY))
794 return -EINVAL;
795 acc = oflag2acc[oflag & O_ACCMODE];
796 return inode_permission(d_inode(dentry), acc);
797}
798
799static int do_mq_open(const char __user *u_name, int oflag, umode_t mode,
800 struct mq_attr *attr)
801{
802 struct vfsmount *mnt = current->nsproxy->ipc_ns->mq_mnt;
803 struct dentry *root = mnt->mnt_root;
804 struct filename *name;
805 struct path path;
806 int fd, error;
807 int ro;
808
809 audit_mq_open(oflag, mode, attr);
810
811 if (IS_ERR(name = getname(u_name)))
812 return PTR_ERR(name);
813
814 fd = get_unused_fd_flags(O_CLOEXEC);
815 if (fd < 0)
816 goto out_putname;
817
818 ro = mnt_want_write(mnt);
819 inode_lock(d_inode(root));
820 path.dentry = lookup_one_len(name->name, root, strlen(name->name));
821 if (IS_ERR(path.dentry)) {
822 error = PTR_ERR(path.dentry);
823 goto out_putfd;
824 }
825 path.mnt = mntget(mnt);
826 error = prepare_open(path.dentry, oflag, ro, mode, name, attr);
827 if (!error) {
828 struct file *file = dentry_open(&path, oflag, current_cred());
829 if (!IS_ERR(file))
830 fd_install(fd, file);
831 else
832 error = PTR_ERR(file);
833 }
834 path_put(&path);
835out_putfd:
836 if (error) {
837 put_unused_fd(fd);
838 fd = error;
839 }
840 inode_unlock(d_inode(root));
841 if (!ro)
842 mnt_drop_write(mnt);
843out_putname:
844 putname(name);
845 return fd;
846}
847
848SYSCALL_DEFINE4(mq_open, const char __user *, u_name, int, oflag, umode_t, mode,
849 struct mq_attr __user *, u_attr)
850{
851 struct mq_attr attr;
852 if (u_attr && copy_from_user(&attr, u_attr, sizeof(struct mq_attr)))
853 return -EFAULT;
854
855 return do_mq_open(u_name, oflag, mode, u_attr ? &attr : NULL);
856}
857
858SYSCALL_DEFINE1(mq_unlink, const char __user *, u_name)
859{
860 int err;
861 struct filename *name;
862 struct dentry *dentry;
863 struct inode *inode = NULL;
864 struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns;
865 struct vfsmount *mnt = ipc_ns->mq_mnt;
866
867 name = getname(u_name);
868 if (IS_ERR(name))
869 return PTR_ERR(name);
870
871 audit_inode_parent_hidden(name, mnt->mnt_root);
872 err = mnt_want_write(mnt);
873 if (err)
874 goto out_name;
875 inode_lock_nested(d_inode(mnt->mnt_root), I_MUTEX_PARENT);
876 dentry = lookup_one_len(name->name, mnt->mnt_root,
877 strlen(name->name));
878 if (IS_ERR(dentry)) {
879 err = PTR_ERR(dentry);
880 goto out_unlock;
881 }
882
883 inode = d_inode(dentry);
884 if (!inode) {
885 err = -ENOENT;
886 } else {
887 ihold(inode);
888 err = vfs_unlink(d_inode(dentry->d_parent), dentry, NULL);
889 }
890 dput(dentry);
891
892out_unlock:
893 inode_unlock(d_inode(mnt->mnt_root));
894 if (inode)
895 iput(inode);
896 mnt_drop_write(mnt);
897out_name:
898 putname(name);
899
900 return err;
901}
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925static inline void pipelined_send(struct wake_q_head *wake_q,
926 struct mqueue_inode_info *info,
927 struct msg_msg *message,
928 struct ext_wait_queue *receiver)
929{
930 receiver->msg = message;
931 list_del(&receiver->list);
932 wake_q_add(wake_q, receiver->task);
933
934
935
936
937
938
939
940
941 receiver->state = STATE_READY;
942}
943
944
945
946static inline void pipelined_receive(struct wake_q_head *wake_q,
947 struct mqueue_inode_info *info)
948{
949 struct ext_wait_queue *sender = wq_get_first_waiter(info, SEND);
950
951 if (!sender) {
952
953 wake_up_interruptible(&info->wait_q);
954 return;
955 }
956 if (msg_insert(sender->msg, info))
957 return;
958
959 list_del(&sender->list);
960 wake_q_add(wake_q, sender->task);
961 sender->state = STATE_READY;
962}
963
964static int do_mq_timedsend(mqd_t mqdes, const char __user *u_msg_ptr,
965 size_t msg_len, unsigned int msg_prio,
966 struct timespec64 *ts)
967{
968 struct fd f;
969 struct inode *inode;
970 struct ext_wait_queue wait;
971 struct ext_wait_queue *receiver;
972 struct msg_msg *msg_ptr;
973 struct mqueue_inode_info *info;
974 ktime_t expires, *timeout = NULL;
975 struct posix_msg_tree_node *new_leaf = NULL;
976 int ret = 0;
977 DEFINE_WAKE_Q(wake_q);
978
979 if (unlikely(msg_prio >= (unsigned long) MQ_PRIO_MAX))
980 return -EINVAL;
981
982 if (ts) {
983 expires = timespec64_to_ktime(*ts);
984 timeout = &expires;
985 }
986
987 audit_mq_sendrecv(mqdes, msg_len, msg_prio, ts);
988
989 f = fdget(mqdes);
990 if (unlikely(!f.file)) {
991 ret = -EBADF;
992 goto out;
993 }
994
995 inode = file_inode(f.file);
996 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
997 ret = -EBADF;
998 goto out_fput;
999 }
1000 info = MQUEUE_I(inode);
1001 audit_file(f.file);
1002
1003 if (unlikely(!(f.file->f_mode & FMODE_WRITE))) {
1004 ret = -EBADF;
1005 goto out_fput;
1006 }
1007
1008 if (unlikely(msg_len > info->attr.mq_msgsize)) {
1009 ret = -EMSGSIZE;
1010 goto out_fput;
1011 }
1012
1013
1014
1015 msg_ptr = load_msg(u_msg_ptr, msg_len);
1016 if (IS_ERR(msg_ptr)) {
1017 ret = PTR_ERR(msg_ptr);
1018 goto out_fput;
1019 }
1020 msg_ptr->m_ts = msg_len;
1021 msg_ptr->m_type = msg_prio;
1022
1023
1024
1025
1026
1027
1028 if (!info->node_cache)
1029 new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL);
1030
1031 spin_lock(&info->lock);
1032
1033 if (!info->node_cache && new_leaf) {
1034
1035 INIT_LIST_HEAD(&new_leaf->msg_list);
1036 info->node_cache = new_leaf;
1037 new_leaf = NULL;
1038 } else {
1039 kfree(new_leaf);
1040 }
1041
1042 if (info->attr.mq_curmsgs == info->attr.mq_maxmsg) {
1043 if (f.file->f_flags & O_NONBLOCK) {
1044 ret = -EAGAIN;
1045 } else {
1046 wait.task = current;
1047 wait.msg = (void *) msg_ptr;
1048 wait.state = STATE_NONE;
1049 ret = wq_sleep(info, SEND, timeout, &wait);
1050
1051
1052
1053
1054 goto out_free;
1055 }
1056 } else {
1057 receiver = wq_get_first_waiter(info, RECV);
1058 if (receiver) {
1059 pipelined_send(&wake_q, info, msg_ptr, receiver);
1060 } else {
1061
1062 ret = msg_insert(msg_ptr, info);
1063 if (ret)
1064 goto out_unlock;
1065 __do_notify(info);
1066 }
1067 inode->i_atime = inode->i_mtime = inode->i_ctime =
1068 current_time(inode);
1069 }
1070out_unlock:
1071 spin_unlock(&info->lock);
1072 wake_up_q(&wake_q);
1073out_free:
1074 if (ret)
1075 free_msg(msg_ptr);
1076out_fput:
1077 fdput(f);
1078out:
1079 return ret;
1080}
1081
1082static int do_mq_timedreceive(mqd_t mqdes, char __user *u_msg_ptr,
1083 size_t msg_len, unsigned int __user *u_msg_prio,
1084 struct timespec64 *ts)
1085{
1086 ssize_t ret;
1087 struct msg_msg *msg_ptr;
1088 struct fd f;
1089 struct inode *inode;
1090 struct mqueue_inode_info *info;
1091 struct ext_wait_queue wait;
1092 ktime_t expires, *timeout = NULL;
1093 struct posix_msg_tree_node *new_leaf = NULL;
1094
1095 if (ts) {
1096 expires = timespec64_to_ktime(*ts);
1097 timeout = &expires;
1098 }
1099
1100 audit_mq_sendrecv(mqdes, msg_len, 0, ts);
1101
1102 f = fdget(mqdes);
1103 if (unlikely(!f.file)) {
1104 ret = -EBADF;
1105 goto out;
1106 }
1107
1108 inode = file_inode(f.file);
1109 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
1110 ret = -EBADF;
1111 goto out_fput;
1112 }
1113 info = MQUEUE_I(inode);
1114 audit_file(f.file);
1115
1116 if (unlikely(!(f.file->f_mode & FMODE_READ))) {
1117 ret = -EBADF;
1118 goto out_fput;
1119 }
1120
1121
1122 if (unlikely(msg_len < info->attr.mq_msgsize)) {
1123 ret = -EMSGSIZE;
1124 goto out_fput;
1125 }
1126
1127
1128
1129
1130
1131
1132 if (!info->node_cache)
1133 new_leaf = kmalloc(sizeof(*new_leaf), GFP_KERNEL);
1134
1135 spin_lock(&info->lock);
1136
1137 if (!info->node_cache && new_leaf) {
1138
1139 INIT_LIST_HEAD(&new_leaf->msg_list);
1140 info->node_cache = new_leaf;
1141 } else {
1142 kfree(new_leaf);
1143 }
1144
1145 if (info->attr.mq_curmsgs == 0) {
1146 if (f.file->f_flags & O_NONBLOCK) {
1147 spin_unlock(&info->lock);
1148 ret = -EAGAIN;
1149 } else {
1150 wait.task = current;
1151 wait.state = STATE_NONE;
1152 ret = wq_sleep(info, RECV, timeout, &wait);
1153 msg_ptr = wait.msg;
1154 }
1155 } else {
1156 DEFINE_WAKE_Q(wake_q);
1157
1158 msg_ptr = msg_get(info);
1159
1160 inode->i_atime = inode->i_mtime = inode->i_ctime =
1161 current_time(inode);
1162
1163
1164 pipelined_receive(&wake_q, info);
1165 spin_unlock(&info->lock);
1166 wake_up_q(&wake_q);
1167 ret = 0;
1168 }
1169 if (ret == 0) {
1170 ret = msg_ptr->m_ts;
1171
1172 if ((u_msg_prio && put_user(msg_ptr->m_type, u_msg_prio)) ||
1173 store_msg(u_msg_ptr, msg_ptr, msg_ptr->m_ts)) {
1174 ret = -EFAULT;
1175 }
1176 free_msg(msg_ptr);
1177 }
1178out_fput:
1179 fdput(f);
1180out:
1181 return ret;
1182}
1183
1184SYSCALL_DEFINE5(mq_timedsend, mqd_t, mqdes, const char __user *, u_msg_ptr,
1185 size_t, msg_len, unsigned int, msg_prio,
1186 const struct __kernel_timespec __user *, u_abs_timeout)
1187{
1188 struct timespec64 ts, *p = NULL;
1189 if (u_abs_timeout) {
1190 int res = prepare_timeout(u_abs_timeout, &ts);
1191 if (res)
1192 return res;
1193 p = &ts;
1194 }
1195 return do_mq_timedsend(mqdes, u_msg_ptr, msg_len, msg_prio, p);
1196}
1197
1198SYSCALL_DEFINE5(mq_timedreceive, mqd_t, mqdes, char __user *, u_msg_ptr,
1199 size_t, msg_len, unsigned int __user *, u_msg_prio,
1200 const struct __kernel_timespec __user *, u_abs_timeout)
1201{
1202 struct timespec64 ts, *p = NULL;
1203 if (u_abs_timeout) {
1204 int res = prepare_timeout(u_abs_timeout, &ts);
1205 if (res)
1206 return res;
1207 p = &ts;
1208 }
1209 return do_mq_timedreceive(mqdes, u_msg_ptr, msg_len, u_msg_prio, p);
1210}
1211
1212
1213
1214
1215
1216
1217static int do_mq_notify(mqd_t mqdes, const struct sigevent *notification)
1218{
1219 int ret;
1220 struct fd f;
1221 struct sock *sock;
1222 struct inode *inode;
1223 struct mqueue_inode_info *info;
1224 struct sk_buff *nc;
1225
1226 audit_mq_notify(mqdes, notification);
1227
1228 nc = NULL;
1229 sock = NULL;
1230 if (notification != NULL) {
1231 if (unlikely(notification->sigev_notify != SIGEV_NONE &&
1232 notification->sigev_notify != SIGEV_SIGNAL &&
1233 notification->sigev_notify != SIGEV_THREAD))
1234 return -EINVAL;
1235 if (notification->sigev_notify == SIGEV_SIGNAL &&
1236 !valid_signal(notification->sigev_signo)) {
1237 return -EINVAL;
1238 }
1239 if (notification->sigev_notify == SIGEV_THREAD) {
1240 long timeo;
1241
1242
1243 nc = alloc_skb(NOTIFY_COOKIE_LEN, GFP_KERNEL);
1244 if (!nc) {
1245 ret = -ENOMEM;
1246 goto out;
1247 }
1248 if (copy_from_user(nc->data,
1249 notification->sigev_value.sival_ptr,
1250 NOTIFY_COOKIE_LEN)) {
1251 ret = -EFAULT;
1252 goto out;
1253 }
1254
1255
1256 skb_put(nc, NOTIFY_COOKIE_LEN);
1257
1258retry:
1259 f = fdget(notification->sigev_signo);
1260 if (!f.file) {
1261 ret = -EBADF;
1262 goto out;
1263 }
1264 sock = netlink_getsockbyfilp(f.file);
1265 fdput(f);
1266 if (IS_ERR(sock)) {
1267 ret = PTR_ERR(sock);
1268 sock = NULL;
1269 goto out;
1270 }
1271
1272 timeo = MAX_SCHEDULE_TIMEOUT;
1273 ret = netlink_attachskb(sock, nc, &timeo, NULL);
1274 if (ret == 1) {
1275 sock = NULL;
1276 goto retry;
1277 }
1278 if (ret) {
1279 sock = NULL;
1280 nc = NULL;
1281 goto out;
1282 }
1283 }
1284 }
1285
1286 f = fdget(mqdes);
1287 if (!f.file) {
1288 ret = -EBADF;
1289 goto out;
1290 }
1291
1292 inode = file_inode(f.file);
1293 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
1294 ret = -EBADF;
1295 goto out_fput;
1296 }
1297 info = MQUEUE_I(inode);
1298
1299 ret = 0;
1300 spin_lock(&info->lock);
1301 if (notification == NULL) {
1302 if (info->notify_owner == task_tgid(current)) {
1303 remove_notification(info);
1304 inode->i_atime = inode->i_ctime = current_time(inode);
1305 }
1306 } else if (info->notify_owner != NULL) {
1307 ret = -EBUSY;
1308 } else {
1309 switch (notification->sigev_notify) {
1310 case SIGEV_NONE:
1311 info->notify.sigev_notify = SIGEV_NONE;
1312 break;
1313 case SIGEV_THREAD:
1314 info->notify_sock = sock;
1315 info->notify_cookie = nc;
1316 sock = NULL;
1317 nc = NULL;
1318 info->notify.sigev_notify = SIGEV_THREAD;
1319 break;
1320 case SIGEV_SIGNAL:
1321 info->notify.sigev_signo = notification->sigev_signo;
1322 info->notify.sigev_value = notification->sigev_value;
1323 info->notify.sigev_notify = SIGEV_SIGNAL;
1324 break;
1325 }
1326
1327 info->notify_owner = get_pid(task_tgid(current));
1328 info->notify_user_ns = get_user_ns(current_user_ns());
1329 inode->i_atime = inode->i_ctime = current_time(inode);
1330 }
1331 spin_unlock(&info->lock);
1332out_fput:
1333 fdput(f);
1334out:
1335 if (sock)
1336 netlink_detachskb(sock, nc);
1337 else if (nc)
1338 dev_kfree_skb(nc);
1339
1340 return ret;
1341}
1342
1343SYSCALL_DEFINE2(mq_notify, mqd_t, mqdes,
1344 const struct sigevent __user *, u_notification)
1345{
1346 struct sigevent n, *p = NULL;
1347 if (u_notification) {
1348 if (copy_from_user(&n, u_notification, sizeof(struct sigevent)))
1349 return -EFAULT;
1350 p = &n;
1351 }
1352 return do_mq_notify(mqdes, p);
1353}
1354
1355static int do_mq_getsetattr(int mqdes, struct mq_attr *new, struct mq_attr *old)
1356{
1357 struct fd f;
1358 struct inode *inode;
1359 struct mqueue_inode_info *info;
1360
1361 if (new && (new->mq_flags & (~O_NONBLOCK)))
1362 return -EINVAL;
1363
1364 f = fdget(mqdes);
1365 if (!f.file)
1366 return -EBADF;
1367
1368 if (unlikely(f.file->f_op != &mqueue_file_operations)) {
1369 fdput(f);
1370 return -EBADF;
1371 }
1372
1373 inode = file_inode(f.file);
1374 info = MQUEUE_I(inode);
1375
1376 spin_lock(&info->lock);
1377
1378 if (old) {
1379 *old = info->attr;
1380 old->mq_flags = f.file->f_flags & O_NONBLOCK;
1381 }
1382 if (new) {
1383 audit_mq_getsetattr(mqdes, new);
1384 spin_lock(&f.file->f_lock);
1385 if (new->mq_flags & O_NONBLOCK)
1386 f.file->f_flags |= O_NONBLOCK;
1387 else
1388 f.file->f_flags &= ~O_NONBLOCK;
1389 spin_unlock(&f.file->f_lock);
1390
1391 inode->i_atime = inode->i_ctime = current_time(inode);
1392 }
1393
1394 spin_unlock(&info->lock);
1395 fdput(f);
1396 return 0;
1397}
1398
1399SYSCALL_DEFINE3(mq_getsetattr, mqd_t, mqdes,
1400 const struct mq_attr __user *, u_mqstat,
1401 struct mq_attr __user *, u_omqstat)
1402{
1403 int ret;
1404 struct mq_attr mqstat, omqstat;
1405 struct mq_attr *new = NULL, *old = NULL;
1406
1407 if (u_mqstat) {
1408 new = &mqstat;
1409 if (copy_from_user(new, u_mqstat, sizeof(struct mq_attr)))
1410 return -EFAULT;
1411 }
1412 if (u_omqstat)
1413 old = &omqstat;
1414
1415 ret = do_mq_getsetattr(mqdes, new, old);
1416 if (ret || !old)
1417 return ret;
1418
1419 if (copy_to_user(u_omqstat, old, sizeof(struct mq_attr)))
1420 return -EFAULT;
1421 return 0;
1422}
1423
1424#ifdef CONFIG_COMPAT
1425
1426struct compat_mq_attr {
1427 compat_long_t mq_flags;
1428 compat_long_t mq_maxmsg;
1429 compat_long_t mq_msgsize;
1430 compat_long_t mq_curmsgs;
1431 compat_long_t __reserved[4];
1432};
1433
1434static inline int get_compat_mq_attr(struct mq_attr *attr,
1435 const struct compat_mq_attr __user *uattr)
1436{
1437 struct compat_mq_attr v;
1438
1439 if (copy_from_user(&v, uattr, sizeof(*uattr)))
1440 return -EFAULT;
1441
1442 memset(attr, 0, sizeof(*attr));
1443 attr->mq_flags = v.mq_flags;
1444 attr->mq_maxmsg = v.mq_maxmsg;
1445 attr->mq_msgsize = v.mq_msgsize;
1446 attr->mq_curmsgs = v.mq_curmsgs;
1447 return 0;
1448}
1449
1450static inline int put_compat_mq_attr(const struct mq_attr *attr,
1451 struct compat_mq_attr __user *uattr)
1452{
1453 struct compat_mq_attr v;
1454
1455 memset(&v, 0, sizeof(v));
1456 v.mq_flags = attr->mq_flags;
1457 v.mq_maxmsg = attr->mq_maxmsg;
1458 v.mq_msgsize = attr->mq_msgsize;
1459 v.mq_curmsgs = attr->mq_curmsgs;
1460 if (copy_to_user(uattr, &v, sizeof(*uattr)))
1461 return -EFAULT;
1462 return 0;
1463}
1464
1465COMPAT_SYSCALL_DEFINE4(mq_open, const char __user *, u_name,
1466 int, oflag, compat_mode_t, mode,
1467 struct compat_mq_attr __user *, u_attr)
1468{
1469 struct mq_attr attr, *p = NULL;
1470 if (u_attr && oflag & O_CREAT) {
1471 p = &attr;
1472 if (get_compat_mq_attr(&attr, u_attr))
1473 return -EFAULT;
1474 }
1475 return do_mq_open(u_name, oflag, mode, p);
1476}
1477
1478COMPAT_SYSCALL_DEFINE2(mq_notify, mqd_t, mqdes,
1479 const struct compat_sigevent __user *, u_notification)
1480{
1481 struct sigevent n, *p = NULL;
1482 if (u_notification) {
1483 if (get_compat_sigevent(&n, u_notification))
1484 return -EFAULT;
1485 if (n.sigev_notify == SIGEV_THREAD)
1486 n.sigev_value.sival_ptr = compat_ptr(n.sigev_value.sival_int);
1487 p = &n;
1488 }
1489 return do_mq_notify(mqdes, p);
1490}
1491
1492COMPAT_SYSCALL_DEFINE3(mq_getsetattr, mqd_t, mqdes,
1493 const struct compat_mq_attr __user *, u_mqstat,
1494 struct compat_mq_attr __user *, u_omqstat)
1495{
1496 int ret;
1497 struct mq_attr mqstat, omqstat;
1498 struct mq_attr *new = NULL, *old = NULL;
1499
1500 if (u_mqstat) {
1501 new = &mqstat;
1502 if (get_compat_mq_attr(new, u_mqstat))
1503 return -EFAULT;
1504 }
1505 if (u_omqstat)
1506 old = &omqstat;
1507
1508 ret = do_mq_getsetattr(mqdes, new, old);
1509 if (ret || !old)
1510 return ret;
1511
1512 if (put_compat_mq_attr(old, u_omqstat))
1513 return -EFAULT;
1514 return 0;
1515}
1516#endif
1517
1518#ifdef CONFIG_COMPAT_32BIT_TIME
1519static int compat_prepare_timeout(const struct old_timespec32 __user *p,
1520 struct timespec64 *ts)
1521{
1522 if (get_old_timespec32(ts, p))
1523 return -EFAULT;
1524 if (!timespec64_valid(ts))
1525 return -EINVAL;
1526 return 0;
1527}
1528
1529SYSCALL_DEFINE5(mq_timedsend_time32, mqd_t, mqdes,
1530 const char __user *, u_msg_ptr,
1531 unsigned int, msg_len, unsigned int, msg_prio,
1532 const struct old_timespec32 __user *, u_abs_timeout)
1533{
1534 struct timespec64 ts, *p = NULL;
1535 if (u_abs_timeout) {
1536 int res = compat_prepare_timeout(u_abs_timeout, &ts);
1537 if (res)
1538 return res;
1539 p = &ts;
1540 }
1541 return do_mq_timedsend(mqdes, u_msg_ptr, msg_len, msg_prio, p);
1542}
1543
1544SYSCALL_DEFINE5(mq_timedreceive_time32, mqd_t, mqdes,
1545 char __user *, u_msg_ptr,
1546 unsigned int, msg_len, unsigned int __user *, u_msg_prio,
1547 const struct old_timespec32 __user *, u_abs_timeout)
1548{
1549 struct timespec64 ts, *p = NULL;
1550 if (u_abs_timeout) {
1551 int res = compat_prepare_timeout(u_abs_timeout, &ts);
1552 if (res)
1553 return res;
1554 p = &ts;
1555 }
1556 return do_mq_timedreceive(mqdes, u_msg_ptr, msg_len, u_msg_prio, p);
1557}
1558#endif
1559
1560static const struct inode_operations mqueue_dir_inode_operations = {
1561 .lookup = simple_lookup,
1562 .create = mqueue_create,
1563 .unlink = mqueue_unlink,
1564};
1565
1566static const struct file_operations mqueue_file_operations = {
1567 .flush = mqueue_flush_file,
1568 .poll = mqueue_poll_file,
1569 .read = mqueue_read_file,
1570 .llseek = default_llseek,
1571};
1572
1573static const struct super_operations mqueue_super_ops = {
1574 .alloc_inode = mqueue_alloc_inode,
1575 .free_inode = mqueue_free_inode,
1576 .evict_inode = mqueue_evict_inode,
1577 .statfs = simple_statfs,
1578};
1579
1580static const struct fs_context_operations mqueue_fs_context_ops = {
1581 .free = mqueue_fs_context_free,
1582 .get_tree = mqueue_get_tree,
1583};
1584
1585static struct file_system_type mqueue_fs_type = {
1586 .name = "mqueue",
1587 .init_fs_context = mqueue_init_fs_context,
1588 .kill_sb = kill_litter_super,
1589 .fs_flags = FS_USERNS_MOUNT,
1590};
1591
1592int mq_init_ns(struct ipc_namespace *ns)
1593{
1594 struct vfsmount *m;
1595
1596 ns->mq_queues_count = 0;
1597 ns->mq_queues_max = DFLT_QUEUESMAX;
1598 ns->mq_msg_max = DFLT_MSGMAX;
1599 ns->mq_msgsize_max = DFLT_MSGSIZEMAX;
1600 ns->mq_msg_default = DFLT_MSG;
1601 ns->mq_msgsize_default = DFLT_MSGSIZE;
1602
1603 m = mq_create_mount(ns);
1604 if (IS_ERR(m))
1605 return PTR_ERR(m);
1606 ns->mq_mnt = m;
1607 return 0;
1608}
1609
1610void mq_clear_sbinfo(struct ipc_namespace *ns)
1611{
1612 ns->mq_mnt->mnt_sb->s_fs_info = NULL;
1613}
1614
1615void mq_put_mnt(struct ipc_namespace *ns)
1616{
1617 kern_unmount(ns->mq_mnt);
1618}
1619
1620static int __init init_mqueue_fs(void)
1621{
1622 int error;
1623
1624 mqueue_inode_cachep = kmem_cache_create("mqueue_inode_cache",
1625 sizeof(struct mqueue_inode_info), 0,
1626 SLAB_HWCACHE_ALIGN|SLAB_ACCOUNT, init_once);
1627 if (mqueue_inode_cachep == NULL)
1628 return -ENOMEM;
1629
1630
1631 mq_sysctl_table = mq_register_sysctl_table();
1632
1633 error = register_filesystem(&mqueue_fs_type);
1634 if (error)
1635 goto out_sysctl;
1636
1637 spin_lock_init(&mq_lock);
1638
1639 error = mq_init_ns(&init_ipc_ns);
1640 if (error)
1641 goto out_filesystem;
1642
1643 return 0;
1644
1645out_filesystem:
1646 unregister_filesystem(&mqueue_fs_type);
1647out_sysctl:
1648 if (mq_sysctl_table)
1649 unregister_sysctl_table(mq_sysctl_table);
1650 kmem_cache_destroy(mqueue_inode_cachep);
1651 return error;
1652}
1653
1654device_initcall(init_mqueue_fs);
1655