2 * Fast Userspace Mutexes (which I call "Futexes!").
3 * (C) Rusty Russell, IBM 2002
5 * Generalized futexes, futex requeueing, misc fixes by Ingo Molnar
6 * (C) Copyright 2003 Red Hat Inc, All Rights Reserved
8 * Removed page pinning, fix privately mapped COW pages and other cleanups
9 * (C) Copyright 2003, 2004 Jamie Lokier
11 * Robust futex support started by Ingo Molnar
12 * (C) Copyright 2006 Red Hat Inc, All Rights Reserved
13 * Thanks to Thomas Gleixner for suggestions, analysis and fixes.
15 * PI-futex support started by Ingo Molnar and Thomas Gleixner
16 * Copyright (C) 2006 Red Hat, Inc., Ingo Molnar <mingo@redhat.com>
17 * Copyright (C) 2006 Timesys Corp., Thomas Gleixner <tglx@timesys.com>
19 * PRIVATE futexes by Eric Dumazet
20 * Copyright (C) 2007 Eric Dumazet <dada1@cosmosbay.com>
22 * Thanks to Ben LaHaise for yelling "hashed waitqueues" loudly
23 * enough at me, Linus for the original (flawed) idea, Matthew
24 * Kirkwood for proof-of-concept implementation.
26 * "The futexes are also cursed."
27 * "But they come in a choice of three flavours!"
29 * This program is free software; you can redistribute it and/or modify
30 * it under the terms of the GNU General Public License as published by
31 * the Free Software Foundation; either version 2 of the License, or
32 * (at your option) any later version.
34 * This program is distributed in the hope that it will be useful,
35 * but WITHOUT ANY WARRANTY; without even the implied warranty of
36 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
37 * GNU General Public License for more details.
39 * You should have received a copy of the GNU General Public License
40 * along with this program; if not, write to the Free Software
41 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
43 #include <linux/slab.h>
44 #include <linux/poll.h>
46 #include <linux/file.h>
47 #include <linux/jhash.h>
48 #include <linux/init.h>
49 #include <linux/futex.h>
50 #include <linux/mount.h>
51 #include <linux/pagemap.h>
52 #include <linux/syscalls.h>
53 #include <linux/signal.h>
54 #include <linux/module.h>
55 #include <linux/magic.h>
56 #include <linux/pid.h>
57 #include <linux/nsproxy.h>
59 #include <asm/futex.h>
61 #include "rtmutex_common.h"
63 int __read_mostly futex_cmpxchg_enabled
;
65 #define FUTEX_HASHBITS (CONFIG_BASE_SMALL ? 4 : 8)
68 * Priority Inheritance state:
70 struct futex_pi_state
{
72 * list of 'owned' pi_state instances - these have to be
73 * cleaned up in do_exit() if the task exits prematurely:
75 struct list_head list
;
80 struct rt_mutex pi_mutex
;
82 struct task_struct
*owner
;
89 * We use this hashed waitqueue instead of a normal wait_queue_t, so
90 * we can wake only the relevant ones (hashed queues may be shared).
92 * A futex_q has a woken state, just like tasks have TASK_RUNNING.
93 * It is considered woken when plist_node_empty(&q->list) || q->lock_ptr == 0.
94 * The order of wakup is always to make the first condition true, then
95 * wake up q->waiters, then make the second condition true.
98 struct plist_node list
;
99 wait_queue_head_t waiters
;
101 /* Which hash list lock to use: */
102 spinlock_t
*lock_ptr
;
104 /* Key which the futex is hashed on: */
107 /* Optional priority inheritance state: */
108 struct futex_pi_state
*pi_state
;
109 struct task_struct
*task
;
111 /* Bitset for the optional bitmasked wakeup */
116 * Split the global futex_lock into every hash list lock.
118 struct futex_hash_bucket
{
120 struct plist_head chain
;
123 static struct futex_hash_bucket futex_queues
[1<<FUTEX_HASHBITS
];
126 * Take mm->mmap_sem, when futex is shared
128 static inline void futex_lock_mm(struct rw_semaphore
*fshared
)
135 * Release mm->mmap_sem, when the futex is shared
137 static inline void futex_unlock_mm(struct rw_semaphore
*fshared
)
144 * We hash on the keys returned from get_futex_key (see below).
146 static struct futex_hash_bucket
*hash_futex(union futex_key
*key
)
148 u32 hash
= jhash2((u32
*)&key
->both
.word
,
149 (sizeof(key
->both
.word
)+sizeof(key
->both
.ptr
))/4,
151 return &futex_queues
[hash
& ((1 << FUTEX_HASHBITS
)-1)];
155 * Return 1 if two futex_keys are equal, 0 otherwise.
157 static inline int match_futex(union futex_key
*key1
, union futex_key
*key2
)
159 return (key1
->both
.word
== key2
->both
.word
160 && key1
->both
.ptr
== key2
->both
.ptr
161 && key1
->both
.offset
== key2
->both
.offset
);
165 * get_futex_key - Get parameters which are the keys for a futex.
166 * @uaddr: virtual address of the futex
167 * @shared: NULL for a PROCESS_PRIVATE futex,
168 * ¤t->mm->mmap_sem for a PROCESS_SHARED futex
169 * @key: address where result is stored.
171 * Returns a negative error code or 0
172 * The key words are stored in *key on success.
174 * For shared mappings, it's (page->index, vma->vm_file->f_path.dentry->d_inode,
175 * offset_within_page). For private mappings, it's (uaddr, current->mm).
176 * We can usually work out the index without swapping in the page.
178 * fshared is NULL for PROCESS_PRIVATE futexes
179 * For other futexes, it points to ¤t->mm->mmap_sem and
180 * caller must have taken the reader lock. but NOT any spinlocks.
182 static int get_futex_key(u32 __user
*uaddr
, struct rw_semaphore
*fshared
,
183 union futex_key
*key
)
185 unsigned long address
= (unsigned long)uaddr
;
186 struct mm_struct
*mm
= current
->mm
;
187 struct vm_area_struct
*vma
;
192 * The futex address must be "naturally" aligned.
194 key
->both
.offset
= address
% PAGE_SIZE
;
195 if (unlikely((address
% sizeof(u32
)) != 0))
197 address
-= key
->both
.offset
;
200 * PROCESS_PRIVATE futexes are fast.
201 * As the mm cannot disappear under us and the 'key' only needs
202 * virtual address, we dont even have to find the underlying vma.
203 * Note : We do have to check 'uaddr' is a valid user address,
204 * but access_ok() should be faster than find_vma()
207 if (unlikely(!access_ok(VERIFY_WRITE
, uaddr
, sizeof(u32
))))
209 key
->private.mm
= mm
;
210 key
->private.address
= address
;
214 * The futex is hashed differently depending on whether
215 * it's in a shared or private mapping. So check vma first.
217 vma
= find_extend_vma(mm
, address
);
224 if (unlikely((vma
->vm_flags
& (VM_IO
|VM_READ
)) != VM_READ
))
225 return (vma
->vm_flags
& VM_IO
) ? -EPERM
: -EACCES
;
228 * Private mappings are handled in a simple way.
230 * NOTE: When userspace waits on a MAP_SHARED mapping, even if
231 * it's a read-only handle, it's expected that futexes attach to
232 * the object not the particular process. Therefore we use
233 * VM_MAYSHARE here, not VM_SHARED which is restricted to shared
234 * mappings of _writable_ handles.
236 if (likely(!(vma
->vm_flags
& VM_MAYSHARE
))) {
237 key
->both
.offset
|= FUT_OFF_MMSHARED
; /* reference taken on mm */
238 key
->private.mm
= mm
;
239 key
->private.address
= address
;
244 * Linear file mappings are also simple.
246 key
->shared
.inode
= vma
->vm_file
->f_path
.dentry
->d_inode
;
247 key
->both
.offset
|= FUT_OFF_INODE
; /* inode-based key. */
248 if (likely(!(vma
->vm_flags
& VM_NONLINEAR
))) {
249 key
->shared
.pgoff
= (((address
- vma
->vm_start
) >> PAGE_SHIFT
)
255 * We could walk the page table to read the non-linear
256 * pte, and get the page index without fetching the page
257 * from swap. But that's a lot of code to duplicate here
258 * for a rare case, so we simply fetch the page.
260 err
= get_user_pages(current
, mm
, address
, 1, 0, 0, &page
, NULL
);
263 page
->index
<< (PAGE_CACHE_SHIFT
- PAGE_SHIFT
);
271 * Take a reference to the resource addressed by a key.
272 * Can be called while holding spinlocks.
275 static void get_futex_key_refs(union futex_key
*key
)
277 if (key
->both
.ptr
== NULL
)
279 switch (key
->both
.offset
& (FUT_OFF_INODE
|FUT_OFF_MMSHARED
)) {
281 atomic_inc(&key
->shared
.inode
->i_count
);
283 case FUT_OFF_MMSHARED
:
284 atomic_inc(&key
->private.mm
->mm_count
);
290 * Drop a reference to the resource addressed by a key.
291 * The hash bucket spinlock must not be held.
293 static void drop_futex_key_refs(union futex_key
*key
)
297 switch (key
->both
.offset
& (FUT_OFF_INODE
|FUT_OFF_MMSHARED
)) {
299 iput(key
->shared
.inode
);
301 case FUT_OFF_MMSHARED
:
302 mmdrop(key
->private.mm
);
307 static u32
cmpxchg_futex_value_locked(u32 __user
*uaddr
, u32 uval
, u32 newval
)
312 curval
= futex_atomic_cmpxchg_inatomic(uaddr
, uval
, newval
);
318 static int get_futex_value_locked(u32
*dest
, u32 __user
*from
)
323 ret
= __copy_from_user_inatomic(dest
, from
, sizeof(u32
));
326 return ret
? -EFAULT
: 0;
331 * if fshared is non NULL, current->mm->mmap_sem is already held
333 static int futex_handle_fault(unsigned long address
,
334 struct rw_semaphore
*fshared
, int attempt
)
336 struct vm_area_struct
* vma
;
337 struct mm_struct
*mm
= current
->mm
;
344 down_read(&mm
->mmap_sem
);
345 vma
= find_vma(mm
, address
);
346 if (vma
&& address
>= vma
->vm_start
&&
347 (vma
->vm_flags
& VM_WRITE
)) {
349 fault
= handle_mm_fault(mm
, vma
, address
, 1);
350 if (unlikely((fault
& VM_FAULT_ERROR
))) {
352 /* XXX: let's do this when we verify it is OK */
353 if (ret
& VM_FAULT_OOM
)
358 if (fault
& VM_FAULT_MAJOR
)
365 up_read(&mm
->mmap_sem
);
372 static int refill_pi_state_cache(void)
374 struct futex_pi_state
*pi_state
;
376 if (likely(current
->pi_state_cache
))
379 pi_state
= kzalloc(sizeof(*pi_state
), GFP_KERNEL
);
384 INIT_LIST_HEAD(&pi_state
->list
);
385 /* pi_mutex gets initialized later */
386 pi_state
->owner
= NULL
;
387 atomic_set(&pi_state
->refcount
, 1);
389 current
->pi_state_cache
= pi_state
;
394 static struct futex_pi_state
* alloc_pi_state(void)
396 struct futex_pi_state
*pi_state
= current
->pi_state_cache
;
399 current
->pi_state_cache
= NULL
;
404 static void free_pi_state(struct futex_pi_state
*pi_state
)
406 if (!atomic_dec_and_test(&pi_state
->refcount
))
410 * If pi_state->owner is NULL, the owner is most probably dying
411 * and has cleaned up the pi_state already
413 if (pi_state
->owner
) {
414 spin_lock_irq(&pi_state
->owner
->pi_lock
);
415 list_del_init(&pi_state
->list
);
416 spin_unlock_irq(&pi_state
->owner
->pi_lock
);
418 rt_mutex_proxy_unlock(&pi_state
->pi_mutex
, pi_state
->owner
);
421 if (current
->pi_state_cache
)
425 * pi_state->list is already empty.
426 * clear pi_state->owner.
427 * refcount is at 0 - put it back to 1.
429 pi_state
->owner
= NULL
;
430 atomic_set(&pi_state
->refcount
, 1);
431 current
->pi_state_cache
= pi_state
;
436 * Look up the task based on what TID userspace gave us.
439 static struct task_struct
* futex_find_get_task(pid_t pid
)
441 struct task_struct
*p
;
442 uid_t euid
= current_euid();
445 p
= find_task_by_vpid(pid
);
446 if (!p
|| (euid
!= p
->cred
->euid
&&
447 euid
!= p
->cred
->uid
))
458 * This task is holding PI mutexes at exit time => bad.
459 * Kernel cleans up PI-state, but userspace is likely hosed.
460 * (Robust-futex cleanup is separate and might save the day for userspace.)
462 void exit_pi_state_list(struct task_struct
*curr
)
464 struct list_head
*next
, *head
= &curr
->pi_state_list
;
465 struct futex_pi_state
*pi_state
;
466 struct futex_hash_bucket
*hb
;
469 if (!futex_cmpxchg_enabled
)
472 * We are a ZOMBIE and nobody can enqueue itself on
473 * pi_state_list anymore, but we have to be careful
474 * versus waiters unqueueing themselves:
476 spin_lock_irq(&curr
->pi_lock
);
477 while (!list_empty(head
)) {
480 pi_state
= list_entry(next
, struct futex_pi_state
, list
);
482 hb
= hash_futex(&key
);
483 spin_unlock_irq(&curr
->pi_lock
);
485 spin_lock(&hb
->lock
);
487 spin_lock_irq(&curr
->pi_lock
);
489 * We dropped the pi-lock, so re-check whether this
490 * task still owns the PI-state:
492 if (head
->next
!= next
) {
493 spin_unlock(&hb
->lock
);
497 WARN_ON(pi_state
->owner
!= curr
);
498 WARN_ON(list_empty(&pi_state
->list
));
499 list_del_init(&pi_state
->list
);
500 pi_state
->owner
= NULL
;
501 spin_unlock_irq(&curr
->pi_lock
);
503 rt_mutex_unlock(&pi_state
->pi_mutex
);
505 spin_unlock(&hb
->lock
);
507 spin_lock_irq(&curr
->pi_lock
);
509 spin_unlock_irq(&curr
->pi_lock
);
513 lookup_pi_state(u32 uval
, struct futex_hash_bucket
*hb
,
514 union futex_key
*key
, struct futex_pi_state
**ps
)
516 struct futex_pi_state
*pi_state
= NULL
;
517 struct futex_q
*this, *next
;
518 struct plist_head
*head
;
519 struct task_struct
*p
;
520 pid_t pid
= uval
& FUTEX_TID_MASK
;
524 plist_for_each_entry_safe(this, next
, head
, list
) {
525 if (match_futex(&this->key
, key
)) {
527 * Another waiter already exists - bump up
528 * the refcount and return its pi_state:
530 pi_state
= this->pi_state
;
532 * Userspace might have messed up non PI and PI futexes
534 if (unlikely(!pi_state
))
537 WARN_ON(!atomic_read(&pi_state
->refcount
));
538 WARN_ON(pid
&& pi_state
->owner
&&
539 pi_state
->owner
->pid
!= pid
);
541 atomic_inc(&pi_state
->refcount
);
549 * We are the first waiter - try to look up the real owner and attach
550 * the new pi_state to it, but bail out when TID = 0
554 p
= futex_find_get_task(pid
);
559 * We need to look at the task state flags to figure out,
560 * whether the task is exiting. To protect against the do_exit
561 * change of the task flags, we do this protected by
564 spin_lock_irq(&p
->pi_lock
);
565 if (unlikely(p
->flags
& PF_EXITING
)) {
567 * The task is on the way out. When PF_EXITPIDONE is
568 * set, we know that the task has finished the
571 int ret
= (p
->flags
& PF_EXITPIDONE
) ? -ESRCH
: -EAGAIN
;
573 spin_unlock_irq(&p
->pi_lock
);
578 pi_state
= alloc_pi_state();
581 * Initialize the pi_mutex in locked state and make 'p'
584 rt_mutex_init_proxy_locked(&pi_state
->pi_mutex
, p
);
586 /* Store the key for possible exit cleanups: */
587 pi_state
->key
= *key
;
589 WARN_ON(!list_empty(&pi_state
->list
));
590 list_add(&pi_state
->list
, &p
->pi_state_list
);
592 spin_unlock_irq(&p
->pi_lock
);
602 * The hash bucket lock must be held when this is called.
603 * Afterwards, the futex_q must not be accessed.
605 static void wake_futex(struct futex_q
*q
)
607 plist_del(&q
->list
, &q
->list
.plist
);
609 * The lock in wake_up_all() is a crucial memory barrier after the
610 * plist_del() and also before assigning to q->lock_ptr.
612 wake_up_all(&q
->waiters
);
614 * The waiting task can free the futex_q as soon as this is written,
615 * without taking any locks. This must come last.
617 * A memory barrier is required here to prevent the following store
618 * to lock_ptr from getting ahead of the wakeup. Clearing the lock
619 * at the end of wake_up_all() does not prevent this store from
626 static int wake_futex_pi(u32 __user
*uaddr
, u32 uval
, struct futex_q
*this)
628 struct task_struct
*new_owner
;
629 struct futex_pi_state
*pi_state
= this->pi_state
;
635 spin_lock(&pi_state
->pi_mutex
.wait_lock
);
636 new_owner
= rt_mutex_next_owner(&pi_state
->pi_mutex
);
639 * This happens when we have stolen the lock and the original
640 * pending owner did not enqueue itself back on the rt_mutex.
641 * Thats not a tragedy. We know that way, that a lock waiter
642 * is on the fly. We make the futex_q waiter the pending owner.
645 new_owner
= this->task
;
648 * We pass it to the next owner. (The WAITERS bit is always
649 * kept enabled while there is PI state around. We must also
650 * preserve the owner died bit.)
652 if (!(uval
& FUTEX_OWNER_DIED
)) {
655 newval
= FUTEX_WAITERS
| task_pid_vnr(new_owner
);
657 curval
= cmpxchg_futex_value_locked(uaddr
, uval
, newval
);
659 if (curval
== -EFAULT
)
661 else if (curval
!= uval
)
664 spin_unlock(&pi_state
->pi_mutex
.wait_lock
);
669 spin_lock_irq(&pi_state
->owner
->pi_lock
);
670 WARN_ON(list_empty(&pi_state
->list
));
671 list_del_init(&pi_state
->list
);
672 spin_unlock_irq(&pi_state
->owner
->pi_lock
);
674 spin_lock_irq(&new_owner
->pi_lock
);
675 WARN_ON(!list_empty(&pi_state
->list
));
676 list_add(&pi_state
->list
, &new_owner
->pi_state_list
);
677 pi_state
->owner
= new_owner
;
678 spin_unlock_irq(&new_owner
->pi_lock
);
680 spin_unlock(&pi_state
->pi_mutex
.wait_lock
);
681 rt_mutex_unlock(&pi_state
->pi_mutex
);
686 static int unlock_futex_pi(u32 __user
*uaddr
, u32 uval
)
691 * There is no waiter, so we unlock the futex. The owner died
692 * bit has not to be preserved here. We are the owner:
694 oldval
= cmpxchg_futex_value_locked(uaddr
, uval
, 0);
696 if (oldval
== -EFAULT
)
705 * Express the locking dependencies for lockdep:
708 double_lock_hb(struct futex_hash_bucket
*hb1
, struct futex_hash_bucket
*hb2
)
711 spin_lock(&hb1
->lock
);
713 spin_lock_nested(&hb2
->lock
, SINGLE_DEPTH_NESTING
);
714 } else { /* hb1 > hb2 */
715 spin_lock(&hb2
->lock
);
716 spin_lock_nested(&hb1
->lock
, SINGLE_DEPTH_NESTING
);
721 * Wake up all waiters hashed on the physical page that is mapped
722 * to this virtual address:
724 static int futex_wake(u32 __user
*uaddr
, struct rw_semaphore
*fshared
,
725 int nr_wake
, u32 bitset
)
727 struct futex_hash_bucket
*hb
;
728 struct futex_q
*this, *next
;
729 struct plist_head
*head
;
736 futex_lock_mm(fshared
);
738 ret
= get_futex_key(uaddr
, fshared
, &key
);
739 if (unlikely(ret
!= 0))
742 hb
= hash_futex(&key
);
743 spin_lock(&hb
->lock
);
746 plist_for_each_entry_safe(this, next
, head
, list
) {
747 if (match_futex (&this->key
, &key
)) {
748 if (this->pi_state
) {
753 /* Check if one of the bits is set in both bitsets */
754 if (!(this->bitset
& bitset
))
758 if (++ret
>= nr_wake
)
763 spin_unlock(&hb
->lock
);
765 futex_unlock_mm(fshared
);
770 * Wake up all waiters hashed on the physical page that is mapped
771 * to this virtual address:
774 futex_wake_op(u32 __user
*uaddr1
, struct rw_semaphore
*fshared
,
776 int nr_wake
, int nr_wake2
, int op
)
778 union futex_key key1
, key2
;
779 struct futex_hash_bucket
*hb1
, *hb2
;
780 struct plist_head
*head
;
781 struct futex_q
*this, *next
;
782 int ret
, op_ret
, attempt
= 0;
785 futex_lock_mm(fshared
);
787 ret
= get_futex_key(uaddr1
, fshared
, &key1
);
788 if (unlikely(ret
!= 0))
790 ret
= get_futex_key(uaddr2
, fshared
, &key2
);
791 if (unlikely(ret
!= 0))
794 hb1
= hash_futex(&key1
);
795 hb2
= hash_futex(&key2
);
798 double_lock_hb(hb1
, hb2
);
800 op_ret
= futex_atomic_op_inuser(op
, uaddr2
);
801 if (unlikely(op_ret
< 0)) {
804 spin_unlock(&hb1
->lock
);
806 spin_unlock(&hb2
->lock
);
810 * we don't get EFAULT from MMU faults if we don't have an MMU,
811 * but we might get them from range checking
817 if (unlikely(op_ret
!= -EFAULT
)) {
823 * futex_atomic_op_inuser needs to both read and write
824 * *(int __user *)uaddr2, but we can't modify it
825 * non-atomically. Therefore, if get_user below is not
826 * enough, we need to handle the fault ourselves, while
827 * still holding the mmap_sem.
830 ret
= futex_handle_fault((unsigned long)uaddr2
,
838 * If we would have faulted, release mmap_sem,
839 * fault it in and start all over again.
841 futex_unlock_mm(fshared
);
843 ret
= get_user(dummy
, uaddr2
);
852 plist_for_each_entry_safe(this, next
, head
, list
) {
853 if (match_futex (&this->key
, &key1
)) {
855 if (++ret
>= nr_wake
)
864 plist_for_each_entry_safe(this, next
, head
, list
) {
865 if (match_futex (&this->key
, &key2
)) {
867 if (++op_ret
>= nr_wake2
)
874 spin_unlock(&hb1
->lock
);
876 spin_unlock(&hb2
->lock
);
878 futex_unlock_mm(fshared
);
884 * Requeue all waiters hashed on one physical page to another
887 static int futex_requeue(u32 __user
*uaddr1
, struct rw_semaphore
*fshared
,
889 int nr_wake
, int nr_requeue
, u32
*cmpval
)
891 union futex_key key1
, key2
;
892 struct futex_hash_bucket
*hb1
, *hb2
;
893 struct plist_head
*head1
;
894 struct futex_q
*this, *next
;
895 int ret
, drop_count
= 0;
898 futex_lock_mm(fshared
);
900 ret
= get_futex_key(uaddr1
, fshared
, &key1
);
901 if (unlikely(ret
!= 0))
903 ret
= get_futex_key(uaddr2
, fshared
, &key2
);
904 if (unlikely(ret
!= 0))
907 hb1
= hash_futex(&key1
);
908 hb2
= hash_futex(&key2
);
910 double_lock_hb(hb1
, hb2
);
912 if (likely(cmpval
!= NULL
)) {
915 ret
= get_futex_value_locked(&curval
, uaddr1
);
918 spin_unlock(&hb1
->lock
);
920 spin_unlock(&hb2
->lock
);
923 * If we would have faulted, release mmap_sem, fault
924 * it in and start all over again.
926 futex_unlock_mm(fshared
);
928 ret
= get_user(curval
, uaddr1
);
935 if (curval
!= *cmpval
) {
942 plist_for_each_entry_safe(this, next
, head1
, list
) {
943 if (!match_futex (&this->key
, &key1
))
945 if (++ret
<= nr_wake
) {
949 * If key1 and key2 hash to the same bucket, no need to
952 if (likely(head1
!= &hb2
->chain
)) {
953 plist_del(&this->list
, &hb1
->chain
);
954 plist_add(&this->list
, &hb2
->chain
);
955 this->lock_ptr
= &hb2
->lock
;
956 #ifdef CONFIG_DEBUG_PI_LIST
957 this->list
.plist
.lock
= &hb2
->lock
;
961 get_futex_key_refs(&key2
);
964 if (ret
- nr_wake
>= nr_requeue
)
970 spin_unlock(&hb1
->lock
);
972 spin_unlock(&hb2
->lock
);
974 /* drop_futex_key_refs() must be called outside the spinlocks. */
975 while (--drop_count
>= 0)
976 drop_futex_key_refs(&key1
);
979 futex_unlock_mm(fshared
);
983 /* The key must be already stored in q->key. */
984 static inline struct futex_hash_bucket
*queue_lock(struct futex_q
*q
)
986 struct futex_hash_bucket
*hb
;
988 init_waitqueue_head(&q
->waiters
);
990 get_futex_key_refs(&q
->key
);
991 hb
= hash_futex(&q
->key
);
992 q
->lock_ptr
= &hb
->lock
;
994 spin_lock(&hb
->lock
);
998 static inline void queue_me(struct futex_q
*q
, struct futex_hash_bucket
*hb
)
1003 * The priority used to register this element is
1004 * - either the real thread-priority for the real-time threads
1005 * (i.e. threads with a priority lower than MAX_RT_PRIO)
1006 * - or MAX_RT_PRIO for non-RT threads.
1007 * Thus, all RT-threads are woken first in priority order, and
1008 * the others are woken last, in FIFO order.
1010 prio
= min(current
->normal_prio
, MAX_RT_PRIO
);
1012 plist_node_init(&q
->list
, prio
);
1013 #ifdef CONFIG_DEBUG_PI_LIST
1014 q
->list
.plist
.lock
= &hb
->lock
;
1016 plist_add(&q
->list
, &hb
->chain
);
1018 spin_unlock(&hb
->lock
);
1022 queue_unlock(struct futex_q
*q
, struct futex_hash_bucket
*hb
)
1024 spin_unlock(&hb
->lock
);
1025 drop_futex_key_refs(&q
->key
);
1029 * queue_me and unqueue_me must be called as a pair, each
1030 * exactly once. They are called with the hashed spinlock held.
1033 /* Return 1 if we were still queued (ie. 0 means we were woken) */
1034 static int unqueue_me(struct futex_q
*q
)
1036 spinlock_t
*lock_ptr
;
1039 /* In the common case we don't take the spinlock, which is nice. */
1041 lock_ptr
= q
->lock_ptr
;
1043 if (lock_ptr
!= NULL
) {
1044 spin_lock(lock_ptr
);
1046 * q->lock_ptr can change between reading it and
1047 * spin_lock(), causing us to take the wrong lock. This
1048 * corrects the race condition.
1050 * Reasoning goes like this: if we have the wrong lock,
1051 * q->lock_ptr must have changed (maybe several times)
1052 * between reading it and the spin_lock(). It can
1053 * change again after the spin_lock() but only if it was
1054 * already changed before the spin_lock(). It cannot,
1055 * however, change back to the original value. Therefore
1056 * we can detect whether we acquired the correct lock.
1058 if (unlikely(lock_ptr
!= q
->lock_ptr
)) {
1059 spin_unlock(lock_ptr
);
1062 WARN_ON(plist_node_empty(&q
->list
));
1063 plist_del(&q
->list
, &q
->list
.plist
);
1065 BUG_ON(q
->pi_state
);
1067 spin_unlock(lock_ptr
);
1071 drop_futex_key_refs(&q
->key
);
1076 * PI futexes can not be requeued and must remove themself from the
1077 * hash bucket. The hash bucket lock (i.e. lock_ptr) is held on entry
1080 static void unqueue_me_pi(struct futex_q
*q
)
1082 WARN_ON(plist_node_empty(&q
->list
));
1083 plist_del(&q
->list
, &q
->list
.plist
);
1085 BUG_ON(!q
->pi_state
);
1086 free_pi_state(q
->pi_state
);
1089 spin_unlock(q
->lock_ptr
);
1091 drop_futex_key_refs(&q
->key
);
1095 * Fixup the pi_state owner with the new owner.
1097 * Must be called with hash bucket lock held and mm->sem held for non
1100 static int fixup_pi_state_owner(u32 __user
*uaddr
, struct futex_q
*q
,
1101 struct task_struct
*newowner
,
1102 struct rw_semaphore
*fshared
)
1104 u32 newtid
= task_pid_vnr(newowner
) | FUTEX_WAITERS
;
1105 struct futex_pi_state
*pi_state
= q
->pi_state
;
1106 struct task_struct
*oldowner
= pi_state
->owner
;
1107 u32 uval
, curval
, newval
;
1108 int ret
, attempt
= 0;
1111 if (!pi_state
->owner
)
1112 newtid
|= FUTEX_OWNER_DIED
;
1115 * We are here either because we stole the rtmutex from the
1116 * pending owner or we are the pending owner which failed to
1117 * get the rtmutex. We have to replace the pending owner TID
1118 * in the user space variable. This must be atomic as we have
1119 * to preserve the owner died bit here.
1121 * Note: We write the user space value _before_ changing the
1122 * pi_state because we can fault here. Imagine swapped out
1123 * pages or a fork, which was running right before we acquired
1124 * mmap_sem, that marked all the anonymous memory readonly for
1127 * Modifying pi_state _before_ the user space value would
1128 * leave the pi_state in an inconsistent state when we fault
1129 * here, because we need to drop the hash bucket lock to
1130 * handle the fault. This might be observed in the PID check
1131 * in lookup_pi_state.
1134 if (get_futex_value_locked(&uval
, uaddr
))
1138 newval
= (uval
& FUTEX_OWNER_DIED
) | newtid
;
1140 curval
= cmpxchg_futex_value_locked(uaddr
, uval
, newval
);
1142 if (curval
== -EFAULT
)
1150 * We fixed up user space. Now we need to fix the pi_state
1153 if (pi_state
->owner
!= NULL
) {
1154 spin_lock_irq(&pi_state
->owner
->pi_lock
);
1155 WARN_ON(list_empty(&pi_state
->list
));
1156 list_del_init(&pi_state
->list
);
1157 spin_unlock_irq(&pi_state
->owner
->pi_lock
);
1160 pi_state
->owner
= newowner
;
1162 spin_lock_irq(&newowner
->pi_lock
);
1163 WARN_ON(!list_empty(&pi_state
->list
));
1164 list_add(&pi_state
->list
, &newowner
->pi_state_list
);
1165 spin_unlock_irq(&newowner
->pi_lock
);
1169 * To handle the page fault we need to drop the hash bucket
1170 * lock here. That gives the other task (either the pending
1171 * owner itself or the task which stole the rtmutex) the
1172 * chance to try the fixup of the pi_state. So once we are
1173 * back from handling the fault we need to check the pi_state
1174 * after reacquiring the hash bucket lock and before trying to
1175 * do another fixup. When the fixup has been done already we
1179 spin_unlock(q
->lock_ptr
);
1181 ret
= futex_handle_fault((unsigned long)uaddr
, fshared
, attempt
++);
1183 spin_lock(q
->lock_ptr
);
1186 * Check if someone else fixed it for us:
1188 if (pi_state
->owner
!= oldowner
)
1198 * In case we must use restart_block to restart a futex_wait,
1199 * we encode in the 'flags' shared capability
1201 #define FLAGS_SHARED 1
1203 static long futex_wait_restart(struct restart_block
*restart
);
1205 static int futex_wait(u32 __user
*uaddr
, struct rw_semaphore
*fshared
,
1206 u32 val
, ktime_t
*abs_time
, u32 bitset
)
1208 struct task_struct
*curr
= current
;
1209 DECLARE_WAITQUEUE(wait
, curr
);
1210 struct futex_hash_bucket
*hb
;
1214 struct hrtimer_sleeper t
;
1223 futex_lock_mm(fshared
);
1225 ret
= get_futex_key(uaddr
, fshared
, &q
.key
);
1226 if (unlikely(ret
!= 0))
1227 goto out_release_sem
;
1229 hb
= queue_lock(&q
);
1232 * Access the page AFTER the futex is queued.
1233 * Order is important:
1235 * Userspace waiter: val = var; if (cond(val)) futex_wait(&var, val);
1236 * Userspace waker: if (cond(var)) { var = new; futex_wake(&var); }
1238 * The basic logical guarantee of a futex is that it blocks ONLY
1239 * if cond(var) is known to be true at the time of blocking, for
1240 * any cond. If we queued after testing *uaddr, that would open
1241 * a race condition where we could block indefinitely with
1242 * cond(var) false, which would violate the guarantee.
1244 * A consequence is that futex_wait() can return zero and absorb
1245 * a wakeup when *uaddr != val on entry to the syscall. This is
1248 * for shared futexes, we hold the mmap semaphore, so the mapping
1249 * cannot have changed since we looked it up in get_futex_key.
1251 ret
= get_futex_value_locked(&uval
, uaddr
);
1253 if (unlikely(ret
)) {
1254 queue_unlock(&q
, hb
);
1257 * If we would have faulted, release mmap_sem, fault it in and
1258 * start all over again.
1260 futex_unlock_mm(fshared
);
1262 ret
= get_user(uval
, uaddr
);
1270 goto out_unlock_release_sem
;
1272 /* Only actually queue if *uaddr contained val. */
1276 * Now the futex is queued and we have checked the data, we
1277 * don't want to hold mmap_sem while we sleep.
1279 futex_unlock_mm(fshared
);
1282 * There might have been scheduling since the queue_me(), as we
1283 * cannot hold a spinlock across the get_user() in case it
1284 * faults, and we cannot just set TASK_INTERRUPTIBLE state when
1285 * queueing ourselves into the futex hash. This code thus has to
1286 * rely on the futex_wake() code removing us from hash when it
1290 /* add_wait_queue is the barrier after __set_current_state. */
1291 __set_current_state(TASK_INTERRUPTIBLE
);
1292 add_wait_queue(&q
.waiters
, &wait
);
1294 * !plist_node_empty() is safe here without any lock.
1295 * q.lock_ptr != 0 is not safe, because of ordering against wakeup.
1297 if (likely(!plist_node_empty(&q
.list
))) {
1301 unsigned long slack
;
1302 slack
= current
->timer_slack_ns
;
1303 if (rt_task(current
))
1305 hrtimer_init_on_stack(&t
.timer
, CLOCK_MONOTONIC
,
1307 hrtimer_init_sleeper(&t
, current
);
1308 hrtimer_set_expires_range_ns(&t
.timer
, *abs_time
, slack
);
1310 hrtimer_start_expires(&t
.timer
, HRTIMER_MODE_ABS
);
1311 if (!hrtimer_active(&t
.timer
))
1315 * the timer could have already expired, in which
1316 * case current would be flagged for rescheduling.
1317 * Don't bother calling schedule.
1322 hrtimer_cancel(&t
.timer
);
1324 /* Flag if a timeout occured */
1325 rem
= (t
.task
== NULL
);
1327 destroy_hrtimer_on_stack(&t
.timer
);
1330 __set_current_state(TASK_RUNNING
);
1333 * NOTE: we don't remove ourselves from the waitqueue because
1334 * we are the only user of it.
1337 /* If we were woken (and unqueued), we succeeded, whatever. */
1338 if (!unqueue_me(&q
))
1344 * We expect signal_pending(current), but another thread may
1345 * have handled it for us already.
1348 return -ERESTARTSYS
;
1350 struct restart_block
*restart
;
1351 restart
= ¤t_thread_info()->restart_block
;
1352 restart
->fn
= futex_wait_restart
;
1353 restart
->futex
.uaddr
= (u32
*)uaddr
;
1354 restart
->futex
.val
= val
;
1355 restart
->futex
.time
= abs_time
->tv64
;
1356 restart
->futex
.bitset
= bitset
;
1357 restart
->futex
.flags
= 0;
1360 restart
->futex
.flags
|= FLAGS_SHARED
;
1361 return -ERESTART_RESTARTBLOCK
;
1364 out_unlock_release_sem
:
1365 queue_unlock(&q
, hb
);
1368 futex_unlock_mm(fshared
);
1373 static long futex_wait_restart(struct restart_block
*restart
)
1375 u32 __user
*uaddr
= (u32 __user
*)restart
->futex
.uaddr
;
1376 struct rw_semaphore
*fshared
= NULL
;
1379 t
.tv64
= restart
->futex
.time
;
1380 restart
->fn
= do_no_restart_syscall
;
1381 if (restart
->futex
.flags
& FLAGS_SHARED
)
1382 fshared
= ¤t
->mm
->mmap_sem
;
1383 return (long)futex_wait(uaddr
, fshared
, restart
->futex
.val
, &t
,
1384 restart
->futex
.bitset
);
1389 * Userspace tried a 0 -> TID atomic transition of the futex value
1390 * and failed. The kernel side here does the whole locking operation:
1391 * if there are waiters then it will block, it does PI, etc. (Due to
1392 * races the kernel might see a 0 value of the futex too.)
1394 static int futex_lock_pi(u32 __user
*uaddr
, struct rw_semaphore
*fshared
,
1395 int detect
, ktime_t
*time
, int trylock
)
1397 struct hrtimer_sleeper timeout
, *to
= NULL
;
1398 struct task_struct
*curr
= current
;
1399 struct futex_hash_bucket
*hb
;
1400 u32 uval
, newval
, curval
;
1402 int ret
, lock_taken
, ownerdied
= 0, attempt
= 0;
1404 if (refill_pi_state_cache())
1409 hrtimer_init_on_stack(&to
->timer
, CLOCK_REALTIME
,
1411 hrtimer_init_sleeper(to
, current
);
1412 hrtimer_set_expires(&to
->timer
, *time
);
1417 futex_lock_mm(fshared
);
1419 ret
= get_futex_key(uaddr
, fshared
, &q
.key
);
1420 if (unlikely(ret
!= 0))
1421 goto out_release_sem
;
1424 hb
= queue_lock(&q
);
1427 ret
= lock_taken
= 0;
1430 * To avoid races, we attempt to take the lock here again
1431 * (by doing a 0 -> TID atomic cmpxchg), while holding all
1432 * the locks. It will most likely not succeed.
1434 newval
= task_pid_vnr(current
);
1436 curval
= cmpxchg_futex_value_locked(uaddr
, 0, newval
);
1438 if (unlikely(curval
== -EFAULT
))
1442 * Detect deadlocks. In case of REQUEUE_PI this is a valid
1443 * situation and we return success to user space.
1445 if (unlikely((curval
& FUTEX_TID_MASK
) == task_pid_vnr(current
))) {
1447 goto out_unlock_release_sem
;
1451 * Surprise - we got the lock. Just return to userspace:
1453 if (unlikely(!curval
))
1454 goto out_unlock_release_sem
;
1459 * Set the WAITERS flag, so the owner will know it has someone
1460 * to wake at next unlock
1462 newval
= curval
| FUTEX_WAITERS
;
1465 * There are two cases, where a futex might have no owner (the
1466 * owner TID is 0): OWNER_DIED. We take over the futex in this
1467 * case. We also do an unconditional take over, when the owner
1468 * of the futex died.
1470 * This is safe as we are protected by the hash bucket lock !
1472 if (unlikely(ownerdied
|| !(curval
& FUTEX_TID_MASK
))) {
1473 /* Keep the OWNER_DIED bit */
1474 newval
= (curval
& ~FUTEX_TID_MASK
) | task_pid_vnr(current
);
1479 curval
= cmpxchg_futex_value_locked(uaddr
, uval
, newval
);
1481 if (unlikely(curval
== -EFAULT
))
1483 if (unlikely(curval
!= uval
))
1487 * We took the lock due to owner died take over.
1489 if (unlikely(lock_taken
))
1490 goto out_unlock_release_sem
;
1493 * We dont have the lock. Look up the PI state (or create it if
1494 * we are the first waiter):
1496 ret
= lookup_pi_state(uval
, hb
, &q
.key
, &q
.pi_state
);
1498 if (unlikely(ret
)) {
1503 * Task is exiting and we just wait for the
1506 queue_unlock(&q
, hb
);
1507 futex_unlock_mm(fshared
);
1513 * No owner found for this futex. Check if the
1514 * OWNER_DIED bit is set to figure out whether
1515 * this is a robust futex or not.
1517 if (get_futex_value_locked(&curval
, uaddr
))
1521 * We simply start over in case of a robust
1522 * futex. The code above will take the futex
1525 if (curval
& FUTEX_OWNER_DIED
) {
1530 goto out_unlock_release_sem
;
1535 * Only actually queue now that the atomic ops are done:
1540 * Now the futex is queued and we have checked the data, we
1541 * don't want to hold mmap_sem while we sleep.
1543 futex_unlock_mm(fshared
);
1545 WARN_ON(!q
.pi_state
);
1547 * Block on the PI mutex:
1550 ret
= rt_mutex_timed_lock(&q
.pi_state
->pi_mutex
, to
, 1);
1552 ret
= rt_mutex_trylock(&q
.pi_state
->pi_mutex
);
1553 /* Fixup the trylock return value: */
1554 ret
= ret
? 0 : -EWOULDBLOCK
;
1557 futex_lock_mm(fshared
);
1558 spin_lock(q
.lock_ptr
);
1562 * Got the lock. We might not be the anticipated owner
1563 * if we did a lock-steal - fix up the PI-state in
1566 if (q
.pi_state
->owner
!= curr
)
1567 ret
= fixup_pi_state_owner(uaddr
, &q
, curr
, fshared
);
1570 * Catch the rare case, where the lock was released
1571 * when we were on the way back before we locked the
1574 if (q
.pi_state
->owner
== curr
) {
1576 * Try to get the rt_mutex now. This might
1577 * fail as some other task acquired the
1578 * rt_mutex after we removed ourself from the
1579 * rt_mutex waiters list.
1581 if (rt_mutex_trylock(&q
.pi_state
->pi_mutex
))
1585 * pi_state is incorrect, some other
1586 * task did a lock steal and we
1587 * returned due to timeout or signal
1588 * without taking the rt_mutex. Too
1589 * late. We can access the
1590 * rt_mutex_owner without locking, as
1591 * the other task is now blocked on
1592 * the hash bucket lock. Fix the state
1595 struct task_struct
*owner
;
1598 owner
= rt_mutex_owner(&q
.pi_state
->pi_mutex
);
1599 res
= fixup_pi_state_owner(uaddr
, &q
, owner
,
1602 /* propagate -EFAULT, if the fixup failed */
1608 * Paranoia check. If we did not take the lock
1609 * in the trylock above, then we should not be
1610 * the owner of the rtmutex, neither the real
1611 * nor the pending one:
1613 if (rt_mutex_owner(&q
.pi_state
->pi_mutex
) == curr
)
1614 printk(KERN_ERR
"futex_lock_pi: ret = %d "
1615 "pi-mutex: %p pi-state %p\n", ret
,
1616 q
.pi_state
->pi_mutex
.owner
,
1621 /* Unqueue and drop the lock */
1623 futex_unlock_mm(fshared
);
1626 destroy_hrtimer_on_stack(&to
->timer
);
1627 return ret
!= -EINTR
? ret
: -ERESTARTNOINTR
;
1629 out_unlock_release_sem
:
1630 queue_unlock(&q
, hb
);
1633 futex_unlock_mm(fshared
);
1635 destroy_hrtimer_on_stack(&to
->timer
);
1640 * We have to r/w *(int __user *)uaddr, but we can't modify it
1641 * non-atomically. Therefore, if get_user below is not
1642 * enough, we need to handle the fault ourselves, while
1643 * still holding the mmap_sem.
1645 * ... and hb->lock. :-) --ANK
1647 queue_unlock(&q
, hb
);
1650 ret
= futex_handle_fault((unsigned long)uaddr
, fshared
,
1653 goto out_release_sem
;
1654 goto retry_unlocked
;
1657 futex_unlock_mm(fshared
);
1659 ret
= get_user(uval
, uaddr
);
1660 if (!ret
&& (uval
!= -EFAULT
))
1664 destroy_hrtimer_on_stack(&to
->timer
);
1669 * Userspace attempted a TID -> 0 atomic transition, and failed.
1670 * This is the in-kernel slowpath: we look up the PI state (if any),
1671 * and do the rt-mutex unlock.
1673 static int futex_unlock_pi(u32 __user
*uaddr
, struct rw_semaphore
*fshared
)
1675 struct futex_hash_bucket
*hb
;
1676 struct futex_q
*this, *next
;
1678 struct plist_head
*head
;
1679 union futex_key key
;
1680 int ret
, attempt
= 0;
1683 if (get_user(uval
, uaddr
))
1686 * We release only a lock we actually own:
1688 if ((uval
& FUTEX_TID_MASK
) != task_pid_vnr(current
))
1691 * First take all the futex related locks:
1693 futex_lock_mm(fshared
);
1695 ret
= get_futex_key(uaddr
, fshared
, &key
);
1696 if (unlikely(ret
!= 0))
1699 hb
= hash_futex(&key
);
1701 spin_lock(&hb
->lock
);
1704 * To avoid races, try to do the TID -> 0 atomic transition
1705 * again. If it succeeds then we can return without waking
1708 if (!(uval
& FUTEX_OWNER_DIED
))
1709 uval
= cmpxchg_futex_value_locked(uaddr
, task_pid_vnr(current
), 0);
1712 if (unlikely(uval
== -EFAULT
))
1715 * Rare case: we managed to release the lock atomically,
1716 * no need to wake anyone else up:
1718 if (unlikely(uval
== task_pid_vnr(current
)))
1722 * Ok, other tasks may need to be woken up - check waiters
1723 * and do the wakeup if necessary:
1727 plist_for_each_entry_safe(this, next
, head
, list
) {
1728 if (!match_futex (&this->key
, &key
))
1730 ret
= wake_futex_pi(uaddr
, uval
, this);
1732 * The atomic access to the futex value
1733 * generated a pagefault, so retry the
1734 * user-access and the wakeup:
1741 * No waiters - kernel unlocks the futex:
1743 if (!(uval
& FUTEX_OWNER_DIED
)) {
1744 ret
= unlock_futex_pi(uaddr
, uval
);
1750 spin_unlock(&hb
->lock
);
1752 futex_unlock_mm(fshared
);
1758 * We have to r/w *(int __user *)uaddr, but we can't modify it
1759 * non-atomically. Therefore, if get_user below is not
1760 * enough, we need to handle the fault ourselves, while
1761 * still holding the mmap_sem.
1763 * ... and hb->lock. --ANK
1765 spin_unlock(&hb
->lock
);
1768 ret
= futex_handle_fault((unsigned long)uaddr
, fshared
,
1773 goto retry_unlocked
;
1776 futex_unlock_mm(fshared
);
1778 ret
= get_user(uval
, uaddr
);
1779 if (!ret
&& (uval
!= -EFAULT
))
1786 * Support for robust futexes: the kernel cleans up held futexes at
1789 * Implementation: user-space maintains a per-thread list of locks it
1790 * is holding. Upon do_exit(), the kernel carefully walks this list,
1791 * and marks all locks that are owned by this thread with the
1792 * FUTEX_OWNER_DIED bit, and wakes up a waiter (if any). The list is
1793 * always manipulated with the lock held, so the list is private and
1794 * per-thread. Userspace also maintains a per-thread 'list_op_pending'
1795 * field, to allow the kernel to clean up if the thread dies after
1796 * acquiring the lock, but just before it could have added itself to
1797 * the list. There can only be one such pending lock.
1801 * sys_set_robust_list - set the robust-futex list head of a task
1802 * @head: pointer to the list-head
1803 * @len: length of the list-head, as userspace expects
1806 sys_set_robust_list(struct robust_list_head __user
*head
,
1809 if (!futex_cmpxchg_enabled
)
1812 * The kernel knows only one size for now:
1814 if (unlikely(len
!= sizeof(*head
)))
1817 current
->robust_list
= head
;
1823 * sys_get_robust_list - get the robust-futex list head of a task
1824 * @pid: pid of the process [zero for current task]
1825 * @head_ptr: pointer to a list-head pointer, the kernel fills it in
1826 * @len_ptr: pointer to a length field, the kernel fills in the header size
1829 sys_get_robust_list(int pid
, struct robust_list_head __user
* __user
*head_ptr
,
1830 size_t __user
*len_ptr
)
1832 struct robust_list_head __user
*head
;
1834 uid_t euid
= current_euid();
1836 if (!futex_cmpxchg_enabled
)
1840 head
= current
->robust_list
;
1842 struct task_struct
*p
;
1846 p
= find_task_by_vpid(pid
);
1850 if (euid
!= p
->cred
->euid
&&
1851 euid
!= p
->cred
->uid
&&
1852 !capable(CAP_SYS_PTRACE
))
1854 head
= p
->robust_list
;
1858 if (put_user(sizeof(*head
), len_ptr
))
1860 return put_user(head
, head_ptr
);
1869 * Process a futex-list entry, check whether it's owned by the
1870 * dying task, and do notification if so:
1872 int handle_futex_death(u32 __user
*uaddr
, struct task_struct
*curr
, int pi
)
1874 u32 uval
, nval
, mval
;
1877 if (get_user(uval
, uaddr
))
1880 if ((uval
& FUTEX_TID_MASK
) == task_pid_vnr(curr
)) {
1882 * Ok, this dying thread is truly holding a futex
1883 * of interest. Set the OWNER_DIED bit atomically
1884 * via cmpxchg, and if the value had FUTEX_WAITERS
1885 * set, wake up a waiter (if any). (We have to do a
1886 * futex_wake() even if OWNER_DIED is already set -
1887 * to handle the rare but possible case of recursive
1888 * thread-death.) The rest of the cleanup is done in
1891 mval
= (uval
& FUTEX_WAITERS
) | FUTEX_OWNER_DIED
;
1892 nval
= futex_atomic_cmpxchg_inatomic(uaddr
, uval
, mval
);
1894 if (nval
== -EFAULT
)
1901 * Wake robust non-PI futexes here. The wakeup of
1902 * PI futexes happens in exit_pi_state():
1904 if (!pi
&& (uval
& FUTEX_WAITERS
))
1905 futex_wake(uaddr
, &curr
->mm
->mmap_sem
, 1,
1906 FUTEX_BITSET_MATCH_ANY
);
1912 * Fetch a robust-list pointer. Bit 0 signals PI futexes:
1914 static inline int fetch_robust_entry(struct robust_list __user
**entry
,
1915 struct robust_list __user
* __user
*head
,
1918 unsigned long uentry
;
1920 if (get_user(uentry
, (unsigned long __user
*)head
))
1923 *entry
= (void __user
*)(uentry
& ~1UL);
1930 * Walk curr->robust_list (very carefully, it's a userspace list!)
1931 * and mark any locks found there dead, and notify any waiters.
1933 * We silently return on any sign of list-walking problem.
1935 void exit_robust_list(struct task_struct
*curr
)
1937 struct robust_list_head __user
*head
= curr
->robust_list
;
1938 struct robust_list __user
*entry
, *next_entry
, *pending
;
1939 unsigned int limit
= ROBUST_LIST_LIMIT
, pi
, next_pi
, pip
;
1940 unsigned long futex_offset
;
1943 if (!futex_cmpxchg_enabled
)
1947 * Fetch the list head (which was registered earlier, via
1948 * sys_set_robust_list()):
1950 if (fetch_robust_entry(&entry
, &head
->list
.next
, &pi
))
1953 * Fetch the relative futex offset:
1955 if (get_user(futex_offset
, &head
->futex_offset
))
1958 * Fetch any possibly pending lock-add first, and handle it
1961 if (fetch_robust_entry(&pending
, &head
->list_op_pending
, &pip
))
1964 next_entry
= NULL
; /* avoid warning with gcc */
1965 while (entry
!= &head
->list
) {
1967 * Fetch the next entry in the list before calling
1968 * handle_futex_death:
1970 rc
= fetch_robust_entry(&next_entry
, &entry
->next
, &next_pi
);
1972 * A pending lock might already be on the list, so
1973 * don't process it twice:
1975 if (entry
!= pending
)
1976 if (handle_futex_death((void __user
*)entry
+ futex_offset
,
1984 * Avoid excessively long or circular lists:
1993 handle_futex_death((void __user
*)pending
+ futex_offset
,
1997 long do_futex(u32 __user
*uaddr
, int op
, u32 val
, ktime_t
*timeout
,
1998 u32 __user
*uaddr2
, u32 val2
, u32 val3
)
2001 int cmd
= op
& FUTEX_CMD_MASK
;
2002 struct rw_semaphore
*fshared
= NULL
;
2004 if (!(op
& FUTEX_PRIVATE_FLAG
))
2005 fshared
= ¤t
->mm
->mmap_sem
;
2009 val3
= FUTEX_BITSET_MATCH_ANY
;
2010 case FUTEX_WAIT_BITSET
:
2011 ret
= futex_wait(uaddr
, fshared
, val
, timeout
, val3
);
2014 val3
= FUTEX_BITSET_MATCH_ANY
;
2015 case FUTEX_WAKE_BITSET
:
2016 ret
= futex_wake(uaddr
, fshared
, val
, val3
);
2019 ret
= futex_requeue(uaddr
, fshared
, uaddr2
, val
, val2
, NULL
);
2021 case FUTEX_CMP_REQUEUE
:
2022 ret
= futex_requeue(uaddr
, fshared
, uaddr2
, val
, val2
, &val3
);
2025 ret
= futex_wake_op(uaddr
, fshared
, uaddr2
, val
, val2
, val3
);
2028 if (futex_cmpxchg_enabled
)
2029 ret
= futex_lock_pi(uaddr
, fshared
, val
, timeout
, 0);
2031 case FUTEX_UNLOCK_PI
:
2032 if (futex_cmpxchg_enabled
)
2033 ret
= futex_unlock_pi(uaddr
, fshared
);
2035 case FUTEX_TRYLOCK_PI
:
2036 if (futex_cmpxchg_enabled
)
2037 ret
= futex_lock_pi(uaddr
, fshared
, 0, timeout
, 1);
2046 asmlinkage
long sys_futex(u32 __user
*uaddr
, int op
, u32 val
,
2047 struct timespec __user
*utime
, u32 __user
*uaddr2
,
2051 ktime_t t
, *tp
= NULL
;
2053 int cmd
= op
& FUTEX_CMD_MASK
;
2055 if (utime
&& (cmd
== FUTEX_WAIT
|| cmd
== FUTEX_LOCK_PI
||
2056 cmd
== FUTEX_WAIT_BITSET
)) {
2057 if (copy_from_user(&ts
, utime
, sizeof(ts
)) != 0)
2059 if (!timespec_valid(&ts
))
2062 t
= timespec_to_ktime(ts
);
2063 if (cmd
== FUTEX_WAIT
)
2064 t
= ktime_add_safe(ktime_get(), t
);
2068 * requeue parameter in 'utime' if cmd == FUTEX_REQUEUE.
2069 * number of waiters to wake in 'utime' if cmd == FUTEX_WAKE_OP.
2071 if (cmd
== FUTEX_REQUEUE
|| cmd
== FUTEX_CMP_REQUEUE
||
2072 cmd
== FUTEX_WAKE_OP
)
2073 val2
= (u32
) (unsigned long) utime
;
2075 return do_futex(uaddr
, op
, val
, tp
, uaddr2
, val2
, val3
);
2078 static int __init
futex_init(void)
2084 * This will fail and we want it. Some arch implementations do
2085 * runtime detection of the futex_atomic_cmpxchg_inatomic()
2086 * functionality. We want to know that before we call in any
2087 * of the complex code paths. Also we want to prevent
2088 * registration of robust lists in that case. NULL is
2089 * guaranteed to fault and we get -EFAULT on functional
2090 * implementation, the non functional ones will return
2093 curval
= cmpxchg_futex_value_locked(NULL
, 0, 0);
2094 if (curval
== -EFAULT
)
2095 futex_cmpxchg_enabled
= 1;
2097 for (i
= 0; i
< ARRAY_SIZE(futex_queues
); i
++) {
2098 plist_head_init(&futex_queues
[i
].chain
, &futex_queues
[i
].lock
);
2099 spin_lock_init(&futex_queues
[i
].lock
);
2104 __initcall(futex_init
);