Merge branch 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
[safe/jmp/linux-2.6] / kernel / futex.c
index 8af1002..f89d373 100644 (file)
@@ -92,11 +92,12 @@ struct futex_pi_state {
  * A futex_q has a woken state, just like tasks have TASK_RUNNING.
  * It is considered woken when plist_node_empty(&q->list) || q->lock_ptr == 0.
  * The order of wakup is always to make the first condition true, then
- * wake up q->waiters, then make the second condition true.
+ * wake up q->waiter, then make the second condition true.
  */
 struct futex_q {
        struct plist_node list;
-       wait_queue_head_t waiters;
+       /* There can only be a single waiter */
+       wait_queue_head_t waiter;
 
        /* Which hash list lock to use: */
        spinlock_t *lock_ptr;
@@ -123,24 +124,6 @@ struct futex_hash_bucket {
 static struct futex_hash_bucket futex_queues[1<<FUTEX_HASHBITS];
 
 /*
- * Take mm->mmap_sem, when futex is shared
- */
-static inline void futex_lock_mm(struct rw_semaphore *fshared)
-{
-       if (fshared)
-               down_read(fshared);
-}
-
-/*
- * Release mm->mmap_sem, when the futex is shared
- */
-static inline void futex_unlock_mm(struct rw_semaphore *fshared)
-{
-       if (fshared)
-               up_read(fshared);
-}
-
-/*
  * We hash on the keys returned from get_futex_key (see below).
  */
 static struct futex_hash_bucket *hash_futex(union futex_key *key)
@@ -161,6 +144,48 @@ static inline int match_futex(union futex_key *key1, union futex_key *key2)
                && key1->both.offset == key2->both.offset);
 }
 
+/*
+ * Take a reference to the resource addressed by a key.
+ * Can be called while holding spinlocks.
+ *
+ */
+static void get_futex_key_refs(union futex_key *key)
+{
+       if (!key->both.ptr)
+               return;
+
+       switch (key->both.offset & (FUT_OFF_INODE|FUT_OFF_MMSHARED)) {
+       case FUT_OFF_INODE:
+               atomic_inc(&key->shared.inode->i_count);
+               break;
+       case FUT_OFF_MMSHARED:
+               atomic_inc(&key->private.mm->mm_count);
+               break;
+       }
+}
+
+/*
+ * Drop a reference to the resource addressed by a key.
+ * The hash bucket spinlock must not be held.
+ */
+static void drop_futex_key_refs(union futex_key *key)
+{
+       if (!key->both.ptr) {
+               /* If we're here then we tried to put a key we failed to get */
+               WARN_ON_ONCE(1);
+               return;
+       }
+
+       switch (key->both.offset & (FUT_OFF_INODE|FUT_OFF_MMSHARED)) {
+       case FUT_OFF_INODE:
+               iput(key->shared.inode);
+               break;
+       case FUT_OFF_MMSHARED:
+               mmdrop(key->private.mm);
+               break;
+       }
+}
+
 /**
  * get_futex_key - Get parameters which are the keys for a futex.
  * @uaddr: virtual address of the futex
@@ -179,12 +204,10 @@ static inline int match_futex(union futex_key *key1, union futex_key *key2)
  * For other futexes, it points to &current->mm->mmap_sem and
  * caller must have taken the reader lock. but NOT any spinlocks.
  */
-static int get_futex_key(u32 __user *uaddr, struct rw_semaphore *fshared,
-                        union futex_key *key)
+static int get_futex_key(u32 __user *uaddr, int fshared, union futex_key *key)
 {
        unsigned long address = (unsigned long)uaddr;
        struct mm_struct *mm = current->mm;
-       struct vm_area_struct *vma;
        struct page *page;
        int err;
 
@@ -208,100 +231,50 @@ static int get_futex_key(u32 __user *uaddr, struct rw_semaphore *fshared,
                        return -EFAULT;
                key->private.mm = mm;
                key->private.address = address;
+               get_futex_key_refs(key);
                return 0;
        }
-       /*
-        * The futex is hashed differently depending on whether
-        * it's in a shared or private mapping.  So check vma first.
-        */
-       vma = find_extend_vma(mm, address);
-       if (unlikely(!vma))
-               return -EFAULT;
 
-       /*
-        * Permissions.
-        */
-       if (unlikely((vma->vm_flags & (VM_IO|VM_READ)) != VM_READ))
-               return (vma->vm_flags & VM_IO) ? -EPERM : -EACCES;
+again:
+       err = get_user_pages_fast(address, 1, 0, &page);
+       if (err < 0)
+               return err;
+
+       lock_page(page);
+       if (!page->mapping) {
+               unlock_page(page);
+               put_page(page);
+               goto again;
+       }
 
        /*
         * Private mappings are handled in a simple way.
         *
         * NOTE: When userspace waits on a MAP_SHARED mapping, even if
         * it's a read-only handle, it's expected that futexes attach to
-        * the object not the particular process.  Therefore we use
-        * VM_MAYSHARE here, not VM_SHARED which is restricted to shared
-        * mappings of _writable_ handles.
+        * the object not the particular process.
         */
-       if (likely(!(vma->vm_flags & VM_MAYSHARE))) {
-               key->both.offset |= FUT_OFF_MMSHARED; /* reference taken on mm */
+       if (PageAnon(page)) {
+               key->both.offset |= FUT_OFF_MMSHARED; /* ref taken on mm */
                key->private.mm = mm;
                key->private.address = address;
-               return 0;
-       }
-
-       /*
-        * Linear file mappings are also simple.
-        */
-       key->shared.inode = vma->vm_file->f_path.dentry->d_inode;
-       key->both.offset |= FUT_OFF_INODE; /* inode-based key. */
-       if (likely(!(vma->vm_flags & VM_NONLINEAR))) {
-               key->shared.pgoff = (((address - vma->vm_start) >> PAGE_SHIFT)
-                                    + vma->vm_pgoff);
-               return 0;
+       } else {
+               key->both.offset |= FUT_OFF_INODE; /* inode-based key */
+               key->shared.inode = page->mapping->host;
+               key->shared.pgoff = page->index;
        }
 
-       /*
-        * We could walk the page table to read the non-linear
-        * pte, and get the page index without fetching the page
-        * from swap.  But that's a lot of code to duplicate here
-        * for a rare case, so we simply fetch the page.
-        */
-       err = get_user_pages(current, mm, address, 1, 0, 0, &page, NULL);
-       if (err >= 0) {
-               key->shared.pgoff =
-                       page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
-               put_page(page);
-               return 0;
-       }
-       return err;
-}
+       get_futex_key_refs(key);
 
-/*
- * Take a reference to the resource addressed by a key.
- * Can be called while holding spinlocks.
- *
- */
-static void get_futex_key_refs(union futex_key *key)
-{
-       if (key->both.ptr == NULL)
-               return;
-       switch (key->both.offset & (FUT_OFF_INODE|FUT_OFF_MMSHARED)) {
-               case FUT_OFF_INODE:
-                       atomic_inc(&key->shared.inode->i_count);
-                       break;
-               case FUT_OFF_MMSHARED:
-                       atomic_inc(&key->private.mm->mm_count);
-                       break;
-       }
+       unlock_page(page);
+       put_page(page);
+       return 0;
 }
 
-/*
- * Drop a reference to the resource addressed by a key.
- * The hash bucket spinlock must not be held.
- */
-static void drop_futex_key_refs(union futex_key *key)
+static inline
+void put_futex_key(int fshared, union futex_key *key)
 {
-       if (!key->both.ptr)
-               return;
-       switch (key->both.offset & (FUT_OFF_INODE|FUT_OFF_MMSHARED)) {
-               case FUT_OFF_INODE:
-                       iput(key->shared.inode);
-                       break;
-               case FUT_OFF_MMSHARED:
-                       mmdrop(key->private.mm);
-                       break;
-       }
+       drop_futex_key_refs(key);
 }
 
 static u32 cmpxchg_futex_value_locked(u32 __user *uaddr, u32 uval, u32 newval)
@@ -328,10 +301,8 @@ static int get_futex_value_locked(u32 *dest, u32 __user *from)
 
 /*
  * Fault handling.
- * if fshared is non NULL, current->mm->mmap_sem is already held
  */
-static int futex_handle_fault(unsigned long address,
-                             struct rw_semaphore *fshared, int attempt)
+static int futex_handle_fault(unsigned long address, int attempt)
 {
        struct vm_area_struct * vma;
        struct mm_struct *mm = current->mm;
@@ -340,8 +311,7 @@ static int futex_handle_fault(unsigned long address,
        if (attempt > 2)
                return ret;
 
-       if (!fshared)
-               down_read(&mm->mmap_sem);
+       down_read(&mm->mmap_sem);
        vma = find_vma(mm, address);
        if (vma && address >= vma->vm_start &&
            (vma->vm_flags & VM_WRITE)) {
@@ -361,8 +331,7 @@ static int futex_handle_fault(unsigned long address,
                                current->min_flt++;
                }
        }
-       if (!fshared)
-               up_read(&mm->mmap_sem);
+       up_read(&mm->mmap_sem);
        return ret;
 }
 
@@ -385,6 +354,7 @@ static int refill_pi_state_cache(void)
        /* pi_mutex gets initialized later */
        pi_state->owner = NULL;
        atomic_set(&pi_state->refcount, 1);
+       pi_state->key = FUTEX_KEY_INIT;
 
        current->pi_state_cache = pi_state;
 
@@ -439,13 +409,20 @@ static void free_pi_state(struct futex_pi_state *pi_state)
 static struct task_struct * futex_find_get_task(pid_t pid)
 {
        struct task_struct *p;
+       const struct cred *cred = current_cred(), *pcred;
 
        rcu_read_lock();
        p = find_task_by_vpid(pid);
-       if (!p || ((current->euid != p->euid) && (current->euid != p->uid)))
+       if (!p) {
                p = ERR_PTR(-ESRCH);
-       else
-               get_task_struct(p);
+       } else {
+               pcred = __task_cred(p);
+               if (cred->euid != pcred->euid &&
+                   cred->euid != pcred->uid)
+                       p = ERR_PTR(-ESRCH);
+               else
+                       get_task_struct(p);
+       }
 
        rcu_read_unlock();
 
@@ -462,7 +439,7 @@ void exit_pi_state_list(struct task_struct *curr)
        struct list_head *next, *head = &curr->pi_state_list;
        struct futex_pi_state *pi_state;
        struct futex_hash_bucket *hb;
-       union futex_key key;
+       union futex_key key = FUTEX_KEY_INIT;
 
        if (!futex_cmpxchg_enabled)
                return;
@@ -607,7 +584,7 @@ static void wake_futex(struct futex_q *q)
         * The lock in wake_up_all() is a crucial memory barrier after the
         * plist_del() and also before assigning to q->lock_ptr.
         */
-       wake_up_all(&q->waiters);
+       wake_up(&q->waiter);
        /*
         * The waiting task can free the futex_q as soon as this is written,
         * without taking any locks.  This must come last.
@@ -719,20 +696,17 @@ double_lock_hb(struct futex_hash_bucket *hb1, struct futex_hash_bucket *hb2)
  * Wake up all waiters hashed on the physical page that is mapped
  * to this virtual address:
  */
-static int futex_wake(u32 __user *uaddr, struct rw_semaphore *fshared,
-                     int nr_wake, u32 bitset)
+static int futex_wake(u32 __user *uaddr, int fshared, int nr_wake, u32 bitset)
 {
        struct futex_hash_bucket *hb;
        struct futex_q *this, *next;
        struct plist_head *head;
-       union futex_key key;
+       union futex_key key = FUTEX_KEY_INIT;
        int ret;
 
        if (!bitset)
                return -EINVAL;
 
-       futex_lock_mm(fshared);
-
        ret = get_futex_key(uaddr, fshared, &key);
        if (unlikely(ret != 0))
                goto out;
@@ -759,8 +733,8 @@ static int futex_wake(u32 __user *uaddr, struct rw_semaphore *fshared,
        }
 
        spin_unlock(&hb->lock);
+       put_futex_key(fshared, &key);
 out:
-       futex_unlock_mm(fshared);
        return ret;
 }
 
@@ -769,25 +743,22 @@ out:
  * to this virtual address:
  */
 static int
-futex_wake_op(u32 __user *uaddr1, struct rw_semaphore *fshared,
-             u32 __user *uaddr2,
+futex_wake_op(u32 __user *uaddr1, int fshared, u32 __user *uaddr2,
              int nr_wake, int nr_wake2, int op)
 {
-       union futex_key key1, key2;
+       union futex_key key1 = FUTEX_KEY_INIT, key2 = FUTEX_KEY_INIT;
        struct futex_hash_bucket *hb1, *hb2;
        struct plist_head *head;
        struct futex_q *this, *next;
        int ret, op_ret, attempt = 0;
 
 retryfull:
-       futex_lock_mm(fshared);
-
        ret = get_futex_key(uaddr1, fshared, &key1);
        if (unlikely(ret != 0))
                goto out;
        ret = get_futex_key(uaddr2, fshared, &key2);
        if (unlikely(ret != 0))
-               goto out;
+               goto out_put_key1;
 
        hb1 = hash_futex(&key1);
        hb2 = hash_futex(&key2);
@@ -809,12 +780,12 @@ retry:
                 * but we might get them from range checking
                 */
                ret = op_ret;
-               goto out;
+               goto out_put_keys;
 #endif
 
                if (unlikely(op_ret != -EFAULT)) {
                        ret = op_ret;
-                       goto out;
+                       goto out_put_keys;
                }
 
                /*
@@ -826,18 +797,12 @@ retry:
                 */
                if (attempt++) {
                        ret = futex_handle_fault((unsigned long)uaddr2,
-                                                fshared, attempt);
+                                                attempt);
                        if (ret)
-                               goto out;
+                               goto out_put_keys;
                        goto retry;
                }
 
-               /*
-                * If we would have faulted, release mmap_sem,
-                * fault it in and start all over again.
-                */
-               futex_unlock_mm(fshared);
-
                ret = get_user(dummy, uaddr2);
                if (ret)
                        return ret;
@@ -872,9 +837,11 @@ retry:
        spin_unlock(&hb1->lock);
        if (hb1 != hb2)
                spin_unlock(&hb2->lock);
+out_put_keys:
+       put_futex_key(fshared, &key2);
+out_put_key1:
+       put_futex_key(fshared, &key1);
 out:
-       futex_unlock_mm(fshared);
-
        return ret;
 }
 
@@ -882,25 +849,22 @@ out:
  * Requeue all waiters hashed on one physical page to another
  * physical page.
  */
-static int futex_requeue(u32 __user *uaddr1, struct rw_semaphore *fshared,
-                        u32 __user *uaddr2,
+static int futex_requeue(u32 __user *uaddr1, int fshared, u32 __user *uaddr2,
                         int nr_wake, int nr_requeue, u32 *cmpval)
 {
-       union futex_key key1, key2;
+       union futex_key key1 = FUTEX_KEY_INIT, key2 = FUTEX_KEY_INIT;
        struct futex_hash_bucket *hb1, *hb2;
        struct plist_head *head1;
        struct futex_q *this, *next;
        int ret, drop_count = 0;
 
- retry:
-       futex_lock_mm(fshared);
-
+retry:
        ret = get_futex_key(uaddr1, fshared, &key1);
        if (unlikely(ret != 0))
                goto out;
        ret = get_futex_key(uaddr2, fshared, &key2);
        if (unlikely(ret != 0))
-               goto out;
+               goto out_put_key1;
 
        hb1 = hash_futex(&key1);
        hb2 = hash_futex(&key2);
@@ -917,18 +881,12 @@ static int futex_requeue(u32 __user *uaddr1, struct rw_semaphore *fshared,
                        if (hb1 != hb2)
                                spin_unlock(&hb2->lock);
 
-                       /*
-                        * If we would have faulted, release mmap_sem, fault
-                        * it in and start all over again.
-                        */
-                       futex_unlock_mm(fshared);
-
                        ret = get_user(curval, uaddr1);
 
                        if (!ret)
                                goto retry;
 
-                       return ret;
+                       goto out_put_keys;
                }
                if (curval != *cmpval) {
                        ret = -EAGAIN;
@@ -973,8 +931,11 @@ out_unlock:
        while (--drop_count >= 0)
                drop_futex_key_refs(&key1);
 
+out_put_keys:
+       put_futex_key(fshared, &key2);
+out_put_key1:
+       put_futex_key(fshared, &key1);
 out:
-       futex_unlock_mm(fshared);
        return ret;
 }
 
@@ -983,7 +944,7 @@ static inline struct futex_hash_bucket *queue_lock(struct futex_q *q)
 {
        struct futex_hash_bucket *hb;
 
-       init_waitqueue_head(&q->waiters);
+       init_waitqueue_head(&q->waiter);
 
        get_futex_key_refs(&q->key);
        hb = hash_futex(&q->key);
@@ -1035,7 +996,7 @@ static int unqueue_me(struct futex_q *q)
        int ret = 0;
 
        /* In the common case we don't take the spinlock, which is nice. */
- retry:
+retry:
        lock_ptr = q->lock_ptr;
        barrier();
        if (lock_ptr != NULL) {
@@ -1096,8 +1057,7 @@ static void unqueue_me_pi(struct futex_q *q)
  * private futexes.
  */
 static int fixup_pi_state_owner(u32 __user *uaddr, struct futex_q *q,
-                               struct task_struct *newowner,
-                               struct rw_semaphore *fshared)
+                               struct task_struct *newowner, int fshared)
 {
        u32 newtid = task_pid_vnr(newowner) | FUTEX_WAITERS;
        struct futex_pi_state *pi_state = q->pi_state;
@@ -1176,7 +1136,7 @@ retry:
 handle_fault:
        spin_unlock(q->lock_ptr);
 
-       ret = futex_handle_fault((unsigned long)uaddr, fshared, attempt++);
+       ret = futex_handle_fault((unsigned long)uaddr, attempt++);
 
        spin_lock(q->lock_ptr);
 
@@ -1196,12 +1156,13 @@ handle_fault:
  * In case we must use restart_block to restart a futex_wait,
  * we encode in the 'flags' shared capability
  */
-#define FLAGS_SHARED  1
+#define FLAGS_SHARED           0x01
+#define FLAGS_CLOCKRT          0x02
 
 static long futex_wait_restart(struct restart_block *restart);
 
-static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared,
-                     u32 val, ktime_t *abs_time, u32 bitset)
+static int futex_wait(u32 __user *uaddr, int fshared,
+                     u32 val, ktime_t *abs_time, u32 bitset, int clockrt)
 {
        struct task_struct *curr = current;
        DECLARE_WAITQUEUE(wait, curr);
@@ -1217,12 +1178,11 @@ static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared,
 
        q.pi_state = NULL;
        q.bitset = bitset;
- retry:
-       futex_lock_mm(fshared);
-
+retry:
+       q.key = FUTEX_KEY_INIT;
        ret = get_futex_key(uaddr, fshared, &q.key);
        if (unlikely(ret != 0))
-               goto out_release_sem;
+               goto out;
 
        hb = queue_lock(&q);
 
@@ -1250,12 +1210,7 @@ static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared,
 
        if (unlikely(ret)) {
                queue_unlock(&q, hb);
-
-               /*
-                * If we would have faulted, release mmap_sem, fault it in and
-                * start all over again.
-                */
-               futex_unlock_mm(fshared);
+               put_futex_key(fshared, &q.key);
 
                ret = get_user(uval, uaddr);
 
@@ -1265,18 +1220,12 @@ static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared,
        }
        ret = -EWOULDBLOCK;
        if (uval != val)
-               goto out_unlock_release_sem;
+               goto out_unlock_put_key;
 
        /* Only actually queue if *uaddr contained val.  */
        queue_me(&q, hb);
 
        /*
-        * Now the futex is queued and we have checked the data, we
-        * don't want to hold mmap_sem while we sleep.
-        */
-       futex_unlock_mm(fshared);
-
-       /*
         * There might have been scheduling since the queue_me(), as we
         * cannot hold a spinlock across the get_user() in case it
         * faults, and we cannot just set TASK_INTERRUPTIBLE state when
@@ -1287,7 +1236,7 @@ static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared,
 
        /* add_wait_queue is the barrier after __set_current_state. */
        __set_current_state(TASK_INTERRUPTIBLE);
-       add_wait_queue(&q.waiters, &wait);
+       add_wait_queue(&q.waiter, &wait);
        /*
         * !plist_node_empty() is safe here without any lock.
         * q.lock_ptr != 0 is not safe, because of ordering against wakeup.
@@ -1300,8 +1249,10 @@ static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared,
                        slack = current->timer_slack_ns;
                        if (rt_task(current))
                                slack = 0;
-                       hrtimer_init_on_stack(&t.timer, CLOCK_MONOTONIC,
-                                               HRTIMER_MODE_ABS);
+                       hrtimer_init_on_stack(&t.timer,
+                                             clockrt ? CLOCK_REALTIME :
+                                             CLOCK_MONOTONIC,
+                                             HRTIMER_MODE_ABS);
                        hrtimer_init_sleeper(&t, current);
                        hrtimer_set_expires_range_ns(&t.timer, *abs_time, slack);
 
@@ -1356,14 +1307,16 @@ static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared,
 
                if (fshared)
                        restart->futex.flags |= FLAGS_SHARED;
+               if (clockrt)
+                       restart->futex.flags |= FLAGS_CLOCKRT;
                return -ERESTART_RESTARTBLOCK;
        }
 
- out_unlock_release_sem:
+out_unlock_put_key:
        queue_unlock(&q, hb);
+       put_futex_key(fshared, &q.key);
 
- out_release_sem:
-       futex_unlock_mm(fshared);
+out:
        return ret;
 }
 
@@ -1371,15 +1324,16 @@ static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared,
 static long futex_wait_restart(struct restart_block *restart)
 {
        u32 __user *uaddr = (u32 __user *)restart->futex.uaddr;
-       struct rw_semaphore *fshared = NULL;
+       int fshared = 0;
        ktime_t t;
 
        t.tv64 = restart->futex.time;
        restart->fn = do_no_restart_syscall;
        if (restart->futex.flags & FLAGS_SHARED)
-               fshared = &current->mm->mmap_sem;
+               fshared = 1;
        return (long)futex_wait(uaddr, fshared, restart->futex.val, &t,
-                               restart->futex.bitset);
+                               restart->futex.bitset,
+                               restart->futex.flags & FLAGS_CLOCKRT);
 }
 
 
@@ -1389,7 +1343,7 @@ static long futex_wait_restart(struct restart_block *restart)
  * if there are waiters then it will block, it does PI, etc. (Due to
  * races the kernel might see a 0 value of the futex too.)
  */
-static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared,
+static int futex_lock_pi(u32 __user *uaddr, int fshared,
                         int detect, ktime_t *time, int trylock)
 {
        struct hrtimer_sleeper timeout, *to = NULL;
@@ -1411,17 +1365,16 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared,
        }
 
        q.pi_state = NULL;
- retry:
-       futex_lock_mm(fshared);
-
+retry:
+       q.key = FUTEX_KEY_INIT;
        ret = get_futex_key(uaddr, fshared, &q.key);
        if (unlikely(ret != 0))
-               goto out_release_sem;
+               goto out;
 
- retry_unlocked:
+retry_unlocked:
        hb = queue_lock(&q);
 
- retry_locked:
+retry_locked:
        ret = lock_taken = 0;
 
        /*
@@ -1442,14 +1395,14 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared,
         */
        if (unlikely((curval & FUTEX_TID_MASK) == task_pid_vnr(current))) {
                ret = -EDEADLK;
-               goto out_unlock_release_sem;
+               goto out_unlock_put_key;
        }
 
        /*
         * Surprise - we got the lock. Just return to userspace:
         */
        if (unlikely(!curval))
-               goto out_unlock_release_sem;
+               goto out_unlock_put_key;
 
        uval = curval;
 
@@ -1485,7 +1438,7 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared,
         * We took the lock due to owner died take over.
         */
        if (unlikely(lock_taken))
-               goto out_unlock_release_sem;
+               goto out_unlock_put_key;
 
        /*
         * We dont have the lock. Look up the PI state (or create it if
@@ -1502,7 +1455,6 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared,
                         * exit to complete.
                         */
                        queue_unlock(&q, hb);
-                       futex_unlock_mm(fshared);
                        cond_resched();
                        goto retry;
 
@@ -1525,7 +1477,7 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared,
                                goto retry_locked;
                        }
                default:
-                       goto out_unlock_release_sem;
+                       goto out_unlock_put_key;
                }
        }
 
@@ -1534,12 +1486,6 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared,
         */
        queue_me(&q, hb);
 
-       /*
-        * Now the futex is queued and we have checked the data, we
-        * don't want to hold mmap_sem while we sleep.
-        */
-       futex_unlock_mm(fshared);
-
        WARN_ON(!q.pi_state);
        /*
         * Block on the PI mutex:
@@ -1552,7 +1498,6 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared,
                ret = ret ? 0 : -EWOULDBLOCK;
        }
 
-       futex_lock_mm(fshared);
        spin_lock(q.lock_ptr);
 
        if (!ret) {
@@ -1618,44 +1563,40 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared,
 
        /* Unqueue and drop the lock */
        unqueue_me_pi(&q);
-       futex_unlock_mm(fshared);
 
        if (to)
                destroy_hrtimer_on_stack(&to->timer);
        return ret != -EINTR ? ret : -ERESTARTNOINTR;
 
- out_unlock_release_sem:
+out_unlock_put_key:
        queue_unlock(&q, hb);
 
- out_release_sem:
-       futex_unlock_mm(fshared);
+out_put_key:
+       put_futex_key(fshared, &q.key);
+out:
        if (to)
                destroy_hrtimer_on_stack(&to->timer);
        return ret;
 
- uaddr_faulted:
+uaddr_faulted:
        /*
-        * We have to r/w  *(int __user *)uaddr, but we can't modify it
-        * non-atomically.  Therefore, if get_user below is not
-        * enough, we need to handle the fault ourselves, while
-        * still holding the mmap_sem.
-        *
-        * ... and hb->lock. :-) --ANK
+        * We have to r/w  *(int __user *)uaddr, and we have to modify it
+        * atomically.  Therefore, if we continue to fault after get_user()
+        * below, we need to handle the fault ourselves, while still holding
+        * the mmap_sem.  This can occur if the uaddr is under contention as
+        * we have to drop the mmap_sem in order to call get_user().
         */
        queue_unlock(&q, hb);
 
        if (attempt++) {
-               ret = futex_handle_fault((unsigned long)uaddr, fshared,
-                                        attempt);
+               ret = futex_handle_fault((unsigned long)uaddr, attempt);
                if (ret)
-                       goto out_release_sem;
+                       goto out_put_key;
                goto retry_unlocked;
        }
 
-       futex_unlock_mm(fshared);
-
        ret = get_user(uval, uaddr);
-       if (!ret && (uval != -EFAULT))
+       if (!ret)
                goto retry;
 
        if (to)
@@ -1668,13 +1609,13 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared,
  * This is the in-kernel slowpath: we look up the PI state (if any),
  * and do the rt-mutex unlock.
  */
-static int futex_unlock_pi(u32 __user *uaddr, struct rw_semaphore *fshared)
+static int futex_unlock_pi(u32 __user *uaddr, int fshared)
 {
        struct futex_hash_bucket *hb;
        struct futex_q *this, *next;
        u32 uval;
        struct plist_head *head;
-       union futex_key key;
+       union futex_key key = FUTEX_KEY_INIT;
        int ret, attempt = 0;
 
 retry:
@@ -1685,10 +1626,6 @@ retry:
         */
        if ((uval & FUTEX_TID_MASK) != task_pid_vnr(current))
                return -EPERM;
-       /*
-        * First take all the futex related locks:
-        */
-       futex_lock_mm(fshared);
 
        ret = get_futex_key(uaddr, fshared, &key);
        if (unlikely(ret != 0))
@@ -1746,35 +1683,31 @@ retry_unlocked:
 
 out_unlock:
        spin_unlock(&hb->lock);
-out:
-       futex_unlock_mm(fshared);
+       put_futex_key(fshared, &key);
 
+out:
        return ret;
 
 pi_faulted:
        /*
-        * We have to r/w  *(int __user *)uaddr, but we can't modify it
-        * non-atomically.  Therefore, if get_user below is not
-        * enough, we need to handle the fault ourselves, while
-        * still holding the mmap_sem.
-        *
-        * ... and hb->lock. --ANK
+        * We have to r/w  *(int __user *)uaddr, and we have to modify it
+        * atomically.  Therefore, if we continue to fault after get_user()
+        * below, we need to handle the fault ourselves, while still holding
+        * the mmap_sem.  This can occur if the uaddr is under contention as
+        * we have to drop the mmap_sem in order to call get_user().
         */
        spin_unlock(&hb->lock);
 
        if (attempt++) {
-               ret = futex_handle_fault((unsigned long)uaddr, fshared,
-                                        attempt);
+               ret = futex_handle_fault((unsigned long)uaddr, attempt);
                if (ret)
                        goto out;
                uval = 0;
                goto retry_unlocked;
        }
 
-       futex_unlock_mm(fshared);
-
        ret = get_user(uval, uaddr);
-       if (!ret && (uval != -EFAULT))
+       if (!ret)
                goto retry;
 
        return ret;
@@ -1800,9 +1733,8 @@ pi_faulted:
  * @head: pointer to the list-head
  * @len: length of the list-head, as userspace expects
  */
-asmlinkage long
-sys_set_robust_list(struct robust_list_head __user *head,
-                   size_t len)
+SYSCALL_DEFINE2(set_robust_list, struct robust_list_head __user *, head,
+               size_t, len)
 {
        if (!futex_cmpxchg_enabled)
                return -ENOSYS;
@@ -1823,12 +1755,13 @@ sys_set_robust_list(struct robust_list_head __user *head,
  * @head_ptr: pointer to a list-head pointer, the kernel fills it in
  * @len_ptr: pointer to a length field, the kernel fills in the header size
  */
-asmlinkage long
-sys_get_robust_list(int pid, struct robust_list_head __user * __user *head_ptr,
-                   size_t __user *len_ptr)
+SYSCALL_DEFINE3(get_robust_list, int, pid,
+               struct robust_list_head __user * __user *, head_ptr,
+               size_t __user *, len_ptr)
 {
        struct robust_list_head __user *head;
        unsigned long ret;
+       const struct cred *cred = current_cred(), *pcred;
 
        if (!futex_cmpxchg_enabled)
                return -ENOSYS;
@@ -1844,8 +1777,10 @@ sys_get_robust_list(int pid, struct robust_list_head __user * __user *head_ptr,
                if (!p)
                        goto err_unlock;
                ret = -EPERM;
-               if ((current->euid != p->euid) && (current->euid != p->uid) &&
-                               !capable(CAP_SYS_PTRACE))
+               pcred = __task_cred(p);
+               if (cred->euid != pcred->euid &&
+                   cred->euid != pcred->uid &&
+                   !capable(CAP_SYS_PTRACE))
                        goto err_unlock;
                head = p->robust_list;
                rcu_read_unlock();
@@ -1898,8 +1833,7 @@ retry:
                 * PI futexes happens in exit_pi_state():
                 */
                if (!pi && (uval & FUTEX_WAITERS))
-                       futex_wake(uaddr, &curr->mm->mmap_sem, 1,
-                                  FUTEX_BITSET_MATCH_ANY);
+                       futex_wake(uaddr, 1, 1, FUTEX_BITSET_MATCH_ANY);
        }
        return 0;
 }
@@ -1993,18 +1927,22 @@ void exit_robust_list(struct task_struct *curr)
 long do_futex(u32 __user *uaddr, int op, u32 val, ktime_t *timeout,
                u32 __user *uaddr2, u32 val2, u32 val3)
 {
-       int ret = -ENOSYS;
+       int clockrt, ret = -ENOSYS;
        int cmd = op & FUTEX_CMD_MASK;
-       struct rw_semaphore *fshared = NULL;
+       int fshared = 0;
 
        if (!(op & FUTEX_PRIVATE_FLAG))
-               fshared = &current->mm->mmap_sem;
+               fshared = 1;
+
+       clockrt = op & FUTEX_CLOCK_REALTIME;
+       if (clockrt && cmd != FUTEX_WAIT_BITSET)
+               return -ENOSYS;
 
        switch (cmd) {
        case FUTEX_WAIT:
                val3 = FUTEX_BITSET_MATCH_ANY;
        case FUTEX_WAIT_BITSET:
-               ret = futex_wait(uaddr, fshared, val, timeout, val3);
+               ret = futex_wait(uaddr, fshared, val, timeout, val3, clockrt);
                break;
        case FUTEX_WAKE:
                val3 = FUTEX_BITSET_MATCH_ANY;
@@ -2039,9 +1977,9 @@ long do_futex(u32 __user *uaddr, int op, u32 val, ktime_t *timeout,
 }
 
 
-asmlinkage long sys_futex(u32 __user *uaddr, int op, u32 val,
-                         struct timespec __user *utime, u32 __user *uaddr2,
-                         u32 val3)
+SYSCALL_DEFINE6(futex, u32 __user *, uaddr, int, op, u32, val,
+               struct timespec __user *, utime, u32 __user *, uaddr2,
+               u32, val3)
 {
        struct timespec ts;
        ktime_t t, *tp = NULL;