From 38d47c1b7075bd7ec3881141bb3629da58f88dab Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 26 Sep 2008 19:32:20 +0200 Subject: futex: rely on get_user_pages() for shared futexes On the way of getting rid of the mmap_sem requirement for shared futexes, start by relying on get_user_pages(). Signed-off-by: Peter Zijlstra Acked-by: Nick Piggin Signed-off-by: Ingo Molnar --- kernel/futex.c | 162 ++++++++++++++++++++++++++++----------------------------- 1 file changed, 80 insertions(+), 82 deletions(-) (limited to 'kernel/futex.c') diff --git a/kernel/futex.c b/kernel/futex.c index 7d1136e97c14..a4c39fa0a7a3 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -161,6 +161,45 @@ 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) + 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 @@ -184,7 +223,6 @@ static int get_futex_key(u32 __user *uaddr, struct rw_semaphore *fshared, { unsigned long address = (unsigned long)uaddr; struct mm_struct *mm = current->mm; - struct vm_area_struct *vma; struct page *page; int err; @@ -210,98 +248,47 @@ static int get_futex_key(u32 __user *uaddr, struct rw_semaphore *fshared, key->private.address = address; 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(current, mm, address, 1, 0, 0, &page, NULL); + 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(struct rw_semaphore *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) @@ -385,6 +372,7 @@ 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; @@ -462,7 +450,7 @@ 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; @@ -725,7 +713,7 @@ static int futex_wake(u32 __user *uaddr, struct rw_semaphore *fshared, 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) @@ -760,6 +748,7 @@ static int futex_wake(u32 __user *uaddr, struct rw_semaphore *fshared, spin_unlock(&hb->lock); out: + put_futex_key(fshared, &key); futex_unlock_mm(fshared); return ret; } @@ -773,7 +762,7 @@ futex_wake_op(u32 __user *uaddr1, struct rw_semaphore *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; @@ -873,6 +862,8 @@ retry: if (hb1 != hb2) spin_unlock(&hb2->lock); out: + put_futex_key(fshared, &key2); + put_futex_key(fshared, &key1); futex_unlock_mm(fshared); return ret; @@ -886,7 +877,7 @@ static int futex_requeue(u32 __user *uaddr1, struct rw_semaphore *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; @@ -974,6 +965,8 @@ out_unlock: drop_futex_key_refs(&key1); out: + put_futex_key(fshared, &key2); + put_futex_key(fshared, &key1); futex_unlock_mm(fshared); return ret; } @@ -1220,6 +1213,7 @@ static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared, retry: futex_lock_mm(fshared); + q.key = FUTEX_KEY_INIT; ret = get_futex_key(uaddr, fshared, &q.key); if (unlikely(ret != 0)) goto out_release_sem; @@ -1360,6 +1354,7 @@ static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared, queue_unlock(&q, hb); out_release_sem: + put_futex_key(fshared, &q.key); futex_unlock_mm(fshared); return ret; } @@ -1411,6 +1406,7 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared, retry: futex_lock_mm(fshared); + q.key = FUTEX_KEY_INIT; ret = get_futex_key(uaddr, fshared, &q.key); if (unlikely(ret != 0)) goto out_release_sem; @@ -1625,6 +1621,7 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared, queue_unlock(&q, hb); out_release_sem: + put_futex_key(fshared, &q.key); futex_unlock_mm(fshared); if (to) destroy_hrtimer_on_stack(&to->timer); @@ -1671,7 +1668,7 @@ static int futex_unlock_pi(u32 __user *uaddr, struct rw_semaphore *fshared) 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: @@ -1744,6 +1741,7 @@ retry_unlocked: out_unlock: spin_unlock(&hb->lock); out: + put_futex_key(fshared, &key); futex_unlock_mm(fshared); return ret; -- cgit v1.2.3 From 61270708ecf1cda148e84fbf6e0703ee5aa81814 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 26 Sep 2008 19:32:21 +0200 Subject: futex: reduce mmap_sem usage now that we rely on get_user_pages() for the shared key handling move all the mmap_sem stuff closely around the slow paths. Signed-off-by: Peter Zijlstra Acked-by: Nick Piggin Signed-off-by: Ingo Molnar --- kernel/futex.c | 83 +++------------------------------------------------------- 1 file changed, 4 insertions(+), 79 deletions(-) (limited to 'kernel/futex.c') diff --git a/kernel/futex.c b/kernel/futex.c index a4c39fa0a7a3..6a726684217e 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -122,24 +122,6 @@ struct futex_hash_bucket { static struct futex_hash_bucket futex_queues[1<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). */ @@ -250,7 +232,9 @@ static int get_futex_key(u32 __user *uaddr, struct rw_semaphore *fshared, } again: + down_read(&mm->mmap_sem); err = get_user_pages(current, mm, address, 1, 0, 0, &page, NULL); + up_read(&mm->mmap_sem); if (err < 0) return err; @@ -327,8 +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)) { @@ -348,8 +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; } @@ -719,8 +701,6 @@ static int futex_wake(u32 __user *uaddr, struct rw_semaphore *fshared, if (!bitset) return -EINVAL; - futex_lock_mm(fshared); - ret = get_futex_key(uaddr, fshared, &key); if (unlikely(ret != 0)) goto out; @@ -749,7 +729,6 @@ static int futex_wake(u32 __user *uaddr, struct rw_semaphore *fshared, spin_unlock(&hb->lock); out: put_futex_key(fshared, &key); - futex_unlock_mm(fshared); return ret; } @@ -769,8 +748,6 @@ futex_wake_op(u32 __user *uaddr1, struct rw_semaphore *fshared, int ret, op_ret, attempt = 0; retryfull: - futex_lock_mm(fshared); - ret = get_futex_key(uaddr1, fshared, &key1); if (unlikely(ret != 0)) goto out; @@ -821,12 +798,6 @@ retry: 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; @@ -864,7 +835,6 @@ retry: out: put_futex_key(fshared, &key2); put_futex_key(fshared, &key1); - futex_unlock_mm(fshared); return ret; } @@ -884,8 +854,6 @@ static int futex_requeue(u32 __user *uaddr1, struct rw_semaphore *fshared, int ret, drop_count = 0; retry: - futex_lock_mm(fshared); - ret = get_futex_key(uaddr1, fshared, &key1); if (unlikely(ret != 0)) goto out; @@ -908,12 +876,6 @@ 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) @@ -967,7 +929,6 @@ out_unlock: out: put_futex_key(fshared, &key2); put_futex_key(fshared, &key1); - futex_unlock_mm(fshared); return ret; } @@ -1211,8 +1172,6 @@ static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared, q.pi_state = NULL; q.bitset = bitset; retry: - futex_lock_mm(fshared); - q.key = FUTEX_KEY_INIT; ret = get_futex_key(uaddr, fshared, &q.key); if (unlikely(ret != 0)) @@ -1245,12 +1204,6 @@ 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); - ret = get_user(uval, uaddr); if (!ret) @@ -1264,12 +1217,6 @@ static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared, /* 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 @@ -1355,7 +1302,6 @@ static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared, out_release_sem: put_futex_key(fshared, &q.key); - futex_unlock_mm(fshared); return ret; } @@ -1404,8 +1350,6 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared, q.pi_state = NULL; retry: - futex_lock_mm(fshared); - q.key = FUTEX_KEY_INIT; ret = get_futex_key(uaddr, fshared, &q.key); if (unlikely(ret != 0)) @@ -1495,7 +1439,6 @@ 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; @@ -1527,12 +1470,6 @@ 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: @@ -1545,7 +1482,6 @@ 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) { @@ -1611,7 +1547,6 @@ 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); @@ -1622,7 +1557,6 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared, out_release_sem: put_futex_key(fshared, &q.key); - futex_unlock_mm(fshared); if (to) destroy_hrtimer_on_stack(&to->timer); return ret; @@ -1646,8 +1580,6 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared, goto retry_unlocked; } - futex_unlock_mm(fshared); - ret = get_user(uval, uaddr); if (!ret && (uval != -EFAULT)) goto retry; @@ -1679,10 +1611,6 @@ 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)) @@ -1742,7 +1670,6 @@ out_unlock: spin_unlock(&hb->lock); out: put_futex_key(fshared, &key); - futex_unlock_mm(fshared); return ret; @@ -1766,8 +1693,6 @@ pi_faulted: goto retry_unlocked; } - futex_unlock_mm(fshared); - ret = get_user(uval, uaddr); if (!ret && (uval != -EFAULT)) goto retry; -- cgit v1.2.3 From 734b05b10e51d4ba38c8fc3ee02e846aab09eedf Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 26 Sep 2008 19:32:22 +0200 Subject: futex: use fast_gup() Change the get_user_pages() call with fast_gup() which doesn't require holding the mmap_sem thereby removing the mmap_sem from all fast paths. Signed-off-by: Peter Zijlstra Acked-by: Nick Piggin Signed-off-by: Ingo Molnar --- kernel/futex.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'kernel/futex.c') diff --git a/kernel/futex.c b/kernel/futex.c index 6a726684217e..facf17d1a705 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -232,9 +232,7 @@ static int get_futex_key(u32 __user *uaddr, struct rw_semaphore *fshared, } again: - down_read(&mm->mmap_sem); - err = get_user_pages(current, mm, address, 1, 0, 0, &page, NULL); - up_read(&mm->mmap_sem); + err = get_user_pages_fast(address, 1, 0, &page); if (err < 0) return err; -- cgit v1.2.3 From c2f9f20154bfb137ccdf8c9159992429a40dfe20 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 26 Sep 2008 19:32:23 +0200 Subject: futex: cleanup fshared fshared doesn't need to be a rw_sem pointer anymore, so clean that up. Signed-off-by: Peter Zijlstra Acked-by: Nick Piggin Signed-off-by: Ingo Molnar --- kernel/futex.c | 48 +++++++++++++++++++----------------------------- 1 file changed, 19 insertions(+), 29 deletions(-) (limited to 'kernel/futex.c') diff --git a/kernel/futex.c b/kernel/futex.c index facf17d1a705..60b47bb9e3dd 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -200,8 +200,7 @@ static void drop_futex_key_refs(union futex_key *key) * For other futexes, it points to ¤t->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; @@ -268,7 +267,7 @@ again: } static inline -void put_futex_key(struct rw_semaphore *fshared, union futex_key *key) +void put_futex_key(int fshared, union futex_key *key) { drop_futex_key_refs(key); } @@ -297,10 +296,8 @@ 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; @@ -687,8 +684,7 @@ 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; @@ -735,8 +731,7 @@ 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 = FUTEX_KEY_INIT, key2 = FUTEX_KEY_INIT; @@ -790,7 +785,7 @@ retry: */ if (attempt++) { ret = futex_handle_fault((unsigned long)uaddr2, - fshared, attempt); + attempt); if (ret) goto out; goto retry; @@ -841,8 +836,7 @@ 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 = FUTEX_KEY_INIT, key2 = FUTEX_KEY_INIT; @@ -1048,8 +1042,7 @@ 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; @@ -1128,7 +1121,7 @@ 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); @@ -1152,7 +1145,7 @@ handle_fault: static long futex_wait_restart(struct restart_block *restart); -static int futex_wait(u32 __user *uaddr, struct rw_semaphore *fshared, +static int futex_wait(u32 __user *uaddr, int fshared, u32 val, ktime_t *abs_time, u32 bitset) { struct task_struct *curr = current; @@ -1307,13 +1300,13 @@ 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 = ¤t->mm->mmap_sem; + fshared = 1; return (long)futex_wait(uaddr, fshared, restart->futex.val, &t, restart->futex.bitset); } @@ -1325,7 +1318,7 @@ 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; @@ -1571,8 +1564,7 @@ static int futex_lock_pi(u32 __user *uaddr, struct rw_semaphore *fshared, 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 retry_unlocked; @@ -1592,7 +1584,7 @@ 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; @@ -1683,8 +1675,7 @@ pi_faulted: 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; @@ -1816,8 +1807,7 @@ 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; } @@ -1913,10 +1903,10 @@ long do_futex(u32 __user *uaddr, int op, u32 val, ktime_t *timeout, { int ret = -ENOSYS; int cmd = op & FUTEX_CMD_MASK; - struct rw_semaphore *fshared = NULL; + int fshared = 0; if (!(op & FUTEX_PRIVATE_FLAG)) - fshared = ¤t->mm->mmap_sem; + fshared = 1; switch (cmd) { case FUTEX_WAIT: -- cgit v1.2.3 From 42569c39917a08e8de1e8b5685463be7b74baebd Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 30 Sep 2008 12:33:07 +0200 Subject: futex: fixup get_futex_key() for private futexes With the get_user_pages_fast() patches we made get_futex_key() obtain a reference on the returned key, but failed to do so for private futexes. Signed-off-by: Peter Zijlstra Acked-by: Nick Piggin Signed-off-by: Ingo Molnar --- kernel/futex.c | 1 + 1 file changed, 1 insertion(+) (limited to 'kernel/futex.c') diff --git a/kernel/futex.c b/kernel/futex.c index 60b47bb9e3dd..62cbd648e28a 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -227,6 +227,7 @@ static int get_futex_key(u32 __user *uaddr, int fshared, union futex_key *key) return -EFAULT; key->private.mm = mm; key->private.address = address; + get_futex_key_refs(key); return 0; } -- cgit v1.2.3 From 1acdac104668a0834cfa267de9946fac7764d486 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Thu, 20 Nov 2008 10:02:53 -0800 Subject: futex: make clock selectable for FUTEX_WAIT_BITSET FUTEX_WAIT_BITSET could be used instead of FUTEX_WAIT by setting the bit set to FUTEX_BITSET_MATCH_ANY, but FUTEX_WAIT uses CLOCK_REALTIME while FUTEX_WAIT_BITSET uses CLOCK_MONOTONIC. Add a flag to select CLOCK_REALTIME for FUTEX_WAIT_BITSET so glibc can replace the FUTEX_WAIT logic which needs to do gettimeofday() calls before and after the syscall to convert the absolute timeout to a relative timeout for FUTEX_WAIT. Signed-off-by: Thomas Gleixner Cc: Ulrich Drepper --- include/linux/futex.h | 3 ++- kernel/futex.c | 24 +++++++++++++++++------- 2 files changed, 19 insertions(+), 8 deletions(-) (limited to 'kernel/futex.c') diff --git a/include/linux/futex.h b/include/linux/futex.h index 8f627b9ae2b1..3bf5bb5a34f9 100644 --- a/include/linux/futex.h +++ b/include/linux/futex.h @@ -25,7 +25,8 @@ union ktime; #define FUTEX_WAKE_BITSET 10 #define FUTEX_PRIVATE_FLAG 128 -#define FUTEX_CMD_MASK ~FUTEX_PRIVATE_FLAG +#define FUTEX_CLOCK_REALTIME 256 +#define FUTEX_CMD_MASK ~(FUTEX_PRIVATE_FLAG | FUTEX_CLOCK_REALTIME) #define FUTEX_WAIT_PRIVATE (FUTEX_WAIT | FUTEX_PRIVATE_FLAG) #define FUTEX_WAKE_PRIVATE (FUTEX_WAKE | FUTEX_PRIVATE_FLAG) diff --git a/kernel/futex.c b/kernel/futex.c index e10c5c8786a6..ba0d3b83c091 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -1142,12 +1142,13 @@ 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, int fshared, - u32 val, ktime_t *abs_time, u32 bitset) + u32 val, ktime_t *abs_time, u32 bitset, int clockrt) { struct task_struct *curr = current; DECLARE_WAITQUEUE(wait, curr); @@ -1233,8 +1234,10 @@ static int futex_wait(u32 __user *uaddr, int 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); @@ -1289,6 +1292,8 @@ static int futex_wait(u32 __user *uaddr, int fshared, if (fshared) restart->futex.flags |= FLAGS_SHARED; + if (clockrt) + restart->futex.flags |= FLAGS_CLOCKRT; return -ERESTART_RESTARTBLOCK; } @@ -1312,7 +1317,8 @@ static long futex_wait_restart(struct restart_block *restart) if (restart->futex.flags & FLAGS_SHARED) fshared = 1; return (long)futex_wait(uaddr, fshared, restart->futex.val, &t, - restart->futex.bitset); + restart->futex.bitset, + restart->futex.flags & FLAGS_CLOCKRT); } @@ -1905,18 +1911,22 @@ 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; int fshared = 0; if (!(op & FUTEX_PRIVATE_FLAG)) 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; -- cgit v1.2.3 From 73500ac545d24610eb2cf8579ffc88957e9c5847 Mon Sep 17 00:00:00 2001 From: Darren Hart Date: Wed, 17 Dec 2008 17:29:56 -0800 Subject: futex: rename field in futex_q to clarify single waiter semantics Impact: simplify code I've tripped over the naming of this field a couple times. The futex_q uses a "waiters" list to represent a single blocked task and then calles wake_up_all(). This can lead to confusion in trying to understand the intent of the code, which is to have a single futex_q for every task waiting on a futex. This patch corrects the problem, using a single pointer to the waiting task, and an appropriate call to wake_up, rather than wake_up_all. Compile and boot tested on an 8way x86_64 machine. Signed-off-by: Darren Hart Acked-by: Thomas Gleixner Signed-off-by: Ingo Molnar --- kernel/futex.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) (limited to 'kernel/futex.c') diff --git a/kernel/futex.c b/kernel/futex.c index ba0d3b83c091..99f8acce08bf 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -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; @@ -573,7 +574,7 @@ 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. @@ -930,7 +931,7 @@ 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); @@ -1221,7 +1222,7 @@ static int futex_wait(u32 __user *uaddr, int 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. -- cgit v1.2.3 From b56863630ddbdea6e22df8835f78f0b1da037103 Mon Sep 17 00:00:00 2001 From: Darren Hart Date: Thu, 18 Dec 2008 15:06:34 -0800 Subject: futex: clean up futex_(un)lock_pi fault handling Impact: cleanup Some apparently left over cruft code was complicating the fault logic: Testing if uval != -EFAULT doesn't have any meaning, get_user() sets ret to either 0 or -EFAULT, there's no need to compare uval, especially not against EFAULT which it will never be. This patch removes the superfluous test and clarifies the comment blocks. Build and boot tested on an 8way x86_64 system. Signed-off-by: Darren Hart Signed-off-by: Ingo Molnar --- kernel/futex.c | 26 ++++++++++++-------------- 1 file changed, 12 insertions(+), 14 deletions(-) (limited to 'kernel/futex.c') diff --git a/kernel/futex.c b/kernel/futex.c index 99f8acce08bf..b4f87bac91c1 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -1565,12 +1565,11 @@ static int futex_lock_pi(u32 __user *uaddr, int fshared, 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); @@ -1582,7 +1581,7 @@ static int futex_lock_pi(u32 __user *uaddr, int fshared, } ret = get_user(uval, uaddr); - if (!ret && (uval != -EFAULT)) + if (!ret) goto retry; if (to) @@ -1676,12 +1675,11 @@ out: 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); @@ -1694,7 +1692,7 @@ pi_faulted: } ret = get_user(uval, uaddr); - if (!ret && (uval != -EFAULT)) + if (!ret) goto retry; return ret; -- cgit v1.2.3