Merge branch 'fix/waitqueue-barriers' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux into for-linus-4.4

12 files changed, 62 insertions, 7 deletions

diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c
index 57ee8ca29b06..3a9317ce67f8 100644
--- a/fs/btrfs/compression.c
+++ b/fs/btrfs/compression.c
@@ -839,6 +839,9 @@ static void free_workspace(int type, struct list_head *workspace)
         btrfs_compress_op[idx]->free_workspace(workspace);
         atomic_dec(alloc_workspace);
 wake:
+        /*
+         * Make sure counter is updated before we wake up waiters.
+         */
         smp_mb();
         if (waitqueue_active(workspace_wait))
                 wake_up(workspace_wait);
diff --git a/fs/btrfs/delayed-inode.c b/fs/btrfs/delayed-inode.c
index a2ae42720a6a..e0941fbb913c 100644
--- a/fs/btrfs/delayed-inode.c
+++ b/fs/btrfs/delayed-inode.c
@@ -463,6 +463,10 @@ static int __btrfs_add_delayed_deletion_item(struct btrfs_delayed_node *node,
 static void finish_one_item(struct btrfs_delayed_root *delayed_root)
 {
         int seq = atomic_inc_return(&delayed_root->items_seq);
+
+        /*
+         * atomic_dec_return implies a barrier for waitqueue_active
+         */
         if ((atomic_dec_return(&delayed_root->items) <
             BTRFS_DELAYED_BACKGROUND || seq % BTRFS_DELAYED_BATCH == 0) &&
             waitqueue_active(&delayed_root->wait))
diff --git a/fs/btrfs/dev-replace.c b/fs/btrfs/dev-replace.c
index a64ca942b63f..1e668fb7dd4c 100644
--- a/fs/btrfs/dev-replace.c
+++ b/fs/btrfs/dev-replace.c
@@ -454,8 +454,7 @@ static void btrfs_rm_dev_replace_blocked(struct btrfs_fs_info *fs_info)
 static void btrfs_rm_dev_replace_unblocked(struct btrfs_fs_info *fs_info)
 {
         clear_bit(BTRFS_FS_STATE_DEV_REPLACING, &fs_info->fs_state);
-        if (waitqueue_active(&fs_info->replace_wait))
-                wake_up(&fs_info->replace_wait);
+        wake_up(&fs_info->replace_wait);
 }
 
 static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info,
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 7f934cccc98d..a42c1f575ef7 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -801,6 +801,9 @@ static void run_one_async_done(struct btrfs_work *work)
         limit = btrfs_async_submit_limit(fs_info);
         limit = limit * 2 / 3;
 
+        /*
+         * atomic_dec_return implies a barrier for waitqueue_active
+         */
         if (atomic_dec_return(&fs_info->nr_async_submits) < limit &&
             waitqueue_active(&fs_info->async_submit_wait))
                 wake_up(&fs_info->async_submit_wait);
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 92a5b4cad2ab..522fb45d472a 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -10372,8 +10372,7 @@ void btrfs_end_write_no_snapshoting(struct btrfs_root *root)
 {
         percpu_counter_dec(&root->subv_writers->counter);
         /*
-         * Make sure counter is updated before we wake up
-         * waiters.
+         * Make sure counter is updated before we wake up waiters.
          */
         smp_mb();
         if (waitqueue_active(&root->subv_writers->wait))
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 25eb814c302f..208db4e835f0 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -1096,6 +1096,9 @@ static noinline void async_cow_submit(struct btrfs_work *work)
         nr_pages = (async_cow->end - async_cow->start + PAGE_CACHE_SIZE) >>
                 PAGE_CACHE_SHIFT;
 
+        /*
+         * atomic_sub_return implies a barrier for waitqueue_active
+         */
         if (atomic_sub_return(nr_pages, &root->fs_info->async_delalloc_pages) <
             5 * 1024 * 1024 &&
             waitqueue_active(&root->fs_info->async_submit_wait))
diff --git a/fs/btrfs/locking.c b/fs/btrfs/locking.c
index d7e6baf1b205..8077461fc56a 100644
--- a/fs/btrfs/locking.c
+++ b/fs/btrfs/locking.c
@@ -79,6 +79,9 @@ void btrfs_clear_lock_blocking_rw(struct extent_buffer *eb, int rw)
                 write_lock(&eb->lock);
                 WARN_ON(atomic_read(&eb->spinning_writers));
                 atomic_inc(&eb->spinning_writers);
+                /*
+                 * atomic_dec_and_test implies a barrier for waitqueue_active
+                 */
                 if (atomic_dec_and_test(&eb->blocking_writers) &&
                     waitqueue_active(&eb->write_lock_wq))
                         wake_up(&eb->write_lock_wq);
@@ -86,6 +89,9 @@ void btrfs_clear_lock_blocking_rw(struct extent_buffer *eb, int rw)
                 BUG_ON(atomic_read(&eb->blocking_readers) == 0);
                 read_lock(&eb->lock);
                 atomic_inc(&eb->spinning_readers);
+                /*
+                 * atomic_dec_and_test implies a barrier for waitqueue_active
+                 */
                 if (atomic_dec_and_test(&eb->blocking_readers) &&
                     waitqueue_active(&eb->read_lock_wq))
                         wake_up(&eb->read_lock_wq);
@@ -229,6 +235,9 @@ void btrfs_tree_read_unlock_blocking(struct extent_buffer *eb)
         }
         btrfs_assert_tree_read_locked(eb);
         WARN_ON(atomic_read(&eb->blocking_readers) == 0);
+        /*
+         * atomic_dec_and_test implies a barrier for waitqueue_active
+         */
         if (atomic_dec_and_test(&eb->blocking_readers) &&
             waitqueue_active(&eb->read_lock_wq))
                 wake_up(&eb->read_lock_wq);
@@ -280,6 +289,9 @@ void btrfs_tree_unlock(struct extent_buffer *eb)
         if (blockers) {
                 WARN_ON(atomic_read(&eb->spinning_writers));
                 atomic_dec(&eb->blocking_writers);
+                /*
+                 * Make sure counter is updated before we wake up waiters.
+                 */
                 smp_mb();
                 if (waitqueue_active(&eb->write_lock_wq))
                         wake_up(&eb->write_lock_wq);
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
index 52170cf1757e..071005f008c1 100644
--- a/fs/btrfs/ordered-data.c
+++ b/fs/btrfs/ordered-data.c
@@ -345,6 +345,9 @@ int btrfs_dec_test_first_ordered_pending(struct inode *inode,
 
         if (entry->bytes_left == 0) {
                 ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags);
+                /*
+                 * Implicit memory barrier after test_and_set_bit
+                 */
                 if (waitqueue_active(&entry->wait))
                         wake_up(&entry->wait);
         } else {
@@ -409,6 +412,9 @@ have_entry:
 
         if (entry->bytes_left == 0) {
                 ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags);
+                /*
+                 * Implicit memory barrier after test_and_set_bit
+                 */
                 if (waitqueue_active(&entry->wait))
                         wake_up(&entry->wait);
         } else {
diff --git a/fs/btrfs/raid56.c b/fs/btrfs/raid56.c
index fcf7265ca46f..1a33d3eb36de 100644
--- a/fs/btrfs/raid56.c
+++ b/fs/btrfs/raid56.c
@@ -810,7 +810,11 @@ static noinline void unlock_stripe(struct btrfs_raid_bio *rbio)
                         }
 
                         goto done_nolock;
-                } else  if (waitqueue_active(&h->wait)) {
+                        /*
+                         * The barrier for this waitqueue_active is not needed,
+                         * we're protected by h->lock and can't miss a wakeup.
+                         */
+                } else if (waitqueue_active(&h->wait)) {
                         spin_unlock(&rbio->bio_list_lock);
                         spin_unlock_irqrestore(&h->lock, flags);
                         wake_up(&h->wait);
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 76354bb1312c..47fd27031a20 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -82,6 +82,12 @@ void btrfs_put_transaction(struct btrfs_transaction *transaction)
 static void clear_btree_io_tree(struct extent_io_tree *tree)
 {
         spin_lock(&tree->lock);
+        /*
+         * Do a single barrier for the waitqueue_active check here, the state
+         * of the waitqueue should not change once clear_btree_io_tree is
+         * called.
+         */
+        smp_mb();
         while (!RB_EMPTY_ROOT(&tree->state)) {
                 struct rb_node *node;
                 struct extent_state *state;
@@ -856,6 +862,9 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
         atomic_dec(&cur_trans->num_writers);
         extwriter_counter_dec(cur_trans, trans->type);
 
+        /*
+         * Make sure counter is updated before we wake up waiters.
+         */
         smp_mb();
         if (waitqueue_active(&cur_trans->writer_wait))
                 wake_up(&cur_trans->writer_wait);
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index c3f9a9c71f28..1fffe8845c09 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -229,7 +229,9 @@ int btrfs_pin_log_trans(struct btrfs_root *root)
 void btrfs_end_log_trans(struct btrfs_root *root)
 {
         if (atomic_dec_and_test(&root->log_writers)) {
-                smp_mb();
+                /*
+                 * Implicit memory barrier after atomic_dec_and_test
+                 */
                 if (waitqueue_active(&root->log_writer_wait))
                         wake_up(&root->log_writer_wait);
         }
@@ -2820,7 +2822,9 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 
         mutex_lock(&log_root_tree->log_mutex);
         if (atomic_dec_and_test(&log_root_tree->log_writers)) {
-                smp_mb();
+                /*
+                 * Implicit memory barrier after atomic_dec_and_test
+                 */
                 if (waitqueue_active(&log_root_tree->log_writer_wait))
                         wake_up(&log_root_tree->log_writer_wait);
         }
@@ -2950,6 +2954,9 @@ out_wake_log_root:
         atomic_set(&log_root_tree->log_commit[index2], 0);
         mutex_unlock(&log_root_tree->log_mutex);
 
+        /*
+         * The barrier before waitqueue_active is implied by mutex_unlock
+         */
         if (waitqueue_active(&log_root_tree->log_commit_wait[index2]))
                 wake_up(&log_root_tree->log_commit_wait[index2]);
 out:
@@ -2961,6 +2968,9 @@ out:
         atomic_set(&root->log_commit[index1], 0);
         mutex_unlock(&root->log_mutex);
 
+        /*
+         * The barrier before waitqueue_active is implied by mutex_unlock
+         */
         if (waitqueue_active(&root->log_commit_wait[index1]))
                 wake_up(&root->log_commit_wait[index1]);
         return ret;
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index dd644d767782..cc1cc3559045 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -344,6 +344,9 @@ loop_lock:
                 pending = pending->bi_next;
                 cur->bi_next = NULL;
 
+                /*
+                 * atomic_dec_return implies a barrier for waitqueue_active
+                 */
                 if (atomic_dec_return(&fs_info->nr_async_bios) < limit &&
                     waitqueue_active(&fs_info->async_submit_wait))
                         wake_up(&fs_info->async_submit_wait);