Btrfs: only use the existing eb if it's count isn't 0

We can run into a problem where we find an eb for our existing page already on
the radix tree but it has a ref count of 0.  It hasn't yet been removed by RCU
yet so this can cause issues where we will use the EB after free.  So do
atomic_inc_not_zero on the exists->refs and if it is zero just do
synchronize_rcu() and try again.  We won't have to worry about new allocators
coming in since they will block on the page lock at this point.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 0381b6007ae45d15250351b49d5770f923014b95..0f74262911bee77941708c8709e3a70e0e9f6dc8 100644 (file)
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -3750,7 +3750,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
        }
        if (uptodate)
                set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
-
+again:
        ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
        if (ret)
                goto free_eb;
@@ -3760,7 +3760,13 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
        if (ret == -EEXIST) {
                exists = radix_tree_lookup(&tree->buffer,
                                                start >> PAGE_CACHE_SHIFT);
-               atomic_inc(&exists->refs);
+               if (!atomic_inc_not_zero(&exists->refs)) {
+                       spin_unlock(&tree->buffer_lock);
+                       radix_tree_preload_end();
+                       synchronize_rcu();
+                       exists = NULL;
+                       goto again;
+               }
                spin_unlock(&tree->buffer_lock);
                radix_tree_preload_end();
                goto free_eb;