On Wed, Apr 03, 2019 at 03:21:22PM +1100, Tobin C. Harding wrote:
> +void xa_object_migrate(struct xa_node *node, int numa_node)
> +{
> + struct xarray *xa = READ_ONCE(node->array);
> + void __rcu **slot;
> + struct xa_node *new_node;
> + int i;
> +
> + /* Freed or not yet in tree then skip */
> + if (!xa || xa == XA_RCU_FREE)
> + return;
> +
> + new_node = kmem_cache_alloc_node(radix_tree_node_cachep,
> + GFP_KERNEL, numa_node);
> + if (!new_node)
> + return;
> +
> + xa_lock_irq(xa);
> +
> + /* Check again..... */
> + if (xa != node->array || !list_empty(&node->private_list)) {
> + node = new_node;
> + goto unlock;
> + }
> +
> + memcpy(new_node, node, sizeof(struct xa_node));
> +
> + /* Move pointers to new node */
> + INIT_LIST_HEAD(&new_node->private_list);
Surely we can do something more clever, like ...
if (xa != node->array) {
...
if (list_empty(&node->private_list))
INIT_LIST_HEAD(&new_node->private_list);
else
list_replace(&node->private_list, &new_node->private_list);
BTW, the raidx tree nodes / xa_nodes share the same slab cache; we need
to finish converting all radix tree & IDR users to the XArray before
this series can go in.
