Re: [PATCH 03/10] rbtree: Provide rb_find_rcu() / rb_find_add_rcu()

From: Peter Zijlstra
Date: Wed Jul 10 2024 - 03:56:12 EST


On Wed, Jul 10, 2024 at 10:29:59AM +0900, Masami Hiramatsu wrote:
> On Mon, 08 Jul 2024 11:12:44 +0200
> Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>
> > Much like latch_tree, add two RCU methods for the regular RB-tree,
> > which can be used in conjunction with a seqcount to provide lockless
> > lookups.
> >
> > Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
> > ---
> > include/linux/rbtree.h | 67 +++++++++++++++++++++++++++++++++++++++++++++++++
> > 1 file changed, 67 insertions(+)
> >
> > --- a/include/linux/rbtree.h
> > +++ b/include/linux/rbtree.h
> > @@ -245,6 +245,42 @@ rb_find_add(struct rb_node *node, struct
> > }
> >
> > /**
> > + * rb_find_add_rcu() - find equivalent @node in @tree, or add @node
> > + * @node: node to look-for / insert
> > + * @tree: tree to search / modify
> > + * @cmp: operator defining the node order
> > + *
> > + * Adds a Store-Release for link_node.
> > + *
> > + * Returns the rb_node matching @node, or NULL when no match is found and @node
> > + * is inserted.
> > + */
> > +static __always_inline struct rb_node *
> > +rb_find_add_rcu(struct rb_node *node, struct rb_root *tree,
> > + int (*cmp)(struct rb_node *, const struct rb_node *))
> > +{
> > + struct rb_node **link = &tree->rb_node;
> > + struct rb_node *parent = NULL;
> > + int c;
> > +
> > + while (*link) {
>
> Don't we need to use rcu_dereference_raw(*link) here?

This is a modifying operation and as such we can assume operation under
the exclusive lock. IOW the tree should be stable here.

> > + parent = *link;
> > + c = cmp(node, parent);
> > +
> > + if (c < 0)
> > + link = &parent->rb_left;
> > + else if (c > 0)
> > + link = &parent->rb_right;
> > + else
> > + return parent;
> > + }
> > +
> > + rb_link_node_rcu(node, parent, link);

Only the link operation needs the rcu_assign_pointer() thing for
publishing our new node.

> > + rb_insert_color(node, tree);

The rotations use WRITE_ONCE() to avoid tearing.

> > + return NULL;
> > +}