Re: [PATCH -mm v2 01/11] mm: track free size between VMAs in VMA rbtree

[Peter Zijlstra <peterz@infradead.org>]

On Tue, 2012-06-26 at 11:49 -0400, Rik van Riel wrote:
> 
> However, doing an insert or delete changes the
> gap size for the _next_ vma, and potentially a
> change in the maximum gap size for the parent
> node, so both insert and delete cause two tree
> walks :( 

Right,.. don't have anything smart for that :/

I guess there's nothing to it but create a number of variants of
rb_insert/rb_erase, possibly using Daniel's 'template' stuff so we don't
actually have to maintain multiple copies of the code.

Maybe something simple like:

static void __always_inline
__rb_insert(struct rb_node *node, struct rb_root *root, rb_augment_f func, bool threaded)
{
	/* all the fancy code */
}

void rb_insert(struct rb_node *node, struct rb_root *root)
{
	__rb_insert(node, root, NULL, false);
}

void rb_insert_threaded(struct rb_node *node, struct rb_root *root)
{
	__rb_insert(node, root, NULL, true);
}

void rb_insert_augment(struct rb_node *node, struct rb_root *root, rb_augment_f func)
{
	__rb_insert(node, root, func, false);
}

void rb_insert_augment_threaded(struct rb_node *node, struct rb_root *root, rb_augment_f func)
{
	__rb_insert(node, root, func, true);
}

Would do, except it wouldn't be able to inline the augment function. For
that to happen we'd need to move __rb_insert() and the
__rb_insert_augment*() variants into rbtree.h.

But it would create clean variants without augmentation/threading
without too much duplicate code.


BTW, is there a reason rb_link_node() and rb_insert_color() are separate
functions? They seem to always be used together in sequence.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>