On Tue, Jan 17, 2023 at 12:27:59PM +0100, Pablo Neira Ayuso wrote: [...] > Otherwise, with large sets, starting from rb_first() is slow (it takes > 30s if I use rb_first() instead of this approach), because of the linear > list walk. This should be instead: Otherwise, with large sets, starting from rb_first() is slow (it takes 30s if I use rb_first() instead of 3s), because of the linear list walk. so it is 3 seconds with the speed up vs. 30 seconds starting from rb_first().