On 2021/5/7 18:19, Jesper Dangaard Brouer wrote: > On Fri, 7 May 2021 16:28:30 +0800 > Yunsheng Lin <linyunsheng@xxxxxxxxxx> wrote: > [...] > > I'm actually not against storing page_pool object ptr directly in > struct-page. It is a nice optimization. Perhaps we should implement > this optimization outside this patchset first, and let __xdp_return() > for XDP-redirected packets also take advantage to this optimization? > > Then it would feel more natural to also used this optimization in this > patchset, right? Yes, it would be good if XDP can take advantage of this optimization too. Then it seems we can remove the "mem_id_ht"? if we want to support different type of page pool in the future, the type field is in the page pool too when page_pool object ptr directly in struct-page. >
