Yunsheng Lin <yunshenglin0825@xxxxxxxxx> writes: > On 3/14/2025 6:10 PM, Toke Høiland-Jørgensen wrote: >> Change the single-bit booleans for dma_sync into an unsigned long with >> BIT() definitions so that a subsequent patch can write them both with a >> singe WRITE_ONCE() on teardown. Also move the check for the sync_cpu >> side into __page_pool_dma_sync_for_cpu() so it can be disabled for >> non-netmem providers as well. > > I guess this patch is for the preparation of disabling the > page_pool_dma_sync_for_cpu() related API on teardown? > > It seems unnecessary that page_pool_dma_sync_for_cpu() related API need > to be disabled on teardown as page_pool_dma_sync_for_cpu() has the same > calling assumption as the alloc API, which is not supposed to be called > by the drivers when page_pool_destroy() is called. Sure, we could keep it to the dma_sync_for_dev() direction only, but making both directions use the same variable to store the state, and just resetting it at once, seemed simpler and more consistent. -Toke
