On 03/20/24 at 11:36am, Mike Rapoport wrote: > On Wed, Mar 20, 2024 at 03:52:52PM +0800, Baoquan He wrote: > > On 03/19/24 at 05:49pm, Mike Rapoport wrote: > > > Hi Baoquan, > > > > > > On Mon, Mar 18, 2024 at 10:21:34PM +0800, Baoquan He wrote: > > > > This is not needed any more. > > > > > > I'd swap this and the first patch, so that the first patch would remove > > > memblock_find_dma_reserve() and it's changelog will explain why it's not > > > needed and then the second patch will simply drop unused set_dma_reserve() > > > > Thanks, Mike. > > > > My thought on the patch 1/2 splitting is: > > patch 1 is removing all relevant codes in mm, including the usage of > > dma_reserve in free_area_init_core() and exporting set_dma_reserve() > > to any ARCH which want to subtract the dma_reserve from DMA zone. > > > > Patch 2 purely remove the code in x86 ARCH about how to get dma_reserve. > > I think it's better first to remove the usage of set_dma_reserve() in x86 > and then clean up the unused code. OK, firslty remove the only user, that sounds reasonable. Will change. Thanks.
