On Mon, 10 Jul 2023 15:49:44 +0800 Yajun Deng <yajun.deng@xxxxxxxxx> wrote: > The kernel parameter 'cma_pernuma=' only supports reserving the same > size of CMA area for each node. We need to reserve different sizes of > CMA area for specified nodes if these devices belong to different nodes. > > Change kernel parameter from 'cma_pernuma=' to 'numa_cma='. If node > isn't specified, it will reserve CMA area for per-numa node. if the node > is specified, it will reserve CMA area for the specified node. > > At the same time, print the node id in cma_declare_contiguous_nid() if > enabled CONFIG_NUMA. Changing an existing kernel parameter will break existing setups. It would be preferable to add a new parameter for this?
