Re: [PATCH V2] LoongArch: Introduce the numa_memblks conversion

[Mike Rapoport <rppt@kernel.org>]

Hi Huacai,

On Tue, Apr 15, 2025 at 03:21:18PM +0800, Huacai Chen wrote:
> Commit 87482708210ff3333a ("mm: introduce numa_memblks") has moved
> numa_memblks from x86 to the generic code, but LoongArch was left out
> of this conversion.
> 
> This patch introduces the generic numa_memblks for LoongArch.
> 
> In detail:
> 1. Enable NUMA_MEMBLKS (but disable NUMA_EMU) in Kconfig;
> 2. Use generic definition for numa_memblk and numa_meminfo;
> 3. Use generic implementation for numa_add_memblk() and its friends;
> 4. Use generic implementation for numa_set_distance() and its friends;
> 5. Use generic implementation for memory_add_physaddr_to_nid() and its
>    friends.
>
> Note: Disable NUMA_EMU because it needs more efforts and no obvious
> demand now.
> 
> Tested-by: Binbin Zhou <zhoubinbin@loongson.cn>
> Signed-off-by: Yuquan Wang <wangyuquan1236@phytium.com.cn>
> Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
> ---
>  arch/loongarch/Kconfig                 |   1 +
>  arch/loongarch/include/asm/numa.h      |  14 ----
>  arch/loongarch/include/asm/sparsemem.h |   5 --
>  arch/loongarch/include/asm/topology.h  |   7 +-
>  arch/loongarch/kernel/acpi.c           |  11 ---
>  arch/loongarch/kernel/numa.c           | 108 +++----------------------
>  arch/loongarch/mm/init.c               |   8 --
>  mm/Kconfig                             |   1 +
>  8 files changed, 16 insertions(+), 139 deletions(-)

This is really cool simplification!

I wonder if maybe we can take a step or two further and make loongarch to
share more code with other architectures.

Is there a strong reason not to call numa_memblks_init() from
init_numa_memory()? 
And more broadly, maybe loongarch can be even switched to arch_numa?

-- 
Sincerely yours,
Mike.