Re: [PATCH] LoongArch: Add sparse memory vmemmap support

[Huacai Chen <chenhuacai@kernel.org>]

Hi, Arnd,

On Mon, Jun 20, 2022 at 7:27 PM Arnd Bergmann <arnd@arndb.de> wrote:
>
> On Mon, Jun 20, 2022 at 12:24 PM Huacai Chen <chenhuacai@kernel.org> wrote:
> > On Fri, Jun 17, 2022 at 11:42 PM Arnd Bergmann <arnd@arndb.de> wrote:
> > >
> > > On Fri, Jun 17, 2022 at 4:58 PM Huacai Chen <chenhuacai@loongson.cn> wrote:
> > > >
> > > > Add sparse memory vmemmap support for LoongArch. SPARSEMEM_VMEMMAP
> > > > uses a virtually mapped memmap to optimise pfn_to_page and page_to_pfn
> > > > operations. This is the most efficient option when sufficient kernel
> > > > resources are available.
> > >
> > > I have not looked at this in detail, but from a high-level perspective, it
> > > seems very similar to the corresponding code in arch/arm64 and arch/x86.
> > >
> > > Can you try to merge the three copies into a generic helper and add that
> > > to mm/sparse-vmemmap.c? If this does not work, can you describe in the
> > > changelog text why these have to be architecture specific?
> >
> > It is difficult to merge, because LoongArch needs to call pud_init(),
> > pmd_init() and other similar things which are unnecessary on ARM64 and
> > X86.
>
>  I can see that this part of the code is the same as mips. What is the
> initialization
>  for? Can it be encapsulated in an architecture specific inline
> function (for mips
> and loongarch), with a generic version like this?
>
> #ifndef pud_init_invalid
> static inline void pud_init_invalid(void *) {}
> #endif
OK, let's have a try.

Huacai
>
>        Arnd