Re: [PATCH v2 6/7] x86/Kconfig: Introduce ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE

[David Hildenbrand]

On 26.02.21 13:04, Oscar Salvador wrote:
> On Thu, Feb 25, 2021 at 07:29:07PM +0100, David Hildenbrand wrote:
> > On 09.02.21 14:38, Oscar Salvador wrote:
> > > Enable x86_64 platform to use the MHP_MEMMAP_ON_MEMORY feature.
> > >
> > > Signed-off-by: Oscar Salvador <osalvador@xxxxxxx>
> > > ---
> > >    arch/x86/Kconfig | 4 ++++
> > >    1 file changed, 4 insertions(+)
> > >
> > > diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
> > > index 72663de8b04c..81046b7adb10 100644
> > > --- a/arch/x86/Kconfig
> > > +++ b/arch/x86/Kconfig
> > > @@ -2440,6 +2440,10 @@ config ARCH_ENABLE_MEMORY_HOTREMOVE
> > >    	def_bool y
> > >    	depends on MEMORY_HOTPLUG
> > > +config ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
> > > +	def_bool y
> > > +	depends on X86_64 && MEMORY_HOTPLUG && SPARSEMEM_VMEMMAP_ENABLE
> >
> > It depends on SPARSEMEM_VMEMMAP, no?
> >
> > I think we could have
> >
> > SPARSEMEM_VMEMMAP_ENABLE=y
> > and
> > SPARSEMEM_VMEMMAP=n
> >
> > on manually tuned configs.
>
> I do not think this can happen:
>
> from mm/Kconfig:
>
> config SPARSEMEM_VMEMMAP_ENABLE
>          bool
>
> config SPARSEMEM_VMEMMAP
>          bool "Sparse Memory virtual memmap"
>          depends on SPARSEMEM && SPARSEMEM_VMEMMAP_ENABLE
>          default y
>          help
>            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
>
> and from arch/x86/Kconfig:
>
> config ARCH_SPARSEMEM_ENABLE
>          def_bool y
>          depends on X86_64 || NUMA || X86_32 || X86_32_NON_STANDARD
>          select SPARSEMEM_STATIC if X86_32
>          select SPARSEMEM_VMEMMAP_ENABLE if X86_64
>
>
> So, if I read this correctly, for SPARSEMEM_VMEMMAP to be true,
> SPARSEMEM_VMEMMAP_ENABLE needs to be true as well.
>
> Am I missing something?

Take your config and set
	X86_5LEVEL=n
(because it enforces SPARSEMEM_VMEMMAP)
and
	SPARSEMEM_VMEMMAP=n

When compiling, you'll end up with a config like
	CONFIG_SPARSEMEM_VMEMMAP_ENABLE=y
	# CONFIG_SPARSEMEM_VMEMMAP is not set

Yet, with your patch you would get

ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE=y

And it would not get fenced off in the code, right?


I think you either have to check (IS_ENABLED(CONFIG_SPARSEMEM_VMEMMAP)) 
in addition in your code, or enforce it differently. Like


config MHP_MEMMAP_ON_MEMORY
	depends on SPARSEMEM_VMEMMAP && ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
	bool


Then you can simplify the arch Kconfig settings, removing the sparesemem 
dependency there.

--
Thanks,

David / dhildenb