Re: [PATCH] mm/Kconfig: allow user to select MIGRATION if MEMORY_FAILURE is enabled

["David Hildenbrand (Arm)" <david@kernel.org>]

On 8/13/26 09:23, Xie Yuanbin wrote:
> On Thu, 13 Aug 2026 09:06:07 +0200, David Hildenbrand wrote:
>> Why add a new config option?
> 
> I think this provides more flexibility in configuration, when:
> 	CONFIG_COMPACTION=n
> 	CONFIG_MEMORY_HOTREMOVE=n
> 	CONFIG_NUMA_MIGRATION=n
> 	CONFIG_CMA=n
> 	CONFIG_MEMORY_FAILURE=y
> Users can choose whether to enable migration, just like what commit
> 6ebf98d71f9b509e833e ("mm: introduce CONFIG_NUMA_MIGRATION and simplify
> CONFIG_MIGRATION") done. However, it does add a new config, which may
> confuse the users...
> 
> I think that just making MEMORY_FAILURE select MIGRATION is also a very
> good solution, and I respect the maintainers' opinion.

There must be a good reason to do something fine grained like
MEMORY_FAILURE_MIGRATON, really.

So if there is a use case out there that absolutely doesn't want
CONFIG_MIGRATION but does want CONFIG_MEMORY_FAILURE, we could discuss it.

As really only softdirty needs page migration (IIRC), we could also just put
that under a separate config that implies CONFIG_MIGRATION, like
CONFIG_MEMORY_FAILURE_SOFT_OFFLINE. But I'd rather avoid that unless really
required.

-- 
Cheers,

David