On Thu, Jun 12, 2025 at 05:00:51PM +0100, Ryan Roberts wrote:

On 12/06/2025 15:59, Will Deacon wrote:

quoted

To be honest, I don't think the proposal in this series is really
improving what we have. Either we support nested lazy mode or we don't;
having __kernel_map_pages() mess around with the lazy mmu state because
it somehow knows that set_memory_valid() is going to use it is fragile
and ugly.

So I'm incined to leave the current code as-is, unless we can remove it
in favour of teaching the core code how to handle it instead.

Yeah fair enough. I'm not going to have time to do the proper nesting support
thing. But I'll see if I can find someone internally that I might be able to
convince. If not, we'll just leave as is.

Sounds fine by me, thanks Ryan.

Will