Re: [PATCH v4 02/12] x86/xen: simplify flush_lazy_mmu()

[Kevin Brodsky <kevin.brodsky@xxxxxxx>]

On 07/11/2025 12:31, Ryan Roberts wrote:
> On 29/10/2025 10:08, Kevin Brodsky wrote:
>> arch_flush_lazy_mmu_mode() is called when outstanding batched
>> pgtable operations must be completed immediately. There should
>> however be no need to leave and re-enter lazy MMU completely. The
>> only part of that sequence that we really need is xen_mc_flush();
>> call it directly.
>>
>> Signed-off-by: Kevin Brodsky <kevin.brodsky@xxxxxxx>
> This looks functionally equivalent to me, so:
>
> Reviewed-by: Ryan Roberts <ryan.roberts@xxxxxxx>
>
> But I don't think this tidy up is strictly necessary for your series to work?
> (perhaps I'll change my mind on that as I go through it).

I initially thought it might be, but in the end I think you're right -
it should still work fine without this patch.

Still, I'd rather avoid unnecessary calls to arch_enter() and
arch_leave() as it makes it harder to reason about what is called where.
Namely, keeping them here means that a nested call to
lazy_mmu_mode_disable() would cause arch_leave() then arch_enter() to be
called - rather unexpected.

The only calls to arch_enter() and arch_leave() that are left after this
series are the ones in <linux/pgtable.h> and the Xen context-switching
logic (the one case where calling arch hooks directly is justified, see
discussion on v3 [1]).

- Kevin

[1]
https://lore.kernel.org/all/390e41ae-4b66-40c1-935f-7a1794ba0b71@xxxxxxx/