Re: [PATCH v2 2/7] mm: introduce local state for lazy_mmu sections

[Kevin Brodsky <kevin.brodsky@xxxxxxx>]

On 11/09/2025 14:06, Alexander Gordeev wrote:
> On Wed, Sep 10, 2025 at 06:11:54PM +0200, Kevin Brodsky wrote:
>
> Hi Kevin,
>
>> On 09/09/2025 16:38, Alexander Gordeev wrote:
>>>>>>> Would that integrate well with LAZY_MMU_DEFAULT etc?
>>>>>> Hmm... I though the idea is to use LAZY_MMU_* by architectures that
>>>>>> want to use it - at least that is how I read the description above.
>>>>>>
>>>>>> It is only kasan_populate|depopulate_vmalloc_pte() in generic code
>>>>>> that do not follow this pattern, and it looks as a problem to me.
>>>> This discussion also made me realise that this is problematic, as the
>>>> LAZY_MMU_{DEFAULT,NESTED} macros were meant only for architectures'
>>>> convenience, not for generic code (where lazy_mmu_state_t should ideally
>>>> be an opaque type as mentioned above). It almost feels like the kasan
>>>> case deserves a different API, because this is not how enter() and
>>>> leave() are meant to be used. This would mean quite a bit of churn
>>>> though, so maybe just introduce another arch-defined value to pass to
>>>> leave() for such a situation - for instance,
>>>> arch_leave_lazy_mmu_mode(LAZY_MMU_FLUSH)?
>>> What about to adjust the semantics of apply_to_page_range() instead?
>>>
>>> It currently assumes any caller is fine with apply_to_pte_range() to
>>> enter the lazy mode. By contrast, kasan_(de)populate_vmalloc_pte() are
>>> not fine at all and must leave the lazy mode. That literally suggests
>>> the original assumption is incorrect.
>>>
>>> We could change int apply_to_pte_range(..., bool create, ...) to e.g.
>>> apply_to_pte_range(..., unsigned int flags, ...) and introduce a flag
>>> that simply skips entering the lazy mmu mode.
>> This is pretty much what Ryan proposed [1r] some time ago, although for
>> a different purpose (avoiding nesting). There wasn't much appetite for
>> it then, but I agree that this would be a more logical way to go about it.
>>
>> - Kevin
>>
>> [1r]
>> https://lore.kernel.org/all/20250530140446.2387131-4-ryan.roberts@xxxxxxx/
> May be I missing the point, but I read it as an opposition to the whole
> series in general and to the way apply_to_pte_range() would be altered
> in particular:
>
>  static int apply_to_pte_range(struct mm_struct *mm, pmd_t *pmd,
>                                    unsigned long addr, unsigned long end,
>                                    pte_fn_t fn, void *data, bool create,
> -                                  pgtbl_mod_mask *mask)
> +                                  pgtbl_mod_mask *mask, bool lazy_mmu)
>
> The idea of instructing apply_to_page_range() to skip the lazy mmu mode
> was not countered. Quite opposite, Liam suggested exactly the same:

Yes that's a fair point. It would be sensible to post a new series
trying to eliminate the leave()/enter() calls in mm/kasan as you
suggested. Still I think that it makes sense to define an API to handle
that situation ("pausing" lazy_mmu), as discussed with David H.

- Kevin

>
> <quote>
> Could we do something like the pgtbl_mod_mask or zap_details and pass
> through a struct or one unsigned int for create and lazy_mmu?
>
> These wrappers are terrible for readability and annoying for argument
> lists too.
>
> Could we do something like the pgtbl_mod_mask or zap_details and pass
> through a struct or one unsigned int for create and lazy_mmu?
>
> At least we'd have better self-documenting code in the wrappers.. and if
> we ever need a third boolean, we could avoid multiplying the wrappers
> again.
> <quote>
>
> Thanks!