Re: [PATCH 2/2] arm64: mm: relax VM_ALLOW_HUGE_VMAP if BBML2_NOABORT is supported

[Dev Jain <dev.jain@arm.com>]

On 17/10/25 12:20 am, Yang Shi wrote:
>
>
> On 10/14/25 11:50 PM, Dev Jain wrote:
>>
>> On 14/10/25 4:57 am, Yang Shi wrote:
>>> When changing permissions for vmalloc area, VM_ALLOW_HUGE_VMAP area is
>>> exclueded because kernel can't split the va mapping if it is called on
>>> partial range.
>>> It is no longer true if the machines support BBML2_NOABORT after commit
>>> a166563e7ec3 ("arm64: mm: support large block mapping when 
>>> rodata=full").
>>> So we can relax this restriction and update the comments accordingly.
>>>
>>> Fixes: a166563e7ec3 ("arm64: mm: support large block mapping when 
>>> rodata=full")
>>> Signed-off-by: Yang Shi <yang@os.amperecomputing.com>
>>> ---
>>>   arch/arm64/mm/pageattr.c | 13 +++++++------
>>>   1 file changed, 7 insertions(+), 6 deletions(-)
>>>
>>> diff --git a/arch/arm64/mm/pageattr.c b/arch/arm64/mm/pageattr.c
>>> index c21a2c319028..b4dcae6273a8 100644
>>> --- a/arch/arm64/mm/pageattr.c
>>> +++ b/arch/arm64/mm/pageattr.c
>>> @@ -157,13 +157,13 @@ static int change_memory_common(unsigned long 
>>> addr, int numpages,
>>>         /*
>>>        * Kernel VA mappings are always live, and splitting live section
>>> -     * mappings into page mappings may cause TLB conflicts. This means
>>> -     * we have to ensure that changing the permission bits of the 
>>> range
>>> -     * we are operating on does not result in such splitting.
>>> +     * mappings into page mappings may cause TLB conflicts on the 
>>> machines
>>> +     * which don't support BBML2_NOABORT.
>>>        *
>>>        * Let's restrict ourselves to mappings created by vmalloc (or 
>>> vmap).
>>> -     * Disallow VM_ALLOW_HUGE_VMAP mappings to guarantee that only 
>>> page
>>> -     * mappings are updated and splitting is never needed.
>>> +     * Disallow VM_ALLOW_HUGE_VMAP mappings if the systems don't 
>>> support
>>> +     * BBML2_NOABORT to guarantee that only page mappings are 
>>> updated and
>>> +     * splitting is never needed on those machines.
>>>        *
>>>        * So check whether the [addr, addr + size) interval is entirely
>>>        * covered by precisely one VM area that has the VM_ALLOC flag 
>>> set.
>>> @@ -171,7 +171,8 @@ static int change_memory_common(unsigned long 
>>> addr, int numpages,
>>>       area = find_vm_area((void *)addr);
>>>       if (!area ||
>>>           end > (unsigned long)kasan_reset_tag(area->addr) + 
>>> area->size ||
>>> -        ((area->flags & (VM_ALLOC | VM_ALLOW_HUGE_VMAP)) != VM_ALLOC))
>>> +        !(area->flags & VM_ALLOC) || ((area->flags & 
>>> VM_ALLOW_HUGE_VMAP) &&
>>> +        !system_supports_bbml2_noabort()))
>>>           return -EINVAL;
>>>         if (!numpages)
>>
>> This will conflict with my upcoming vmalloc-huge series, so best to 
>> leave it to me,
>> I already have this included :)
>
> My point is that I hope this can be merged as a hotfix for 6.18. I 
> have no strong opinion on either the maintainers take this one or from 
> your series. But if this will go into 6.18 as a hotfix, it should be 
> also a prerequisite patch (standalone)  in your series, and the rest 
> of your series should be based on top of it. Of course this argument 
> will not stand if we don't care to have it fixed for 6.18.

I see what you mean, but I don't think this patch should be treated as a 
hotfix. We forgot to relax a

restriction - that's fine. That is not an incorrectness in the linear 
map series. A fix usually

fixes an incorrectness.


>
> Thanks,
> Yang
>
>