Re: [PATCH 2/4] arm64: hide more compat_vdso code

[Dmitry Safonov]

On 10/29/20 1:35 PM, Arnd Bergmann wrote:
> On Mon, Oct 26, 2020 at 5:55 PM Mark Rutland <mark.rutland@xxxxxxx> wrote:
>> On Mon, Oct 26, 2020 at 05:03:29PM +0100, Arnd Bergmann wrote:
>>> From: Arnd Bergmann <arnd@xxxxxxxx>
>>>
>>> When CONFIG_COMPAT_VDSO is disabled, we get a warning
>>> about a potential out-of-bounds access:
>>>
>>> arch/arm64/kernel/vdso.c: In function 'aarch32_vdso_mremap':
>>> arch/arm64/kernel/vdso.c:86:37: warning: array subscript 1 is above array bounds of 'struct vdso_abi_info[1]' [-Warray-bounds]
>>>    86 |  unsigned long vdso_size = vdso_info[abi].vdso_code_end -
>>>       |                            ~~~~~~~~~^~~~~
>>>
>>> This is all in dead code however that the compiler is unable to
>>> eliminate by itself.
>>>
>>> Change the array to individual local variables that can be
>>> dropped in dead code elimination to let the compiler understand
>>> this better.
>>>
>>> Fixes: 0cbc2659123e ("arm64: vdso32: Remove a bunch of #ifdef CONFIG_COMPAT_VDSO guards")
>>> Signed-off-by: Arnd Bergmann <arnd@xxxxxxxx>
>>
>> This looks like a nice cleanup to me! I agree we don't need the array
>> here.
>>
>> Reviewed-by: Mark Rutland <mark.rutland@xxxxxxx>
> 
> Thanks!
> 
> I see the patch now conflicts with "mm: forbid splitting special mappings"
> in -mm, by Dmitry Safonov. I have rebased my patch on top, should
> I send it to Andrew for inclusion in -mm then?

Makes sense to me.
I plan to add some more patches on top that will make tracking of user
landing (on vdso/sigpage/etc) common between architectures in generic code.
So, I think it's probably good idea to keep it in one place, -mm tree
seems like a proper place for it.

Thanks,
          Dmitry