Re: [PATCH] mm/kasan: map KASAN zero page read only

[Ard Biesheuvel <ard.biesheuvel@linaro.org>]

On 7 January 2016 at 10:51, Catalin Marinas <catalin.marinas@arm.com> wrote:
> On Wed, Jan 06, 2016 at 09:18:03PM +0100, Ard Biesheuvel wrote:
>> On 6 January 2016 at 20:48, Andrey Ryabinin <ryabinin.a.a@gmail.com> wrote:
>> > 2016-01-06 18:54 GMT+03:00 Ard Biesheuvel <ard.biesheuvel@linaro.org>:
>> >> The original x86_64-only version of KASAN mapped its zero page
>> >> read-only, but this got lost when the code was generalised and
>> >> ported to arm64, since, at the time, the PAGE_KERNEL_RO define
>> >> did not exist. It has been added to arm64 in the mean time, so
>> >> let's use it.
>> >>
>> >
>> > Read-only wasn't lost. Just look at the next line:
>> >      zero_pte = pte_wrprotect(zero_pte);
>> >
>> > PAGE_KERNEL_RO is not available on all architectures, thus it would be better
>> > to not use it in generic code.
>>
>> OK, I didn't see that. For some reason, it is not working for me on
>> arm64, though.
>
> It's because the arm64 set_pte_at() doesn't bother checking for
> !PTE_WRITE to set PTE_RDONLY when mapping kernel pages. It works fine
> for user though. That's because usually all read-only kernel mappings
> already have PTE_RDONLY set via PAGE_KERNEL_RO.
>
> We may need to change the set_pte_at logic a bit to cover the above
> case.
>

Yes, that would be useful. I had an interesting dive down a rabbit
hole yesterday due to the fact that the kasan zero page (which backs a
substantial chunk of the shadow area) was getting written to by one
mapping, and reporting KAsan errors via another.

-- 
Ard.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>