Re: [PATCH RESEND] mm, kasan: don't poison boot memory

[Andrey Konovalov]

On Thu, Feb 18, 2021 at 11:46 AM Catalin Marinas
<catalin.marinas@xxxxxxx> wrote:
>
> The approach looks fine to me. If you don't like the trade-off, I think
> you could still leave the kasan poisoning in if CONFIG_DEBUG_KERNEL.

This won't work, Android enables CONFIG_DEBUG_KERNEL in GKI as it turns out :)

> Reviewed-by: Catalin Marinas <catalin.marinas@xxxxxxx>
>
> Just curious, have you noticed any issue booting a KASAN_SW_TAGS-enabled
> kernel on a system with sufficiently large RAM? Is the boot slow-down
> significant?

When booting KASAN_SW_TAGS in QEMU with 40G there's a noticeable
start-up delay compared to 2G, but it doesn't seem to be caused by
this memblock->page_alloc poisoning, as removing it makes no
noticeable difference.

I also don't see a noticeable "hang" when booting KASAN_SW_TAGS in
FVP, compared to the one I see with KASAN_HW_TAGS. But I do see a
"hang" in QEMU when going from 2G to 40G with KASAN_HW_TAGS.

It seems that doing STG is much more expensive than writing to the
shadow memory.

> For MTE, we could look at optimising the poisoning code for page size to
> use STGM or DC GZVA but I don't think we can make it unnoticeable for
> large systems (especially with DC GZVA, that's like zeroing the whole
> RAM at boot).

https://bugzilla.kernel.org/show_bug.cgi?id=211817