Re: [PATCH v7 4/4] powerpc: Book3S 64-bit "heavyweight" KASAN support

[Daniel Axtens]

Christophe Leroy <christophe.leroy@xxxxxx> writes:

> Le 13/02/2020 Ã 01:47, Daniel Axtens a ÃcritÂ:
>> diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
>> index 497b7d0b2d7e..f1c54c08a88e 100644
>> --- a/arch/powerpc/Kconfig
>> +++ b/arch/powerpc/Kconfig
>> @@ -169,7 +169,9 @@ config PPC
>>   	select HAVE_ARCH_HUGE_VMAP		if PPC_BOOK3S_64 && PPC_RADIX_MMU
>>   	select HAVE_ARCH_JUMP_LABEL
>>   	select HAVE_ARCH_KASAN			if PPC32
>> +	select HAVE_ARCH_KASAN			if PPC_BOOK3S_64 && PPC_RADIX_MMU
>
> That's probably detail, but as it is necessary to deeply define the HW 
> when selecting that (I mean giving the exact amount of memory and with 
> restrictions like having a wholeblock memory), should it also depend of 
> EXPERT ?

If it weren't a debug feature I would definitely agree with you, but I
think being a debug feature it's not so necessary. Also anything with
more memory than the config option specifies will still boot - it's just
less memory that won't boot. I've set the default to 1024MB: I know
that's a lot of memory for an embedded system but I think for anything
with the Radix MMU it's an OK default.

I'm sure if mpe disagrees he can add EXPERT when he's merging :)

> Maybe we could have
>
> -  	select HAVE_ARCH_KASAN_VMALLOC		if PPC32
> +	select HAVE_ARCH_KASAN_VMALLOC		if HAVE_ARCH_KASAN

That's a good idea. Done.

Thanks for the review!

Regards,
Daniel

>
>>   	select HAVE_ARCH_KGDB
>>   	select HAVE_ARCH_MMAP_RND_BITS
>>   	select HAVE_ARCH_MMAP_RND_COMPAT_BITS	if COMPAT
>
>
> Christophe