Don't give errors about CONFIG_PAGE_POISONING when using an alternative about kernel-hardening-checker

[anthraxx]

I would love this :P

[a13xp0p0v]

As I remember, all these features are different in some sense.
Are you sure that they are alternative to each other?

[madaidan]

As far as I know, they all have the same goal which is to overwrite memory to prevent use-after-free but they have some slight differences as PAGE_POISONING forces debugging bloat (as it is actually a debugging feature) which makes init_on_{,free,alloc} or PAGE_SANITIZE (which was dropped in newer linux-hardened versions for init_on_{,free,alloc}) better.

init_on_{,free,alloc} actually disables itself when PAGE_POISONING is being used to prevent conflict.

torvalds/linux@6471384

If either SLUB poisoning or page poisoning is enabled, those options take
precedence over init_on_alloc and init_on_free: initialization is only
applied to unpoisoned allocations.

Also notice that linux-hardened and ClipOS do not enable PAGE_POISONING but use the others instead.

[a13xp0p0v]

@madaidan, thanks for the details.
So yes, PAGE_POISONING is a debugging feature.
It provides less erasing than INIT_ON_FREE_DEFAULT_ON.

I joined these checks with OR giving preference to INIT_ON_FREE_DEFAULT_ON.
Please see the linked commit.

[madaidan]

Great, thanks.