Re: Linux 6.6-rc3 (DEBUG_VIRTUAL is unhappy on x86)

[Peter Zijlstra]

On Sun, Oct 01, 2023 at 11:17:21PM +0900, Hyeonggon Yoo wrote:

> OK, I looked into this a little bit, and it turns out that the problematic
> address here is from cleanup_trusted() in
> security/keys/trusted-keys/trusted_core.c.
> (and it's builtin due to CONFIG_TRUSTED_KEYS=y)

That code is insane.. wth would you want to use an explicit
static_call() in either __init or __exit ?!?

I think the reason we support init was because it was just really hard
to avoid throughout the abstraction layers etc.. But this seems to be
the only __exit user, and it is really quite daft.

> The function is marked as __exit, so it does not fall within the
> 'core kernel text address range,' which is between _stext and _etext
> (or between _sinittext and _einittext). and thus __text_poke() thinks that
> it's vmalloc/module area.
> 
> I think __text_poke() should be taught that functions marked as __exit
> also belong to kernel code just like __init.

Should we not do something like:

#ifdef MODULE
#define __exit __section(".exit.text") __exitused __cold notrace
#else
#define __exit __section(".discard.exit.text")
#endif

It's not like that code should ever be ran or referenced when built-in.