[RFC] UML/x86_64 module loader

[RFC] UML/x86_64 module loader

[Richard Weinberger]

Hi!

UserModeLinux is using the arch specific module functions from 
arch/x86/kernel/module.c.
On i386 this works perfectly fine but on x86_64 it causes problems.

apply_relocate_add() assumes modules compiled with -mcmodel=kernel 
because the kernel lives in the negative 2GiB of the address space.
This assumption is not true for UML.
On an UML instance with more than 512MiB of memory no modules can be 
loaded because vmalloc() locates the module near the 2GiB limit and the
ELF relocations causes an overflow. (Detected by "if ((s64)val != *(s32 
*)loc)" in apply_relocate_add()).

Now I'm not sure how to fix this.
Mostly because I'm not a module loader nor an ELF expert. 8-)

Is there an easy way to adjust the current x86_64 module loader to deal 
with UML "positive" relocations?

Thanks,
//richard

P.s: Just removing the overflow-check does not fix the problem. ;-)

Re: [RFC] UML/x86_64 module loader

[Rusty Russell]

On Mon, 22 Aug 2011 23:38:43 +0200, Richard Weinberger <richard@nod.at> wrote:
> Hi!
> 
> UserModeLinux is using the arch specific module functions from 
> arch/x86/kernel/module.c.
> On i386 this works perfectly fine but on x86_64 it causes problems.
> 
> apply_relocate_add() assumes modules compiled with -mcmodel=kernel 
> because the kernel lives in the negative 2GiB of the address space.
> This assumption is not true for UML.
> On an UML instance with more than 512MiB of memory no modules can be 
> loaded because vmalloc() locates the module near the 2GiB limit and the
> ELF relocations causes an overflow. (Detected by "if ((s64)val != *(s32 
> *)loc)" in apply_relocate_add()).
> 
> Now I'm not sure how to fix this.
> Mostly because I'm not a module loader nor an ELF expert. 8-)

I think you need to write your own routines.  It shouldn't be that hard,
just keep implementing relocations until you're done :)

Cheers,
Rusty.