Re: [PATCH 1/3] module: deal with alignment issues in built-in module versions

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

On Thu, 2011-02-17 at 15:08 -0800, Linus Torvalds wrote:
> On Thu, Feb 17, 2011 at 2:48 PM, Linus Torvalds
> <torvalds@linux-foundation.org> wrote:
> >
> > Gaah, I didn't realize that we've never had to do anything like this
> > before, and that the exception table is all arch-specific code. So we
> > don't have any way to "output that damned pointer and stop whining
> > about it" model at all.
> 
> Actually, I don't think the problem is about ".word" vs ".xword". We
> should be able to just use ".long" everywhere.
> 
> But the symbol _name_ may have different prefixes, and when we use
> "asm()" at the top level, we can't use the expressions to fix it up.
> So a
> 
>   asm(".long %0":'i" (symbol))
> 
> doesn't work (ignore the lack of section naming, that's not
> important), and neither can we just do something like
> 
>   #define output_asm_pointer(section, symbol) \
>        asm(".long " #symbol)
> 
> portably, because some linker formats want to see prepended underscores etc.

Except that .long is 32-bit on ppc64 :-( You need .llong for 64-bit.

Cheers,
Ben.