Re: [PATCH] i386: fix stack alignment for signal handlers

[Markus F.X.J. Oberhumer]

Linus Torvalds wrote:
> On Wed, 14 Sep 2005, Markus F.X.J. Oberhumer wrote:
>
> > > You seem to be expecting that the address be aligned "before the return 
> > > address push", which is a totally different thing. Quite frankly, I don't 
> > > know which one gcc prefers or whether there's an ABI specifying any 
> > > preferences.
> >
> > I'm pretty sure that on both amd64 and i386 the alignment has to be 
> > _before_ the address push from the call, though I cannot find any exact ABI 
> > specs at the moment. Experts please advise.
> >
> > What do you get when running this slightly modified version of your test 
> > program? My patch would fix the alignment of Aligned16 here.
>
>
> Your test program does seems to imply that gcc wants the alignment before
> the return address (ie it prints out an address that is 4 bytes offset),
> but on the other hand I'm not even sure how careful gcc is about this
> alignment thing at all.
>
> In the "main()" function, gcc will actually generate a "andl $-16,%esp" to 
> force the alignment, but ot in the handler function. Just a gcc special 
> case? Random luck?

I think that main() is a known name and therefore gets a special treatment 
- if you rename main() to foo() and then compare the disassembly you will 
see that the "andl $-16,%esp" has vanished.

OTOS the "andl" in main() exactly does show how gcc wants the stack to be 
aligned, i.e. _before_ the call-address push.

Another argument would be the 16-byte aligned stack-setup of glibc - please 
try runing this tiny program under gdb and look at "info reg":

    asm(".globl main\n main:\n int $3\n");

All of this would indicate that the kernel should get fixed.

~Markus

> Andi - you know the gcc people, is there some documented rules somewhere? 
> How does gcc itself try to align the stack when it generates the calls?
>
> 		Linus

--
Markus Oberhumer, <markus@xxxxxxxxxxxxx>, http://www.oberhumer.com/
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/