busybox segfault in __uClibc_main

[Joakim Tjernlund]

> 
> On Sun, 11 Dec 2005, Joakim Tjernlund wrote:
> 
> >  
> > > 
> > > On Sun, 11 Dec 2005, Joakim Tjernlund wrote:
> > > 
> > > >  >
> > > > > Joakim Tjernlund <joakim.tjernlund at transmode.se> wrote:
> > > > > >
> > > > > > Curious, does replacing r7 with r9 in
> > > > > > 	lwz	r7,_NIP(r1)
> > > > > > 	lwz	r8,_MSR(r1)
> > > > > > 	FIX_SRR1(r8, r0)
> > > > > > 	lwz	r2,GPR2(r1)
> > > > > > 	lwz	r1,GPR1(r1)
> > > > > > 	mtspr	SPRN_SRR0,r7
> > > > > > 	mtspr	SPRN_SRR1,r8
> > > > > > 	SYNC
> > > > > > 	RFI
> > > > > > work for you?
> > > > > 
> > > > > I'm at home atm, but I'm damned sure it would work.
> > > > 
> > > > hmm, one could also fix it in __uClibc_main() by addin 
> > > #ifdef SHARED 
> > > > .. #endif, not tested
> > > > 
> > > > #ifdef SHARED
> > > >   __rtld_fini = rtld_fini;
> > > > #endif
> > > > 
> > > > will probably only work on current svn, not on .28
> > > 
> > > can this somehow be checked on other arch as well?
> > 
> > Don't quite understand, the above code covers all archs. What else
> > needs to be checked?
> 
> I have thought that the failure exhibits only on ppc

As far as I know, yes. You mean that that the above #ifdef SHARED should
depend and PPC as well?
Maybe, I am not sure yet.

> 
> > 
> > > I am sure that __uClibc_main "offers" more places that 
> can be guarded w/ SHARED
> > 
> > Well, one could eliminate __rtld_fini completely.
> > 
> > Can't rember now, are there 2 versions of crt1, one for 
> shared and one for static that
> > lives side by side? if so one can force rtld_fini to zero 
> in the static one.
> 
> no, there is only one crt1.o both for static and for "normal" 
> binaries, 
> the other one (Scrt1.o) is only for ET_DYN/PIE.

OK, thanks.

> 
> Peter