Compiling Fiasco UX problems under Fedora 8

[Adam Lackorzynski]

On Wed Mar 12, 2008 at 13:02:46 +0100, Adam Lackorzynski wrote:
> 
> On Mon Mar 10, 2008 at 18:16:41 +0100, Olivier.Landemarre at utbm.fr wrote:
> > >
> > > This seems to be very specific to FC8 as this doesn't happen to me and
> > > to our automatic builds. Can you discover what causes the segfaults?
> > > You can also send me one of the binaries that segfault to analyze.
> > 
> > with gdb I have this informations running fiasco:
> > 
> > Program received signal SIGSEGV, Segmentation fault.
> > 0x00039bee in __libc_setup_tls ()
> >     at /home/olivier/l4/kernel/fiasco/src/drivers/ux/processor-ux.cpp:47
> > 47        asm volatile (" .byte 0xf3, 0x90 #pause \n" );
> > 
> > 
> > I have currently an AMD Athlon processor
> > 
> > If I try with mapdb_t, I have this
> > 
> > Program received signal SIGSEGV, Segmentation fault.
> > 0x00043b8e in __libc_setup_tls ()
> >     at /home/olivier/l4/kernel/fiasco/src/kern/kmem_alloc.cpp:74
> > 74        assert(o>=8 /*NEW INTERFACE PARANIOIA*/);
> > 
> > 
> > I can send binary
> 
> I've looked a bit into this and it's a pointer deref that goes wrong
> (AFAICS) inside the glibc code. I'd guess a FC8 system is now needed...

I can reproduce this in FC8 and _dl_phdr seems to be initialized with
the wrong value very early in the glibc startup. I don't know why. The
next step would probably be to recompile glibc with some more debugging
code added and see what's happening.



Adam
-- 
Adam                 adam at os.inf.tu-dresden.de
  Lackorzynski         http://os.inf.tu-dresden.de/~adam/