[uClibc]2.5.30-rmk1-pxa1 uClibc failures

Kyle Harris kharris at nexus-tech.net
Wed Sep 18 01:14:35 UTC 2002


On Tuesday 17 September 2002 01:38 pm, Erik Andersen wrote:

> On Tue Sep 17, 2002 at 01:09:31PM -0400, Kyle Harris wrote:
> > On Tuesday 17 September 2002 01:07 pm, Erik Andersen wrote:
> > > On Tue Sep 17, 2002 at 12:36:43PM -0400, Kyle Harris wrote:
> > > > With help from Russell (thank you), I have determined that my init
> > > > failures are related to running shared code from uClibc-0.9.15. I can
> > > > statically build a simple hello program and it runs fine as init. If
> > > > I rebuild dynamic, it fails (dump at end of message). Is anyone
> > > > running with uClibc and 2.5 on any arm platforms? Or know of any
> > > > issues?
> > >
> > > I have not tried uClibc with a 2.5 kernel on any platform.  I use
> > > uClibc on arm with 2.4 kernels all the time and it works just fine
> > > for me...
> >
> > Yes, it works fine with 2.4 for me also. Just not with 2.5 :(
>
> If shared libraries are broken, but static linking works, then I
> will bet you $1 US that the kernel binfmt_elf has changed in 2.5
> and the stack layout is now different.  Can you try the attached
> program and let me know what you see?   This is a mini version of
> what I do in the uClibc shared lib loader.  Does removing the arm
> specific 'stack += 1' make a difference?

Oh yeh, I left out one important fact. uClibc shared works fine with 2.5 when 
using an NFS root. Which I think indicates a problem in the kernel. Your 
stack_test works with stack += 1. Removing this results in a seg fault.

Kyle.



More information about the uClibc mailing list