RE: [uml-devel] [PATCH] PPC 2.4.13-um1 problems...

[McMechan, J. <McM...@na...>]

Your *add = 1; SIGSEGV looks very suspicious *add is not assigned in any
of the other __uml_setup function calls and so may be doing bad things.

at the moment it might be better to leave the *add = 1 out and let jail
show up in the command line since it is already 1 from linux_main()?
and will show up anyway now, I think it was supposed to hide it by *add = 0;

also do you have a compiled version of the PPC port available.  I still run
MkLinux on PPC and have not had much luck compiling under
host 2.0.36/egcs 1.1.2

> From: Rusty Russell <ru...@ru...>
> To: use...@li...
> Date: Sun, 28 Oct 2001 15:16:33 +1100
> Subject: [uml-devel] [PATCH] PPC 2.4.13-um1 problems...
> 
> Hi all,
> 
> 	Jeff, found the following problems (also in 2.4.12 patches):
> 
> 1) the new system calls are not defined for all architectures.  See
>    patch.
> 
> 2) The two #defines of pt_regs (ptrace.h: pt_regs_subarch,
>     sigcontext-ppc.h: sys_pt_regs) make for lots of warnings, but I'm
>     not sure if it is fatal...
> 
> 3) The #ifdef CONFIG_SMP in arch/um/kernel/trap_user.c is testing the
>    compiling hosts' configuration (ie. /usr/include/linux/autoconf.h),
>    not the UML kernel configuration.  I just #if 0'd it out in the
>    patch.
> 
> OK, now it compiles, and I get a segv on return from an __init
> function (according to GDB):
> 
> Starting program: /home/rusty/devel/kernel/working-2.4.13-uml-proc/linux 
> 
> Program received signal SIGSEGV, Segmentation fault.
> 0xa000b000 in _start () at process_kern.c:754
> (gdb) 
> 
> That is the return line of the following function:
> 	static int __init jail_setup(char *line, int *add)
> 	{
> 		jail = 1;
> 		*add = 1;
> 		return(0);
> 	}
> 
> Any clues?
> Rusty.
> --
> Premature optmztion is rt of all evl. --DK
>