On Thu, Dec 18, 2003 at 11:19:29AM -0800, Andrey Petrov wrote:
> On Thu, Dec 18, 2003 at 05:45:28PM +0100, R. Braun wrote:
> > Hi,
> > 
> > One of the users of my sparc64 shell server has tested some sparc64 assembly
> > code. He created a process which keeps running in RE+ state :
> > 
> >  9138 p2- RE+   6:03.52 (h)
> > 
> > Later, the process moved to the RNE state (nobody niced it) :
> > 
> >  9138 p2- RNE  13:03.16 (h)
> > 
> > here is the source code :
> > 
> > .global main
> > main:
> >  sub %sp,4,%sp
> >  mov 42,%o0    // exit(42)
> >  mov 1,%g1     // 1 = SYS_exit
> >  ta  9         // get the syscall
> > 
> > I guess the problem comes from the sub %sp,4,%sp line. Though it is a
> > programming error (in this case it was used to see how the machine would handle
> > such a programming error), is there no way to prevent this ? And is it related
> > to the architecture or can NetBSD handle it ?
> > 
> 
> It should but it not always does it properly. Thanks for test.
> And yes, it is sparc specific problem due to register windows lazy save
> operations.
> 

And -current DEBUG kernel handles it correctly. *sigh* unfortunately.

	Andrey