Re: NetBSD on Sun T5220

Palle Lyckegaard <[email protected]> Fri, 10 Oct 2025 18:56:26 +0000 (UTC)
Newsgroups gmane.os.netbsd.ports.sparc64
Message-ID <[email protected]>
On Thu, 9 Oct 2025, Sad Clouds wrote:

>
> This is what I get with -current /bin/sh and your patch:
>
> ...
> [  10.2521573] init path (default /sbin/init):
> [  10.2827453] init: copying out flags `-s' 3
> [  10.2827453] init: copying out path `/sbin/init' 11
> [  10.2827453] entropy: best effort
> Enter pathname of shell or RETURN for /bin/sh:
> [  50.1619768] Alignment error: pid=31.31 comm=sh pc=4099a12c
> [  50.1619768] dfa = 00000000e0002000
> [  50.1619768] dfc = 0000000000000000
> [  50.1619768] pid 31 (sh), uid 0: exited on signal 10 (core not
> dumped, err = 30)
>
> It doesn't get as far as a single user mode. I think the patch made
> things a bit worse, since the kernel now hangs and doesn't even drop
> into debugger.
>
> If the issue is confined to the kernel then the older /bin/sh
> binary should not make any difference as it still complies with V9 ABI
> and should be backward compatible with NetBSD system libraries.
>
> It's a bit easier to debug in single user mode.
>

Thank you for testing - you are right that the provided patch is not 
solving the issue - I have something that looks like a regression with 
my -current based setup, so I'm back at looking at the code in locore.s 
again... :-)