Subject: Re: panic: cache error -- on R5k INDY
To: Christopher SEKIYA <wileyc@rezrov.net>
From: sgimips NetBSD list <sgimips@mrynet.com>
List: port-sgimips
Date: 04/24/2004 20:00:12
On Sun, Apr 25, 2004 at  9:30am, Christopher SEKIYA wrote:
> Subject: Re: panic: cache error -- on R5k INDY

Update:

I've just upgrade the root to a recent (a few days ago) userland from -current
(which includes the libc.so.12.116 and respective kernel).  This was a build
done a few days ago...

All seems to be working with this .116 libc and kernel.

I'm performing a full build of -current now and will perform a clean install
to ensure the current MAIN branch is good after the switch to .116 libc.
Regardless, all sgimips people should avoid builds from main and netbsd-2-0
which use the .114 libc.  I think this pretty much covers everything from
2004/03/25 to 2004/04/21 at least.

I'll report back on the current MAIN branch status on this in a few hours
when the build and install complete.

-scott

> (port-mips added to distribution list, as we're now in MI land)
> 
> On Sat, Apr 24, 2004 at 06:01:22PM -0500, sgimips NetBSD list wrote:
> 
> > Moved things around, and libc.so.12.114 is the culprit.  libc.so.12.114
> > and the currently-built libc.so.12.116 both cause the problem.
> > When libc.so.12.113 (which is pre-2.0) is placed in /lib (and all /lib
> > and /usr/lib symlinks to it are adjusted) all seems fine.
> 
> That narrows it down to a MIPS-specific change in libc committed between
> 2004/03/04 23:43:56 and 2004/03/25 00:02:46 (the former date marking .113,
> the latter .114).  Maybe the sigtrampoline code?
> 
> (to bring port-mips up to speed, 2.0 and -current experience bus errors and
> cache errors at random intervals with libc.so.12.114 and newer.  libc.so.12.113
> is fine, resulting in a stable system.  This is going to be a showstopper for
> sgimips for 2.0 :( )
> 
> -- Chris
> 	GPG key FEB9DE7F (91AF 4534 4529 4BCC 31A5  938E 023E EEFB FEB9 DE7F)
>-- End of excerpt from Christopher SEKIYA