Subject: Re: 8k
To: None <glass@sun-lamp.cs.berkeley.edu, mw@eunet.ch>
From: Mike Hibler <mike@cs.utah.edu>
List: m68k
Date: 02/14/1994 17:12:25
> From: mw@eunet.ch
> Subject: Re: 8k
> To: glass@sun-lamp.cs.berkeley.edu (Adam Glass)
> Date: Tue, 15 Feb 1994 00:58:43 +0100 (MET)
> Cc: m68k@sun-lamp.cs.berkeley.edu
> 
> > Before I setup a cross-compileation environment to compile 1-2 files:
> > 
> > Anybody have an environment that builds MID_68K (i.e m68k NetBSD
> > binary with 8k page sizes)...
> 
> Eh, sure:-) I'll upload new binary dist for amiga RSN, hopefully this
> week still. You should be able to use those binaries easily (even with
> gcc-2.5.8, not the slightly antique gcc-2.4.5 in the bmake'd tree). I'll
> tell you when I upload them to ftp.eunet.ch (bin-site for amiga NetBSD).
> 
We found one nasty bug in 2.5.8 which made our 4.3 68k kernels crash.  It
had to do with using negative offsets from the beginning of a structure.
As an example:

	struct foo {
		short foo1;
		...
	}

a read of foo.foo1 would generate a longword memory reference at "reg@(-2)"
in some cases (e.g. where it is known that the high 2 bytes will be masked
off later).  However, if your structure is the first thing on a page, and
the previous page is not accessible...

Jeff Law can post the official fix for people.

------------------------------------------------------------------------------