Re: cpu error 4 (was: Re: O2 framebuffer problems with install kernel)
Jorge Acereda Maciá <[email protected]>
| Newsgroups | gmane.os.netbsd.ports.sgimips |
|---|---|
| Message-ID | <[email protected]> |
Hi, On Jul 29, 2008, at 8:57 PM, Michael Lorenz wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Hello, > > On Jul 29, 2008, at 12:15, Jorge Acereda Maciá wrote: > >> On Jul 29, 2008, at 5:42 PM, Michael Lorenz wrote: >> >>> On Jul 29, 2008, at 11:23, Jorge Acereda Maciá wrote: >>> >>>> With the first patch (the one using BUS_DMA_COHERENT), I get >>>> constant reboots, usually under heavy IO load (probably an >>>> unrelated problem). >>> >>> Lots of network traffic on the onboard interface? >> >> Yes, I'll try to isolate if it's due to network IO or disk IO. > > My bet is on mec - even if it works right it's not very fast and > doesn't do stuff like checksums in hw. If you've got a PCI ethernet > board sitting around - I can personally vouch for 3com 3c905 based > cards to work just fine, others used DEC 2114x clones. I'm right now building a kernel on the O2 and installing some pre- fetched stuff from pkgsrc at the same time. It has been running for quite a while and no hangs so far, so you're probably right. After that I'll try to do some heavy network IO. > In theory anything supported by NetBSD should work but since the O2 > is a little more ticklish than - say, macppc and sparc - regarding > DMA and cache synchronization ( macppc for instance is always DMA > coherent ) I'd expect occasional trouble. > >>> Any change with the 2nd patch? >> >> Well, I wanted to keep that machine self-hosting and so far I've >> been unable to checkout the kernel sources :-) > > Eh, you won't be impressed by build times. Indeed. My 68030@50 Amiga compiled faster (well, probably the reason is that gcc2 is much faster than gcc4)