At 08:42 PM 10/12/2002, Izumi Tsutsui wrote:
>In article <200210121856.g9CIuxr13201@unixag-kl.fh-kl.de>
>jkunz@unixag-kl.fh-kl.de wrote:
>
> > I mounted a DEFPA FDDI card in my IBM RS/6000 43P-100 (PPC/prep machine)
> > runing NetBSD 1.6. At the moment when the first packet is transmitted
> > the machine hangs. No panic, no core dump, absolutely nothing. It just
> > freezes. Someone warnd me about that already some time ago. I decided
> > to give it a try in the hope that 1.6 does not suffer from this bug.
> > But no... :-(
> >
> > This is a "Should Never Happen" bug I will try to hunt. Any sugestions
> > where and how to start?
>
>I guess this is endianness problem. Maybe we should add more
>htole32() and le32toh() functions where the driver passes
>TX/RX descriptors via DMA, but pdq.c uses some bit field
>definitions in the descriptors, so it would be a bit complicated
>to fix them. sys/dev/ic/i82557.c (and other devices listed in
>macppc/conf/GENERIC) would help to see how to handle them.

That might work but the driver is telling the card to swap
endianness for DMA so that would confuse it even more.

I'll try to look at it in a week or two (in a sparc64 machine).

-- 
Matt Thomas               Internet:   matt@3am-software.com
3am Software Foundry      WWW URL:    http://www.3am-software.com/bio/matt/
Cupertino, CA             Disclaimer: I avow all knowledge of this message