> On Sat, 31 Aug 1996, 0xdeadbeef wrote:
>
> > It appears that 2.0.16 fixes many bugs, but creates a few new ones of its
> > own. Also, I see that the latest 3c59x.c (from Donald Becker, who else :-)
> > has not yet been integrated into the kernel. The latest version fixes
> > crashes that would otherwise occur with my card (a very recent 3c590TPO
> > PCI, mfg. aug 1996). The latest versions of the drivers have been
> > extremely stable under heavy load (about 300-400packets/sec, mainly
> > because RPI Dormnet is getting netbios/netbeui broadcast storms for badly
> > configured win95 machines). Both boomerang.c (with support for 3c90x
> > cards as well as 3c59x) and 3c59x.c off of the FTP site work without
> > problems and really should be included in the latest kernels before 2.1
> > comes out, as we really don't want bad drivers for these cards sticking
> > around forever in 2.0.x. Also, 3c90x support would not be a bad idea,
> > especially since the cards seem to be far less brain-damaged (and faster)
> > than the 3c59x cards, as well as cheaper.
>
> boomerang.c (v.28c) didnt work with my 3c590 with 8k buffer. Please dont
> incorporate it into the kernel before it has been made to work so we wont
> have the problems we had when .25 was implemented before it was tested
> alright.
>
>
My 3c590 also claims to have the 8k buffer, but has been rock solid under
both 3c59x.c (0.28c) and boomerang.c (0.31, 0.32, and 0.33). It appears
that there is a pretty large variation between the 3com boards of the
same model. Let's just make sure that the final 2.0.x release has a 3c59x
drive that's stable for all of the versions of the card.
Simon Karpen
karpes@rpi.edu