On Fri, 2005-08-12 at 16:06 +0200, Magnus wrote: > > David Hollis <[EMAIL PROTECTED]> wrote: > > There were some fixes incorporated not too long ago that deal with the > > speed negotiation, but those should already be incorporated into the > > kernel you are running. It is possible that there are endian issues but > > I don't have any non-x86 hardware to test on unfortunately. I'm > > currently looking into making use of the PHY abstraction layer that was > > recently submitted and that may help with some of these issues. Do you > > know if your device ever worked properly? Could you possibly try it > > with Windows and see if it works properly? The main issues I've seen > > with the speed handling has been going from 100-full to 10-half and that > > sort of thing. If you auto to 100-full and stay there, the device works > > just fine. > > -- > > David Hollis <[EMAIL PROTECTED]> > > I just tried the device on Windows XP (driver v1.30 from dlink.com), works > like a charm (both in auto-sense and forced modes). My switches did > negotiate 100FDX, throughput was about 7-8mb/s. > When I connect it to the Mac the switch goes to HDX, as indictated by the > switch FDX led and port info. > > So this is definately a bug, possibly only on PPC (Don't have any x86 to > test on).
Interesting, I wonder if anyone else is using these devices on other PPC or like systems? I suppose they may be trying to use them but not having much success and just not reporting it. I wouldn't be too surprised if it's an endian issue, though the PHY handling has been a bit suspect as of late. Hopefully the work I'm doing now will get the PHYs working much more predictably. -- David Hollis <[EMAIL PROTECTED]>
signature.asc
Description: This is a digitally signed message part
