Well, here is the stupid and interesting thing.
I've got more than 5 boxes here, all using the SAME realtek cards. Now if
it was some negotiation error between card and switch, then it, in theory,
should occur with all of them. Also, I mentioned that I had SWITCH cards
around, so a functioning card from a "stable debian" box was put into the
"unstable debian" box, and the same fallback to 10Mb occurred.
So I'm pretty sure that its a software issue. I was wondering if there
could be anything... ANYTHING software related that could force the card
into a fallback like that. Keep in mind that during bootup, and before
Debian loads up the network code, it is still in 100Mb mode. It ISN'T a
kernel problem either. The reason I say that is because if it was, then
when kernel loads the RT8139 (yes, i finally checked ;-) ) the fallback
should occur. However, it is STILL in 100Mb mode when it detects the
Realtek cards (2 of them, and yes, i've tried booting with only 1, and
switch those two around, put them in different PCI slots, etc.).
Could it, but some chance, maybe be the cards going into promiscuous mode
causing the card to fall back to 100Mb? I've never seen it happen... but
Heres an interesting thing I tried. The cards came with a software disk
(needed msdos to boot) that allowed me to switch between 10Mb, 100Mb, and
Auto-neg. The cards are set by default to Auto-neg, as they should be. I
FORCED it to 100Mb to see what would happen. Sure enough, I successfully
got it to stay at 100Mb, but then the switch automatically disabled the
port after around 5-10 minutes, and said it shut the port down due to
"conflict". Thats it. Thats all it said (oh how helpful). This is a Cisco
Please... ANY suggestions and help would be greatly appreciated.