> Paul Slootman <paul@wau.mis.ah.nl> wrote:
> : card (at 10Mb/s), and have *never* had any problems. This is with
> : heavy NFS and TCP traffic.
>
> : As far as I can see, the 0.31 version in the pre-patch -2 with the
> : various -jul29 etc. patches also works fine.
>
> I am glad to hear that since I have just bought a Dell PowerEdge server
> and given the option of choosing an Intel EtherExpress Pro 10/100 and
> a 3C905 I chose the 3C905. I'll be running at 10 MBits/sec.
>
> My question is: is it possible that all these people having problems
> with the 3C9065 driver are running at 100 MBits?
Some clarification: Those of us who have had dreadful trouble with 3c59x.c
driver are those with the 8k (I think) 3c590 boards, 10Mbits, with
possibly some hardware bug. Basically, under 0.25, 0.30-all, 0.40 and
0.41, things are fine until you hit a transmit timed out, then you're
shafted. Tx dies. The hastily hacked v0.31 3c59x.c driver in pre-2.0.31-2
seems _rock solid_ on these boards, but does not include the bus master
support for 3c905 cards.
The alternative is 0.42 which by all accounts sounds like it has the
stability _and_ the bus mastering capabilities..? Something needs to be
done about the 2.0.30 driver, at any rate.
Chris