Re: 3c590 _again_

Chris Evans (
Thu, 27 Feb 1997 21:54:11 +0000 (GMT)

On Thu, 27 Feb 1997, Alan Cox wrote:

[re: 3c590]

> [Stuff about packet sizes then]
> "This has the side effect of scaling all threshold registers by a factor
> of 4 when PIO code is running on PCI/EISA bus master adapters"
> Its right

Ah well. Here's another random thought; if some of the 3c590 problems are
caused by faulty interrupt generation on certain 8k 3c590 boards, do these
problems persist if you force the board into bus master mode? By default
the board is always run on PIO mode since bus mastering is only a win with
very large packets.

And have 3Com admitted a hardware bug?? If so they can bloody well replace
the board in our server :-)

Incidentally, I am running on a patched 3c590 driver which now resets the
card and sets dev->tbusy to 0 if a "transmitter access conflict" is
detected. I also printk() a suitable message. In about 2 weeks, I have had
two messages to say "your network was about to disappear, but I've saved
the day". For the time being I'll live with that!