On Thu, 26 Feb 1998, Mark Brinicombe wrote:
> The failed bit problem is one I need to look into. It appears that the
> interface does not turn off during the packet buffer test and thus the
> test can be trashed if a packet is received during the test. The more
> network activity present the more likely this is to occur during driver
> initialisation. This may be a peculiarity of the chipset being used but I
> should have more information soon.

It is on a busy network. ISTR that the RISC OS driver has the same
problem.

> Alternatively the packet buffer testing (MI) needs to be modified so that
> retries can occur or the driver can continue to attach even if the memory
> did not complete test (i.e. possibly with a smaller buffer in case it is a
> real memory problem).

Perhaps with a flag akin to ehbug. It seems the only reliable card is the
EtherB (or Ether3 if you've IRQ modded it) and even that spews timeout
messages on quiet networks.

> The round trip time of >1000ms is one that I have observed occasionally
> but don't have a means to recreated it. I am not sure what is happening
> there but just doing
> 
> ifconfig em0 down
> ifconfig em0 up
> 
> should fix the problem.
> Do you have a reliable way of triggering this behaviour ?

Unfortunately not. I think it's something to do with not having the
network plugged in at boot up and then attaching it. 

Stephen Borrill