[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [linux-tr] Resend: tr0: bad OPEN response: fe -Reply-Reply


> Well, i compiled the driver with #define OLYMPIC_DEBUG 1 and the error didn't 
> appear! Then i installed the new driver 0.1.9, installed a second card and the error > appeared exactly one time ... i  rebooted the machine several times and it didn't 
> reappear ... very strange! I hope this error doesn't reappear again.

Hmmm, this does not sound like a driver problem more like a hardware / memory setup / accessing  - bios self-configuring problem. (I run two ibm cards on the same interrupt no trouble, actually I also have a olicom card in there as well on the same int.)

>Is the current release of the driver suitable for a production machine? If no, what's
> missing? I would like to replace the ISA TokenRing Cards as soon as possible,
> they are making troubles. (see other message with "unrecoverable error: 
> errorcode  = 0059")

All that's missing from the driver for production purposes is some cleaned up error trapping and reporting.  There are probably a couple of error conditions that can arise on a token ring network that will upset the driver (but by then the ring has died anyway.)  So I think it is robust enough to use for production, you may want to increase the rx_ring number from 16 to a larger figure and you really want to adjust mtu sizes to match the rest of the network for maximum performance. If mtu > buffer size (default 4096) then more than one receive buffer will be used by the card and slower code must be used within the driver to get the packets to the upper layers. I like to leave some slack between mtu and buffer size and would probably set mtu to 4000.  

Also, the driver hasn't been sanitized for smp installations yet. (although it is only the open and closing of the adapter that is possibly smp un-safe).

There are other features that will make it into the driver before it is finalized (such as multicasting and more network analysis functions.)