[vortex] old problem? with vortex driver

David Dunlop djd@WPI.EDU
Fri Mar 22 12:24:05 2002


Donald Becker wrote:
> On Fri, 22 Mar 2002, David Dunlop wrote:
> 
> 
>>I have a problem simuler to the one in these archives:
>>http://www.tux.org/hypermail/linux-vortex-bug/1998-Jan/0002.html
>>http://www.uwsg.iu.edu/hypermail/linux/net/0005.2/0171.html
>>
>>When I Download at High speeds I get this into my logs:
>>Mar 21 20:49:55 zaphod kernel: eth0: Transmit error, Tx status register 90.
> 
> 
> Transmit underrun.  This is rare.
> 
> 
>>[root@zaphod log]# dmesg |grep 3c
>>3c59x: Donald Becker and others. www.scyld.com/network/vortex.html
>>02:0c.0: 3Com PCI 3c590 Vortex 10Mbps at 0xdf20. Vers LK1.1.16
> 
> 
> Two interesting details: this is a really old 3c590, and it's 10Mbps
> only.
> 
> Is there anything unusual about your system?

I don't think there is anything unusual about my system.
It's a 1GHz Dell 256M RDRAMM w/ECC
PCI cards are:
Soundblaster Live
Promise Ultra 100 TX2 IDE card
Adaptech SCSI card

I should also node that when i am getting these i am writting through 
the Promise card


> 
> My driver releases default to programmed-I/O for the 3c590, and thus can not
> encounter this error.  I'm not certain what the LK releases default to,
> but at 10Mbps you should never get an underrun..
> 
> The user can set an option to use PCI bus mastering with my driver, but
> the 3c590 series only had the hardware for weak single-linear-region
> transfers.  Much like ISA DMA, setting up a short transfer takes more
> time than just pushing the data out with PIO.  Using PIO was fine when the
> 3c590 was new and most PCI implementations had problems with bus
> masters, but today you will get far better performance from any $10 10/100
> NIC.

I've slated the card to be replaced some time

DD