Tulip.c->tulip_timer->check_duplex()

[Gideon Riddell]

Hello All,
>From working on embedded systems, we have established that the function
check_duplex() in the tulip driver needs improvement. Behaviour: Does
not communicate reliably. Problem: Mii Phy negociates full duplex,
driver fails to set 21143 MAC to full duplex, result: both NICs transmit
simulaneously while ignoring receive (half duplex behaviour) while phy
does not generate collision (full duplex behaviour) so both frames get
lost!

Datasheet spells out in appendix D how to set full duplex, and
stipulates in part 6 that both transmit and receive processes must be
stopped, and waited on till they stop before changing the port settings.
There is no indication of incorrect settings other than strange
behaviour.

Gideon.
--
Gideon Riddell
Storage Area Networks
Rustat House, 62 Clifton Road,
Cambridge, CB1 7EG
UK                          www.san.com
Tel:+44 1223 566111 Fax:+44 1223 566313


-------------------------------------------------------------------
To unsubscribe send a message body containing "unsubscribe"
to linux-tulip-bug-request@beowulf.org