[realtek-bug] Transmit time out on eth1 (rtl8139) / dirty entry in queue [again]

Axel Siebenwirth axel@rayfun.org
Wed, 16 May 2001 19:25:12 +0200 (CEST)


hi,

this old problem I had been faced with had been solved with 2.4.3-ac13/14,
but now with kernel 2.4.4-ac9 and all other 2.4.4-acx it came up again.
It's a Realtek 8139C chip on a AT2500 (allied telesyn or sumpin like that)

Instead the former

  Apr 24 16:16:57 bello kernel: eth1: Setting half-duplex based on
    auto-negotiated partner ability 0000.

I have now an unconnected cable which is not the fact:)

  May 16 15:20:26 bello kernel: eth1: media is unconnected, link down, or
    incompatible connection
  May 16 15:20:44 bello kernel: NETDEV WATCHDOG: eth1: transmit timed out
  May 16 15:20:44 bello kernel: eth1: Tx queue start entry 783  dirty
    entry 779.
  May 16 15:20:44 bello kernel: eth1:  Tx descriptor 0 is 00002000.
  May 16 15:20:44 bello kernel: eth1:  Tx descriptor 1 is 00002000.
  May 16 15:20:44 bello kernel: eth1:  Tx descriptor 2 is 00002000.
  May 16 15:20:44 bello kernel: eth1:  Tx descriptor 3 is 00002000. (queue
  head)
  May 16 15:20:44 bello kernel: eth1: media is unconnected, link down, or
    incompatible connection

Could I adjust some values, like tx buffer to get it to work?
Please excuse my unscientific nature.
I can as well supply any more information about the scenario...

Regards,
Axel Siebenwirth