Hi, On Fri, Jan 14, 2000 at 10:16:30AM -0600, Nathan E Norman wrote: > On Fri, 14 Jan 2000, Dean Allen Provins wrote: > > : System is Debian 2.1 (up to date) on Pentium, 64 MB, with NIC (from > : dmesg), > : > : eth0: PCnet/ISA 79C960 at 0x340, 00 00 f4 b1 84 e7, probed IRQ 12, DMA 7. > : lance.c:v1.14ac 1998/11/20 [EMAIL PROTECTED], [EMAIL PROTECTED] > : > : The error message that appears in 'dmesg' at various times after > : booting is: > : > : eth0: transmit timed out, status 0062, resetting. > : Ring data dump: dirty_tx 118 cur_tx 134 (full) cur_rx 58. > > I think it's a driver problem - I saw this quite a bit on one of our > servers with onboard Lance. I installed a NetGear tulip card and > haven't seen the problem since :) I do have to admit that I never > actually noticed a "problem" other than the error message, but since it > was a server I wasn't taking any chances. >
Yes, I also think it might be a driver problem. I have a 3c507 that worked fine on an old motherboard but gave 'timed out' errors on a faster k6 system. As I understand the problem, (but may be completely wrong of course) the skeleton NIC driver had a common problem that allowed the Tx ring buffer to be overwritten before the transmit status of each sent packet was checked (reaping). This had something to do with upper network layers not honouring the NIC's dev->tbusy bit. If you feel like mucking around in the driver source, I can send details of what I did to get the 3c507 driver working (ugly hack, but it fixes the problem). HTH Mark.
