Greetings, all, As it turned out, the age of my kernel (2.4.14-pre3) was the source of the problem; I had the older pci allocation functions that don't allocate non-cacheable memory. When I updated the kernel to use a more recent pci-dma.c, the eepro100.c driver I'd been using worked fine. (It's not identical to the one in the latest distribution, but it's not far off.)
Thanks, -- David Wright, InfiniSwitch Corp. > -----Original Message----- > On Tue, Apr 16, 2002 at 09:08:41PM -0400, Wright, David wrote: > > > > I've been trying to get an Intel eepro100 card working on a > > Walnut board. I've tried both the 1.09 and 1.19 versions of > > the driver (eepro100.c). Initially, both failed the self- > > test step. I was able to get past that one by invalidating > > the dcache after issuing the self-test directive. However, > > I now find that the card is neither transmitting nor receiving > > reliably. I get messages like: > > > > eth1: Unknown receiver error, status=0x5048. [lots of these] > > > > and > > > > eth1: Transmit timed out: status 0050 0000 at 2/8 commands > 00a00300 00a003000. > > eth1: Restarting the chip... > > eth1: Command unit failed to mark command 00030000 as complete at 6. > > > > > > Any suggestions (other than junking this card, which is not > > an option for me) most welcome. > > Sounds like you have a buggy kernel since it looks like cache > coherency problems. By the revision of the eepro100 driver you > are using, it sounds ancient too. > > Try running the 2_4_devel kernel. The current eepro100 driver > works fine for me on a 440 so you shouldn't have a problem > assuming that the walnut port is working. ** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/
