Attempting to do a fresh install of Gentoo on a machine after replacing failed hard disk. Hardware with problem is a Linksys Network Everywhere Fast Ethernet 10/100 (rev 11) according to lspci (commonly referred to as LNE100TX), based on the tulip series of chipsets. Some combination of settings in the LiveCD support-everything kernel results in a failure to get an address via DHCP: (Taken from the logging console after boot) eth0: ADMtek Comet rev 17 at e08aec00, EEPROM not present ... 0000:01:00.0: tulip_stop_rxtx() failed (above line repeats three more times) dhcpcd [7972]: terminating on signal 1 dhcpcd [7283]: recvfrom: Network is down Trying to restart eth0 via the initscript results in the following: * Starting eth0 * Bringing up eth0 * dhcp * Running dhcpcd... [ !! ] Meanwhile, syslog reports the same message about tulip_stop_rxtx() failing four times and dhcpcd exits with "timed out waiting for a valid DHCP server response." Same results in syslog if I attempt to manually bring up the interface with "dhcpcd eth0". Unable to test interface with a static IP as my router doesn't recognize static IPs and appears to have no mode in which to allocate blocks for static use. Reproducible: Always Steps to Reproduce: 1. Have an x86 box with an LNE100TX 2. Use install-x86-minimal-2005.1.iso 3. Actual Results: eth0 did not get address via DHCP, no connectivity Expected Results: eth0 should have been configured via DHCP at boot time Definitely unable to get all of 'emerge info' because no way of copying/pasting mass amounts of text off of networkless system (above log info was copied by hand!)
Kernel team: seen anything about this sort of thing? If so, has it been fixed? If so, at what version was it fixed? Thanks... =]
Googling seems to turn up that a patch introduced in 2.6.10 designed to make tulip_stop_rxtx() wait for DMA to finish makes things break on certain systems. More info: http://oss.sgi.com/projects/netdev/archive/2004-11/msg00628.html Also, I have the same NIC (LNE100TX rev 11) in another desktop that produces this message to the console, but it does not break DHCP or IP like it does on the other machine. The totally-broken machine is a P3-1000 with an i810-based mainboard, and the working machine is an Athlon XP 2600+ on an Asus A7V880 (Via KT880 + VT8237) board.
Can you try out the 2005.1-r1 CD, please? It has an updated kernel revision that might resolve this issue.
Tried the new LiveCD, same error message.
Assigning to kernel team. If this has been fixed since 2.6.12-r10, feel free to shoot this back to release@ and let me know what version fixed it. Also, once you get this resolved (if it isn't already) shoot it back to release@ so I can track it until our next release. Thanks
So do you now have a working install on this system? If so, the next step is to test a newer kernel such as 2.6.14
No, currently there is not a working install since the LiveCD will not get a network connection. I'll do a networkless install sometime this week and post the results back here.
Please reopen when you are able to test this further. I understand that this is awkward for you to work on. If possible, it would probably be easier if you could temporarily install another network card in that system?