Bugzilla – Bug 153669
net install from eth1 instead of eth0
Last modified: 2006-05-08 12:45:35 UTC
using "beta-5" factory (time stamps and MD5 below) boot.iso image on CD I tried to install via NFS on a AMD Opteron server with 2 broadcom gbit-ethernet ports (tg3 driver) -- this worked fine with beta-3. only eth0 is connected, eth1 is open. now with beta-5, network setup got a correct IP address for our internal network (dhcpcd only can get this data via eth0!!), but NFS startup didn't work. after some checks on VC2 I realized, that eth1 got setup with the "correct" ip address (which is not connected) and eth0 is down. how comes ???? after "ifconfig eth1 down ; dhcpcd eth0" I could continue with NFS installation (see next bug report for yast2 :-((( these are the boot images used: -rw-r--r-- 1 koenig s+c 75476992 Feb 23 12:54 inst-source/boot/boot.iso -rw-r--r-- 1 koenig s+c 6102735 Feb 22 18:50 inst-source/suseboot/initrd64 -rwxr-xr-x 1 koenig s+c 8493720 Feb 22 18:57 inst-source/suseboot/inst64* 4ea0d97534d447af1ae6947972eeba8c inst-source/boot/boot.iso 785def173f15d9d765fe40a03410dcd4 inst-source/suseboot/initrd64 359a818a6b9c469d389e0f1abc361ffd inst-source/suseboot/inst64
Does this happen in the first stage of installation?
(In reply to comment #1) > Does this happen in the first stage of installation? I'm not 100% sure what is "the first stage of installation" but most likely the answer is "yes". I boot from CD, press F3, select installation via NFS (change VESA resolution to 1024x768, select german language) and then start installation (which boots kernel form CD). then I got a read error box in text mode which mentions some problem mounting NFS installation directory... there was a shell available (on VC2?) where I could run ifconfig/dhcpcd/route/... to check setup. in y2start.log I found the following entries for "eth" (no eth0 entry!) Stage [1]: YaST2 startup level (1)... Stage [1]: ========================== |-- export Alias="eth1 tg3" |-- export BOOT_IMAGE="linux" |-- export BootDisk="0" |-- export Broadcast="10.0.3.255" |-- export Cmdline="splash=silent vga=0x317" ... |-- export Nameserver="10.0.0.148" |-- export NetConfig="static" |-- export NetUniqueID="JNkJ.WBuuuXHFfS7" |-- export Netdevice="eth1" |-- export Netmask="255.255.255.0" |-- export Network="10.0.3.0" ,,,
Was there any output from the tg3 module or on one of the consoles like Looking for a network server... OK, that didn't work, see if we can activate another network device... Does acpi=off change anything?
(In reply to comment #3) > Was there any output from the tg3 module or on one of the consoles like > > Looking for a network server... > OK, that didn't work, see if we can activate another network device... > > Does acpi=off change anything? > sorry, but I don't have more information or console logs. I don't remember any kernel driver output -- I'd guess that usually I won't see any driver output anyway when starting a net installation from CDROM. unfortuneately I'm running out of time and can't do any further tests or report because I'm leaving tomorrow for a 3 weeks holiday -- most likely totally offline (and without any 64bit server with 2 ethernet devices;) in the rain forrests of Thailand...
(In reply to comment #3) > Was there any output from the tg3 module or on one of the consoles like > > Looking for a network server... > OK, that didn't work, see if we can activate another network device... > > Does acpi=off change anything? > hmm, one more detail which just pops up my mind from my tests on the text console, maybe this can be any help ? both eth0 and eth1 devices have been there and could be checked with ifconfig eth0 ifconfig eth1 and both eth[01] showed the correct mac address (eth0 is ...:F4 and eth1 is ...:F5 (IIRC) as usual)
Don't think it's a problem in linuxrc as it worked up to beta-3. Maybe a driver problem. BTW, a shell for debugging runs on console 9. Unfortunately, as Harald is off for some weeks, there's nothing I can do presently.
reopen if this can be reproduced with rc5