Bugzilla – Bug 106704
SuSE Linux 10.0 loading scpm profile from NLD 9 partition
Last modified: 2008-07-17 13:29:27 UTC
I am dual booting with NLD 9 and SuSE Linux 10.0 I have an scpm profile with NLD. I have no profiles with SuSE Linux 10. Networking does not work on SuSE Linux 10. Upon searching the /var/log/messages file it is attempting to load my scpm profile from NLD 9. They are on different partitions and I do not mount the NLD 9 filesystem when booting to SuSE Linux 10.
Please attach the appropriate snippet of your /var/log/messages.
Created attachment 47262 [details] lan logs stripped from /var/log/messages
Created attachment 47265 [details] Another snippet of /var/log/messages
Both attachments contain unique snippets from /var/log/messages
So what makes you think your system is trying to restore a scpm profile? There is nothing in the log indicating that, and frankly, it's not possible at all. The log shows you have a driver loaded for your ethernet card and ndiswrapper for you wlan card, which looks good so far. Please describe more in detail what is not working as expected.
Dave, please provide Joachim with the needed info.
I agree that my LAN configuration looks good, but it does not work. I have attempted to set it up identically on Code10 as it is on NLD9, where it is working, but I can't get it to grab a DHCP address. When I use the dmesg command, I see a PROFILE= parameter being passed to the kernel. Whatever my current SCPM profile name is in NLD9, is the name passed to the kernel via this parameter. See attached boot.msg file, line 47. When I change the profile in use on NLD9, and reboot to Code10 beta2, the name of the PROFILE= parameter changes to match the name of the profile in NLD9. NLD9 and Code10 are installed on separate partitions, and this second install also has separate swap files as well. Neither OS is configured to mount the other's partition on bootup-if I need access to the other partition, I have to mount it manually. I see no way for them to communicate the profile name, but it is clearly happening.
Created attachment 47427 [details] boot.msg file
I will also attach GRUB's menu.lst file, so you can see I'm not manually passing the PROFILE parameter.
Created attachment 47428 [details] menu.lst
Regarding the LAN setup: Does calling 'ifup eth0 -o debug' manually shows any errors? Regarding the profile stuff: Now I see where the problem is. Grub always adds the PROFILE parameter when scpm is in use. It reads the list of profiles from /boot/message, and you probably have two /boot/message files, one on each installation. But grub uses only one configuration and one message file, regardless of what scpm writes to the other one. So this is not easily fixable. grub would have need to mount the root partition of the selected entry and look there for configured profiles. So actually you need multiple grubs for multi-boot scenarios with scpm, that means, you need to install a grub into the boot section each partition containing an installation using scpm, and configure the grub in the master boot record to just chainload these. I think it would be a good idea to disable the profile support in grub when there are multiple bootable linux installation configured, so it can't do any harm. I don't know whether this is hard to implement. Torsten, can you help?
I've added a section about 'booting into profiles' to the SCPM documentation, which contains a warning and the hint to use chained grubs in this case.