boot errors - slack 11.0 - kernels > 2.6.19?
I've looked just about everywhere for an answer to this.
With any kernel >2.6.19, during boot, after entering mulituser mode and following detection of scsi devices, I get these lines/errors: chmod: changing permissions of /tmp: read only file system chmod: cannot access /var/tmp : input output error INIT: cannot execute /sbin/agetty (repeated many times) INIT: Id c4 respawning too fast: disabled for 5 minutes ...and so on for c1, c2, c3, c5 ,c6 INIT: no more processes left in this runlevel. What could have changed post 2.6.19 to create this behaviour. I have been over my kernel configuration again and again looking for clues and googled looking for an answer. Most of these errors in isolation have an explanation on the net, but the combination and hence the possible cause seems to evade me at this stage. When booting 2.6.19; the line following scsi detection is "Starting portmapper" then the boot continues normally. When booting >2.6.19 kernels; the system is attempting to change the permissions of /tmp and access /var/tmp. The difference in kernel behaviour prior to this error is that kernels >2.6.19 are searching for NEW scsi interfaces/devices, whereas 2.6.19 does not. Don't know whether that is significant or not. This is a bit beyond me at this stage, and would be grateful of any advice. I realise that posting in the kernel sub forum may be appropriate, but felt that this was slackware specific, not wanting to double post. Cheers, |
To atleast get the thing to boot, you might change the permissions of /etc/rc.d/rc.portmap(per) and /etc/rc.d/rc.scanluns and make them NON-executable.
It is surely a weird looking situation. Another thing to try, is to boot the working kernel, locate and extract /proc/config.gz to a file somewhere. Now run xconfig to compile your NEW kernel, and from xconfig, load up the config file you just grabbed from the running kernel and save it as .config, then compile, (make, make modules_install, fix lilo etc..) and reboot it. Does this help? |
Thanks grapefruitgirl.
As you suggested, disabling rc.scanluns allowed the system to boot without problems. On checking both kernel configs 2.6.19 & 21.1 - Device Drivers>SCSI Devices>Probe all LUNS on each SCSI device, is enabled. In fact both SCSI configs are identical? The noscanluns option referred to in the rc.scanluns script may be redundant. Running the rc.scanluns script was only necessary if noscanluns was NOT a selected kernel option? Anyway, as Probe all LUNS etc is compiled, running rc.scanluns doesn't seem necessary. Please correct me if I am wrong. I still can't figure out why apparently identical configurations behave differently. If I run rc.scanluns locally, it works fine with 2.6.19 but produces the same errors with 2.6.21.1. |
Looks like you are on the right track in diagnosing the situation. I'm no expert by any means, so I may be wrong too, but your idea about the scanluns thing being redundant is reasonable, and may be correct.
Personally, I have had the rc.scanluns script disabled since installation, and have never used it. If everything in your machine works and there are no problems or issues with any devices, then leave it disabled. Thankfully. slackware and its scripts tend to be very well commented :) which is often beneficial. As for what the differences are between your otherwise identical kernels, either check, check, check again for something you've missed, or consult the kernel release changelog.txt for indications of maybe a coding change affecting this. |
Yes thanks again, I'll have a look at the kernel changelogs.
|
All times are GMT -5. The time now is 03:03 PM. |