Bugzilla – Bug 612143
i686 LiveCDs failed to detect CD - rebooting in 120 seconds
Last modified: 2012-03-06 08:08:37 UTC
User-Agent: Mozilla/5.0 (X11; U; Linux i686; de; rv:1.9.2.3) Gecko/20100401 SUSE/3.6.3-1.2 Firefox/3.6.3 In my automated KVM testing setup, i686 LiveCDs often (75%) stop right after bootloader. I found that the ata_piix driver for IDE-CD is not loaded. Reproducible: Sometimes Steps to Reproduce: 1. mkdir raid ; for i in 1 2 3 4 ; do kvm-img create raid/$i 7G ; done 2. /usr/bin/kvm -m 1024 -net user -monitor tcp:127.0.0.1:15222,server,nowait -net nic,model=virtio,macaddr=52:54:00:12:34:56 -drive file=raid/1,if=virtio,boot=on -drive file=raid/2,if=virtio -drive file=raid/3,if=virtio -drive file=raid/4,if=virtio -boot dc -cdrom openSUSE-KDE-LiveCD-i686-Build0663-Media.iso -vnc :99 3. give nohz=off on command line, choose installation Actual Results: error message: failed to detect CD - rebooting in 120 seconds Expected Results: should load ata_piix, find CD and boot up This issue did not occur in earlier tests when I had -smp 4 as parameter. I was able to trigger the error message on physical hardware (Phenom X4 with USB-Stick) by using nohz=off maxcpus=1 mem=1000M (res at 800x600)
Created attachment 367491 [details] Kiwi debug log
Steffen can you take over that one ? Thanks
Marcus, how are the modules loaded? udev?
if /lib/modules/*/modules.order exists I let udev load the modules The following modules are loaded manually though: rd brd edd dm-mod xennet xenblk squashfs fuse but that shouldn't be related to this report
btw, I use hwinfo to detect the device don't know if there is any module loading involved by hwinfo itself, I think not
I found that hwinfo --cdrom only detected the DVD drive if the driver was already loaded: http://www3.zq1.de/opensuse/video/openSUSE-KDE-LiveCD-i686-Build0625-cdrom-test.ogv Is it supposed to be that way?
Sure. hwinfo shows what's there. It is not responsible for system setup. In any case this is more a driver issue. Assigning to kernel.
FWIW, I've seen this in virtualbox too.
This problem is still present in current LiveCD-Build0685. Occurrence on my KVM tests is around 50-80% so can be quite annoying. Is there any I can do? More information or testing needed so that someone could reproduce and fix this?
I just booted 10 times (without raid setup) and it found the cd every time.
On IRC, coolo mentioned that we can debug this with kiwidebug=1 as boot option. I don't see anything weird in dmesg, but there's no /dev/sr0 device and there's nothing in /sys/class/block/ except ramX files. Kay, would you have any idea what could be happening, there?
Hmm, I have no idea, what's going wrong here. The sr_mod kernel module is loaded at that time? Does: grep . /sys/bus/scsi/devices/*/* print anything that looks like a CD drive?
(In reply to comment #12) > Hmm, I have no idea, what's going wrong here. > > The sr_mod kernel module is loaded at that time? Yes. > Does: > grep . /sys/bus/scsi/devices/*/* > print anything that looks like a CD drive? /sys/bus/scsi/devices/ is empty.
Kay and I made some progress: 15:04 < kay> vuntz: there seems something very wrong, i don't really have an idea. is ahci loaded? 15:05 < vuntz> kay: no, it wasn't 15:07 < kay> vuntz: does the cd appear when you load it? 15:08 < vuntz> kay: no 15:08 * kay wonders who made all that stuff modules again ... 15:08 < kay> vuntz: does loading ata_piix make it appear? 15:08 < kay> vuntz: it's all in kvm, right? 15:09 < vuntz> kay: yes it does make it appear 15:09 < vuntz> kay: virtualbox for me 15:10 < kay> vuntz: if you run 'udevadm trigger' does ata_piix get loaded? 15:10 < vuntz> kay: hrm, let me reboot since I now have the module 15:12 < vuntz> kay: yes, it does 15:13 < kay> vuntz: this happens on real boxes too? 15:14 < vuntz> kay: I only saw that on virtualbox 15:14 < vuntz> and the other person in the bug in kvm 15:15 < kay> vuntz: the bootup script should do exactly that trigger, which loads the module. sounds like a weird timing problem with virtualization 15:16 < vuntz> fwiw, I've never seen this with 11.2 and earlier. So it started this cycle 15:17 < kay> vuntz: i think they have been compiled-in in 11.2, like many other distros do too these days 15:17 < kay> vuntz: but i'm not sure, there too many kernel configs :) 15:19 < kay> vuntz: the moblin stuff has all the needed stuff compiled-in, i thought that we used this config for 11.2, but as said, i'm not sure if that really was the case
openSUSE 11.3 is out of maintenance. If you're able to reproduce this with 11.4 or 12.1, please re-open with an updated Product field.