Description of problem: When installing rawhide on a Fedora 9 kvm host, installer freezes during partitions creation. I was told that the thing happening at that time is udevsettle, hence the description. The following message is displayed at the freeze time: Activating new partitions. Version-Release number of selected component (if applicable): kvm-65-7.fc9.x86_64 (host) anaconda-11.4.1.34-1.x86_64 (guest) udev-127-1.fc10.x86_64 (guest) How reproducible: always Steps to Reproduce: 1. Set up a kvm host in F-9 2. Attempt to install rawhide. I used the following command: virt-install --hvm --os-type=linux --location=http://sunsite.icm.edu.pl/pub/Linux/fedora/linux/development/i386/os/ --name=rawhide-fullvirt --vcpus=1 --file=/var/lib/libvirt/images/rawhide-fullvirt.img --accelerate --ram=512 --file-size=15 3. Proceed as normal preserving default options Actual results: Installer freezes Expected results: Installer installs rawhide to the end Additional info: I tried rawhide trees from 20080719, 20080912 and 20080913, which all fail. Fedora 9 installs fine.
virt-install --hvm --os-type=linux --location=http://sunsite.icm.edu.pl/pub/Linux/fedora/linux/development/i386/os/ --name=rawhide-fullvirt --vcpus=1 --file=/var/lib/libvirt/images/rawhide-fullvirt.img --accelerate --ram=512 --file-size=15 --extra-args="console=ttyS0 console=tty0 ignore_loglevel initcall_debug" made it work. More info to follow tomorrow.
Of course the guest packages are i386.
Erm, that's what happens when you work on one thing for too long... It turns out that i386 guest installs on x86_64 host even without any parameters specified, it's x86_64 guest that stalls. I'll post some logs ASAP.
Created attachment 316683 [details] Log file of the rawhide installation, ignore_loglevel This is the stuff that gets output to the serial console with the following command: virt-install --hvm --os-type=linux --location=http://sunsite.icm.edu.pl/pub/Linux/fedora/linux/development/x86_64/os/ --name=rawhide-x86_64 --vcpus=1 --file=/var/lib/libvirt/images/rawhide-x86_64.img --accelerate --ram=512 --file-size=15 --extra-args="console=ttyS0 console=tty0 ignore_loglevel" At this point, anaconda is stalled.
Created attachment 316684 [details] Same log with ignore_loglevel initcall_debug
I don't know how to debug anaconda.. udevsettle just waits for all udev events to be processed. What I would need is a list of processes running at that moment, or setting udev in debug mode. Does anaconda use "start_udev" ? I don't think so. Putting udev in debug mode would be: /sbin/udevadm control --log_priority=debug
*** Bug 465529 has been marked as a duplicate of this bug. ***
There is no easy way to do this, no. It requires rebuilding the loader and making new boot media. Is this still an issue?
I just tried and anaconda manages to pass the "Activating new partitions" screen, so it seems this problem is fixed (unless udevsettle is run at different point).
udevsettle does not do bad things!
I have no idea what was locking up anaconda, blaming udevsettle is a result of a chat with folks in #fedora-devel, who said that this is what is happening when "Activating new partitions" is displayed. Anyway, it seems to work now so I believe this report can be closed.