Fedora Account System
Red Hat Associate
Red Hat Customer
Description of problem: When installing RHEL-5.9 guest using PXE on Fedora, whole PC freezes. I have to shut it down using power button to restart it. Version-Release number of selected component (if applicable): libvirt-0.10.2.4-1.fc18 virt-manager-0.9.5-1.fc18 How reproducible: Always and also on different Hosts with separate Fedora 18 instalation Steps to Reproduce: 1. open virt-manager 2. create new guest 3. turn it off, set up network to be bridged with Host ethernet controller 4. turn on the Guest 5. Boot RHEL-5.9 installation using PXE /* this may have no effect on the bug */ 6. use default language and keyboard settings, turn off IPv6 7. after that the initial text mode switches to graphical installation UI 8. whole Host system freezes after switching to graphical UI Actual results: Whole Host system freezes after switching to graphical UI Expected results: Host system will NOT freeze after switching to graphical UI Additional info: Once it freezed right before switching to the graphical UI. The installation output showed something like: "initializing video device 0100 Red Hat Inc." So maybe this has something to do with video device in libvirt (my guess).
Hey Tomas, libvirt's a userspace component, so I'd be surprised if it was the thing locking up the host, but let's see if we can figure out what's going on here. A few questions: can you provide the guest XML; does the guest OS make a difference; can you ssh into the host while it's in this state?
(In reply to comment #1) > Hey Tomas, libvirt's a userspace component, so I'd be surprised if it was > the thing locking up the host, but let's see if we can figure out what's > going on here.A few questions: can you provide the guest XML I'll attach it to this Bug. > does the guest OS make a difference Well yes. I have problems only when installing RHEL-5-9 (and also with latest RHEL-5.10 nightly snapshot). With the latest nightly snapshot I was able to get further than with RHEL-5.9 but instead of freezing my host kernel crashed. I'll attach picture of what I got on my screen. I'm using kernel-3.9.2-200.fc18.x86_64 > can you ssh into the host while it's in this state? ssh is not working after the host freezes, but I was still able to ping the host.
Created attachment 749296 [details] Guest XML I replaced MAC address of the network interface with "xx:xx:xx:xx:xx:xx" since it is my physical interface's MAC.
Created attachment 749327 [details] Screenshot 1
Created attachment 749328 [details] Screenshot 2
(In reply to comment #4) > Created attachment 749327 [details] > Screenshot 1 (In reply to comment #5) > Created attachment 749328 [details] > Screenshot 2 Sorry for the bad quality of pictures, but this is the best I'm able to take with my phone's camera. ABRT is reporting that there was a kernel oops but the Kernel was tainted (flags: GFD). G - Proprietary module has been loaded F - Module has been forcibly loaded D - Kernel has oopsed before tainted modules are: tpm_tis, joydev Hope this helps to find out what might be causing this, but I'm not sure if this is libvirt Bug or kernel Bug.
(In reply to comment #6) > Hope this helps to find out what might be causing this, but I'm not sure if > this is libvirt Bug or kernel Bug. This is definitely not a libvirt bug; I'm changing the component to kernel, but I suspect you're not going to get a lot of interest from the kernel folks unless you can reproduce on an untainted kernel.
(In reply to comment #7) > (In reply to comment #6) > > Hope this helps to find out what might be causing this, but I'm not sure if > > this is libvirt Bug or kernel Bug. > > This is definitely not a libvirt bug; I'm changing the component to kernel, > but I suspect you're not going to get a lot of interest from the kernel > folks unless you can reproduce on an untainted kernel. Thanks! I'm using *default* Fedora kernel without any 3rd party modules and it is happening on any Fedora 18 installation (I tested on 3 different hosts) so I think they should be interested...
I should also mention that this issue happens also if I unload "tpm_tis" and "joydev" modules from kernel.
Based on the stack trace involving macvlan, macvtap, and vhost_net, I'm going to mark this as a duplicate of 954181 *** This bug has been marked as a duplicate of bug 954181 ***