Bug 296061
| Summary: | No network in 32 bit fullvirt guests | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Product: | [Fedora] Fedora | Reporter: | Richard W.M. Jones <rjones> | ||||||||
| Component: | xen | Assignee: | Daniel Berrangé <berrange> | ||||||||
| Status: | CLOSED RAWHIDE | QA Contact: | Fedora Extras Quality Assurance <extras-qa> | ||||||||
| Severity: | low | Docs Contact: | |||||||||
| Priority: | low | ||||||||||
| Version: | 8 | ||||||||||
| Target Milestone: | --- | ||||||||||
| Target Release: | --- | ||||||||||
| Hardware: | x86_64 | ||||||||||
| OS: | Linux | ||||||||||
| Whiteboard: | |||||||||||
| Fixed In Version: | Doc Type: | Bug Fix | |||||||||
| Doc Text: | Story Points: | --- | |||||||||
| Clone Of: | Environment: | ||||||||||
| Last Closed: | 2007-09-19 21:22:30 UTC | Type: | --- | ||||||||
| Regression: | --- | Mount Type: | --- | ||||||||
| Documentation: | --- | CRM: | |||||||||
| Verified Versions: | Category: | --- | |||||||||
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |||||||||
| Cloudforms Team: | --- | Target Upstream Version: | |||||||||
| Embargoed: | |||||||||||
| Attachments: |
|
||||||||||
|
Description
Richard W.M. Jones
2007-09-19 10:05:58 UTC
# /usr/sbin/brctl show
bridge name bridge id STP enabled interfaces
eth0 8000.00e081740228 no peth0
vif1.0
vif2.0
vif3.0
vif4.0
vif5.0
virbr0 8000.000000000000 no
I'm using Xen networking. Some of those other vifX.0 devices
belong to other working paravirt guests.
Created attachment 199491 [details]
xm list --long
Output of xm list --long.
The domains debian32fv (domid 9) and gentoo32fv (domid 4)
are the afflicted ones.
Created attachment 199511 [details]
xend.log
This is a section from xend.log which shows the
domain debian32fv being shut down and then restarted.
These are the command lines of the qemu-dm processes. Paravirt (working) domains: /usr/lib64/xen/bin/qemu-dm -M xenpv -d 1 -domain-name fc6_0 -vnc 127.0.0.1:2 /usr/lib64/xen/bin/qemu-dm -M xenpv -d 3 -domain-name fc6_1 -vnc 127.0.0.1:3 /usr/lib64/xen/bin/qemu-dm -M xenpv -d 5 -domain-name f764pv -vnc 127.0.0.1:0 -vncunused Fullvirt (non-working) domains: /usr/lib64/xen/bin/qemu-dm -d 4 -vcpus 4 -boot c -serial pty -acpi -usb -domain-name gentoo32fv -vnc 127.0.0.1:0 -vncunused /usr/lib64/xen/bin/qemu-dm -d 9 -vcpus 1 -boot c -acpi -usb -domain-name debian32fv -vnc 127.0.0.1:0 -vncunused I edited:
/var/lib/xend/domains/<uuid>/config.sxp
and changed:
(device
(vif
(bridge eth0)
(uuid eff1d1f0-2a91-e370-9afc-9f584db28948)
(script vif-bridge)
(mac 00:16:3e:36:61:80)
(type ioemu)
^^^^^ was netfront, changed to ioemu
(backend 0)
)
)
which fixes the network.
Created attachment 199941 [details]
Don't clobber 'type' field
There are 3 possible values for the 'type' field in XenD vif devices
- netfront - paravirt drivers
- ioemu - QEMU emulation
- None - paravirt & QEMU
libvirt creates HVM guests with 'None' for type, so that they can easily be
configured to use paravirt drivers post-install. This works initially when
starting the guest. But at the time you start the first guest, XenD re-writes
its config file out to disk, and in doing so converts 'None' into 'netfront'.
This is clearly bogus - it should not change it. The attached patch removes the
code which clobbers it.
With the patch in comment 6 it appears that a guest which starts with no (type ..) in its vif device gets rewritten as (type ioemu). I don't think that is what you intended, but at least it doesn't break the vif configuration as before. (I will also note for myself that xend does the rewriting just after the domain starts up). Very strange, I just retested it and now it doesn't add (type ioemu) but leaves it with no (type ...). Not sure what was going on before. Built into rawhide * Wed Sep 19 2007 Daniel P. Berrange <berrange> - 3.1.0-6.fc8 - Don't clobber the VIF type attribute in FV guests (rhbz #296061) Thanks Daniel! This bug's been biting me lately on F7/x86_64 with a FC6/i386 guest. Thought I was losing my mind. :-) |