Move to POST:
http://post-office.corp.redhat.com/archives/rhvirt-patches/2011-February/msg00431.html

Hi Eric,

I tried to verify this bug in libvirt-0.8.7-6.el6. 
But I am not sure if your updated code was covered. Please check the following steps.

Steps:

1.Start a guest
# virsh start rhel6

2.Create a device with 'ide' controller and attach it to guest
# cat c1.xml 
<controller type='ide' index='0'>
      <alias name='ide0'/>
    </controller>

# virsh attach-device rhel6 c1.xml 
error: Failed to attach device from c1.xml
error: unsupported configuration: disk controller bus 'ide' cannot be hotplugged.

3.Recreate a device with the invaild controller and attach it to guest
# cat c2.xml 
<controller type='fdsfside' index='0'>
      <alias name='ide0'/>
    </controller>

# virsh attach-device rhel6 c1.xml 
error: Failed to attach device from c2.xml
error: internal error Unknown controller type 'fdsfside'


** In step3, ("Unknown controller type: %s") error message can be shown.

Unfortunately, the same error message (in step3) also can be shown in libvirt-0.8.7-4.el6. 

 
Would you please give some suggestions on this bugs?


Thanks,
Chun Cui

(In reply to comment #4)
> I tried to verify this bug in libvirt-0.8.7-6.el6. 
> But I am not sure if your updated code was covered. Please check the following
> steps.

I've asked the original reporter, Hu Tao, for how he tested his upstream patch.

Meanwhile, note that:

> 3.Recreate a device with the invaild controller and attach it to guest
> # cat c2.xml 
> <controller type='fdsfside' index='0'>
>       <alias name='ide0'/>
>     </controller>
> 
> # virsh attach-device rhel6 c1.xml 
> error: Failed to attach device from c2.xml
> error: internal error Unknown controller type 'fdsfside'

There are two different error messages:

src/conf/domain_conf.c:                                 _("Unknown controller type '%s'"), type);
src/qemu/qemu_command.c:                        _("Unknown controller type: %s"),

You hit the pre-existing one in domain_conf.c (no colon), not the new message in qemu_command.c (with a colon), so this test did not tickle the changed code path.  So I think it has to be a known existing controller type to get past the parsing in domain_conf.c, which leaves 'ide', 'fdc', 'scsi', 'sata', 'virtio-serial', and 'ccid'.  Then, in qemu_command.c, the new message can only be triggered if you hit the 'ide' or default: case label; your testing showed that 'ide' got filtered early, and the switch statement covers 'scsi', 'virtio-serial', and 'ccid', but that leaves 'fdc' and 'sata'.  So maybe trying your step 2, but with 'sata' or 'fdc' instead of 'ide', will trigger things?

Have a try with this xml:

<controller type='scsi' index='1'>
  <address type='pci' domain='0x0000' bus='0x00' slot='0x09' function='0x1'/>
</controller>

where slot 0x09 is an unused slot, you can specify any unused slot.

With commit cdbba1c4960a2 the error message is:
error: internal error Only PCI device addresses with function=0 are supported

Without commit cdbba1c4960a2 the error message is:
error: out of memory

Tried the xml mentioned in comment6. The error message is same with first error (commit cdbba1c4960a2 the error message)

Test environment:
libvirt-0.8.7-6.el6
qemu-kvm-0.12.1.2-2.144.el6
kernel-2.6.32-113.el6

cat c3.xml 
<controller type='scsi' index='1'>
  <address type='pci' domain='0x0000' bus='0x00' slot='0x09' function='0x1'/>
</controller>


# virsh attach-device rhel6 c3.xml 
error: Failed to attach device from c3.xml
error: internal error Only PCI device addresses with function=0 are supported

That error message should be fine.

Have you tried the xml with commit cdbba1c4960a2 reverted? If you do not use git, you can manually edit c source file to remove changes introduced by commit cdbba1c4960a2 then rebuild libvirt-0.8.7-6.el6 then repeat steps in comment 7. If you get 'error: out of memory', then commit cdbba1c4960a2 is in effect. Because we're not actually out of memory, but don't support PCI device addresses with function=1.

(In reply to comment #8)
> That error message should be fine.
> 
> Have you tried the xml with commit cdbba1c4960a2 reverted? If you do not use
> git, you can manually edit c source file to remove changes introduced by commit
> cdbba1c4960a2 then rebuild libvirt-0.8.7-6.el6 then repeat steps in comment 7.
> If you get 'error: out of memory', then commit cdbba1c4960a2 is in effect.
> Because we're not actually out of memory, but don't support PCI device
> addresses with function=1.

Yes, I have checked on libvirt-0.8.7-4.el6.x86_64 without commit cdbba1c4960a2 and get "error: out of memory".

-------
# virsh attach-device rhel6 c3.xml 
error: Failed to attach device from c3.xml
error: out of memory
------

So this bug is fixed. Can I change it's status?

Fine. Thanks for your verification.

Thanks for Hu Tao's great help.

Changed it to verified status.

Adding the following xml to a shutoff guest xml, then start the guest will causes 'out of memory' error. For this issue, I report a separate bug 698197 
<controller type='scsi' index='1'>
  <address type='pci' domain='0x0000' bus='0x00' slot='0x09' function='0x1'/>
</controller>

An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on therefore solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.

http://rhn.redhat.com/errata/RHBA-2011-0596.html