Re: [libvirt] [PATCH] Add support for 'passthru' mode for direct network interfaces
Thank you for picking this up !
I've got a question: I want to change the link device's MAC address
prior to starting the VM, and I want to re-set the MAC to it's original
value afterwards.
Setting the MAC in the IF is no issue, but is it ok if I add a MAC
variable to
the direct structure in virDomainNetDef to keep the linkdev's original
MAC ?
The issue with these MACs is that some adapters with VFs come up with
randomized MAC addresses, so resetting them to a pre-set value is not
possible. Also it would make migration leave a trail of identical MACs
throughout a datacenter.
Regards,
Dirk
On May 18, 2011, at 4:22 PM, Eric Blake wrote:
On 05/18/2011 05:32 AM, D. Herrendoerfer wrote:
>>>> destination
>>>> device is in <code>private</code> mode.</dd>
>>>> + <dt><code>passthru</code></dt>
>>>> + <dd>This feature allows to attach a virtual function of a
>>>> SRIOV
>>>> capable
>>>
>> That patch looks fine to me, ACK,
>> The only unresolved uncertainty to me is what happen if you try to
>> use
>> it on a kernel without the support ?
>>
>
> This looks fine, except in formatdomain.html.in the feature is
> still called
> passthru.
>
> Used on a system without passthru support this raises an error
> indicating a
> failed macvtap setup.
I fixed that one last place, and found one more issue:
>
> --- a/src/util/macvtap.h
> +++ b/src/util/macvtap.h
> @@ -92,6 +92,7 @@ void delMacvtap(const char *ifname,
> # define MACVTAP_MODE_PRIVATE_STR "private"
> # define MACVTAP_MODE_VEPA_STR "vepa"
> # define MACVTAP_MODE_BRIDGE_STR "bridge"
> +# define MACVTAP_MODE_PASSTHRU_STR "passthru"
No one was using these defines (they were leftovers from before we
switched to virEnum), so I deleted them.
Now pushed. Thanks for the submission!
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library http://libvirt.org