libvirt users - Jun 2014 - Re: [netcf]IFF_RUNNING flag on a bridge device

Hi All,

I have one netcf question, please help me to resolve it, thanks.

I can set a IFF_RUNNING flag to a bridge device which are no interface device
attached. What status of a flag on a bridge device in current kernel?(w/o
interface), is this a new change in kernel or other component? In netcf, but
there is a patch to fix that flag issue, I can't understand it.

fix wrong status of bridge devices
The IFF_RUNNING flag is not set on a bridge device if there are no devices
attached to it. So bridges without attached devices are reported as inactive.
Destroying inactive devices leads to an error and thus the device can't be
removed (easily) from the operating system. This patch checks if the device is a
bridge and if so ignores the IFF_RUNNING flag.

https://git.fedorahosted.org/cgit/netcf.git/commit/?id=7fa5fa5a1e43e6e2afec68c58b5f43bc916703a6


My testing results of IFF_RUNNING on non-device bridge:

Version:
[root@rhel6 boot]# rpm -q kernel netcf
kernel-2.6.32-468.el6.x86_64
netcf-0.2.4-1.el6.x86_64

[root@rhel6 ~]# brctl show br0
bridge name	bridge id		STP enabled	interfaces
br0		8000.000000000000	no		
[root@rhel6 ~]# ifconfig br0
br0       Link encap:Ethernet  HWaddr 6A:3B:6F:DF:17:A9  
          BROADCAST MULTICAST  MTU:1500  Metric:1
          RX packets:0 errors:0 dropped:0 overruns:0 frame:0
          TX packets:6 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0 
          RX bytes:0 (0.0 b)  TX bytes:468 (468.0 b)

[root@rhel6 ~]# ifconfig br0 up
[root@rhel6 ~]# brctl show br0
bridge name	bridge id		STP enabled	interfaces
br0		8000.000000000000	no		
[root@rhel6 ~]# ifconfig br0
br0       Link encap:Ethernet  HWaddr 6A:3B:6F:DF:17:A9  
          inet6 addr: fe80::683b:6fff:fedf:17a9/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:0 errors:0 dropped:0 overruns:0 frame:0
          TX packets:11 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0 
          RX bytes:0 (0.0 b)  TX bytes:866 (866.0 b)

On 05/27/2014 09:07 AM, Jianwei Hu wrote:
> Hi All,
>
> I have one netcf question, please help me to resolve it, thanks.
>
> I can set a IFF_RUNNING flag to a bridge device which are no interface
device attached. What status of a flag on a bridge device in current kernel?(w/o
interface), is this a new change in kernel or other component? In netcf, but
there is a patch to fix that flag issue, I can't understand it.
>
> fix wrong status of bridge devices
> The IFF_RUNNING flag is not set on a bridge device if there are no devices
attached to it. So bridges without attached devices are reported as inactive.
Destroying inactive devices leads to an error and thus the device can't be
removed (easily) from the operating system. This patch checks if the device is a
bridge and if so ignores the IFF_RUNNING flag.
Well, I took the patch submitter (Hendrik Schwartke <hendrik@os-t.de>)
at his word about the brokenness, and the patch didn't *hurt* anything,
so I pushed it. But now that I try it out I see that, at least on the
kernel version in Fedora 20 (and RHEL6.6) IFF_RUNNING *is* set when a
bridge device is ifconfiged up.

So I'm not sure what had caused the original problem, maybe Hendrik can
eluminate us on that, but it makes sense (since the whole reason for
adding the check of IFF_RUNNING in the first place was to notice when an
ethernet is unplugged, but a bridge is *never* plugged or unplugged in
the sense that we're looking for) and doesn't harm anything (and
apparently does fix something for Hendrik).

BTW, when you have netcf-specific questions of patches, you can send
them to netcf-devel@lists.fedorahosted.org

On 28.05.2014 15:27, Laine Stump wrote:
> On 05/27/2014 09:07 AM, Jianwei Hu wrote:
>> Hi All,
>>
>> I have one netcf question, please help me to resolve it, thanks.
>>
>> I can set a IFF_RUNNING flag to a bridge device which are no interface
device attached. What status of a flag on a bridge device in current kernel?(w/o
interface), is this a new change in kernel or other component? In netcf, but
there is a patch to fix that flag issue, I can't understand it.
>>
>> fix wrong status of bridge devices
>> The IFF_RUNNING flag is not set on a bridge device if there are no
devices attached to it. So bridges without attached devices are reported as
inactive. Destroying inactive devices leads to an error and thus the device
can't be removed (easily) from the operating system. This patch checks if
the device is a bridge and if so ignores the IFF_RUNNING flag.
> Well, I took the patch submitter (Hendrik Schwartke
<hendrik@os-t.de>)
> at his word about the brokenness, and the patch didn't *hurt* anything,
> so I pushed it. But now that I try it out I see that, at least on the
> kernel version in Fedora 20 (and RHEL6.6) IFF_RUNNING *is* set when a
> bridge device is ifconfiged up.
Sorry for my late reply. I was on vacation. I saw this behaviour on
Ubuntu 14.04. I think the kernel version was 3.13.0, but I'm not quite sure.
The IFF_RUNNING flag is not set on bridges without attached interfaces
even if the bridge is up.
>
> So I'm not sure what had caused the original problem, maybe Hendrik can
> eluminate us on that, but it makes sense (since the whole reason for
The problem is that my application create bridge devices which are
connected to virtual machines. If all vms are destroyed then there are
no devices attached to the bridge anymore. Now I can't destroy the
bridge device because it seems already to be down. If I undefine the
bridge device netcf removes it from /etc/network/interfaces but doesn't
remove the actual device because netcf thinks it's already
removed.
> adding the check of IFF_RUNNING in the first place was to notice when an
> ethernet is unplugged, but a bridge is *never* plugged or unplugged in
> the sense that we're looking for) and doesn't harm anything (and
> apparently does fix something for Hendrik).
Yes, that's right. It doesn't make any sense to check for IFF_RUNNING on
*bridges*. So I'm pretty sure that this patch doesn't harm anything.
>
> BTW, when you have netcf-specific questions of patches, you can send
> them to netcf-devel@lists.fedorahosted.org
>
>

-- 
Hendrik Schwartke
OpenSource Security Ralf Spenneberg     http://www.os-t.de
Am Bahnhof 3-5                          48565 Steinfurt         Germany
Fon:   +49(0)2552 638 755               Fax: +49(0)2552 638 757