[RFC PATCH] bridge: remove !IFF_UP restriction when deleting bridge from ioctl

[RFC PATCH] bridge: remove !IFF_UP restriction when deleting bridge from ioctl

[Dan Williams]

netlink doesn't care about IFF_UP when deleting a bridge, so "ip link
del br0" works just fine.  The ioctl does care, which means that brctl
complains "bridge br0 still up; can't delete it".  Make things
consistent by always allowing bridge deletion even if the bridge is up.

Signed-off-by: Dan Williams <dcbw@redhat.com>
---
Question: does anyone consider the ioctl behavior API?  Can we change it
even though it's been this way forever?  It means that bridges that
previously would not have been deleted by brctl will now be deleted.
But anyone using /sbin/ip could already delete them.

diff --git a/net/bridge/br_if.c b/net/bridge/br_if.c
index 54d207d..33cd83f 100644
--- a/net/bridge/br_if.c
+++ b/net/bridge/br_if.c
@@ -264,19 +264,14 @@ int br_del_bridge(struct net *net, const char *name)
 		ret =  -ENXIO; 	/* Could not find device */
 
 	else if (!(dev->priv_flags & IFF_EBRIDGE)) {
 		/* Attempt to delete non bridge device! */
 		ret = -EPERM;
 	}
 
-	else if (dev->flags & IFF_UP) {
-		/* Not shutdown yet. */
-		ret = -EBUSY;
-	}
-
 	else
 		br_dev_delete(dev, NULL);
 
 	rtnl_unlock();
 	return ret;
 }
 

Re: [RFC PATCH] bridge: remove !IFF_UP restriction when deleting bridge from ioctl

[David Miller]

From: Dan Williams <dcbw@redhat.com>
Date: Fri, 04 Apr 2014 09:56:22 -0500

> netlink doesn't care about IFF_UP when deleting a bridge, so "ip link
> del br0" works just fine.  The ioctl does care, which means that brctl
> complains "bridge br0 still up; can't delete it".  Make things
> consistent by always allowing bridge deletion even if the bridge is up.
> 
> Signed-off-by: Dan Williams <dcbw@redhat.com>
> ---
> Question: does anyone consider the ioctl behavior API?  Can we change it
> even though it's been this way forever?  It means that bridges that
> previously would not have been deleted by brctl will now be deleted.
> But anyone using /sbin/ip could already delete them.

Unfortunately I think we're stuck with the existing behavior.

And since, as you say, iproute2 always does the removal when
the device is up, the behavior you desire is available.

Re: [RFC PATCH] bridge: remove !IFF_UP restriction when deleting bridge from ioctl

[Dan Williams]

On Fri, 2014-04-04 at 11:08 -0400, David Miller wrote:
> From: Dan Williams <dcbw@redhat.com>
> Date: Fri, 04 Apr 2014 09:56:22 -0500
> 
> > netlink doesn't care about IFF_UP when deleting a bridge, so "ip link
> > del br0" works just fine.  The ioctl does care, which means that brctl
> > complains "bridge br0 still up; can't delete it".  Make things
> > consistent by always allowing bridge deletion even if the bridge is up.
> > 
> > Signed-off-by: Dan Williams <dcbw@redhat.com>
> > ---
> > Question: does anyone consider the ioctl behavior API?  Can we change it
> > even though it's been this way forever?  It means that bridges that
> > previously would not have been deleted by brctl will now be deleted.
> > But anyone using /sbin/ip could already delete them.
> 
> Unfortunately I think we're stuck with the existing behavior.

Ok, fair enough.

Dan

Re: [RFC PATCH] bridge: remove !IFF_UP restriction when deleting bridge from ioctl

[Stephen Hemminger]

On Fri, 04 Apr 2014 10:15:50 -0500
Dan Williams <dcbw@redhat.com> wrote:

> On Fri, 2014-04-04 at 11:08 -0400, David Miller wrote:
> > From: Dan Williams <dcbw@redhat.com>
> > Date: Fri, 04 Apr 2014 09:56:22 -0500
> > 
> > > netlink doesn't care about IFF_UP when deleting a bridge, so "ip link
> > > del br0" works just fine.  The ioctl does care, which means that brctl
> > > complains "bridge br0 still up; can't delete it".  Make things
> > > consistent by always allowing bridge deletion even if the bridge is up.
> > > 
> > > Signed-off-by: Dan Williams <dcbw@redhat.com>
> > > ---
> > > Question: does anyone consider the ioctl behavior API?  Can we change it
> > > even though it's been this way forever?  It means that bridges that
> > > previously would not have been deleted by brctl will now be deleted.
> > > But anyone using /sbin/ip could already delete them.
> > 
> > Unfortunately I think we're stuck with the existing behavior.
> 
> Ok, fair enough.
> 
> Dan
> 
> 

That is the way it has been all the way back to 2.4.