Stephen Hemminger <[email protected]> writes:

> On Thu, 09 Dec 2010 11:09:34 -0800
> [email protected] (Eric W. Biederman) wrote:
> So let's please keep this a line that we can easily remove.  It isn't
> like interfaces coming up and down are a fast path where every cycle
> counts.  We just need to be reasonably efficient.

> No but since removing address propagates up to user space daemons
> like Quagga please analyze and fix the problem, don't just look
> for band aid.

You fix the problem.  You introduced the regression, and you didn't test
that keeping addresses actually worked.  This is not the first patch
that has been applied to fix regressions in this area.

I introduced a targeted revert of your broken change that only preserves
the one address people are least likely to change.

I know people are not downing the ipv6 loopback interface in practice
and bringing it back up because in practice on running systems because
that breaks ipv6 networking.  So quagga should not see this issue
in practice.

Right now misplaced perfectionism is being a huge enemy of creating
a kernel that actually works.

Eric

_______________________________________________
stable mailing list
[email protected]
http://linux.kernel.org/mailman/listinfo/stable

Reply via email to