Hello Alexey,
On Thu, 26 Jul 2001, Alexey Kuznetsov wrote:
> Heh... The first thing which we need is to fix the bug with holding
> references to dead devices. All the rest is easy and have to be carefully
> audited anyway.
OK, I'm attaching a draft version (against 2.4.7) for
your backlog :) It works, I tested it with permanent devices (eth0) and
with dummy module. nh_dev is now cleared. Now it survives dummy module
unload with dead nexthop in multipath route without printing
emergency messages from unregister_netdevice.
Additionally I touched something for RTN_NAT in fib_semantic_match
The funny part: iproute now does not show the "dead" text for
single-path routes because the nexthop flag is not checked. Someone
have to change the user space if that is fatal :)))
> Alexey
Regards
--
Julian Anastasov <ja@xxxxxx>
dead-2.4.7-1.diff
Description: the static routes now survive device state change
|