Re: IGMPv3 hot interface detach panics?

Weongyo Jeong wrote:
> ...
> This is one I have got from "Paul B. Mahol" <onemda_at_gmail.com>
> yesterday and I think he might help you to get full backtrace:
>   
Thanks for this.  I really do need a full backtrace to work out what to 
go on, however,
as it looks like the compiler is inlining static functions here.

Is IGMPv2 in use on this network? The stack will default to IGMPv3, it 
is possible that
the call to igmp_ifdetach() needs to be forced to happen before 
in_purgemaddrs() causes
the IP stack's reference to the in_multi to be freed.

Are any multicast applications in use?

Sam is seeing a very different condition with 802.11 VAPs, even though 
the symptoms are similar.

I appreciate that it isn't always possible to test with all combinations 
of network drivers
that people may be using, vlans vaps etc. and some of them do use parts 
of the network stack
in different ways, i.e. stacking the use of otherwise refcounted 
structures, etc.

This looks like it is coming from USB2 and NDIS in some way. I am not at 
all familiar with
those subsystems and don't use them, so help from others would be very 
appreciated,
however I will try my best to help fix, the root cause may be more general.

thanks,
BMS