Re: kern/50933: can not enable the interface capabilities at ixg0

[David Young <dyoung%pobox.com@localhost>]

On Thu, Mar 10, 2016 at 01:35:00PM +0000, 6bone%6bone.informatik.uni-leipzig.de@localhost wrote:
> >Number:         50933
> >Category:       kern
> >Synopsis:       can not enable the interface capabilities at ixg0
> >Confidential:   no
> >Severity:       serious
> >Priority:       medium
> >Responsible:    kern-bug-people
> >State:          open
> >Class:          sw-bug
> >Submitter-Id:   net
> >Arrival-Date:   Thu Mar 10 13:35:00 +0000 2016
> >Originator:     Uwe Toenjes
> >Release:        netbsd-7
> >Organization:
> University of Leipzig
> >Environment:
> NetBSD 6bone.informatik.uni-leipzig.de 7.0_STABLE NetBSD 7.0_STABLE (MYCONF7.gdb) #0: Sat Feb 13 12:32:12 CET 2016  root%6bone.informatik.uni-leipzig.de@localhost:/usr/obj/sys/arch/amd64/compile/MYCONF7.gdb amd64
> 
> >Description:
> ifconfig ixg0 reports
> 
> ixg0: flags=8802<BROADCAST,SIMPLEX,MULTICAST> mtu 1500
>         capabilities=fff80<TSO4,IP4CSUM_Rx,IP4CSUM_Tx,TCP4CSUM_Rx>
>         capabilities=fff80<TCP4CSUM_Tx,UDP4CSUM_Rx,UDP4CSUM_Tx,TCP6CSUM_Rx>
>         capabilities=fff80<TCP6CSUM_Tx,UDP6CSUM_Rx,UDP6CSUM_Tx,TSO6,LRO>
>         enabled=42b80<TSO4,IP4CSUM_Rx,IP4CSUM_Tx,TCP4CSUM_Tx>
>         enabled=42b80<UDP4CSUM_Tx,TSO6>
>         ec_capabilities=7<VLAN_MTU,VLAN_HWTAGGING,JUMBO_MTU>
>         ec_enabled=7<VLAN_MTU,VLAN_HWTAGGING,JUMBO_MTU>
>         address: a0:36:9f:26:8e:7c
>         media: Ethernet autoselect
> 
> I try to activate the capabilities.
> 
> I try to activate the capabilities.
> 
> ifconfig ixg0 tso4 tso6 - works
> ifconfig ixg0 ip4csum - works
> 
> ifconfig ixg0 tcp4csum
> ifconfig: SIOCSIFCAP: Invalid argument

On ixg, all of the L4 checksum-offload capabilities (tcp4csum, tcp6csum,
udp4csum, udp6csum) have to be activated/deactivated together, because
that is how the hardware works.  If that is not already mentioned in the
manual page, my apologies.

I would argue that given the limitations of SIOCSIFCAP, today, this is
the least bad way for ixg offload capabilities to work.  I'd be happy to
talk on tech-net about how to bring the ioctl interface up to snuff.

Dave

-- 
David Young
dyoung%pobox.com@localhost    Urbana, IL    (217) 721-9981