mpd on sparc64 problem (maybe netgraph involved)

[Julian Elischer]

Volodymyr Kostyrko wrote:

> Can't make use of mpd on sparc64. After going up link goes down almost 
> instantly.
>
> On sparc64(client):
> Aug 29 16:32:04 blade mpd: [pppoe0] IFACE: Up event
> Aug 29 16:32:10 blade mpd: [pppoe0] PPPoE connection closed
> Aug 29 16:32:10 blade mpd: [pppoe0] device: DOWN event in state UP
> Aug 29 16:32:10 blade mpd: [pppoe0] device is now in state DOWN
>
> On i386(server):
> Aug 29 19:33:19 nora mpd: [pppoe2] LCP: magic number is wrong: 
> 0xc4b86a38 != 0x00000000
> Aug 29 19:33:19 nora mpd: [pppoe2] LCP: received an invalid magic number
> Aug 29 19:33:19 nora mpd: [pppoe2] LCP: LayerFinish
> Aug 29 19:33:19 nora mpd: [pppoe2] LCP: LayerStart
> Aug 29 19:33:19 nora mpd: [pppoe2] LCP: state change Opened --> Starting
> Aug 29 19:33:19 nora mpd: [pppoe2] LCP: phase shift NETWORK --> DEAD
>
> It goes on and on, and nothing changes...
>
> > uname -a
> FreeBSD blade.lan 6.1-STABLE FreeBSD 6.1-STABLE #16: Wed Aug  9 
> 12:14:57 UTC 2006     arcade at blade.lan:/usr/obj/usr/src/sys/BLADE  
> sparc64
>
> Mpd version: mpd-3.18_4

0xc4b86a38  looks like a kernel address


use tcpdump on the ethernet interace with 2 x i386 machines and with the 
failing setup and see what the difference is..
quite possibly sparc64 is getting the pppoe header differently due to 
allignment or something..