James Carlson wrote:
> Not being able to reset it seems right to me, but being able to set it
> at all sounds like a bug.
>
> The POINTOPOINT flag is a property of the underlying medium. There's
> no reasonable way that this can be specified by the user, and we
> really shouldn't allow it. (I suspect there are other skeletons
> buried here ...)
>
It's pretty easy to duplicate (initial run on snv_89):
# ifconfig dnet0 plumb
# ifconfig dnet0 10.0.0.1 255.255.255.0 up #missing netmask
# ifconfig dnet0
dnet0: flags=200000851<UP,POINTOPOINT,RUNNING,MULTICAST,CoS> mtu 1500
inet 10.0.0.1 --> 255.255.255.0 netmask ff000000
ether 0:c0:ca:13:a4:9e
And, an attempt to reset the interface (forcibly setting the broadcast):
# ifconfig dnet0 down
# ifconfig dnet0 10.0.0.1 netmask 255.255.255.0 broadcast 10.0.0.255 up
ifconfig: SIOCSLIFBRDADDR: dnet0: Cannot assign requested address
Something else interesting to point out, is after the first ifconfig
sets the POINTOPOINT flag, a subsequent ifconfig with a correct netmask
set keeps the POINTOPOINT flag - it seems that the only way to try
'forcing' it into BROADCAST is to explicitly define a broadcast, which
fails (see above).
Steve
--
Yet magic and hierarchy
arise from the same source,
and this source has a null pointer.
Reference the NULL within NULL,
it is the gateway to all wizardry.
_______________________________________________
networking-discuss mailing list
[email protected]
