part of commit 0dcffac1 upstream
(the upstream multislice patch contains the same fix within
 myri10ge_alloc_slices() which does this for every slice)

Even if we don't have multiple slices in myri10ge in 2.6.26,
we already use some multislice-aware routines that need the
slice state pointers to mgp and dev to be valid.

This patch fixes a regression where configuring the interface up
would oops in myri10ge_allocate_rings() when using ss->mgp
(and later when using ss->dev).

Oops reported and patch tested by Lukas Hejtmanek at
http://lkml.org/lkml/2008/7/23/101Signed-off-by: Brice Goglin <brice@myri.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>