On Mon, Jul 24, 2017 at 03:39:12PM +0200, Martin Pieuchot wrote:
> On 23/07/17(Sun) 17:23, Florian Obser wrote:
> > On Sat, Jul 22, 2017 at 02:05:27PM -0000, Christian Weisgerber wrote:
> > > On 2017-07-22, Florian Obser <[email protected]> wrote:
> > [...]
> > # ifconfig vio0
> > vio0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
> >         lladdr 52:54:00:c7:09:1b
> >         llprio 3
> >         groups: egress
> >         media: Ethernet autoselect
> >         status: active
> >         inet 192.168.178.178 netmask 0xffffff00 broadcast 192.168.178.255
> >         inet6 fe80::5054:ff:fec7:91b%vio0 prefixlen 64 scopeid 0x1
> >         inet6 2001:db8::1 prefixlen 64
> >         inet6 2001:db8::2 prefixlen 64
> > 
> > As a work around you could put in inet6 -autoconfprivacy into the
> > hostname.if since the privacy address is in the same /64
> > 
> > Btw. with kernel based slaac this used to work since the
> > kernel didn't go through the ioctl path but just fiddled directly with
> > the interface and skipped rt_ifa_add() which is failing here.
> 
> Because the kernel based slaac was inserting a single prefix.  This was
> a limitation of the code.
> 

that's one way of putting it...

> The correct fix is to enable MPATH on ramdisk.  This is mostly a matter
> of changing all the "#ifdef ART" into "#ifndef RADIX" and deal with the
> ramdisk grow.

All the mpath code in ART is ifndef SMALL_KERNEL as far as I can work out,
so this isn't helping any.

However, if I put

option  ART

into arch/amd64/conf/RAMDISK_CD

solves naddy's second problem, not being able to ping6 things that are
not on-link. This suggests some bit rot in the radix code...

I'd rather not hunt for that but enable ART at least where we have
INET6, since I got slaacd onto those installers without too much space
problems it probably just fits. Note that none of the floppies have
INET6. 

-- 
I'm not entirely sure you are real.

Reply via email to