inet6 broken pretty badly in -current post 20080828

[S.P.Zeidler]

Hi,

somewhen between 20080809 and 20080828 inet6 broke thusly:
- rtsol fails
- ping6 <linklocal-of-host> from another host on local link doesn't work,
nor do connects to open tcp6 ports on linklocal.
- after setting global v6 address and default gate manually, ping6 to that
address won't work, neither will connects to open tcp6 ports for that address.
But ping6 and other connects to linklocal suddenly start working.
rtsol will still fail.
- setting a further IPv6 address doesn't make the middle one work
- dns will work (needs the global address)
- outgoing connections work (obviously also not using linklocal)

IPv4 it completely fine, otherwise there'd be more noise about it I bet :)

Anyone have a bright idea? I'll build myself kernels of dates between 0809
and 0828 to pinpoint the change that caused this, but if someone has an
'oh, ooops, I forgot' moment that might spare me some work :-P

BTW, sys/net/rtsock.c can't be the culprit because I had still 1.109 in
the first kernel that gave me this problem.

regards,
spz

[Matthias Scheler]

Post by S.P.Zeidler
- rtsol fails

I'm not using rtsol.

Post by S.P.Zeidler
- ping6 <linklocal-of-host> from another host on local link doesn't work,
nor do connects to open tcp6 ports on linklocal.

I just tested IPv6 on two machines:
1.) Sun ULTRA60 with wm(4) interface running NetBSD/sparc64.
2.) HP Proliant ML110 G4 with a bge(4) interface running NetBSD/amd64.
Both machines run NetBSD 4.99.72 built from 2008-08-29 sources.

There are absolutely no problems on the ULTRA60.

IPv6 connections to the Proliant don't work unless the Proliant has sent
IPv6 packets to the remote host first:

1.) Power Mac G5 to Proliant:

***@excalibur:~>ping6 fe80::21c:c4ff:fe5f:d9b%en0
PING6(56=40+8+8 bytes) fe80::216:cbff:fe37:999a%en0 -->
fe80::21c:c4ff:fe5f:d9b%en0
^C
--- fe80::21c:c4ff:fe5f:d9b%en0 ping6 statistics ---
3 packets transmitted, 0 packets received, 100% packet loss

2.) Proliant to Power Mac G5:

***@lyssa:~>ping6 fe80::216:cbff:fe37:999a%bge0
PING6(56=40+8+8 bytes) fe80::21c:c4ff:fe5f:d9b%bge0 -->
fe80::216:cbff:fe37:999a%bge0
16 bytes from fe80::216:cbff:fe37:999a%bge0, icmp_seq=0 hlim=64
time=0.657 ms
16 bytes from fe80::216:cbff:fe37:999a%bge0, icmp_seq=1 hlim=64
time=0.406 ms
^C
--- fe80::216:cbff:fe37:999a%bge0 ping6 statistics ---
2 packets transmitted, 2 packets received, 0.0% packet loss
round-trip min/avg/max/std-dev = 0.406/0.532/0.657/0.177 ms

3.) Power Mac G5 to Proliant after (2):

***@excalibur:~>ping6 fe80::21c:c4ff:fe5f:d9b%en0
PING6(56=40+8+8 bytes) fe80::216:cbff:fe37:999a%en0 -->
fe80::21c:c4ff:fe5f:d9b%en0
16 bytes from fe80::21c:c4ff:fe5f:d9b%en0, icmp_seq=0 hlim=64
time=0.578 ms
16 bytes from fe80::21c:c4ff:fe5f:d9b%en0, icmp_seq=1 hlim=64 time=0.5
ms
^C
--- fe80::21c:c4ff:fe5f:d9b%en0 ping6 statistics ---
2 packets transmitted, 2 packets received, 0% packet loss
round-trip min/avg/max = 0.500/0.539/0.578 ms

Post by S.P.Zeidler
Anyone have a bright idea?

Yes, it is a problem in bge(4) caused by this commit:

http://mail-index.netbsd.org/source-changes/2008/08/25/msg209337.html

After downgrading src/sys/dev/pci/if_bge.c to revision 1.150 I can no
longer
reproduce the problem.

Kind regards

[Matthias Drochner]

There seems to be a multicast reception problem
in the bge driver. If the interface is in promisc
mode, rtsol works.

best regards
Matthias




-------------------------------------------------------------------
-------------------------------------------------------------------
Forschungszentrum Juelich GmbH
52425 Juelich

Sitz der Gesellschaft: Juelich
Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
Vorsitzende des Aufsichtsrats: MinDir'in Baerbel Brumme-Bothe
Geschaeftsfuehrung: Prof. Dr. Achim Bachem (Vorsitzender),
Dr. Ulrich Krafft (stellv. Vorsitzender), Prof. Dr. Harald Bolt,
Dr. Sebastian M. Schmidt
-------------------------------------------------------------------
-------------------------------------------------------------------

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de

[Matthias Scheler]

Post by Matthias Drochner
There seems to be a multicast reception problem
in the bge driver. If the interface is in promisc
mode, rtsol works.

I've reported this in PR kern/39439.

Kind regards