Hi,
The patch is working. I have patched both the local testing setup and
the production pilot. I tcpdumped the interface and got a nice IPv6
withdraw-packet:
No. Time Source Destination Protocol
Info
101 27.955719 2001:db8:1::a500:6777:1 2001:db8:1::a506:5506:1
BGP UPDATE Message
Frame 101 (112 bytes on wire, 112 bytes captured)
Ethernet II, Src: DellComp_8d:1a:c2 (00:06:5b:8d:1a:c2), Dst:
FoundryN_9f:c5:ae (00:12:f2:9f:c5:ae)
Internet Protocol Version 6
Transmission Control Protocol, Src Port: bgp (179), Dst Port: 8172
(8172), Seq: 227, Ack: 155, Len: 38
Border Gateway Protocol
UPDATE Message
Marker: 16 bytes
Length: 38 bytes
Type: UPDATE Message (2)
Unfeasible routes length: 0 bytes
Total path attribute length: 15 bytes
Path attributes
MP_UNREACH_NLRI (15 bytes)
Flags: 0x80 (Optional, Non-transitive, Complete)
Type code: MP_UNREACH_NLRI (15)
Length: 12 bytes
Address family: IPv6 (2)
Subsequent address family identifier: Unicast (1)
Withdrawn routes (9 bytes)
2001:db8:97::/64
0000 00 12 f2 9f c5 ae 00 06 5b 8d 1a c2 86 dd 60 09 ........[.....`.
0010 97 b5 00 3a 06 01 20 01 0d b8 00 01 00 00 00 00 ...:.. .........
0020 a5 00 67 77 00 01 20 01 0d b8 00 01 00 00 00 00 ..gw.. .........
0030 a5 06 55 06 00 01 00 b3 1f ec ad 53 e5 05 42 31 ..U........S..B1
0040 1a e0 50 18 43 80 a9 dd 00 00 ff ff ff ff ff ff ..P.C...........
0050 ff ff ff ff ff ff ff ff ff ff 00 26 02 00 00 00 ...........&....
0060 0f 80 0f 0c 00 02 01 40 20 01 0d b8 00 97 00 00 .......@ .......
Thanks for the patch!
Arnoud Vermeer
On 3/11/09 10:11 AM, Claudio Jeker wrote:
> On Tue, Mar 10, 2009 at 02:46:56PM +0100, Arnoud Vermeer wrote:
>
>> Hi,
>>
>> Elisa and I were looking at the production-pilot logs last night and
>> noticed the following:
>>
>>
>
> I finally found some time to look into this and your dumps. The problem is
> actually with withdraws that are still totaly fucked up. So the following
> diff should fix this issue -- beforehands we just never managed to
> correctly withdraw IPv6 prefixes.
>
> This diff fixes two bugs and makes bgpd do the same paranoic checking as
> juniper and bails out if there is crap after an empty update.
> With this I can correctly withdraw IPv6 prefixes.
