[hipl-users] Re: beet kernel patch summary
- From: Miika Komu <miika@xxxxxx>
- To: hipl-users@xxxxxxxxxxxxx
- Date: Thu, 19 Apr 2007 19:45:12 +0300 (EEST)
On Thu, 19 Apr 2007, Henderson, Thomas R wrote:
(Kazunori and Yoshifugi, there is a question for you in the middle, please
reply if you have some time)
-----Original Message-----
From: Miika Komu [mailto:miika@xxxxxx]
Sent: Wednesday, April 18, 2007 10:34 PM
To: hipl-users@xxxxxxxxxxxxx
Subject: [hipl-users] Re: beet kernel patch summary
On Wed, 18 Apr 2007, Henderson, Thomas R wrote:
Hi all,
I was poking around your web site today looking for this
(interfamily)
patch; can you send a pointer? I downloaded the nightly tarball but
found one patch that was labeled BROKEN in the 2.6.19.0 directory.
The latest working patches are located in patches/2.6.17.14
directory.
2.6.11+ are still work in progress. More specifically, 2.6.11 fails
to set-up BEET mode SPs properly.
I am not sure about the above statement. 2.6.17.14 works but 2.6.11+
are still work in progress?
Sorry, I meant 2.6.19.0+, not 2.6.11+
I am not sure that I want patches that are pre-2.6.19. I see that
2.6.20.7 (today's version) has BEET, and so does 2.6.19+ I understand.
Is there a separate patch on top of 2.6.19+ that does the interfamily
piece of BEET? Where is that patch, and is it stable?
There is no stable patch yet for 2.6.19 and above. I am currently working
on 2.6.19 patches (on the request of guys in Achen) and then moving
towards 2.6.20. Feel free to participate the kernel fun if you have
resources :)
More generally, I want to explore again a Linux implementation that
requires no patches to run HIP, yet has kernel-space data processing.
Loadable kernel modules are OK, as long as CONFIG_*=m in the typical
distribution. This is to facilitate deployment; I think that asking
testers to rebuild their kernel is too cumbersome. In the absence of
something like BEET module, we would resort to custom iptables mangle
modules.
We are providing kernel deb and rpm packages in our latest release for
2.6.17.14. When we get the next kernel versions to work, we will surely
provide new binary kernels at the infrahip website.
- is it true that 2.6.19+ as is requires no patches to do base/esp/mm
HIP processing, if one does not want to do interfamily roaming?
Yes, although I have not tested this myself. I am mostly interested on the
interfamily patch myself because we want to use IPv4 addressing
for real networks and we don't support LSIs in our implementation,
although we provide different IPv4 support through opportunistic mode.
- will it be true that 2.6.2? (with interfamily) will be the same?
Not for 2.6.20. Maybe for 2.6.21, but I don't know the release schedule.
We gave the latest interfamily to Yoshifugi and Kazunori for review. Could
either of you tell about the status of the review?
That is, even if BEET is supported in the future, will there still be a
need to apply some patch like policy-sleep to get it to work right?
No. 2.6.21-rc7 includes already a patch that just drops packets until SAs
are established. This is suitable for most cases as packet tranmissions
are handles usually at the application layer for UDP and at the transport
layer with TCP.
I understand that hipmod will be still experimental (a patch) for now,
but that seems to be a separable piece.
Yes. It implements the native API draft (but is somewhat out-of-date):
http://www.ietf.org/internet-drafts/draft-ietf-hip-native-api-01.txt
--
Miika Komu http://www.iki.fi/miika/
Other related posts:
