[Openswan dev] New/final steps to build KLIPS on 2.6

pw at linuxops.net pw at linuxops.net
Mon Jul 5 10:06:22 CEST 2004


Quoting Michael Richardson <mcr at sandelman.ottawa.on.ca>:

> -----BEGIN PGP SIGNED MESSAGE-----
> 
> 
> >>>>> "Ferdinand" == Ferdinand O Tempel <pw at linuxops.net> writes:
>     Ferdinand> This Makefile builds the required library components in
>     Ferdinand> place, removing the requirement to move the files into
>     Ferdinand> the module build directory. What 
>     Ferdinand> this does when you 'make programs' before or after the
>     Ferdinand> build I haven't 
> 
>   If I understand you right, you have arranged to build kernel
> components during "make programs"
>   (I can't tell from the Makefiles attached)

Nope, that Makefile is to be called seperately. I guess if you call 'make
module' from the openswan root directory it'll try to build it, but fail
miserably. That can be fixed too, of course.

No, what I meant is the situation that it might happen that someone did a 'make
programs' against some kernel version, and tries to build the KLIPS module
against another without a 'make clean' first. The nature of the Makefile I
provided is that it simply links whatever it finds in the libfreeswan and
libcrypto directories, which in that case would be wrong. That's why forcing a
cleanup in those directories before building the module seems like a good idea.

>   That just doesn't work in general.  (consider cross compilation)
>   You get the wrong compilation flags, and maybe even the wrong compiler.

Yes, all true. Task would be to integrate the module build process into the
regular build process. Shouldn't be too hard, just extend whatever is used for
'make module' a bit. Maybe introduce a target called 'module26' or something.

On a side note, I've spent the better part of last saturday playing with NAT-T.
I got to the point that the kernel calls the ipsec_rcv function when it runs
into an UDPinUDP packet. Too bad ipsec_rcv skips the packet as there are some
checks around the code to process it. When I force ipsec_rcv to deal with the
UDPinUDP packet, the kernel oopses. Oops.

I think I hit the limits of my current kernel programming knowledge :-) I'll
post some patches about what I did sofar when I get back home to create them, so
people smarter than I am can have a looksee at it. Maybe I'll stick to what I
know somewhat, and try to integrate the KLIPS build into the regular openswan
build process :-)


----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.


More information about the Dev mailing list