[OpenAFS-devel] take 2: openafs 1.3.64 on linux 2.6 -- what am I missing now?

[Matthew Miller]

Okay, figured out not to do "--enable-redhat-buildsys". Still having
problems, though. I'm building on Fedora Core 2 testN, where N is somewhere
between 2 and the current development tree. (I'm using the very latest
kernel package, kernel-2.6.5-1.358.)

First, I needed to make some minor tweaks to get the thing to build at all
-- in src/rx/LINUX/rx_knet.c, the sock_create call is missing its final
parameter. I'm just guessing, so I put in a 0. This may be horribly wrong,
but at this point, I'm still trying to get the thing to _build_, and I can
worry about _build correctly_ after that. :)

Then, src/afs/LINUX/osi_module.c seemed to want a 
"#if !defined(AFS_LINUX24_ENV)" put around the "sys_settimeofdayp = 0;".

Then, I needed to have a kernel tree to which I had write access -- it wants
to make some symlinks there. (That doesn't seem so good....)

And perhaps most troubling, the whole thing didn't seem happy at all unless
I made a change to ./configure -- there's no "i386_linux26" section, so I
created one by just copying the i386_linux24 part. This is probably not
sufficient, and is probably what leads to my current state of affairs:

With the above-mentioned changes, everything builds, but the kernel modules
generated are 'libafs-2.6.5-1.358custom.o' and
'libafs-2.6.5-1.358custom.mp.o'. They're not the fancy new .ko format, and
insmod tells me:

 error inserting 'libafs-2.6.5-1.358custom.o': -1 Invalid module format

which seems to go along with the theory that it's still trying to do
something the 2.4 way.

I admit that this is the first time I've tried to build an external module
for the 2.6 kernel. I hope that the problem is something trivial that I'm
missing. In any case, your help is much appreciated.

-- 
Matthew Miller           mattdm@mattdm.org        <http://www.mattdm.org/>
Boston University Linux      ------>                <http://linux.bu.edu/>