Subject: Re: status of AFS support?
From: Wes Hardaker (wjhardaker@ucdavis.edu)
Date: Thu Jun 08 2000 - 17:29:24 EDT
>>>>> On Mon, 5 Jun 2000 14:26:54 -0400 (EDT), Stephen Joyce <stephen@physics.unc.edu> said:
Stephen> Anyone know the current status of AFS support in asun's
Stephen> netatalk?
Yes, but its severely broken as is in many ways. Plus, I ended up
using DHX encrypted passwords to a afs pam module.
Here's the README file I've built for myself on the subject:
Problems:
1) netatalk doesn't build shared objects on solaris properly.
a) remove CFLAGS from link lines in etc/{afpd,atalkd,papd}/Makefile
b) fix sys/solaris/Makefile adding appropriate -fPIC --shared, etc.
Patch file: patch.makefiles
2) new openssl versions require you to either have /dev/urandom
(linux) or to have called RAND_seed(STRING, LENGTH) first.
etc/uams/uams_dhx_pam.c doesn't do this before calling
DH_generate_key(). Patch forth coming
(RAND_seed("1234123412341234",16); works but is so very very wrong).
3) etc/afpd/*.c isn't compiled with -DAFS, etc. I'm not sure why as
this completely [swear word dropped for mailling to this list] up
build process isn't easy to follow who's calling what and with
where the make arguments come from. In the mean time, I've added:
-DAFS -I/usr/afsws/include to CFLAGS in etc/afpd/Makefile.
Furthermore, -lucb is needed since a afs library needs sigvec() and
-lc needs to be linked first, since readdir() is screwed up too in
libucb. (symptoms: files returned by readdir() are missing first 2
characters if -lc is linked second).
4) sigh... "make kinstall" doesn't build a driver properly anymore,
so I installed one using "make kinstall" before applying any of the
above patches and it works.
Patch file: patch.afpd
Stephen> I played around a bit and finally got netatalk
Stephen> 1.4b2+asun2.1.4-37b to compile on Solaris 7 with AFS support,
Stephen> but I'm getting an error when trying to load the uam's:
Solaris plugin support is broken as well, as indicated above.
The patch files described above are included below (though not all of
the above problems are fixed in the following two patches). I've
mailed one of them to this list before but no one ever did anything
with it to get it included is the source so I never mailed the rest of
the above descriptions.
The nice thing is that once you actually do get the thing compiled,
it's fairly stable and I'm using the very latest test sources...
-- Wes Hardaker Distributed Computing Analysis and Support University of California at Davis
This archive was generated by hypermail 2b28 : Wed Jan 17 2001 - 14:31:01 EST