[OpenAFS] Re: OpenAFS for x86 solaris 8

Curt Freeland curt@grumpy.cse.nd.edu
Sun, 30 Jun 2002 19:07:49 -0500


I have attempted to build Openafs-1.2.{2,3,4,5} client for Solaris 8 
x86 without much success.  I did get a version of 1.2.2 to compile with
gcc 2.95.4 , but the only useful part was the pam_afs.so.1 libraries 
(kernel modules did not work but provided some amusing crash and 
burn sessions).  I used the pam_afs libraries with Arla 0.35.4 and 
limped along.  Arla will compile and "run", but we've had a lot of 
trouble with it crashing the kernel module.  Arla releases since 0.35.4 
compile, but have more serious problems (in my experience).  I've sent 
them some fixes, and crash traces, but never heard back from them...

In order to get Openafs to compile at all on Solaris 8 x86, I had to 
copy several Makefiles (look for Makefile.sunx86_54.in and other files
with the x86_54 in the filename) and edit them with changes as required 
for the later Solaris releases.  This was not trivial task, as I had to 
compare sun4x_5* files for each Solaris release, then when I understood 
the Sparc changes, I had to port those to the hand-made sunx86_58 files.  
I also had to create some .h files that are OS specific.  Using the Sparc
(4x_5*) files will not work due to processor architecture differences (I 
imagine big vs little endian if nothing else causes problems)

Unfortunately, all of my Openafs-1.2.2 work went up in smoke one night 
when the gods of lightening decided that I shouldn't be running Solaris 
x86...the disk wasn't on the dump list (yeah I know.... dumb), and I 
couldn't glue all of the itty-bitty pieces of chips (that I found on 
the floor) back together into a working drive.

As for Solaris 9 x86, at the Usenix Technical Conference, Sun seemed to 
waffle.  My assistant said that they got beat up pretty bad....over half of
the room at one of the BOF's were there to yell at Sun about Solaris 9 x86.
As I understand it, Solaris 9 x86 was available during Beta, but dropped 
before FCS.  My assistant left the BOF with the idea that Sun was 
reconsidering their decision.

Derrick, if you need access to a Solaris 8 x86 system, (single and/or 
multiprocessor) drop me a note.  I can arrange this for you if it helps
me get a working Solaris 8 x86 Openafs.  ;-)

--curt

Curt Freeland (curt@cse.nd.edu) GCIA #0223
Director of Facilities, Computer Science and Engineering Department
323A Cushing Hall,  The University of Notre Dame
Voice: (574) 631-5893 / FAX: (574) 631-9260   <---(NEW AREA CODE!!)