[OpenAFS] Re: "unable to authenticate to AFS" error- linux distribution specific?

[Christof Hanke]

I've debugged my broken klog (SLES9 on openafs-1.3.74)  using -O2 and 
found out that in authclient.c:ka_Authenticate() the char *label gets 
corrupted, when KAA_AuthenticateV2 is called (as Derrick pointed out 
earlier).
I have no fix for that, but I found out when you declare the char *label 
as static, the klog works (at least for me).
in src/kauth :
diff authclient.c authclient_new.c
526c526
<     char *ans_label;
---
 >     static char *ans_label;

What this means in detail, I do not know, but I guess it prevents some 
overoptimization.

Hope this helps anyway.

Christof Hanke

Joseph Kiniry wrote:

> On Wed, 17 Nov 2004 10:33:16 -0500 (EST) Derrick J Brashear wrote:
>
> > On Wed, 17 Nov 2004, Thomas Bertram wrote:
> >
> > > Under SuSE 9.1 (2.6 Kernel) I encountered the same problem. 
> However, on two
> > > installations with SuSE 9.0 with an own compiled 2.6.7 kernel and 
> a compiled
> > > AFS 1.3.71 client,  klog is working properly -- also with the same 
> old
> > > server.
> >
> > Someone claimed compiling the entire kauth directory without 
> optimization
> > "fixed" the problem. Ideally someone would figure out which file, or
> > better yet which function, needs to be compiled this way, so the 
> problem
> > could be fixed for real. no one has.
>
> In setting up my new cell here at UCD (my old cell is at the 
> University of Nijmegen), I have attempted a build of OpenAFS 1.3.74 on 
> a clean Fedora Core 3 install (fully updated) and I experienced this 
> problem as well.  Additional system information: kernel is stock 
> "Linux secure" 2.6.9-1.667, gcc (GCC) 3.4.2 20041017 (Red Hat 
> 3.4.2-6.fc3), glibc-2.3.3-74, and SELinux is running in the default 
> targeted policy mode.
>
> I've setup the initial server and everything seems to be kosher except 
> I cannot authenticate.  Seeing the earlier post about turning off 
> optimization, I attempted to replicate their reported success.  I 
> initially tried to only recompile klog and its dependencies with -O0, 
> but that resulted in a klog that would segfault.  Thinking I wasn't 
> aggressive enough, I recompiled the whole of OpenAFS with -O0.  That 
> build performs exactly as the -O2 version (everything works but 
> klog).  Thus, I suspect there is something more subtle going on than 
> just an optimizer issue.
>
> Any suggestions would be welcome.  I'm a very experienced AFS guy (I'm 
> an old OFS employee) and developer, so I'm happy to hack to help.
>
> Thanks,
> Joe
> ---
> Joseph R. Kiniry                                ID 78860581      ICQ 
> 4344804
> University College Dublin / National University of Ireland, Dublin
> KindSoftware, LLC                             
> http://www.kindsoftware.com/
> Board Chair: NICE                                    
> http://www.eiffel-nice.org/
>
> _______________________________________________
> OpenAFS-info mailing list
> OpenAFS-info@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-info