On Feb 25, 2005, at 3:05 PM, Hartmut Reuter wrote:
Michael Niksch wrote:Michael: You might want to try my patch and see how much things improve for you.I'll try that.Given that the thing works better with larger stack I'd assume that some assumption that's valid on AIX4 isn't valid now which causes the thing to overwrite memory in kernel space and things goes downhill from there.I compiled 1.3.79 also on AIX 4.3.3, where I have been using 1.2.10 without problems for a long time now. With 1.3.79, however, I continue to have the problem that klog dies with 'Illegal instruction' as it did already with previous version of 1.3.xx. If I use klog from 1.2.10, I can obtain a token, but after doing so, the machine dies just like the AIX 5.2 machine, even though kdb stat (attached) looks different.
There are probably two problems:
1) klog does not work unless you compiled it without -O (I think specially the des libraries, but better if you do it for all). This is obviously a bug of the IBM compiler.
2) afs_pioctl.c had been changed between 1.2 1nd 1.3 at some point to support the giant tickets created by Microsoft's active directory (12000 byte). AIX has a problem with this. Therefore I came back to the old code which limits the ticket size to e reasonable value.
OpenAFS doesn't compile the des library with optimization on.
The version Michael is using has code without that large ticket allocation At least he is supposed to use that, and I'm pretty sure Niklas is using the same since he has some source from Harald Barth which we developed on my visit to Stockholm for the Hackathon.
But for ensuring that, could you take a look into the file src/rxkad/rxkad.p.h and see what you have defined for MAXKTCTICKETLEN, Michael??
Horst
_______________________________________________ OpenAFS-devel mailing list [email protected] https://lists.openafs.org/mailman/listinfo/openafs-devel
