[OpenAFS] OpenAFS & Linux kernel 2.6: please have a trackable bug report / task ?

[Todd M. Lewis]

Horst Birthelmer wrote:
>> Is it complicated to rewrite those 2 hooks just for AFS (a little
>> kernel-patch) or is there an OpenAFS-policy demanding not to change
>> the kernel-sources at all?
> 
> 
> It's neither complicated nor is there any 'codec' (at least I'm not 
> knowing of). BUT how many people you know they would install some
> 'dubious' kernel patch (that was the way the ppc64 patch was called
> by some people) just to run AFS??

It isn't "just to run AFS", it's to run 2.6.  AFS is not an option for 
me; if I can't run AFS/OpenAFS there's no point in my turning on the 
computer.

Sure, I'd prefer to run OpenAFS on the latest stock kernel. If I can't 
do that, then if I had the choice between running OpenAFS on a modified 
2.6 or a stock 2.4, well, at least I've got a choice.

As it stands, there's nothing to choose: run 2.4, period.

It may be instructive to play with different ways of incorporating 
OpenAFS into 2.6.  On the other hand, it's hard to imaging a method of 
making PAGs behave correctly other than through the groups mechanism. 
The semantics are just right. So right in fact, that it should be 
generalized. Don't think of PAGs as a bag-on-the-side of the groups 
mechanism, but rather implement group inheritance through a generalized 
PAG mechanism.

Linus likes to generalize whenever it makes sense, so AFS PAGs will have 
to play nice with NFS4 PAGs to get into his kernel. It occurs to me that 
the inherited process uid is just a poor man's (though efficiently 
implemented) credential that should be handled through a real PAG 
mechanism like any other cred.  Maybe the new Linux generalized PAG 
mechanism will need to handle local credentials the same as remotely 
authenticated creds.  Just a thought...
-- 
    +--------------------------------------------------------------+
   / Todd_Lewis@unc.edu  919-962-5273  http://www.unc.edu/~utoddl /
  /      A man's home is his castle, in a manor of speaking.     /
+--------------------------------------------------------------+