[OpenAFS] OpenAFS on Linux 2.5.x

[Derrick J Brashear]

On Wed, 16 Apr 2003, chas williams wrote:

> In message <Pine.LNX.4.53L-031.0304151607430.14506@trafford.andrew.cmu.edu>,Der
> rick J Brashear writes:
> >> You mean in terms of inserting ourselves into the sys_call_table?  Or
> >> are you referring to something else?
>
> i think the syscall table can be handled easily enough via a different
> interface.  did the kernel people have any suggestion as the to 'proper'
> scheme?

i know we talked about it here before. making all of the entry points for
daemons being kernel threads means we could reduce startup entry to an
ioctl done on a file in /proc, or somesuch. then the afsdb handler also
needs a way to pass messages.

> >That's one example. Something needs to happen for PAGs also, and us
> >hooking the setgroups/getgroups calls is really the wrong answer.
>
> again, did they have any suggestions?  how about tampering with struct
> task directly (that would probably make the kernel folks real happy of
> course :) -- or getting a field added to struct task so we can hang the
> afs credentials from it?

right, well, i made a side contact who was going to look into this, but
last i knew, no progress.

tampering with task_struct in addition to being ugly seems not likely to
win us any friends, but more importantly, perhaps (or perhaps not) it
seems not useful: there appear to be no fields not used that we can
piggyback on.