[OpenAFS-devel] Re: OpenAFS 1.4.7pre1 Oops

[Marc Dionne]

> A quick grep suggests that this is used primarily by osi_probe.c, both 
> as a symbol to use to try and find the syscall table, and to define 
> probe limits. Could we turn it into a weak reference, and just disable 
> syscall probing where it doesn't exist? (I'd be highly surprised if 
> distribution kernels which don't have init_mm have a writable syscall 
> table).
> 
> S.

That's pretty much the logic for the original patch I submitted for 2.6.25, where I ifdef'ed out the probing 
code based on init_mm not being exported.  As I said in the comments (RT 88000), I'm not sure that's the most 
elegant solution, but it seemed pointless to try to tweak the probing code for new kernels where the syscall 
table is probably not writable and keyrings are available.

Marc