[OpenAFS-devel] protection against clock damage

[Derrick Brashear]

For paranoia, this does not change to using interval timers if we  
already weren't. I believe there to be some pthreads implementations  
which do not handle interval timers correctly. Comments to that vein  
as well as on this implementation would be welcome; In particular I am  
worried about performance impact and I think this is as close to  
minimal as possible.

Background:
if your fileserver's clock goes backwards it is possible for e.g. rx  
retransmit events to be lost. This attempts in rx to notice that the  
clock went backwards and arrange for all events to run in a timely  
manner.

/afs/andrew.cmu.edu/usr/shadow/clock3.diff