Johan Danielsson writes:
> Greg Oster <oster@cs.usask.ca> writes:
> 
> > There are some serious consequences to doing *ANYTHING* to data on a
> > RAID set before the parity rewrite completes.  If the parity is not
> > correct for a given stripe, and you write real data to a portion of
> > that stripe, old (and *incorrect*) parity will be used to create the
> > new parity, and that new parity will also be *incorrect*.
> 
> Shouldn't the correct parity be calculated on demand then? That is,
> fix the parity of the stripes you write to, before you write to them.

Since having incorrect parity isn't the "normal mode of operation", the 
question becomes "how expensive is the test to see whether or not we
need to fix the parity on demand?"  Given all the other stuff that 
goes on, it actually might not be *that* expensive to do... 

Later...

Greg Oster