RFC: GEOM MULTIPATH rewrite

[John-Mark Gurney]

Alexander Motin wrote this message on Tue, Nov 01, 2011 at 15:05 +0200:
> > 2. In active/active mode do you do anything to handle possible
> >    reordering? Ie. if you have overlapping writes and send both of them
> >    using different paths, you cannot be sure that order will be
> >    preserved. Most of the time that's not a problem, as file systems
> >    rarely if at all send overlapping writes to device, but this is weak
> >    assumption.
> 
> No, I don't. I have doubt that it is sane to send even dependent I/O
> simultaneously without waiting for completion, not speaking about
> overlapping. When most of present devices support command queuing and so
> officially justify reordering simultaneous commands in custom way, I am
> not sure why above layers should be more strict, especially in cases
> when it is problematic. If somebody have ideas why and how to implement
> it, I am ready to discuss.

I know that phk and others have an idea what the contract for writes like
this, but I just checked geom(4) and the IO section doesn't describe it..

I believe that you should not submit overlapping writes unless it's
preceded or is an _ORDERED write, and that reads can be satisifed w/
stale data if it is submitted after a write of the same location, but
it's not in geom(4).   Hmm... turns out BIO_ORDERED isn't even documented
in geom(4)...

I'm willing to put some of this in the man page if someone comes up w/
a good list of points.  Are the ones I listed above enough?

-- 
  John-Mark Gurney				Voice: +1 415 225 5579

     "All that I will do, has been done, All that I have, has not."