K 10
svn:author
V 3
imp
K 8
svn:date
V 27
2011-01-11T20:52:34.603180Z
K 7
svn:log
V 1118
Implement range locking.  When a range of a volume is locked, any
writes to that range will be deferred.  Range locking either succeeds
right away, or is deferred until all the in-flight writes complete.
Once a range is locked, write requests are queued.  These requests are
resumed when the range is unlocked.  This is intended to be used while
rebuilding a drive, or when doing bad-sector recovery via writing disk
sectors that fail to read properly.  Writes to areas of the volume
that aren't locked are unaffected.

Users of this facility should ensure that they lock small ranges for
short periods to keep too many requests from queueing up (possiblely
forcing a resource shortage that may prevent the range from being
unlocked).  I'm not sure how to trigger this problem, nor what
remediation is necessary to prevent/reduce this problem.

All the hooks are in place.  Nothing that I'm checking in uses this
facility yet, but in testing so far it doesn't affect anything by it
being there unused.

Also, create a callback for the RAID1 transform so I can start using
it for bad sector recovery and then rebuilds.

END