On Mon, 3 Feb 2014, NeilBrown wrote:

We only try to fix a read error when all reads failed rather than when any
read fails.
Most of the time there is only one read so this makes no difference.
However for 'check' and 'repair' we read all devices so the current code will
only try to repair a read error if *every*  device failed, which of course
would be pointless.

Is this the same problem I hit and documented in my thread a few weeks 
back? ("read errors not corrected when doing check on RAID6")

Is the same problem is valid for RAID6 as well and it fixes that as well?

-- 
Mikael Abrahamsson    email: swmike@swm.pp.se