Feature suggestion: immediate repair attempt on failed read?

On 13/11/2008 23:04, Mikael Abrahamsson wrote (in the "OT: Tips for good 
hard drives for a home server" thread):

Looking at the md code in linux, would it be possible to do some work 
around here in that if a drive read operationstalls, let's try to read 
from the other drives and then re-write the block on the drive that 
stalled?

I'd like to second this suggestion; my coding skills are WAY too rusty 
to attempt it myself, but it sounds like an excellent idea.

We could also have a timeout setting so that if a read stalls, at least 
the application can get its data (by reading the other discs) and the md 
subsystem can carry on waiting for the read and/or just doing the 
rewrite in the background.

What happens at the moment - could one slow read cause the array to be 
switched into degraded mode immediately?

Cheers,

John.