mpt(4) hickups (was: RAIDframe component scrubbing)

[Edgar Fuß <ef%math.uni-bonn.de@localhost>]

> I have been working on improving error handling for the mpt(4) driver.
[...]
> The fixes address issues when the IOC hickups and help to get the IOC back
> on-line if it goes out to lunch.
The problem I faced (second time today, but I don't have logs -- see below) was

Sep  7 03:21:38 donau /netbsd: sd3(mpt0:0:3:0): command timeout
Sep  7 03:21:38 donau /netbsd: mpt0: timeout on request index = 0x8a, seq = 
0x1655bf37
Sep  7 03:21:38 donau /netbsd: mpt0: Status 0x00000000, Mask 0x00000001, 
Doorbell 0x24000000
Sep  7 03:21:38 donau /netbsd: mpt0: request state: On Chip

Is this the type of problem your fix adresses?
What worries me is that today, I got the same error again, on the same disc.
So I'm afraid it might be more of a cabling or disc electronics problem.


This time however, I was fast enough rushing down to the server room and 
halting the machine before it could declare sd3 failed. So I only had about 
one hour of parity re-build (thanks to parity maps!) instead of twenty hours 
of reconstruction.