Subject: Re: kern/9857: wddone() omits block numbers from soft errors
To: Manuel Bouyer <bouyer@antioche.lip6.fr>
From: John Hawkinson <jhawk@MIT.EDU>
List: netbsd-bugs
Date: 04/11/2000 12:02:54
>> wd0e:  uncorrectable data error reading fsbn 7509888 of 7509872-7509919 (wd0
> bn 12954033; cn 13707 tn 14 sn 36), retrying
>> wd0e:  uncorrectable data error reading fsbn 7509888 of 7509872-7509919 (wd0
> bn 12954033; cn 13707 tn 14 sn 36), retrying
>> wd0: soft error (corrected)

>> Now, the absence of the "soft error" output there would make block
>> 7509888 indistinguishable from 7490056 or 7510128.
>
>Ok, so the text should maybe be changed ('retry succeeded', or something
>like this).

Yes, that sounds good to me, it would be more clear and would save
a documentation update.

> But I don't think it's worth repeating the block number.

If you can convince me that they will always be in-order, than
that seems reasonable.

>Yes. diconnect/reselect and tagged command queueing would make this even
>more attractive.

IDE has tagged command queueing? If so, might there not be a problem
with two bad blocks being read simultaneously and not knowing
which one is associated with the successful retry?

--jhawk