Subject: Re: Dealing with bad blocks on fixed disks
To: Darren Reed <>
From: Manuel Bouyer <>
List: tech-kern
Date: 04/16/2003 21:48:57
On Wed, Apr 16, 2003 at 10:04:43AM +1000, Darren Reed wrote:
> In some email I received from Jason Thorpe, sie wrote:
> > 
> > bp->b_bcount is a byte count, not a block count.  So, the calculation 
> > of maxblk is incorrect.
> Yes, I found that out and corrected it.
> My last remaining thoughts on this "feature" are that it would be nice
> if in cases where an error has been experienced for b_bcount > DEV_BSIZE
> for the kernel to be able to schedule reading each of the blocks itself
> and use the feedback from that to populate the "bad sector" list.  If
> this were possible, then the list would only ever contain known problematic
> blocks, without any incrimination of others within the same range being
> fetched.

The read/write backed of wd on IDE, ata_wdc.c already does this.
This should probably be moved to wd.c, so that it works with others
backend too.
One more thing on my todo list :)

Manuel Bouyer <>
     NetBSD: 24 ans d'experience feront toujours la difference