g_vfs_done with offset greater than disk size

[Eric Anderson]

Steven Hartland wrote:
> I've been trying to get a problem fixed with a Highpoint 1820a crashing
> out with errors like:
> kernel: IAL: COMPLETION ERROR, adapter 0, channel 2, flags=104
> kernel: ATA regs: error 10, sector count 20, LBA low ff, LBA mid ff, 
> LBA high ff, device 4f, status 51
> which results in the disk being dropped from the array.
>
> A sharp eyed engineer at Highpoint has just spotted the fact that the
> read error reported later is well beyond the end of the array ( 750Gb ):
> kernel: 
> g_vfs_done():da0s1h[READ(offset=535260184576,length=131072)]error = 5
>
> As such is it possible there is a problem in the FS that fsck is not 
> detecting
> which could be causing this behaviour when an rsync ( read only ) is
> performed against it?
>
> If there is indeed a vfs error which fsck is not detecting how would I
> go about:
> 1. finding it
> 2. fixing it

I recently sent a very similar issue to freebsd-geom list, with no 
responses yet (just sent it last week).  I've got a system with ICH6 (or 
ICH7 maybe) controller, and a SATA disk.  I don't believe it's a driver 
issue, and I don't think it's an fs issue, since I was dd'ing an image 
onto the disk, and the image contained linux partitioning and an ext2 
fs.  I think it has to do with the tasting (or re-tasting) of the GEOM 
devices, but that's pretty much a guess.


Eric




-- 
------------------------------------------------------------------------
Eric Anderson        Sr. Systems Administrator        Centaur Technology
Anything that works is better than anything that doesn't.
------------------------------------------------------------------------