Subject: Re: CVS commit: src/sys/dev/scsipi
To: Jason Thorpe <>
From: Brett Lymn <>
List: source-changes
Date: 08/08/2005 09:27:00
On Sun, Aug 07, 2005 at 10:28:49AM -0700, Jason Thorpe wrote:
> I sure wish this patch hadn't been checked in.  Obviously the correct  
> thing would have been to make a generic subsystem, rather than one  
> that only works for st(4).

Actually, the changes for doing other tapes is not much.  The stats
are kept in their own linked list and the driver just keeps a pointer
to the appropriate linked list entry so gathering stats in other
drivers is a matter of linking in a new stats structure on attach and
updating the stats fields when i/o is done.  After doing that the
stats would be picked up.

> Either extending disk stats to also measure meaningful stuff for  
> tapes, or making a generic tape stats framework.

I thought that it was generic enough - if I read the comments
correctly in st.c that handles ATAPI drives as well as true SCSI
ones.  I didn't want to munge them in with disks because they are not
disks which may be meaningful in some circumstances (I cannot think of
an example but I was just being cautious.

Brett Lymn