Subject: Re: kern/2679: addition of seek stat
To: None <firstname.lastname@example.org>
From: Jason Thorpe <email@example.com>
Date: 08/09/1996 11:37:23
On Thu, 8 Aug 1996 02:41:30 -0700 (PDT)
John M Vinopal <firstname.lastname@example.org> wrote:
> >Synopsis: addition of seek stat
> dk_seek is not updated in subr_disk.c
> A seek could be defined as any xfer or zero byte xfer, in this
> case the following patch is appropriate.
> *** subr_disk.c Sun Mar 17 04:28:11 1996
> --- /tmp/a Thu Aug 8 02:37:14 1996
> *** 385,390 ****
> --- 385,391 ----
> diskp->dk_bytes += bcount;
> + diskp->dk_seek++;
This is not really correct. May disks do seeks independent of transfers
(this is true for HP-IB disks, at least, and SMD, if I'm not mistaken).
Thus, adding this would skew the seek metric, since these disks call
disk_busy() to calculate time spent seeking (and call disk_unbusy() with
a transfer length of 0 upon seek completion).
Charles Hannum mentioned to be in passing some months ago that an implied
seek in disksort() might be appropriate (when disksort() rewinds to the
beginning of the queue). However, this too would skew the metrics of
disks that do independent seeks.
Probably the best place to put this is in the individual drivers that
don't do independent seeks or the drivers (like the sparc floppy driver)
which do implied seeks.
-- save the ancient forests - http://www.bayarea.net/~thorpej/forest/ --
Jason R. Thorpe email@example.com
NASA Ames Research Center Home: 408.866.1912
NAS: M/S 258-6 Work: 415.604.0935
Moffett Field, CA 94035 Pager: 415.428.6939