In article <20041004002244.GA25563@che.ojctech.com>,
	dyoung@pobox.com (David Young) writes:
> A local NetBSD user tells me this system is only able to write the
> disk at about 600kB/s.  Is there any known bug in ahc for 1.6.2 that
> would cause this?  Any reason to think 2.0_BETA improves the situation?

The driver in NetBSD 2.0 is practically a complete rewrite of the one
in NetBSD 1.6.X (both imported from FreeBSD), so yes, it's likely to
do quite a bit better. The man page still says that U160 is limited to
40 or 80 MB/s, but I'm not sure it that's still true:

 # dmesg | grep ahc
 ahc1 at pci1 dev 3 function 0: Adaptec aic7892 Ultra160 SCSI adapter
 ahc1: interrupting at irq 9
 ahc1: aic7892: Ultra160 Wide Channel A, SCSI Id=7, 32/253 SCBs
 scsibus0 at ahc1: 16 targets, 8 luns per target

 # dmesg | grep ^sd
 sd0 at scsibus0 target 0 lun 0: <IBM-PSG, DDYS-T09170N  M, SA2A> disk fixed
 sd0: 8678 MB, 15110 cyl, 6 head, 196 sec, 512 bytes/sect x 17774160 sectors
 sd0: sync (12.50ns offset 63), 16-bit (160.000MB/s) transfers, tagged queueing
 sd1 at scsibus0 target 1 lun 0: <IBM-PSG, DDYS-T09170N  M, SA2A> disk fixed
 sd1: 8678 MB, 15110 cyl, 6 head, 196 sec, 512 bytes/sect x 17774160 sectors
 sd1: sync (12.50ns offset 63), 16-bit (160.000MB/s) transfers, tagged queueing

Then again, cat'ing the raw partition to "/dev/null" seems to be limited to
no higher than 40 MB/s. That's still a lot better than your user's 600kB/s.


Frederick