Date: Wed, 29 Jul 1998 19:47:13 -0300 (EST) From: Carlos Carvalho <carlos@fisica.ufpr.br> To: aic7xxx Mailing List <AIC7xxx@FreeBSD.ORG> Subject: tagged queuing problem Message-ID: <13759.42609.481448.897476@hoggar.fisica.ufpr.br>
next in thread | raw e-mail | index | archive | help
I reported some days ago that some disks were not using the
prescribed queue depth. I've done some more checks with scsiinfo and
found nothing different between the disks that use tagged queuing and
those that don't, so I suspect a driver problem. Here are the details.
>From /proc/scsi/aic7xxx, first controller:
Default Tag Queue Depth: 24
Tagged Queue By Device array for aic7xxx host instance 0:
{32,24,32,255,255,255,255,255,255,255,255,255,255,255,255,255}
Actual queue depth per device for aic7xxx host instance 0:
{1,24,32,1,1,1,1,1,1,1,1,1,1,1,1,1}
^
Note that /dev/sda doesn't use tagged queueing! For the second
controller:
Default Tag Queue Depth: 24
Tagged Queue By Device array for aic7xxx host instance 1:
{32,32,255,255,255,255,255,255,255,255,255,255,255,255,255,255}
Actual queue depth per device for aic7xxx host instance 1:
{32,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}
^
Here /dev/sde isn't using tagged queueing, but /dev/sdd is.
>From /proc/scsi/scsi:
/dev/sda
Host: scsi0 Channel: 00 Id: 01 Lun: 00
Vendor: SEAGATE Model: ST34501W Rev: 0013
Type: Direct-Access ANSI SCSI revision: 02
/dev/sdd
Host: scsi1 Channel: 00 Id: 00 Lun: 00
Vendor: SEAGATE Model: ST34501W Rev: 0013
Type: Direct-Access ANSI SCSI revision: 02
/dev/sde
Host: scsi1 Channel: 00 Id: 02 Lun: 00
Vendor: SEAGATE Model: ST34501W Rev: 0013
Type: Direct-Access ANSI SCSI revision: 02
These are Cheetah's, 10000rpm ultra-wide disks. They should be
identical. /dev/sda is in the middle of a mixed wide-narrow bus, with
a Fujitsu and a Quantum disks. /dev/sdd is also in the middle, but of
a pure ultra-wide bus. It's together with another Cheetah, /dev/sde.
/dev/sde terminates the bus.
Only /dev/sdd is using the assigned queue depth, 32. You see that
being at the end of the bus or not doesn't seem to be a problem, since
we have both cases NOT using queuing.
I checked the Cheetahs with scsiinfo -a. Here's the diff between the
outputs for /dev/sdd, which uses the correct queue depth, and
/dev/sd[ae], which don't:
--- sda Tue Jul 28 21:31:03 1998
+++ sdd Tue Jul 28 21:31:09 1998
@@ -20,9 +20,9 @@
Response Data Format 2
Vendor: SEAGATE
Product: ST34501W
-Revision level: 0013NG012451
+Revision level: 0013NG012643
-Serial Number 'NG012451'
+Serial Number 'NG012643'
Data from Rigid Disk Drive Geometry Page
----------------------------------------
Number of cylinders 6576
@@ -104,26 +104,17 @@
Data from Defect Lists
----------------------
-79 entries in manufacturer table.
+35 entries in manufacturer table.
The other differences are only in sector numbers. The same happens for
sdd/sde diffs.
Here's the output for /dev/sda, which does NOT queue:
Inquiry command
---------------
Relative Address 0
Wide bus 32 0
Wide bus 16 1
Synchronous neg. 1
Linked Commands 1
Command Queueing 1 <============
SftRe 0
Device Type 0
Peripheral Qualifier 0
Removable? 0
Device Type Modifier 0
ISO Version 0
ECMA Version 0
ANSI Version 2
AENC 0
TrmIOP 0
Response Data Format 2
Vendor: SEAGATE
Product: ST34501W
Revision level: 0013NG012451
[snip]
Data from Control Page
----------------------
RLEC 0
QErr 0
DQue 0
EECA 0
RAENP 0
UUAENP 0
EAENP 0
Queue Algorithm Modifier 0
Ready AEN Holdoff Period 0
Any clues?
Carlos
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe aic7xxx" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?13759.42609.481448.897476>