Subject: port-sparc/13260: Tagged Queueing SCSI hanging on Sparc sun4m (Sparc 5, 10 and 20 reported)
To: None <gnats-bugs@gnats.netbsd.org>
From: None <netbsd@mrynet.com>
List: netbsd-bugs
Date: 06/20/2001 07:30:43
>Number:         13260
>Category:       port-sparc
>Synopsis:       Tagged Queueing SCSI hanging on Sparc sun4m (Sparc 5, 10 and 20 reported)
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    port-sparc-maintainer
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Jun 20 07:29:00 PDT 2001
>Closed-Date:
>Last-Modified:
>Originator:     Scott Taylor
>Release:        1.5W from cvs
>Organization:
>Environment:
NetBSD dudley 1.5W NetBSD 1.5W (MRYSPARC) #1: Wed Jun 20 07:38:47 CDT 2001     root@dudley:/usr/src/sys/arch/sparc/compile/MRYSPARC sparc

>Description:
Since sometime during the 1.5V incantation, increased usage of disk would
cause the machine to halt all disk access.  It has also been noticed that
"Tagged Queueing" is now being enabled on SCSI disks on the esp scsi
driver.  Prior to this problem, Tagged Queueing was being rejected.
No error messages or other indications are produced, except for the
hung disk access.
>How-To-Repeat:
Boot a 1.5W kernel and perform moderate disk I/O.  "cvs update -d -P"
of /usr/src will accomplish the hang in less than 5 seconds on a
Sparc 5.
>Fix:
A workaround for disabling the tagged queueing:
--- /sys/dev/ic/ncr53c9x.c.old  Wed Jun 20 09:27:59 2001
+++ /sys/dev/ic/ncr53c9x.c      Wed Jun 20 07:37:07 2001
@@ -951,8 +951,10 @@
        if (ti->width)
                xm.xm_mode |= PERIPH_CAP_WIDE16;
 
+/*
        if ((ti->flags & (T_RSELECTOFF|T_TAG)) == T_TAG)
                xm.xm_mode |= PERIPH_CAP_TQING;
+*/
 
        scsipi_async_event(&sc->sc_channel, ASYNC_EVENT_XFER_MODE, &xm);

>Release-Note:
>Audit-Trail:
>Unformatted: