Subject: port-hpcmips/30748: Screen freezes after suspend/resume with ITE8181_LCD_CONTROL_ENABLE
To: None <port-hpcmips-maintainer@netbsd.org, gnats-admin@netbsd.org,>
From: None <ssorensen@gmail.com>
List: netbsd-bugs
Date: 07/14/2005 16:54:01
>Number: 30748
>Category: port-hpcmips
>Synopsis: Screen freezes after suspend/resume with ITE8181_LCD_CONTROL_ENABLE
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: port-hpcmips-maintainer
>State: open
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Thu Jul 14 16:54:00 +0000 2005
>Originator: Svend Sorensen
>Release: NetBSD 2.0.2
>Organization:
>Environment:
NetBSD ink.fakedomain.com 2.0.0_STABLE NetBSD 2.0.2_STABLE (IBMZ50) #0 Tue Jul5 16:27:47 PDT 2005 sorensen@sverige:/tmp/sorensen/src/sys/arch/hpcmips/compile/IBMz50 hpcmips
>Description:
With ITE8181_LCD_CONTROL_ENABLE option set, the screen on the IBM Z50 will blank when the system suspends. The system normally resumes without any problems. The power button initiates a suspend, and wakes the system up if it is suspended.
Occasionally, when resuming, the screen will display a set of vertical bars, and the system will be unresponsive. The power button does not re-suspend the machine, and a hard reset (remove power cable, remove battery) is required.
The port-hpcmips mailing list has several posts regarding this issue. Search the list for the string "vertical bars" (http://search.gmane.org/?query=vertical+bars&group=gmane.os.netbsd.ports.hpcmips).
I patched my kernel with the 'TLB out of universe' fix at http://www.city-net.com/~darkstar/netbsd/hpcmips.diff (see http://mail-index.netbsd.org/port-hpcmips/2004/12/30/0003.html). I added the line:
options ITE8181_LCD_CONTROL_ENABLE
to the GENERIC kernel config and rebuilt it.
>How-To-Repeat:
Enable ITE8181_LCD_CONTROL_ENABLE (see http://netbsd.org/Ports/hpcmips/faq.html#z50-lcd). Suspend the IBM Z50 by pressing the power button. Resume the IBM Z50 by pressing the power button again. Repeat until the screen displays vertical bars on resume.
>Fix:
Members of the port-hpcmips list suggested that running sync before suspending greatly reduced the probability of the issue arrising, but I did not find this to be the case. When 'sync' did not work, I set up all partitions as read only, with mfs for /tmp and /var. This eliminated any writes to the CF card, but the problem still arose.