NetBSD Problem Report #47224

From www@NetBSD.org  Wed Nov 21 05:43:45 2012
Return-Path: <www@NetBSD.org>
Received: from mail.netbsd.org (mail.netbsd.org [149.20.53.66])
	by www.NetBSD.org (Postfix) with ESMTP id 51BBF63DCB2
	for <gnats-bugs@gnats.NetBSD.org>; Wed, 21 Nov 2012 05:43:45 +0000 (UTC)
Message-Id: <20121121054344.50FAA63DCB2@www.NetBSD.org>
Date: Wed, 21 Nov 2012 05:43:44 +0000 (UTC)
From: bonhagjeffrey@gmail.com
Reply-To: bonhagjeffrey@gmail.com
To: gnats-bugs@NetBSD.org
Subject: Dell Latitude D531 
X-Send-Pr-Version: www-1.0

>Number:         47224
>Category:       kern
>Synopsis:       Dell Latitude D531
>Confidential:   no
>Severity:       non-critical
>Priority:       medium
>Responsible:    kern-bug-people
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Nov 21 05:45:00 +0000 2012
>Closed-Date:    Sun Dec 30 17:37:13 +0000 2012
>Last-Modified:  Sun Dec 30 17:37:13 +0000 2012
>Originator:     Jeffrey Bonhag
>Release:        6.0 GENERIC
>Organization:
>Environment:
NetBSD  6.0 NetBSD 6.0 (GENERIC) amd64
>Description:
When I booted from the install CD for the first time, the system froze right after my uhub devices were loaded.  I was able to install NetBSD by booting with ACPI disabled.

The problem continues after NetBSD is installed, so I'm just booting with ACPI disabled and things seem to be working fine.

I don't know if this is important, but I'm using Dell's BIOS version A10.
>How-To-Repeat:
1. Get a Dell Latitude D531
2. Boot the NetBSD installation disc normally
3. Watch it freeze
>Fix:
Boot with ACPI disabled.

>Release-Note:

>Audit-Trail:
From: Imre Vadasz <imre@vdsz.com>
To: gnats-bugs@NetBSD.org
Cc: 
Subject: Re: install/47224
Date: Thu, 6 Dec 2012 01:26:18 +0100

 a) Try disabling C1E in the bios if that is possible.
 If C1E cannot be disabled in the bios, you can try the appended patch,
 which disables C1E on AMD k8 cpus (like the tl-60) which belong to
 family 0x0f.
 That patch fixes a problem with identical symptoms on my HP 6715s
 notebook, which has almost the same hardware.

 When your system freezes during bootup, try pressing random keys for
 a while, if NetBSD slowly continues to boot, then your problem is
 definitely caused by an enabled C1E mode.
 When the cpu goes into C1E power-saving mode, no more timer-interrupts
 are received by the OS. Pressing keys on the keyboard or otherwise
 putting load on the system causes some of the interrupts to come through.

 b)I also experienced problems on boot when the ohci driver was enabled,
 using 6.99.15 on my HP 6715s notebook.
 The problem caused by ohci manifested in a blanked screen and a frozen
 system on some bootups.
 I worked around that problem by disabling ohci.

 A patch for the C1E problem, which also disables C1E on AMD K8 cpus
 (family 0x0f) (current code in HEAD only disables C1E on families 0x10
 and 0x11).

 Index: sys/arch/x86/acpi/acpi_cpu_md.c
 ===================================================================
 RCS file: /cvsroot/src/sys/arch/x86/acpi/acpi_cpu_md.c,v
 retrieving revision 1.71
 diff -u -r1.71 acpi_cpu_md.c
 --- sys/arch/x86/acpi/acpi_cpu_md.c	11 Feb 2012 22:09:47 -0000	1.71
 +++ sys/arch/x86/acpi/acpi_cpu_md.c	5 Dec 2012 16:09:37 -0000
 @@ -252,6 +252,9 @@

  		case 0x0f:

 +			if (rdmsr_safe(MSR_CMPHALT, &msr) != EFAULT)
 +				val |= ACPICPU_FLAG_C_C1E;
 +
  			/*
  			 * Evaluate support for the "FID/VID
  			 * algorithm" also used by powernow(4).

From: "Jukka Ruohonen" <jruoho@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/47224 CVS commit: src/sys/arch/x86/acpi
Date: Thu, 6 Dec 2012 04:43:29 +0000

 Module Name:	src
 Committed By:	jruoho
 Date:		Thu Dec  6 04:43:29 UTC 2012

 Modified Files:
 	src/sys/arch/x86/acpi: acpi_cpu_md.c

 Log Message:
 Disable C1E also on K8, if present. From Imre Vadasz <imre@vdsz.com>
 in PR install/47224.


 To generate a diff of this commit:
 cvs rdiff -u -r1.71 -r1.72 src/sys/arch/x86/acpi/acpi_cpu_md.c

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

From: bonhagjeffrey@gmail.com
To: gnats-bugs@NetBSD.org
Cc: 
Subject: Re: install/47224
Date: Tue, 11 Dec 2012 04:46:48 +0000

 On Thu, Dec 06, 2012 at 12:30:12AM +0000, Imre Vadasz wrote:
 > The following reply was made to PR install/47224; it has been noted by GNATS.
 > 
 > From: Imre Vadasz <imre@vdsz.com>
 > To: gnats-bugs@NetBSD.org
 > Cc: 
 > Subject: Re: install/47224
 > Date: Thu, 6 Dec 2012 01:26:18 +0100
 > 
 >  a) Try disabling C1E in the bios if that is possible.
 >  If C1E cannot be disabled in the bios, you can try the appended patch,
 >  which disables C1E on AMD k8 cpus (like the tl-60) which belong to
 >  family 0x0f.
 >  That patch fixes a problem with identical symptoms on my HP 6715s
 >  notebook, which has almost the same hardware.
 >  

 While I was playing around, I compiled a version of the kernel with ACPI taken
 out completely, and the system booted happily with no intervention required on
 my part.

 On the Dell, I fonud a setting for "PowerNow!" in the bios.  I disabled it,
 recompiled a GENERIC kernel, and sure enough, the system booted without any
 problems.

 >  b)I also experienced problems on boot when the ohci driver was enabled,
 >  using 6.99.15 on my HP 6715s notebook.
 >  The problem caused by ohci manifested in a blanked screen and a frozen
 >  system on some bootups.
 >  I worked around that problem by disabling ohci.

 Did you disable ohci in the bios or in your kernel configuration somewhere?

 >  
 >  A patch for the C1E problem, which also disables C1E on AMD K8 cpus
 >  (family 0x0f) (current code in HEAD only disables C1E on families 0x10
 >  and 0x11).
 >  

 I grabbed the patch (ran cvs update before compiling the GENERIC kernel).  Is
 this also used on the amd64 architecture?

 Thanks a million for your help.

 Jeff

From: David Laight <david@l8s.co.uk>
To: gnats-bugs@NetBSD.org
Cc: 
Subject: Re: install/47224
Date: Tue, 11 Dec 2012 08:30:35 +0000

 On Tue, Dec 11, 2012 at 04:50:08AM +0000, bonhagjeffrey@gmail.com wrote:
 >  
 >  On the Dell, I fonud a setting for "PowerNow!" in the bios.  I disabled it,
 >  recompiled a GENERIC kernel, and sure enough, the system booted without any
 >  problems.

 I wonder what difference that makes to the ACPI tables?
 Might be a hint to the faulty code.

 	David

 -- 
 David Laight: david@l8s.co.uk

From: Imre Vadasz <imre@vdsz.com>
To: gnats-bugs@NetBSD.org
Cc: 
Subject: Re: install/47224
Date: Tue, 11 Dec 2012 18:38:14 +0100

 Sorry for the confusion, disabling ohci doesn't make any difference.
 Since my 6715s sometimes crashes on bootup I mistakenly attributed
 that problem to ohci. Since that problem still occurs for me it is
 probably not caused by ohci, for now I assume that my bootup problem
 is specific to the 6715s (and thus doesn't concern this bug).
 Again sorry for that confusion.

 The C1E patch should fix NetBSD on AMD K8 notebook cpus, when C1E
 cannot be disabled in the BIOS.
 Disabling PowerNow! in the BIOS shouldn't be necessary, since
 that will probably also disable frequency scaling.

From: "Jeff Rizzo" <riz@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/47224 CVS commit: [netbsd-6] src/sys/arch/x86/acpi
Date: Thu, 13 Dec 2012 23:41:57 +0000

 Module Name:	src
 Committed By:	riz
 Date:		Thu Dec 13 23:41:57 UTC 2012

 Modified Files:
 	src/sys/arch/x86/acpi [netbsd-6]: acpi_cpu_md.c

 Log Message:
 Pull up following revision(s) (requested by jruoho in ticket #741):
 	sys/arch/x86/acpi/acpi_cpu_md.c: revision 1.72
 Disable C1E also on K8, if present. From Imre Vadasz <imre@vdsz.com>
 in PR install/47224.


 To generate a diff of this commit:
 cvs rdiff -u -r1.71 -r1.71.2.1 src/sys/arch/x86/acpi/acpi_cpu_md.c

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

From: "Jeff Rizzo" <riz@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/47224 CVS commit: [netbsd-6-0] src/sys/arch/x86/acpi
Date: Thu, 13 Dec 2012 23:42:16 +0000

 Module Name:	src
 Committed By:	riz
 Date:		Thu Dec 13 23:42:16 UTC 2012

 Modified Files:
 	src/sys/arch/x86/acpi [netbsd-6-0]: acpi_cpu_md.c

 Log Message:
 Pull up following revision(s) (requested by jruoho in ticket #741):
 	sys/arch/x86/acpi/acpi_cpu_md.c: revision 1.72
 Disable C1E also on K8, if present. From Imre Vadasz <imre@vdsz.com>
 in PR install/47224.


 To generate a diff of this commit:
 cvs rdiff -u -r1.71 -r1.71.8.1 src/sys/arch/x86/acpi/acpi_cpu_md.c

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

From: bonhagjeffrey@gmail.com
To: gnats-bugs@NetBSD.org
Cc: 
Subject: Re: install/47224
Date: Mon, 17 Dec 2012 02:03:00 +0000

 On Tue, Dec 11, 2012 at 05:45:05PM +0000, Imre Vadasz wrote:
 > The following reply was made to PR install/47224; it has been noted by GNATS.
 > 
 > From: Imre Vadasz <imre@vdsz.com>
 > To: gnats-bugs@NetBSD.org
 > Cc: 
 > Subject: Re: install/47224
 > Date: Tue, 11 Dec 2012 18:38:14 +0100
 > 
 >  Sorry for the confusion, disabling ohci doesn't make any difference.
 >  Since my 6715s sometimes crashes on bootup I mistakenly attributed
 >  that problem to ohci. Since that problem still occurs for me it is
 >  probably not caused by ohci, for now I assume that my bootup problem
 >  is specific to the 6715s (and thus doesn't concern this bug).
 >  Again sorry for that confusion.
 >  
 >  The C1E patch should fix NetBSD on AMD K8 notebook cpus, when C1E
 >  cannot be disabled in the BIOS.
 >  Disabling PowerNow! in the BIOS shouldn't be necessary, since
 >  that will probably also disable frequency scaling.
 >  

 Happy to report that even with PowerNow! enabled, the kernel boots beautifully
 thanks to your patch.  Thanks again for working through the problem with me.

 Jeff

Responsible-Changed-From-To: install-manager->kern-bug-people
Responsible-Changed-By: chs@NetBSD.org
Responsible-Changed-When: Sun, 30 Dec 2012 17:37:13 +0000
Responsible-Changed-Why:
kernel bug, not installer bug


State-Changed-From-To: open->closed
State-Changed-By: chs@NetBSD.org
State-Changed-When: Sun, 30 Dec 2012 17:37:13 +0000
State-Changed-Why:
confirmed fixed


>Unformatted:

Home	PR Database Search