[PATCH] x86: Add reboot quirk for Dell Precision M4600

[PATCH] x86: Add reboot quirk for Dell Precision M4600

[Mel Gorman]

ACPI reboot does not work for Dell Precision M4600. Detect this
machine via DMI and force the use of the PCI reboot method.
    
Reported-and-tested-by: Bruno Friedmann <bruno@ioda-net.ch>
Signed-off-by: Mel Gorman <mgorman@suse.de>
---
 arch/x86/kernel/reboot.c |    9 +++++++++
 1 files changed, 9 insertions(+), 0 deletions(-)

diff --git a/arch/x86/kernel/reboot.c b/arch/x86/kernel/reboot.c
index 9242436..7ca2ec4 100644
--- a/arch/x86/kernel/reboot.c
+++ b/arch/x86/kernel/reboot.c
@@ -443,6 +443,15 @@ static struct dmi_system_id __initdata pci_reboot_dmi_table[] = {
 			DMI_MATCH(DMI_PRODUCT_NAME, "Latitude E6420"),
 		},
 	},
+	{	/* Handle problems with rebooting on Dell Precision M4600's */
+		.callback = set_pci_reboot,
+		.ident = "Dell Precision M4600",
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc."),
+			DMI_MATCH(DMI_PRODUCT_NAME, "Precision M4600"),
+		},
+	},
+
 	{ }
 };
 

Re: [PATCH] x86: Add reboot quirk for Dell Precision M4600

[Ingo Molnar]

* Mel Gorman <mgorman@suse.de> wrote:

> ACPI reboot does not work for Dell Precision M4600. Detect this
> machine via DMI and force the use of the PCI reboot method.
>     
> Reported-and-tested-by: Bruno Friedmann <bruno@ioda-net.ch>
> Signed-off-by: Mel Gorman <mgorman@suse.de>
> ---
>  arch/x86/kernel/reboot.c |    9 +++++++++
>  1 files changed, 9 insertions(+), 0 deletions(-)

Weird - ACPI is the default Windows method of reboot - how come it 
does not work on that box? Does Windows work around this specific 
machine (unlikely), or is our ACPI boot sequence not 100% compatible 
with the Windows method (more likely IMO)?

Thanks,

	Ingo

Re: [PATCH] x86: Add reboot quirk for Dell Precision M4600

[H. Peter Anvin]

Ingo Molnar <mingo@elte.hu> wrote:

>
>* Mel Gorman <mgorman@suse.de> wrote:
>
>> ACPI reboot does not work for Dell Precision M4600. Detect this
>> machine via DMI and force the use of the PCI reboot method.
>>     
>> Reported-and-tested-by: Bruno Friedmann <bruno@ioda-net.ch>
>> Signed-off-by: Mel Gorman <mgorman@suse.de>
>> ---
>>  arch/x86/kernel/reboot.c |    9 +++++++++
>>  1 files changed, 9 insertions(+), 0 deletions(-)
>
>Weird - ACPI is the default Windows method of reboot - how come it 
>does not work on that box? Does Windows work around this specific 
>machine (unlikely), or is our ACPI boot sequence not 100% compatible 
>with the Windows method (more likely IMO)?
>
>Thanks,
>
>	Ingo

It seems almost all the affected machines are Dell, too... very odd, no?
-- 
Sent from my mobile phone. Please excuse my brevity and lack of formatting.

Re: [PATCH] x86: Add reboot quirk for Dell Precision M4600

[Matthew Garrett]

On Sat, Aug 06, 2011 at 08:56:35AM +0200, Ingo Molnar wrote:
> Weird - ACPI is the default Windows method of reboot - how come it 
> does not work on that box? Does Windows work around this specific 
> machine (unlikely), or is our ACPI boot sequence not 100% compatible 
> with the Windows method (more likely IMO)?

The ACPI reboot method on these machines hits the keyboard controller 
rather than PCI, which is unusual. What I suspect is that the write 
triggers some SMM or EC code that's making assumptions about system 
state, and we're not quite identical. Bouncing on the keyboard 
controller from userspace with init=/bin/bash wedges in the same way.

-- 
Matthew Garrett | mjg59@srcf.ucam.org

Re: [PATCH] x86: Add reboot quirk for Dell Precision M4600

[Len Brown]

On Sat, 6 Aug 2011, H. Peter Anvin wrote:

> Ingo Molnar <mingo@elte.hu> wrote:
> 
> >* Mel Gorman <mgorman@suse.de> wrote:
> >
> >> ACPI reboot does not work for Dell Precision M4600. Detect this
> >> machine via DMI and force the use of the PCI reboot method.
> >>     
> >> Reported-and-tested-by: Bruno Friedmann <bruno@ioda-net.ch>
> >> Signed-off-by: Mel Gorman <mgorman@suse.de>
> >> ---
> >>  arch/x86/kernel/reboot.c |    9 +++++++++
> >>  1 files changed, 9 insertions(+), 0 deletions(-)
> >
> >Weird - ACPI is the default Windows method of reboot - how come it 
> >does not work on that box? Does Windows work around this specific 
> >machine (unlikely), or is our ACPI boot sequence not 100% compatible 
> >with the Windows method (more likely IMO)?
> 
> It seems almost all the affected machines are Dell, too... very odd, no?

Dell website says that the M4600 is current,
and ships with Windows 7.  It is a "mobile workstation",
so it is more likely to have a windows "feature" driver,
say, than Dell's servers.

Matthew,
Did you do your reboot comopatibility test using Win7,
or something older?

BTW, it would be nice if ACPI incompatibilities and workarounds
somehow got stuck in the ACPI code rather than the arch code.
Unfortunately this kind sets a working reboot method instead
of disabling a non-working one -- so what it overrides has
changed when the default changed.

thanks,
-Len Brown, Intel Open Source Technolgy Center

Re: [PATCH] x86: Add reboot quirk for Dell Precision M4600

[Bruno Friedmann]

On 08/06/2011 05:31 PM, Len Brown wrote:
> On Sat, 6 Aug 2011, H. Peter Anvin wrote:
> 
>> Ingo Molnar <mingo@elte.hu> wrote:
>>
>>> * Mel Gorman <mgorman@suse.de> wrote:
>>>
>>>> ACPI reboot does not work for Dell Precision M4600. Detect this
>>>> machine via DMI and force the use of the PCI reboot method.
>>>>     
>>>> Reported-and-tested-by: Bruno Friedmann <bruno@ioda-net.ch>
>>>> Signed-off-by: Mel Gorman <mgorman@suse.de>
>>>> ---
>>>>  arch/x86/kernel/reboot.c |    9 +++++++++
>>>>  1 files changed, 9 insertions(+), 0 deletions(-)
>>>
>>> Weird - ACPI is the default Windows method of reboot - how come it 
>>> does not work on that box? Does Windows work around this specific 
>>> machine (unlikely), or is our ACPI boot sequence not 100% compatible 
>>> with the Windows method (more likely IMO)?
>>
>> It seems almost all the affected machines are Dell, too... very odd, no?
> 
> Dell website says that the M4600 is current,
> and ships with Windows 7.  It is a "mobile workstation",
> so it is more likely to have a windows "feature" driver,
> say, than Dell's servers.

And this computer can be buyed with RedHat 6.0 by phone call.
If you check the support website download and drivers you will some drivers for Redhat
(mainly the new version of alsa)
http://dell.to/r1IKhp

> 
> Matthew,
> Did you do your reboot comopatibility test using Win7,
> or something older?
> 
> BTW, it would be nice if ACPI incompatibilities and workarounds
> somehow got stuck in the ACPI code rather than the arch code.
> Unfortunately this kind sets a working reboot method instead
> of disabling a non-working one -- so what it overrides has
> changed when the default changed.
> 
> thanks,
> -Len Brown, Intel Open Source Technolgy Center
> 

In any case, if you want anymore testing, and dumps of something, just drop me an email
and the howto get it formatted as you want it.

Actually I'm in the desktop summit with the computer until 9th August night.

I don't know if we can find an owner of the M6600 (the 17") version to get feedback about that hardware which share
a lot with the M4600.

Also Dell can upgrade their Bios.

The only error or significant warning I saw in boot.msg is
<3>[   36.851649] sdhci-pci 0000:0b:00.1: Invalid iomem size. You may experience problems.




-- 

Bruno Friedmann
Ioda-Net Sàrl www.ioda-net.ch

openSUSE Member & Ambassador
GPG KEY : D5C9B751C4653227
irc: tigerfoot

Re: [PATCH] x86: Add reboot quirk for Dell Precision M4600

[Mel Gorman]

On Sat, Aug 06, 2011 at 08:56:35AM +0200, Ingo Molnar wrote:
> 
> * Mel Gorman <mgorman@suse.de> wrote:
> 
> > ACPI reboot does not work for Dell Precision M4600. Detect this
> > machine via DMI and force the use of the PCI reboot method.
> >     
> > Reported-and-tested-by: Bruno Friedmann <bruno@ioda-net.ch>
> > Signed-off-by: Mel Gorman <mgorman@suse.de>
> > ---
> >  arch/x86/kernel/reboot.c |    9 +++++++++
> >  1 files changed, 9 insertions(+), 0 deletions(-)
> 
> Weird - ACPI is the default Windows method of reboot - how come it 
> does not work on that box? Does Windows work around this specific 
> machine (unlikely), or is our ACPI boot sequence not 100% compatible 
> with the Windows method (more likely IMO)?
> 

I have no idea and am not well versed on how this sort of problem is
normally investigated.

As Matthew is the person most likely to be able to fix this, I'll defer
to him. Matthew, would you prefer a quirk be added or hold off until
there is time to investigate an alternative?

-- 
Mel Gorman
SUSE Labs

Re: [PATCH] x86: Add reboot quirk for Dell Precision M4600

[H. Peter Anvin]

On 08/06/2011 10:30 AM, Matthew Garrett wrote:
> On Sat, Aug 06, 2011 at 08:56:35AM +0200, Ingo Molnar wrote:
>> Weird - ACPI is the default Windows method of reboot - how come it 
>> does not work on that box? Does Windows work around this specific 
>> machine (unlikely), or is our ACPI boot sequence not 100% compatible 
>> with the Windows method (more likely IMO)?
> 
> The ACPI reboot method on these machines hits the keyboard controller 
> rather than PCI, which is unusual. What I suspect is that the write 
> triggers some SMM or EC code that's making assumptions about system 
> state, and we're not quite identical. Bouncing on the keyboard 
> controller from userspace with init=/bin/bash wedges in the same way.
> 

I have a vague suspicion that Windows might have a meta-quirk for this
class of machines, i.e. if the ACPI reboot method, say, points at the
keyboard controller or points at the "PCI" reboot port then it actually
uses its own special-purpose routines rather than general routines.  It
would be extremely interesting to know if "reboot=kbd" works on these boxen.

	-hpa

Re: [PATCH] x86: Add reboot quirk for Dell Precision M4600

[Bruno Friedmann]

On 08/08/2011 06:40 PM, H. Peter Anvin wrote:
> On 08/06/2011 10:30 AM, Matthew Garrett wrote:
>> On Sat, Aug 06, 2011 at 08:56:35AM +0200, Ingo Molnar wrote:
>>> Weird - ACPI is the default Windows method of reboot - how come it 
>>> does not work on that box? Does Windows work around this specific 
>>> machine (unlikely), or is our ACPI boot sequence not 100% compatible 
>>> with the Windows method (more likely IMO)?
>>
>> The ACPI reboot method on these machines hits the keyboard controller 
>> rather than PCI, which is unusual. What I suspect is that the write 
>> triggers some SMM or EC code that's making assumptions about system 
>> state, and we're not quite identical. Bouncing on the keyboard 
>> controller from userspace with init=/bin/bash wedges in the same way.
>>
> 
> I have a vague suspicion that Windows might have a meta-quirk for this
> class of machines, i.e. if the ACPI reboot method, say, points at the
> keyboard controller or points at the "PCI" reboot port then it actually
> uses its own special-purpose routines rather than general routines.  It
> would be extremely interesting to know if "reboot=kbd" works on these boxen.
> 
> 	-hpa
> 

Ok I will try to activate reboot=kbd tomorrow morning and report the result about it.


-- 

Bruno Friedmann
Ioda-Net Sàrl www.ioda-net.ch

openSUSE Member & Ambassador
GPG KEY : D5C9B751C4653227
irc: tigerfoot

Re: [PATCH] x86: Add reboot quirk for Dell Precision M4600

[Bruno Friedmann]

On 08/08/2011 06:40 PM, H. Peter Anvin wrote:
> On 08/06/2011 10:30 AM, Matthew Garrett wrote:
>> On Sat, Aug 06, 2011 at 08:56:35AM +0200, Ingo Molnar wrote:
>>> Weird - ACPI is the default Windows method of reboot - how come it 
>>> does not work on that box? Does Windows work around this specific 
>>> machine (unlikely), or is our ACPI boot sequence not 100% compatible 
>>> with the Windows method (more likely IMO)?
>>
>> The ACPI reboot method on these machines hits the keyboard controller 
>> rather than PCI, which is unusual. What I suspect is that the write 
>> triggers some SMM or EC code that's making assumptions about system 
>> state, and we're not quite identical. Bouncing on the keyboard 
>> controller from userspace with init=/bin/bash wedges in the same way.
>>
> 
> I have a vague suspicion that Windows might have a meta-quirk for this
> class of machines, i.e. if the ACPI reboot method, say, points at the
> keyboard controller or points at the "PCI" reboot port then it actually
> uses its own special-purpose routines rather than general routines.  It
> would be extremely interesting to know if "reboot=kbd" works on these boxen.
> 
> 	-hpa
> 

Test done with reboot=pci and unfortunately doesn't work.


-- 

Bruno Friedmann
Ioda-Net Sàrl www.ioda-net.ch

openSUSE Member & Ambassador
GPG KEY : D5C9B751C4653227
irc: tigerfoot

Re: [PATCH] x86: Add reboot quirk for Dell Precision M4600

[Ingo Molnar]

* Bruno Friedmann <bruno@ioda-net.ch> wrote:

> On 08/08/2011 06:40 PM, H. Peter Anvin wrote:
> > On 08/06/2011 10:30 AM, Matthew Garrett wrote:
> >> On Sat, Aug 06, 2011 at 08:56:35AM +0200, Ingo Molnar wrote:
> >>> Weird - ACPI is the default Windows method of reboot - how come it 
> >>> does not work on that box? Does Windows work around this specific 
> >>> machine (unlikely), or is our ACPI boot sequence not 100% compatible 
> >>> with the Windows method (more likely IMO)?
> >>
> >> The ACPI reboot method on these machines hits the keyboard controller 
> >> rather than PCI, which is unusual. What I suspect is that the write 
> >> triggers some SMM or EC code that's making assumptions about system 
> >> state, and we're not quite identical. Bouncing on the keyboard 
> >> controller from userspace with init=/bin/bash wedges in the same way.
> >>
> > 
> > I have a vague suspicion that Windows might have a meta-quirk for this
> > class of machines, i.e. if the ACPI reboot method, say, points at the
> > keyboard controller or points at the "PCI" reboot port then it actually
> > uses its own special-purpose routines rather than general routines.  It
> > would be extremely interesting to know if "reboot=kbd" works on these boxen.
> > 
> > 	-hpa
> > 
> 
> Test done with reboot=pci and unfortunately doesn't work.

Didn't hpa suggest reboot=kbd?

Thanks,

	Ingo

Re: [PATCH] x86: Add reboot quirk for Dell Precision M4600

[Bruno Friedmann]

On 08/09/2011 05:02 PM, Ingo Molnar wrote:
> 
> * Bruno Friedmann <bruno@ioda-net.ch> wrote:
> 
>> On 08/08/2011 06:40 PM, H. Peter Anvin wrote:
>>> On 08/06/2011 10:30 AM, Matthew Garrett wrote:
>>>> On Sat, Aug 06, 2011 at 08:56:35AM +0200, Ingo Molnar wrote:
>>>>> Weird - ACPI is the default Windows method of reboot - how come it 
>>>>> does not work on that box? Does Windows work around this specific 
>>>>> machine (unlikely), or is our ACPI boot sequence not 100% compatible 
>>>>> with the Windows method (more likely IMO)?
>>>>
>>>> The ACPI reboot method on these machines hits the keyboard controller 
>>>> rather than PCI, which is unusual. What I suspect is that the write 
>>>> triggers some SMM or EC code that's making assumptions about system 
>>>> state, and we're not quite identical. Bouncing on the keyboard 
>>>> controller from userspace with init=/bin/bash wedges in the same way.
>>>>
>>>
>>> I have a vague suspicion that Windows might have a meta-quirk for this
>>> class of machines, i.e. if the ACPI reboot method, say, points at the
>>> keyboard controller or points at the "PCI" reboot port then it actually
>>> uses its own special-purpose routines rather than general routines.  It
>>> would be extremely interesting to know if "reboot=kbd" works on these boxen.
>>>
>>> 	-hpa
>>>
>>
>> Test done with reboot=pci and unfortunately doesn't work.
> 
> Didn't hpa suggest reboot=kbd?
> 
> Thanks,
> 
> 	Ingo

Ingo yeap and reboot=kbd doesn't work ...
I must really be awake before writing

Once the system has synchronised the harddrive, it just stay at now rebooting ...


-- 

Bruno Friedmann
Ioda-Net Sàrl www.ioda-net.ch

openSUSE Member & Ambassador
GPG KEY : D5C9B751C4653227
irc: tigerfoot