Could the k8temp driver be interfering with ACPI?

Could the k8temp driver be interfering with ACPI?

[Chuck Ebbert]

Recently my notebook has started shutting down with
these messages in the logs:

	ACPI: Critical trip point
	Critical temperature reached (128 C), shutting down.

But it didn't seem hot at all to me, so I wrote a script to
cat /proc/acpi/thermal_zone/THRM/temperature once a second
and eventually caught this (but no shutdown):

temperature:             47 C
temperature:             47 C
temperature:             128 C
temperature:             48 C
temperature:             47 C

Google found several people reporting problems like mine
after installing lm-sensors, and when I looked at the list
of loaded modules I found k8temp and hwmon there. Then I
realized my problems had started after installing a 2.6.19
kernel that had the new k8temp driver.

So, could ACPI and the k8temp driver be at odds?

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Chuck Ebbert]

Recently my notebook has started shutting down with
these messages in the logs:

	ACPI: Critical trip point
	Critical temperature reached (128 C), shutting down.

But it didn't seem hot at all to me, so I wrote a script to
cat /proc/acpi/thermal_zone/THRM/temperature once a second
and eventually caught this (but no shutdown):

temperature:             47 C
temperature:             47 C
temperature:             128 C
temperature:             48 C
temperature:             47 C

Google found several people reporting problems like mine
after installing lm-sensors, and when I looked at the list
of loaded modules I found k8temp and hwmon there. Then I
realized my problems had started after installing a 2.6.19
kernel that had the new k8temp driver.

So, could ACPI and the k8temp driver be at odds?

Re: Could the k8temp driver be interfering with ACPI?

[Len Brown]

On Friday 16 February 2007 12:31, Chuck Ebbert wrote:
> Recently my notebook has started shutting down with
> these messages in the logs:
> 
> 	ACPI: Critical trip point
> 	Critical temperature reached (128 C), shutting down.
> 
> But it didn't seem hot at all to me, so I wrote a script to
> cat /proc/acpi/thermal_zone/THRM/temperature once a second
> and eventually caught this (but no shutdown):
> 
> temperature:             47 C
> temperature:             47 C
> temperature:             128 C
> temperature:             48 C
> temperature:             47 C
> 
> Google found several people reporting problems like mine
> after installing lm-sensors, and when I looked at the list
> of loaded modules I found k8temp and hwmon there. Then I
> realized my problems had started after installing a 2.6.19
> kernel that had the new k8temp driver.
> 
> So, could ACPI and the k8temp driver be at odds?


Yes.
-Len

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Len Brown]

On Friday 16 February 2007 12:31, Chuck Ebbert wrote:
> Recently my notebook has started shutting down with
> these messages in the logs:
> 
> 	ACPI: Critical trip point
> 	Critical temperature reached (128 C), shutting down.
> 
> But it didn't seem hot at all to me, so I wrote a script to
> cat /proc/acpi/thermal_zone/THRM/temperature once a second
> and eventually caught this (but no shutdown):
> 
> temperature:             47 C
> temperature:             47 C
> temperature:             128 C
> temperature:             48 C
> temperature:             47 C
> 
> Google found several people reporting problems like mine
> after installing lm-sensors, and when I looked at the list
> of loaded modules I found k8temp and hwmon there. Then I
> realized my problems had started after installing a 2.6.19
> kernel that had the new k8temp driver.
> 
> So, could ACPI and the k8temp driver be at odds?


Yes.
-Len

Re: Could the k8temp driver be interfering with ACPI?

[Chuck Ebbert]

Len Brown wrote:
>> So, could ACPI and the k8temp driver be at odds?
> 
> 
> Yes.

Hmm, now it's showing 130 degrees once every five seconds
and 54 degrees for the other four.

And while I was typing this on another machine I heard a
click from the notebook -- it shut down again.

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Chuck Ebbert]

Len Brown wrote:
>> So, could ACPI and the k8temp driver be at odds?
> 
> 
> Yes.

Hmm, now it's showing 130 degrees once every five seconds
and 54 degrees for the other four.

And while I was typing this on another machine I heard a
click from the notebook -- it shut down again.

Re: Could the k8temp driver be interfering with ACPI?

[Chuck Ebbert]

Andi Kleen wrote:
> Chuck Ebbert <cebbert@redhat.com> writes:
>> So, could ACPI and the k8temp driver be at odds?
> 
> Yes, there is no locking between ACPI and Linux drivers for register access.
> e.g. if there is a indexed register both try to access (and temperature
> sensors tend to use these things) they can race and get corrupted data.

I blacklisted the k8temp driver and everything looks OK now.

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Chuck Ebbert]

Andi Kleen wrote:
> Chuck Ebbert <cebbert at redhat.com> writes:
>> So, could ACPI and the k8temp driver be at odds?
> 
> Yes, there is no locking between ACPI and Linux drivers for register access.
> e.g. if there is a indexed register both try to access (and temperature
> sensors tend to use these things) they can race and get corrupted data.

I blacklisted the k8temp driver and everything looks OK now.

Re: Could the k8temp driver be interfering with ACPI?

[Andi Kleen]

Chuck Ebbert <cebbert@redhat.com> writes:
> 
> So, could ACPI and the k8temp driver be at odds?

Yes, there is no locking between ACPI and Linux drivers for register access.
e.g. if there is a indexed register both try to access (and temperature
sensors tend to use these things) they can race and get corrupted data.

-Andi

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Andi Kleen]

Chuck Ebbert <cebbert at redhat.com> writes:
> 
> So, could ACPI and the k8temp driver be at odds?

Yes, there is no locking between ACPI and Linux drivers for register access.
e.g. if there is a indexed register both try to access (and temperature
sensors tend to use these things) they can race and get corrupted data.

-Andi

Re: Could the k8temp driver be interfering with ACPI?

[Rudolf Marek]

Hello Chuck,

I'm the author of K8temp. Please can you share with us your DSDT table?
(cat /proc/acpi/dsdt > /tmp/dsdt.bin)

> So, could ACPI and the k8temp driver be at odds?

Yes because ACPI AML code has no synchronization with Linux drivers. Second
reason is that ACPI AML code assign resource regions to itself but with cleared
busy flag - so other drivers could bind and might possibly interfere  with ACPI.

This is very long term problem, I already proposed some possible solutions to
this problem
(http://lists.lm-sensors.org/pipermail/lm-sensors/2007-February/018788.html)

There are some ideas, but none is implemented yet. As you already wrote, best
solution is to stop using k8temp driver.

I will check the DSDT table to confirm this fact.

Thanks,
Rudolf

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Rudolf Marek]

Hello Chuck,

I'm the author of K8temp. Please can you share with us your DSDT table?
(cat /proc/acpi/dsdt > /tmp/dsdt.bin)

> So, could ACPI and the k8temp driver be at odds?

Yes because ACPI AML code has no synchronization with Linux drivers. Second
reason is that ACPI AML code assign resource regions to itself but with cleared
busy flag - so other drivers could bind and might possibly interfere  with ACPI.

This is very long term problem, I already proposed some possible solutions to
this problem
(http://lists.lm-sensors.org/pipermail/lm-sensors/2007-February/018788.html)

There are some ideas, but none is implemented yet. As you already wrote, best
solution is to stop using k8temp driver.

I will check the DSDT table to confirm this fact.

Thanks,
Rudolf

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Rudolf Marek]

Hello Chuck,

I'm the author of K8temp. Please can you share with us your DSDT table?
(cat /proc/acpi/dsdt > /tmp/dsdt.bin)

> So, could ACPI and the k8temp driver be at odds?

Yes because ACPI AML code has no synchronization with Linux drivers. Second
reason is that ACPI AML code assign resource regions to itself but with cleared
busy flag - so other drivers could bind and might possibly interfere  with ACPI.

This is very long term problem, I already proposed some possible solutions to
this problem
(http://lists.lm-sensors.org/pipermail/lm-sensors/2007-February/018788.html)

There are some ideas, but none is implemented yet. As you already wrote, best
solution is to stop using k8temp driver.

I will check the DSDT table to confirm this fact.

Thanks,
Rudolf

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Chuck Ebbert]

Rudolf Marek wrote:
> Hello Chuck,
> 
> I'm the author of K8temp. Please can you share with us your DSDT table?
> (cat /proc/acpi/dsdt > /tmp/dsdt.bin)
> 

The system is Compaq Presario V2300 series notebook.  I won't be able
to get the DSDT until tomorrow or Monday.

>> So, could ACPI and the k8temp driver be at odds?
> 
> Yes because ACPI AML code has no synchronization with Linux drivers. Second
> reason is that ACPI AML code assign resource regions to itself but with cleared
> busy flag - so other drivers could bind and might possibly interfere  with ACPI.

Well I had an idea after looking at k8temp -- why not make it default to
doing only reads from the sensor?  You'd only get information from whatever
core/sensor combination that ACPI had last used, but it would be safe.

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Chuck Ebbert]

Rudolf Marek wrote:
> Hello Chuck,
> 
> I'm the author of K8temp. Please can you share with us your DSDT table?
> (cat /proc/acpi/dsdt > /tmp/dsdt.bin)
> 

The system is Compaq Presario V2300 series notebook.  I won't be able
to get the DSDT until tomorrow or Monday.

>> So, could ACPI and the k8temp driver be at odds?
> 
> Yes because ACPI AML code has no synchronization with Linux drivers. Second
> reason is that ACPI AML code assign resource regions to itself but with cleared
> busy flag - so other drivers could bind and might possibly interfere  with ACPI.

Well I had an idea after looking at k8temp -- why not make it default to
doing only reads from the sensor?  You'd only get information from whatever
core/sensor combination that ACPI had last used, but it would be safe.

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Chuck,

On Fri, 16 Feb 2007 14:31:44 -0500, Chuck Ebbert wrote:
> Andi Kleen wrote:
> > Chuck Ebbert wrote:
> > > So, could ACPI and the k8temp driver be at odds?
> > 
> > Yes, there is no locking between ACPI and Linux drivers for register access.
> > e.g. if there is a indexed register both try to access (and temperature
> > sensors tend to use these things) they can race and get corrupted data.
> 
> I blacklisted the k8temp driver and everything looks OK now.

You could blacklist the ACPI "thermal" module instead. If you're
interested in monitoring your CPU temperature, k8temp is IMHO more
convenient to use than ACPI, as it interfaces properly with libsensors
and all hardware monitoring user interfaces.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Chuck,

On Fri, 16 Feb 2007 14:31:44 -0500, Chuck Ebbert wrote:
> Andi Kleen wrote:
> > Chuck Ebbert wrote:
> > > So, could ACPI and the k8temp driver be at odds?
> > 
> > Yes, there is no locking between ACPI and Linux drivers for register access.
> > e.g. if there is a indexed register both try to access (and temperature
> > sensors tend to use these things) they can race and get corrupted data.
> 
> I blacklisted the k8temp driver and everything looks OK now.

You could blacklist the ACPI "thermal" module instead. If you're
interested in monitoring your CPU temperature, k8temp is IMHO more
convenient to use than ACPI, as it interfaces properly with libsensors
and all hardware monitoring user interfaces.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Sat, 17 Feb 2007 13:14:50 -0500, Chuck Ebbert wrote:
> Well I had an idea after looking at k8temp -- why not make it default to
> doing only reads from the sensor?  You'd only get information from whatever
> core/sensor combination that ACPI had last used, but it would be safe.

ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
doesn't conflict with only k8temp, but with virtually all hardware
monitoring drivers, all I2C/SMBus drivers, and probably other types of
drivers too. We just can't restrict or blacklist all these drivers
because ACPI misbehaves.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Sat, 17 Feb 2007 13:14:50 -0500, Chuck Ebbert wrote:
> Well I had an idea after looking at k8temp -- why not make it default to
> doing only reads from the sensor?  You'd only get information from whatever
> core/sensor combination that ACPI had last used, but it would be safe.

ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
doesn't conflict with only k8temp, but with virtually all hardware
monitoring drivers, all I2C/SMBus drivers, and probably other types of
drivers too. We just can't restrict or blacklist all these drivers
because ACPI misbehaves.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Rudolf Marek]

Hello all,

I got the DSDT from chuck and it seems there is nothing interesting - no
declaration of PCI_config for the registers. If someone wants to check it I can
send him the DSDT.

_TMP looks like this:

 Store (\_SB.PCI0.LPC0.EC0.RTMP, Local0)
                    Store ("Current temp is: ", Debug)

                    Store (Local0, Debug)

                    Store (Local0, \_SB.CM25)

                    Return (Add (0x0AAC, Multiply (Local0, 0x0A)))

This looks quite OK LPC0.EC0 is embedded controller IO RAM at 0x62. Nothing
special. I guess some SMM interrupt is reading the the PCI regs and sends it to EC.


Chuck, please can you try to use your script which reads the temps from acpi
once a second and in other console run following command: (as root)

 watch -d -n 1 lspci -xxx -s 18.3

This will list the pci registers every second and marks the diff. You should see
changes at 0xe6 which is your temperature. However if you see 0xe4 changing and
the k8temp driver is NOT loaded - it means something else is doing that.
If you spot the 0xe4 changes, please stop your script and check if it is still
changing.


Maybe you will not be able to see any changes at all - then the SMM is doing
reads only. Also you did not say how many cores, or places temps you see using
the k8temp driver. You have dual core with two possible places?

Thanks
Rudolf

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Rudolf Marek]

Hello all,

I got the DSDT from chuck and it seems there is nothing interesting - no
declaration of PCI_config for the registers. If someone wants to check it I can
send him the DSDT.

_TMP looks like this:

 Store (\_SB.PCI0.LPC0.EC0.RTMP, Local0)
                    Store ("Current temp is: ", Debug)

                    Store (Local0, Debug)

                    Store (Local0, \_SB.CM25)

                    Return (Add (0x0AAC, Multiply (Local0, 0x0A)))

This looks quite OK LPC0.EC0 is embedded controller IO RAM at 0x62. Nothing
special. I guess some SMM interrupt is reading the the PCI regs and sends it to EC.


Chuck, please can you try to use your script which reads the temps from acpi
once a second and in other console run following command: (as root)

 watch -d -n 1 lspci -xxx -s 18.3

This will list the pci registers every second and marks the diff. You should see
changes at 0xe6 which is your temperature. However if you see 0xe4 changing and
the k8temp driver is NOT loaded - it means something else is doing that.
If you spot the 0xe4 changes, please stop your script and check if it is still
changing.


Maybe you will not be able to see any changes at all - then the SMM is doing
reads only. Also you did not say how many cores, or places temps you see using
the k8temp driver. You have dual core with two possible places?

Thanks
Rudolf

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Andi Kleen]

> You could blacklist the ACPI "thermal" module instead. If you're
> interested in monitoring your CPU temperature, k8temp is IMHO more
> convenient to use than ACPI, as it interfaces properly with libsensors
> and all hardware monitoring user interfaces.

That would be somewhat dangerous because ACPI thermal does more than
just displaying the temperature. Especially laptops often need it.

-Andi

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Andi Kleen]

> You could blacklist the ACPI "thermal" module instead. If you're
> interested in monitoring your CPU temperature, k8temp is IMHO more
> convenient to use than ACPI, as it interfaces properly with libsensors
> and all hardware monitoring user interfaces.

That would be somewhat dangerous because ACPI thermal does more than
just displaying the temperature. Especially laptops often need it.

-Andi

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Chuck Ebbert]

Rudolf Marek wrote:
> Hello all,
> 
> I got the DSDT from chuck and it seems there is nothing interesting - no
> declaration of PCI_config for the registers. If someone wants to check it I can
> send him the DSDT.
> 
> _TMP looks like this:
> 
>  Store (\_SB.PCI0.LPC0.EC0.RTMP, Local0)
>                     Store ("Current temp is: ", Debug)
> 
>                     Store (Local0, Debug)
> 
>                     Store (Local0, \_SB.CM25)
> 
>                     Return (Add (0x0AAC, Multiply (Local0, 0x0A)))
> 
> This looks quite OK LPC0.EC0 is embedded controller IO RAM at 0x62. Nothing
> special. I guess some SMM interrupt is reading the the PCI regs and sends it to EC.

I blacklisted the k8temp driver (and the out-of-tree k8_edac driver
in Fedora) and the temps were still volatile, so that's not causing
it. Since then I've upgraded the system BIOS from F.06 to F.27 and
the problems _may_ have gone away. My own custom 2.6.19 kernel has
never been a problem, so I'm thinking it's one of these drivers
loaded by Fedora that I never even compile:

	i2c_core
	i2c_ec
	i2c_piix4
	asus_acpi (on a Compaq???)
	sbs

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Chuck Ebbert]

Rudolf Marek wrote:
> Hello all,
> 
> I got the DSDT from chuck and it seems there is nothing interesting - no
> declaration of PCI_config for the registers. If someone wants to check it I can
> send him the DSDT.
> 
> _TMP looks like this:
> 
>  Store (\_SB.PCI0.LPC0.EC0.RTMP, Local0)
>                     Store ("Current temp is: ", Debug)
> 
>                     Store (Local0, Debug)
> 
>                     Store (Local0, \_SB.CM25)
> 
>                     Return (Add (0x0AAC, Multiply (Local0, 0x0A)))
> 
> This looks quite OK LPC0.EC0 is embedded controller IO RAM at 0x62. Nothing
> special. I guess some SMM interrupt is reading the the PCI regs and sends it to EC.

I blacklisted the k8temp driver (and the out-of-tree k8_edac driver
in Fedora) and the temps were still volatile, so that's not causing
it. Since then I've upgraded the system BIOS from F.06 to F.27 and
the problems _may_ have gone away. My own custom 2.6.19 kernel has
never been a problem, so I'm thinking it's one of these drivers
loaded by Fedora that I never even compile:

	i2c_core
	i2c_ec
	i2c_piix4
	asus_acpi (on a Compaq???)
	sbs

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Matthew Garrett]

On Sun, Feb 18, 2007 at 06:38:05PM +0100, Jean Delvare wrote:

> ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
> doesn't conflict with only k8temp, but with virtually all hardware
> monitoring drivers, all I2C/SMBus drivers, and probably other types of
> drivers too. We just can't restrict or blacklist all these drivers
> because ACPI misbehaves.

No, the simple fact of the matter is that if you're running on an ACPI 
platform you need to change some of your assumptions. ACPI owns the 
hardware. The OS doesn't. To an extent this has always been true on 
laptops and servers /anyway/ - the BIOS is free to have a wide variety 
of SMM insanity that invalidates basic assumptions like "If I hold this 
lock, nothing can interrupt me between this write and this read". That's 
simply not true.

So this isn't about fixing ACPI. It's about trying to find a mechanism 
that allows ACPI and raw hardware drivers to coexist, which is made 
somewhat harder by it not being a situation that the platform designers 
have considered in the slightest. The suggested low-level driver for 
io-port arbitration would certainly be a step forward in making this 
work better.

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

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Matthew Garrett]

On Sun, Feb 18, 2007 at 06:38:05PM +0100, Jean Delvare wrote:

> ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
> doesn't conflict with only k8temp, but with virtually all hardware
> monitoring drivers, all I2C/SMBus drivers, and probably other types of
> drivers too. We just can't restrict or blacklist all these drivers
> because ACPI misbehaves.

No, the simple fact of the matter is that if you're running on an ACPI 
platform you need to change some of your assumptions. ACPI owns the 
hardware. The OS doesn't. To an extent this has always been true on 
laptops and servers /anyway/ - the BIOS is free to have a wide variety 
of SMM insanity that invalidates basic assumptions like "If I hold this 
lock, nothing can interrupt me between this write and this read". That's 
simply not true.

So this isn't about fixing ACPI. It's about trying to find a mechanism 
that allows ACPI and raw hardware drivers to coexist, which is made 
somewhat harder by it not being a situation that the platform designers 
have considered in the slightest. The suggested low-level driver for 
io-port arbitration would certainly be a step forward in making this 
work better.

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

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Luca Tettamanti]

On 2/20/07, Matthew Garrett <mjg59@srcf.ucam.org> wrote:
> On Sun, Feb 18, 2007 at 06:38:05PM +0100, Jean Delvare wrote:
>
> > ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
> > doesn't conflict with only k8temp, but with virtually all hardware
> > monitoring drivers, all I2C/SMBus drivers, and probably other types of
> > drivers too. We just can't restrict or blacklist all these drivers
> > because ACPI misbehaves.
>
> No, the simple fact of the matter is that if you're running on an ACPI
> platform you need to change some of your assumptions. ACPI owns the
> hardware. The OS doesn't. To an extent this has always been true on
> laptops and servers /anyway/ - the BIOS is free to have a wide variety
> of SMM insanity that invalidates basic assumptions like "If I hold this
> lock, nothing can interrupt me between this write and this read". That's
> simply not true.
>
> So this isn't about fixing ACPI. It's about trying to find a mechanism
> that allows ACPI and raw hardware drivers to coexist, which is made
> somewhat harder by it not being a situation that the platform designers
> have considered in the slightest.

Motherboard vendors usually provide tools for $(TheOtherOS) that can
read from all thermal  / fan / voltage / whatever sensors, so I guess
it's possible to make the ACPI driver and the "raw" one play nice with
each other[1].

Luca
[1] Unless their solution is "poke at the hardware and hope that ACPI
doesn't blow up", that is.

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Luca Tettamanti]

On 2/20/07, Matthew Garrett <mjg59 at srcf.ucam.org> wrote:
> On Sun, Feb 18, 2007 at 06:38:05PM +0100, Jean Delvare wrote:
>
> > ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
> > doesn't conflict with only k8temp, but with virtually all hardware
> > monitoring drivers, all I2C/SMBus drivers, and probably other types of
> > drivers too. We just can't restrict or blacklist all these drivers
> > because ACPI misbehaves.
>
> No, the simple fact of the matter is that if you're running on an ACPI
> platform you need to change some of your assumptions. ACPI owns the
> hardware. The OS doesn't. To an extent this has always been true on
> laptops and servers /anyway/ - the BIOS is free to have a wide variety
> of SMM insanity that invalidates basic assumptions like "If I hold this
> lock, nothing can interrupt me between this write and this read". That's
> simply not true.
>
> So this isn't about fixing ACPI. It's about trying to find a mechanism
> that allows ACPI and raw hardware drivers to coexist, which is made
> somewhat harder by it not being a situation that the platform designers
> have considered in the slightest.

Motherboard vendors usually provide tools for $(TheOtherOS) that can
read from all thermal  / fan / voltage / whatever sensors, so I guess
it's possible to make the ACPI driver and the "raw" one play nice with
each other[1].

Luca
[1] Unless their solution is "poke at the hardware and hope that ACPI
doesn't blow up", that is.

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Dave Jones]

On Tue, Feb 20, 2007 at 10:08:26AM -0500, Chuck Ebbert wrote:
 
 > 	i2c_core
 > 	i2c_ec
 > 	i2c_piix4
 > 	asus_acpi (on a Compaq???)
 > 	sbs

Something is pulling in asus_acpi as a dependancy. I've never
figured out what the cause is.  For a long time I was thinking
that we had an explicit modprobe for it in an initscript, but
grepping for it in /etc turns up zip.

		Dave

-- 
http://www.codemonkey.org.uk

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Dave Jones]

On Tue, Feb 20, 2007 at 10:08:26AM -0500, Chuck Ebbert wrote:
 
 > 	i2c_core
 > 	i2c_ec
 > 	i2c_piix4
 > 	asus_acpi (on a Compaq???)
 > 	sbs

Something is pulling in asus_acpi as a dependancy. I've never
figured out what the cause is.  For a long time I was thinking
that we had an explicit modprobe for it in an initscript, but
grepping for it in /etc turns up zip.

		Dave

-- 
http://www.codemonkey.org.uk

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Chuck,

On Tue, 20 Feb 2007 10:08:26 -0500, Chuck Ebbert wrote:
> I blacklisted the k8temp driver (and the out-of-tree k8_edac driver
> in Fedora) and the temps were still volatile, so that's not causing
> it. Since then I've upgraded the system BIOS from F.06 to F.27 and
> the problems _may_ have gone away. My own custom 2.6.19 kernel has
> never been a problem, so I'm thinking it's one of these drivers
> loaded by Fedora that I never even compile:
> 
> 	i2c_core

i2c-core doesn't touch the hardware by itself.

> 	i2c_ec

Presumably autoloaded by the ACPI subsystem, I guess your ACPI
implementation includes an SMBus 2.0 controller.

> 	i2c_piix4

i2c-piix4 will autoload if a supported PCI device is found on your
system. Assuming this is the same physical bus as i2c_ec is exposing,
it's no good to load both i2c-piix4 and i2c_ec at the same time.
Unfortunately i2c_ec doesn't request the I/O resources it uses so this
kind of conflict cannot be avoided currently.

Can you try to load the i2c-dev driver, then run the following commands
and report the results:
$ i2cdetect -l
For each bus listed:
$ i2cdetect N
(where N is the i2c bus number)

> 	asus_acpi (on a Compaq???)
> 	sbs

This is a new battery driver used in conjunction with i2c_ec. I guess
you have a smart battery in your laptop which is accessed through
the SMBus. I found that this driver bypasses the i2c-core locking,
which is really bad. I reported it one week ago:
http://marc.theaimsgroup.com/?l=linux-acpi&m=117160531631100&w=2
(for some reason my original post wasn't archived)
My patch wasn't applied, but the problems you describe could well be
caused by this locking issue. So I suggest that you unload the sbs
driver and see if things get better. If they do, you could try to apply
my patch and load sbs again, and see if it fixes it.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Chuck,

On Tue, 20 Feb 2007 10:08:26 -0500, Chuck Ebbert wrote:
> I blacklisted the k8temp driver (and the out-of-tree k8_edac driver
> in Fedora) and the temps were still volatile, so that's not causing
> it. Since then I've upgraded the system BIOS from F.06 to F.27 and
> the problems _may_ have gone away. My own custom 2.6.19 kernel has
> never been a problem, so I'm thinking it's one of these drivers
> loaded by Fedora that I never even compile:
> 
> 	i2c_core

i2c-core doesn't touch the hardware by itself.

> 	i2c_ec

Presumably autoloaded by the ACPI subsystem, I guess your ACPI
implementation includes an SMBus 2.0 controller.

> 	i2c_piix4

i2c-piix4 will autoload if a supported PCI device is found on your
system. Assuming this is the same physical bus as i2c_ec is exposing,
it's no good to load both i2c-piix4 and i2c_ec at the same time.
Unfortunately i2c_ec doesn't request the I/O resources it uses so this
kind of conflict cannot be avoided currently.

Can you try to load the i2c-dev driver, then run the following commands
and report the results:
$ i2cdetect -l
For each bus listed:
$ i2cdetect N
(where N is the i2c bus number)

> 	asus_acpi (on a Compaq???)
> 	sbs

This is a new battery driver used in conjunction with i2c_ec. I guess
you have a smart battery in your laptop which is accessed through
the SMBus. I found that this driver bypasses the i2c-core locking,
which is really bad. I reported it one week ago:
http://marc.theaimsgroup.com/?l=linux-acpi&m\x117160531631100&w=2
(for some reason my original post wasn't archived)
My patch wasn't applied, but the problems you describe could well be
caused by this locking issue. So I suggest that you unload the sbs
driver and see if things get better. If they do, you could try to apply
my patch and load sbs again, and see if it fixes it.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Luca,

On Tue, 20 Feb 2007 16:33:56 +0100, Luca Tettamanti wrote:
> Motherboard vendors usually provide tools for $(TheOtherOS) that can
> read from all thermal  / fan / voltage / whatever sensors, so I guess
> it's possible to make the ACPI driver and the "raw" one play nice with
> each other[1].
> 
> Luca
> [1] Unless their solution is "poke at the hardware and hope that ACPI
> doesn't blow up", that is.

Without the sources it's hard to tell. And all these applications are
vendor-specific, so if they indeed have ways to avoid conflicting
accesses between ACPI and the rest of the system, these ways are likely
to be vendor-specific as well, and not documented.

Either way, this means we need the support from hardware vendors to
solve this concurrent access problem, and unfortunately I doubt this
happens anytime soon :(

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Luca,

On Tue, 20 Feb 2007 16:33:56 +0100, Luca Tettamanti wrote:
> Motherboard vendors usually provide tools for $(TheOtherOS) that can
> read from all thermal  / fan / voltage / whatever sensors, so I guess
> it's possible to make the ACPI driver and the "raw" one play nice with
> each other[1].
> 
> Luca
> [1] Unless their solution is "poke at the hardware and hope that ACPI
> doesn't blow up", that is.

Without the sources it's hard to tell. And all these applications are
vendor-specific, so if they indeed have ways to avoid conflicting
accesses between ACPI and the rest of the system, these ways are likely
to be vendor-specific as well, and not documented.

Either way, this means we need the support from hardware vendors to
solve this concurrent access problem, and unfortunately I doubt this
happens anytime soon :(

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Matthew,

On Tue, 20 Feb 2007 15:18:13 +0000, Matthew Garrett wrote:
> On Sun, Feb 18, 2007 at 06:38:05PM +0100, Jean Delvare wrote:
> 
> > ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
> > doesn't conflict with only k8temp, but with virtually all hardware
> > monitoring drivers, all I2C/SMBus drivers, and probably other types of
> > drivers too. We just can't restrict or blacklist all these drivers
> > because ACPI misbehaves.
> 
> No, the simple fact of the matter is that if you're running on an ACPI 
> platform you need to change some of your assumptions. ACPI owns the 
> hardware. The OS doesn't. To an extent this has always been true on 

The Linux device driver model assumes that it owns the hardware. If
this is not true, then should we prevent any non-ACPI driver from
loading as soon as ACPI is enabled?

> laptops and servers /anyway/ - the BIOS is free to have a wide variety 
> of SMM insanity that invalidates basic assumptions like "If I hold this 
> lock, nothing can interrupt me between this write and this read". That's 
> simply not true.

Yeah, this is correct, and just as unfortunate. It's amazingly sad that
hardware vendors as a whole are still repeating the same design
mistakes over and over again :(

> So this isn't about fixing ACPI. It's about trying to find a mechanism 
> that allows ACPI and raw hardware drivers to coexist, which is made 

Exactly what I said, you're only rewording it to make it sound nicer ;)

> somewhat harder by it not being a situation that the platform designers 
> have considered in the slightest. The suggested low-level driver for 
> io-port arbitration would certainly be a step forward in making this 
> work better.

I sure hope we can find a solution, by as your said yourself, nothing
is going to prevent SMM and similar oddities from messing up the drivers
assumptions.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Matthew,

On Tue, 20 Feb 2007 15:18:13 +0000, Matthew Garrett wrote:
> On Sun, Feb 18, 2007 at 06:38:05PM +0100, Jean Delvare wrote:
> 
> > ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
> > doesn't conflict with only k8temp, but with virtually all hardware
> > monitoring drivers, all I2C/SMBus drivers, and probably other types of
> > drivers too. We just can't restrict or blacklist all these drivers
> > because ACPI misbehaves.
> 
> No, the simple fact of the matter is that if you're running on an ACPI 
> platform you need to change some of your assumptions. ACPI owns the 
> hardware. The OS doesn't. To an extent this has always been true on 

The Linux device driver model assumes that it owns the hardware. If
this is not true, then should we prevent any non-ACPI driver from
loading as soon as ACPI is enabled?

> laptops and servers /anyway/ - the BIOS is free to have a wide variety 
> of SMM insanity that invalidates basic assumptions like "If I hold this 
> lock, nothing can interrupt me between this write and this read". That's 
> simply not true.

Yeah, this is correct, and just as unfortunate. It's amazingly sad that
hardware vendors as a whole are still repeating the same design
mistakes over and over again :(

> So this isn't about fixing ACPI. It's about trying to find a mechanism 
> that allows ACPI and raw hardware drivers to coexist, which is made 

Exactly what I said, you're only rewording it to make it sound nicer ;)

> somewhat harder by it not being a situation that the platform designers 
> have considered in the slightest. The suggested low-level driver for 
> io-port arbitration would certainly be a step forward in making this 
> work better.

I sure hope we can find a solution, by as your said yourself, nothing
is going to prevent SMM and similar oddities from messing up the drivers
assumptions.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Chuck Ebbert]

Jean Delvare wrote:
> Can you try to load the i2c-dev driver, then run the following commands
> and report the results:
> $ i2cdetect -l
> For each bus listed:
> $ i2cdetect N

FWIW it's really an ATIIXP chipset, but supposedly PIIX4 compatible:

# i2cdetect -l
i2c-0   smbus           SMBus PIIX4 adapter at 8400             Non-I2C SMBus adapter
# i2cdetect 0
WARNING! This program can confuse your I2C bus, cause data loss and worse!
I will probe file /dev/i2c-0.
I will probe address range 0x03-0x77.
Continue? [Y/n] y
     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
00:          XX XX XX XX XX XX XX XX XX XX XX XX XX
10: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
20: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
30: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
40: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
50: 50 51 XX XX XX XX XX XX XX XX XX XX XX XX XX XX
60: XX XX XX XX XX XX XX XX XX 69 XX XX XX XX XX XX
70: XX XX XX XX XX XX XX XX

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Chuck Ebbert]

Jean Delvare wrote:
> Can you try to load the i2c-dev driver, then run the following commands
> and report the results:
> $ i2cdetect -l
> For each bus listed:
> $ i2cdetect N

FWIW it's really an ATIIXP chipset, but supposedly PIIX4 compatible:

# i2cdetect -l
i2c-0   smbus           SMBus PIIX4 adapter at 8400             Non-I2C SMBus adapter
# i2cdetect 0
WARNING! This program can confuse your I2C bus, cause data loss and worse!
I will probe file /dev/i2c-0.
I will probe address range 0x03-0x77.
Continue? [Y/n] y
     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
00:          XX XX XX XX XX XX XX XX XX XX XX XX XX
10: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
20: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
30: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
40: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
50: 50 51 XX XX XX XX XX XX XX XX XX XX XX XX XX XX
60: XX XX XX XX XX XX XX XX XX 69 XX XX XX XX XX XX
70: XX XX XX XX XX XX XX XX

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Tue, 20 Feb 2007 14:11:42 -0500, Dave Jones wrote:
> On Tue, Feb 20, 2007 at 10:08:26AM -0500, Chuck Ebbert wrote:
>  
>  > 	i2c_core
>  > 	i2c_ec
>  > 	i2c_piix4
>  > 	asus_acpi (on a Compaq???)
>  > 	sbs
> 
> Something is pulling in asus_acpi as a dependancy. I've never
> figured out what the cause is.  For a long time I was thinking
> that we had an explicit modprobe for it in an initscript, but
> grepping for it in /etc turns up zip.

How could it be, given that asus_acpi doesn't export any symbol?

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Tue, 20 Feb 2007 14:11:42 -0500, Dave Jones wrote:
> On Tue, Feb 20, 2007 at 10:08:26AM -0500, Chuck Ebbert wrote:
>  
>  > 	i2c_core
>  > 	i2c_ec
>  > 	i2c_piix4
>  > 	asus_acpi (on a Compaq???)
>  > 	sbs
> 
> Something is pulling in asus_acpi as a dependancy. I've never
> figured out what the cause is.  For a long time I was thinking
> that we had an explicit modprobe for it in an initscript, but
> grepping for it in /etc turns up zip.

How could it be, given that asus_acpi doesn't export any symbol?

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Wed, 21 Feb 2007 11:03:07 -0500, Chuck Ebbert wrote:
> Jean Delvare wrote:
> > Can you try to load the i2c-dev driver, then run the following commands
> > and report the results:
> > $ i2cdetect -l
> > For each bus listed:
> > $ i2cdetect N
> 
> FWIW it's really an ATIIXP chipset, but supposedly PIIX4 compatible:
> 
> # i2cdetect -l
> i2c-0   smbus           SMBus PIIX4 adapter at 8400             Non-I2C SMBus adapter

No i2c_ec. Maybe your distribution is loading it by default for
everyone then.

Either way, it means you can forget right away about sbs, if i2c_ec
didn't register a device, sbs cannot be used.

> # i2cdetect 0
> WARNING! This program can confuse your I2C bus, cause data loss and worse!
> I will probe file /dev/i2c-0.
> I will probe address range 0x03-0x77.
> Continue? [Y/n] y
>      0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
> 00:          XX XX XX XX XX XX XX XX XX XX XX XX XX
> 10: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
> 20: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
> 30: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
> 40: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
> 50: 50 51 XX XX XX XX XX XX XX XX XX XX XX XX XX XX
> 60: XX XX XX XX XX XX XX XX XX 69 XX XX XX XX XX XX
> 70: XX XX XX XX XX XX XX XX

Only a couple EEPROMs and a clock chip on your SMBus, it's very
unlikely that ACPI accesses this at all. So I'd be surprised that
i2c-piix4 is causing any trouble.

This leaves asus_acpi as the only candidate? Better unload _all_ the
drivers you consider as suspects, and see if it changes anything. I
guess not.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Wed, 21 Feb 2007 11:03:07 -0500, Chuck Ebbert wrote:
> Jean Delvare wrote:
> > Can you try to load the i2c-dev driver, then run the following commands
> > and report the results:
> > $ i2cdetect -l
> > For each bus listed:
> > $ i2cdetect N
> 
> FWIW it's really an ATIIXP chipset, but supposedly PIIX4 compatible:
> 
> # i2cdetect -l
> i2c-0   smbus           SMBus PIIX4 adapter at 8400             Non-I2C SMBus adapter

No i2c_ec. Maybe your distribution is loading it by default for
everyone then.

Either way, it means you can forget right away about sbs, if i2c_ec
didn't register a device, sbs cannot be used.

> # i2cdetect 0
> WARNING! This program can confuse your I2C bus, cause data loss and worse!
> I will probe file /dev/i2c-0.
> I will probe address range 0x03-0x77.
> Continue? [Y/n] y
>      0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
> 00:          XX XX XX XX XX XX XX XX XX XX XX XX XX
> 10: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
> 20: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
> 30: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
> 40: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX
> 50: 50 51 XX XX XX XX XX XX XX XX XX XX XX XX XX XX
> 60: XX XX XX XX XX XX XX XX XX 69 XX XX XX XX XX XX
> 70: XX XX XX XX XX XX XX XX

Only a couple EEPROMs and a clock chip on your SMBus, it's very
unlikely that ACPI accesses this at all. So I'd be surprised that
i2c-piix4 is causing any trouble.

This leaves asus_acpi as the only candidate? Better unload _all_ the
drivers you consider as suspects, and see if it changes anything. I
guess not.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Dave Jones]

On Wed, Feb 21, 2007 at 05:17:37PM +0100, Jean Delvare wrote:
 > On Tue, 20 Feb 2007 14:11:42 -0500, Dave Jones wrote:
 > > On Tue, Feb 20, 2007 at 10:08:26AM -0500, Chuck Ebbert wrote:
 > >  
 > >  > 	i2c_core
 > >  > 	i2c_ec
 > >  > 	i2c_piix4
 > >  > 	asus_acpi (on a Compaq???)
 > >  > 	sbs
 > > 
 > > Something is pulling in asus_acpi as a dependancy. I've never
 > > figured out what the cause is.  For a long time I was thinking
 > > that we had an explicit modprobe for it in an initscript, but
 > > grepping for it in /etc turns up zip.
 > 
 > How could it be, given that asus_acpi doesn't export any symbol?
 
If I knew I'd have fixed it by now.

		Dave

-- 
http://www.codemonkey.org.uk

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Dave Jones]

On Wed, Feb 21, 2007 at 05:17:37PM +0100, Jean Delvare wrote:
 > On Tue, 20 Feb 2007 14:11:42 -0500, Dave Jones wrote:
 > > On Tue, Feb 20, 2007 at 10:08:26AM -0500, Chuck Ebbert wrote:
 > >  
 > >  > 	i2c_core
 > >  > 	i2c_ec
 > >  > 	i2c_piix4
 > >  > 	asus_acpi (on a Compaq???)
 > >  > 	sbs
 > > 
 > > Something is pulling in asus_acpi as a dependancy. I've never
 > > figured out what the cause is.  For a long time I was thinking
 > > that we had an explicit modprobe for it in an initscript, but
 > > grepping for it in /etc turns up zip.
 > 
 > How could it be, given that asus_acpi doesn't export any symbol?
 
If I knew I'd have fixed it by now.

		Dave

-- 
http://www.codemonkey.org.uk

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Dave Jones]

On Wed, Feb 21, 2007 at 12:37:40PM -0500, Dave Jones wrote:
 > On Wed, Feb 21, 2007 at 05:17:37PM +0100, Jean Delvare wrote:
 >  > On Tue, 20 Feb 2007 14:11:42 -0500, Dave Jones wrote:
 >  > > On Tue, Feb 20, 2007 at 10:08:26AM -0500, Chuck Ebbert wrote:
 >  > >  
 >  > >  > 	i2c_core
 >  > >  > 	i2c_ec
 >  > >  > 	i2c_piix4
 >  > >  > 	asus_acpi (on a Compaq???)
 >  > >  > 	sbs
 >  > > 
 >  > > Something is pulling in asus_acpi as a dependancy. I've never
 >  > > figured out what the cause is.  For a long time I was thinking
 >  > > that we had an explicit modprobe for it in an initscript, but
 >  > > grepping for it in /etc turns up zip.
 >  > 
 >  > How could it be, given that asus_acpi doesn't export any symbol?
 >  
 > If I knew I'd have fixed it by now.

Ah, Fedora has this horror in its initscripts (which explains why I missed
it in my grep)..

# Initialize ACPI bits
if [ -d /proc/acpi ]; then
    for module in /lib/modules/$unamer/kernel/drivers/acpi/* ; do
        module=${module##*/}
        module=${module%.ko}
        modprobe $module >/dev/null 2>&1
    done
fi


This is there because there's no clean way for userspace to know whether
to load the system specific stuff right now.   Bill Nottingham pointed
out that we could add a /sys/class/dmi/modalias and appropriate MODULE_DMI
tags to the various modules like asus_acpi to make udev autoload them.

		Dave

-- 
http://www.codemonkey.org.uk

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Dave Jones]

On Wed, Feb 21, 2007 at 12:37:40PM -0500, Dave Jones wrote:
 > On Wed, Feb 21, 2007 at 05:17:37PM +0100, Jean Delvare wrote:
 >  > On Tue, 20 Feb 2007 14:11:42 -0500, Dave Jones wrote:
 >  > > On Tue, Feb 20, 2007 at 10:08:26AM -0500, Chuck Ebbert wrote:
 >  > >  
 >  > >  > 	i2c_core
 >  > >  > 	i2c_ec
 >  > >  > 	i2c_piix4
 >  > >  > 	asus_acpi (on a Compaq???)
 >  > >  > 	sbs
 >  > > 
 >  > > Something is pulling in asus_acpi as a dependancy. I've never
 >  > > figured out what the cause is.  For a long time I was thinking
 >  > > that we had an explicit modprobe for it in an initscript, but
 >  > > grepping for it in /etc turns up zip.
 >  > 
 >  > How could it be, given that asus_acpi doesn't export any symbol?
 >  
 > If I knew I'd have fixed it by now.

Ah, Fedora has this horror in its initscripts (which explains why I missed
it in my grep)..

# Initialize ACPI bits
if [ -d /proc/acpi ]; then
    for module in /lib/modules/$unamer/kernel/drivers/acpi/* ; do
        module=${module##*/}
        module=${module%.ko}
        modprobe $module >/dev/null 2>&1
    done
fi


This is there because there's no clean way for userspace to know whether
to load the system specific stuff right now.   Bill Nottingham pointed
out that we could add a /sys/class/dmi/modalias and appropriate MODULE_DMI
tags to the various modules like asus_acpi to make udev autoload them.

		Dave

-- 
http://www.codemonkey.org.uk

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Wed, 21 Feb 2007 15:19:44 -0500, Dave Jones wrote:
> Ah, Fedora has this horror in its initscripts (which explains why I missed
> it in my grep)..
> 
> # Initialize ACPI bits
> if [ -d /proc/acpi ]; then
>     for module in /lib/modules/$unamer/kernel/drivers/acpi/* ; do
>         module=${module##*/}
>         module=${module%.ko}
>         modprobe $module >/dev/null 2>&1
>     done
> fi

Ah, this also explains why the i2c_ec and sbs drivers were loaded on
Chuck's system, although they were not needed.

> This is there because there's no clean way for userspace to know whether
> to load the system specific stuff right now.   Bill Nottingham pointed
> out that we could add a /sys/class/dmi/modalias and appropriate MODULE_DMI
> tags to the various modules like asus_acpi to make udev autoload them.

Something similiar should be doable for i2c_ec, as it's only useful if a
given ACPI object is present. sbs, in turn, is only useful if i2c_ec is
loaded.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Wed, 21 Feb 2007 15:19:44 -0500, Dave Jones wrote:
> Ah, Fedora has this horror in its initscripts (which explains why I missed
> it in my grep)..
> 
> # Initialize ACPI bits
> if [ -d /proc/acpi ]; then
>     for module in /lib/modules/$unamer/kernel/drivers/acpi/* ; do
>         module=${module##*/}
>         module=${module%.ko}
>         modprobe $module >/dev/null 2>&1
>     done
> fi

Ah, this also explains why the i2c_ec and sbs drivers were loaded on
Chuck's system, although they were not needed.

> This is there because there's no clean way for userspace to know whether
> to load the system specific stuff right now.   Bill Nottingham pointed
> out that we could add a /sys/class/dmi/modalias and appropriate MODULE_DMI
> tags to the various modules like asus_acpi to make udev autoload them.

Something similiar should be doable for i2c_ec, as it's only useful if a
given ACPI object is present. sbs, in turn, is only useful if i2c_ec is
loaded.

-- 
Jean Delvare

Re: Could the k8temp driver be interfering with ACPI?

[Hans de Goede]

Jean Delvare wrote:
> On Wed, 21 Feb 2007 15:19:44 -0500, Dave Jones wrote:
>> Ah, Fedora has this horror in its initscripts (which explains why I missed
>> it in my grep)..
>>
>> # Initialize ACPI bits
>> if [ -d /proc/acpi ]; then
>>     for module in /lib/modules/$unamer/kernel/drivers/acpi/* ; do
>>         module=${module##*/}
>>         module=${module%.ko}
>>         modprobe $module >/dev/null 2>&1
>>     done
>> fi
> 
> Ah, this also explains why the i2c_ec and sbs drivers were loaded on
> Chuck's system, although they were not needed.
> 
>> This is there because there's no clean way for userspace to know whether
>> to load the system specific stuff right now.   Bill Nottingham pointed
>> out that we could add a /sys/class/dmi/modalias and appropriate MODULE_DMI
>> tags to the various modules like asus_acpi to make udev autoload them.
> 
> Something similiar should be doable for i2c_ec, as it's only useful if a
> given ACPI object is present. sbs, in turn, is only useful if i2c_ec is
> loaded.
> 

I'm thinking that it might be an idea to also use this idea of udev autoloading 
through DMI info for the abituguru and abituguru3 driver (review please). The 
both only support about 12 motherboards. For the abituguru driver, dmi info 
could also be used to automatically set the module options needed on the 2 
oldest uguru featuring abit motherboards. What do you think about this?

Regards,

Hans



_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Hans de Goede]

Jean Delvare wrote:
> On Wed, 21 Feb 2007 15:19:44 -0500, Dave Jones wrote:
>> Ah, Fedora has this horror in its initscripts (which explains why I missed
>> it in my grep)..
>>
>> # Initialize ACPI bits
>> if [ -d /proc/acpi ]; then
>>     for module in /lib/modules/$unamer/kernel/drivers/acpi/* ; do
>>         module=${module##*/}
>>         module=${module%.ko}
>>         modprobe $module >/dev/null 2>&1
>>     done
>> fi
> 
> Ah, this also explains why the i2c_ec and sbs drivers were loaded on
> Chuck's system, although they were not needed.
> 
>> This is there because there's no clean way for userspace to know whether
>> to load the system specific stuff right now.   Bill Nottingham pointed
>> out that we could add a /sys/class/dmi/modalias and appropriate MODULE_DMI
>> tags to the various modules like asus_acpi to make udev autoload them.
> 
> Something similiar should be doable for i2c_ec, as it's only useful if a
> given ACPI object is present. sbs, in turn, is only useful if i2c_ec is
> loaded.
> 

I'm thinking that it might be an idea to also use this idea of udev autoloading 
through DMI info for the abituguru and abituguru3 driver (review please). The 
both only support about 12 motherboards. For the abituguru driver, dmi info 
could also be used to automatically set the module options needed on the 2 
oldest uguru featuring abit motherboards. What do you think about this?

Regards,

Hans

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Hans de Goede]

Jean Delvare wrote:
> On Wed, 21 Feb 2007 15:19:44 -0500, Dave Jones wrote:
>> Ah, Fedora has this horror in its initscripts (which explains why I missed
>> it in my grep)..
>>
>> # Initialize ACPI bits
>> if [ -d /proc/acpi ]; then
>>     for module in /lib/modules/$unamer/kernel/drivers/acpi/* ; do
>>         module=${module##*/}
>>         module=${module%.ko}
>>         modprobe $module >/dev/null 2>&1
>>     done
>> fi
> 
> Ah, this also explains why the i2c_ec and sbs drivers were loaded on
> Chuck's system, although they were not needed.
> 
>> This is there because there's no clean way for userspace to know whether
>> to load the system specific stuff right now.   Bill Nottingham pointed
>> out that we could add a /sys/class/dmi/modalias and appropriate MODULE_DMI
>> tags to the various modules like asus_acpi to make udev autoload them.
> 
> Something similiar should be doable for i2c_ec, as it's only useful if a
> given ACPI object is present. sbs, in turn, is only useful if i2c_ec is
> loaded.
> 

I'm thinking that it might be an idea to also use this idea of udev autoloading 
through DMI info for the abituguru and abituguru3 driver (review please). The 
both only support about 12 motherboards. For the abituguru driver, dmi info 
could also be used to automatically set the module options needed on the 2 
oldest uguru featuring abit motherboards. What do you think about this?

Regards,

Hans

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Fri, 23 Feb 2007 08:13:30 +0100, Hans de Goede wrote:
> I'm thinking that it might be an idea to also use this idea of udev autoloading 
> through DMI info for the abituguru and abituguru3 driver (review please). The 
> both only support about 12 motherboards. For the abituguru driver, dmi info 
> could also be used to automatically set the module options needed on the 2 
> oldest uguru featuring abit motherboards. What do you think about this?

Given that the uguru chips are hard (impossible) to detect and only a
small number of boards need it, yes, I think it's a good idea.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Fri, 23 Feb 2007 08:13:30 +0100, Hans de Goede wrote:
> I'm thinking that it might be an idea to also use this idea of udev autoloading 
> through DMI info for the abituguru and abituguru3 driver (review please). The 
> both only support about 12 motherboards. For the abituguru driver, dmi info 
> could also be used to automatically set the module options needed on the 2 
> oldest uguru featuring abit motherboards. What do you think about this?

Given that the uguru chips are hard (impossible) to detect and only a
small number of boards need it, yes, I think it's a good idea.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Hans de Goede]

Jean Delvare wrote:
> On Fri, 23 Feb 2007 08:13:30 +0100, Hans de Goede wrote:
>> I'm thinking that it might be an idea to also use this idea of udev autoloading 
>> through DMI info for the abituguru and abituguru3 driver (review please). The 
>> both only support about 12 motherboards. For the abituguru driver, dmi info 
>> could also be used to automatically set the module options needed on the 2 
>> oldest uguru featuring abit motherboards. What do you think about this?
> 
> Given that the uguru chips are hard (impossible) to detect and only a
> small number of boards need it, yes, I think it's a good idea.
> 

<snipping CC-list down a bit>

I'm glad you like it, but how does this relate to using dmi info in 
sensors-detect, if we add that capability to sensors-detect and add a database 
with known motherboards and there configs, then that would be a doublure with 
having the info in the drivers, this is the main reason why I hesitate.

Regards,

Hans

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Hans,

On Fri, 23 Feb 2007 09:08:30 +0100, Hans de Goede wrote:
> I'm glad you like it, but how does this relate to using dmi info in 
> sensors-detect, if we add that capability to sensors-detect and add a database 
> with known motherboards and there configs, then that would be a doublure with 
> having the info in the drivers, this is the main reason why I hesitate.

Very good point.

I believe that the abituguru case is a bit special. The DMI data is not
only useful to find out which boards need the driver. It is also useful
to prevent the driver from being loaded on random hardware. Given that
the chip is impossible to detect, it would sound sensible to at least
prevent the abituguru driver from being loaded on non-Abit systems. If
the list of supported boards is short and well known, including it too
can't hurt.

But we can't do that for the rest of the hwmon drivers. The list of
supported boards for each driver is too long and growing too fast to
live in the kernel. For other drivers my plan is to just detect them
and trigger the module load. For Super-I/O chips, they have ID
registers, we can read them and have a table that match IDs to drivers.
For SMBus chip, once we have David Brownell's new i2c-core
infrastructure, we should be able to write a hwmon-smbus-probe drivers
which would probe the SMBus for chips and trigger the right module
loading.

But it might take a long time before we manage to implement all this.
In the meantime we have to let user-space deal with it, which is the
case at the moment, and it doesn't work so bad if you ask me.

As for sensors-detect itself, there's much more to do with the DMI data
than just finding which drivers need to be loaded. I don't think we
would bother adding DMI support to sensors-detect just for that, except
for the few boards with undetectable chips (uGuru) or which are
otherwise known to cause trouble when probed. The true value of using
DMI data in sensors-detect is to have a repository of configuration
files identified by DMI data, so that sensors-detect can fetch and
install a custom configuration file for your system. Being able to find
out the right drivers without probing is more or less a side effect.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Hans de Goede]

Jean Delvare wrote:
> Hi Hans,
> 
> On Fri, 23 Feb 2007 09:08:30 +0100, Hans de Goede wrote:
>> I'm glad you like it, but how does this relate to using dmi info in 
>> sensors-detect, if we add that capability to sensors-detect and add a database 
>> with known motherboards and there configs, then that would be a doublure with 
>> having the info in the drivers, this is the main reason why I hesitate.
> 
> Very good point.
> 
> I believe that the abituguru case is a bit special. The DMI data is not
> only useful to find out which boards need the driver. It is also useful
> to prevent the driver from being loaded on random hardware. Given that
> the chip is impossible to detect, it would sound sensible to at least
> prevent the abituguru driver from being loaded on non-Abit systems. If
> the list of supported boards is short and well known, including it too
> can't hurt.
> 

Ok, notice though that although the detection of the abituguru is indeed
very weak, it will probably cause tons of false positives, that the
driver will not actually load as the driver probe function will do a
full test-read, which will fail on anything but an uguru (1 in a million
chance this doesn't happen). The problems here are:
1) the read cycle does isa outb's which could upset non uguru hardware
2) at this point in the mod-loading stage, the insmod has succeeded
   confusing users, as the see the module but get no sysfs hwmon entries
   (the driver didn't load but the platform device was created)

Now 2 can be fixed by doing the test-read before creating the platform
device, but I don't know if thats worth it.

> As for sensors-detect itself, there's much more to do with the DMI data
> than just finding which drivers need to be loaded. I don't think we
> would bother adding DMI support to sensors-detect just for that, except
> for the few boards with undetectable chips (uGuru) or which are
> otherwise known to cause trouble when probed. The true value of using
> DMI data in sensors-detect is to have a repository of configuration
> files identified by DMI data, so that sensors-detect can fetch and
> install a custom configuration file for your system. Being able to find
> out the right drivers without probing is more or less a side effect.
> 

Agreed 100% this is _exactly_ the asignment I have given my students,
I've even told them to create a web interface to allow people to upload
correct sensors.conf settings for a certain motherboatd together with
the DMI data. Please read the document they have posted, it is in
English and it should contain a product description exactly matching
what you describe / want. If you want to fix the requirements I laid
down for this now would be THE time.

Regards,

Hans

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Mon, 26 Feb 2007 20:03:59 +0100, Hans de Goede wrote:
> Ok, notice though that although the detection of the abituguru is indeed
> very weak, it will probably cause tons of false positives, that the
> driver will not actually load as the driver probe function will do a
> full test-read, which will fail on anything but an uguru (1 in a million
> chance this doesn't happen). The problems here are:
> 1) the read cycle does isa outb's which could upset non uguru hardware

outb is, by definition, a write, which is why we should prevent it from
happening on systems without the relevant hardware.

> 2) at this point in the mod-loading stage, the insmod has succeeded
>    confusing users, as the see the module but get no sysfs hwmon entries
>    (the driver didn't load but the platform device was created)
> 
> Now 2 can be fixed by doing the test-read before creating the platform
> device, but I don't know if thats worth it.

Well there are two approaches:
1* Keep the platform driver approach, and have the platform device
created outside the abituguru driver. In this case it is OK to let the
abituguru driver load when no device is found, same as PCI drivers do.
* Think of the uGuru as a legacy ISA device. In that case the driver
should fail if the uGuru isn't found. Maybe you can look into the new
isa driver type (drivers/base/isa.c). I must look into it myself, it
might be the right solution to move the lm78 and w83781d drivers out of
using i2c-isa.

> Agreed 100% this is _exactly_ the asignment I have given my students,
> I've even told them to create a web interface to allow people to upload
> correct sensors.conf settings for a certain motherboatd together with
> the DMI data. Please read the document they have posted, it is in
> English and it should contain a product description exactly matching
> what you describe / want. If you want to fix the requirements I laid
> down for this now would be THE time.

I've read the "Background Information" and "Objectives" parts, overall
it looks good, except that a proofread of the document would be very
welcome. I'll comment in the adequate thread.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > Well I had an idea after looking at k8temp -- why not make it default to
> > doing only reads from the sensor?  You'd only get information from whatever
> > core/sensor combination that ACPI had last used, but it would be safe.
> 
> ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
> doesn't conflict with only k8temp, but with virtually all hardware
> monitoring drivers, all I2C/SMBus drivers, and probably other types of
> drivers too. We just can't restrict or blacklist all these drivers
> because ACPI misbehaves.

Oops, sorry about that but no, that will not work.

There's piece of paper, called ACPI specification, and we are
following it.

Bug is not in our implementation.

Bug is in the ACPI specs... it does not explicitely allow you to go
out and bitbang i2c, and you do it, and you get problems.

Now, you may try to change specs to be hwmon-friendly... good luck.

But currently hw manufacturers follow ACPI specs, so we have to follow
it, too; bad luck for hwmon. BIOS hiding smbus from you is good hint
you are doing something wrong...?
							Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > Well I had an idea after looking at k8temp -- why not make it default to
> > doing only reads from the sensor?  You'd only get information from whatever
> > core/sensor combination that ACPI had last used, but it would be safe.
> 
> ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
> doesn't conflict with only k8temp, but with virtually all hardware
> monitoring drivers, all I2C/SMBus drivers, and probably other types of
> drivers too. We just can't restrict or blacklist all these drivers
> because ACPI misbehaves.

Oops, sorry about that but no, that will not work.

There's piece of paper, called ACPI specification, and we are
following it.

Bug is not in our implementation.

Bug is in the ACPI specs... it does not explicitely allow you to go
out and bitbang i2c, and you do it, and you get problems.

Now, you may try to change specs to be hwmon-friendly... good luck.

But currently hw manufacturers follow ACPI specs, so we have to follow
it, too; bad luck for hwmon. BIOS hiding smbus from you is good hint
you are doing something wrong...?
							Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Pavel,

On Wed, 28 Feb 2007 21:38:04 +0000, Pavel Machek wrote:
> > > Well I had an idea after looking at k8temp -- why not make it default to
> > > doing only reads from the sensor?  You'd only get information from whatever
> > > core/sensor combination that ACPI had last used, but it would be safe.
> > 
> > ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
> > doesn't conflict with only k8temp, but with virtually all hardware
> > monitoring drivers, all I2C/SMBus drivers, and probably other types of
> > drivers too. We just can't restrict or blacklist all these drivers
> > because ACPI misbehaves.
> 
> Oops, sorry about that but no, that will not work.
> 
> There's piece of paper, called ACPI specification, and we are
> following it.

I never suggested otherwise. But the Linux 2.6 device driver model is
based, in part, on the fact that each driver must request the
resources it needs before actually using them. The acpi subsystem fails
to do that, so it is, in that sense, misbehaving. The is the root cause
of the problems people are reporting these days.

> Bug is not in our implementation.
> 
> Bug is in the ACPI specs... it does not explicitely allow you to go
> out and bitbang i2c, and you do it, and you get problems.

It really doesn't have anything to do with bit-banged I2C. It's about
SMBus master chips (most often embedded in south bridges), hardware
monitoring chips (be them SMBus-based or not), or basically any driver
which uses a resource ACPI is using as well (virtually any driver
accessing a Super-IO/LPC device is affected, for example.)

Secondly, your logic is somewhat broken. Just because one specification
doesn't explicitely allow me to do something, I must stop doing it?
There must be two dozen different specifications modern PC hardware
supposedly conforms to. If we stop doing everything which isn't
explicitely allowed by all of them, I fear there won't be much left ;)
A better question would be IMHO: Does the ACPI specification
explicitely _prohibits_ accessing directly to some categories of
hardware?

Lastly, I wholeheartedly agree that the core problem appears to be in
the ACPI design rather than in the Linux implementation. As a matter of
fact, other operating systems are facing the same problem [1]. The
problem isn't less real, though.

> Now, you may try to change specs to be hwmon-friendly... good luck.

I would like them to be driver-model-friendly, that's even a broader
challenge ;)

> But currently hw manufacturers follow ACPI specs, so we have to follow
> it, too; bad luck for hwmon. BIOS hiding smbus from you is good hint
> you are doing something wrong...?

I would love things to be that easy, but unfortunately they are not.

Firstly, the first records of hidden SMBus, in September 2000, predate
ACPI. All the early boards where the SMBus was hidden did not have ACPI
code poking at it at all, so this is definitely not the reason why it
was removed. The Asus P4 series is a good example of that. Unhiding the
SMBus on these boards actually let the user take benefit of the
hardware they had paid for.

Secondly, only a few south bridges are capable of hiding the SMBus. And
an even smaller number of systems have their SMBus actually hidden by
the BIOS. A much large number of systems, especially recent ones, have
ACPI poking at the hardware monitoring chip (again, SMBus-based or
not.) So there is no correlation between hidden SMBus and ACPI poking
at the hardware monitoring chip. So, even though we are very careful
these days when adding a quirk to unhide the SMBus on a new system (we
make sure that ACPI isn't accessing it), this is not going to solve the
general problem.

I would be happy to prevent SMBus and/or hardware monitoring drivers
from being loaded on ACPI-based system if we had a way to know which
systems do have ACPI code accessing these chips and which do not, and if
ACPI was offering a level of functionality comparable to what
individual hardware monitoring drivers offer. Unfortunately:

1* As far as I know, we currently have no way to know if the ACPI code
plans to ever access the hardware monitoring chip. If the acpi
subsystem could export this information somehow, it would help a lot.
But I'm not familiar with ACPI, so I don't know if this is feasable or
not. We just can't prevent the SMBus and hardware monitoring drivers
drivers from being loaded as soon as ACPI is enabled. This would
prevent a majority of users from using them, while they work fine for
most of them.

2* The hardware monitoring features offered by ACPI today are one level
of magnitude weaker than what lm_sensors was already offering back in
1999. The monitoring chips can do much but unfortunately ACPI only
exposes a very small subset of the chip features. ACPI doesn't handle
voltage monitoring at all. It usually reports no more than one fan, and
in my experience, more often than not, the speed is reported as a
boolean (spinning or not), when lm_sensors gives you the exact speed of
all your fans. ACPI thermal zones are not so bad, but the interface to
control them is ugly, and lm_sensors usually gives more details. And
more importantly, on almost all desktop and server motherboards, ACPI
does _not_ provide any hardware monitoring feature. So we really can't
go tell our users: give up the individual hardware monitoring drivers
and lm_sensors on all ACPI-enabled systems and use whatever ACPI offers
instead. This would be a major functionality loss for a vast majority
of users.

What do you propose?

[1] http://www.microsoft.com/whdc/system/pnppwr/powermgmt/BIOSAML.mspx

-- 
Jean Delvare

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Pavel,

On Wed, 28 Feb 2007 21:38:04 +0000, Pavel Machek wrote:
> > > Well I had an idea after looking at k8temp -- why not make it default to
> > > doing only reads from the sensor?  You'd only get information from whatever
> > > core/sensor combination that ACPI had last used, but it would be safe.
> > 
> > ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
> > doesn't conflict with only k8temp, but with virtually all hardware
> > monitoring drivers, all I2C/SMBus drivers, and probably other types of
> > drivers too. We just can't restrict or blacklist all these drivers
> > because ACPI misbehaves.
> 
> Oops, sorry about that but no, that will not work.
> 
> There's piece of paper, called ACPI specification, and we are
> following it.

I never suggested otherwise. But the Linux 2.6 device driver model is
based, in part, on the fact that each driver must request the
resources it needs before actually using them. The acpi subsystem fails
to do that, so it is, in that sense, misbehaving. The is the root cause
of the problems people are reporting these days.

> Bug is not in our implementation.
> 
> Bug is in the ACPI specs... it does not explicitely allow you to go
> out and bitbang i2c, and you do it, and you get problems.

It really doesn't have anything to do with bit-banged I2C. It's about
SMBus master chips (most often embedded in south bridges), hardware
monitoring chips (be them SMBus-based or not), or basically any driver
which uses a resource ACPI is using as well (virtually any driver
accessing a Super-IO/LPC device is affected, for example.)

Secondly, your logic is somewhat broken. Just because one specification
doesn't explicitely allow me to do something, I must stop doing it?
There must be two dozen different specifications modern PC hardware
supposedly conforms to. If we stop doing everything which isn't
explicitely allowed by all of them, I fear there won't be much left ;)
A better question would be IMHO: Does the ACPI specification
explicitely _prohibits_ accessing directly to some categories of
hardware?

Lastly, I wholeheartedly agree that the core problem appears to be in
the ACPI design rather than in the Linux implementation. As a matter of
fact, other operating systems are facing the same problem [1]. The
problem isn't less real, though.

> Now, you may try to change specs to be hwmon-friendly... good luck.

I would like them to be driver-model-friendly, that's even a broader
challenge ;)

> But currently hw manufacturers follow ACPI specs, so we have to follow
> it, too; bad luck for hwmon. BIOS hiding smbus from you is good hint
> you are doing something wrong...?

I would love things to be that easy, but unfortunately they are not.

Firstly, the first records of hidden SMBus, in September 2000, predate
ACPI. All the early boards where the SMBus was hidden did not have ACPI
code poking at it at all, so this is definitely not the reason why it
was removed. The Asus P4 series is a good example of that. Unhiding the
SMBus on these boards actually let the user take benefit of the
hardware they had paid for.

Secondly, only a few south bridges are capable of hiding the SMBus. And
an even smaller number of systems have their SMBus actually hidden by
the BIOS. A much large number of systems, especially recent ones, have
ACPI poking at the hardware monitoring chip (again, SMBus-based or
not.) So there is no correlation between hidden SMBus and ACPI poking
at the hardware monitoring chip. So, even though we are very careful
these days when adding a quirk to unhide the SMBus on a new system (we
make sure that ACPI isn't accessing it), this is not going to solve the
general problem.

I would be happy to prevent SMBus and/or hardware monitoring drivers
from being loaded on ACPI-based system if we had a way to know which
systems do have ACPI code accessing these chips and which do not, and if
ACPI was offering a level of functionality comparable to what
individual hardware monitoring drivers offer. Unfortunately:

1* As far as I know, we currently have no way to know if the ACPI code
plans to ever access the hardware monitoring chip. If the acpi
subsystem could export this information somehow, it would help a lot.
But I'm not familiar with ACPI, so I don't know if this is feasable or
not. We just can't prevent the SMBus and hardware monitoring drivers
drivers from being loaded as soon as ACPI is enabled. This would
prevent a majority of users from using them, while they work fine for
most of them.

2* The hardware monitoring features offered by ACPI today are one level
of magnitude weaker than what lm_sensors was already offering back in
1999. The monitoring chips can do much but unfortunately ACPI only
exposes a very small subset of the chip features. ACPI doesn't handle
voltage monitoring at all. It usually reports no more than one fan, and
in my experience, more often than not, the speed is reported as a
boolean (spinning or not), when lm_sensors gives you the exact speed of
all your fans. ACPI thermal zones are not so bad, but the interface to
control them is ugly, and lm_sensors usually gives more details. And
more importantly, on almost all desktop and server motherboards, ACPI
does _not_ provide any hardware monitoring feature. So we really can't
go tell our users: give up the individual hardware monitoring drivers
and lm_sensors on all ACPI-enabled systems and use whatever ACPI offers
instead. This would be a major functionality loss for a vast majority
of users.

What do you propose?

[1] http://www.microsoft.com/whdc/system/pnppwr/powermgmt/BIOSAML.mspx

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Pavel,

On Wed, 28 Feb 2007 21:38:04 +0000, Pavel Machek wrote:
> > > Well I had an idea after looking at k8temp -- why not make it default to
> > > doing only reads from the sensor?  You'd only get information from whatever
> > > core/sensor combination that ACPI had last used, but it would be safe.
> > 
> > ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
> > doesn't conflict with only k8temp, but with virtually all hardware
> > monitoring drivers, all I2C/SMBus drivers, and probably other types of
> > drivers too. We just can't restrict or blacklist all these drivers
> > because ACPI misbehaves.
> 
> Oops, sorry about that but no, that will not work.
> 
> There's piece of paper, called ACPI specification, and we are
> following it.

I never suggested otherwise. But the Linux 2.6 device driver model is
based, in part, on the fact that each driver must request the
resources it needs before actually using them. The acpi subsystem fails
to do that, so it is, in that sense, misbehaving. The is the root cause
of the problems people are reporting these days.

> Bug is not in our implementation.
> 
> Bug is in the ACPI specs... it does not explicitely allow you to go
> out and bitbang i2c, and you do it, and you get problems.

It really doesn't have anything to do with bit-banged I2C. It's about
SMBus master chips (most often embedded in south bridges), hardware
monitoring chips (be them SMBus-based or not), or basically any driver
which uses a resource ACPI is using as well (virtually any driver
accessing a Super-IO/LPC device is affected, for example.)

Secondly, your logic is somewhat broken. Just because one specification
doesn't explicitely allow me to do something, I must stop doing it?
There must be two dozen different specifications modern PC hardware
supposedly conforms to. If we stop doing everything which isn't
explicitely allowed by all of them, I fear there won't be much left ;)
A better question would be IMHO: Does the ACPI specification
explicitely _prohibits_ accessing directly to some categories of
hardware?

Lastly, I wholeheartedly agree that the core problem appears to be in
the ACPI design rather than in the Linux implementation. As a matter of
fact, other operating systems are facing the same problem [1]. The
problem isn't less real, though.

> Now, you may try to change specs to be hwmon-friendly... good luck.

I would like them to be driver-model-friendly, that's even a broader
challenge ;)

> But currently hw manufacturers follow ACPI specs, so we have to follow
> it, too; bad luck for hwmon. BIOS hiding smbus from you is good hint
> you are doing something wrong...?

I would love things to be that easy, but unfortunately they are not.

Firstly, the first records of hidden SMBus, in September 2000, predate
ACPI. All the early boards where the SMBus was hidden did not have ACPI
code poking at it at all, so this is definitely not the reason why it
was removed. The Asus P4 series is a good example of that. Unhiding the
SMBus on these boards actually let the user take benefit of the
hardware they had paid for.

Secondly, only a few south bridges are capable of hiding the SMBus. And
an even smaller number of systems have their SMBus actually hidden by
the BIOS. A much large number of systems, especially recent ones, have
ACPI poking at the hardware monitoring chip (again, SMBus-based or
not.) So there is no correlation between hidden SMBus and ACPI poking
at the hardware monitoring chip. So, even though we are very careful
these days when adding a quirk to unhide the SMBus on a new system (we
make sure that ACPI isn't accessing it), this is not going to solve the
general problem.

I would be happy to prevent SMBus and/or hardware monitoring drivers
from being loaded on ACPI-based system if we had a way to know which
systems do have ACPI code accessing these chips and which do not, and if
ACPI was offering a level of functionality comparable to what
individual hardware monitoring drivers offer. Unfortunately:

1* As far as I know, we currently have no way to know if the ACPI code
plans to ever access the hardware monitoring chip. If the acpi
subsystem could export this information somehow, it would help a lot.
But I'm not familiar with ACPI, so I don't know if this is feasable or
not. We just can't prevent the SMBus and hardware monitoring drivers
drivers from being loaded as soon as ACPI is enabled. This would
prevent a majority of users from using them, while they work fine for
most of them.

2* The hardware monitoring features offered by ACPI today are one level
of magnitude weaker than what lm_sensors was already offering back in
1999. The monitoring chips can do much but unfortunately ACPI only
exposes a very small subset of the chip features. ACPI doesn't handle
voltage monitoring at all. It usually reports no more than one fan, and
in my experience, more often than not, the speed is reported as a
boolean (spinning or not), when lm_sensors gives you the exact speed of
all your fans. ACPI thermal zones are not so bad, but the interface to
control them is ugly, and lm_sensors usually gives more details. And
more importantly, on almost all desktop and server motherboards, ACPI
does _not_ provide any hardware monitoring feature. So we really can't
go tell our users: give up the individual hardware monitoring drivers
and lm_sensors on all ACPI-enabled systems and use whatever ACPI offers
instead. This would be a major functionality loss for a vast majority
of users.

What do you propose?

[1] http://www.microsoft.com/whdc/system/pnppwr/powermgmt/BIOSAML.mspx

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Dave Jones]

On Thu, Mar 01, 2007 at 03:26:55PM +0100, Jean Delvare wrote:

 > Firstly, the first records of hidden SMBus, in September 2000, predate
 > ACPI.

The earliest ACPI spec I have handy is 1.0b, which came out in Feb 2 1999
so this isn't true. The all knowing (and always accurate :) wikipedia
claims it was first released in 1996, though I believe that all the pre 1.0b
machines were using acpi implementations before the standard was finalised.

I certainly remember seeing ACPI capable machines circa 1997/1998.

		Dave

-- 
http://www.codemonkey.org.uk

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Dave Jones]

On Thu, Mar 01, 2007 at 03:26:55PM +0100, Jean Delvare wrote:

 > Firstly, the first records of hidden SMBus, in September 2000, predate
 > ACPI.

The earliest ACPI spec I have handy is 1.0b, which came out in Feb 2 1999
so this isn't true. The all knowing (and always accurate :) wikipedia
claims it was first released in 1996, though I believe that all the pre 1.0b
machines were using acpi implementations before the standard was finalised.

I certainly remember seeing ACPI capable machines circa 1997/1998.

		Dave

-- 
http://www.codemonkey.org.uk

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Thu, 1 Mar 2007 12:48:06 -0500, Dave Jones wrote:
> On Thu, Mar 01, 2007 at 03:26:55PM +0100, Jean Delvare wrote:
> 
>  > Firstly, the first records of hidden SMBus, in September 2000, predate
>  > ACPI.
> 
> The earliest ACPI spec I have handy is 1.0b, which came out in Feb 2 1999
> so this isn't true. The all knowing (and always accurate :) wikipedia
> claims it was first released in 1996, though I believe that all the pre 1.0b
> machines were using acpi implementations before the standard was finalised.
> 
> I certainly remember seeing ACPI capable machines circa 1997/1998.

Yeah, and these early ACPI implementations were so good that we have an
option to blacklist them.
arch/i386/defconfig:CONFIG_ACPI_BLACKLIST_YEAR=2001

My point (which you didn't quote) was that there is no correlation
between the SMBus being hidden and ACPI accessing the hardware
monitoring chip, contrary to what Pavel was suggesting.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Thu, 1 Mar 2007 12:48:06 -0500, Dave Jones wrote:
> On Thu, Mar 01, 2007 at 03:26:55PM +0100, Jean Delvare wrote:
> 
>  > Firstly, the first records of hidden SMBus, in September 2000, predate
>  > ACPI.
> 
> The earliest ACPI spec I have handy is 1.0b, which came out in Feb 2 1999
> so this isn't true. The all knowing (and always accurate :) wikipedia
> claims it was first released in 1996, though I believe that all the pre 1.0b
> machines were using acpi implementations before the standard was finalised.
> 
> I certainly remember seeing ACPI capable machines circa 1997/1998.

Yeah, and these early ACPI implementations were so good that we have an
option to blacklist them.
arch/i386/defconfig:CONFIG_ACPI_BLACKLIST_YEAR 01

My point (which you didn't quote) was that there is no correlation
between the SMBus being hidden and ACPI accessing the hardware
monitoring chip, contrary to what Pavel was suggesting.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> >  > Firstly, the first records of hidden SMBus, in September 2000, predate
> >  > ACPI.
> > 
> > The earliest ACPI spec I have handy is 1.0b, which came out in Feb 2 1999
> > so this isn't true. The all knowing (and always accurate :) wikipedia
> > claims it was first released in 1996, though I believe that all the pre 1.0b
> > machines were using acpi implementations before the standard was finalised.
> > 
> > I certainly remember seeing ACPI capable machines circa 1997/1998.
> 
> Yeah, and these early ACPI implementations were so good that we have an
> option to blacklist them.
> arch/i386/defconfig:CONFIG_ACPI_BLACKLIST_YEAR=2001
> 
> My point (which you didn't quote) was that there is no correlation
> between the SMBus being hidden and ACPI accessing the hardware
> monitoring chip, contrary to what Pavel was suggesting.

It may not be correlated with ACPI, but BIOS authors clearly want to
keep you away from their SMBus controllers....
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> >  > Firstly, the first records of hidden SMBus, in September 2000, predate
> >  > ACPI.
> > 
> > The earliest ACPI spec I have handy is 1.0b, which came out in Feb 2 1999
> > so this isn't true. The all knowing (and always accurate :) wikipedia
> > claims it was first released in 1996, though I believe that all the pre 1.0b
> > machines were using acpi implementations before the standard was finalised.
> > 
> > I certainly remember seeing ACPI capable machines circa 1997/1998.
> 
> Yeah, and these early ACPI implementations were so good that we have an
> option to blacklist them.
> arch/i386/defconfig:CONFIG_ACPI_BLACKLIST_YEAR 01
> 
> My point (which you didn't quote) was that there is no correlation
> between the SMBus being hidden and ACPI accessing the hardware
> monitoring chip, contrary to what Pavel was suggesting.

It may not be correlated with ACPI, but BIOS authors clearly want to
keep you away from their SMBus controllers....
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > > > Well I had an idea after looking at k8temp -- why not make it default to
> > > > doing only reads from the sensor?  You'd only get information from whatever
> > > > core/sensor combination that ACPI had last used, but it would be safe.
> > > 
> > > ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
> > > doesn't conflict with only k8temp, but with virtually all hardware
> > > monitoring drivers, all I2C/SMBus drivers, and probably other types of
> > > drivers too. We just can't restrict or blacklist all these drivers
> > > because ACPI misbehaves.
> > 
> > Oops, sorry about that but no, that will not work.
> > 
> > There's piece of paper, called ACPI specification, and we are
> > following it.
> 
> I never suggested otherwise. But the Linux 2.6 device driver model is
> based, in part, on the fact that each driver must request the
> resources it needs before actually using them. The acpi subsystem fails
> to do that, so it is, in that sense, misbehaving. The is the root cause
> of the problems people are reporting these days.

Ok. You are right that ACPI is an ugly piece of mess. But I'm pretty
sure that 90%+ of ACPI notebook implementations *will* want to talk to
their monitoring chips... for temperature readings.

So even if we fixed ACPI to reserve the ports, you'd be still unhappy;
lm-sensors would break at least on all the notebooks.

> > Now, you may try to change specs to be hwmon-friendly... good luck.
> 
> I would like them to be driver-model-friendly, that's even a broader
> challenge ;)

:-).

> > But currently hw manufacturers follow ACPI specs, so we have to follow
> > it, too; bad luck for hwmon. BIOS hiding smbus from you is good hint
> > you are doing something wrong...?
> 
> I would love things to be that easy, but unfortunately they are not.
> 
> Firstly, the first records of hidden SMBus, in September 2000, predate
> ACPI. All the early boards where the SMBus was hidden did not have ACPI
> code poking at it at all, so this is definitely not the reason why it
> was removed. The Asus P4 series is a good example of that. Unhiding the
> SMBus on these boards actually let the user take benefit of the
> hardware they had paid for.

...against wishes of the manufacturers. Which sometimes know what they
are doing. (Sometimes not :-).

> I would be happy to prevent SMBus and/or hardware monitoring drivers
> from being loaded on ACPI-based system if we had a way to know which
> systems do have ACPI code accessing these chips and which do not, and if
> ACPI was offering a level of functionality comparable to what
> individual hardware monitoring drivers offer. Unfortunately:

Well, I'm afraid you should assume all recent notebooks touch sensors
from ACPI.

> 1* As far as I know, we currently have no way to know if the ACPI code
> plans to ever access the hardware monitoring chip. If the acpi
> subsystem could export this information somehow, it would help a lot.
> But I'm not familiar with ACPI, so I don't know if this is feasable or
> not. We just can't prevent the SMBus and hardware monitoring drivers
> drivers from being loaded as soon as ACPI is enabled. This would
> prevent a majority of users from using them, while they work fine for
> most of them.

What about whitelist? DMI-based? That's only way to do it, I'm afraid.

> 2* The hardware monitoring features offered by ACPI today are one level
> of magnitude weaker than what lm_sensors was already offering back in
> 1999. The monitoring chips can do much but unfortunately ACPI only
> exposes a very small subset of the chip features. ACPI doesn't
> handle

Yes, I know ACPI sucks at hardware monitoring. Unfortunately, we can't
go without ACPI.

> voltage monitoring at all. It usually reports no more than one fan, and
> in my experience, more often than not, the speed is reported as a
> boolean (spinning or not), when lm_sensors gives you the exact speed of
> all your fans. ACPI thermal zones are not so bad, but the interface to
> control them is ugly, and lm_sensors usually gives more details. And

Fix the interface? ;-). Actually that move may be underway as we are
moving out of /proc.

> What do you propose?

Whitelist seems like a way to go :(.
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > > > Well I had an idea after looking at k8temp -- why not make it default to
> > > > doing only reads from the sensor?  You'd only get information from whatever
> > > > core/sensor combination that ACPI had last used, but it would be safe.
> > > 
> > > ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
> > > doesn't conflict with only k8temp, but with virtually all hardware
> > > monitoring drivers, all I2C/SMBus drivers, and probably other types of
> > > drivers too. We just can't restrict or blacklist all these drivers
> > > because ACPI misbehaves.
> > 
> > Oops, sorry about that but no, that will not work.
> > 
> > There's piece of paper, called ACPI specification, and we are
> > following it.
> 
> I never suggested otherwise. But the Linux 2.6 device driver model is
> based, in part, on the fact that each driver must request the
> resources it needs before actually using them. The acpi subsystem fails
> to do that, so it is, in that sense, misbehaving. The is the root cause
> of the problems people are reporting these days.

Ok. You are right that ACPI is an ugly piece of mess. But I'm pretty
sure that 90%+ of ACPI notebook implementations *will* want to talk to
their monitoring chips... for temperature readings.

So even if we fixed ACPI to reserve the ports, you'd be still unhappy;
lm-sensors would break at least on all the notebooks.

> > Now, you may try to change specs to be hwmon-friendly... good luck.
> 
> I would like them to be driver-model-friendly, that's even a broader
> challenge ;)

:-).

> > But currently hw manufacturers follow ACPI specs, so we have to follow
> > it, too; bad luck for hwmon. BIOS hiding smbus from you is good hint
> > you are doing something wrong...?
> 
> I would love things to be that easy, but unfortunately they are not.
> 
> Firstly, the first records of hidden SMBus, in September 2000, predate
> ACPI. All the early boards where the SMBus was hidden did not have ACPI
> code poking at it at all, so this is definitely not the reason why it
> was removed. The Asus P4 series is a good example of that. Unhiding the
> SMBus on these boards actually let the user take benefit of the
> hardware they had paid for.

...against wishes of the manufacturers. Which sometimes know what they
are doing. (Sometimes not :-).

> I would be happy to prevent SMBus and/or hardware monitoring drivers
> from being loaded on ACPI-based system if we had a way to know which
> systems do have ACPI code accessing these chips and which do not, and if
> ACPI was offering a level of functionality comparable to what
> individual hardware monitoring drivers offer. Unfortunately:

Well, I'm afraid you should assume all recent notebooks touch sensors
from ACPI.

> 1* As far as I know, we currently have no way to know if the ACPI code
> plans to ever access the hardware monitoring chip. If the acpi
> subsystem could export this information somehow, it would help a lot.
> But I'm not familiar with ACPI, so I don't know if this is feasable or
> not. We just can't prevent the SMBus and hardware monitoring drivers
> drivers from being loaded as soon as ACPI is enabled. This would
> prevent a majority of users from using them, while they work fine for
> most of them.

What about whitelist? DMI-based? That's only way to do it, I'm afraid.

> 2* The hardware monitoring features offered by ACPI today are one level
> of magnitude weaker than what lm_sensors was already offering back in
> 1999. The monitoring chips can do much but unfortunately ACPI only
> exposes a very small subset of the chip features. ACPI doesn't
> handle

Yes, I know ACPI sucks at hardware monitoring. Unfortunately, we can't
go without ACPI.

> voltage monitoring at all. It usually reports no more than one fan, and
> in my experience, more often than not, the speed is reported as a
> boolean (spinning or not), when lm_sensors gives you the exact speed of
> all your fans. ACPI thermal zones are not so bad, but the interface to
> control them is ugly, and lm_sensors usually gives more details. And

Fix the interface? ;-). Actually that move may be underway as we are
moving out of /proc.

> What do you propose?

Whitelist seems like a way to go :(.
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Matthew Garrett]

On Fri, Mar 02, 2007 at 12:40:23PM +0100, Pavel Machek wrote:

> Whitelist seems like a way to go :(.

The DSDT code clearly can't touch the hardware itself - hardware access 
is carried out by the kernel. If we can identify cases where ACPI reads 
and writes would touch resources claimed by other drivers, that would be 
a good starting point for working out what's going on.

Of course, this ignores the case where the DSDT just traps into SMM 
code. That one is clearly unsolvable.
-- 
Matthew Garrett | mjg59@srcf.ucam.org

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Matthew Garrett]

On Fri, Mar 02, 2007 at 12:40:23PM +0100, Pavel Machek wrote:

> Whitelist seems like a way to go :(.

The DSDT code clearly can't touch the hardware itself - hardware access 
is carried out by the kernel. If we can identify cases where ACPI reads 
and writes would touch resources claimed by other drivers, that would be 
a good starting point for working out what's going on.

Of course, this ignores the case where the DSDT just traps into SMM 
code. That one is clearly unsolvable.
-- 
Matthew Garrett | mjg59 at srcf.ucam.org

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Fri, 2 Mar 2007 12:31:58 +0100, Pavel Machek wrote:
> > My point (which you didn't quote) was that there is no correlation
> > between the SMBus being hidden and ACPI accessing the hardware
> > monitoring chip, contrary to what Pavel was suggesting.
> 
> It may not be correlated with ACPI, but BIOS authors clearly want to
> keep you away from their SMBus controllers....

drivers/pci/quirks.c is full of things we do against the BIOS authors
intent. You don't plan to remove them all, do you?

(And as a side note, this is really the board's owner SMBus controller.
The hardware doesn't belong to the BIOS author.)

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Fri, 2 Mar 2007 12:31:58 +0100, Pavel Machek wrote:
> > My point (which you didn't quote) was that there is no correlation
> > between the SMBus being hidden and ACPI accessing the hardware
> > monitoring chip, contrary to what Pavel was suggesting.
> 
> It may not be correlated with ACPI, but BIOS authors clearly want to
> keep you away from their SMBus controllers....

drivers/pci/quirks.c is full of things we do against the BIOS authors
intent. You don't plan to remove them all, do you?

(And as a side note, this is really the board's owner SMBus controller.
The hardware doesn't belong to the BIOS author.)

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Henrique de Moraes Holschuh]

On Fri, 02 Mar 2007, Jean Delvare wrote:
> (And as a side note, this is really the board's owner SMBus controller.
> The hardware doesn't belong to the BIOS author.)

Especially when they won't do their job right, and won't fix it later
either...

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Henrique de Moraes Holschuh]

On Fri, 02 Mar 2007, Jean Delvare wrote:
> (And as a side note, this is really the board's owner SMBus controller.
> The hardware doesn't belong to the BIOS author.)

Especially when they won't do their job right, and won't fix it later
either...

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

On Fri 2007-03-02 14:37:20, Jean Delvare wrote:
> On Fri, 2 Mar 2007 12:31:58 +0100, Pavel Machek wrote:
> > > My point (which you didn't quote) was that there is no correlation
> > > between the SMBus being hidden and ACPI accessing the hardware
> > > monitoring chip, contrary to what Pavel was suggesting.
> > 
> > It may not be correlated with ACPI, but BIOS authors clearly want to
> > keep you away from their SMBus controllers....
> 
> drivers/pci/quirks.c is full of things we do against the BIOS authors
> intent. You don't plan to remove them all, do you?

Notice how quirks.c is careful to name machines where given quirk is
used.

If you do whitelist "it is okay to do sensors accesses on this board",
that is okay with me. But having quirk "on all future Intel chipsets,
do foo" would be stupid. Enabling sensors blindly is pretty similar.

> (And as a side note, this is really the board's owner SMBus controller.
> The hardware doesn't belong to the BIOS author.)

True.
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

On Fri 2007-03-02 14:37:20, Jean Delvare wrote:
> On Fri, 2 Mar 2007 12:31:58 +0100, Pavel Machek wrote:
> > > My point (which you didn't quote) was that there is no correlation
> > > between the SMBus being hidden and ACPI accessing the hardware
> > > monitoring chip, contrary to what Pavel was suggesting.
> > 
> > It may not be correlated with ACPI, but BIOS authors clearly want to
> > keep you away from their SMBus controllers....
> 
> drivers/pci/quirks.c is full of things we do against the BIOS authors
> intent. You don't plan to remove them all, do you?

Notice how quirks.c is careful to name machines where given quirk is
used.

If you do whitelist "it is okay to do sensors accesses on this board",
that is okay with me. But having quirk "on all future Intel chipsets,
do foo" would be stupid. Enabling sensors blindly is pretty similar.

> (And as a side note, this is really the board's owner SMBus controller.
> The hardware doesn't belong to the BIOS author.)

True.
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

On Fri 2007-03-02 11:47:47, Matthew Garrett wrote:
> On Fri, Mar 02, 2007 at 12:40:23PM +0100, Pavel Machek wrote:
> 
> > Whitelist seems like a way to go :(.
> 
> The DSDT code clearly can't touch the hardware itself - hardware access 
> is carried out by the kernel. If we can identify cases where ACPI reads 
> and writes would touch resources claimed by other drivers, that would be 
> a good starting point for working out what's going on.
> 
> Of course, this ignores the case where the DSDT just traps into SMM 
> code. That one is clearly unsolvable.

We can't solve SMM stuff. (Whitelist needed :-).

Actually for the acpi stuff... we might wrap ACPI interpretter with a
semaphore that needs to be taken before starting any AML code. Then
just make sure sensors take same semaphore?
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

On Fri 2007-03-02 11:47:47, Matthew Garrett wrote:
> On Fri, Mar 02, 2007 at 12:40:23PM +0100, Pavel Machek wrote:
> 
> > Whitelist seems like a way to go :(.
> 
> The DSDT code clearly can't touch the hardware itself - hardware access 
> is carried out by the kernel. If we can identify cases where ACPI reads 
> and writes would touch resources claimed by other drivers, that would be 
> a good starting point for working out what's going on.
> 
> Of course, this ignores the case where the DSDT just traps into SMM 
> code. That one is clearly unsolvable.

We can't solve SMM stuff. (Whitelist needed :-).

Actually for the acpi stuff... we might wrap ACPI interpretter with a
semaphore that needs to be taken before starting any AML code. Then
just make sure sensors take same semaphore?
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Pavel,

On Fri, 2 Mar 2007 12:40:23 +0100, Pavel Machek wrote:
> Ok. You are right that ACPI is an ugly piece of mess. But I'm pretty
> sure that 90%+ of ACPI notebook implementations *will* want to talk to
> their monitoring chips... for temperature readings.
>
> So even if we fixed ACPI to reserve the ports, you'd be still unhappy;
> lm-sensors would break at least on all the notebooks.

That's a secondary problem. The primary issue is the concurrent access
to resources, which cause lots of trouble which are hard to investigate.
If ACPI reserves the ports, then the SMBus or hardware monitoring
drivers (or any other conflicting driver) will cleanly fail to load,
which would be a move in the right direction. Ideally we would be able
to synchronize the accesses between ACPI and the other drivers, but if
we can't, I'd already be _very happy_ to just prevent conflicting
drivers from being loaded at the same time.

So, can ACPI actually reserve the ports it accesses?

> > I would be happy to prevent SMBus and/or hardware monitoring drivers
> > from being loaded on ACPI-based system if we had a way to know which
> > systems do have ACPI code accessing these chips and which do not, and if
> > ACPI was offering a level of functionality comparable to what
> > individual hardware monitoring drivers offer. Unfortunately:
> 
> Well, I'm afraid you should assume all recent notebooks touch sensors
> from ACPI.

Correct, and on most notebooks, traditional hardware monitoring
drivers do not work anyway. Or I should say, used to not work. The
recent CPUs have embedded sensors which can be read using Rudolf
Marek's k8temp and coretemp drivers, this works on laptops as well.
Chances are good that future laptops will not include a separate
temperature sensor but will read the temperature from the CPU directly,
which will cause conflicts with our drivers.

Now the problem is that we can't blacklist SMBus and hardware
monitoring drivers on all laptops by default. There may be other
devices on the SMBus which the user can legitimately want to access,
such as EEPROMs or RTCs.

> > 1* As far as I know, we currently have no way to know if the ACPI code
> > plans to ever access the hardware monitoring chip. If the acpi
> > subsystem could export this information somehow, it would help a lot.
> > But I'm not familiar with ACPI, so I don't know if this is feasable or
> > not. We just can't prevent the SMBus and hardware monitoring drivers
> > drivers from being loaded as soon as ACPI is enabled. This would
> > prevent a majority of users from using them, while they work fine for
> > most of them.
> 
> What about whitelist? DMI-based? That's only way to do it, I'm afraid.

What kind of whitelist do you have in mind? We can't realistically
maintain an ever-growing whitelist of hundreds of entries in the
kernel. We could block all laptops by default and maintain a white list
only for them, and a black list for other systems, the would probably
limit the maintenance work, maybe not to an acceptable level though.

Anyway I would definitely prefer the resource conflicts approach, as it
wouldn't need maintenance at all.

> > 2* The hardware monitoring features offered by ACPI today are one level
> > of magnitude weaker than what lm_sensors was already offering back in
> > 1999. The monitoring chips can do much but unfortunately ACPI only
> > exposes a very small subset of the chip features. ACPI doesn't
> > handle
> 
> Yes, I know ACPI sucks at hardware monitoring. Unfortunately, we can't
> go without ACPI.

Agreed.

> > voltage monitoring at all. It usually reports no more than one fan, and
> > in my experience, more often than not, the speed is reported as a
> > boolean (spinning or not), when lm_sensors gives you the exact speed of
> > all your fans. ACPI thermal zones are not so bad, but the interface to
> > control them is ugly, and lm_sensors usually gives more details. And
> 
> Fix the interface? ;-). Actually that move may be underway as we are
> moving out of /proc.

Great, looking forward.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Pavel,

On Fri, 2 Mar 2007 12:40:23 +0100, Pavel Machek wrote:
> Ok. You are right that ACPI is an ugly piece of mess. But I'm pretty
> sure that 90%+ of ACPI notebook implementations *will* want to talk to
> their monitoring chips... for temperature readings.
>
> So even if we fixed ACPI to reserve the ports, you'd be still unhappy;
> lm-sensors would break at least on all the notebooks.

That's a secondary problem. The primary issue is the concurrent access
to resources, which cause lots of trouble which are hard to investigate.
If ACPI reserves the ports, then the SMBus or hardware monitoring
drivers (or any other conflicting driver) will cleanly fail to load,
which would be a move in the right direction. Ideally we would be able
to synchronize the accesses between ACPI and the other drivers, but if
we can't, I'd already be _very happy_ to just prevent conflicting
drivers from being loaded at the same time.

So, can ACPI actually reserve the ports it accesses?

> > I would be happy to prevent SMBus and/or hardware monitoring drivers
> > from being loaded on ACPI-based system if we had a way to know which
> > systems do have ACPI code accessing these chips and which do not, and if
> > ACPI was offering a level of functionality comparable to what
> > individual hardware monitoring drivers offer. Unfortunately:
> 
> Well, I'm afraid you should assume all recent notebooks touch sensors
> from ACPI.

Correct, and on most notebooks, traditional hardware monitoring
drivers do not work anyway. Or I should say, used to not work. The
recent CPUs have embedded sensors which can be read using Rudolf
Marek's k8temp and coretemp drivers, this works on laptops as well.
Chances are good that future laptops will not include a separate
temperature sensor but will read the temperature from the CPU directly,
which will cause conflicts with our drivers.

Now the problem is that we can't blacklist SMBus and hardware
monitoring drivers on all laptops by default. There may be other
devices on the SMBus which the user can legitimately want to access,
such as EEPROMs or RTCs.

> > 1* As far as I know, we currently have no way to know if the ACPI code
> > plans to ever access the hardware monitoring chip. If the acpi
> > subsystem could export this information somehow, it would help a lot.
> > But I'm not familiar with ACPI, so I don't know if this is feasable or
> > not. We just can't prevent the SMBus and hardware monitoring drivers
> > drivers from being loaded as soon as ACPI is enabled. This would
> > prevent a majority of users from using them, while they work fine for
> > most of them.
> 
> What about whitelist? DMI-based? That's only way to do it, I'm afraid.

What kind of whitelist do you have in mind? We can't realistically
maintain an ever-growing whitelist of hundreds of entries in the
kernel. We could block all laptops by default and maintain a white list
only for them, and a black list for other systems, the would probably
limit the maintenance work, maybe not to an acceptable level though.

Anyway I would definitely prefer the resource conflicts approach, as it
wouldn't need maintenance at all.

> > 2* The hardware monitoring features offered by ACPI today are one level
> > of magnitude weaker than what lm_sensors was already offering back in
> > 1999. The monitoring chips can do much but unfortunately ACPI only
> > exposes a very small subset of the chip features. ACPI doesn't
> > handle
> 
> Yes, I know ACPI sucks at hardware monitoring. Unfortunately, we can't
> go without ACPI.

Agreed.

> > voltage monitoring at all. It usually reports no more than one fan, and
> > in my experience, more often than not, the speed is reported as a
> > boolean (spinning or not), when lm_sensors gives you the exact speed of
> > all your fans. ACPI thermal zones are not so bad, but the interface to
> > control them is ugly, and lm_sensors usually gives more details. And
> 
> Fix the interface? ;-). Actually that move may be underway as we are
> moving out of /proc.

Great, looking forward.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Matthew,

On Fri, 2 Mar 2007 11:47:47 +0000, Matthew Garrett wrote:
> The DSDT code clearly can't touch the hardware itself - hardware access 
> is carried out by the kernel. If we can identify cases where ACPI reads 
> and writes would touch resources claimed by other drivers, that would be 
> a good starting point for working out what's going on.

I'm not familiar with APCI at all so I didn't know, but what you write
here brings some hope. Would it be possible to parse all the DSDT code
at boot time and deduce all the ports which ACPI would need to request
to be safe? Or do we have to wait for the accesses to actually happen?

> Of course, this ignores the case where the DSDT just traps into SMM 
> code. That one is clearly unsolvable.

Yeah, SMM is an even more complex problem :(

Do we know in advance when we are going to SMM mode and back? If we do,
I'd be happy with a mutex every interested driver could use to protect
relevant parts of its code. SMBus master drivers for example could
request that mutex during SMBus transactions. Of course we don't know
if SMM will actually touch the SMBus, but better safe than sorry I
guess. And SMM calls aren't happening so frequently, are they?

Thanks,
-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Matthew,

On Fri, 2 Mar 2007 11:47:47 +0000, Matthew Garrett wrote:
> The DSDT code clearly can't touch the hardware itself - hardware access 
> is carried out by the kernel. If we can identify cases where ACPI reads 
> and writes would touch resources claimed by other drivers, that would be 
> a good starting point for working out what's going on.

I'm not familiar with APCI at all so I didn't know, but what you write
here brings some hope. Would it be possible to parse all the DSDT code
at boot time and deduce all the ports which ACPI would need to request
to be safe? Or do we have to wait for the accesses to actually happen?

> Of course, this ignores the case where the DSDT just traps into SMM 
> code. That one is clearly unsolvable.

Yeah, SMM is an even more complex problem :(

Do we know in advance when we are going to SMM mode and back? If we do,
I'd be happy with a mutex every interested driver could use to protect
relevant parts of its code. SMBus master drivers for example could
request that mutex during SMBus transactions. Of course we don't know
if SMM will actually touch the SMBus, but better safe than sorry I
guess. And SMM calls aren't happening so frequently, are they?

Thanks,
-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Matthew Garrett]

On Fri, Mar 02, 2007 at 03:10:55PM +0100, Jean Delvare wrote:

> I'm not familiar with APCI at all so I didn't know, but what you write
> here brings some hope. Would it be possible to parse all the DSDT code
> at boot time and deduce all the ports which ACPI would need to request
> to be safe? Or do we have to wait for the accesses to actually happen?

In theory I /think/ so, but it would probably end up being an 
overestimate of the coverage actually needed. Trapping at runtime is 
arguably more elegant?

> Do we know in advance when we are going to SMM mode and back? If we do,
> I'd be happy with a mutex every interested driver could use to protect
> relevant parts of its code. SMBus master drivers for example could
> request that mutex during SMBus transactions. Of course we don't know
> if SMM will actually touch the SMBus, but better safe than sorry I
> guess. And SMM calls aren't happening so frequently, are they?

My understanding is that pretty much arbitrary hardware access can cause 
SMM transitions without OS notification, though this is getting outside 
the areas I know about.
-- 
Matthew Garrett | mjg59@srcf.ucam.org

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Matthew Garrett]

On Fri, Mar 02, 2007 at 03:10:55PM +0100, Jean Delvare wrote:

> I'm not familiar with APCI at all so I didn't know, but what you write
> here brings some hope. Would it be possible to parse all the DSDT code
> at boot time and deduce all the ports which ACPI would need to request
> to be safe? Or do we have to wait for the accesses to actually happen?

In theory I /think/ so, but it would probably end up being an 
overestimate of the coverage actually needed. Trapping at runtime is 
arguably more elegant?

> Do we know in advance when we are going to SMM mode and back? If we do,
> I'd be happy with a mutex every interested driver could use to protect
> relevant parts of its code. SMBus master drivers for example could
> request that mutex during SMBus transactions. Of course we don't know
> if SMM will actually touch the SMBus, but better safe than sorry I
> guess. And SMM calls aren't happening so frequently, are they?

My understanding is that pretty much arbitrary hardware access can cause 
SMM transitions without OS notification, though this is getting outside 
the areas I know about.
-- 
Matthew Garrett | mjg59 at srcf.ucam.org

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > The DSDT code clearly can't touch the hardware itself - hardware access 
> > is carried out by the kernel. If we can identify cases where ACPI reads 
> > and writes would touch resources claimed by other drivers, that would be 
> > a good starting point for working out what's going on.
> 
> I'm not familiar with APCI at all so I didn't know, but what you write
> here brings some hope. Would it be possible to parse all the DSDT code
> at boot time and deduce all the ports which ACPI would need to request
> to be safe?

I'm afraid it is about as hard as disassembling SMM BIOS to figure out
what it accesses.

> > Of course, this ignores the case where the DSDT just traps into SMM 
> > code. That one is clearly unsolvable.
> 
> Yeah, SMM is an even more complex problem :(
> 
> Do we know in advance when we are going to SMM mode and back? If we

SMM is often invoked by timer, AC unplug, etc.
								Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > The DSDT code clearly can't touch the hardware itself - hardware access 
> > is carried out by the kernel. If we can identify cases where ACPI reads 
> > and writes would touch resources claimed by other drivers, that would be 
> > a good starting point for working out what's going on.
> 
> I'm not familiar with APCI at all so I didn't know, but what you write
> here brings some hope. Would it be possible to parse all the DSDT code
> at boot time and deduce all the ports which ACPI would need to request
> to be safe?

I'm afraid it is about as hard as disassembling SMM BIOS to figure out
what it accesses.

> > Of course, this ignores the case where the DSDT just traps into SMM 
> > code. That one is clearly unsolvable.
> 
> Yeah, SMM is an even more complex problem :(
> 
> Do we know in advance when we are going to SMM mode and back? If we

SMM is often invoked by timer, AC unplug, etc.
								Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > Ok. You are right that ACPI is an ugly piece of mess. But I'm pretty
> > sure that 90%+ of ACPI notebook implementations *will* want to talk to
> > their monitoring chips... for temperature readings.
> >
> > So even if we fixed ACPI to reserve the ports, you'd be still unhappy;
> > lm-sensors would break at least on all the notebooks.
> 
> That's a secondary problem. The primary issue is the concurrent access
> to resources, which cause lots of trouble which are hard to investigate.
> If ACPI reserves the ports, then the SMBus or hardware monitoring
> drivers (or any other conflicting driver) will cleanly fail to load,
> which would be a move in the right direction. Ideally we would be able
> to synchronize the accesses between ACPI and the other drivers, but if
> we can't, I'd already be _very happy_ to just prevent conflicting
> drivers from being loaded at the same time.
> 
> So, can ACPI actually reserve the ports it accesses?

No idea, talk to Len Brown (or start reading the code) :-(. I don't
think it will be easy.

> > > 1* As far as I know, we currently have no way to know if the ACPI code
> > > plans to ever access the hardware monitoring chip. If the acpi
> > > subsystem could export this information somehow, it would help a lot.
> > > But I'm not familiar with ACPI, so I don't know if this is feasable or
> > > not. We just can't prevent the SMBus and hardware monitoring drivers
> > > drivers from being loaded as soon as ACPI is enabled. This would
> > > prevent a majority of users from using them, while they work fine for
> > > most of them.
> > 
> > What about whitelist? DMI-based? That's only way to do it, I'm afraid.
> 
> What kind of whitelist do you have in mind? We can't realistically
> maintain an ever-growing whitelist of hundreds of entries in the
> kernel. We could block all laptops by default and maintain a white list
> only for them, and a black list for other systems, the would probably
> limit the maintenance work, maybe not to an acceptable level though.

I'm afraid something like that is way to go.

									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > Ok. You are right that ACPI is an ugly piece of mess. But I'm pretty
> > sure that 90%+ of ACPI notebook implementations *will* want to talk to
> > their monitoring chips... for temperature readings.
> >
> > So even if we fixed ACPI to reserve the ports, you'd be still unhappy;
> > lm-sensors would break at least on all the notebooks.
> 
> That's a secondary problem. The primary issue is the concurrent access
> to resources, which cause lots of trouble which are hard to investigate.
> If ACPI reserves the ports, then the SMBus or hardware monitoring
> drivers (or any other conflicting driver) will cleanly fail to load,
> which would be a move in the right direction. Ideally we would be able
> to synchronize the accesses between ACPI and the other drivers, but if
> we can't, I'd already be _very happy_ to just prevent conflicting
> drivers from being loaded at the same time.
> 
> So, can ACPI actually reserve the ports it accesses?

No idea, talk to Len Brown (or start reading the code) :-(. I don't
think it will be easy.

> > > 1* As far as I know, we currently have no way to know if the ACPI code
> > > plans to ever access the hardware monitoring chip. If the acpi
> > > subsystem could export this information somehow, it would help a lot.
> > > But I'm not familiar with ACPI, so I don't know if this is feasable or
> > > not. We just can't prevent the SMBus and hardware monitoring drivers
> > > drivers from being loaded as soon as ACPI is enabled. This would
> > > prevent a majority of users from using them, while they work fine for
> > > most of them.
> > 
> > What about whitelist? DMI-based? That's only way to do it, I'm afraid.
> 
> What kind of whitelist do you have in mind? We can't realistically
> maintain an ever-growing whitelist of hundreds of entries in the
> kernel. We could block all laptops by default and maintain a white list
> only for them, and a black list for other systems, the would probably
> limit the maintenance work, maybe not to an acceptable level though.

I'm afraid something like that is way to go.

									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Matthew Garrett]

How about this? It's informational only, but ought to result in 
complaints whenever ACPI tries to touch something that other hardware 
has reserved. We can't fail these accesses, but in theory we could 
consider some sort of locking layer that made it possible to interact 
anyway. I haven't even checked if this builds, but I think the concept 
is reasonable.

diff --git a/drivers/acpi/osl.c b/drivers/acpi/osl.c
index d870fb2..088abd1 100644
--- a/drivers/acpi/osl.c
+++ b/drivers/acpi/osl.c
@@ -35,6 +35,7 @@
 #include <linux/kmod.h>
 #include <linux/delay.h>
 #include <linux/workqueue.h>
+#include <linux/ioport.h>
 #include <linux/nmi.h>
 #include <acpi/acpi.h>
 #include <asm/io.h>
@@ -338,6 +339,10 @@ acpi_status acpi_os_read_port(acpi_io_address port, u32 * value, u32 width)
 {
 	u32 dummy;
 
+	if (__check_region(&ioport_resource, port, width/8))
+		printk (KERN_INFO "ACPI read from allocated ioport %lx\n", 
+			port);
+
 	if (!value)
 		value = &dummy;
 
@@ -362,6 +367,10 @@ EXPORT_SYMBOL(acpi_os_read_port);
 
 acpi_status acpi_os_write_port(acpi_io_address port, u32 value, u32 width)
 {
+	if (__check_region(&ioport_resource, port, width/8))
+		printk (KERN_INFO "ACPI write to allocated ioport %lx\n", 
+			port);
+
 	switch (width) {
 	case 8:
 		outb(value, port);
@@ -387,6 +396,10 @@ acpi_os_read_memory(acpi_physical_address phys_addr, u32 * value, u32 width)
 	u32 dummy;
 	void __iomem *virt_addr;
 
+	if (__check_region(&iomem_resource, phys_addr, width/8))
+		printk (KERN_INFO "ACPI read from allocated iomem %lx\n", 
+			phys_addr);
+
 	virt_addr = ioremap(phys_addr, width);
 	if (!value)
 		value = &dummy;
@@ -415,6 +428,10 @@ acpi_os_write_memory(acpi_physical_address phys_addr, u32 value, u32 width)
 {
 	void __iomem *virt_addr;
 
+	if (__check_region(&iomem_resource, phys_addr, width/8))
+		printk (KERN_INFO "ACPI write to allocated iomem %lx\n", 
+			phys_addr);
+
 	virt_addr = ioremap(phys_addr, width);
 
 	switch (width) {

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

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Matthew Garrett]

How about this? It's informational only, but ought to result in 
complaints whenever ACPI tries to touch something that other hardware 
has reserved. We can't fail these accesses, but in theory we could 
consider some sort of locking layer that made it possible to interact 
anyway. I haven't even checked if this builds, but I think the concept 
is reasonable.

diff --git a/drivers/acpi/osl.c b/drivers/acpi/osl.c
index d870fb2..088abd1 100644
--- a/drivers/acpi/osl.c
+++ b/drivers/acpi/osl.c
@@ -35,6 +35,7 @@
 #include <linux/kmod.h>
 #include <linux/delay.h>
 #include <linux/workqueue.h>
+#include <linux/ioport.h>
 #include <linux/nmi.h>
 #include <acpi/acpi.h>
 #include <asm/io.h>
@@ -338,6 +339,10 @@ acpi_status acpi_os_read_port(acpi_io_address port, u32 * value, u32 width)
 {
 	u32 dummy;
 
+	if (__check_region(&ioport_resource, port, width/8))
+		printk (KERN_INFO "ACPI read from allocated ioport %lx\n", 
+			port);
+
 	if (!value)
 		value = &dummy;
 
@@ -362,6 +367,10 @@ EXPORT_SYMBOL(acpi_os_read_port);
 
 acpi_status acpi_os_write_port(acpi_io_address port, u32 value, u32 width)
 {
+	if (__check_region(&ioport_resource, port, width/8))
+		printk (KERN_INFO "ACPI write to allocated ioport %lx\n", 
+			port);
+
 	switch (width) {
 	case 8:
 		outb(value, port);
@@ -387,6 +396,10 @@ acpi_os_read_memory(acpi_physical_address phys_addr, u32 * value, u32 width)
 	u32 dummy;
 	void __iomem *virt_addr;
 
+	if (__check_region(&iomem_resource, phys_addr, width/8))
+		printk (KERN_INFO "ACPI read from allocated iomem %lx\n", 
+			phys_addr);
+
 	virt_addr = ioremap(phys_addr, width);
 	if (!value)
 		value = &dummy;
@@ -415,6 +428,10 @@ acpi_os_write_memory(acpi_physical_address phys_addr, u32 value, u32 width)
 {
 	void __iomem *virt_addr;
 
+	if (__check_region(&iomem_resource, phys_addr, width/8))
+		printk (KERN_INFO "ACPI write to allocated iomem %lx\n", 
+			phys_addr);
+
 	virt_addr = ioremap(phys_addr, width);
 
 	switch (width) {

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

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Pavel,

On Fri, 2 Mar 2007 14:58:10 +0100, Pavel Machek wrote:
> Actually for the acpi stuff... we might wrap ACPI interpretter with a
> semaphore that needs to be taken before starting any AML code. Then
> just make sure sensors take same semaphore?

I like the idea, it should work as long as we are guaranteed that all
the hwmon device accesses happen in the AML code? I'm not familiar with
ACPI, so you tell me.

In practice it's rather the SMBus drivers which will need to take the
lock, as this is what the AML code will access (for SMBus-based
hardware monitoring drivers.) For non-SMBus based hardware monitoring
drivers, indeed the driver itself should take it. We will have to pay
attention to deadlocks on systems with multiple hardware monitoring
chips and/or SMBus channels.

Can you please provide a patch implementing your proposal in acpi? Then
I could implement the i2c/hwmon part on some selected drivers and start
testing real world scenarii.

Thanks,
-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Pavel,

On Fri, 2 Mar 2007 14:58:10 +0100, Pavel Machek wrote:
> Actually for the acpi stuff... we might wrap ACPI interpretter with a
> semaphore that needs to be taken before starting any AML code. Then
> just make sure sensors take same semaphore?

I like the idea, it should work as long as we are guaranteed that all
the hwmon device accesses happen in the AML code? I'm not familiar with
ACPI, so you tell me.

In practice it's rather the SMBus drivers which will need to take the
lock, as this is what the AML code will access (for SMBus-based
hardware monitoring drivers.) For non-SMBus based hardware monitoring
drivers, indeed the driver itself should take it. We will have to pay
attention to deadlocks on systems with multiple hardware monitoring
chips and/or SMBus channels.

Can you please provide a patch implementing your proposal in acpi? Then
I could implement the i2c/hwmon part on some selected drivers and start
testing real world scenarii.

Thanks,
-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Fri, 2 Mar 2007 14:18:40 +0000, Matthew Garrett wrote:
> On Fri, Mar 02, 2007 at 03:10:55PM +0100, Jean Delvare wrote:
> 
> > I'm not familiar with APCI at all so I didn't know, but what you write
> > here brings some hope. Would it be possible to parse all the DSDT code
> > at boot time and deduce all the ports which ACPI would need to request
> > to be safe? Or do we have to wait for the accesses to actually happen?
> 
> In theory I /think/ so, but it would probably end up being an 
> overestimate of the coverage actually needed. Trapping at runtime is 
> arguably more elegant?

It might be more elegant but it won't work. We don't want to prevent
ACPI from accessing these I/O ports. If we need to choose only one
"driver" accessing the I/O port, it must be acpi, at leat for now,
despite the fact that acpi provides very weak hardware monitoring
capabilities compared to the specific drivers.

Why would we end up with an overestimation if we check the I/O ports at
boot time? Do you have concrete cases in mind?

Thanks,
-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Fri, 2 Mar 2007 14:18:40 +0000, Matthew Garrett wrote:
> On Fri, Mar 02, 2007 at 03:10:55PM +0100, Jean Delvare wrote:
> 
> > I'm not familiar with APCI at all so I didn't know, but what you write
> > here brings some hope. Would it be possible to parse all the DSDT code
> > at boot time and deduce all the ports which ACPI would need to request
> > to be safe? Or do we have to wait for the accesses to actually happen?
> 
> In theory I /think/ so, but it would probably end up being an 
> overestimate of the coverage actually needed. Trapping at runtime is 
> arguably more elegant?

It might be more elegant but it won't work. We don't want to prevent
ACPI from accessing these I/O ports. If we need to choose only one
"driver" accessing the I/O port, it must be acpi, at leat for now,
despite the fact that acpi provides very weak hardware monitoring
capabilities compared to the specific drivers.

Why would we end up with an overestimation if we check the I/O ports at
boot time? Do you have concrete cases in mind?

Thanks,
-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Matthew Garrett]

On Fri, Mar 02, 2007 at 10:04:54PM +0100, Jean Delvare wrote:
> On Fri, 2 Mar 2007 14:18:40 +0000, Matthew Garrett wrote:
> > In theory I /think/ so, but it would probably end up being an 
> > overestimate of the coverage actually needed. Trapping at runtime is 
> > arguably more elegant?
> 
> It might be more elegant but it won't work. We don't want to prevent
> ACPI from accessing these I/O ports. If we need to choose only one
> "driver" accessing the I/O port, it must be acpi, at leat for now,
> despite the fact that acpi provides very weak hardware monitoring
> capabilities compared to the specific drivers.

Assuming arbitration of access, what's the problem with having two 
drivers accessing the same hardware? Do these chips generally have any 
significant internal state other than trip points and the like?

> Why would we end up with an overestimation if we check the I/O ports at
> boot time? Do you have concrete cases in mind?

ACPI will often describe large operation regions, but won't necessarily 
touch all of them. Effectively, every codepath would have to be walked 
through at boot time and checked for io access.

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

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Matthew Garrett]

On Fri, Mar 02, 2007 at 10:04:54PM +0100, Jean Delvare wrote:
> On Fri, 2 Mar 2007 14:18:40 +0000, Matthew Garrett wrote:
> > In theory I /think/ so, but it would probably end up being an 
> > overestimate of the coverage actually needed. Trapping at runtime is 
> > arguably more elegant?
> 
> It might be more elegant but it won't work. We don't want to prevent
> ACPI from accessing these I/O ports. If we need to choose only one
> "driver" accessing the I/O port, it must be acpi, at leat for now,
> despite the fact that acpi provides very weak hardware monitoring
> capabilities compared to the specific drivers.

Assuming arbitration of access, what's the problem with having two 
drivers accessing the same hardware? Do these chips generally have any 
significant internal state other than trip points and the like?

> Why would we end up with an overestimation if we check the I/O ports at
> boot time? Do you have concrete cases in mind?

ACPI will often describe large operation regions, but won't necessarily 
touch all of them. Effectively, every codepath would have to be walked 
through at boot time and checked for io access.

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

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Henrique de Moraes Holschuh]

On Fri, 02 Mar 2007, Jean Delvare wrote:
> I like the idea, it should work as long as we are guaranteed that all
> the hwmon device accesses happen in the AML code? I'm not familiar with
> ACPI, so you tell me.

Some fubar machines do it in SMM mode, and the AML code just reads memory
regions that were updated by the SMBIOS.  Or, if one is lucky, the AML
triggers the SMI (ThinkPads do this, for example -- but I am not sure that's
the only way a ThinkPad will cause a SMI), so it could be made to run
protected by the lock.

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Henrique de Moraes Holschuh]

On Fri, 02 Mar 2007, Jean Delvare wrote:
> I like the idea, it should work as long as we are guaranteed that all
> the hwmon device accesses happen in the AML code? I'm not familiar with
> ACPI, so you tell me.

Some fubar machines do it in SMM mode, and the AML code just reads memory
regions that were updated by the SMBIOS.  Or, if one is lucky, the AML
triggers the SMI (ThinkPads do this, for example -- but I am not sure that's
the only way a ThinkPad will cause a SMI), so it could be made to run
protected by the lock.

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Matthew,

On Fri, 2 Mar 2007 14:57:09 +0000, Matthew Garrett wrote:
> How about this? It's informational only, but ought to result in 
> complaints whenever ACPI tries to touch something that other hardware 
> has reserved. We can't fail these accesses, but in theory we could 
> consider some sort of locking layer that made it possible to interact 
> anyway. I haven't even checked if this builds, but I think the concept 
> is reasonable.

I like the patch, after adding some casts to the printf args it
compiles fine. However you print warnings each time a resource has been
reserved... without checking if it hasn't been reserved by ACPI itself!
My machine looks like this:

1000-107f : 0000:00:1f.0
  1000-1003 : ACPI PM1a_EVT_BLK
  1004-1005 : ACPI PM1a_CNT_BLK
  1008-100b : ACPI PM_TMR
  1010-1015 : ACPI CPU throttle
  1020-1020 : ACPI PM2_CNT_BLK
  1028-102b : ACPI GPE0_BLK
  102c-102f : ACPI GPE1_BLK

Given that these ports were reserved by ACPI it is perfectly legitimate
that ACPI accesses it, so we must not print a warning in this case. We
need to exclude from the test the regions those "name" starts with
"ACPI", but I'm not sure how we can do that.

Thanks,
-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Matthew,

On Fri, 2 Mar 2007 14:57:09 +0000, Matthew Garrett wrote:
> How about this? It's informational only, but ought to result in 
> complaints whenever ACPI tries to touch something that other hardware 
> has reserved. We can't fail these accesses, but in theory we could 
> consider some sort of locking layer that made it possible to interact 
> anyway. I haven't even checked if this builds, but I think the concept 
> is reasonable.

I like the patch, after adding some casts to the printf args it
compiles fine. However you print warnings each time a resource has been
reserved... without checking if it hasn't been reserved by ACPI itself!
My machine looks like this:

1000-107f : 0000:00:1f.0
  1000-1003 : ACPI PM1a_EVT_BLK
  1004-1005 : ACPI PM1a_CNT_BLK
  1008-100b : ACPI PM_TMR
  1010-1015 : ACPI CPU throttle
  1020-1020 : ACPI PM2_CNT_BLK
  1028-102b : ACPI GPE0_BLK
  102c-102f : ACPI GPE1_BLK

Given that these ports were reserved by ACPI it is perfectly legitimate
that ACPI accesses it, so we must not print a warning in this case. We
need to exclude from the test the regions those "name" starts with
"ACPI", but I'm not sure how we can do that.

Thanks,
-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Matthew Garrett]

On Fri, Mar 02, 2007 at 10:41:55PM +0100, Jean Delvare wrote:

> I like the patch, after adding some casts to the printf args it
> compiles fine. However you print warnings each time a resource has been
> reserved... without checking if it hasn't been reserved by ACPI itself!
> My machine looks like this:

Oops! I'll look into fixing that. Thanks, that's an excellent point...

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

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Matthew Garrett]

On Fri, Mar 02, 2007 at 10:41:55PM +0100, Jean Delvare wrote:

> I like the patch, after adding some casts to the printf args it
> compiles fine. However you print warnings each time a resource has been
> reserved... without checking if it hasn't been reserved by ACPI itself!
> My machine looks like this:

Oops! I'll look into fixing that. Thanks, that's an excellent point...

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

RE: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

Port (and memory) addresses can be dynamically generated by the AML code
and thus, there is no way that the ACPI subsystem can statically predict
any addresses that will be accessed by the AML.


> -----Original Message-----
> From: linux-acpi-owner@vger.kernel.org [mailto:linux-acpi-
> owner@vger.kernel.org] On Behalf Of Jean Delvare
> Sent: Friday, March 02, 2007 1:05 PM
> To: Matthew Garrett
> Cc: Pavel Machek; Chuck Ebbert; Rudolf Marek;
linux-acpi@vger.kernel.org;
> linux-kernel; lm-sensors@lm-sensors.org
> Subject: Re: [lm-sensors] Could the k8temp driver be interfering with
> ACPI?
> 
> On Fri, 2 Mar 2007 14:18:40 +0000, Matthew Garrett wrote:
> > On Fri, Mar 02, 2007 at 03:10:55PM +0100, Jean Delvare wrote:
> >
> > > I'm not familiar with APCI at all so I didn't know, but what you
write
> > > here brings some hope. Would it be possible to parse all the DSDT
code
> > > at boot time and deduce all the ports which ACPI would need to
request
> > > to be safe? Or do we have to wait for the accesses to actually
happen?
> >
> > In theory I /think/ so, but it would probably end up being an
> > overestimate of the coverage actually needed. Trapping at runtime is
> > arguably more elegant?
> 
> It might be more elegant but it won't work. We don't want to prevent
> ACPI from accessing these I/O ports. If we need to choose only one
> "driver" accessing the I/O port, it must be acpi, at leat for now,
> despite the fact that acpi provides very weak hardware monitoring
> capabilities compared to the specific drivers.
> 
> Why would we end up with an overestimation if we check the I/O ports
at
> boot time? Do you have concrete cases in mind?
> 
> Thanks,
> --
> Jean Delvare
> -
> To unsubscribe from this list: send the line "unsubscribe linux-acpi"
in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

Port (and memory) addresses can be dynamically generated by the AML code
and thus, there is no way that the ACPI subsystem can statically predict
any addresses that will be accessed by the AML.


> -----Original Message-----
> From: linux-acpi-owner@vger.kernel.org [mailto:linux-acpi-
> owner@vger.kernel.org] On Behalf Of Jean Delvare
> Sent: Friday, March 02, 2007 1:05 PM
> To: Matthew Garrett
> Cc: Pavel Machek; Chuck Ebbert; Rudolf Marek;
linux-acpi@vger.kernel.org;
> linux-kernel; lm-sensors@lm-sensors.org
> Subject: Re: [lm-sensors] Could the k8temp driver be interfering with
> ACPI?
> 
> On Fri, 2 Mar 2007 14:18:40 +0000, Matthew Garrett wrote:
> > On Fri, Mar 02, 2007 at 03:10:55PM +0100, Jean Delvare wrote:
> >
> > > I'm not familiar with APCI at all so I didn't know, but what you
write
> > > here brings some hope. Would it be possible to parse all the DSDT
code
> > > at boot time and deduce all the ports which ACPI would need to
request
> > > to be safe? Or do we have to wait for the accesses to actually
happen?
> >
> > In theory I /think/ so, but it would probably end up being an
> > overestimate of the coverage actually needed. Trapping at runtime is
> > arguably more elegant?
> 
> It might be more elegant but it won't work. We don't want to prevent
> ACPI from accessing these I/O ports. If we need to choose only one
> "driver" accessing the I/O port, it must be acpi, at leat for now,
> despite the fact that acpi provides very weak hardware monitoring
> capabilities compared to the specific drivers.
> 
> Why would we end up with an overestimation if we check the I/O ports
at
> boot time? Do you have concrete cases in mind?
> 
> Thanks,
> --
> Jean Delvare
> -
> To unsubscribe from this list: send the line "unsubscribe linux-acpi"
in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

Port (and memory) addresses can be dynamically generated by the AML code
and thus, there is no way that the ACPI subsystem can statically predict
any addresses that will be accessed by the AML.


> -----Original Message-----
> From: linux-acpi-owner at vger.kernel.org [mailto:linux-acpi-
> owner at vger.kernel.org] On Behalf Of Jean Delvare
> Sent: Friday, March 02, 2007 1:05 PM
> To: Matthew Garrett
> Cc: Pavel Machek; Chuck Ebbert; Rudolf Marek;
linux-acpi at vger.kernel.org;
> linux-kernel; lm-sensors at lm-sensors.org
> Subject: Re: [lm-sensors] Could the k8temp driver be interfering with
> ACPI?
> 
> On Fri, 2 Mar 2007 14:18:40 +0000, Matthew Garrett wrote:
> > On Fri, Mar 02, 2007 at 03:10:55PM +0100, Jean Delvare wrote:
> >
> > > I'm not familiar with APCI at all so I didn't know, but what you
write
> > > here brings some hope. Would it be possible to parse all the DSDT
code
> > > at boot time and deduce all the ports which ACPI would need to
request
> > > to be safe? Or do we have to wait for the accesses to actually
happen?
> >
> > In theory I /think/ so, but it would probably end up being an
> > overestimate of the coverage actually needed. Trapping at runtime is
> > arguably more elegant?
> 
> It might be more elegant but it won't work. We don't want to prevent
> ACPI from accessing these I/O ports. If we need to choose only one
> "driver" accessing the I/O port, it must be acpi, at leat for now,
> despite the fact that acpi provides very weak hardware monitoring
> capabilities compared to the specific drivers.
> 
> Why would we end up with an overestimation if we check the I/O ports
at
> boot time? Do you have concrete cases in mind?
> 
> Thanks,
> --
> Jean Delvare
> -
> To unsubscribe from this list: send the line "unsubscribe linux-acpi"
in
> the body of a message to majordomo at vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Fri, 2 Mar 2007 14:57:06 +0100, Pavel Machek wrote:
> On Fri 2007-03-02 14:37:20, Jean Delvare wrote:
> > drivers/pci/quirks.c is full of things we do against the BIOS authors
> > intent. You don't plan to remove them all, do you?
> 
> Notice how quirks.c is careful to name machines where given quirk is
> used.

Notice how we unhide the SMBus only on a few named machines as well.

> If you do whitelist "it is okay to do sensors accesses on this board",
> that is okay with me. But having quirk "on all future Intel chipsets,
> do foo" would be stupid.

We don't do that, never did, and never planed to do.

>                    (...) Enabling sensors blindly is pretty similar.

This is the view of someone obviously not interested in sensors. Go
tell that to all the lm-sensors users and they'll reply: "Disabling
sensors blindly would be stupid." Please realize that we're not writing
hardware monitoring drivers just for the fun of conflicting with ACPI
on some machines. We're writing them because users are asking for these
features.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Fri, 2 Mar 2007 14:57:06 +0100, Pavel Machek wrote:
> On Fri 2007-03-02 14:37:20, Jean Delvare wrote:
> > drivers/pci/quirks.c is full of things we do against the BIOS authors
> > intent. You don't plan to remove them all, do you?
> 
> Notice how quirks.c is careful to name machines where given quirk is
> used.

Notice how we unhide the SMBus only on a few named machines as well.

> If you do whitelist "it is okay to do sensors accesses on this board",
> that is okay with me. But having quirk "on all future Intel chipsets,
> do foo" would be stupid.

We don't do that, never did, and never planed to do.

>                    (...) Enabling sensors blindly is pretty similar.

This is the view of someone obviously not interested in sensors. Go
tell that to all the lm-sensors users and they'll reply: "Disabling
sensors blindly would be stupid." Please realize that we're not writing
hardware monitoring drivers just for the fun of conflicting with ACPI
on some machines. We're writing them because users are asking for these
features.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Matthew,

On Fri, 2 Mar 2007 21:12:51 +0000, Matthew Garrett wrote:
> On Fri, Mar 02, 2007 at 10:04:54PM +0100, Jean Delvare wrote:
> > It might be more elegant but it won't work. We don't want to prevent
> > ACPI from accessing these I/O ports. If we need to choose only one
> > "driver" accessing the I/O port, it must be acpi, at leat for now,
> > despite the fact that acpi provides very weak hardware monitoring
> > capabilities compared to the specific drivers.
> 
> Assuming arbitration of access, what's the problem with having two 
> drivers accessing the same hardware? Do these chips generally have any 
> significant internal state other than trip points and the like?

The "assuming arbitration of access" is the key part of your
sentence ;) The problem is that currently no arbitration is done. If it
was done, then state would probably not be a problem. Most hardware
monitoring drivers don't assume any state is preserved between
accesses, and those which do can easily be changed not to. The ACPI
side is another story though, I guess we can't assume anything about
the AML code's assumption on states, as it differs from one machine to
the next. But we can try to be cooperative and restore the sensible
registers (e.g. bank selector) in the same state we found them.

Anyway, just because we can't get things right on 100% of the machines
is no reason not to try anything at all. The current situation is bad,
any improvement would be welcome.

> > Why would we end up with an overestimation if we check the I/O ports at
> > boot time? Do you have concrete cases in mind?
> 
> ACPI will often describe large operation regions, but won't necessarily 
> touch all of them. Effectively, every codepath would have to be walked 
> through at boot time and checked for io access.

Is there anything preventing us from doing such a walk and pre-allocate
all the I/O ranges? I am not familiar with the ACPI code at all, would
you possibly propose a patch doing that?

If we can't do that, the overestimation approach might still work. I
wonder if it would cause problems in practice. If it does, we're back
to Pavel's AML lock.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Matthew,

On Fri, 2 Mar 2007 21:12:51 +0000, Matthew Garrett wrote:
> On Fri, Mar 02, 2007 at 10:04:54PM +0100, Jean Delvare wrote:
> > It might be more elegant but it won't work. We don't want to prevent
> > ACPI from accessing these I/O ports. If we need to choose only one
> > "driver" accessing the I/O port, it must be acpi, at leat for now,
> > despite the fact that acpi provides very weak hardware monitoring
> > capabilities compared to the specific drivers.
> 
> Assuming arbitration of access, what's the problem with having two 
> drivers accessing the same hardware? Do these chips generally have any 
> significant internal state other than trip points and the like?

The "assuming arbitration of access" is the key part of your
sentence ;) The problem is that currently no arbitration is done. If it
was done, then state would probably not be a problem. Most hardware
monitoring drivers don't assume any state is preserved between
accesses, and those which do can easily be changed not to. The ACPI
side is another story though, I guess we can't assume anything about
the AML code's assumption on states, as it differs from one machine to
the next. But we can try to be cooperative and restore the sensible
registers (e.g. bank selector) in the same state we found them.

Anyway, just because we can't get things right on 100% of the machines
is no reason not to try anything at all. The current situation is bad,
any improvement would be welcome.

> > Why would we end up with an overestimation if we check the I/O ports at
> > boot time? Do you have concrete cases in mind?
> 
> ACPI will often describe large operation regions, but won't necessarily 
> touch all of them. Effectively, every codepath would have to be walked 
> through at boot time and checked for io access.

Is there anything preventing us from doing such a walk and pre-allocate
all the I/O ranges? I am not familiar with the ACPI code at all, would
you possibly propose a patch doing that?

If we can't do that, the overestimation approach might still work. I
wonder if it would cause problems in practice. If it does, we're back
to Pavel's AML lock.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[David Hubbard]

Hi Jean,

> Is there anything preventing us from doing such a walk and pre-allocate
> all the I/O ranges? I am not familiar with the ACPI code at all, would
> you possibly propose a patch doing that?

Here's a random idea -- what do you think of it?

ACPI already allocates some I/O ranges, which was a surprise to me:

> My machine looks like this:
>
> 1000-107f : 0000:00:1f.0
>  1000-1003 : ACPI PM1a_EVT_BLK
>  1004-1005 : ACPI PM1a_CNT_BLK
>  1008-100b : ACPI PM_TMR
>  1010-1015 : ACPI CPU throttle
>  1020-1020 : ACPI PM2_CNT_BLK
>  1028-102b : ACPI GPE0_BLK
>  102c-102f : ACPI GPE1_BLK

For I/O and memory that ACPI accesses and has not reserved, the AML
interpreter could allocate at run-time.

I'm not sure how to implement exactly. For example, it would be bad to
have a /proc/ioports that had a lot of single ports allocated, for
example:
1000-107f : 0000:00:1f.0
 1000-1000 : ACPI PM1a_EVT_BLK
 1001-1001 : ACPI PM1a_EVT_BLK
 1002-1002 : ACPI PM1a_EVT_BLK
 1003-1003 : ACPI PM1a_EVT_BLK

Thus the AML interpreter would need to have some reasonable
intelligence when allocating regions. Conflict resolution would also
be more difficult, e.g. if a hwmon driver was loaded first and then
acpi as a module, ACPI could not allocate the region. Maybe run-time
allocating won't work.

And then, how would ACPI release a region after it has used it? The
easiest method would be to never release anything used even once.

Thoughts?

David

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[David Hubbard]

Hi Jean,

> Is there anything preventing us from doing such a walk and pre-allocate
> all the I/O ranges? I am not familiar with the ACPI code at all, would
> you possibly propose a patch doing that?

Here's a random idea -- what do you think of it?

ACPI already allocates some I/O ranges, which was a surprise to me:

> My machine looks like this:
>
> 1000-107f : 0000:00:1f.0
>  1000-1003 : ACPI PM1a_EVT_BLK
>  1004-1005 : ACPI PM1a_CNT_BLK
>  1008-100b : ACPI PM_TMR
>  1010-1015 : ACPI CPU throttle
>  1020-1020 : ACPI PM2_CNT_BLK
>  1028-102b : ACPI GPE0_BLK
>  102c-102f : ACPI GPE1_BLK

For I/O and memory that ACPI accesses and has not reserved, the AML
interpreter could allocate at run-time.

I'm not sure how to implement exactly. For example, it would be bad to
have a /proc/ioports that had a lot of single ports allocated, for
example:
1000-107f : 0000:00:1f.0
 1000-1000 : ACPI PM1a_EVT_BLK
 1001-1001 : ACPI PM1a_EVT_BLK
 1002-1002 : ACPI PM1a_EVT_BLK
 1003-1003 : ACPI PM1a_EVT_BLK

Thus the AML interpreter would need to have some reasonable
intelligence when allocating regions. Conflict resolution would also
be more difficult, e.g. if a hwmon driver was loaded first and then
acpi as a module, ACPI could not allocate the region. Maybe run-time
allocating won't work.

And then, how would ACPI release a region after it has used it? The
easiest method would be to never release anything used even once.

Thoughts?

David

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Matthew Garrett]

On Sat, Mar 03, 2007 at 08:47:21AM -0700, David Hubbard wrote:

> For I/O and memory that ACPI accesses and has not reserved, the AML
> interpreter could allocate at run-time.

Not ideal. ACPI's already fiddling with ranges that have been reserved 
by other drivers.

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

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Matthew Garrett]

On Sat, Mar 03, 2007 at 08:47:21AM -0700, David Hubbard wrote:

> For I/O and memory that ACPI accesses and has not reserved, the AML
> interpreter could allocate at run-time.

Not ideal. ACPI's already fiddling with ranges that have been reserved 
by other drivers.

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

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Rudolf Marek]

Hello all,

I was already thinking about some very general port forwarder,
(http://lists.lm-sensors.org/pipermail/lm-sensors/2007-February/018788.html)

But I'm not able to find out how to do that painlessly. Even don't know if to
use the classes or even some new "bus" in driver model :/ some suggestions on
this would be very welcome.

Let start with some very naive 1:1 solution which would try out if we can handle
"transactions" (typically write to one port read from other operation)
and preserve the "banks". As was already written some virtualization in driver
would be required, but it may be done because we know how the hardware is
supposed to work.

It seems that most simple solution to try out would be to add a possible
callback structure into a IORESOURCE structure. Please don't blame me for this,
it is just for testing purpose, we must think of better place, but as I said I
dont know where or how :/

Here is some concept:

When the request_region fails, ACPI could call request_region_acpi
which will return a pointer to some structure with defined callbacks
like do->outb do->inb. ACPI will first chekc if the pointer is not NULL ;)

This callback would end up in let say other linux driver.

For the typical hwmon chip things works similar way as CMOS access. Port is
written 0x295 with register address and port 0x296 is read or written.

Therefore:
void outb(port, val) {

if (port == 0x295) {
	reg_pointer=val;
} else if (port == 0x296) {
	// ACPI wants to write to the chip
	if (reg_pointer == BANKREG)
		reg_bank=val;
	}
	do_chip_write(reg_pointer, reg_bank_val);
}
}

void inb(port, val) {

if (port == 0x296) {
	

	return do_read_chip(reg_pointer, reg_bank);
}

}

The code above shows how the IO access could be synchronized with the ACPI, to
make everyone happy again. The driver could snoop what ACPI tries to change, so
it for example force reloading the cached values...

The second device which might be a challange is the SMBus controller. The
virtualization could be done same/similar way.

For PCI memory regions, the situation is same as for IO, however for the
pci_config space I cannot see any straight solution.


Regards
Rudolf

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Rudolf Marek]

Hello all,

I was already thinking about some very general port forwarder,
(http://lists.lm-sensors.org/pipermail/lm-sensors/2007-February/018788.html)

But I'm not able to find out how to do that painlessly. Even don't know if to
use the classes or even some new "bus" in driver model :/ some suggestions on
this would be very welcome.

Let start with some very naive 1:1 solution which would try out if we can handle
"transactions" (typically write to one port read from other operation)
and preserve the "banks". As was already written some virtualization in driver
would be required, but it may be done because we know how the hardware is
supposed to work.

It seems that most simple solution to try out would be to add a possible
callback structure into a IORESOURCE structure. Please don't blame me for this,
it is just for testing purpose, we must think of better place, but as I said I
dont know where or how :/

Here is some concept:

When the request_region fails, ACPI could call request_region_acpi
which will return a pointer to some structure with defined callbacks
like do->outb do->inb. ACPI will first chekc if the pointer is not NULL ;)

This callback would end up in let say other linux driver.

For the typical hwmon chip things works similar way as CMOS access. Port is
written 0x295 with register address and port 0x296 is read or written.

Therefore:
void outb(port, val) {

if (port = 0x295) {
	reg_pointer=val;
} else if (port = 0x296) {
	// ACPI wants to write to the chip
	if (reg_pointer = BANKREG)
		reg_bank=val;
	}
	do_chip_write(reg_pointer, reg_bank_val);
}
}

void inb(port, val) {

if (port = 0x296) {
	

	return do_read_chip(reg_pointer, reg_bank);
}

}

The code above shows how the IO access could be synchronized with the ACPI, to
make everyone happy again. The driver could snoop what ACPI tries to change, so
it for example force reloading the cached values...

The second device which might be a challange is the SMBus controller. The
virtualization could be done same/similar way.

For PCI memory regions, the situation is same as for IO, however for the
pci_config space I cannot see any straight solution.


Regards
Rudolf

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi David,

On Sat, 3 Mar 2007 08:47:21 -0700, David Hubbard wrote:
> > Is there anything preventing us from doing such a walk and pre-allocate
> > all the I/O ranges? I am not familiar with the ACPI code at all, would
> > you possibly propose a patch doing that?
> 
> Here's a random idea -- what do you think of it?
> 
> ACPI already allocates some I/O ranges, which was a surprise to me:
> 
> > My machine looks like this:
> >
> > 1000-107f : 0000:00:1f.0
> >  1000-1003 : ACPI PM1a_EVT_BLK
> >  1004-1005 : ACPI PM1a_CNT_BLK
> >  1008-100b : ACPI PM_TMR
> >  1010-1015 : ACPI CPU throttle
> >  1020-1020 : ACPI PM2_CNT_BLK
> >  1028-102b : ACPI GPE0_BLK
> >  102c-102f : ACPI GPE1_BLK
> 
> For I/O and memory that ACPI accesses and has not reserved, the AML
> interpreter could allocate at run-time.
> 
> I'm not sure how to implement exactly. For example, it would be bad to
> have a /proc/ioports that had a lot of single ports allocated, for
> example:
> 1000-107f : 0000:00:1f.0
>  1000-1000 : ACPI PM1a_EVT_BLK
>  1001-1001 : ACPI PM1a_EVT_BLK
>  1002-1002 : ACPI PM1a_EVT_BLK
>  1003-1003 : ACPI PM1a_EVT_BLK
> 
> Thus the AML interpreter would need to have some reasonable
> intelligence when allocating regions. Conflict resolution would also
> be more difficult, e.g. if a hwmon driver was loaded first and then
> acpi as a module, ACPI could not allocate the region. Maybe run-time
> allocating won't work.
> 
> And then, how would ACPI release a region after it has used it? The
> easiest method would be to never release anything used even once.

I've been thinking about it too, but I could convince myself quickly
that it would never work, so I didn't bother posting about it ;)

As you found out yourself, it isn't trivial to allocate the ports
dynamically. Either you'll end up with lots of 1-address ranges, or
you'll allocate more than is actually needed, or you'll need a
buddy-allocator like mechanism to merge contiguous ranges. This sounds
quite complex to get right. And very costly, too. I don't know the
exact algorithm to find out whether an I/O range is already allocated
or not, and by who, but having to do it for each port access from AML,
forever, sounds like a huge overhead.

On top of that, there is a risk that another driver already requested
the I/O range. And we do not want to prevent ACPI from accessing it!
OK, it would be cleaner that the current situation in a way, but it
could also have bad consequences as was underlined elsewhere in this
thread.

What we want is to grant access to the resources to at least ACPI (and
ACPI only if we can't do better), or if possible to both ACPI and
individual drivers but with some mechanism avoiding concurrent access
(be it a mutex or a port forwarder.)

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi David,

On Sat, 3 Mar 2007 08:47:21 -0700, David Hubbard wrote:
> > Is there anything preventing us from doing such a walk and pre-allocate
> > all the I/O ranges? I am not familiar with the ACPI code at all, would
> > you possibly propose a patch doing that?
> 
> Here's a random idea -- what do you think of it?
> 
> ACPI already allocates some I/O ranges, which was a surprise to me:
> 
> > My machine looks like this:
> >
> > 1000-107f : 0000:00:1f.0
> >  1000-1003 : ACPI PM1a_EVT_BLK
> >  1004-1005 : ACPI PM1a_CNT_BLK
> >  1008-100b : ACPI PM_TMR
> >  1010-1015 : ACPI CPU throttle
> >  1020-1020 : ACPI PM2_CNT_BLK
> >  1028-102b : ACPI GPE0_BLK
> >  102c-102f : ACPI GPE1_BLK
> 
> For I/O and memory that ACPI accesses and has not reserved, the AML
> interpreter could allocate at run-time.
> 
> I'm not sure how to implement exactly. For example, it would be bad to
> have a /proc/ioports that had a lot of single ports allocated, for
> example:
> 1000-107f : 0000:00:1f.0
>  1000-1000 : ACPI PM1a_EVT_BLK
>  1001-1001 : ACPI PM1a_EVT_BLK
>  1002-1002 : ACPI PM1a_EVT_BLK
>  1003-1003 : ACPI PM1a_EVT_BLK
> 
> Thus the AML interpreter would need to have some reasonable
> intelligence when allocating regions. Conflict resolution would also
> be more difficult, e.g. if a hwmon driver was loaded first and then
> acpi as a module, ACPI could not allocate the region. Maybe run-time
> allocating won't work.
> 
> And then, how would ACPI release a region after it has used it? The
> easiest method would be to never release anything used even once.

I've been thinking about it too, but I could convince myself quickly
that it would never work, so I didn't bother posting about it ;)

As you found out yourself, it isn't trivial to allocate the ports
dynamically. Either you'll end up with lots of 1-address ranges, or
you'll allocate more than is actually needed, or you'll need a
buddy-allocator like mechanism to merge contiguous ranges. This sounds
quite complex to get right. And very costly, too. I don't know the
exact algorithm to find out whether an I/O range is already allocated
or not, and by who, but having to do it for each port access from AML,
forever, sounds like a huge overhead.

On top of that, there is a risk that another driver already requested
the I/O range. And we do not want to prevent ACPI from accessing it!
OK, it would be cleaner that the current situation in a way, but it
could also have bad consequences as was underlined elsewhere in this
thread.

What we want is to grant access to the resources to at least ACPI (and
ACPI only if we can't do better), or if possible to both ACPI and
individual drivers but with some mechanism avoiding concurrent access
(be it a mutex or a port forwarder.)

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Rudolf Marek]

[-- Attachment #1: Type: text/plain, Size: 2675 bytes --]

Hello again,

I produced some code which I proposed in mail above. The patch is not for review
it is just a PoC to better show what I'm trying to do. It is a test case for my
motherboard which has W83627EHF chip and ACPI thermal method and w83627ehf
compete the device.

When no driver is loaded, acpi is doing its own raw access to device. However
when the w83627ehf driver is loaded, the ACPI access to ports (0x295/0x296) is
forwarded to the w83627ehf driver.

How it works? I added one pointer to the struct resource which will contain a
pointer to the structure with callback functions. I know this is not an ideal
solution, but for a test it works fine. Everytime ACPI wants do IO it will ask
via ____request_resource(conflict, &res) if some driver claimed the resource,
if so, it iterates the resource structure to the last entry, if the callback
structure exists it is called (and device driver context is passed too), if not,
raw hw access is performed.

The routines in EHF driver partly emulates the chip, the address of which next
acpi access will want to read or write is saved in the data->reg_pointer.

The actual access to the chip is done only when the data port of the chip is
accessed, and driver generic io function is called. Bank register writes are
faked to the data->reg_bank. All access to the chip which is done in these
routines respects the "virtual" bank register which was previously set by ACPI.
(current emulation is not 100% perfect, but this could be easily fixed)
If something was written to the chip, driver's register cache is invalidated.

This approach does not need any external locking because the actual device
access is done by one function of driver (which already locks).

As from ACPI point of view the device behaves same way as real HW, the big plus
is that the driver actually KNOWs what is ACPI trying to do to the chip. Of
course if the HW access is done in SMM some other countermeasures must be
invented like GBL lock or the "take ownership" stuff.

Maybe this forwarder may be used by other drivers, which may want to know what
is ACPI doing to hardware during suspend/resume methods...

As I wrote earlier the SMBus access could be also virtualized same way, there
will be only more virtual registers, real transaction will start when the ACPI
sets the "begin transaction" bits ;).

Maybe the pointer to driver context could be removed, and container_of could be
used instead to reach it. I will need some advise how to implement this
randevouz between ACPI and one Linux driver... Exploiting the resource structure
  is nice but pehaps not acceptable??? Some class maybe? We have only 1:1
relation...

Thanks,
Rudolf



[-- Attachment #2: acpi-io-sync.patch --]
[-- Type: text/x-diff, Size: 7824 bytes --]

Index: linux-2.6.20/include/linux/ioport.h
===================================================================
--- linux-2.6.20.orig/include/linux/ioport.h	2007-03-04 10:31:03.457166952 +0100
+++ linux-2.6.20/include/linux/ioport.h	2007-03-04 15:55:03.458260761 +0100
@@ -14,10 +14,19 @@
  * Resources are tree-like, allowing
  * nesting etc..
  */
+
+struct resource_ops {
+	int (*read_io)(u32 size, u32 port, u32 *where, void *priv);
+	int (*write_io)(u32 size, u32 port, u32 val, void *priv);
+	void *priv_data;
+};
+
 struct resource {
 	resource_size_t start;
 	resource_size_t end;
 	const char *name;
+	//void *resource_data;
+	struct resource_ops *res_ops;
 	unsigned long flags;
 	struct resource *parent, *sibling, *child;
 };
Index: linux-2.6.20/drivers/acpi/osl.c
===================================================================
--- linux-2.6.20.orig/drivers/acpi/osl.c	2007-03-04 10:31:30.626715256 +0100
+++ linux-2.6.20/drivers/acpi/osl.c	2007-03-04 16:18:29.958412636 +0100
@@ -36,6 +36,7 @@
 #include <linux/delay.h>
 #include <linux/workqueue.h>
 #include <linux/nmi.h>
+#include <linux/ioport.h>
 #include <acpi/acpi.h>
 #include <asm/io.h>
 #include <acpi/acpi_bus.h>
@@ -331,13 +332,8 @@
 	return ++t;
 }
 
-acpi_status acpi_os_read_port(acpi_io_address port, u32 * value, u32 width)
+static acpi_status acpi_os_read_port_raw(acpi_io_address port, u32 * value, u32 width)
 {
-	u32 dummy;
-
-	if (!value)
-		value = &dummy;
-
 	switch (width) {
 	case 8:
 		*(u8 *) value = inb(port);
@@ -355,10 +351,51 @@
 	return AE_OK;
 }
 
+acpi_status acpi_os_read_port(acpi_io_address port, u32 * value, u32 width)
+{
+	acpi_status ret = AE_OK;
+	struct resource *conflict, res;
+	u32 dummy;
+	int retval;
+
+	if (!value)
+		value = &dummy;
+
+	printk("%s: Port %x\n", __FUNCTION__ , (int) port);
+	res.flags = 0;
+	res.name = "ACPI Access";
+	res.start = port;
+	res.end = port + width/8 - 1;
+	
+	conflict = ____request_resource(&ioport_resource, &res);
+
+	while ((conflict) && (conflict->child))
+		conflict = ____request_resource(conflict, &res);
+
+	if ((conflict) && (conflict->res_ops) && (conflict->res_ops->write_io)) {
+		printk("Redir called\n");
+		retval = conflict->res_ops->read_io(width, port, value, conflict->res_ops->priv_data);
+
+		//this needs coding fix
+		if (retval < 0) {
+			ret = acpi_os_read_port_raw(port, value, width);
+		} else {
+			ret = AE_OK;
+		}
+	} else {
+		printk("Ordinary job\n");
+		ret = acpi_os_read_port_raw(port, value, width);
+		if (conflict == NULL)
+			release_resource(&res);
+	}
+
+	return ret;
+}
+
 EXPORT_SYMBOL(acpi_os_read_port);
 
-acpi_status acpi_os_write_port(acpi_io_address port, u32 value, u32 width)
-{
+static acpi_status acpi_os_write_port_raw(acpi_io_address port, u32 value, u32 width) {
+
 	switch (width) {
 	case 8:
 		outb(value, port);
@@ -376,6 +413,41 @@
 	return AE_OK;
 }
 
+acpi_status acpi_os_write_port(acpi_io_address port, u32 value, u32 width)
+{
+	struct resource *conflict, res;
+	acpi_status ret = AE_OK;
+	int retval;
+	res.flags = 0;
+	res.start = port;
+	res.end = port + width/8 - 1;
+	res.name = "ACPI Access";
+	printk("%s: Port %x\n", __FUNCTION__ , (int) port);
+	
+	conflict = ____request_resource(&ioport_resource, &res);
+
+	while ((conflict) && (conflict->child))
+		conflict = ____request_resource(conflict, &res);
+
+	if ((conflict) && (conflict->res_ops) && (conflict->res_ops->write_io)) {
+		printk("Redir called\n");
+		retval = conflict->res_ops->write_io(width, port, value,
+							conflict->res_ops->priv_data);
+		//this needs coding fix
+		if (retval < 0) {
+			ret = acpi_os_write_port_raw(port, value, width);
+		} else {
+			ret = AE_OK;
+		}
+	} else {
+		printk("Ordinary job\n");
+		ret = acpi_os_write_port_raw(port, value, width);
+		if (conflict == NULL)
+			release_resource(&res);
+	}
+	return ret;
+}
+
 EXPORT_SYMBOL(acpi_os_write_port);
 
 acpi_status
Index: linux-2.6.20/drivers/hwmon/w83627ehf.c
===================================================================
--- linux-2.6.20.orig/drivers/hwmon/w83627ehf.c	2007-03-04 10:31:03.689180173 +0100
+++ linux-2.6.20/drivers/hwmon/w83627ehf.c	2007-03-04 15:51:23.737739602 +0100
@@ -245,6 +245,9 @@
 	char valid;		/* !=0 if following fields are valid */
 	unsigned long last_updated;	/* In jiffies */
 
+	u8 reg_pointer;
+	u8 reg_bank;
+
 	/* Register values */
 	u8 in[10];		/* Register value */
 	u8 in_max[10];		/* Register value */
@@ -270,6 +273,7 @@
 
 	u8 fan_min_output[4]; /* minimum fan speed */
 	u8 fan_stop_time[4];
+	struct resource_ops res_ops;
 };
 
 static inline int is_word_sized(u16 reg)
@@ -344,6 +348,74 @@
 	return 0;
 }
 
+
+//we may return error/unsuppored and legacy acpi outb/inb stuff might take a chance
+
+static int acpi_emul_read(u32 size, u32 port, u32 *where, void *priv_data) {
+	struct i2c_client *client =  (struct i2c_client *) priv_data;
+	struct w83627ehf_data *data =  i2c_get_clientdata(client);
+	u16 tmp, reg = (data->reg_bank << 8) | data->reg_pointer;
+
+	if (size != 8) {
+		printk("this io port size is not emulated\n");
+		return -EINVAL;
+	}
+
+	if (port != (client->addr + DATA_REG_OFFSET)) {
+		printk("this port %x is not emulated\n", port);
+		return -EINVAL;
+	}
+
+	printk("%s: client addr %x port: %x \n", __FUNCTION__ , client->addr, port);
+	printk("Calling read with bank %x reg %x (%x)\n",
+			data->reg_bank, data->reg_pointer,reg);
+
+	tmp = w83627ehf_read_value(client, reg);
+	*((u8 *) where) = (is_word_sized(reg)) ? tmp >> 8 : tmp;
+	return 0;
+}
+
+static int acpi_emul_write(u32 size, u32 port, u32 val, void *priv_data) {
+	struct i2c_client *client = (struct i2c_client *) priv_data;
+	struct w83627ehf_data *data = i2c_get_clientdata(client);
+	u16 reg;
+
+	if (size != 8) {
+		printk("this io port size is not emulated\n");
+		return -EINVAL;
+	}
+
+	printk("%s client addr %x port %x\n", __FUNCTION__ , client->addr, port);
+
+	if (port == (client->addr + ADDR_REG_OFFSET)) {
+		data->reg_pointer = val;
+	} else if (port == (client->addr + DATA_REG_OFFSET)) {
+
+		if (data->reg_pointer == W83627EHF_REG_BANK) {
+			data->reg_bank = val & 0x3;
+			printk("REG bank changed %x\n", data->reg_bank);
+		} else { 
+			reg = (data->reg_bank << 8) | data->reg_pointer;
+			printk("Calling write with bank %x reg %x value %x\n",
+				data->reg_bank, data->reg_pointer, val);
+			
+			if (is_word_sized(reg)) {
+				u16 tmp = w83627ehf_read_value(client, reg);
+				tmp &= 0xff;
+				w83627ehf_write_value(client, reg, (val << 8) | tmp);
+			} else {
+				w83627ehf_write_value(client, reg, val);
+			}
+			data->valid = 0; /* invalidate cache */
+		}
+	} else {
+		printk("this port %x is not emulated\n", port);
+		return -EINVAL;	
+	}
+	return 0;
+}
+
+
 /* This function assumes that the caller holds data->update_lock */
 static void w83627ehf_write_fan_div(struct i2c_client *client, int nr)
 {
@@ -1173,11 +1245,12 @@
 	struct i2c_client *client;
 	struct w83627ehf_data *data;
 	struct device *dev;
+	struct resource *res;
 	u8 fan4pin, fan5pin;
 	int i, err = 0;
 
-	if (!request_region(address + REGION_OFFSET, REGION_LENGTH,
-	                    w83627ehf_driver.driver.name)) {
+	if (!(res = request_region(address + REGION_OFFSET, REGION_LENGTH,
+	                    w83627ehf_driver.driver.name))) {
 		err = -EBUSY;
 		goto exit;
 	}
@@ -1292,6 +1365,13 @@
 		goto exit_remove;
 	}
 
+	/* there is race condition right now */
+	
+	printk("RES %p res->res_ops %p client %p\n",res, res->res_ops, client);
+	data->res_ops.write_io = &acpi_emul_write;
+	data->res_ops.read_io = &acpi_emul_read;
+	data->res_ops.priv_data = client;
+	res->res_ops=&data->res_ops;
 	return 0;
 
 exit_remove:
@@ -1315,6 +1395,7 @@
 
 	if ((err = i2c_detach_client(client)))
 		return err;
+
 	release_region(client->addr + REGION_OFFSET, REGION_LENGTH);
 	kfree(data);
 

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Rudolf Marek]

Hello again,

I produced some code which I proposed in mail above. The patch is not for review
it is just a PoC to better show what I'm trying to do. It is a test case for my
motherboard which has W83627EHF chip and ACPI thermal method and w83627ehf
compete the device.

When no driver is loaded, acpi is doing its own raw access to device. However
when the w83627ehf driver is loaded, the ACPI access to ports (0x295/0x296) is
forwarded to the w83627ehf driver.

How it works? I added one pointer to the struct resource which will contain a
pointer to the structure with callback functions. I know this is not an ideal
solution, but for a test it works fine. Everytime ACPI wants do IO it will ask
via ____request_resource(conflict, &res) if some driver claimed the resource,
if so, it iterates the resource structure to the last entry, if the callback
structure exists it is called (and device driver context is passed too), if not,
raw hw access is performed.

The routines in EHF driver partly emulates the chip, the address of which next
acpi access will want to read or write is saved in the data->reg_pointer.

The actual access to the chip is done only when the data port of the chip is
accessed, and driver generic io function is called. Bank register writes are
faked to the data->reg_bank. All access to the chip which is done in these
routines respects the "virtual" bank register which was previously set by ACPI.
(current emulation is not 100% perfect, but this could be easily fixed)
If something was written to the chip, driver's register cache is invalidated.

This approach does not need any external locking because the actual device
access is done by one function of driver (which already locks).

As from ACPI point of view the device behaves same way as real HW, the big plus
is that the driver actually KNOWs what is ACPI trying to do to the chip. Of
course if the HW access is done in SMM some other countermeasures must be
invented like GBL lock or the "take ownership" stuff.

Maybe this forwarder may be used by other drivers, which may want to know what
is ACPI doing to hardware during suspend/resume methods...

As I wrote earlier the SMBus access could be also virtualized same way, there
will be only more virtual registers, real transaction will start when the ACPI
sets the "begin transaction" bits ;).

Maybe the pointer to driver context could be removed, and container_of could be
used instead to reach it. I will need some advise how to implement this
randevouz between ACPI and one Linux driver... Exploiting the resource structure
  is nice but pehaps not acceptable??? Some class maybe? We have only 1:1
relation...

Thanks,
Rudolf


-------------- next part --------------
A non-text attachment was scrubbed...
Name: acpi-io-sync.patch
Type: text/x-diff
Size: 7824 bytes
Desc: not available
Url : http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20070304/8d85322d/attachment.bin 

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Rudolf,

On Sun, 04 Mar 2007 18:29:12 +0100, Rudolf Marek wrote:
> I produced some code which I proposed in mail above. The patch is not for review
> it is just a PoC to better show what I'm trying to do. It is a test case for my
> motherboard which has W83627EHF chip and ACPI thermal method and w83627ehf
> compete the device.
> 
> When no driver is loaded, acpi is doing its own raw access to device. However
> when the w83627ehf driver is loaded, the ACPI access to ports (0x295/0x296) is
> forwarded to the w83627ehf driver.
> 
> How it works? I added one pointer to the struct resource which will contain a
> pointer to the structure with callback functions. I know this is not an ideal
> solution, but for a test it works fine. Everytime ACPI wants do IO it will ask
> via ____request_resource(conflict, &res) if some driver claimed the resource,
> if so, it iterates the resource structure to the last entry, if the callback
> structure exists it is called (and device driver context is passed too), if not,
> raw hw access is performed.
> 
> The routines in EHF driver partly emulates the chip, the address of which next
> acpi access will want to read or write is saved in the data->reg_pointer.
> 
> The actual access to the chip is done only when the data port of the chip is
> accessed, and driver generic io function is called. Bank register writes are
> faked to the data->reg_bank. All access to the chip which is done in these
> routines respects the "virtual" bank register which was previously set by ACPI.
> (current emulation is not 100% perfect, but this could be easily fixed)
> If something was written to the chip, driver's register cache is invalidated.
> 
> This approach does not need any external locking because the actual device
> access is done by one function of driver (which already locks).
> 
> As from ACPI point of view the device behaves same way as real HW, the big plus
> is that the driver actually KNOWs what is ACPI trying to do to the chip. Of
> course if the HW access is done in SMM some other countermeasures must be
> invented like GBL lock or the "take ownership" stuff.
> 
> Maybe this forwarder may be used by other drivers, which may want to know what
> is ACPI doing to hardware during suspend/resume methods...
> 
> As I wrote earlier the SMBus access could be also virtualized same way, there
> will be only more virtual registers, real transaction will start when the ACPI
> sets the "begin transaction" bits ;).
> 
> Maybe the pointer to driver context could be removed, and container_of could be
> used instead to reach it. I will need some advise how to implement this
> randevouz between ACPI and one Linux driver... Exploiting the resource structure
>   is nice but pehaps not acceptable??? Some class maybe? We have only 1:1
> relation...

I like your implementation, it's nice. Thanks for coming up with this
idea. The only change I would propose is to make the last parameter of
the read_io and write_io callback functions a device * instead of a
void *. This would let the compiler check the types and avoid improper
casts.

One bug in your code:

> +acpi_status acpi_os_read_port(acpi_io_address port, u32 * value, u32 width)
> +{
> +	acpi_status ret = AE_OK;
> +	struct resource *conflict, res;
> +	u32 dummy;
> +	int retval;
> +
> +	if (!value)
> +		value = &dummy;
> +
> +	printk("%s: Port %x\n", __FUNCTION__ , (int) port);
> +	res.flags = 0;
> +	res.name = "ACPI Access";
> +	res.start = port;
> +	res.end = port + width/8 - 1;
> +	
> +	conflict = ____request_resource(&ioport_resource, &res);
> +
> +	while ((conflict) && (conflict->child))
> +		conflict = ____request_resource(conflict, &res);
> +
> +	if ((conflict) && (conflict->res_ops) && (conflict->res_ops->write_io)) {

You want to check that conflict->res_ops->read_io exists, not write_io.

> +		printk("Redir called\n");
> +		retval = conflict->res_ops->read_io(width, port, value, conflict->res_ops->priv_data);
> +
> +		//this needs coding fix
> +		if (retval < 0) {
> +			ret = acpi_os_read_port_raw(port, value, width);
> +		} else {
> +			ret = AE_OK;
> +		}
> +	} else {
> +		printk("Ordinary job\n");
> +		ret = acpi_os_read_port_raw(port, value, width);
> +		if (conflict == NULL)
> +			release_resource(&res);
> +	}
> +
> +	return ret;
> +}

This port forwarding approach is interesting, maybe even beyond the
ACPI case we are trying to solve here. However, it has some drawbacks:

1* It requires that we modify each driver individually. It's quite a
shame that we have to update drivers all around the kernel tree with
specific code to workaround a problem which is originally located in a
single place (the ACPI subsystem.) That being said this isn't a blocker
point, if this is the only way, we'll do it. But that's a rather great
amount of work.

2* It seems to incur a signficant performance penalty.
____request_resource isn't cheap as far as I know. And in the
absence of conflict, you are even allocating and releasing the resource
on _each_ I/O operation, which certainly isn't cheap either. Again, it
is not a blocker point, after all this is a workaround for an improper
behavior of the acpi subsystem, this performance penalty is the price to
pay. But there is also a performance penalty for legitimate I/O access,
which worries me more.

3* What about concurrent accesses from ACPI itself? Unless we have a
guarantee that only one kernel thread can run AML code at a given
moment, I can imagine a conflict happening, as your code protects us
well against ACPI and driver concurrent accesses, but not against two
concurrent ACPI accesses. But I'm not familiar with ACPI - maybe we do
have this guarantee? OTOH, if this ever happens, well, it would happen
even without your code.

So for now I tend to think that the idea of a global AML lock proposed
by Pavel Machek and Bodo Eggert would be more efficient. And it
wouldn't need any driver-specific code, so it would be much more simple
to implement. The drawback being that we serialize more than we really
need to (e.g. the hardware monitoring driver will not be allowed to
access the chip when _any_ AML code is running, not just when the AML
code accesses the hardware monitoring chip, and if two drivers want to
check the AML lock, they'll exclude each other as well.) But I wonder
if this is a problem in practice. Maybe we'll have to make some
profiling on both solutions and compare.

Thanks,
-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Rudolf,

On Sun, 04 Mar 2007 18:29:12 +0100, Rudolf Marek wrote:
> I produced some code which I proposed in mail above. The patch is not for review
> it is just a PoC to better show what I'm trying to do. It is a test case for my
> motherboard which has W83627EHF chip and ACPI thermal method and w83627ehf
> compete the device.
> 
> When no driver is loaded, acpi is doing its own raw access to device. However
> when the w83627ehf driver is loaded, the ACPI access to ports (0x295/0x296) is
> forwarded to the w83627ehf driver.
> 
> How it works? I added one pointer to the struct resource which will contain a
> pointer to the structure with callback functions. I know this is not an ideal
> solution, but for a test it works fine. Everytime ACPI wants do IO it will ask
> via ____request_resource(conflict, &res) if some driver claimed the resource,
> if so, it iterates the resource structure to the last entry, if the callback
> structure exists it is called (and device driver context is passed too), if not,
> raw hw access is performed.
> 
> The routines in EHF driver partly emulates the chip, the address of which next
> acpi access will want to read or write is saved in the data->reg_pointer.
> 
> The actual access to the chip is done only when the data port of the chip is
> accessed, and driver generic io function is called. Bank register writes are
> faked to the data->reg_bank. All access to the chip which is done in these
> routines respects the "virtual" bank register which was previously set by ACPI.
> (current emulation is not 100% perfect, but this could be easily fixed)
> If something was written to the chip, driver's register cache is invalidated.
> 
> This approach does not need any external locking because the actual device
> access is done by one function of driver (which already locks).
> 
> As from ACPI point of view the device behaves same way as real HW, the big plus
> is that the driver actually KNOWs what is ACPI trying to do to the chip. Of
> course if the HW access is done in SMM some other countermeasures must be
> invented like GBL lock or the "take ownership" stuff.
> 
> Maybe this forwarder may be used by other drivers, which may want to know what
> is ACPI doing to hardware during suspend/resume methods...
> 
> As I wrote earlier the SMBus access could be also virtualized same way, there
> will be only more virtual registers, real transaction will start when the ACPI
> sets the "begin transaction" bits ;).
> 
> Maybe the pointer to driver context could be removed, and container_of could be
> used instead to reach it. I will need some advise how to implement this
> randevouz between ACPI and one Linux driver... Exploiting the resource structure
>   is nice but pehaps not acceptable??? Some class maybe? We have only 1:1
> relation...

I like your implementation, it's nice. Thanks for coming up with this
idea. The only change I would propose is to make the last parameter of
the read_io and write_io callback functions a device * instead of a
void *. This would let the compiler check the types and avoid improper
casts.

One bug in your code:

> +acpi_status acpi_os_read_port(acpi_io_address port, u32 * value, u32 width)
> +{
> +	acpi_status ret = AE_OK;
> +	struct resource *conflict, res;
> +	u32 dummy;
> +	int retval;
> +
> +	if (!value)
> +		value = &dummy;
> +
> +	printk("%s: Port %x\n", __FUNCTION__ , (int) port);
> +	res.flags = 0;
> +	res.name = "ACPI Access";
> +	res.start = port;
> +	res.end = port + width/8 - 1;
> +	
> +	conflict = ____request_resource(&ioport_resource, &res);
> +
> +	while ((conflict) && (conflict->child))
> +		conflict = ____request_resource(conflict, &res);
> +
> +	if ((conflict) && (conflict->res_ops) && (conflict->res_ops->write_io)) {

You want to check that conflict->res_ops->read_io exists, not write_io.

> +		printk("Redir called\n");
> +		retval = conflict->res_ops->read_io(width, port, value, conflict->res_ops->priv_data);
> +
> +		//this needs coding fix
> +		if (retval < 0) {
> +			ret = acpi_os_read_port_raw(port, value, width);
> +		} else {
> +			ret = AE_OK;
> +		}
> +	} else {
> +		printk("Ordinary job\n");
> +		ret = acpi_os_read_port_raw(port, value, width);
> +		if (conflict = NULL)
> +			release_resource(&res);
> +	}
> +
> +	return ret;
> +}

This port forwarding approach is interesting, maybe even beyond the
ACPI case we are trying to solve here. However, it has some drawbacks:

1* It requires that we modify each driver individually. It's quite a
shame that we have to update drivers all around the kernel tree with
specific code to workaround a problem which is originally located in a
single place (the ACPI subsystem.) That being said this isn't a blocker
point, if this is the only way, we'll do it. But that's a rather great
amount of work.

2* It seems to incur a signficant performance penalty.
____request_resource isn't cheap as far as I know. And in the
absence of conflict, you are even allocating and releasing the resource
on _each_ I/O operation, which certainly isn't cheap either. Again, it
is not a blocker point, after all this is a workaround for an improper
behavior of the acpi subsystem, this performance penalty is the price to
pay. But there is also a performance penalty for legitimate I/O access,
which worries me more.

3* What about concurrent accesses from ACPI itself? Unless we have a
guarantee that only one kernel thread can run AML code at a given
moment, I can imagine a conflict happening, as your code protects us
well against ACPI and driver concurrent accesses, but not against two
concurrent ACPI accesses. But I'm not familiar with ACPI - maybe we do
have this guarantee? OTOH, if this ever happens, well, it would happen
even without your code.

So for now I tend to think that the idea of a global AML lock proposed
by Pavel Machek and Bodo Eggert would be more efficient. And it
wouldn't need any driver-specific code, so it would be much more simple
to implement. The drawback being that we serialize more than we really
need to (e.g. the hardware monitoring driver will not be allowed to
access the chip when _any_ AML code is running, not just when the AML
code accesses the hardware monitoring chip, and if two drivers want to
check the AML lock, they'll exclude each other as well.) But I wonder
if this is a problem in practice. Maybe we'll have to make some
profiling on both solutions and compare.

Thanks,
-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[David Hubbard]

Hi Jean, Rudolf,

> 1* It requires that we modify each driver individually. It's quite a
> shame that we have to update drivers all around the kernel tree with
> specific code to workaround a problem which is originally located in a
> single place (the ACPI subsystem.) That being said this isn't a blocker
> point, if this is the only way, we'll do it. But that's a rather great
> amount of work.

These additions to drivers might help with the kernel virtualization.

> 2* It seems to incur a signficant performance penalty.
> ____request_resource isn't cheap as far as I know. And in the
> absence of conflict, you are even allocating and releasing the resource
> on _each_ I/O operation, which certainly isn't cheap either. Again, it
> is not a blocker point, after all this is a workaround for an improper
> behavior of the acpi subsystem, this performance penalty is the price to
> pay. But there is also a performance penalty for legitimate I/O access,
> which worries me more.

I thought Rudolf's patch allocated the resource in the driver
(w83627ehf) and ACPI contacted the driver when it could not allocate
the resource. Since ACPI never *really* wants to allocate the
resource, is there a fast-path check it could do? This would help
performance.

> 3* What about concurrent accesses from ACPI itself? Unless we have a
> guarantee that only one kernel thread can run AML code at a given
> moment, I can imagine a conflict happening, as your code protects us
> well against ACPI and driver concurrent accesses, but not against two
> concurrent ACPI accesses. But I'm not familiar with ACPI - maybe we do
> have this guarantee? OTOH, if this ever happens, well, it would happen
> even without your code.

All of ACPI uses the same "virtualized" access mechanism, so even if
two threads are poking at the bank select register in a race condition
(that would be awful code, I think) they would see the same virtual
behavior.

> So for now I tend to think that the idea of a global AML lock proposed
> by Pavel Machek and Bodo Eggert would be more efficient. And it
> wouldn't need any driver-specific code, so it would be much more simple
> to implement. The drawback being that we serialize more than we really
> need to (e.g. the hardware monitoring driver will not be allowed to
> access the chip when _any_ AML code is running, not just when the AML
> code accesses the hardware monitoring chip, and if two drivers want to
> check the AML lock, they'll exclude each other as well.) But I wonder
> if this is a problem in practice. Maybe we'll have to make some
> profiling on both solutions and compare.

I like the virtualized driver method (if it wasn't obvious!) but the
global AML lock works also. It will be interesting to see profiling of
both solutions.

David

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[David Hubbard]

Hi Jean, Rudolf,

> 1* It requires that we modify each driver individually. It's quite a
> shame that we have to update drivers all around the kernel tree with
> specific code to workaround a problem which is originally located in a
> single place (the ACPI subsystem.) That being said this isn't a blocker
> point, if this is the only way, we'll do it. But that's a rather great
> amount of work.

These additions to drivers might help with the kernel virtualization.

> 2* It seems to incur a signficant performance penalty.
> ____request_resource isn't cheap as far as I know. And in the
> absence of conflict, you are even allocating and releasing the resource
> on _each_ I/O operation, which certainly isn't cheap either. Again, it
> is not a blocker point, after all this is a workaround for an improper
> behavior of the acpi subsystem, this performance penalty is the price to
> pay. But there is also a performance penalty for legitimate I/O access,
> which worries me more.

I thought Rudolf's patch allocated the resource in the driver
(w83627ehf) and ACPI contacted the driver when it could not allocate
the resource. Since ACPI never *really* wants to allocate the
resource, is there a fast-path check it could do? This would help
performance.

> 3* What about concurrent accesses from ACPI itself? Unless we have a
> guarantee that only one kernel thread can run AML code at a given
> moment, I can imagine a conflict happening, as your code protects us
> well against ACPI and driver concurrent accesses, but not against two
> concurrent ACPI accesses. But I'm not familiar with ACPI - maybe we do
> have this guarantee? OTOH, if this ever happens, well, it would happen
> even without your code.

All of ACPI uses the same "virtualized" access mechanism, so even if
two threads are poking at the bank select register in a race condition
(that would be awful code, I think) they would see the same virtual
behavior.

> So for now I tend to think that the idea of a global AML lock proposed
> by Pavel Machek and Bodo Eggert would be more efficient. And it
> wouldn't need any driver-specific code, so it would be much more simple
> to implement. The drawback being that we serialize more than we really
> need to (e.g. the hardware monitoring driver will not be allowed to
> access the chip when _any_ AML code is running, not just when the AML
> code accesses the hardware monitoring chip, and if two drivers want to
> check the AML lock, they'll exclude each other as well.) But I wonder
> if this is a problem in practice. Maybe we'll have to make some
> profiling on both solutions and compare.

I like the virtualized driver method (if it wasn't obvious!) but the
global AML lock works also. It will be interesting to see profiling of
both solutions.

David

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > > Why would we end up with an overestimation if we check the I/O ports at
> > > boot time? Do you have concrete cases in mind?
> > 
> > ACPI will often describe large operation regions, but won't necessarily 
> > touch all of them. Effectively, every codepath would have to be walked 
> > through at boot time and checked for io access.
> 
> Is there anything preventing us from doing such a walk and pre-allocate
> all the I/O ranges? I am not familiar with the ACPI code at all, would
> you possibly propose a patch doing that?

ACPI AML is probably turing-complete: I'm afraid you are trying to
solve the halting problem (-> impossible).
						Pavel

> 
> If we can't do that, the overestimation approach might still work. I
> wonder if it would cause problems in practice. If it does, we're back
> to Pavel's AML lock.
> 
> -- 
> Jean Delvare

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > > Why would we end up with an overestimation if we check the I/O ports at
> > > boot time? Do you have concrete cases in mind?
> > 
> > ACPI will often describe large operation regions, but won't necessarily 
> > touch all of them. Effectively, every codepath would have to be walked 
> > through at boot time and checked for io access.
> 
> Is there anything preventing us from doing such a walk and pre-allocate
> all the I/O ranges? I am not familiar with the ACPI code at all, would
> you possibly propose a patch doing that?

ACPI AML is probably turing-complete: I'm afraid you are trying to
solve the halting problem (-> impossible).
						Pavel

> 
> If we can't do that, the overestimation approach might still work. I
> wonder if it would cause problems in practice. If it does, we're back
> to Pavel's AML lock.
> 
> -- 
> Jean Delvare

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Benny Amorsen]

>>>>> "PM" == Pavel Machek <pavel@ucw.cz> writes:

PM> ACPI AML is probably turing-complete: I'm afraid you are trying to
PM> solve the halting problem (-> impossible).

If you can restrict the virtual machine which AML runs in to a limited
amount of memory/storage, you can solve halting for it in exponential
time.

Unfortunately the problem here is even harder than halting. With the
halting problem you get all inputs beforehand. For AML, the locations
it accesses could be dependent on what it reads off of the hardware.


/Benny

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Benny Amorsen]

>>>>> "PM" = Pavel Machek <pavel at ucw.cz> writes:

PM> ACPI AML is probably turing-complete: I'm afraid you are trying to
PM> solve the halting problem (-> impossible).

If you can restrict the virtual machine which AML runs in to a limited
amount of memory/storage, you can solve halting for it in exponential
time.

Unfortunately the problem here is even harder than halting. With the
halting problem you get all inputs beforehand. For AML, the locations
it accesses could be dependent on what it reads off of the hardware.


/Benny

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi David,

On Mon, 5 Mar 2007 13:35:20 -0800, David Hubbard wrote:
> > 2* It seems to incur a signficant performance penalty.
> > ____request_resource isn't cheap as far as I know. And in the
> > absence of conflict, you are even allocating and releasing the resource
> > on _each_ I/O operation, which certainly isn't cheap either. Again, it
> > is not a blocker point, after all this is a workaround for an improper
> > behavior of the acpi subsystem, this performance penalty is the price to
> > pay. But there is also a performance penalty for legitimate I/O access,
> > which worries me more.
> 
> I thought Rudolf's patch allocated the resource in the driver
> (w83627ehf) and ACPI contacted the driver when it could not allocate
> the resource. Since ACPI never *really* wants to allocate the
> resource, is there a fast-path check it could do? This would help
> performance.

Alas, I don't think it can work safely. If ACPI doesn't actually
allocate the resource, there is a risk that another driver does so
between the conflict check and the I/O access. I believe this is the
reason why Rudolf used ____request_resource and not __check_region.

> I like the virtualized driver method (if it wasn't obvious!) but the
> global AML lock works also. It will be interesting to see profiling of
> both solutions.

In fact Rudolf's solution is nice for LPC chips, however I don't think
it can work with SMBus chips. It might intercept the accesses to the
SMBus master and be able to emulate it, even though this will be more
complex that the W83627EHF case, and I'm not sure about the PCI config
space. But ultimately we need to emulate all the chips behind the
SMBus, too. The drivers for these chips won't know if they are accessed
for real of through the emulation layer, so there is no way they'll
remember states, while they might have to (e.g. the W83781D has a bank
register too.)

The AML lock approach, OTOH, should work fine in all cases as long as
the context doesn't need to be remembered across AML "sections".

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi David,

On Mon, 5 Mar 2007 13:35:20 -0800, David Hubbard wrote:
> > 2* It seems to incur a signficant performance penalty.
> > ____request_resource isn't cheap as far as I know. And in the
> > absence of conflict, you are even allocating and releasing the resource
> > on _each_ I/O operation, which certainly isn't cheap either. Again, it
> > is not a blocker point, after all this is a workaround for an improper
> > behavior of the acpi subsystem, this performance penalty is the price to
> > pay. But there is also a performance penalty for legitimate I/O access,
> > which worries me more.
> 
> I thought Rudolf's patch allocated the resource in the driver
> (w83627ehf) and ACPI contacted the driver when it could not allocate
> the resource. Since ACPI never *really* wants to allocate the
> resource, is there a fast-path check it could do? This would help
> performance.

Alas, I don't think it can work safely. If ACPI doesn't actually
allocate the resource, there is a risk that another driver does so
between the conflict check and the I/O access. I believe this is the
reason why Rudolf used ____request_resource and not __check_region.

> I like the virtualized driver method (if it wasn't obvious!) but the
> global AML lock works also. It will be interesting to see profiling of
> both solutions.

In fact Rudolf's solution is nice for LPC chips, however I don't think
it can work with SMBus chips. It might intercept the accesses to the
SMBus master and be able to emulate it, even though this will be more
complex that the W83627EHF case, and I'm not sure about the PCI config
space. But ultimately we need to emulate all the chips behind the
SMBus, too. The drivers for these chips won't know if they are accessed
for real of through the emulation layer, so there is no way they'll
remember states, while they might have to (e.g. the W83781D has a bank
register too.)

The AML lock approach, OTOH, should work fine in all cases as long as
the context doesn't need to be remembered across AML "sections".

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Pavel,

On Mon, 5 Mar 2007 22:25:48 +0000, Pavel Machek wrote:
> > Is there anything preventing us from doing such a walk and pre-allocate
> > all the I/O ranges? I am not familiar with the ACPI code at all, would
> > you possibly propose a patch doing that?
> 
> ACPI AML is probably turing-complete: I'm afraid you are trying to
> solve the halting problem (-> impossible).

Can you please translate this into something mere humans like myself
have a chance to understand?

Thanks,
-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Pavel,

On Mon, 5 Mar 2007 22:25:48 +0000, Pavel Machek wrote:
> > Is there anything preventing us from doing such a walk and pre-allocate
> > all the I/O ranges? I am not familiar with the ACPI code at all, would
> > you possibly propose a patch doing that?
> 
> ACPI AML is probably turing-complete: I'm afraid you are trying to
> solve the halting problem (-> impossible).

Can you please translate this into something mere humans like myself
have a chance to understand?

Thanks,
-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Stefan Monnier]

>> > Is there anything preventing us from doing such a walk and pre-allocate
>> > all the I/O ranges? I am not familiar with the ACPI code at all, would
>> > you possibly propose a patch doing that?
>> 
>> ACPI AML is probably turing-complete: I'm afraid you are trying to
>> solve the halting problem (-> impossible).

> Can you please translate this into something mere humans like myself
> have a chance to understand?

Try wikipedia "Halting problem".


        Stefan

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Stefan Monnier]

>> > Is there anything preventing us from doing such a walk and pre-allocate
>> > all the I/O ranges? I am not familiar with the ACPI code at all, would
>> > you possibly propose a patch doing that?
>> 
>> ACPI AML is probably turing-complete: I'm afraid you are trying to
>> solve the halting problem (-> impossible).

> Can you please translate this into something mere humans like myself
> have a chance to understand?

Try wikipedia "Halting problem".


        Stefan

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Stefan Monnier]

>> > Is there anything preventing us from doing such a walk and pre-allocate
>> > all the I/O ranges? I am not familiar with the ACPI code at all, would
>> > you possibly propose a patch doing that?
>> 
>> ACPI AML is probably turing-complete: I'm afraid you are trying to
>> solve the halting problem (-> impossible).

> Can you please translate this into something mere humans like myself
> have a chance to understand?

Try wikipedia "Halting problem".


        Stefan

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > > Is there anything preventing us from doing such a walk and pre-allocate
> > > all the I/O ranges? I am not familiar with the ACPI code at all, would
> > > you possibly propose a patch doing that?
> > 
> > ACPI AML is probably turing-complete: I'm afraid you are trying to
> > solve the halting problem (-> impossible).
> 
> Can you please translate this into something mere humans like myself
> have a chance to understand?

ACPI AML is turing-complete -- that means it is as powerful any
programming language. It can do arbitrary computation. That means it
is theoretically impossible to analyze its accesses using any program.

Now... may be possible to introduce _some_ ACPI BIOSes, but doing it
would certainly be very complex -- we are talking "put gcc into
kernel" here.

So no, it is not possible to preallocate the ranges.
								Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > > Is there anything preventing us from doing such a walk and pre-allocate
> > > all the I/O ranges? I am not familiar with the ACPI code at all, would
> > > you possibly propose a patch doing that?
> > 
> > ACPI AML is probably turing-complete: I'm afraid you are trying to
> > solve the halting problem (-> impossible).
> 
> Can you please translate this into something mere humans like myself
> have a chance to understand?

ACPI AML is turing-complete -- that means it is as powerful any
programming language. It can do arbitrary computation. That means it
is theoretically impossible to analyze its accesses using any program.

Now... may be possible to introduce _some_ ACPI BIOSes, but doing it
would certainly be very complex -- we are talking "put gcc into
kernel" here.

So no, it is not possible to preallocate the ranges.
								Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

RE: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

In other words, as per my earlier message:

Port addresses can be dynamically generated by the AML code and thus,
there is no way that the ACPI subsystem can statically predict any
addresses that will be accessed by the AML.

Bob


> -----Original Message-----
> From: linux-acpi-owner@vger.kernel.org [mailto:linux-acpi-
> owner@vger.kernel.org] On Behalf Of Pavel Machek
> Sent: Tuesday, March 06, 2007 1:21 PM
> To: Jean Delvare
> Cc: Matthew Garrett; Chuck Ebbert; Rudolf Marek; linux-
> acpi@vger.kernel.org; linux-kernel; lm-sensors@lm-sensors.org
> Subject: Re: [lm-sensors] Could the k8temp driver be interfering with
> ACPI?
> 
> Hi!
> 
> > > > Is there anything preventing us from doing such a walk and pre-
> allocate
> > > > all the I/O ranges? I am not familiar with the ACPI code at all,
> would
> > > > you possibly propose a patch doing that?
> > >
> > > ACPI AML is probably turing-complete: I'm afraid you are trying to
> > > solve the halting problem (-> impossible).
> >
> > Can you please translate this into something mere humans like myself
> > have a chance to understand?
> 
> ACPI AML is turing-complete -- that means it is as powerful any
> programming language. It can do arbitrary computation. That means it
> is theoretically impossible to analyze its accesses using any program.
> 
> Now... may be possible to introduce _some_ ACPI BIOSes, but doing it
> would certainly be very complex -- we are talking "put gcc into
> kernel" here.
> 
> So no, it is not possible to preallocate the ranges.
> 								Pavel
> --
> (english) http://www.livejournal.com/~pavelmachek
> (cesky, pictures)
> http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
> -
> To unsubscribe from this list: send the line "unsubscribe linux-acpi"
in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

In other words, as per my earlier message:

Port addresses can be dynamically generated by the AML code and thus,
there is no way that the ACPI subsystem can statically predict any
addresses that will be accessed by the AML.

Bob


> -----Original Message-----
> From: linux-acpi-owner@vger.kernel.org [mailto:linux-acpi-
> owner@vger.kernel.org] On Behalf Of Pavel Machek
> Sent: Tuesday, March 06, 2007 1:21 PM
> To: Jean Delvare
> Cc: Matthew Garrett; Chuck Ebbert; Rudolf Marek; linux-
> acpi@vger.kernel.org; linux-kernel; lm-sensors@lm-sensors.org
> Subject: Re: [lm-sensors] Could the k8temp driver be interfering with
> ACPI?
> 
> Hi!
> 
> > > > Is there anything preventing us from doing such a walk and pre-
> allocate
> > > > all the I/O ranges? I am not familiar with the ACPI code at all,
> would
> > > > you possibly propose a patch doing that?
> > >
> > > ACPI AML is probably turing-complete: I'm afraid you are trying to
> > > solve the halting problem (-> impossible).
> >
> > Can you please translate this into something mere humans like myself
> > have a chance to understand?
> 
> ACPI AML is turing-complete -- that means it is as powerful any
> programming language. It can do arbitrary computation. That means it
> is theoretically impossible to analyze its accesses using any program.
> 
> Now... may be possible to introduce _some_ ACPI BIOSes, but doing it
> would certainly be very complex -- we are talking "put gcc into
> kernel" here.
> 
> So no, it is not possible to preallocate the ranges.
> 								Pavel
> --
> (english) http://www.livejournal.com/~pavelmachek
> (cesky, pictures)
> http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
> -
> To unsubscribe from this list: send the line "unsubscribe linux-acpi"
in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

In other words, as per my earlier message:

Port addresses can be dynamically generated by the AML code and thus,
there is no way that the ACPI subsystem can statically predict any
addresses that will be accessed by the AML.

Bob


> -----Original Message-----
> From: linux-acpi-owner at vger.kernel.org [mailto:linux-acpi-
> owner at vger.kernel.org] On Behalf Of Pavel Machek
> Sent: Tuesday, March 06, 2007 1:21 PM
> To: Jean Delvare
> Cc: Matthew Garrett; Chuck Ebbert; Rudolf Marek; linux-
> acpi at vger.kernel.org; linux-kernel; lm-sensors at lm-sensors.org
> Subject: Re: [lm-sensors] Could the k8temp driver be interfering with
> ACPI?
> 
> Hi!
> 
> > > > Is there anything preventing us from doing such a walk and pre-
> allocate
> > > > all the I/O ranges? I am not familiar with the ACPI code at all,
> would
> > > > you possibly propose a patch doing that?
> > >
> > > ACPI AML is probably turing-complete: I'm afraid you are trying to
> > > solve the halting problem (-> impossible).
> >
> > Can you please translate this into something mere humans like myself
> > have a chance to understand?
> 
> ACPI AML is turing-complete -- that means it is as powerful any
> programming language. It can do arbitrary computation. That means it
> is theoretically impossible to analyze its accesses using any program.
> 
> Now... may be possible to introduce _some_ ACPI BIOSes, but doing it
> would certainly be very complex -- we are talking "put gcc into
> kernel" here.
> 
> So no, it is not possible to preallocate the ranges.
> 								Pavel
> --
> (english) http://www.livejournal.com/~pavelmachek
> (cesky, pictures)
> http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
> -
> To unsubscribe from this list: send the line "unsubscribe linux-acpi"
in
> the body of a message to majordomo at vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Matthew,

On Fri, 2 Mar 2007 21:46:43 +0000, Matthew Garrett wrote:
> On Fri, Mar 02, 2007 at 10:41:55PM +0100, Jean Delvare wrote:
> 
> > I like the patch, after adding some casts to the printf args it
> > compiles fine. However you print warnings each time a resource has been
> > reserved... without checking if it hasn't been reserved by ACPI itself!
> > My machine looks like this:
> 
> Oops! I'll look into fixing that. Thanks, that's an excellent point...

Here is what I have come up with, by mixing your patch with Rudolf
Marek's one. Again this is only a reporting patch, but now it only
reports real unreserved accesses. I plan to use it for debugging
purposes.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
---
 drivers/acpi/osl.c |   72 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 72 insertions(+)

--- linux-2.6.21-rc2.orig/drivers/acpi/osl.c	2007-03-06 20:59:00.000000000 +0100
+++ linux-2.6.21-rc2/drivers/acpi/osl.c	2007-03-06 21:33:13.000000000 +0100
@@ -35,6 +35,7 @@
 #include <linux/kmod.h>
 #include <linux/delay.h>
 #include <linux/workqueue.h>
+#include <linux/ioport.h>
 #include <linux/nmi.h>
 #include <linux/acpi.h>
 #include <acpi/acpi.h>
@@ -370,6 +371,7 @@ u64 acpi_os_get_timer(void)
 acpi_status acpi_os_read_port(acpi_io_address port, u32 * value, u32 width)
 {
 	u32 dummy;
+	struct resource *conflict, res;
 
 	if (!value)
 		value = &dummy;
@@ -388,6 +390,23 @@ acpi_status acpi_os_read_port(acpi_io_ad
 		BUG();
 	}
 
+	res.flags = IORESOURCE_IO;
+	res.name = "_ACPI Access";
+	res.start = port;
+	res.end = port + width/8 - 1;
+
+	conflict = ____request_resource(&ioport_resource, &res);
+	while (conflict && conflict->child)
+		conflict = ____request_resource(conflict, &res);
+
+	if (conflict && strncmp(conflict->name, "ACPI ", 5)) {
+		printk (KERN_INFO "ACPI read from allocated ioport %lx, value %lx, width %d\n",
+			(unsigned long)port, (unsigned long)(*value), (int)width);
+	}
+
+	if (conflict == NULL)
+		release_resource(&res);
+
 	return AE_OK;
 }
 
@@ -395,6 +414,25 @@ EXPORT_SYMBOL(acpi_os_read_port);
 
 acpi_status acpi_os_write_port(acpi_io_address port, u32 value, u32 width)
 {
+	struct resource *conflict, res;
+
+	res.flags = IORESOURCE_IO;
+	res.name = "_ACPI Access";
+	res.start = port;
+	res.end = port + width/8 - 1;
+
+	conflict = ____request_resource(&ioport_resource, &res);
+	while (conflict && conflict->child)
+		conflict = ____request_resource(conflict, &res);
+
+	if (conflict && strncmp(conflict->name, "ACPI ", 5)) {
+		printk (KERN_INFO "ACPI write to allocated ioport %lx, value %lx, width %d\n",
+			(unsigned long)port, (unsigned long)(value), (int)width);
+	}
+
+	if (conflict == NULL)
+		release_resource(&res);
+
 	switch (width) {
 	case 8:
 		outb(value, port);
@@ -419,6 +457,7 @@ acpi_os_read_memory(acpi_physical_addres
 {
 	u32 dummy;
 	void __iomem *virt_addr;
+	struct resource *conflict, res;
 
 	virt_addr = ioremap(phys_addr, width);
 	if (!value)
@@ -440,6 +479,22 @@ acpi_os_read_memory(acpi_physical_addres
 
 	iounmap(virt_addr);
 
+	res.flags = IORESOURCE_MEM;
+	res.name = "_ACPI Access";
+	res.start = phys_addr;
+	res.end = phys_addr + width/8 - 1;
+
+	conflict = ____request_resource(&iomem_resource, &res);
+	while (conflict && conflict->child)
+		conflict = ____request_resource(conflict, &res);
+
+	if (conflict && strncmp(conflict->name, "ACPI ", 5))
+		pr_info("ACPI read from allocated iomem %lx, value %lx, width %d\n",
+			(unsigned long)phys_addr, (unsigned long)(*value), (int)width);
+
+	if (conflict == NULL)
+		release_resource(&res);
+
 	return AE_OK;
 }
 
@@ -447,6 +502,23 @@ acpi_status
 acpi_os_write_memory(acpi_physical_address phys_addr, u32 value, u32 width)
 {
 	void __iomem *virt_addr;
+	struct resource *conflict, res;
+
+	res.flags = IORESOURCE_MEM;
+	res.name = "_ACPI Access";
+	res.start = phys_addr;
+	res.end = phys_addr + width/8 - 1;
+
+	conflict = ____request_resource(&iomem_resource, &res);
+	while (conflict && conflict->child)
+		conflict = ____request_resource(conflict, &res);
+
+	if (conflict && strncmp(conflict->name, "ACPI ", 5))
+		pr_info("ACPI write to allocated iomem %lx, value %lx, width %d\n",
+			(unsigned long)phys_addr, (unsigned long)value, (int)width);
+
+	if (conflict == NULL)
+		release_resource(&res);
 
 	virt_addr = ioremap(phys_addr, width);
 


-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Matthew,

On Fri, 2 Mar 2007 21:46:43 +0000, Matthew Garrett wrote:
> On Fri, Mar 02, 2007 at 10:41:55PM +0100, Jean Delvare wrote:
> 
> > I like the patch, after adding some casts to the printf args it
> > compiles fine. However you print warnings each time a resource has been
> > reserved... without checking if it hasn't been reserved by ACPI itself!
> > My machine looks like this:
> 
> Oops! I'll look into fixing that. Thanks, that's an excellent point...

Here is what I have come up with, by mixing your patch with Rudolf
Marek's one. Again this is only a reporting patch, but now it only
reports real unreserved accesses. I plan to use it for debugging
purposes.

Signed-off-by: Jean Delvare <khali at linux-fr.org>
---
 drivers/acpi/osl.c |   72 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 72 insertions(+)

--- linux-2.6.21-rc2.orig/drivers/acpi/osl.c	2007-03-06 20:59:00.000000000 +0100
+++ linux-2.6.21-rc2/drivers/acpi/osl.c	2007-03-06 21:33:13.000000000 +0100
@@ -35,6 +35,7 @@
 #include <linux/kmod.h>
 #include <linux/delay.h>
 #include <linux/workqueue.h>
+#include <linux/ioport.h>
 #include <linux/nmi.h>
 #include <linux/acpi.h>
 #include <acpi/acpi.h>
@@ -370,6 +371,7 @@ u64 acpi_os_get_timer(void)
 acpi_status acpi_os_read_port(acpi_io_address port, u32 * value, u32 width)
 {
 	u32 dummy;
+	struct resource *conflict, res;
 
 	if (!value)
 		value = &dummy;
@@ -388,6 +390,23 @@ acpi_status acpi_os_read_port(acpi_io_ad
 		BUG();
 	}
 
+	res.flags = IORESOURCE_IO;
+	res.name = "_ACPI Access";
+	res.start = port;
+	res.end = port + width/8 - 1;
+
+	conflict = ____request_resource(&ioport_resource, &res);
+	while (conflict && conflict->child)
+		conflict = ____request_resource(conflict, &res);
+
+	if (conflict && strncmp(conflict->name, "ACPI ", 5)) {
+		printk (KERN_INFO "ACPI read from allocated ioport %lx, value %lx, width %d\n",
+			(unsigned long)port, (unsigned long)(*value), (int)width);
+	}
+
+	if (conflict = NULL)
+		release_resource(&res);
+
 	return AE_OK;
 }
 
@@ -395,6 +414,25 @@ EXPORT_SYMBOL(acpi_os_read_port);
 
 acpi_status acpi_os_write_port(acpi_io_address port, u32 value, u32 width)
 {
+	struct resource *conflict, res;
+
+	res.flags = IORESOURCE_IO;
+	res.name = "_ACPI Access";
+	res.start = port;
+	res.end = port + width/8 - 1;
+
+	conflict = ____request_resource(&ioport_resource, &res);
+	while (conflict && conflict->child)
+		conflict = ____request_resource(conflict, &res);
+
+	if (conflict && strncmp(conflict->name, "ACPI ", 5)) {
+		printk (KERN_INFO "ACPI write to allocated ioport %lx, value %lx, width %d\n",
+			(unsigned long)port, (unsigned long)(value), (int)width);
+	}
+
+	if (conflict = NULL)
+		release_resource(&res);
+
 	switch (width) {
 	case 8:
 		outb(value, port);
@@ -419,6 +457,7 @@ acpi_os_read_memory(acpi_physical_addres
 {
 	u32 dummy;
 	void __iomem *virt_addr;
+	struct resource *conflict, res;
 
 	virt_addr = ioremap(phys_addr, width);
 	if (!value)
@@ -440,6 +479,22 @@ acpi_os_read_memory(acpi_physical_addres
 
 	iounmap(virt_addr);
 
+	res.flags = IORESOURCE_MEM;
+	res.name = "_ACPI Access";
+	res.start = phys_addr;
+	res.end = phys_addr + width/8 - 1;
+
+	conflict = ____request_resource(&iomem_resource, &res);
+	while (conflict && conflict->child)
+		conflict = ____request_resource(conflict, &res);
+
+	if (conflict && strncmp(conflict->name, "ACPI ", 5))
+		pr_info("ACPI read from allocated iomem %lx, value %lx, width %d\n",
+			(unsigned long)phys_addr, (unsigned long)(*value), (int)width);
+
+	if (conflict = NULL)
+		release_resource(&res);
+
 	return AE_OK;
 }
 
@@ -447,6 +502,23 @@ acpi_status
 acpi_os_write_memory(acpi_physical_address phys_addr, u32 value, u32 width)
 {
 	void __iomem *virt_addr;
+	struct resource *conflict, res;
+
+	res.flags = IORESOURCE_MEM;
+	res.name = "_ACPI Access";
+	res.start = phys_addr;
+	res.end = phys_addr + width/8 - 1;
+
+	conflict = ____request_resource(&iomem_resource, &res);
+	while (conflict && conflict->child)
+		conflict = ____request_resource(conflict, &res);
+
+	if (conflict && strncmp(conflict->name, "ACPI ", 5))
+		pr_info("ACPI write to allocated iomem %lx, value %lx, width %d\n",
+			(unsigned long)phys_addr, (unsigned long)value, (int)width);
+
+	if (conflict = NULL)
+		release_resource(&res);
 
 	virt_addr = ioremap(phys_addr, width);
 


-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> Port (and memory) addresses can be dynamically generated by the AML code
> and thus, there is no way that the ACPI subsystem can statically predict
> any addresses that will be accessed by the AML.

Can you take this as a wishlist item?

It would be nice if next version of acpi specs supported table

'AML / SMM BIOS will access these ports'

...so we can get it correct with acpi4 or something..?

							Pavel

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> Port (and memory) addresses can be dynamically generated by the AML code
> and thus, there is no way that the ACPI subsystem can statically predict
> any addresses that will be accessed by the AML.

Can you take this as a wishlist item?

It would be nice if next version of acpi specs supported table

'AML / SMM BIOS will access these ports'

...so we can get it correct with acpi4 or something..?

							Pavel

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Fri, 9 Mar 2007 07:18:56 +0000, Pavel Machek wrote:
> > Port (and memory) addresses can be dynamically generated by the AML code
> > and thus, there is no way that the ACPI subsystem can statically predict
> > any addresses that will be accessed by the AML.
> 
> Can you take this as a wishlist item?
> 
> It would be nice if next version of acpi specs supported table
> 
> 'AML / SMM BIOS will access these ports'
> 
> ...so we can get it correct with acpi4 or something..?

I can only second Pavel's wish here. This would be highly convenient
for OS developers to at least know which resources are accessed by AML
and SMM. Without this information, we can never be sure that OS-level
code won't conflict with ACPI or SMM.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

On Fri, 9 Mar 2007 07:18:56 +0000, Pavel Machek wrote:
> > Port (and memory) addresses can be dynamically generated by the AML code
> > and thus, there is no way that the ACPI subsystem can statically predict
> > any addresses that will be accessed by the AML.
> 
> Can you take this as a wishlist item?
> 
> It would be nice if next version of acpi specs supported table
> 
> 'AML / SMM BIOS will access these ports'
> 
> ...so we can get it correct with acpi4 or something..?

I can only second Pavel's wish here. This would be highly convenient
for OS developers to at least know which resources are accessed by AML
and SMM. Without this information, we can never be sure that OS-level
code won't conflict with ACPI or SMM.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Alexey Starikovskiy]

Jean Delvare wrote:
> On Fri, 9 Mar 2007 07:18:56 +0000, Pavel Machek wrote:
>   
>>> Port (and memory) addresses can be dynamically generated by the AML code
>>> and thus, there is no way that the ACPI subsystem can statically predict
>>> any addresses that will be accessed by the AML.
>>>       
>> Can you take this as a wishlist item?
>>
>> It would be nice if next version of acpi specs supported table
>>
>> 'AML / SMM BIOS will access these ports'
>>
>> ...so we can get it correct with acpi4 or something..?
>>     
>
> I can only second Pavel's wish here. This would be highly convenient
> for OS developers to at least know which resources are accessed by AML
> and SMM. Without this information, we can never be sure that OS-level
> code won't conflict with ACPI or SMM.
>
>   
BIOS vendors are not required to support latest and greatest ACPI spec. 
So even if some future spec version
will include this ports description, we will still have majority of 
hardware not exporting it...

Regards,
    Alex.

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Alexey Starikovskiy]

Jean Delvare wrote:
> On Fri, 9 Mar 2007 07:18:56 +0000, Pavel Machek wrote:
>   
>>> Port (and memory) addresses can be dynamically generated by the AML code
>>> and thus, there is no way that the ACPI subsystem can statically predict
>>> any addresses that will be accessed by the AML.
>>>       
>> Can you take this as a wishlist item?
>>
>> It would be nice if next version of acpi specs supported table
>>
>> 'AML / SMM BIOS will access these ports'
>>
>> ...so we can get it correct with acpi4 or something..?
>>     
>
> I can only second Pavel's wish here. This would be highly convenient
> for OS developers to at least know which resources are accessed by AML
> and SMM. Without this information, we can never be sure that OS-level
> code won't conflict with ACPI or SMM.
>
>   
BIOS vendors are not required to support latest and greatest ACPI spec. 
So even if some future spec version
will include this ports description, we will still have majority of 
hardware not exporting it...

Regards,
    Alex.

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> >>Can you take this as a wishlist item?
> >>
> >>It would be nice if next version of acpi specs supported table
> >>
> >>'AML / SMM BIOS will access these ports'
> >>
> >>...so we can get it correct with acpi4 or something..?
> >>    
> >
> >I can only second Pavel's wish here. This would be highly convenient
> >for OS developers to at least know which resources are accessed by AML
> >and SMM. Without this information, we can never be sure that OS-level
> >code won't conflict with ACPI or SMM.
> >
> >  
> BIOS vendors are not required to support latest and greatest ACPI spec. 
> So even if some future spec version
> will include this ports description, we will still have majority of 
> hardware not exporting it...

That's okay... vendors are not required to support _ACPI_, but they
mostly do. Can we get the "ports used by BIOS" table to the spec?
								Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> >>Can you take this as a wishlist item?
> >>
> >>It would be nice if next version of acpi specs supported table
> >>
> >>'AML / SMM BIOS will access these ports'
> >>
> >>...so we can get it correct with acpi4 or something..?
> >>    
> >
> >I can only second Pavel's wish here. This would be highly convenient
> >for OS developers to at least know which resources are accessed by AML
> >and SMM. Without this information, we can never be sure that OS-level
> >code won't conflict with ACPI or SMM.
> >
> >  
> BIOS vendors are not required to support latest and greatest ACPI spec. 
> So even if some future spec version
> will include this ports description, we will still have majority of 
> hardware not exporting it...

That's okay... vendors are not required to support _ACPI_, but they
mostly do. Can we get the "ports used by BIOS" table to the spec?
								Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Alexey,

On Fri, 09 Mar 2007 13:39:33 +0300, Alexey Starikovskiy wrote:
> Jean Delvare wrote:
> > I can only second Pavel's wish here. This would be highly convenient
> > for OS developers to at least know which resources are accessed by AML
> > and SMM. Without this information, we can never be sure that OS-level
> > code won't conflict with ACPI or SMM.
>
> BIOS vendors are not required to support latest and greatest ACPI spec. 
> So even if some future spec version
> will include this ports description, we will still have majority of 
> hardware not exporting it...

Your reasoning is amazing. So we should refrain from proposing any
improvement which we aren't certain 100% of the systems will support
tomorrow? Then let's all stay away from our keyboards forever, as the
evolution of computer technology is based on exactly that -
improvements which not all systems implement.

It's friday evening, let's have some more for fun. With a similar
logic, ten years ago we'd have come up with the following conclusions:

The majority of computers have a single CPU, there is no point in
adding SMP support to Linux.

Let's not add a new instruction set in our next CPU family. The
majority of systems will not implement it so it will be useless anyway.

There's no point in supporting PnP in Linux, there are a majority of
legacy ISA cards out there which do not support it anyway!

See my point? Just because not every hardware out there supports a
given standard doesn't make that standard necessarily useless.

Just make the next version of ACPI better than the previous one (not
necessarily a challenge) and everyone will embrace it.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Alexey,

On Fri, 09 Mar 2007 13:39:33 +0300, Alexey Starikovskiy wrote:
> Jean Delvare wrote:
> > I can only second Pavel's wish here. This would be highly convenient
> > for OS developers to at least know which resources are accessed by AML
> > and SMM. Without this information, we can never be sure that OS-level
> > code won't conflict with ACPI or SMM.
>
> BIOS vendors are not required to support latest and greatest ACPI spec. 
> So even if some future spec version
> will include this ports description, we will still have majority of 
> hardware not exporting it...

Your reasoning is amazing. So we should refrain from proposing any
improvement which we aren't certain 100% of the systems will support
tomorrow? Then let's all stay away from our keyboards forever, as the
evolution of computer technology is based on exactly that -
improvements which not all systems implement.

It's friday evening, let's have some more for fun. With a similar
logic, ten years ago we'd have come up with the following conclusions:

The majority of computers have a single CPU, there is no point in
adding SMP support to Linux.

Let's not add a new instruction set in our next CPU family. The
majority of systems will not implement it so it will be useless anyway.

There's no point in supporting PnP in Linux, there are a majority of
legacy ISA cards out there which do not support it anyway!

See my point? Just because not every hardware out there supports a
given standard doesn't make that standard necessarily useless.

Just make the next version of ACPI better than the previous one (not
necessarily a challenge) and everyone will embrace it.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Alexey Starikovskiy]

Jean Delvare wrote:
> Hi Alexey,
>
> On Fri, 09 Mar 2007 13:39:33 +0300, Alexey Starikovskiy wrote:
>   
>> Jean Delvare wrote:
>>     
>>> I can only second Pavel's wish here. This would be highly convenient
>>> for OS developers to at least know which resources are accessed by AML
>>> and SMM. Without this information, we can never be sure that OS-level
>>> code won't conflict with ACPI or SMM.
>>>       
>> BIOS vendors are not required to support latest and greatest ACPI spec. 
>> So even if some future spec version
>> will include this ports description, we will still have majority of 
>> hardware not exporting it...
>>     
>
> Your reasoning is amazing. So we should refrain from proposing any
> improvement which we aren't certain 100% of the systems will support
> tomorrow? Then let's all stay away from our keyboards forever, as the
> evolution of computer technology is based on exactly that -
> improvements which not all systems implement.
>
> It's friday evening, let's have some more for fun. With a similar
> logic, ten years ago we'd have come up with the following conclusions:
>
> The majority of computers have a single CPU, there is no point in
> adding SMP support to Linux.
>
> Let's not add a new instruction set in our next CPU family. The
> majority of systems will not implement it so it will be useless anyway.
>
> There's no point in supporting PnP in Linux, there are a majority of
> legacy ISA cards out there which do not support it anyway!
>
> See my point? Just because not every hardware out there supports a
> given standard doesn't make that standard necessarily useless.
>
> Just make the next version of ACPI better than the previous one (not
> necessarily a challenge) and everyone will embrace it.
>
>   
You get me wrong, I'm not against the proposal, so keep your breath.
I'm just saying that you get old waiting for BIOS vendors to export this 
info, even if it's in spec.

Regards,
Alex.

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Alexey Starikovskiy]

Jean Delvare wrote:
> Hi Alexey,
>
> On Fri, 09 Mar 2007 13:39:33 +0300, Alexey Starikovskiy wrote:
>   
>> Jean Delvare wrote:
>>     
>>> I can only second Pavel's wish here. This would be highly convenient
>>> for OS developers to at least know which resources are accessed by AML
>>> and SMM. Without this information, we can never be sure that OS-level
>>> code won't conflict with ACPI or SMM.
>>>       
>> BIOS vendors are not required to support latest and greatest ACPI spec. 
>> So even if some future spec version
>> will include this ports description, we will still have majority of 
>> hardware not exporting it...
>>     
>
> Your reasoning is amazing. So we should refrain from proposing any
> improvement which we aren't certain 100% of the systems will support
> tomorrow? Then let's all stay away from our keyboards forever, as the
> evolution of computer technology is based on exactly that -
> improvements which not all systems implement.
>
> It's friday evening, let's have some more for fun. With a similar
> logic, ten years ago we'd have come up with the following conclusions:
>
> The majority of computers have a single CPU, there is no point in
> adding SMP support to Linux.
>
> Let's not add a new instruction set in our next CPU family. The
> majority of systems will not implement it so it will be useless anyway.
>
> There's no point in supporting PnP in Linux, there are a majority of
> legacy ISA cards out there which do not support it anyway!
>
> See my point? Just because not every hardware out there supports a
> given standard doesn't make that standard necessarily useless.
>
> Just make the next version of ACPI better than the previous one (not
> necessarily a challenge) and everyone will embrace it.
>
>   
You get me wrong, I'm not against the proposal, so keep your breath.
I'm just saying that you get old waiting for BIOS vendors to export this 
info, even if it's in spec.

Regards,
Alex.

RE: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

Included the Intel ACPI spec representative.

I have heard that Windows is somehow restricting the ports and memory
locations that are accessible via AML; I don't know any of the details.
Also, there are fears of an "AML virus" attacking the machine.

Bob


> -----Original Message-----
> From: Pavel Machek [mailto:pavel@ucw.cz]
> Sent: Thursday, March 08, 2007 11:19 PM
> To: Moore, Robert
> Cc: Jean Delvare; Matthew Garrett; Chuck Ebbert; Rudolf Marek; linux-
> acpi@vger.kernel.org; linux-kernel; lm-sensors@lm-sensors.org
> Subject: Re: [lm-sensors] Could the k8temp driver be interfering with
> ACPI?
> 
> Hi!
> 
> > Port (and memory) addresses can be dynamically generated by the AML
code
> > and thus, there is no way that the ACPI subsystem can statically
predict
> > any addresses that will be accessed by the AML.
> 
> Can you take this as a wishlist item?
> 
> It would be nice if next version of acpi specs supported table
> 
> 'AML / SMM BIOS will access these ports'
> 
> ...so we can get it correct with acpi4 or something..?
> 
> 							Pavel
> 
> --
> (english) http://www.livejournal.com/~pavelmachek
> (cesky, pictures)
> http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

RE: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

Included the Intel ACPI spec representative.

I have heard that Windows is somehow restricting the ports and memory
locations that are accessible via AML; I don't know any of the details.
Also, there are fears of an "AML virus" attacking the machine.

Bob


> -----Original Message-----
> From: Pavel Machek [mailto:pavel@ucw.cz]
> Sent: Thursday, March 08, 2007 11:19 PM
> To: Moore, Robert
> Cc: Jean Delvare; Matthew Garrett; Chuck Ebbert; Rudolf Marek; linux-
> acpi@vger.kernel.org; linux-kernel; lm-sensors@lm-sensors.org
> Subject: Re: [lm-sensors] Could the k8temp driver be interfering with
> ACPI?
> 
> Hi!
> 
> > Port (and memory) addresses can be dynamically generated by the AML
code
> > and thus, there is no way that the ACPI subsystem can statically
predict
> > any addresses that will be accessed by the AML.
> 
> Can you take this as a wishlist item?
> 
> It would be nice if next version of acpi specs supported table
> 
> 'AML / SMM BIOS will access these ports'
> 
> ...so we can get it correct with acpi4 or something..?
> 
> 							Pavel
> 
> --
> (english) http://www.livejournal.com/~pavelmachek
> (cesky, pictures)
> http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

Included the Intel ACPI spec representative.

I have heard that Windows is somehow restricting the ports and memory
locations that are accessible via AML; I don't know any of the details.
Also, there are fears of an "AML virus" attacking the machine.

Bob


> -----Original Message-----
> From: Pavel Machek [mailto:pavel at ucw.cz]
> Sent: Thursday, March 08, 2007 11:19 PM
> To: Moore, Robert
> Cc: Jean Delvare; Matthew Garrett; Chuck Ebbert; Rudolf Marek; linux-
> acpi at vger.kernel.org; linux-kernel; lm-sensors at lm-sensors.org
> Subject: Re: [lm-sensors] Could the k8temp driver be interfering with
> ACPI?
> 
> Hi!
> 
> > Port (and memory) addresses can be dynamically generated by the AML
code
> > and thus, there is no way that the ACPI subsystem can statically
predict
> > any addresses that will be accessed by the AML.
> 
> Can you take this as a wishlist item?
> 
> It would be nice if next version of acpi specs supported table
> 
> 'AML / SMM BIOS will access these ports'
> 
> ...so we can get it correct with acpi4 or something..?
> 
> 							Pavel
> 
> --
> (english) http://www.livejournal.com/~pavelmachek
> (cesky, pictures)
> http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

RE: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

No ACPI discussion can be complete without mentioning Microsoft and
Microsoft compatibility -- Windows does not fully support ACPI 2.0 to
this day, even though it was released in the year 2000, and ACPI 3.0 has
been out since 2004.

> -----Original Message-----
> From: Alexey Starikovskiy
[mailto:alexey.y.starikovskiy@linux.intel.com]
> Sent: Friday, March 09, 2007 9:35 AM
> To: Jean Delvare
> Cc: Pavel Machek; Moore, Robert; Matthew Garrett; Chuck Ebbert; Rudolf
> Marek; linux-acpi@vger.kernel.org; linux-kernel;
lm-sensors@lm-sensors.org
> Subject: Re: [lm-sensors] Could the k8temp driver be interfering with
> ACPI?
> 
> Jean Delvare wrote:
> > Hi Alexey,
> >
> > On Fri, 09 Mar 2007 13:39:33 +0300, Alexey Starikovskiy wrote:
> >
> >> Jean Delvare wrote:
> >>
> >>> I can only second Pavel's wish here. This would be highly
convenient
> >>> for OS developers to at least know which resources are accessed by
AML
> >>> and SMM. Without this information, we can never be sure that
OS-level
> >>> code won't conflict with ACPI or SMM.
> >>>
> >> BIOS vendors are not required to support latest and greatest ACPI
spec.
> >> So even if some future spec version
> >> will include this ports description, we will still have majority of
> >> hardware not exporting it...
> >>
> >
> > Your reasoning is amazing. So we should refrain from proposing any
> > improvement which we aren't certain 100% of the systems will support
> > tomorrow? Then let's all stay away from our keyboards forever, as
the
> > evolution of computer technology is based on exactly that -
> > improvements which not all systems implement.
> >
> > It's friday evening, let's have some more for fun. With a similar
> > logic, ten years ago we'd have come up with the following
conclusions:
> >
> > The majority of computers have a single CPU, there is no point in
> > adding SMP support to Linux.
> >
> > Let's not add a new instruction set in our next CPU family. The
> > majority of systems will not implement it so it will be useless
anyway.
> >
> > There's no point in supporting PnP in Linux, there are a majority of
> > legacy ISA cards out there which do not support it anyway!
> >
> > See my point? Just because not every hardware out there supports a
> > given standard doesn't make that standard necessarily useless.
> >
> > Just make the next version of ACPI better than the previous one (not
> > necessarily a challenge) and everyone will embrace it.
> >
> >
> You get me wrong, I'm not against the proposal, so keep your breath.
> I'm just saying that you get old waiting for BIOS vendors to export
this
> info, even if it's in spec.
> 
> Regards,
> Alex.

RE: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

No ACPI discussion can be complete without mentioning Microsoft and
Microsoft compatibility -- Windows does not fully support ACPI 2.0 to
this day, even though it was released in the year 2000, and ACPI 3.0 has
been out since 2004.

> -----Original Message-----
> From: Alexey Starikovskiy
[mailto:alexey.y.starikovskiy@linux.intel.com]
> Sent: Friday, March 09, 2007 9:35 AM
> To: Jean Delvare
> Cc: Pavel Machek; Moore, Robert; Matthew Garrett; Chuck Ebbert; Rudolf
> Marek; linux-acpi@vger.kernel.org; linux-kernel;
lm-sensors@lm-sensors.org
> Subject: Re: [lm-sensors] Could the k8temp driver be interfering with
> ACPI?
> 
> Jean Delvare wrote:
> > Hi Alexey,
> >
> > On Fri, 09 Mar 2007 13:39:33 +0300, Alexey Starikovskiy wrote:
> >
> >> Jean Delvare wrote:
> >>
> >>> I can only second Pavel's wish here. This would be highly
convenient
> >>> for OS developers to at least know which resources are accessed by
AML
> >>> and SMM. Without this information, we can never be sure that
OS-level
> >>> code won't conflict with ACPI or SMM.
> >>>
> >> BIOS vendors are not required to support latest and greatest ACPI
spec.
> >> So even if some future spec version
> >> will include this ports description, we will still have majority of
> >> hardware not exporting it...
> >>
> >
> > Your reasoning is amazing. So we should refrain from proposing any
> > improvement which we aren't certain 100% of the systems will support
> > tomorrow? Then let's all stay away from our keyboards forever, as
the
> > evolution of computer technology is based on exactly that -
> > improvements which not all systems implement.
> >
> > It's friday evening, let's have some more for fun. With a similar
> > logic, ten years ago we'd have come up with the following
conclusions:
> >
> > The majority of computers have a single CPU, there is no point in
> > adding SMP support to Linux.
> >
> > Let's not add a new instruction set in our next CPU family. The
> > majority of systems will not implement it so it will be useless
anyway.
> >
> > There's no point in supporting PnP in Linux, there are a majority of
> > legacy ISA cards out there which do not support it anyway!
> >
> > See my point? Just because not every hardware out there supports a
> > given standard doesn't make that standard necessarily useless.
> >
> > Just make the next version of ACPI better than the previous one (not
> > necessarily a challenge) and everyone will embrace it.
> >
> >
> You get me wrong, I'm not against the proposal, so keep your breath.
> I'm just saying that you get old waiting for BIOS vendors to export
this
> info, even if it's in spec.
> 
> Regards,
> Alex.

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

No ACPI discussion can be complete without mentioning Microsoft and
Microsoft compatibility -- Windows does not fully support ACPI 2.0 to
this day, even though it was released in the year 2000, and ACPI 3.0 has
been out since 2004.

> -----Original Message-----
> From: Alexey Starikovskiy
[mailto:alexey.y.starikovskiy at linux.intel.com]
> Sent: Friday, March 09, 2007 9:35 AM
> To: Jean Delvare
> Cc: Pavel Machek; Moore, Robert; Matthew Garrett; Chuck Ebbert; Rudolf
> Marek; linux-acpi at vger.kernel.org; linux-kernel;
lm-sensors at lm-sensors.org
> Subject: Re: [lm-sensors] Could the k8temp driver be interfering with
> ACPI?
> 
> Jean Delvare wrote:
> > Hi Alexey,
> >
> > On Fri, 09 Mar 2007 13:39:33 +0300, Alexey Starikovskiy wrote:
> >
> >> Jean Delvare wrote:
> >>
> >>> I can only second Pavel's wish here. This would be highly
convenient
> >>> for OS developers to at least know which resources are accessed by
AML
> >>> and SMM. Without this information, we can never be sure that
OS-level
> >>> code won't conflict with ACPI or SMM.
> >>>
> >> BIOS vendors are not required to support latest and greatest ACPI
spec.
> >> So even if some future spec version
> >> will include this ports description, we will still have majority of
> >> hardware not exporting it...
> >>
> >
> > Your reasoning is amazing. So we should refrain from proposing any
> > improvement which we aren't certain 100% of the systems will support
> > tomorrow? Then let's all stay away from our keyboards forever, as
the
> > evolution of computer technology is based on exactly that -
> > improvements which not all systems implement.
> >
> > It's friday evening, let's have some more for fun. With a similar
> > logic, ten years ago we'd have come up with the following
conclusions:
> >
> > The majority of computers have a single CPU, there is no point in
> > adding SMP support to Linux.
> >
> > Let's not add a new instruction set in our next CPU family. The
> > majority of systems will not implement it so it will be useless
anyway.
> >
> > There's no point in supporting PnP in Linux, there are a majority of
> > legacy ISA cards out there which do not support it anyway!
> >
> > See my point? Just because not every hardware out there supports a
> > given standard doesn't make that standard necessarily useless.
> >
> > Just make the next version of ACPI better than the previous one (not
> > necessarily a challenge) and everyone will embrace it.
> >
> >
> You get me wrong, I'm not against the proposal, so keep your breath.
> I'm just saying that you get old waiting for BIOS vendors to export
this
> info, even if it's in spec.
> 
> Regards,
> Alex.

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[richardvoigt]

On 3/3/07, Jean Delvare <khali@linux-fr.org> wrote:
> Hi Matthew,
>
> On Fri, 2 Mar 2007 21:12:51 +0000, Matthew Garrett wrote:
> > On Fri, Mar 02, 2007 at 10:04:54PM +0100, Jean Delvare wrote:
> > > It might be more elegant but it won't work. We don't want to prevent
> > > ACPI from accessing these I/O ports. If we need to choose only one
> > > "driver" accessing the I/O port, it must be acpi, at leat for now,
> > > despite the fact that acpi provides very weak hardware monitoring
> > > capabilities compared to the specific drivers.
> >
> > Assuming arbitration of access, what's the problem with having two
> > drivers accessing the same hardware? Do these chips generally have any
> > significant internal state other than trip points and the like?
>
> The "assuming arbitration of access" is the key part of your
> sentence ;) The problem is that currently no arbitration is done. If it
> was done, then state would probably not be a problem. Most hardware
> monitoring drivers don't assume any state is preserved between
> accesses, and those which do can easily be changed not to. The ACPI
> side is another story though, I guess we can't assume anything about
> the AML code's assumption on states, as it differs from one machine to
> the next. But we can try to be cooperative and restore the sensible
> registers (e.g. bank selector) in the same state we found them.

I recall that one of the stated drawbacks of a lock was that no ACPI
code could execute while the hwmon driver was doing its fairly lengthy
conversation with the hardware.

It seems that using transactional concepts here would solve that
problem.  For example, the hwmon driver issues a start transaction
request.  The first write request to any given location (bank register
for example) causes the previous memory value to be saved.  Then,
instead of locking AML out, AML is allowed to execute, but any access
to the memory/port ranges reserved by the driver (when the transaction
is set up) cause the hwmon transaction to be rolled back so the AML
sees the expected state.  Then AML proceeds as usual.  When hwmon
tries additional operations, they fail with some "transaction
interrupted" error message, indicating to the hwmon driver to start
over.

The only issue with this that I can see, is that if AML isn['t
executed atomically wrt hwmon, then knowing when it is safe for hwmon
to retry is going to be difficult.

This probably requires changes to every hwmon driver, but they can be
updated piecemeal, starting with the ones most commonly found in
notebooks, where ACPI is most important.

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[richardvoigt at gmail.com]

On 3/3/07, Jean Delvare <khali at linux-fr.org> wrote:
> Hi Matthew,
>
> On Fri, 2 Mar 2007 21:12:51 +0000, Matthew Garrett wrote:
> > On Fri, Mar 02, 2007 at 10:04:54PM +0100, Jean Delvare wrote:
> > > It might be more elegant but it won't work. We don't want to prevent
> > > ACPI from accessing these I/O ports. If we need to choose only one
> > > "driver" accessing the I/O port, it must be acpi, at leat for now,
> > > despite the fact that acpi provides very weak hardware monitoring
> > > capabilities compared to the specific drivers.
> >
> > Assuming arbitration of access, what's the problem with having two
> > drivers accessing the same hardware? Do these chips generally have any
> > significant internal state other than trip points and the like?
>
> The "assuming arbitration of access" is the key part of your
> sentence ;) The problem is that currently no arbitration is done. If it
> was done, then state would probably not be a problem. Most hardware
> monitoring drivers don't assume any state is preserved between
> accesses, and those which do can easily be changed not to. The ACPI
> side is another story though, I guess we can't assume anything about
> the AML code's assumption on states, as it differs from one machine to
> the next. But we can try to be cooperative and restore the sensible
> registers (e.g. bank selector) in the same state we found them.

I recall that one of the stated drawbacks of a lock was that no ACPI
code could execute while the hwmon driver was doing its fairly lengthy
conversation with the hardware.

It seems that using transactional concepts here would solve that
problem.  For example, the hwmon driver issues a start transaction
request.  The first write request to any given location (bank register
for example) causes the previous memory value to be saved.  Then,
instead of locking AML out, AML is allowed to execute, but any access
to the memory/port ranges reserved by the driver (when the transaction
is set up) cause the hwmon transaction to be rolled back so the AML
sees the expected state.  Then AML proceeds as usual.  When hwmon
tries additional operations, they fail with some "transaction
interrupted" error message, indicating to the hwmon driver to start
over.

The only issue with this that I can see, is that if AML isn['t
executed atomically wrt hwmon, then knowing when it is safe for hwmon
to retry is going to be difficult.

This probably requires changes to every hwmon driver, but they can be
updated piecemeal, starting with the ones most commonly found in
notebooks, where ACPI is most important.

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Richard,

On Sun, 18 Mar 2007 14:36:08 -0500, Richard Voigt wrote:
> On 3/3/07, Jean Delvare wrote:
> > On Fri, 2 Mar 2007 21:12:51 +0000, Matthew Garrett wrote:
> > > Assuming arbitration of access, what's the problem with having two
> > > drivers accessing the same hardware? Do these chips generally have any
> > > significant internal state other than trip points and the like?
> >
> > The "assuming arbitration of access" is the key part of your
> > sentence ;) The problem is that currently no arbitration is done. If it
> > was done, then state would probably not be a problem. Most hardware
> > monitoring drivers don't assume any state is preserved between
> > accesses, and those which do can easily be changed not to. The ACPI
> > side is another story though, I guess we can't assume anything about
> > the AML code's assumption on states, as it differs from one machine to
> > the next. But we can try to be cooperative and restore the sensible
> > registers (e.g. bank selector) in the same state we found them.
> 
> I recall that one of the stated drawbacks of a lock was that no ACPI
> code could execute while the hwmon driver was doing its fairly lengthy
> conversation with the hardware.
> 
> It seems that using transactional concepts here would solve that
> problem.  For example, the hwmon driver issues a start transaction
> request.  The first write request to any given location (bank register
> for example) causes the previous memory value to be saved.  Then,
> instead of locking AML out, AML is allowed to execute, but any access
> to the memory/port ranges reserved by the driver (when the transaction
> is set up) cause the hwmon transaction to be rolled back so the AML
> sees the expected state.  Then AML proceeds as usual.  When hwmon
> tries additional operations, they fail with some "transaction
> interrupted" error message, indicating to the hwmon driver to start
> over.
>
> The only issue with this that I can see, is that if AML isn't
> executed atomically wrt hwmon, then knowing when it is safe for hwmon
> to retry is going to be difficult.

No. We're not rolling back anything, it's totally unrealistic. These
are device drivers we're talking about here, not a database. The I/O
accesses done by the hardware monitoring drivers are not that long, so
AML gets to wait for them to be finished, and that's it. There is no
valid reason to give the priority to AML over regular device drivers.
Rudolf Marek's approach sounds much better.

> This probably requires changes to every hwmon driver, but they can be
> updated piecemeal, starting with the ones most commonly found in
> notebooks, where ACPI is most important.

Most notebooks don't expose their hardware monitoring chip at all.
Those which do use SMBus devices in majority, where I/O forwarding is
going to be difficult, as it needs to be done at the SMBus controller
level, not the hardware monitoring device level. I want to get my hands
on such a laptop first though, as I need to see what exacly ACPI is
doing before I can think of a solution.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Richard,

On Sun, 18 Mar 2007 14:36:08 -0500, Richard Voigt wrote:
> On 3/3/07, Jean Delvare wrote:
> > On Fri, 2 Mar 2007 21:12:51 +0000, Matthew Garrett wrote:
> > > Assuming arbitration of access, what's the problem with having two
> > > drivers accessing the same hardware? Do these chips generally have any
> > > significant internal state other than trip points and the like?
> >
> > The "assuming arbitration of access" is the key part of your
> > sentence ;) The problem is that currently no arbitration is done. If it
> > was done, then state would probably not be a problem. Most hardware
> > monitoring drivers don't assume any state is preserved between
> > accesses, and those which do can easily be changed not to. The ACPI
> > side is another story though, I guess we can't assume anything about
> > the AML code's assumption on states, as it differs from one machine to
> > the next. But we can try to be cooperative and restore the sensible
> > registers (e.g. bank selector) in the same state we found them.
> 
> I recall that one of the stated drawbacks of a lock was that no ACPI
> code could execute while the hwmon driver was doing its fairly lengthy
> conversation with the hardware.
> 
> It seems that using transactional concepts here would solve that
> problem.  For example, the hwmon driver issues a start transaction
> request.  The first write request to any given location (bank register
> for example) causes the previous memory value to be saved.  Then,
> instead of locking AML out, AML is allowed to execute, but any access
> to the memory/port ranges reserved by the driver (when the transaction
> is set up) cause the hwmon transaction to be rolled back so the AML
> sees the expected state.  Then AML proceeds as usual.  When hwmon
> tries additional operations, they fail with some "transaction
> interrupted" error message, indicating to the hwmon driver to start
> over.
>
> The only issue with this that I can see, is that if AML isn't
> executed atomically wrt hwmon, then knowing when it is safe for hwmon
> to retry is going to be difficult.

No. We're not rolling back anything, it's totally unrealistic. These
are device drivers we're talking about here, not a database. The I/O
accesses done by the hardware monitoring drivers are not that long, so
AML gets to wait for them to be finished, and that's it. There is no
valid reason to give the priority to AML over regular device drivers.
Rudolf Marek's approach sounds much better.

> This probably requires changes to every hwmon driver, but they can be
> updated piecemeal, starting with the ones most commonly found in
> notebooks, where ACPI is most important.

Most notebooks don't expose their hardware monitoring chip at all.
Those which do use SMBus devices in majority, where I/O forwarding is
going to be difficult, as it needs to be done at the SMBus controller
level, not the hardware monitoring device level. I want to get my hands
on such a laptop first though, as I need to see what exacly ACPI is
doing before I can think of a solution.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > Actually for the acpi stuff... we might wrap ACPI interpretter with a
> > semaphore that needs to be taken before starting any AML code. Then
> > just make sure sensors take same semaphore?
> 
> I like the idea, it should work as long as we are guaranteed that all
> the hwmon device accesses happen in the AML code? I'm not familiar with
> ACPI, so you tell me.
> 
> In practice it's rather the SMBus drivers which will need to take the
> lock, as this is what the AML code will access (for SMBus-based
> hardware monitoring drivers.) For non-SMBus based hardware monitoring
> drivers, indeed the driver itself should take it. We will have to pay
> attention to deadlocks on systems with multiple hardware monitoring
> chips and/or SMBus channels.
> 
> Can you please provide a patch implementing your proposal in acpi? Then
> I could implement the i2c/hwmon part on some selected drivers and start
> testing real world scenarii.

I'm sorry, but I do not have time for a patch.... and I'm not really
acpi expert, anyway. Ask Len?
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > Actually for the acpi stuff... we might wrap ACPI interpretter with a
> > semaphore that needs to be taken before starting any AML code. Then
> > just make sure sensors take same semaphore?
> 
> I like the idea, it should work as long as we are guaranteed that all
> the hwmon device accesses happen in the AML code? I'm not familiar with
> ACPI, so you tell me.
> 
> In practice it's rather the SMBus drivers which will need to take the
> lock, as this is what the AML code will access (for SMBus-based
> hardware monitoring drivers.) For non-SMBus based hardware monitoring
> drivers, indeed the driver itself should take it. We will have to pay
> attention to deadlocks on systems with multiple hardware monitoring
> chips and/or SMBus channels.
> 
> Can you please provide a patch implementing your proposal in acpi? Then
> I could implement the i2c/hwmon part on some selected drivers and start
> testing real world scenarii.

I'm sorry, but I do not have time for a patch.... and I'm not really
acpi expert, anyway. Ask Len?
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Pavel, Len,

On Sun, 1 Apr 2007 15:39:52 +0000, Pavel Machek wrote:
> > > Actually for the acpi stuff... we might wrap ACPI interpretter with a
> > > semaphore that needs to be taken before starting any AML code. Then
> > > just make sure sensors take same semaphore?
> > 
> > I like the idea, it should work as long as we are guaranteed that all
> > the hwmon device accesses happen in the AML code? I'm not familiar with
> > ACPI, so you tell me.
> > 
> > In practice it's rather the SMBus drivers which will need to take the
> > lock, as this is what the AML code will access (for SMBus-based
> > hardware monitoring drivers.) For non-SMBus based hardware monitoring
> > drivers, indeed the driver itself should take it. We will have to pay
> > attention to deadlocks on systems with multiple hardware monitoring
> > chips and/or SMBus channels.
> > 
> > Can you please provide a patch implementing your proposal in acpi? Then
> > I could implement the i2c/hwmon part on some selected drivers and start
> > testing real world scenarii.
> 
> I'm sorry, but I do not have time for a patch.... and I'm not really
> acpi expert, anyway. Ask Len?

Meanwhile Alexey Starikovskiy pointed me to the
acpi_ex_enter/exit_interpreter functions, which take an ACPI mutex
(ACPI_MTX_INTERPRETER). As the mutex already exists, it sounds more
efficient to just reuse it rather than introducing a new one. I made an
experiment with the f71805f driver on a machine where ACPI is accessing
the F71805F chip, and it appears to work fine; patch at the end of this
post is someone wants to take a look and/or comment.

Looking at the comment before acpi_ex_exit_interpreter raises two
questions though:

>  * Cases where the interpreter is unlocked:
>  *      1) Completion of the execution of a control method
>  *      2) Method blocked on a Sleep() AML opcode
>  *      3) Method blocked on an Acquire() AML opcode
>  *      4) Method blocked on a Wait() AML opcode
>  *      5) Method blocked to acquire the global lock
>  *      6) Method blocked to execute a serialized control method that is
>  *          already executing
>  *      7) About to invoke a user-installed opregion handler

1* This suggests that the mutex could be released by the AML
interpreter in the middle of an SMBus transaction. If so, and if it
happens in practice, this means that we unfortunately cannot use this
mutex to reliably protect the SMBus drivers from concurrent accesses.
This even suggests that it's simply not possible to have a mutex we
take at the beginning when entering the AML interpreter and we release
when leaving the AML interpreter, as it looks like ACPI itself allows
interlaced execution of AML functions. Len, is this true?

What is the purpose of the ACPI_MTX_INTERPRETER mutex in the first
place, given that it seems it will be released on numerous occasions?
Is it to prevent concurrent AML execution while still allowing
interlaced execution?

2* What are "user-installed opregion handlers"? Are they something that
could help solve the ACPI vs. other drivers problem?

Thanks.

---
 drivers/acpi/utilities/utmutex.c |    2 ++
 drivers/hwmon/f71805f.c          |   32 +++++++++++++++++++++++++++++++-
 2 files changed, 33 insertions(+), 1 deletion(-)

--- linux-2.6.21-rc5.orig/drivers/acpi/utilities/utmutex.c	2007-02-21 08:34:20.000000000 +0100
+++ linux-2.6.21-rc5/drivers/acpi/utilities/utmutex.c	2007-04-02 15:48:41.000000000 +0200
@@ -265,6 +265,7 @@ acpi_status acpi_ut_acquire_mutex(acpi_m
 
 	return (status);
 }
+EXPORT_SYMBOL_GPL(acpi_ut_acquire_mutex);
 
 /*******************************************************************************
  *
@@ -339,3 +340,4 @@ acpi_status acpi_ut_release_mutex(acpi_m
 	acpi_os_release_mutex(acpi_gbl_mutex_info[mutex_id].mutex);
 	return (AE_OK);
 }
+EXPORT_SYMBOL_GPL(acpi_ut_release_mutex);
--- linux-2.6.21-rc5.orig/drivers/hwmon/f71805f.c	2007-04-02 15:20:46.000000000 +0200
+++ linux-2.6.21-rc5/drivers/hwmon/f71805f.c	2007-04-02 17:15:46.000000000 +0200
@@ -37,6 +37,10 @@
 #include <linux/sysfs.h>
 #include <asm/io.h>
 
+#ifdef CONFIG_ACPI
+#include <acpi/acpi.h>
+#endif
+
 static struct platform_device *pdev;
 
 #define DRVNAME "f71805f"
@@ -273,15 +277,29 @@ static inline u8 temp_to_reg(long val)
 /* Must be called with data->update_lock held, except during initialization */
 static u8 f71805f_read8(struct f71805f_data *data, u8 reg)
 {
+	u8  val;
+#ifdef CONFIG_ACPI
+	acpi_ut_acquire_mutex(ACPI_MTX_INTERPRETER);
+#endif
 	outb(reg, data->addr + ADDR_REG_OFFSET);
-	return inb(data->addr + DATA_REG_OFFSET);
+	val = inb(data->addr + DATA_REG_OFFSET);
+#ifdef CONFIG_ACPI
+	acpi_ut_release_mutex(ACPI_MTX_INTERPRETER);
+#endif
+	return val;
 }
 
 /* Must be called with data->update_lock held, except during initialization */
 static void f71805f_write8(struct f71805f_data *data, u8 reg, u8 val)
 {
+#ifdef CONFIG_ACPI
+	acpi_ut_acquire_mutex(ACPI_MTX_INTERPRETER);
+#endif
 	outb(reg, data->addr + ADDR_REG_OFFSET);
 	outb(val, data->addr + DATA_REG_OFFSET);
+#ifdef CONFIG_ACPI
+	acpi_ut_release_mutex(ACPI_MTX_INTERPRETER);
+#endif
 }
 
 /* It is important to read the MSB first, because doing so latches the
@@ -291,10 +309,16 @@ static u16 f71805f_read16(struct f71805f
 {
 	u16 val;
 
+#ifdef CONFIG_ACPI
+	acpi_ut_acquire_mutex(ACPI_MTX_INTERPRETER);
+#endif
 	outb(reg, data->addr + ADDR_REG_OFFSET);
 	val = inb(data->addr + DATA_REG_OFFSET) << 8;
 	outb(++reg, data->addr + ADDR_REG_OFFSET);
 	val |= inb(data->addr + DATA_REG_OFFSET);
+#ifdef CONFIG_ACPI
+	acpi_ut_release_mutex(ACPI_MTX_INTERPRETER);
+#endif
 
 	return val;
 }
@@ -302,10 +326,16 @@ static u16 f71805f_read16(struct f71805f
 /* Must be called with data->update_lock held, except during initialization */
 static void f71805f_write16(struct f71805f_data *data, u8 reg, u16 val)
 {
+#ifdef CONFIG_ACPI
+	acpi_ut_acquire_mutex(ACPI_MTX_INTERPRETER);
+#endif
 	outb(reg, data->addr + ADDR_REG_OFFSET);
 	outb(val >> 8, data->addr + DATA_REG_OFFSET);
 	outb(++reg, data->addr + ADDR_REG_OFFSET);
 	outb(val & 0xff, data->addr + DATA_REG_OFFSET);
+#ifdef CONFIG_ACPI
+	acpi_ut_release_mutex(ACPI_MTX_INTERPRETER);
+#endif
 }
 
 static struct f71805f_data *f71805f_update_device(struct device *dev)


-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Pavel, Len,

On Sun, 1 Apr 2007 15:39:52 +0000, Pavel Machek wrote:
> > > Actually for the acpi stuff... we might wrap ACPI interpretter with a
> > > semaphore that needs to be taken before starting any AML code. Then
> > > just make sure sensors take same semaphore?
> > 
> > I like the idea, it should work as long as we are guaranteed that all
> > the hwmon device accesses happen in the AML code? I'm not familiar with
> > ACPI, so you tell me.
> > 
> > In practice it's rather the SMBus drivers which will need to take the
> > lock, as this is what the AML code will access (for SMBus-based
> > hardware monitoring drivers.) For non-SMBus based hardware monitoring
> > drivers, indeed the driver itself should take it. We will have to pay
> > attention to deadlocks on systems with multiple hardware monitoring
> > chips and/or SMBus channels.
> > 
> > Can you please provide a patch implementing your proposal in acpi? Then
> > I could implement the i2c/hwmon part on some selected drivers and start
> > testing real world scenarii.
> 
> I'm sorry, but I do not have time for a patch.... and I'm not really
> acpi expert, anyway. Ask Len?

Meanwhile Alexey Starikovskiy pointed me to the
acpi_ex_enter/exit_interpreter functions, which take an ACPI mutex
(ACPI_MTX_INTERPRETER). As the mutex already exists, it sounds more
efficient to just reuse it rather than introducing a new one. I made an
experiment with the f71805f driver on a machine where ACPI is accessing
the F71805F chip, and it appears to work fine; patch at the end of this
post is someone wants to take a look and/or comment.

Looking at the comment before acpi_ex_exit_interpreter raises two
questions though:

>  * Cases where the interpreter is unlocked:
>  *      1) Completion of the execution of a control method
>  *      2) Method blocked on a Sleep() AML opcode
>  *      3) Method blocked on an Acquire() AML opcode
>  *      4) Method blocked on a Wait() AML opcode
>  *      5) Method blocked to acquire the global lock
>  *      6) Method blocked to execute a serialized control method that is
>  *          already executing
>  *      7) About to invoke a user-installed opregion handler

1* This suggests that the mutex could be released by the AML
interpreter in the middle of an SMBus transaction. If so, and if it
happens in practice, this means that we unfortunately cannot use this
mutex to reliably protect the SMBus drivers from concurrent accesses.
This even suggests that it's simply not possible to have a mutex we
take at the beginning when entering the AML interpreter and we release
when leaving the AML interpreter, as it looks like ACPI itself allows
interlaced execution of AML functions. Len, is this true?

What is the purpose of the ACPI_MTX_INTERPRETER mutex in the first
place, given that it seems it will be released on numerous occasions?
Is it to prevent concurrent AML execution while still allowing
interlaced execution?

2* What are "user-installed opregion handlers"? Are they something that
could help solve the ACPI vs. other drivers problem?

Thanks.

---
 drivers/acpi/utilities/utmutex.c |    2 ++
 drivers/hwmon/f71805f.c          |   32 +++++++++++++++++++++++++++++++-
 2 files changed, 33 insertions(+), 1 deletion(-)

--- linux-2.6.21-rc5.orig/drivers/acpi/utilities/utmutex.c	2007-02-21 08:34:20.000000000 +0100
+++ linux-2.6.21-rc5/drivers/acpi/utilities/utmutex.c	2007-04-02 15:48:41.000000000 +0200
@@ -265,6 +265,7 @@ acpi_status acpi_ut_acquire_mutex(acpi_m
 
 	return (status);
 }
+EXPORT_SYMBOL_GPL(acpi_ut_acquire_mutex);
 
 /*******************************************************************************
  *
@@ -339,3 +340,4 @@ acpi_status acpi_ut_release_mutex(acpi_m
 	acpi_os_release_mutex(acpi_gbl_mutex_info[mutex_id].mutex);
 	return (AE_OK);
 }
+EXPORT_SYMBOL_GPL(acpi_ut_release_mutex);
--- linux-2.6.21-rc5.orig/drivers/hwmon/f71805f.c	2007-04-02 15:20:46.000000000 +0200
+++ linux-2.6.21-rc5/drivers/hwmon/f71805f.c	2007-04-02 17:15:46.000000000 +0200
@@ -37,6 +37,10 @@
 #include <linux/sysfs.h>
 #include <asm/io.h>
 
+#ifdef CONFIG_ACPI
+#include <acpi/acpi.h>
+#endif
+
 static struct platform_device *pdev;
 
 #define DRVNAME "f71805f"
@@ -273,15 +277,29 @@ static inline u8 temp_to_reg(long val)
 /* Must be called with data->update_lock held, except during initialization */
 static u8 f71805f_read8(struct f71805f_data *data, u8 reg)
 {
+	u8  val;
+#ifdef CONFIG_ACPI
+	acpi_ut_acquire_mutex(ACPI_MTX_INTERPRETER);
+#endif
 	outb(reg, data->addr + ADDR_REG_OFFSET);
-	return inb(data->addr + DATA_REG_OFFSET);
+	val = inb(data->addr + DATA_REG_OFFSET);
+#ifdef CONFIG_ACPI
+	acpi_ut_release_mutex(ACPI_MTX_INTERPRETER);
+#endif
+	return val;
 }
 
 /* Must be called with data->update_lock held, except during initialization */
 static void f71805f_write8(struct f71805f_data *data, u8 reg, u8 val)
 {
+#ifdef CONFIG_ACPI
+	acpi_ut_acquire_mutex(ACPI_MTX_INTERPRETER);
+#endif
 	outb(reg, data->addr + ADDR_REG_OFFSET);
 	outb(val, data->addr + DATA_REG_OFFSET);
+#ifdef CONFIG_ACPI
+	acpi_ut_release_mutex(ACPI_MTX_INTERPRETER);
+#endif
 }
 
 /* It is important to read the MSB first, because doing so latches the
@@ -291,10 +309,16 @@ static u16 f71805f_read16(struct f71805f
 {
 	u16 val;
 
+#ifdef CONFIG_ACPI
+	acpi_ut_acquire_mutex(ACPI_MTX_INTERPRETER);
+#endif
 	outb(reg, data->addr + ADDR_REG_OFFSET);
 	val = inb(data->addr + DATA_REG_OFFSET) << 8;
 	outb(++reg, data->addr + ADDR_REG_OFFSET);
 	val |= inb(data->addr + DATA_REG_OFFSET);
+#ifdef CONFIG_ACPI
+	acpi_ut_release_mutex(ACPI_MTX_INTERPRETER);
+#endif
 
 	return val;
 }
@@ -302,10 +326,16 @@ static u16 f71805f_read16(struct f71805f
 /* Must be called with data->update_lock held, except during initialization */
 static void f71805f_write16(struct f71805f_data *data, u8 reg, u16 val)
 {
+#ifdef CONFIG_ACPI
+	acpi_ut_acquire_mutex(ACPI_MTX_INTERPRETER);
+#endif
 	outb(reg, data->addr + ADDR_REG_OFFSET);
 	outb(val >> 8, data->addr + DATA_REG_OFFSET);
 	outb(++reg, data->addr + ADDR_REG_OFFSET);
 	outb(val & 0xff, data->addr + DATA_REG_OFFSET);
+#ifdef CONFIG_ACPI
+	acpi_ut_release_mutex(ACPI_MTX_INTERPRETER);
+#endif
 }
 
 static struct f71805f_data *f71805f_update_device(struct device *dev)


-- 
Jean Delvare

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: Could the k8temp driver be interfering with ACPI?

[Dave Jones]

On Mon, Apr 02, 2007 at 05:48:59PM +0200, Jean Delvare wrote:
 > +	u8  val;
 > +#ifdef CONFIG_ACPI
 > +	acpi_ut_acquire_mutex(ACPI_MTX_INTERPRETER);
 > +#endif
 >  	outb(reg, data->addr + ADDR_REG_OFFSET);
 > -	return inb(data->addr + DATA_REG_OFFSET);
 > +	val = inb(data->addr + DATA_REG_OFFSET);
 > +#ifdef CONFIG_ACPI
 > +	acpi_ut_release_mutex(ACPI_MTX_INTERPRETER);
 > +#endif
 > +	return val;
 > ... deletia, more of the same.

it'd probably end up a lot cleaner to #define them to empty macros
in the !ACPI case in acpi/acpi.h and just #include it unconditionally.

	Dave

-- 
http://www.codemonkey.org.uk

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Dave Jones]

On Mon, Apr 02, 2007 at 05:48:59PM +0200, Jean Delvare wrote:
 > +	u8  val;
 > +#ifdef CONFIG_ACPI
 > +	acpi_ut_acquire_mutex(ACPI_MTX_INTERPRETER);
 > +#endif
 >  	outb(reg, data->addr + ADDR_REG_OFFSET);
 > -	return inb(data->addr + DATA_REG_OFFSET);
 > +	val = inb(data->addr + DATA_REG_OFFSET);
 > +#ifdef CONFIG_ACPI
 > +	acpi_ut_release_mutex(ACPI_MTX_INTERPRETER);
 > +#endif
 > +	return val;
 > ... deletia, more of the same.

it'd probably end up a lot cleaner to #define them to empty macros
in the !ACPI case in acpi/acpi.h and just #include it unconditionally.

	Dave

-- 
http://www.codemonkey.org.uk

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

RE: Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

The ACPI specification allows concurrent execution of control methods
although methods cannot be preempted. The ACPICA interpreter mutex is
used to implement this model. 

>From section 5.5.2, "Control Method Execution": Interpretation of a
Control Method is not preemptive, but it can block. When a control
method does block, the operating software can initiate or continue the
execution of a different control method. A control method can only
assume that access to global objects is exclusive for any period the
control method does not block.

Therefore, the interpreter lock is acquired and a control method is
allowed to execute to completion unless it blocks on one of the events
described below. If the method blocks, the interpreter is unlocked and
other control methods may execute.

I'm not sure what you mean by "in the middle of an SMBus transaction", I
don't know how long such a transaction is valid. I might guess that a
single transaction can only span a single operation region access, but
I'm not sure of this.

A user-installed operation region handler is an operation region handler
that is installed by a device driver. This feature would probably only
be used for custom (OEM-defined) operation region address spaces. (I
have not seen one yet.) For the standard address spaces (memory, I/O,
etc.), usually only the default handlers are used.

Bob



> -----Original Message-----
> From: linux-acpi-owner@vger.kernel.org [mailto:linux-acpi-
> owner@vger.kernel.org] On Behalf Of Jean Delvare
> Sent: Monday, April 02, 2007 8:49 AM
> To: Pavel Machek; Brown, Len
> Cc: Matthew Garrett; Chuck Ebbert; Rudolf Marek; linux-
> acpi@vger.kernel.org; linux-kernel; lm-sensors@lm-sensors.org
> Subject: Re: Could the k8temp driver be interfering with ACPI?
> 
> Looking at the comment before acpi_ex_exit_interpreter raises two
> questions though:
> 
> >  * Cases where the interpreter is unlocked:
> >  *      1) Completion of the execution of a control method
> >  *      2) Method blocked on a Sleep() AML opcode
> >  *      3) Method blocked on an Acquire() AML opcode
> >  *      4) Method blocked on a Wait() AML opcode
> >  *      5) Method blocked to acquire the global lock
> >  *      6) Method blocked to execute a serialized control method
that is
> >  *          already executing
> >  *      7) About to invoke a user-installed opregion handler
> 
> 1* This suggests that the mutex could be released by the AML
> interpreter in the middle of an SMBus transaction. If so, and if it
> happens in practice, this means that we unfortunately cannot use this
> mutex to reliably protect the SMBus drivers from concurrent accesses.
> This even suggests that it's simply not possible to have a mutex we
> take at the beginning when entering the AML interpreter and we release
> when leaving the AML interpreter, as it looks like ACPI itself allows
> interlaced execution of AML functions. Len, is this true?
> 
> What is the purpose of the ACPI_MTX_INTERPRETER mutex in the first
> place, given that it seems it will be released on numerous occasions?
> Is it to prevent concurrent AML execution while still allowing
> interlaced execution?
> 
> 2* What are "user-installed opregion handlers"? Are they something
that
> could help solve the ACPI vs. other drivers problem?
> 
> Thanks.

RE: Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

The ACPI specification allows concurrent execution of control methods
although methods cannot be preempted. The ACPICA interpreter mutex is
used to implement this model. 

>From section 5.5.2, "Control Method Execution": Interpretation of a
Control Method is not preemptive, but it can block. When a control
method does block, the operating software can initiate or continue the
execution of a different control method. A control method can only
assume that access to global objects is exclusive for any period the
control method does not block.

Therefore, the interpreter lock is acquired and a control method is
allowed to execute to completion unless it blocks on one of the events
described below. If the method blocks, the interpreter is unlocked and
other control methods may execute.

I'm not sure what you mean by "in the middle of an SMBus transaction", I
don't know how long such a transaction is valid. I might guess that a
single transaction can only span a single operation region access, but
I'm not sure of this.

A user-installed operation region handler is an operation region handler
that is installed by a device driver. This feature would probably only
be used for custom (OEM-defined) operation region address spaces. (I
have not seen one yet.) For the standard address spaces (memory, I/O,
etc.), usually only the default handlers are used.

Bob



> -----Original Message-----
> From: linux-acpi-owner@vger.kernel.org [mailto:linux-acpi-
> owner@vger.kernel.org] On Behalf Of Jean Delvare
> Sent: Monday, April 02, 2007 8:49 AM
> To: Pavel Machek; Brown, Len
> Cc: Matthew Garrett; Chuck Ebbert; Rudolf Marek; linux-
> acpi@vger.kernel.org; linux-kernel; lm-sensors@lm-sensors.org
> Subject: Re: Could the k8temp driver be interfering with ACPI?
> 
> Looking at the comment before acpi_ex_exit_interpreter raises two
> questions though:
> 
> >  * Cases where the interpreter is unlocked:
> >  *      1) Completion of the execution of a control method
> >  *      2) Method blocked on a Sleep() AML opcode
> >  *      3) Method blocked on an Acquire() AML opcode
> >  *      4) Method blocked on a Wait() AML opcode
> >  *      5) Method blocked to acquire the global lock
> >  *      6) Method blocked to execute a serialized control method
that is
> >  *          already executing
> >  *      7) About to invoke a user-installed opregion handler
> 
> 1* This suggests that the mutex could be released by the AML
> interpreter in the middle of an SMBus transaction. If so, and if it
> happens in practice, this means that we unfortunately cannot use this
> mutex to reliably protect the SMBus drivers from concurrent accesses.
> This even suggests that it's simply not possible to have a mutex we
> take at the beginning when entering the AML interpreter and we release
> when leaving the AML interpreter, as it looks like ACPI itself allows
> interlaced execution of AML functions. Len, is this true?
> 
> What is the purpose of the ACPI_MTX_INTERPRETER mutex in the first
> place, given that it seems it will be released on numerous occasions?
> Is it to prevent concurrent AML execution while still allowing
> interlaced execution?
> 
> 2* What are "user-installed opregion handlers"? Are they something
that
> could help solve the ACPI vs. other drivers problem?
> 
> Thanks.

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

The ACPI specification allows concurrent execution of control methods
although methods cannot be preempted. The ACPICA interpreter mutex is
used to implement this model. 

From section 5.5.2, "Control Method Execution": Interpretation of a
Control Method is not preemptive, but it can block. When a control
method does block, the operating software can initiate or continue the
execution of a different control method. A control method can only
assume that access to global objects is exclusive for any period the
control method does not block.

Therefore, the interpreter lock is acquired and a control method is
allowed to execute to completion unless it blocks on one of the events
described below. If the method blocks, the interpreter is unlocked and
other control methods may execute.

I'm not sure what you mean by "in the middle of an SMBus transaction", I
don't know how long such a transaction is valid. I might guess that a
single transaction can only span a single operation region access, but
I'm not sure of this.

A user-installed operation region handler is an operation region handler
that is installed by a device driver. This feature would probably only
be used for custom (OEM-defined) operation region address spaces. (I
have not seen one yet.) For the standard address spaces (memory, I/O,
etc.), usually only the default handlers are used.

Bob



> -----Original Message-----
> From: linux-acpi-owner@vger.kernel.org [mailto:linux-acpi-
> owner@vger.kernel.org] On Behalf Of Jean Delvare
> Sent: Monday, April 02, 2007 8:49 AM
> To: Pavel Machek; Brown, Len
> Cc: Matthew Garrett; Chuck Ebbert; Rudolf Marek; linux-
> acpi@vger.kernel.org; linux-kernel; lm-sensors@lm-sensors.org
> Subject: Re: Could the k8temp driver be interfering with ACPI?
> 
> Looking at the comment before acpi_ex_exit_interpreter raises two
> questions though:
> 
> >  * Cases where the interpreter is unlocked:
> >  *      1) Completion of the execution of a control method
> >  *      2) Method blocked on a Sleep() AML opcode
> >  *      3) Method blocked on an Acquire() AML opcode
> >  *      4) Method blocked on a Wait() AML opcode
> >  *      5) Method blocked to acquire the global lock
> >  *      6) Method blocked to execute a serialized control method
that is
> >  *          already executing
> >  *      7) About to invoke a user-installed opregion handler
> 
> 1* This suggests that the mutex could be released by the AML
> interpreter in the middle of an SMBus transaction. If so, and if it
> happens in practice, this means that we unfortunately cannot use this
> mutex to reliably protect the SMBus drivers from concurrent accesses.
> This even suggests that it's simply not possible to have a mutex we
> take at the beginning when entering the AML interpreter and we release
> when leaving the AML interpreter, as it looks like ACPI itself allows
> interlaced execution of AML functions. Len, is this true?
> 
> What is the purpose of the ACPI_MTX_INTERPRETER mutex in the first
> place, given that it seems it will be released on numerous occasions?
> Is it to prevent concurrent AML execution while still allowing
> interlaced execution?
> 
> 2* What are "user-installed opregion handlers"? Are they something
that
> could help solve the ACPI vs. other drivers problem?
> 
> Thanks.

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Dave,

On Mon, 2 Apr 2007 15:22:09 -0400, Dave Jones wrote:
> On Mon, Apr 02, 2007 at 05:48:59PM +0200, Jean Delvare wrote:
>  > +	u8  val;
>  > +#ifdef CONFIG_ACPI
>  > +	acpi_ut_acquire_mutex(ACPI_MTX_INTERPRETER);
>  > +#endif
>  >  	outb(reg, data->addr + ADDR_REG_OFFSET);
>  > -	return inb(data->addr + DATA_REG_OFFSET);
>  > +	val = inb(data->addr + DATA_REG_OFFSET);
>  > +#ifdef CONFIG_ACPI
>  > +	acpi_ut_release_mutex(ACPI_MTX_INTERPRETER);
>  > +#endif
>  > +	return val;
>  > ... deletia, more of the same.
> 
> it'd probably end up a lot cleaner to #define them to empty macros
> in the !ACPI case in acpi/acpi.h and just #include it unconditionally.

Sure, the implementation details can be refined later. I'm only trying
to see what can be done for now.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Dave,

On Mon, 2 Apr 2007 15:22:09 -0400, Dave Jones wrote:
> On Mon, Apr 02, 2007 at 05:48:59PM +0200, Jean Delvare wrote:
>  > +	u8  val;
>  > +#ifdef CONFIG_ACPI
>  > +	acpi_ut_acquire_mutex(ACPI_MTX_INTERPRETER);
>  > +#endif
>  >  	outb(reg, data->addr + ADDR_REG_OFFSET);
>  > -	return inb(data->addr + DATA_REG_OFFSET);
>  > +	val = inb(data->addr + DATA_REG_OFFSET);
>  > +#ifdef CONFIG_ACPI
>  > +	acpi_ut_release_mutex(ACPI_MTX_INTERPRETER);
>  > +#endif
>  > +	return val;
>  > ... deletia, more of the same.
> 
> it'd probably end up a lot cleaner to #define them to empty macros
> in the !ACPI case in acpi/acpi.h and just #include it unconditionally.

Sure, the implementation details can be refined later. I'm only trying
to see what can be done for now.

-- 
Jean Delvare

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Bob,

On Mon, 2 Apr 2007 13:55:49 -0700, Moore, Robert wrote:
> The ACPI specification allows concurrent execution of control methods
> although methods cannot be preempted. The ACPICA interpreter mutex is
> used to implement this model. 
> 
> From section 5.5.2, "Control Method Execution": Interpretation of a
> Control Method is not preemptive, but it can block. When a control
> method does block, the operating software can initiate or continue the
> execution of a different control method. A control method can only
> assume that access to global objects is exclusive for any period the
> control method does not block.

Do I/O regions count as "global objects"?

> Therefore, the interpreter lock is acquired and a control method is
> allowed to execute to completion unless it blocks on one of the events
> described below. If the method blocks, the interpreter is unlocked and
> other control methods may execute.
> 
> I'm not sure what you mean by "in the middle of an SMBus transaction", I
> don't know how long such a transaction is valid. I might guess that a
> single transaction can only span a single operation region access, but
> I'm not sure of this.

Basically an SMBus transaction looks like this:
1* Prepare the transaction.
2* Start the transaction.
3* Wait for the transaction to complete, typically a few ms.
4* Read the result of the transaction.

Steps 1 and 2 require writing to the SMBus I/O region. Step 4 requires
reading from it, and so does step 3 if the wait loop is poll-based. The
transaction is only safe if we have an exclusive access to the I/O
region during all the 4 steps. My fear is that step 3 could be
implemented by ACPI using either a Sleep() or Acquire() or Wait()
opcode. If it is, we're doomed. OTOH, if it does, it is probably not
even safe for itself, unless there's an additional,
implementation-specific mutex to protect SMBus transactions. I yet have
to get my hands on the DSDT of ACPI implementations which actually
access the SMBus to see exactly how they do it.

> A user-installed operation region handler is an operation region handler
> that is installed by a device driver. This feature would probably only
> be used for custom (OEM-defined) operation region address spaces. (I
> have not seen one yet.) For the standard address spaces (memory, I/O,
> etc.), usually only the default handlers are used.

Could regular Linux device drivers install such handlers for a specific
I/O region? I'm asking because Rudolf Marek's proposal [1] to solve the
concurrent access problem involved extending struct resource with
callbacks to driver-specific routines to handle external access to an
I/O region. This sounds somewhat similar to these "user-installed
operation region handler" defined by ACPI, doesn't it? If ACPI already
has an infrastructure to handle this problem, we probably want to use
it rather than implementing our own.

[1] http://marc.info/?l=linux-kernel&m=117302946017204&w=2

-- 
Jean Delvare

Re: Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Bob,

On Mon, 2 Apr 2007 13:55:49 -0700, Moore, Robert wrote:
> The ACPI specification allows concurrent execution of control methods
> although methods cannot be preempted. The ACPICA interpreter mutex is
> used to implement this model. 
> 
> From section 5.5.2, "Control Method Execution": Interpretation of a
> Control Method is not preemptive, but it can block. When a control
> method does block, the operating software can initiate or continue the
> execution of a different control method. A control method can only
> assume that access to global objects is exclusive for any period the
> control method does not block.

Do I/O regions count as "global objects"?

> Therefore, the interpreter lock is acquired and a control method is
> allowed to execute to completion unless it blocks on one of the events
> described below. If the method blocks, the interpreter is unlocked and
> other control methods may execute.
> 
> I'm not sure what you mean by "in the middle of an SMBus transaction", I
> don't know how long such a transaction is valid. I might guess that a
> single transaction can only span a single operation region access, but
> I'm not sure of this.

Basically an SMBus transaction looks like this:
1* Prepare the transaction.
2* Start the transaction.
3* Wait for the transaction to complete, typically a few ms.
4* Read the result of the transaction.

Steps 1 and 2 require writing to the SMBus I/O region. Step 4 requires
reading from it, and so does step 3 if the wait loop is poll-based. The
transaction is only safe if we have an exclusive access to the I/O
region during all the 4 steps. My fear is that step 3 could be
implemented by ACPI using either a Sleep() or Acquire() or Wait()
opcode. If it is, we're doomed. OTOH, if it does, it is probably not
even safe for itself, unless there's an additional,
implementation-specific mutex to protect SMBus transactions. I yet have
to get my hands on the DSDT of ACPI implementations which actually
access the SMBus to see exactly how they do it.

> A user-installed operation region handler is an operation region handler
> that is installed by a device driver. This feature would probably only
> be used for custom (OEM-defined) operation region address spaces. (I
> have not seen one yet.) For the standard address spaces (memory, I/O,
> etc.), usually only the default handlers are used.

Could regular Linux device drivers install such handlers for a specific
I/O region? I'm asking because Rudolf Marek's proposal [1] to solve the
concurrent access problem involved extending struct resource with
callbacks to driver-specific routines to handle external access to an
I/O region. This sounds somewhat similar to these "user-installed
operation region handler" defined by ACPI, doesn't it? If ACPI already
has an infrastructure to handle this problem, we probably want to use
it rather than implementing our own.

[1] http://marc.info/?l=linux-kernel&m=117302946017204&w=2

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Bob,

On Mon, 2 Apr 2007 13:55:49 -0700, Moore, Robert wrote:
> The ACPI specification allows concurrent execution of control methods
> although methods cannot be preempted. The ACPICA interpreter mutex is
> used to implement this model. 
> 
> From section 5.5.2, "Control Method Execution": Interpretation of a
> Control Method is not preemptive, but it can block. When a control
> method does block, the operating software can initiate or continue the
> execution of a different control method. A control method can only
> assume that access to global objects is exclusive for any period the
> control method does not block.

Do I/O regions count as "global objects"?

> Therefore, the interpreter lock is acquired and a control method is
> allowed to execute to completion unless it blocks on one of the events
> described below. If the method blocks, the interpreter is unlocked and
> other control methods may execute.
> 
> I'm not sure what you mean by "in the middle of an SMBus transaction", I
> don't know how long such a transaction is valid. I might guess that a
> single transaction can only span a single operation region access, but
> I'm not sure of this.

Basically an SMBus transaction looks like this:
1* Prepare the transaction.
2* Start the transaction.
3* Wait for the transaction to complete, typically a few ms.
4* Read the result of the transaction.

Steps 1 and 2 require writing to the SMBus I/O region. Step 4 requires
reading from it, and so does step 3 if the wait loop is poll-based. The
transaction is only safe if we have an exclusive access to the I/O
region during all the 4 steps. My fear is that step 3 could be
implemented by ACPI using either a Sleep() or Acquire() or Wait()
opcode. If it is, we're doomed. OTOH, if it does, it is probably not
even safe for itself, unless there's an additional,
implementation-specific mutex to protect SMBus transactions. I yet have
to get my hands on the DSDT of ACPI implementations which actually
access the SMBus to see exactly how they do it.

> A user-installed operation region handler is an operation region handler
> that is installed by a device driver. This feature would probably only
> be used for custom (OEM-defined) operation region address spaces. (I
> have not seen one yet.) For the standard address spaces (memory, I/O,
> etc.), usually only the default handlers are used.

Could regular Linux device drivers install such handlers for a specific
I/O region? I'm asking because Rudolf Marek's proposal [1] to solve the
concurrent access problem involved extending struct resource with
callbacks to driver-specific routines to handle external access to an
I/O region. This sounds somewhat similar to these "user-installed
operation region handler" defined by ACPI, doesn't it? If ACPI already
has an infrastructure to handle this problem, we probably want to use
it rather than implementing our own.

[1] http://marc.info/?l=linux-kernel&m\x117302946017204&w=2

-- 
Jean Delvare

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

RE: Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

> Could regular Linux device drivers install such handlers for a
specific
> I/O region?

No. ACPICA only supports operation region handlers on a
per-address-space basis, not per-region.

> Basically an SMBus transaction looks like this:
> 1* Prepare the transaction.
> 2* Start the transaction.
> 3* Wait for the transaction to complete, typically a few ms.
> 4* Read the result of the transaction.

As far as the AML interpreter is concerned, access to the SMBus is via
an operation region. So, each access to such a region would encompass a
single SMBus transaction. Also, the interpreter remains locked during
this access (I'm fairly sure) even though the EC driver is involved.

> > execution of a different control method. A control method can only
> > assume that access to global objects is exclusive for any period the
> > control method does not block.
> 
> Do I/O regions count as "global objects"?

I think the spec is referring to any global namespace object. This
includes operation regions, so the answer is yes, as long as access to
the region does not block and cause the interpreter to be released. As
far as ACPICA, none of the default handlers for operation regions will
block.

Bob


> -----Original Message-----
> From: Jean Delvare [mailto:khali@linux-fr.org]
> Sent: Tuesday, April 03, 2007 12:21 AM
> To: Moore, Robert
> Cc: Pavel Machek; Brown, Len; Matthew Garrett; Chuck Ebbert; Rudolf
Marek;
> linux-acpi@vger.kernel.org; linux-kernel; lm-sensors@lm-sensors.org
> Subject: Re: Could the k8temp driver be interfering with ACPI?
> 
> Hi Bob,
> 
> On Mon, 2 Apr 2007 13:55:49 -0700, Moore, Robert wrote:
> > The ACPI specification allows concurrent execution of control
methods
> > although methods cannot be preempted. The ACPICA interpreter mutex
is
> > used to implement this model.
> >
> > From section 5.5.2, "Control Method Execution": Interpretation of a
> > Control Method is not preemptive, but it can block. When a control
> > method does block, the operating software can initiate or continue
the
> > execution of a different control method. A control method can only
> > assume that access to global objects is exclusive for any period the
> > control method does not block.
> 
> Do I/O regions count as "global objects"?
> 
> > Therefore, the interpreter lock is acquired and a control method is
> > allowed to execute to completion unless it blocks on one of the
events
> > described below. If the method blocks, the interpreter is unlocked
and
> > other control methods may execute.
> >
> > I'm not sure what you mean by "in the middle of an SMBus
transaction", I
> > don't know how long such a transaction is valid. I might guess that
a
> > single transaction can only span a single operation region access,
but
> > I'm not sure of this.
> 
> Basically an SMBus transaction looks like this:
> 1* Prepare the transaction.
> 2* Start the transaction.
> 3* Wait for the transaction to complete, typically a few ms.
> 4* Read the result of the transaction.
> 
> Steps 1 and 2 require writing to the SMBus I/O region. Step 4 requires
> reading from it, and so does step 3 if the wait loop is poll-based.
The
> transaction is only safe if we have an exclusive access to the I/O
> region during all the 4 steps. My fear is that step 3 could be
> implemented by ACPI using either a Sleep() or Acquire() or Wait()
> opcode. If it is, we're doomed. OTOH, if it does, it is probably not
> even safe for itself, unless there's an additional,
> implementation-specific mutex to protect SMBus transactions. I yet
have
> to get my hands on the DSDT of ACPI implementations which actually
> access the SMBus to see exactly how they do it.
> 
> > A user-installed operation region handler is an operation region
handler
> > that is installed by a device driver. This feature would probably
only
> > be used for custom (OEM-defined) operation region address spaces. (I
> > have not seen one yet.) For the standard address spaces (memory,
I/O,
> > etc.), usually only the default handlers are used.
> 
> Could regular Linux device drivers install such handlers for a
specific
> I/O region? I'm asking because Rudolf Marek's proposal [1] to solve
the
> concurrent access problem involved extending struct resource with
> callbacks to driver-specific routines to handle external access to an
> I/O region. This sounds somewhat similar to these "user-installed
> operation region handler" defined by ACPI, doesn't it? If ACPI already
> has an infrastructure to handle this problem, we probably want to use
> it rather than implementing our own.
> 
> [1] http://marc.info/?l=linux-kernel&m=117302946017204&w=2
> 
> --
> Jean Delvare

RE: Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

> Could regular Linux device drivers install such handlers for a
specific
> I/O region?

No. ACPICA only supports operation region handlers on a
per-address-space basis, not per-region.

> Basically an SMBus transaction looks like this:
> 1* Prepare the transaction.
> 2* Start the transaction.
> 3* Wait for the transaction to complete, typically a few ms.
> 4* Read the result of the transaction.

As far as the AML interpreter is concerned, access to the SMBus is via
an operation region. So, each access to such a region would encompass a
single SMBus transaction. Also, the interpreter remains locked during
this access (I'm fairly sure) even though the EC driver is involved.

> > execution of a different control method. A control method can only
> > assume that access to global objects is exclusive for any period the
> > control method does not block.
> 
> Do I/O regions count as "global objects"?

I think the spec is referring to any global namespace object. This
includes operation regions, so the answer is yes, as long as access to
the region does not block and cause the interpreter to be released. As
far as ACPICA, none of the default handlers for operation regions will
block.

Bob


> -----Original Message-----
> From: Jean Delvare [mailto:khali@linux-fr.org]
> Sent: Tuesday, April 03, 2007 12:21 AM
> To: Moore, Robert
> Cc: Pavel Machek; Brown, Len; Matthew Garrett; Chuck Ebbert; Rudolf
Marek;
> linux-acpi@vger.kernel.org; linux-kernel; lm-sensors@lm-sensors.org
> Subject: Re: Could the k8temp driver be interfering with ACPI?
> 
> Hi Bob,
> 
> On Mon, 2 Apr 2007 13:55:49 -0700, Moore, Robert wrote:
> > The ACPI specification allows concurrent execution of control
methods
> > although methods cannot be preempted. The ACPICA interpreter mutex
is
> > used to implement this model.
> >
> > From section 5.5.2, "Control Method Execution": Interpretation of a
> > Control Method is not preemptive, but it can block. When a control
> > method does block, the operating software can initiate or continue
the
> > execution of a different control method. A control method can only
> > assume that access to global objects is exclusive for any period the
> > control method does not block.
> 
> Do I/O regions count as "global objects"?
> 
> > Therefore, the interpreter lock is acquired and a control method is
> > allowed to execute to completion unless it blocks on one of the
events
> > described below. If the method blocks, the interpreter is unlocked
and
> > other control methods may execute.
> >
> > I'm not sure what you mean by "in the middle of an SMBus
transaction", I
> > don't know how long such a transaction is valid. I might guess that
a
> > single transaction can only span a single operation region access,
but
> > I'm not sure of this.
> 
> Basically an SMBus transaction looks like this:
> 1* Prepare the transaction.
> 2* Start the transaction.
> 3* Wait for the transaction to complete, typically a few ms.
> 4* Read the result of the transaction.
> 
> Steps 1 and 2 require writing to the SMBus I/O region. Step 4 requires
> reading from it, and so does step 3 if the wait loop is poll-based.
The
> transaction is only safe if we have an exclusive access to the I/O
> region during all the 4 steps. My fear is that step 3 could be
> implemented by ACPI using either a Sleep() or Acquire() or Wait()
> opcode. If it is, we're doomed. OTOH, if it does, it is probably not
> even safe for itself, unless there's an additional,
> implementation-specific mutex to protect SMBus transactions. I yet
have
> to get my hands on the DSDT of ACPI implementations which actually
> access the SMBus to see exactly how they do it.
> 
> > A user-installed operation region handler is an operation region
handler
> > that is installed by a device driver. This feature would probably
only
> > be used for custom (OEM-defined) operation region address spaces. (I
> > have not seen one yet.) For the standard address spaces (memory,
I/O,
> > etc.), usually only the default handlers are used.
> 
> Could regular Linux device drivers install such handlers for a
specific
> I/O region? I'm asking because Rudolf Marek's proposal [1] to solve
the
> concurrent access problem involved extending struct resource with
> callbacks to driver-specific routines to handle external access to an
> I/O region. This sounds somewhat similar to these "user-installed
> operation region handler" defined by ACPI, doesn't it? If ACPI already
> has an infrastructure to handle this problem, we probably want to use
> it rather than implementing our own.
> 
> [1] http://marc.info/?l=linux-kernel&m=117302946017204&w=2
> 
> --
> Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Moore, Robert]

> Could regular Linux device drivers install such handlers for a
specific
> I/O region?

No. ACPICA only supports operation region handlers on a
per-address-space basis, not per-region.

> Basically an SMBus transaction looks like this:
> 1* Prepare the transaction.
> 2* Start the transaction.
> 3* Wait for the transaction to complete, typically a few ms.
> 4* Read the result of the transaction.

As far as the AML interpreter is concerned, access to the SMBus is via
an operation region. So, each access to such a region would encompass a
single SMBus transaction. Also, the interpreter remains locked during
this access (I'm fairly sure) even though the EC driver is involved.

> > execution of a different control method. A control method can only
> > assume that access to global objects is exclusive for any period the
> > control method does not block.
> 
> Do I/O regions count as "global objects"?

I think the spec is referring to any global namespace object. This
includes operation regions, so the answer is yes, as long as access to
the region does not block and cause the interpreter to be released. As
far as ACPICA, none of the default handlers for operation regions will
block.

Bob


> -----Original Message-----
> From: Jean Delvare [mailto:khali@linux-fr.org]
> Sent: Tuesday, April 03, 2007 12:21 AM
> To: Moore, Robert
> Cc: Pavel Machek; Brown, Len; Matthew Garrett; Chuck Ebbert; Rudolf
Marek;
> linux-acpi@vger.kernel.org; linux-kernel; lm-sensors@lm-sensors.org
> Subject: Re: Could the k8temp driver be interfering with ACPI?
> 
> Hi Bob,
> 
> On Mon, 2 Apr 2007 13:55:49 -0700, Moore, Robert wrote:
> > The ACPI specification allows concurrent execution of control
methods
> > although methods cannot be preempted. The ACPICA interpreter mutex
is
> > used to implement this model.
> >
> > From section 5.5.2, "Control Method Execution": Interpretation of a
> > Control Method is not preemptive, but it can block. When a control
> > method does block, the operating software can initiate or continue
the
> > execution of a different control method. A control method can only
> > assume that access to global objects is exclusive for any period the
> > control method does not block.
> 
> Do I/O regions count as "global objects"?
> 
> > Therefore, the interpreter lock is acquired and a control method is
> > allowed to execute to completion unless it blocks on one of the
events
> > described below. If the method blocks, the interpreter is unlocked
and
> > other control methods may execute.
> >
> > I'm not sure what you mean by "in the middle of an SMBus
transaction", I
> > don't know how long such a transaction is valid. I might guess that
a
> > single transaction can only span a single operation region access,
but
> > I'm not sure of this.
> 
> Basically an SMBus transaction looks like this:
> 1* Prepare the transaction.
> 2* Start the transaction.
> 3* Wait for the transaction to complete, typically a few ms.
> 4* Read the result of the transaction.
> 
> Steps 1 and 2 require writing to the SMBus I/O region. Step 4 requires
> reading from it, and so does step 3 if the wait loop is poll-based.
The
> transaction is only safe if we have an exclusive access to the I/O
> region during all the 4 steps. My fear is that step 3 could be
> implemented by ACPI using either a Sleep() or Acquire() or Wait()
> opcode. If it is, we're doomed. OTOH, if it does, it is probably not
> even safe for itself, unless there's an additional,
> implementation-specific mutex to protect SMBus transactions. I yet
have
> to get my hands on the DSDT of ACPI implementations which actually
> access the SMBus to see exactly how they do it.
> 
> > A user-installed operation region handler is an operation region
handler
> > that is installed by a device driver. This feature would probably
only
> > be used for custom (OEM-defined) operation region address spaces. (I
> > have not seen one yet.) For the standard address spaces (memory,
I/O,
> > etc.), usually only the default handlers are used.
> 
> Could regular Linux device drivers install such handlers for a
specific
> I/O region? I'm asking because Rudolf Marek's proposal [1] to solve
the
> concurrent access problem involved extending struct resource with
> callbacks to driver-specific routines to handle external access to an
> I/O region. This sounds somewhat similar to these "user-installed
> operation region handler" defined by ACPI, doesn't it? If ACPI already
> has an infrastructure to handle this problem, we probably want to use
> it rather than implementing our own.
> 
> [1] http://marc.info/?l=linux-kernel&m\x117302946017204&w=2
> 
> --
> Jean Delvare

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bjorn Helgaas]

On Friday 02 March 2007 07:03, Jean Delvare wrote:
> ... The primary issue is the concurrent access
> to resources, which cause lots of trouble which are hard to investigate.
> If ACPI reserves the ports, then the SMBus or hardware monitoring
> drivers (or any other conflicting driver) will cleanly fail to load,
> which would be a move in the right direction. ...
> 
> So, can ACPI actually reserve the ports it accesses?

Sorry to join this discussion so late.

ACPI tells us the resources used by devices.  Today, we don't reserve
ACPI resources until a driver claims a device.  PCI does some sort of
reservation up front, before the driver claims devices.  Conceptually,
I think ACPI should do the same thing, and I don't think it's that
hard to do.

But breaking things like lmsensors would make the transition painful.

Bjorn

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bjorn Helgaas]

On Friday 02 March 2007 07:03, Jean Delvare wrote:
> ... The primary issue is the concurrent access
> to resources, which cause lots of trouble which are hard to investigate.
> If ACPI reserves the ports, then the SMBus or hardware monitoring
> drivers (or any other conflicting driver) will cleanly fail to load,
> which would be a move in the right direction. ...
> 
> So, can ACPI actually reserve the ports it accesses?

Sorry to join this discussion so late.

ACPI tells us the resources used by devices.  Today, we don't reserve
ACPI resources until a driver claims a device.  PCI does some sort of
reservation up front, before the driver claims devices.  Conceptually,
I think ACPI should do the same thing, and I don't think it's that
hard to do.

But breaking things like lmsensors would make the transition painful.

Bjorn

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > ... The primary issue is the concurrent access
> > to resources, which cause lots of trouble which are hard to investigate.
> > If ACPI reserves the ports, then the SMBus or hardware monitoring
> > drivers (or any other conflicting driver) will cleanly fail to load,
> > which would be a move in the right direction. ...
> > 
> > So, can ACPI actually reserve the ports it accesses?
> 
> Sorry to join this discussion so late.
> 
> ACPI tells us the resources used by devices.  Today, we don't
> reserve

Problem seems to be that ACPI does _not_ tell us which ports it
accesses from AML code.

But we already found a lock we can take; AFAICT we know how to solve
this problem.
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > ... The primary issue is the concurrent access
> > to resources, which cause lots of trouble which are hard to investigate.
> > If ACPI reserves the ports, then the SMBus or hardware monitoring
> > drivers (or any other conflicting driver) will cleanly fail to load,
> > which would be a move in the right direction. ...
> > 
> > So, can ACPI actually reserve the ports it accesses?
> 
> Sorry to join this discussion so late.
> 
> ACPI tells us the resources used by devices.  Today, we don't
> reserve

Problem seems to be that ACPI does _not_ tell us which ports it
accesses from AML code.

But we already found a lock we can take; AFAICT we know how to solve
this problem.
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bjorn Helgaas]

On Friday 13 April 2007 14:07, Pavel Machek wrote:
> > > ... The primary issue is the concurrent access
> > > to resources, which cause lots of trouble which are hard to investigate.
> > > If ACPI reserves the ports, then the SMBus or hardware monitoring
> > > drivers (or any other conflicting driver) will cleanly fail to load,
> > > which would be a move in the right direction. ...
> > > 
> > > So, can ACPI actually reserve the ports it accesses?
> > 
> > Sorry to join this discussion so late.
> > 
> > ACPI tells us the resources used by devices.  Today, we don't
> > reserve
> 
> Problem seems to be that ACPI does _not_ tell us which ports it
> accesses from AML code.

I think that would violate at least the spirit of the ACPI spec.
The example in section 11.6 of the ACPI 3.0 spec shows a _TMP
method that runs an EC method to read the temp, and the EC ioport
usage is correctly declared in the EC device's _CRS method.

Of course, there are always BIOS defects.  But if we could make a
case that a BIOS that doesn't declare the resources used by the AML
is defective, we could add quirks to reserve the undeclared resources.

Chuck's last update (http://lkml.org/lkml/2007/2/20/136) says his
problem turned out to be unrelated to k8temp and may have gone away
after a BIOS update.

> But we already found a lock we can take; AFAICT we know how to solve
> this problem.

This might solve it, but doesn't seem like a clean way to do it.
I don't like the idea of sharing a lock between drivers and ACPI.
k8temp happens to be x86-dependent, so we'll always have ACPI, but
in principle, we could have the same problem with an arch-independent
PCI driver that only has ACPI on x86 and ia64 platforms.

(BTW, if Chuck's problem was solved by the BIOS update, I assume
there *is* another instance of the problem that we're trying to
solve with this lock.)

Bjorn

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bjorn Helgaas]

On Friday 13 April 2007 14:07, Pavel Machek wrote:
> > > ... The primary issue is the concurrent access
> > > to resources, which cause lots of trouble which are hard to investigate.
> > > If ACPI reserves the ports, then the SMBus or hardware monitoring
> > > drivers (or any other conflicting driver) will cleanly fail to load,
> > > which would be a move in the right direction. ...
> > > 
> > > So, can ACPI actually reserve the ports it accesses?
> > 
> > Sorry to join this discussion so late.
> > 
> > ACPI tells us the resources used by devices.  Today, we don't
> > reserve
> 
> Problem seems to be that ACPI does _not_ tell us which ports it
> accesses from AML code.

I think that would violate at least the spirit of the ACPI spec.
The example in section 11.6 of the ACPI 3.0 spec shows a _TMP
method that runs an EC method to read the temp, and the EC ioport
usage is correctly declared in the EC device's _CRS method.

Of course, there are always BIOS defects.  But if we could make a
case that a BIOS that doesn't declare the resources used by the AML
is defective, we could add quirks to reserve the undeclared resources.

Chuck's last update (http://lkml.org/lkml/2007/2/20/136) says his
problem turned out to be unrelated to k8temp and may have gone away
after a BIOS update.

> But we already found a lock we can take; AFAICT we know how to solve
> this problem.

This might solve it, but doesn't seem like a clean way to do it.
I don't like the idea of sharing a lock between drivers and ACPI.
k8temp happens to be x86-dependent, so we'll always have ACPI, but
in principle, we could have the same problem with an arch-independent
PCI driver that only has ACPI on x86 and ia64 platforms.

(BTW, if Chuck's problem was solved by the BIOS update, I assume
there *is* another instance of the problem that we're trying to
solve with this lock.)

Bjorn

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Bjorn,

On Fri, 13 Apr 2007 14:59:45 -0600, Bjorn Helgaas wrote:
> On Friday 13 April 2007 14:07, Pavel Machek wrote:
> > > > ... The primary issue is the concurrent access
> > > > to resources, which cause lots of trouble which are hard to investigate.
> > > > If ACPI reserves the ports, then the SMBus or hardware monitoring
> > > > drivers (or any other conflicting driver) will cleanly fail to load,
> > > > which would be a move in the right direction. ...
> > > > 
> > > > So, can ACPI actually reserve the ports it accesses?
> > > 
> > > Sorry to join this discussion so late.
> > > 
> > > ACPI tells us the resources used by devices.  Today, we don't
> > > reserve
> > 
> > Problem seems to be that ACPI does _not_ tell us which ports it
> > accesses from AML code.
> 
> I think that would violate at least the spirit of the ACPI spec.
> The example in section 11.6 of the ACPI 3.0 spec shows a _TMP
> method that runs an EC method to read the temp, and the EC ioport
> usage is correctly declared in the EC device's _CRS method.
> 
> Of course, there are always BIOS defects.  But if we could make a
> case that a BIOS that doesn't declare the resources used by the AML
> is defective, we could add quirks to reserve the undeclared resources.

Only realistic if the list of systems needing a quirk is small. Do you
think that would be the case?

> > But we already found a lock we can take; AFAICT we know how to solve
> > this problem.

We even have two solutions, but both have drawbacks.

The AML lock solution has performance issue. It could be improved with a
better semaphore primitive, but that needs to be implemented first. It
also requires to modify all drivers ACPI might conflict with, and
that's a rather long list.

Rudolf Marek's port forwarding idea should perform better, and might
also be useful for other problems than ACPI vs. regular drivers,
however it makes resources bigger, and requires to modify the same long
list of drivers, and with specific code for every driver. I'm not sure
that it will be possible for all driver in practice, as this more or
less requires to forecast the I/O access pattern of ACPI.

> This might solve it, but doesn't seem like a clean way to do it.
> I don't like the idea of sharing a lock between drivers and ACPI.
> k8temp happens to be x86-dependent, so we'll always have ACPI, but
> in principle, we could have the same problem with an arch-independent
> PCI driver that only has ACPI on x86 and ia64 platforms.

We can protect the additional driver code with #ifdef CONFIG_ACPI, as I
did in my proof-of-concept. Not particularly elegant, granted, but it
should work.

If something cleaner can be done, I'm all for it. Could you please
propose an implementation of your idea? It doesn't need to be perfect,
but I would like to see what it would look like.

> (BTW, if Chuck's problem was solved by the BIOS update, I assume
> there *is* another instance of the problem that we're trying to
> solve with this lock.)

Yes, there are several other systems out there which are known to be
affected by the problem. And probably many others where ACPI access I/O
ports reserved by regular drivers and we are simply lucky that nothing
bad happens. My desktop system has such a motherboard (Jetway K8M8MS).
But SMP alternatives are becoming more popular, so we might start
seeing more problems in practice in a near future.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Bjorn,

On Fri, 13 Apr 2007 14:59:45 -0600, Bjorn Helgaas wrote:
> On Friday 13 April 2007 14:07, Pavel Machek wrote:
> > > > ... The primary issue is the concurrent access
> > > > to resources, which cause lots of trouble which are hard to investigate.
> > > > If ACPI reserves the ports, then the SMBus or hardware monitoring
> > > > drivers (or any other conflicting driver) will cleanly fail to load,
> > > > which would be a move in the right direction. ...
> > > > 
> > > > So, can ACPI actually reserve the ports it accesses?
> > > 
> > > Sorry to join this discussion so late.
> > > 
> > > ACPI tells us the resources used by devices.  Today, we don't
> > > reserve
> > 
> > Problem seems to be that ACPI does _not_ tell us which ports it
> > accesses from AML code.
> 
> I think that would violate at least the spirit of the ACPI spec.
> The example in section 11.6 of the ACPI 3.0 spec shows a _TMP
> method that runs an EC method to read the temp, and the EC ioport
> usage is correctly declared in the EC device's _CRS method.
> 
> Of course, there are always BIOS defects.  But if we could make a
> case that a BIOS that doesn't declare the resources used by the AML
> is defective, we could add quirks to reserve the undeclared resources.

Only realistic if the list of systems needing a quirk is small. Do you
think that would be the case?

> > But we already found a lock we can take; AFAICT we know how to solve
> > this problem.

We even have two solutions, but both have drawbacks.

The AML lock solution has performance issue. It could be improved with a
better semaphore primitive, but that needs to be implemented first. It
also requires to modify all drivers ACPI might conflict with, and
that's a rather long list.

Rudolf Marek's port forwarding idea should perform better, and might
also be useful for other problems than ACPI vs. regular drivers,
however it makes resources bigger, and requires to modify the same long
list of drivers, and with specific code for every driver. I'm not sure
that it will be possible for all driver in practice, as this more or
less requires to forecast the I/O access pattern of ACPI.

> This might solve it, but doesn't seem like a clean way to do it.
> I don't like the idea of sharing a lock between drivers and ACPI.
> k8temp happens to be x86-dependent, so we'll always have ACPI, but
> in principle, we could have the same problem with an arch-independent
> PCI driver that only has ACPI on x86 and ia64 platforms.

We can protect the additional driver code with #ifdef CONFIG_ACPI, as I
did in my proof-of-concept. Not particularly elegant, granted, but it
should work.

If something cleaner can be done, I'm all for it. Could you please
propose an implementation of your idea? It doesn't need to be perfect,
but I would like to see what it would look like.

> (BTW, if Chuck's problem was solved by the BIOS update, I assume
> there *is* another instance of the problem that we're trying to
> solve with this lock.)

Yes, there are several other systems out there which are known to be
affected by the problem. And probably many others where ACPI access I/O
ports reserved by regular drivers and we are simply lucky that nothing
bad happens. My desktop system has such a motherboard (Jetway K8M8MS).
But SMP alternatives are becoming more popular, so we might start
seeing more problems in practice in a near future.

-- 
Jean Delvare

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bjorn Helgaas]

On Sunday 15 April 2007 03:41, Jean Delvare wrote:
> On Fri, 13 Apr 2007 14:59:45 -0600, Bjorn Helgaas wrote:
> > Of course, there are always BIOS defects.  But if we could make a
> > case that a BIOS that doesn't declare the resources used by the AML
> > is defective, we could add quirks to reserve the undeclared resources.
> 
> Only realistic if the list of systems needing a quirk is small. Do you
> think that would be the case?

I don't know.  I confess that I don't clearly understand the problem
yet.  It sounds like the sensor drivers want to talk to hardware that
ACPI methods might also use.

But I missed the details, such as the specific devices in question,
which ports they use, how they are described in ACPI, which AML
methods use those ports, and which non-ACPI drivers also use them.

It also sounds like the non-ACPI drivers provide much more
functionality than ACPI exposes.  I'd like to understand this,
too, because an  obvious way to solve the problem would be to
drop the non-ACPI drivers.  Is this extra functionality available
on Windows?  If so, do we know whether Windows uses non-ACPI drivers
or whether they have some smarter way to use ACPI?  In the long
run, I think the easiest, most reliable route would be to use the
system in a similar way.  Then we'd be doing things the way the
manufacturer intended and we'd take advantage of all the Windows-
focused firmware testing.

Bjorn

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bjorn Helgaas]

On Sunday 15 April 2007 03:41, Jean Delvare wrote:
> On Fri, 13 Apr 2007 14:59:45 -0600, Bjorn Helgaas wrote:
> > Of course, there are always BIOS defects.  But if we could make a
> > case that a BIOS that doesn't declare the resources used by the AML
> > is defective, we could add quirks to reserve the undeclared resources.
> 
> Only realistic if the list of systems needing a quirk is small. Do you
> think that would be the case?

I don't know.  I confess that I don't clearly understand the problem
yet.  It sounds like the sensor drivers want to talk to hardware that
ACPI methods might also use.

But I missed the details, such as the specific devices in question,
which ports they use, how they are described in ACPI, which AML
methods use those ports, and which non-ACPI drivers also use them.

It also sounds like the non-ACPI drivers provide much more
functionality than ACPI exposes.  I'd like to understand this,
too, because an  obvious way to solve the problem would be to
drop the non-ACPI drivers.  Is this extra functionality available
on Windows?  If so, do we know whether Windows uses non-ACPI drivers
or whether they have some smarter way to use ACPI?  In the long
run, I think the easiest, most reliable route would be to use the
system in a similar way.  Then we'd be doing things the way the
manufacturer intended and we'd take advantage of all the Windows-
focused firmware testing.

Bjorn

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Luca Tettamanti]

On 4/15/07, Bjorn Helgaas <bjorn.helgaas@hp.com> wrote:
> On Sunday 15 April 2007 03:41, Jean Delvare wrote:
> > On Fri, 13 Apr 2007 14:59:45 -0600, Bjorn Helgaas wrote:
> > > Of course, there are always BIOS defects.  But if we could make a
> > > case that a BIOS that doesn't declare the resources used by the AML
> > > is defective, we could add quirks to reserve the undeclared resources.
> >
> > Only realistic if the list of systems needing a quirk is small. Do you
> > think that would be the case?
>
> I don't know.  I confess that I don't clearly understand the problem
> yet.  It sounds like the sensor drivers want to talk to hardware that
> ACPI methods might also use.

Exactly.

> But I missed the details, such as the specific devices in question,
> which ports they use, how they are described in ACPI, which AML
> methods use those ports, and which non-ACPI drivers also use them.

The original report was about the temperature sensors of K8 cpus. It
happens that ACPI reads the sensors while the linux driver is using it
and gets garbage (and shut down the system). The problem is more
generic though, and applies to all hardware monitoring chips for which
a driver exists.

> It also sounds like the non-ACPI drivers provide much more
> functionality than ACPI exposes.  I'd like to understand this,
> too, because an  obvious way to solve the problem would be to
> drop the non-ACPI drivers.

Problem is that ACPI may access the sensors anyway (e.g. via SMM).

> Is this extra functionality available
> on Windows?  If so, do we know whether Windows uses non-ACPI drivers
> or whether they have some smarter way to use ACPI?

Usually ACPI exposes 1 or 2 temperature readings (CPU and
motherboard), while the hw driver can also provide fans and voltages
measurements.

Vendors usually provides a monitoring utility for Win that also
exposes these information. It's not known whether there's a way to
avoid conflicting accesses between ACPI and the raw driver; it's
possible that it's vendor-specific and not documented.

Luca

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Luca Tettamanti]

On 4/15/07, Bjorn Helgaas <bjorn.helgaas@hp.com> wrote:
> On Sunday 15 April 2007 03:41, Jean Delvare wrote:
> > On Fri, 13 Apr 2007 14:59:45 -0600, Bjorn Helgaas wrote:
> > > Of course, there are always BIOS defects.  But if we could make a
> > > case that a BIOS that doesn't declare the resources used by the AML
> > > is defective, we could add quirks to reserve the undeclared resources.
> >
> > Only realistic if the list of systems needing a quirk is small. Do you
> > think that would be the case?
>
> I don't know.  I confess that I don't clearly understand the problem
> yet.  It sounds like the sensor drivers want to talk to hardware that
> ACPI methods might also use.

Exactly.

> But I missed the details, such as the specific devices in question,
> which ports they use, how they are described in ACPI, which AML
> methods use those ports, and which non-ACPI drivers also use them.

The original report was about the temperature sensors of K8 cpus. It
happens that ACPI reads the sensors while the linux driver is using it
and gets garbage (and shut down the system). The problem is more
generic though, and applies to all hardware monitoring chips for which
a driver exists.

> It also sounds like the non-ACPI drivers provide much more
> functionality than ACPI exposes.  I'd like to understand this,
> too, because an  obvious way to solve the problem would be to
> drop the non-ACPI drivers.

Problem is that ACPI may access the sensors anyway (e.g. via SMM).

> Is this extra functionality available
> on Windows?  If so, do we know whether Windows uses non-ACPI drivers
> or whether they have some smarter way to use ACPI?

Usually ACPI exposes 1 or 2 temperature readings (CPU and
motherboard), while the hw driver can also provide fans and voltages
measurements.

Vendors usually provides a monitoring utility for Win that also
exposes these information. It's not known whether there's a way to
avoid conflicting accesses between ACPI and the raw driver; it's
possible that it's vendor-specific and not documented.

Luca

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bjorn Helgaas]

On Sunday 15 April 2007 14:59, Luca Tettamanti wrote:
> On 4/15/07, Bjorn Helgaas <bjorn.helgaas@hp.com> wrote:
> > But I missed the details, such as the specific devices in question,
> > which ports they use, how they are described in ACPI, which AML
> > methods use those ports, and which non-ACPI drivers also use them.
> 
> The original report was about the temperature sensors of K8 cpus. It
> happens that ACPI reads the sensors while the linux driver is using it
> and gets garbage (and shut down the system). The problem is more
> generic though, and applies to all hardware monitoring chips for which
> a driver exists.

Yes, I saw that much, but that's not enough detail to craft a good
solution.

In the case of k8temp, the driver claims PCI devices with a certain
vendor and device ID.  PCI devices are mostly outside the scope of
ACPI.  There's a standard enumeration protocol, and a driver should
be able to claim any device it recognizes without fear of conflict.

I claim that an AML method that accesses a PCI device is
defective because the AML can't know whether a native driver
has claimed the device.

Sometimes the firmware can hide PCI devices so the OS
enumeration doesn't find them.  In that case, AML might
be able to safely use the PCI device, but the native
driver wouldn't be able to claim the device, so there
would be no conflict.  (Linux sometimes uses quirks to
"unhide" things like this, which could lead to a conflict
of our own making.)

I suspect that other sensor drivers may just probe for devices
at "known" addresses hard-coded in the driver.  This is a
problem because the ACPI model is that the OS learns about
all built-in devices via the ACPI namespace.  If it isn't
in the namespace, it shouldn't exist as far as the OS is
concerned.

So we could easily have the situation where ACPI uses a
sensor and does not expose it to the OS in the namespace.
In that case, the firmware expectation is that the OS
won't touch the device.  If the OS pokes around at the
magic addresses and happens to trip over the device, we
just made ourselves a problem.

> > It also sounds like the non-ACPI drivers provide much more
> > functionality than ACPI exposes.  I'd like to understand this,
> > too, because an  obvious way to solve the problem would be to
> > drop the non-ACPI drivers.
> 
> Problem is that ACPI may access the sensors anyway (e.g. via SMM).

If ACPI accesses sensors but there is no native driver, there
should be no conflict.

> > Is this extra functionality available
> > on Windows?  If so, do we know whether Windows uses non-ACPI drivers
> > or whether they have some smarter way to use ACPI?
> 
> Usually ACPI exposes 1 or 2 temperature readings (CPU and
> motherboard), while the hw driver can also provide fans and voltages
> measurements.
> 
> Vendors usually provides a monitoring utility for Win that also
> exposes these information. It's not known whether there's a way to
> avoid conflicting accesses between ACPI and the raw driver; it's
> possible that it's vendor-specific and not documented.

I'd be surprised if Windows provided interfaces to coordinate
between two drivers.  My impression is that they really want
to have a single owner for a piece of hardware.  It would be
interesting to figure out how these monitoring utilities work.
Maybe the monitor and the AML both go through an embedded
controller driver and coordinate that way?

Bjorn

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bjorn Helgaas]

On Sunday 15 April 2007 14:59, Luca Tettamanti wrote:
> On 4/15/07, Bjorn Helgaas <bjorn.helgaas@hp.com> wrote:
> > But I missed the details, such as the specific devices in question,
> > which ports they use, how they are described in ACPI, which AML
> > methods use those ports, and which non-ACPI drivers also use them.
> 
> The original report was about the temperature sensors of K8 cpus. It
> happens that ACPI reads the sensors while the linux driver is using it
> and gets garbage (and shut down the system). The problem is more
> generic though, and applies to all hardware monitoring chips for which
> a driver exists.

Yes, I saw that much, but that's not enough detail to craft a good
solution.

In the case of k8temp, the driver claims PCI devices with a certain
vendor and device ID.  PCI devices are mostly outside the scope of
ACPI.  There's a standard enumeration protocol, and a driver should
be able to claim any device it recognizes without fear of conflict.

I claim that an AML method that accesses a PCI device is
defective because the AML can't know whether a native driver
has claimed the device.

Sometimes the firmware can hide PCI devices so the OS
enumeration doesn't find them.  In that case, AML might
be able to safely use the PCI device, but the native
driver wouldn't be able to claim the device, so there
would be no conflict.  (Linux sometimes uses quirks to
"unhide" things like this, which could lead to a conflict
of our own making.)

I suspect that other sensor drivers may just probe for devices
at "known" addresses hard-coded in the driver.  This is a
problem because the ACPI model is that the OS learns about
all built-in devices via the ACPI namespace.  If it isn't
in the namespace, it shouldn't exist as far as the OS is
concerned.

So we could easily have the situation where ACPI uses a
sensor and does not expose it to the OS in the namespace.
In that case, the firmware expectation is that the OS
won't touch the device.  If the OS pokes around at the
magic addresses and happens to trip over the device, we
just made ourselves a problem.

> > It also sounds like the non-ACPI drivers provide much more
> > functionality than ACPI exposes.  I'd like to understand this,
> > too, because an  obvious way to solve the problem would be to
> > drop the non-ACPI drivers.
> 
> Problem is that ACPI may access the sensors anyway (e.g. via SMM).

If ACPI accesses sensors but there is no native driver, there
should be no conflict.

> > Is this extra functionality available
> > on Windows?  If so, do we know whether Windows uses non-ACPI drivers
> > or whether they have some smarter way to use ACPI?
> 
> Usually ACPI exposes 1 or 2 temperature readings (CPU and
> motherboard), while the hw driver can also provide fans and voltages
> measurements.
> 
> Vendors usually provides a monitoring utility for Win that also
> exposes these information. It's not known whether there's a way to
> avoid conflicting accesses between ACPI and the raw driver; it's
> possible that it's vendor-specific and not documented.

I'd be surprised if Windows provided interfaces to coordinate
between two drivers.  My impression is that they really want
to have a single owner for a piece of hardware.  It would be
interesting to figure out how these monitoring utilities work.
Maybe the monitor and the AML both go through an embedded
controller driver and coordinate that way?

Bjorn

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Luca Tettamanti]

[-- Attachment #1: Type: text/plain, Size: 6102 bytes --]

Il Sun, Apr 15, 2007 at 06:57:02PM -0600, Bjorn Helgaas ha scritto: 
> On Sunday 15 April 2007 14:59, Luca Tettamanti wrote:
> > On 4/15/07, Bjorn Helgaas <bjorn.helgaas@hp.com> wrote:
> > > But I missed the details, such as the specific devices in question,
> > > which ports they use, how they are described in ACPI, which AML
> > > methods use those ports, and which non-ACPI drivers also use them.
> > 
> > The original report was about the temperature sensors of K8 cpus. It
> > happens that ACPI reads the sensors while the linux driver is using it
> > and gets garbage (and shut down the system). The problem is more
> > generic though, and applies to all hardware monitoring chips for which
> > a driver exists.
> 
> Yes, I saw that much, but that's not enough detail to craft a good
> solution.
> 
> In the case of k8temp, the driver claims PCI devices with a certain
> vendor and device ID.  PCI devices are mostly outside the scope of
> ACPI.  There's a standard enumeration protocol, and a driver should
> be able to claim any device it recognizes without fear of conflict.
>
> I claim that an AML method that accesses a PCI device is
> defective because the AML can't know whether a native driver
> has claimed the device.

k8temp seems a special case tough.

> Sometimes the firmware can hide PCI devices so the OS
> enumeration doesn't find them.  In that case, AML might
> be able to safely use the PCI device, but the native
> driver wouldn't be able to claim the device, so there
> would be no conflict.  (Linux sometimes uses quirks to
> "unhide" things like this, which could lead to a conflict
> of our own making.)
> 
> I suspect that other sensor drivers may just probe for devices
> at "known" addresses hard-coded in the driver.

Usually sensors are attached to SMBUS or available in ISA IO space.
AFAIK they're not enumerated anywhere (at least I2C devices are not, you
poke at various addresses and see if something responds - not sure about
ISA).

> This is a
> problem because the ACPI model is that the OS learns about
> all built-in devices via the ACPI namespace.  If it isn't
> in the namespace, it shouldn't exist as far as the OS is
> concerned.
> 
> So we could easily have the situation where ACPI uses a
> sensor and does not expose it to the OS in the namespace.
> In that case, the firmware expectation is that the OS
> won't touch the device.  If the OS pokes around at the
> magic addresses and happens to trip over the device, we
> just made ourselves a problem.
> 
> > > It also sounds like the non-ACPI drivers provide much more
> > > functionality than ACPI exposes.  I'd like to understand this,
> > > too, because an  obvious way to solve the problem would be to
> > > drop the non-ACPI drivers.
> > 
> > Problem is that ACPI may access the sensors anyway (e.g. via SMM).
> 
> If ACPI accesses sensors but there is no native driver, there
> should be no conflict.

Of course. But we do have native drivers ;-)

> > > Is this extra functionality available
> > > on Windows?  If so, do we know whether Windows uses non-ACPI drivers
> > > or whether they have some smarter way to use ACPI?
> > 
> > Usually ACPI exposes 1 or 2 temperature readings (CPU and
> > motherboard), while the hw driver can also provide fans and voltages
> > measurements.
> > 
> > Vendors usually provides a monitoring utility for Win that also
> > exposes these information. It's not known whether there's a way to
> > avoid conflicting accesses between ACPI and the raw driver; it's
> > possible that it's vendor-specific and not documented.
> 
> I'd be surprised if Windows provided interfaces to coordinate
> between two drivers.  My impression is that they really want
> to have a single owner for a piece of hardware.  It would be
> interesting to figure out how these monitoring utilities work.
> Maybe the monitor and the AML both go through an embedded
> controller driver and coordinate that way?

Hum, the utility I have here (Asus PC Probe) seems to use ACPI:

Pro2.dll:
[Ordinal/Name Pointer] Table
        [  11] OCAPI_ACPI_OC_PresetList
        [  12] OCAPI_ACPI_OC__MB_GetBoardName
        [  23] OCAPI_CheckAiGear
        [   0] OCAPI_CheckIntelPowerSaving
        [   6] OCAPI_CheckWorkable
        [   2] OCAPI_Close
        [   9] OCAPI_EnableQFan
        [  16] OCAPI_GENERAL_GetList
        [  14] OCAPI_GetCpuVoltRange
        [  13] OCAPI_GetCurrentCpuFrequency
        [  15] OCAPI_GetFanStartTemp
        [   3] OCAPI_GetHealthData
        [   4] OCAPI_GetHwSensorData
        [  22] OCAPI_GetMBIF
        [   8] OCAPI_GetQFanInfo
        [   5] OCAPI_HW_EnumerateOption
        [   7] OCAPI_HideQFan
        [   1] OCAPI_Initialization
        [  19] OCAPI_NQFAN_GetData
        [  21] OCAPI_NQFAN_GetList
        [  20] OCAPI_NQFAN_SetData
        [  17] OCAPI_QFAN_GetData
        [  18] OCAPI_QFAN_SetData
        [  10] OCAPI_SetQFanTarget
        [  24] ___CPPdebugHook

ASiO.dll:
[Ordinal/Name Pointer] Table
        [   1] ASIO_Close
        [   9] ASIO_GetCpuID
        [   3] ASIO_InPortB
        [   5] ASIO_InPortD
        [   7] ASIO_MapMem
        [   0] ASIO_Open
        [   4] ASIO_OutPortB
        [   6] ASIO_OutPortD
        [  10] ASIO_ReadMSR
        [   8] ASIO_UnmapMem
        [  11] ASIO_WriteMSR
        [   2] OC_GetCurrentCpuFrequency

It seems that Asus exposes monitorining data using "ATK0110" (enumerated
in DSDT); I see it both on my P5B-E motherboard and on my notebook (L3D)
(they have different methods though). Another motherboard with the same
device may actually call it "FOOBAR123" or "WTFISTHIS".

Problem is that ACPI methods are not documented at all (how am I
supposed to know that "G6T6" is the reading of the 12V rail?) while the
datasheet of hw monitoring chips (w83627ehf in my case) are public (more
or less). Furthermore, sensor driver exposes all the reading of the chip
(e.g. in the DSDT I can't find the VSB or battery voltage).

I'm attacching my DSDT, in case you want to have fun ;-)

Luca
-- 
Collect some stars to shine for you
And start today 'cause there's only a few
A sign of times my friend

[-- Attachment #2: Asus P5B-E DSDT --]
[-- Type: application/octet-stream, Size: 19170 bytes --]

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Luca Tettamanti]

[-- Attachment #1: Type: text/plain, Size: 6102 bytes --]

Il Sun, Apr 15, 2007 at 06:57:02PM -0600, Bjorn Helgaas ha scritto: 
> On Sunday 15 April 2007 14:59, Luca Tettamanti wrote:
> > On 4/15/07, Bjorn Helgaas <bjorn.helgaas@hp.com> wrote:
> > > But I missed the details, such as the specific devices in question,
> > > which ports they use, how they are described in ACPI, which AML
> > > methods use those ports, and which non-ACPI drivers also use them.
> > 
> > The original report was about the temperature sensors of K8 cpus. It
> > happens that ACPI reads the sensors while the linux driver is using it
> > and gets garbage (and shut down the system). The problem is more
> > generic though, and applies to all hardware monitoring chips for which
> > a driver exists.
> 
> Yes, I saw that much, but that's not enough detail to craft a good
> solution.
> 
> In the case of k8temp, the driver claims PCI devices with a certain
> vendor and device ID.  PCI devices are mostly outside the scope of
> ACPI.  There's a standard enumeration protocol, and a driver should
> be able to claim any device it recognizes without fear of conflict.
>
> I claim that an AML method that accesses a PCI device is
> defective because the AML can't know whether a native driver
> has claimed the device.

k8temp seems a special case tough.

> Sometimes the firmware can hide PCI devices so the OS
> enumeration doesn't find them.  In that case, AML might
> be able to safely use the PCI device, but the native
> driver wouldn't be able to claim the device, so there
> would be no conflict.  (Linux sometimes uses quirks to
> "unhide" things like this, which could lead to a conflict
> of our own making.)
> 
> I suspect that other sensor drivers may just probe for devices
> at "known" addresses hard-coded in the driver.

Usually sensors are attached to SMBUS or available in ISA IO space.
AFAIK they're not enumerated anywhere (at least I2C devices are not, you
poke at various addresses and see if something responds - not sure about
ISA).

> This is a
> problem because the ACPI model is that the OS learns about
> all built-in devices via the ACPI namespace.  If it isn't
> in the namespace, it shouldn't exist as far as the OS is
> concerned.
> 
> So we could easily have the situation where ACPI uses a
> sensor and does not expose it to the OS in the namespace.
> In that case, the firmware expectation is that the OS
> won't touch the device.  If the OS pokes around at the
> magic addresses and happens to trip over the device, we
> just made ourselves a problem.
> 
> > > It also sounds like the non-ACPI drivers provide much more
> > > functionality than ACPI exposes.  I'd like to understand this,
> > > too, because an  obvious way to solve the problem would be to
> > > drop the non-ACPI drivers.
> > 
> > Problem is that ACPI may access the sensors anyway (e.g. via SMM).
> 
> If ACPI accesses sensors but there is no native driver, there
> should be no conflict.

Of course. But we do have native drivers ;-)

> > > Is this extra functionality available
> > > on Windows?  If so, do we know whether Windows uses non-ACPI drivers
> > > or whether they have some smarter way to use ACPI?
> > 
> > Usually ACPI exposes 1 or 2 temperature readings (CPU and
> > motherboard), while the hw driver can also provide fans and voltages
> > measurements.
> > 
> > Vendors usually provides a monitoring utility for Win that also
> > exposes these information. It's not known whether there's a way to
> > avoid conflicting accesses between ACPI and the raw driver; it's
> > possible that it's vendor-specific and not documented.
> 
> I'd be surprised if Windows provided interfaces to coordinate
> between two drivers.  My impression is that they really want
> to have a single owner for a piece of hardware.  It would be
> interesting to figure out how these monitoring utilities work.
> Maybe the monitor and the AML both go through an embedded
> controller driver and coordinate that way?

Hum, the utility I have here (Asus PC Probe) seems to use ACPI:

Pro2.dll:
[Ordinal/Name Pointer] Table
        [  11] OCAPI_ACPI_OC_PresetList
        [  12] OCAPI_ACPI_OC__MB_GetBoardName
        [  23] OCAPI_CheckAiGear
        [   0] OCAPI_CheckIntelPowerSaving
        [   6] OCAPI_CheckWorkable
        [   2] OCAPI_Close
        [   9] OCAPI_EnableQFan
        [  16] OCAPI_GENERAL_GetList
        [  14] OCAPI_GetCpuVoltRange
        [  13] OCAPI_GetCurrentCpuFrequency
        [  15] OCAPI_GetFanStartTemp
        [   3] OCAPI_GetHealthData
        [   4] OCAPI_GetHwSensorData
        [  22] OCAPI_GetMBIF
        [   8] OCAPI_GetQFanInfo
        [   5] OCAPI_HW_EnumerateOption
        [   7] OCAPI_HideQFan
        [   1] OCAPI_Initialization
        [  19] OCAPI_NQFAN_GetData
        [  21] OCAPI_NQFAN_GetList
        [  20] OCAPI_NQFAN_SetData
        [  17] OCAPI_QFAN_GetData
        [  18] OCAPI_QFAN_SetData
        [  10] OCAPI_SetQFanTarget
        [  24] ___CPPdebugHook

ASiO.dll:
[Ordinal/Name Pointer] Table
        [   1] ASIO_Close
        [   9] ASIO_GetCpuID
        [   3] ASIO_InPortB
        [   5] ASIO_InPortD
        [   7] ASIO_MapMem
        [   0] ASIO_Open
        [   4] ASIO_OutPortB
        [   6] ASIO_OutPortD
        [  10] ASIO_ReadMSR
        [   8] ASIO_UnmapMem
        [  11] ASIO_WriteMSR
        [   2] OC_GetCurrentCpuFrequency

It seems that Asus exposes monitorining data using "ATK0110" (enumerated
in DSDT); I see it both on my P5B-E motherboard and on my notebook (L3D)
(they have different methods though). Another motherboard with the same
device may actually call it "FOOBAR123" or "WTFISTHIS".

Problem is that ACPI methods are not documented at all (how am I
supposed to know that "G6T6" is the reading of the 12V rail?) while the
datasheet of hw monitoring chips (w83627ehf in my case) are public (more
or less). Furthermore, sensor driver exposes all the reading of the chip
(e.g. in the DSDT I can't find the VSB or battery voltage).

I'm attacching my DSDT, in case you want to have fun ;-)

Luca
-- 
Collect some stars to shine for you
And start today 'cause there's only a few
A sign of times my friend

[-- Attachment #2: Asus P5B-E DSDT --]
[-- Type: application/octet-stream, Size: 19170 bytes --]

[-- Attachment #3: Type: text/plain, Size: 153 bytes --]

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bjorn Helgaas]

On Monday 16 April 2007 15:14, Luca Tettamanti wrote:
> It seems that Asus exposes monitorining data using "ATK0110" (enumerated
> in DSDT); I see it both on my P5B-E motherboard and on my notebook (L3D)
> (they have different methods though). Another motherboard with the same
> device may actually call it "FOOBAR123" or "WTFISTHIS".

Yup, we have the same problem with other devices.  See the long list
of PNP IDs in 8250_pnp.c :-)

> Problem is that ACPI methods are not documented at all (how am I
> supposed to know that "G6T6" is the reading of the 12V rail?) while the
> datasheet of hw monitoring chips (w83627ehf in my case) are public (more
> or less).

Yes, I see that it's attractive to use a single w83627ehf.c driver.
For an ACPI driver, we'd have to build a list of PNP IDs, and possibly
information about which methods read which information.  That's
certainly more work.

On the other hand, the ACPI driver would avoid the synchronization
issues that started this whole thread.  That's a pretty compelling
advantage.

> Furthermore, sensor driver exposes all the reading of the chip
> (e.g. in the DSDT I can't find the VSB or battery voltage).

Maybe Asus didn't hook up those readings on the board.  I would
guess that PC Probe doesn't expose the VSB or battery voltage either.

I'm sure you've seen these:
  http://lists.lm-sensors.org/pipermail/lm-sensors/2005-October/014050.html
  http://www.lm-sensors.org/wiki/AsusFormulaHacking

Looks like nobody took up the challenge, though :-)  It looks fun
to play with, if only I had the time and hardware.

Bjorn

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bjorn Helgaas]

On Monday 16 April 2007 15:14, Luca Tettamanti wrote:
> It seems that Asus exposes monitorining data using "ATK0110" (enumerated
> in DSDT); I see it both on my P5B-E motherboard and on my notebook (L3D)
> (they have different methods though). Another motherboard with the same
> device may actually call it "FOOBAR123" or "WTFISTHIS".

Yup, we have the same problem with other devices.  See the long list
of PNP IDs in 8250_pnp.c :-)

> Problem is that ACPI methods are not documented at all (how am I
> supposed to know that "G6T6" is the reading of the 12V rail?) while the
> datasheet of hw monitoring chips (w83627ehf in my case) are public (more
> or less).

Yes, I see that it's attractive to use a single w83627ehf.c driver.
For an ACPI driver, we'd have to build a list of PNP IDs, and possibly
information about which methods read which information.  That's
certainly more work.

On the other hand, the ACPI driver would avoid the synchronization
issues that started this whole thread.  That's a pretty compelling
advantage.

> Furthermore, sensor driver exposes all the reading of the chip
> (e.g. in the DSDT I can't find the VSB or battery voltage).

Maybe Asus didn't hook up those readings on the board.  I would
guess that PC Probe doesn't expose the VSB or battery voltage either.

I'm sure you've seen these:
  http://lists.lm-sensors.org/pipermail/lm-sensors/2005-October/014050.html
  http://www.lm-sensors.org/wiki/AsusFormulaHacking

Looks like nobody took up the challenge, though :-)  It looks fun
to play with, if only I had the time and hardware.

Bjorn

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Bjorn, Luca,

On Mon, 16 Apr 2007 23:14:11 +0200, Luca Tettamanti wrote:
> Il Sun, Apr 15, 2007 at 06:57:02PM -0600, Bjorn Helgaas ha scritto: 
> > In the case of k8temp, the driver claims PCI devices with a certain
> > vendor and device ID.  PCI devices are mostly outside the scope of
> > ACPI.  There's a standard enumeration protocol, and a driver should
> > be able to claim any device it recognizes without fear of conflict.
> >
> > I claim that an AML method that accesses a PCI device is
> > defective because the AML can't know whether a native driver
> > has claimed the device.
> 
> k8temp seems a special case tough.

Yes, it is, and the latest news are that there were no problems with
the k8temp driver, it was a false alert.

> > Sometimes the firmware can hide PCI devices so the OS
> > enumeration doesn't find them.  In that case, AML might
> > be able to safely use the PCI device, but the native
> > driver wouldn't be able to claim the device, so there
> > would be no conflict.  (Linux sometimes uses quirks to
> > "unhide" things like this, which could lead to a conflict
> > of our own making.)

True, and we are stepping back on this. See:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=9208ee8286ea2c0136a4bc58638b0295bad791c8

However, it's not so easy, as I have been explaining to Pavel lately.
Some machines have the ACPI vs. lm-sensors conflict while the SMBus
wasn't hidden. On others, the SMBus was hidden and there was no
conflict. So we are not going to just stop unhiding the SMBus and fix
all the problems that way. It's neither necessary, nor sufficient. The
decision must be taken independently for every motherboard / laptop.

> > I suspect that other sensor drivers may just probe for devices
> > at "known" addresses hard-coded in the driver.
> 
> Usually sensors are attached to SMBUS or available in ISA IO space.
> AFAIK they're not enumerated anywhere (at least I2C devices are not, you
> poke at various addresses and see if something responds - not sure about
> ISA).

Legacy ISA hardware monitoring chips (W83781D, W83782D, LM78, LM79)
were indeed probed at an arbitrary address (0x290). I pretty much doubt
that these old chips are found on systems with (working) ACPI though.

More recent "ISA" hardware monitoring chips are embedded in Super-I/O
chips, they are in fact LPC devices, not ISA. The base I/O address is
read from the Super-I/O configuration space, (0x2e/0x2f or 0x4e/0x4f),
just like serial ports, parallel ports, floppy disk controllers, IR,
etc. So these devices _are_ enumerated. Sometimes they are also listed
as PNP devices.

As for I2C/SMBus devices, they are indeed not enumerated. But I fail to
see why non-enumerated devices would be the sole properly of ACPI.

> > This is a
> > problem because the ACPI model is that the OS learns about
> > all built-in devices via the ACPI namespace.  If it isn't
> > in the namespace, it shouldn't exist as far as the OS is
> > concerned.
> >
> > So we could easily have the situation where ACPI uses a
> > sensor and does not expose it to the OS in the namespace.
> > In that case, the firmware expectation is that the OS
> > won't touch the device.  If the OS pokes around at the
> > magic addresses and happens to trip over the device, we
> > just made ourselves a problem.

Interesting theory, but how does it fly in practice? Not much, I guess.
I doubt you could do any useful with an ACPI-enabled system today
without any non-ACPI driver.

> > > > Is this extra functionality available
> > > > on Windows?  If so, do we know whether Windows uses non-ACPI drivers
> > > > or whether they have some smarter way to use ACPI?
> > > 
> > > Usually ACPI exposes 1 or 2 temperature readings (CPU and
> > > motherboard), while the hw driver can also provide fans and voltages
> > > measurements.

And fan speed control.

> > > Vendors usually provides a monitoring utility for Win that also
> > > exposes these information. It's not known whether there's a way to
> > > avoid conflicting accesses between ACPI and the raw driver; it's
> > > possible that it's vendor-specific and not documented.
> > 
> > I'd be surprised if Windows provided interfaces to coordinate
> > between two drivers.  My impression is that they really want
> > to have a single owner for a piece of hardware.  It would be
> > interesting to figure out how these monitoring utilities work.
> > Maybe the monitor and the AML both go through an embedded
> > controller driver and coordinate that way?
> 
> Hum, the utility I have here (Asus PC Probe) seems to use ACPI:
> 
> Pro2.dll:
> [Ordinal/Name Pointer] Table
>         [  11] OCAPI_ACPI_OC_PresetList
>         [  12] OCAPI_ACPI_OC__MB_GetBoardName
>         [  23] OCAPI_CheckAiGear
>         [   0] OCAPI_CheckIntelPowerSaving
>         [   6] OCAPI_CheckWorkable
>         [   2] OCAPI_Close
>         [   9] OCAPI_EnableQFan
>         [  16] OCAPI_GENERAL_GetList
>         [  14] OCAPI_GetCpuVoltRange
>         [  13] OCAPI_GetCurrentCpuFrequency
>         [  15] OCAPI_GetFanStartTemp
>         [   3] OCAPI_GetHealthData
>         [   4] OCAPI_GetHwSensorData
>         [  22] OCAPI_GetMBIF
>         [   8] OCAPI_GetQFanInfo
>         [   5] OCAPI_HW_EnumerateOption
>         [   7] OCAPI_HideQFan
>         [   1] OCAPI_Initialization
>         [  19] OCAPI_NQFAN_GetData
>         [  21] OCAPI_NQFAN_GetList
>         [  20] OCAPI_NQFAN_SetData
>         [  17] OCAPI_QFAN_GetData
>         [  18] OCAPI_QFAN_SetData
>         [  10] OCAPI_SetQFanTarget
>         [  24] ___CPPdebugHook
> 
> ASiO.dll:
> [Ordinal/Name Pointer] Table
>         [   1] ASIO_Close
>         [   9] ASIO_GetCpuID
>         [   3] ASIO_InPortB
>         [   5] ASIO_InPortD
>         [   7] ASIO_MapMem
>         [   0] ASIO_Open
>         [   4] ASIO_OutPortB
>         [   6] ASIO_OutPortD
>         [  10] ASIO_ReadMSR
>         [   8] ASIO_UnmapMem
>         [  11] ASIO_WriteMSR
>         [   2] OC_GetCurrentCpuFrequency
> 
> It seems that Asus exposes monitorining data using "ATK0110" (enumerated
> in DSDT); I see it both on my P5B-E motherboard and on my notebook (L3D)
> (they have different methods though). Another motherboard with the same
> device may actually call it "FOOBAR123" or "WTFISTHIS".
> 
> Problem is that ACPI methods are not documented at all (how am I
> supposed to know that "G6T6" is the reading of the 12V rail?) while the
> datasheet of hw monitoring chips (w83627ehf in my case) are public (more
> or less). Furthermore, sensor driver exposes all the reading of the chip
> (e.g. in the DSDT I can't find the VSB or battery voltage).

Asus does things like this, yeah, but I don't remember any other vendor
having such ACPI methods. It would make sense to write a driver for
this Asus stuff. One of the problem we have in hardware monitoring
drivers is that we usually don't know how the chip is wired on a given
motherboard. The ACPI data might help.

For the other vendors, my assumption is the same as yours: they either
ignore the conflict, or have proprietary, undocumented ways to get
around it. After all, there must be a reason why there is no native
sensors support in Windows and every vendor provides it's own tool.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Bjorn, Luca,

On Mon, 16 Apr 2007 23:14:11 +0200, Luca Tettamanti wrote:
> Il Sun, Apr 15, 2007 at 06:57:02PM -0600, Bjorn Helgaas ha scritto: 
> > In the case of k8temp, the driver claims PCI devices with a certain
> > vendor and device ID.  PCI devices are mostly outside the scope of
> > ACPI.  There's a standard enumeration protocol, and a driver should
> > be able to claim any device it recognizes without fear of conflict.
> >
> > I claim that an AML method that accesses a PCI device is
> > defective because the AML can't know whether a native driver
> > has claimed the device.
> 
> k8temp seems a special case tough.

Yes, it is, and the latest news are that there were no problems with
the k8temp driver, it was a false alert.

> > Sometimes the firmware can hide PCI devices so the OS
> > enumeration doesn't find them.  In that case, AML might
> > be able to safely use the PCI device, but the native
> > driver wouldn't be able to claim the device, so there
> > would be no conflict.  (Linux sometimes uses quirks to
> > "unhide" things like this, which could lead to a conflict
> > of our own making.)

True, and we are stepping back on this. See:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h’08ee8286ea2c0136a4bc58638b0295bad791c8

However, it's not so easy, as I have been explaining to Pavel lately.
Some machines have the ACPI vs. lm-sensors conflict while the SMBus
wasn't hidden. On others, the SMBus was hidden and there was no
conflict. So we are not going to just stop unhiding the SMBus and fix
all the problems that way. It's neither necessary, nor sufficient. The
decision must be taken independently for every motherboard / laptop.

> > I suspect that other sensor drivers may just probe for devices
> > at "known" addresses hard-coded in the driver.
> 
> Usually sensors are attached to SMBUS or available in ISA IO space.
> AFAIK they're not enumerated anywhere (at least I2C devices are not, you
> poke at various addresses and see if something responds - not sure about
> ISA).

Legacy ISA hardware monitoring chips (W83781D, W83782D, LM78, LM79)
were indeed probed at an arbitrary address (0x290). I pretty much doubt
that these old chips are found on systems with (working) ACPI though.

More recent "ISA" hardware monitoring chips are embedded in Super-I/O
chips, they are in fact LPC devices, not ISA. The base I/O address is
read from the Super-I/O configuration space, (0x2e/0x2f or 0x4e/0x4f),
just like serial ports, parallel ports, floppy disk controllers, IR,
etc. So these devices _are_ enumerated. Sometimes they are also listed
as PNP devices.

As for I2C/SMBus devices, they are indeed not enumerated. But I fail to
see why non-enumerated devices would be the sole properly of ACPI.

> > This is a
> > problem because the ACPI model is that the OS learns about
> > all built-in devices via the ACPI namespace.  If it isn't
> > in the namespace, it shouldn't exist as far as the OS is
> > concerned.
> >
> > So we could easily have the situation where ACPI uses a
> > sensor and does not expose it to the OS in the namespace.
> > In that case, the firmware expectation is that the OS
> > won't touch the device.  If the OS pokes around at the
> > magic addresses and happens to trip over the device, we
> > just made ourselves a problem.

Interesting theory, but how does it fly in practice? Not much, I guess.
I doubt you could do any useful with an ACPI-enabled system today
without any non-ACPI driver.

> > > > Is this extra functionality available
> > > > on Windows?  If so, do we know whether Windows uses non-ACPI drivers
> > > > or whether they have some smarter way to use ACPI?
> > > 
> > > Usually ACPI exposes 1 or 2 temperature readings (CPU and
> > > motherboard), while the hw driver can also provide fans and voltages
> > > measurements.

And fan speed control.

> > > Vendors usually provides a monitoring utility for Win that also
> > > exposes these information. It's not known whether there's a way to
> > > avoid conflicting accesses between ACPI and the raw driver; it's
> > > possible that it's vendor-specific and not documented.
> > 
> > I'd be surprised if Windows provided interfaces to coordinate
> > between two drivers.  My impression is that they really want
> > to have a single owner for a piece of hardware.  It would be
> > interesting to figure out how these monitoring utilities work.
> > Maybe the monitor and the AML both go through an embedded
> > controller driver and coordinate that way?
> 
> Hum, the utility I have here (Asus PC Probe) seems to use ACPI:
> 
> Pro2.dll:
> [Ordinal/Name Pointer] Table
>         [  11] OCAPI_ACPI_OC_PresetList
>         [  12] OCAPI_ACPI_OC__MB_GetBoardName
>         [  23] OCAPI_CheckAiGear
>         [   0] OCAPI_CheckIntelPowerSaving
>         [   6] OCAPI_CheckWorkable
>         [   2] OCAPI_Close
>         [   9] OCAPI_EnableQFan
>         [  16] OCAPI_GENERAL_GetList
>         [  14] OCAPI_GetCpuVoltRange
>         [  13] OCAPI_GetCurrentCpuFrequency
>         [  15] OCAPI_GetFanStartTemp
>         [   3] OCAPI_GetHealthData
>         [   4] OCAPI_GetHwSensorData
>         [  22] OCAPI_GetMBIF
>         [   8] OCAPI_GetQFanInfo
>         [   5] OCAPI_HW_EnumerateOption
>         [   7] OCAPI_HideQFan
>         [   1] OCAPI_Initialization
>         [  19] OCAPI_NQFAN_GetData
>         [  21] OCAPI_NQFAN_GetList
>         [  20] OCAPI_NQFAN_SetData
>         [  17] OCAPI_QFAN_GetData
>         [  18] OCAPI_QFAN_SetData
>         [  10] OCAPI_SetQFanTarget
>         [  24] ___CPPdebugHook
> 
> ASiO.dll:
> [Ordinal/Name Pointer] Table
>         [   1] ASIO_Close
>         [   9] ASIO_GetCpuID
>         [   3] ASIO_InPortB
>         [   5] ASIO_InPortD
>         [   7] ASIO_MapMem
>         [   0] ASIO_Open
>         [   4] ASIO_OutPortB
>         [   6] ASIO_OutPortD
>         [  10] ASIO_ReadMSR
>         [   8] ASIO_UnmapMem
>         [  11] ASIO_WriteMSR
>         [   2] OC_GetCurrentCpuFrequency
> 
> It seems that Asus exposes monitorining data using "ATK0110" (enumerated
> in DSDT); I see it both on my P5B-E motherboard and on my notebook (L3D)
> (they have different methods though). Another motherboard with the same
> device may actually call it "FOOBAR123" or "WTFISTHIS".
> 
> Problem is that ACPI methods are not documented at all (how am I
> supposed to know that "G6T6" is the reading of the 12V rail?) while the
> datasheet of hw monitoring chips (w83627ehf in my case) are public (more
> or less). Furthermore, sensor driver exposes all the reading of the chip
> (e.g. in the DSDT I can't find the VSB or battery voltage).

Asus does things like this, yeah, but I don't remember any other vendor
having such ACPI methods. It would make sense to write a driver for
this Asus stuff. One of the problem we have in hardware monitoring
drivers is that we usually don't know how the chip is wired on a given
motherboard. The ACPI data might help.

For the other vendors, my assumption is the same as yours: they either
ignore the conflict, or have proprietary, undocumented ways to get
around it. After all, there must be a reason why there is no native
sensors support in Windows and every vendor provides it's own tool.

-- 
Jean Delvare

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Luca Tettamanti]

On 4/17/07, Bjorn Helgaas <bjorn.helgaas@hp.com> wrote:
> On Monday 16 April 2007 15:14, Luca Tettamanti wrote:
> > Problem is that ACPI methods are not documented at all (how am I
> > supposed to know that "G6T6" is the reading of the 12V rail?) while the
> > datasheet of hw monitoring chips (w83627ehf in my case) are public (more
> > or less).
>
> Yes, I see that it's attractive to use a single w83627ehf.c driver.
> For an ACPI driver, we'd have to build a list of PNP IDs, and possibly
> information about which methods read which information.  That's
> certainly more work.
>
> On the other hand, the ACPI driver would avoid the synchronization
> issues that started this whole thread.  That's a pretty compelling
> advantage.
>
> > Furthermore, sensor driver exposes all the reading of the chip
> > (e.g. in the DSDT I can't find the VSB or battery voltage).
>
> Maybe Asus didn't hook up those readings on the board.  I would
> guess that PC Probe doesn't expose the VSB or battery voltage either.

PC Probe does not. But the lines are wired and the readings (from
Linux) are sensible.

> I'm sure you've seen these:
>   http://lists.lm-sensors.org/pipermail/lm-sensors/2005-October/014050.html
>   http://www.lm-sensors.org/wiki/AsusFormulaHacking

Actually I haven't, I've happily ignored ACPI until now ;-) My DSDT
doesn't look too bad, I may give it a try...

Luca

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Luca Tettamanti]

On 4/17/07, Bjorn Helgaas <bjorn.helgaas@hp.com> wrote:
> On Monday 16 April 2007 15:14, Luca Tettamanti wrote:
> > Problem is that ACPI methods are not documented at all (how am I
> > supposed to know that "G6T6" is the reading of the 12V rail?) while the
> > datasheet of hw monitoring chips (w83627ehf in my case) are public (more
> > or less).
>
> Yes, I see that it's attractive to use a single w83627ehf.c driver.
> For an ACPI driver, we'd have to build a list of PNP IDs, and possibly
> information about which methods read which information.  That's
> certainly more work.
>
> On the other hand, the ACPI driver would avoid the synchronization
> issues that started this whole thread.  That's a pretty compelling
> advantage.
>
> > Furthermore, sensor driver exposes all the reading of the chip
> > (e.g. in the DSDT I can't find the VSB or battery voltage).
>
> Maybe Asus didn't hook up those readings on the board.  I would
> guess that PC Probe doesn't expose the VSB or battery voltage either.

PC Probe does not. But the lines are wired and the readings (from
Linux) are sensible.

> I'm sure you've seen these:
>   http://lists.lm-sensors.org/pipermail/lm-sensors/2005-October/014050.html
>   http://www.lm-sensors.org/wiki/AsusFormulaHacking

Actually I haven't, I've happily ignored ACPI until now ;-) My DSDT
doesn't look too bad, I may give it a try...

Luca

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Luca Tettamanti]

Il Wed, Apr 18, 2007 at 01:50:29AM +0200, Luca Tettamanti ha scritto: 
> >I'm sure you've seen these:
> >  http://lists.lm-sensors.org/pipermail/lm-sensors/2005-October/014050.html
> >  http://www.lm-sensors.org/wiki/AsusFormulaHacking
> 
> Actually I haven't, I've happily ignored ACPI until now ;-) My DSDT
> doesn't look too bad, I may give it a try...

It wasn't hard :) Temperature reading works fine. AML code is still a
bit obscure though: the reading are enumerated in two different
sections:

* TSIF, FSIF, VSIF: they have read/write methods, easy to understand (I'm
  using these right now)

* SITM/GITM and GGPR for enumeration: more settings are available and in
  part they overlap the others. The methods are very strange, for e.g.
  temperature GITM (which calls GIT6) returns a hard-coded magic value;
  other methods (e.g. CPU frequency, Q-FAN settings) write a random
  number somewhere in the system memory and return an obscure magic
  value... very cool :)


I'm posting the proof-of-concept driver. A few notes:
- only temperature reading is implemented: I just got my MS degree so
  I've been drun^Wbusy
- sysfs files are still RO
- coding style: yes, I know... will cleanup
- There's a gazillion of debugging printk's :)
- I've extended the hwmon sysfs interface, with a new "temp%d_name";
  ACPI knows how the sensor is wired
- You'll need the following patch (against current Linus' git):

diff --git a/drivers/acpi/namespace/nsutils.c b/drivers/acpi/namespace/nsutils.c
index 90fd059..97e1139 100644
--- a/drivers/acpi/namespace/nsutils.c
+++ b/drivers/acpi/namespace/nsutils.c
@@ -700,6 +700,7 @@ struct acpi_namespace_node *acpi_ns_map_handle_to_node(acpi_handle handle)
 
 	return (ACPI_CAST_PTR(struct acpi_namespace_node, handle));
 }
+EXPORT_SYMBOL(acpi_ns_map_handle_to_node);
 
 /*******************************************************************************
  *
@@ -736,6 +737,7 @@ acpi_handle acpi_ns_convert_entry_to_handle(struct acpi_namespace_node *node)
 	return ((acpi_handle) Node);
 ------------------------------------------------------*/
 }
+EXPORT_SYMBOL(acpi_ns_convert_entry_to_handle);
 
 /*******************************************************************************
  *
@@ -875,6 +877,7 @@ acpi_ns_get_node(struct acpi_namespace_node *prefix_node,
 	ACPI_FREE(internal_path);
 	return_ACPI_STATUS(status);
 }
+EXPORT_SYMBOL(acpi_ns_get_node);
 
 /*******************************************************************************
  *

Now the driver:

/* 
 * Copyright (C) 2007 Luca Tettamanti <kronos.it@gmail.com>
 * Distribute under GPLv2.
 */
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/hwmon.h>

#include <acpi/acpi.h>
#include <acpi/acnamesp.h>
#include <acpi/acpi_drivers.h>
#include <acpi/acpi_bus.h>


#define ATK_HID "ATK0110"
#define ATK_DRV "atk-hwmon"
#define ASOC "_SB.PCI0.SBRG.ASOC"

struct atk_data {
	struct class_device *class_dev;	
	acpi_handle atk_handle;
	struct acpi_device *device;

	acpi_handle rtmp_handle;
} atk_data;


typedef ssize_t (*sysfs_show_func)(struct device *dev, struct device_attribute *attr, char *buf);
typedef ssize_t (*sysfs_store_func)(struct device *dev, struct device_attribute *attr,
			const char *buf, size_t count);


static void atk_init_attribute(struct device_attribute *attr, char *name, mode_t mode,
		sysfs_show_func show, sysfs_store_func store)
{
	attr->attr.name = name;
	attr->attr.mode = mode;
	attr->show = show;
	attr->store = store;
}

#define ATTR_NAME_SIZE 32

struct atk_temp_input {
	struct device_attribute dev_attr;
	/* id is used for the ACPI ID of the temp */
	int id;
	char attr_name[ATTR_NAME_SIZE];
};

struct atk_temp_input_list {
	int count;
	struct atk_temp_input temp_input[];
};

#define to_atk_temp_input(attr) \
	container_of(attr, struct atk_temp_input, dev_attr)

static ssize_t atk_temp_input_show(struct device *dev, struct device_attribute *attr, char *buf) {
	struct atk_temp_input *a = to_atk_temp_input(attr);
	union acpi_object temp;
	struct acpi_buffer ret;
	struct acpi_object_list params;
	union acpi_object id;
	acpi_status status;
	ssize_t count;

	id.type = ACPI_TYPE_INTEGER;
	id.integer.value = a->id;

	params.count = 1;
	params.pointer = &id;

	ret.length = sizeof(temp);
	ret.pointer = &temp;

	status = acpi_evaluate_object(atk_data.rtmp_handle, NULL, &params, &ret);
	if (status != AE_OK) {
		printk("acpi_evaluate_object: %s\n", acpi_format_exception(status));
		return -EIO;
	}

	if (temp.type != ACPI_TYPE_INTEGER) {
		printk("not an integer (%d)\n", temp.type);
		return -EIO;
	}

	/* ACPI returns centidegree */
	count = sprintf(buf, "%llu\n", temp.integer.value * 10);

	return count;
}

struct atk_temp_name {
	struct device_attribute dev_attr;
	char attr_name[ATTR_NAME_SIZE];
	char const *name;
};

struct atk_temp_name_list {
	int count;
	struct atk_temp_name temp_name[];
};

#define to_atk_temp_name(attr) \
	container_of(attr, struct atk_temp_name, dev_attr)

static ssize_t atk_temp_name_show(struct device *dev, struct device_attribute *attr, char *buf) {
	struct atk_temp_name *a = to_atk_temp_name(attr);

	return sprintf(buf, "%s\n", a->name);
}

enum atk_temp_pack_id {
	ATK_TEMP_PACK_MAX = 2,
	ATK_TEMP_PACK_CRIT = 3,
};

static int atk_temp_pack_read(acpi_handle handle, int pack_id, enum atk_temp_pack_id temp_id,
		unsigned long long *temp) {
	struct acpi_buffer ret;
	struct acpi_object_list params;
	union acpi_object id;
	union acpi_object *pack;
	union acpi_object *obj;
	acpi_status status;
	int err = 0;
	
	ret.length = ACPI_ALLOCATE_BUFFER;

	id.type = ACPI_TYPE_INTEGER;
	id.integer.value = pack_id;

	params.count = 1;
	params.pointer = &id;

	status = acpi_evaluate_object(handle, NULL, &params, &ret);
	if (status != AE_OK) {
		printk("atk_temp_pack_read: acpi_evaluate_object: %s\n",
				acpi_format_exception(status));
		return -EIO;
	}

	pack = ret.pointer;

	if (pack->type != ACPI_TYPE_PACKAGE) {
		printk("atk_temp_pack_read: ret object is not a package: %d\n",
			pack->type);

		err = -EIO;
		goto out;
	}
	if (pack->package.count != 5) {
		printk("atk_temp_pack_read: unexpected package size: %d\n",
				pack->package.count);
		err = -EIO;
		goto out;
	}

	obj = &pack->package.elements[temp_id];
	if (obj->type != ACPI_TYPE_INTEGER) {
		printk("atk_temp_pack_read: unexepected type %d\n",
				obj->type);
		err = -EIO;
		goto out;
	}

	*temp = obj->integer.value;
	
out:
	ACPI_FREE(ret.pointer);

	return err;
}

struct atk_temp_max {
	struct device_attribute dev_attr;
	int id;
	acpi_handle handle;
	char attr_name[ATTR_NAME_SIZE];
};

struct atk_temp_max_list {
	int count;
	struct atk_temp_max temp_max[];
};

#define to_atk_temp_max(attr) \
	container_of(attr, struct atk_temp_max, dev_attr)

static ssize_t atk_temp_max_show(struct device *dev, struct device_attribute *attr, char *buf) {
	struct atk_temp_max *a = to_atk_temp_max(attr);
	unsigned long long temp;

	if (atk_temp_pack_read(a->handle, a->id, ATK_TEMP_PACK_MAX, &temp))
		return -EIO;

	return sprintf(buf, "%lld\n", temp * 10);
}

struct atk_temp_crit {
	struct device_attribute dev_attr;
	int id;
	acpi_handle handle;
	char attr_name[ATTR_NAME_SIZE];
};

struct atk_temp_crit_list {
	int count;
	struct atk_temp_crit temp_crit[];
};

#define to_atk_temp_crit(attr) \
	container_of(attr, struct atk_temp_crit, dev_attr)

static ssize_t atk_temp_crit_show(struct device *dev, struct device_attribute *attr, char *buf) {
	struct atk_temp_crit *a = to_atk_temp_crit(attr);
	unsigned long long temp;

	if (atk_temp_pack_read(a->handle, a->id, ATK_TEMP_PACK_CRIT, &temp))
		return -EIO;

	return sprintf(buf, "%lld\n", temp * 10);
}

struct atk_temp_input_list *temp_input_list;
struct atk_temp_name_list *temp_name_list;
struct atk_temp_max_list *temp_max_list;
struct atk_temp_crit_list *temp_crit_list;

static int atk_found;

static int atk_add(struct acpi_device *device);
static int atk_remove(struct acpi_device *device, int type);

static struct acpi_driver atk_driver = {
	.name	= ATK_HID,
	.class	= "hwmon",
	.ids	= ATK_HID,
	.ops	= {
		.add	= atk_add,
		.remove	= atk_remove,
	},
};

static int atk_create_files(struct device *dev) {
	int i;
	int ret;

	/* Temperatures */
	for (i = 0; i < temp_input_list->count; i++) {
		ret = device_create_file(dev, &temp_input_list->temp_input[i].dev_attr);
		if (ret)
			return ret;
	}
	for (i = 0; i < temp_name_list->count; i++) {
		ret = device_create_file(dev, &temp_name_list->temp_name[i].dev_attr);
		if (ret)
			return ret;
	}
	for (i = 0; i < temp_max_list->count; i++) {
		ret = device_create_file(dev, &temp_max_list->temp_max[i].dev_attr);
		if (ret)
			return ret;
	}
	for (i = 0; i < temp_crit_list->count; i++) {
		ret = device_create_file(dev, &temp_crit_list->temp_crit[i].dev_attr);
		if (ret)
			return ret;
	}

	return 0;
}

static void atk_remove_files(struct device *dev) {
	int i;

	/* Temperatures */
	if (temp_input_list) {
		for (i = 0; i < temp_input_list->count; i++)
			device_remove_file(dev, &temp_input_list->temp_input[i].dev_attr);
	}
	kfree(temp_input_list);

	if (temp_name_list) {
		for (i = 0; i < temp_name_list->count; i++) {
			device_remove_file(dev, &temp_name_list->temp_name[i].dev_attr);
			kfree(temp_name_list->temp_name[i].name);
		}
	}
	kfree(temp_name_list);

	if (temp_max_list) {
		for (i = 0; i < temp_max_list->count; i++)
			device_remove_file(dev, &temp_max_list->temp_max[i].dev_attr);
	}
	kfree(temp_max_list);

	if (temp_crit_list) {
		for (i = 0; i < temp_crit_list->count; i++)
			device_remove_file(dev, &temp_crit_list->temp_crit[i].dev_attr);
	}
	kfree(temp_max_list);
}

static int atk_enumerate_temp(struct acpi_buffer *buf) {
	union acpi_object *pack;
	union acpi_object *obj;
	struct atk_temp_input_list *inputs;
	struct atk_temp_name_list *names;
	struct atk_temp_max_list *max;
	struct atk_temp_crit_list *crit;
	int i, ret;

	/* Result must be a package */
	pack = buf->pointer;
	if (pack->type != ACPI_TYPE_PACKAGE) {
		printk("atk_enumerate_temp: not a package\n");
		return -EINVAL;
	}

	if (pack->package.count < 1) {
		printk("atk_enumerate_temp: count < 1\n");
		return -EINVAL;
	}

	/* First field is the number of available readings */
	obj = &pack->package.elements[0];
	if (obj->type != ACPI_TYPE_INTEGER) {
		printk("atk_enumerate_temp: element 0 is not an int (%u)\n", obj->type);
		return -EINVAL;
	}

	/* Sanity check */
	if (pack->package.count != obj->integer.value + 1) {
		printk("atk_enumerate_temp: temperature count (%llu) differs from package count (%u)\n",
				obj->integer.value, pack->package.count);
		return -EINVAL;
	}

	inputs = kzalloc(sizeof(struct atk_temp_input_list) + sizeof(*inputs->temp_input) * obj->integer.value, GFP_KERNEL);
	if (!inputs)
		return -ENOMEM;

	names = kzalloc(sizeof(struct atk_temp_name_list) + sizeof(*names->temp_name) * obj->integer.value, GFP_KERNEL);
	if (!names) {
		ret = -ENOMEM;
		goto cleanup4;
	}

	max = kzalloc(sizeof(*max) + sizeof(*max->temp_max) * obj->integer.value, GFP_KERNEL);
	if (!max) {
		ret = -ENOMEM;
		goto cleanup3;
	}

	crit = kzalloc(sizeof(*crit) + sizeof(*crit->temp_crit) * obj->integer.value, GFP_KERNEL);
	if (!crit) {
		ret = -ENOMEM;
		goto cleanup2;
	}

	inputs->count = names->count = max->count = crit->count = obj->integer.value;
	for (i = 0; i < pack->package.count - 1; i++) {
		struct acpi_buffer buf;
		union acpi_object *temp_pack;
		union acpi_object *name;
		union acpi_object *tmax;
		union acpi_object *tcrit;
		acpi_status status;

		obj = &pack->package.elements[i + 1];

		/* obj is a handle to the temperature package */
		if (obj->type != ACPI_TYPE_ANY) {
			printk("atk_enumerate_temp: unexpected type: %d\n", obj->type);
			ret = -EINVAL;
			goto cleanup;
		}

		buf.length = ACPI_ALLOCATE_BUFFER;
		status = acpi_evaluate_object(obj->reference.handle, NULL, NULL, &buf);
		if (status != AE_OK) {
			printk("ACPI exception on object %u: %s\n", i + 1, acpi_format_exception(status));
			ret = -EINVAL;
			goto cleanup;
		}

		/* Temperature package:
		 * byte buffer?
		 *   [3]: used by GITM/SITM to locate correct GITx/SITx,
		 *        method, same as the other package (GPID)
		 *   [2]: same as the other package, seems unused in
		 *        GITM/STIM
		 *   [1]: type?
		 *        1: cpu freq?
		 *        2: voltage
		 *        3: temperature
		 *        4: fan
		 *        7: Q-FAN (alarm?)
		 *        8: NOS
		 *   [0]: used by GITx/SITx for demux; selects the
		 *        value stored in ASB1 (PRM0)
		 * description
		 * max
		 * critical
		 * unknown
		 */
		temp_pack = buf.pointer;

		if (temp_pack->type != ACPI_TYPE_PACKAGE) {
			printk("atk_enumerate_temp: unexpected type for package %u: %d\n", i + 1, temp_pack->type);
			ret = -EINVAL;
			goto cleanup;
		}

		if (temp_pack->package.count != 5) {
			printk("Invalid package count: %u\n", temp_pack->package.count);
			ret = -EINVAL;
			goto cleanup;
		}

		name = &temp_pack->package.elements[1];
		tmax = &temp_pack->package.elements[2];
		tcrit = &temp_pack->package.elements[3];

		if (name->type != ACPI_TYPE_STRING) {
			printk("atk_enumerate_temp: object %d not a string (%d)\n",
					i, name->type);
			ret = -EINVAL;
			goto cleanup;
		}

		if (tmax->type != ACPI_TYPE_INTEGER) {
			printk("atk_enumerate_temp: object %d int expected (tmax), got: %d\n",
					i, tmax->type);
			ret = -EINVAL;
			goto cleanup;
		}
		if (tcrit->type != ACPI_TYPE_INTEGER) {
			printk("atk_enumerate_temp: object %d int expected (tmax), got: %d\n",
					i, tcrit->type);
			ret = -EINVAL;
			goto cleanup;
		}

		inputs->temp_input[i].id = i;
		snprintf(inputs->temp_input[i].attr_name, ATTR_NAME_SIZE, "temp%d_input", i);
		atk_init_attribute(&inputs->temp_input[i].dev_attr, inputs->temp_input[i].attr_name,
				0444, atk_temp_input_show, NULL);
		names->temp_name[i].name = kstrdup(name->string.pointer, GFP_KERNEL);
		
		snprintf(names->temp_name[i].attr_name, ATTR_NAME_SIZE, "temp%d_name", i);
		atk_init_attribute(&names->temp_name[i].dev_attr, names->temp_name[i].attr_name,
				0444, atk_temp_name_show, NULL);
		
		max->temp_max[i].handle = obj->reference.handle;
		snprintf(max->temp_max[i].attr_name, ATTR_NAME_SIZE, "temp%d_max", i);
		atk_init_attribute(&max->temp_max[i].dev_attr, max->temp_max[i].attr_name,
				0444, atk_temp_max_show, NULL);

		crit->temp_crit[i].handle = obj->reference.handle;
		snprintf(crit->temp_crit[i].attr_name, ATTR_NAME_SIZE, "temp%d_crit", i);
		atk_init_attribute(&crit->temp_crit[i].dev_attr, crit->temp_crit[i].attr_name,
				0444, atk_temp_crit_show, NULL);

		printk("temp %u: %s [%llu-%llu]\n", inputs->temp_input[i].id, names->temp_name[i].name,
				tmax->integer.value, tcrit->integer.value);

		ACPI_FREE(buf.pointer);
	}

	temp_name_list = names;
	temp_input_list = inputs;
	temp_max_list = max;
	temp_crit_list = crit;

	return 0;
cleanup:
	kfree(crit);
cleanup2:
	kfree(max);
cleanup3:
	for (i = 0; i < names->count; i++)
		kfree(names->temp_name[i].name);
	kfree(names);
cleanup4:
	kfree(inputs);

	return ret;
}

static int atk_add(struct acpi_device *device) {
	acpi_status ret;
	int err;
	struct acpi_buffer buf;
	union acpi_object *obj;
	struct acpi_namespace_node *search_ns;
	struct acpi_namespace_node *ns;

	printk("adding...\n");

	atk_data.device = device;
	atk_data.atk_handle = device->handle;
	acpi_driver_data(device) = &atk_data;

	buf.length = ACPI_ALLOCATE_BUFFER;
	ret = acpi_evaluate_object(atk_data.atk_handle, "MBIF", NULL, &buf);
	if (ret != AE_OK)
		return -ENODEV;

	obj = buf.pointer;
	if (obj->type == ACPI_TYPE_PACKAGE) {
		printk("id: %s\n", obj->package.elements[1].string.pointer);
	}
	ACPI_FREE(buf.pointer);

	/* Check for hwmon methods */
	search_ns = acpi_ns_map_handle_to_node(device->handle);
	if (search_ns == NULL) {
		printk("NULL\n");
		return -ENODEV;
	}

	/* RTMP: read temperature */
	ret = acpi_ns_get_node(search_ns, "RTMP", ACPI_NS_NO_UPSEARCH, &ns);
	if (ret != AE_OK) {
		printk(KERN_INFO "atk: RTMP not found\n");
		return -ENODEV;
	}

	atk_data.rtmp_handle = acpi_ns_convert_entry_to_handle(ns);

	/* Enumerate temp data - TSIF */
	buf.length = ACPI_ALLOCATE_BUFFER;
	ret = acpi_evaluate_object(atk_data.atk_handle, "TSIF", NULL, &buf);
	if (ret != AE_OK) {
		printk("TSIF failed: %s\n", acpi_format_exception(ret));
		return -ENODEV;
	}

	err = atk_enumerate_temp(&buf);
	ACPI_FREE(buf.pointer);
	if (err)
		return err;

	atk_found = 1;

	return 0;
}

static int atk_remove(struct acpi_device *device, int type) {
	printk("remove %d\n", type);

	acpi_driver_data(device) = NULL;

	return AE_OK;
}

int atk_init(void) {
	int ret;

	ret = acpi_bus_register_driver(&atk_driver);
	printk("%d\n", ret);
	if (ret)
		return ret;

	if (!atk_data.atk_handle) {
		ret = -ENODEV;
		goto bus_unreg;
	}

	if (!atk_found) {
		printk("atk not found\n");
		ret = -ENODEV;
		goto bus_unreg;
	}

	atk_data.class_dev = hwmon_device_register(&atk_data.device->dev);
	if (IS_ERR(atk_data.class_dev)) {
		ret = PTR_ERR(atk_data.class_dev);
		goto bus_unreg;
	}
	
	ret = atk_create_files(&atk_data.device->dev);
	if (ret)
		goto remove_files;

	return ret;

remove_files:
	atk_remove_files(&atk_data.device->dev);
	hwmon_device_unregister(atk_data.class_dev);
bus_unreg:
	acpi_bus_unregister_driver(&atk_driver);

	return ret;
}

void atk_exit(void) {
	hwmon_device_unregister(atk_data.class_dev);

	atk_remove_files(&atk_data.device->dev);

	acpi_bus_unregister_driver(&atk_driver);


}

module_init(atk_init);
module_exit(atk_exit);

MODULE_LICENSE("GPL");



Luca
-- 
Dicono che  il cane sia  il miglior  amico dell'uomo. Secondo me  non e`
vero. Quanti dei vostri amici avete fatto castrare, recentemente?

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Luca Tettamanti]

Il Wed, Apr 18, 2007 at 01:50:29AM +0200, Luca Tettamanti ha scritto: 
> >I'm sure you've seen these:
> >  http://lists.lm-sensors.org/pipermail/lm-sensors/2005-October/014050.html
> >  http://www.lm-sensors.org/wiki/AsusFormulaHacking
> 
> Actually I haven't, I've happily ignored ACPI until now ;-) My DSDT
> doesn't look too bad, I may give it a try...

It wasn't hard :) Temperature reading works fine. AML code is still a
bit obscure though: the reading are enumerated in two different
sections:

* TSIF, FSIF, VSIF: they have read/write methods, easy to understand (I'm
  using these right now)

* SITM/GITM and GGPR for enumeration: more settings are available and in
  part they overlap the others. The methods are very strange, for e.g.
  temperature GITM (which calls GIT6) returns a hard-coded magic value;
  other methods (e.g. CPU frequency, Q-FAN settings) write a random
  number somewhere in the system memory and return an obscure magic
  value... very cool :)


I'm posting the proof-of-concept driver. A few notes:
- only temperature reading is implemented: I just got my MS degree so
  I've been drun^Wbusy
- sysfs files are still RO
- coding style: yes, I know... will cleanup
- There's a gazillion of debugging printk's :)
- I've extended the hwmon sysfs interface, with a new "temp%d_name";
  ACPI knows how the sensor is wired
- You'll need the following patch (against current Linus' git):

diff --git a/drivers/acpi/namespace/nsutils.c b/drivers/acpi/namespace/nsutils.c
index 90fd059..97e1139 100644
--- a/drivers/acpi/namespace/nsutils.c
+++ b/drivers/acpi/namespace/nsutils.c
@@ -700,6 +700,7 @@ struct acpi_namespace_node *acpi_ns_map_handle_to_node(acpi_handle handle)
 
 	return (ACPI_CAST_PTR(struct acpi_namespace_node, handle));
 }
+EXPORT_SYMBOL(acpi_ns_map_handle_to_node);
 
 /*******************************************************************************
  *
@@ -736,6 +737,7 @@ acpi_handle acpi_ns_convert_entry_to_handle(struct acpi_namespace_node *node)
 	return ((acpi_handle) Node);
 ------------------------------------------------------*/
 }
+EXPORT_SYMBOL(acpi_ns_convert_entry_to_handle);
 
 /*******************************************************************************
  *
@@ -875,6 +877,7 @@ acpi_ns_get_node(struct acpi_namespace_node *prefix_node,
 	ACPI_FREE(internal_path);
 	return_ACPI_STATUS(status);
 }
+EXPORT_SYMBOL(acpi_ns_get_node);
 
 /*******************************************************************************
  *

Now the driver:

/* 
 * Copyright (C) 2007 Luca Tettamanti <kronos.it@gmail.com>
 * Distribute under GPLv2.
 */
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/hwmon.h>

#include <acpi/acpi.h>
#include <acpi/acnamesp.h>
#include <acpi/acpi_drivers.h>
#include <acpi/acpi_bus.h>


#define ATK_HID "ATK0110"
#define ATK_DRV "atk-hwmon"
#define ASOC "_SB.PCI0.SBRG.ASOC"

struct atk_data {
	struct class_device *class_dev;	
	acpi_handle atk_handle;
	struct acpi_device *device;

	acpi_handle rtmp_handle;
} atk_data;


typedef ssize_t (*sysfs_show_func)(struct device *dev, struct device_attribute *attr, char *buf);
typedef ssize_t (*sysfs_store_func)(struct device *dev, struct device_attribute *attr,
			const char *buf, size_t count);


static void atk_init_attribute(struct device_attribute *attr, char *name, mode_t mode,
		sysfs_show_func show, sysfs_store_func store)
{
	attr->attr.name = name;
	attr->attr.mode = mode;
	attr->show = show;
	attr->store = store;
}

#define ATTR_NAME_SIZE 32

struct atk_temp_input {
	struct device_attribute dev_attr;
	/* id is used for the ACPI ID of the temp */
	int id;
	char attr_name[ATTR_NAME_SIZE];
};

struct atk_temp_input_list {
	int count;
	struct atk_temp_input temp_input[];
};

#define to_atk_temp_input(attr) \
	container_of(attr, struct atk_temp_input, dev_attr)

static ssize_t atk_temp_input_show(struct device *dev, struct device_attribute *attr, char *buf) {
	struct atk_temp_input *a = to_atk_temp_input(attr);
	union acpi_object temp;
	struct acpi_buffer ret;
	struct acpi_object_list params;
	union acpi_object id;
	acpi_status status;
	ssize_t count;

	id.type = ACPI_TYPE_INTEGER;
	id.integer.value = a->id;

	params.count = 1;
	params.pointer = &id;

	ret.length = sizeof(temp);
	ret.pointer = &temp;

	status = acpi_evaluate_object(atk_data.rtmp_handle, NULL, &params, &ret);
	if (status != AE_OK) {
		printk("acpi_evaluate_object: %s\n", acpi_format_exception(status));
		return -EIO;
	}

	if (temp.type != ACPI_TYPE_INTEGER) {
		printk("not an integer (%d)\n", temp.type);
		return -EIO;
	}

	/* ACPI returns centidegree */
	count = sprintf(buf, "%llu\n", temp.integer.value * 10);

	return count;
}

struct atk_temp_name {
	struct device_attribute dev_attr;
	char attr_name[ATTR_NAME_SIZE];
	char const *name;
};

struct atk_temp_name_list {
	int count;
	struct atk_temp_name temp_name[];
};

#define to_atk_temp_name(attr) \
	container_of(attr, struct atk_temp_name, dev_attr)

static ssize_t atk_temp_name_show(struct device *dev, struct device_attribute *attr, char *buf) {
	struct atk_temp_name *a = to_atk_temp_name(attr);

	return sprintf(buf, "%s\n", a->name);
}

enum atk_temp_pack_id {
	ATK_TEMP_PACK_MAX = 2,
	ATK_TEMP_PACK_CRIT = 3,
};

static int atk_temp_pack_read(acpi_handle handle, int pack_id, enum atk_temp_pack_id temp_id,
		unsigned long long *temp) {
	struct acpi_buffer ret;
	struct acpi_object_list params;
	union acpi_object id;
	union acpi_object *pack;
	union acpi_object *obj;
	acpi_status status;
	int err = 0;
	
	ret.length = ACPI_ALLOCATE_BUFFER;

	id.type = ACPI_TYPE_INTEGER;
	id.integer.value = pack_id;

	params.count = 1;
	params.pointer = &id;

	status = acpi_evaluate_object(handle, NULL, &params, &ret);
	if (status != AE_OK) {
		printk("atk_temp_pack_read: acpi_evaluate_object: %s\n",
				acpi_format_exception(status));
		return -EIO;
	}

	pack = ret.pointer;

	if (pack->type != ACPI_TYPE_PACKAGE) {
		printk("atk_temp_pack_read: ret object is not a package: %d\n",
			pack->type);

		err = -EIO;
		goto out;
	}
	if (pack->package.count != 5) {
		printk("atk_temp_pack_read: unexpected package size: %d\n",
				pack->package.count);
		err = -EIO;
		goto out;
	}

	obj = &pack->package.elements[temp_id];
	if (obj->type != ACPI_TYPE_INTEGER) {
		printk("atk_temp_pack_read: unexepected type %d\n",
				obj->type);
		err = -EIO;
		goto out;
	}

	*temp = obj->integer.value;
	
out:
	ACPI_FREE(ret.pointer);

	return err;
}

struct atk_temp_max {
	struct device_attribute dev_attr;
	int id;
	acpi_handle handle;
	char attr_name[ATTR_NAME_SIZE];
};

struct atk_temp_max_list {
	int count;
	struct atk_temp_max temp_max[];
};

#define to_atk_temp_max(attr) \
	container_of(attr, struct atk_temp_max, dev_attr)

static ssize_t atk_temp_max_show(struct device *dev, struct device_attribute *attr, char *buf) {
	struct atk_temp_max *a = to_atk_temp_max(attr);
	unsigned long long temp;

	if (atk_temp_pack_read(a->handle, a->id, ATK_TEMP_PACK_MAX, &temp))
		return -EIO;

	return sprintf(buf, "%lld\n", temp * 10);
}

struct atk_temp_crit {
	struct device_attribute dev_attr;
	int id;
	acpi_handle handle;
	char attr_name[ATTR_NAME_SIZE];
};

struct atk_temp_crit_list {
	int count;
	struct atk_temp_crit temp_crit[];
};

#define to_atk_temp_crit(attr) \
	container_of(attr, struct atk_temp_crit, dev_attr)

static ssize_t atk_temp_crit_show(struct device *dev, struct device_attribute *attr, char *buf) {
	struct atk_temp_crit *a = to_atk_temp_crit(attr);
	unsigned long long temp;

	if (atk_temp_pack_read(a->handle, a->id, ATK_TEMP_PACK_CRIT, &temp))
		return -EIO;

	return sprintf(buf, "%lld\n", temp * 10);
}

struct atk_temp_input_list *temp_input_list;
struct atk_temp_name_list *temp_name_list;
struct atk_temp_max_list *temp_max_list;
struct atk_temp_crit_list *temp_crit_list;

static int atk_found;

static int atk_add(struct acpi_device *device);
static int atk_remove(struct acpi_device *device, int type);

static struct acpi_driver atk_driver = {
	.name	= ATK_HID,
	.class	= "hwmon",
	.ids	= ATK_HID,
	.ops	= {
		.add	= atk_add,
		.remove	= atk_remove,
	},
};

static int atk_create_files(struct device *dev) {
	int i;
	int ret;

	/* Temperatures */
	for (i = 0; i < temp_input_list->count; i++) {
		ret = device_create_file(dev, &temp_input_list->temp_input[i].dev_attr);
		if (ret)
			return ret;
	}
	for (i = 0; i < temp_name_list->count; i++) {
		ret = device_create_file(dev, &temp_name_list->temp_name[i].dev_attr);
		if (ret)
			return ret;
	}
	for (i = 0; i < temp_max_list->count; i++) {
		ret = device_create_file(dev, &temp_max_list->temp_max[i].dev_attr);
		if (ret)
			return ret;
	}
	for (i = 0; i < temp_crit_list->count; i++) {
		ret = device_create_file(dev, &temp_crit_list->temp_crit[i].dev_attr);
		if (ret)
			return ret;
	}

	return 0;
}

static void atk_remove_files(struct device *dev) {
	int i;

	/* Temperatures */
	if (temp_input_list) {
		for (i = 0; i < temp_input_list->count; i++)
			device_remove_file(dev, &temp_input_list->temp_input[i].dev_attr);
	}
	kfree(temp_input_list);

	if (temp_name_list) {
		for (i = 0; i < temp_name_list->count; i++) {
			device_remove_file(dev, &temp_name_list->temp_name[i].dev_attr);
			kfree(temp_name_list->temp_name[i].name);
		}
	}
	kfree(temp_name_list);

	if (temp_max_list) {
		for (i = 0; i < temp_max_list->count; i++)
			device_remove_file(dev, &temp_max_list->temp_max[i].dev_attr);
	}
	kfree(temp_max_list);

	if (temp_crit_list) {
		for (i = 0; i < temp_crit_list->count; i++)
			device_remove_file(dev, &temp_crit_list->temp_crit[i].dev_attr);
	}
	kfree(temp_max_list);
}

static int atk_enumerate_temp(struct acpi_buffer *buf) {
	union acpi_object *pack;
	union acpi_object *obj;
	struct atk_temp_input_list *inputs;
	struct atk_temp_name_list *names;
	struct atk_temp_max_list *max;
	struct atk_temp_crit_list *crit;
	int i, ret;

	/* Result must be a package */
	pack = buf->pointer;
	if (pack->type != ACPI_TYPE_PACKAGE) {
		printk("atk_enumerate_temp: not a package\n");
		return -EINVAL;
	}

	if (pack->package.count < 1) {
		printk("atk_enumerate_temp: count < 1\n");
		return -EINVAL;
	}

	/* First field is the number of available readings */
	obj = &pack->package.elements[0];
	if (obj->type != ACPI_TYPE_INTEGER) {
		printk("atk_enumerate_temp: element 0 is not an int (%u)\n", obj->type);
		return -EINVAL;
	}

	/* Sanity check */
	if (pack->package.count != obj->integer.value + 1) {
		printk("atk_enumerate_temp: temperature count (%llu) differs from package count (%u)\n",
				obj->integer.value, pack->package.count);
		return -EINVAL;
	}

	inputs = kzalloc(sizeof(struct atk_temp_input_list) + sizeof(*inputs->temp_input) * obj->integer.value, GFP_KERNEL);
	if (!inputs)
		return -ENOMEM;

	names = kzalloc(sizeof(struct atk_temp_name_list) + sizeof(*names->temp_name) * obj->integer.value, GFP_KERNEL);
	if (!names) {
		ret = -ENOMEM;
		goto cleanup4;
	}

	max = kzalloc(sizeof(*max) + sizeof(*max->temp_max) * obj->integer.value, GFP_KERNEL);
	if (!max) {
		ret = -ENOMEM;
		goto cleanup3;
	}

	crit = kzalloc(sizeof(*crit) + sizeof(*crit->temp_crit) * obj->integer.value, GFP_KERNEL);
	if (!crit) {
		ret = -ENOMEM;
		goto cleanup2;
	}

	inputs->count = names->count = max->count = crit->count = obj->integer.value;
	for (i = 0; i < pack->package.count - 1; i++) {
		struct acpi_buffer buf;
		union acpi_object *temp_pack;
		union acpi_object *name;
		union acpi_object *tmax;
		union acpi_object *tcrit;
		acpi_status status;

		obj = &pack->package.elements[i + 1];

		/* obj is a handle to the temperature package */
		if (obj->type != ACPI_TYPE_ANY) {
			printk("atk_enumerate_temp: unexpected type: %d\n", obj->type);
			ret = -EINVAL;
			goto cleanup;
		}

		buf.length = ACPI_ALLOCATE_BUFFER;
		status = acpi_evaluate_object(obj->reference.handle, NULL, NULL, &buf);
		if (status != AE_OK) {
			printk("ACPI exception on object %u: %s\n", i + 1, acpi_format_exception(status));
			ret = -EINVAL;
			goto cleanup;
		}

		/* Temperature package:
		 * byte buffer?
		 *   [3]: used by GITM/SITM to locate correct GITx/SITx,
		 *        method, same as the other package (GPID)
		 *   [2]: same as the other package, seems unused in
		 *        GITM/STIM
		 *   [1]: type?
		 *        1: cpu freq?
		 *        2: voltage
		 *        3: temperature
		 *        4: fan
		 *        7: Q-FAN (alarm?)
		 *        8: NOS
		 *   [0]: used by GITx/SITx for demux; selects the
		 *        value stored in ASB1 (PRM0)
		 * description
		 * max
		 * critical
		 * unknown
		 */
		temp_pack = buf.pointer;

		if (temp_pack->type != ACPI_TYPE_PACKAGE) {
			printk("atk_enumerate_temp: unexpected type for package %u: %d\n", i + 1, temp_pack->type);
			ret = -EINVAL;
			goto cleanup;
		}

		if (temp_pack->package.count != 5) {
			printk("Invalid package count: %u\n", temp_pack->package.count);
			ret = -EINVAL;
			goto cleanup;
		}

		name = &temp_pack->package.elements[1];
		tmax = &temp_pack->package.elements[2];
		tcrit = &temp_pack->package.elements[3];

		if (name->type != ACPI_TYPE_STRING) {
			printk("atk_enumerate_temp: object %d not a string (%d)\n",
					i, name->type);
			ret = -EINVAL;
			goto cleanup;
		}

		if (tmax->type != ACPI_TYPE_INTEGER) {
			printk("atk_enumerate_temp: object %d int expected (tmax), got: %d\n",
					i, tmax->type);
			ret = -EINVAL;
			goto cleanup;
		}
		if (tcrit->type != ACPI_TYPE_INTEGER) {
			printk("atk_enumerate_temp: object %d int expected (tmax), got: %d\n",
					i, tcrit->type);
			ret = -EINVAL;
			goto cleanup;
		}

		inputs->temp_input[i].id = i;
		snprintf(inputs->temp_input[i].attr_name, ATTR_NAME_SIZE, "temp%d_input", i);
		atk_init_attribute(&inputs->temp_input[i].dev_attr, inputs->temp_input[i].attr_name,
				0444, atk_temp_input_show, NULL);
		names->temp_name[i].name = kstrdup(name->string.pointer, GFP_KERNEL);
		
		snprintf(names->temp_name[i].attr_name, ATTR_NAME_SIZE, "temp%d_name", i);
		atk_init_attribute(&names->temp_name[i].dev_attr, names->temp_name[i].attr_name,
				0444, atk_temp_name_show, NULL);
		
		max->temp_max[i].handle = obj->reference.handle;
		snprintf(max->temp_max[i].attr_name, ATTR_NAME_SIZE, "temp%d_max", i);
		atk_init_attribute(&max->temp_max[i].dev_attr, max->temp_max[i].attr_name,
				0444, atk_temp_max_show, NULL);

		crit->temp_crit[i].handle = obj->reference.handle;
		snprintf(crit->temp_crit[i].attr_name, ATTR_NAME_SIZE, "temp%d_crit", i);
		atk_init_attribute(&crit->temp_crit[i].dev_attr, crit->temp_crit[i].attr_name,
				0444, atk_temp_crit_show, NULL);

		printk("temp %u: %s [%llu-%llu]\n", inputs->temp_input[i].id, names->temp_name[i].name,
				tmax->integer.value, tcrit->integer.value);

		ACPI_FREE(buf.pointer);
	}

	temp_name_list = names;
	temp_input_list = inputs;
	temp_max_list = max;
	temp_crit_list = crit;

	return 0;
cleanup:
	kfree(crit);
cleanup2:
	kfree(max);
cleanup3:
	for (i = 0; i < names->count; i++)
		kfree(names->temp_name[i].name);
	kfree(names);
cleanup4:
	kfree(inputs);

	return ret;
}

static int atk_add(struct acpi_device *device) {
	acpi_status ret;
	int err;
	struct acpi_buffer buf;
	union acpi_object *obj;
	struct acpi_namespace_node *search_ns;
	struct acpi_namespace_node *ns;

	printk("adding...\n");

	atk_data.device = device;
	atk_data.atk_handle = device->handle;
	acpi_driver_data(device) = &atk_data;

	buf.length = ACPI_ALLOCATE_BUFFER;
	ret = acpi_evaluate_object(atk_data.atk_handle, "MBIF", NULL, &buf);
	if (ret != AE_OK)
		return -ENODEV;

	obj = buf.pointer;
	if (obj->type = ACPI_TYPE_PACKAGE) {
		printk("id: %s\n", obj->package.elements[1].string.pointer);
	}
	ACPI_FREE(buf.pointer);

	/* Check for hwmon methods */
	search_ns = acpi_ns_map_handle_to_node(device->handle);
	if (search_ns = NULL) {
		printk("NULL\n");
		return -ENODEV;
	}

	/* RTMP: read temperature */
	ret = acpi_ns_get_node(search_ns, "RTMP", ACPI_NS_NO_UPSEARCH, &ns);
	if (ret != AE_OK) {
		printk(KERN_INFO "atk: RTMP not found\n");
		return -ENODEV;
	}

	atk_data.rtmp_handle = acpi_ns_convert_entry_to_handle(ns);

	/* Enumerate temp data - TSIF */
	buf.length = ACPI_ALLOCATE_BUFFER;
	ret = acpi_evaluate_object(atk_data.atk_handle, "TSIF", NULL, &buf);
	if (ret != AE_OK) {
		printk("TSIF failed: %s\n", acpi_format_exception(ret));
		return -ENODEV;
	}

	err = atk_enumerate_temp(&buf);
	ACPI_FREE(buf.pointer);
	if (err)
		return err;

	atk_found = 1;

	return 0;
}

static int atk_remove(struct acpi_device *device, int type) {
	printk("remove %d\n", type);

	acpi_driver_data(device) = NULL;

	return AE_OK;
}

int atk_init(void) {
	int ret;

	ret = acpi_bus_register_driver(&atk_driver);
	printk("%d\n", ret);
	if (ret)
		return ret;

	if (!atk_data.atk_handle) {
		ret = -ENODEV;
		goto bus_unreg;
	}

	if (!atk_found) {
		printk("atk not found\n");
		ret = -ENODEV;
		goto bus_unreg;
	}

	atk_data.class_dev = hwmon_device_register(&atk_data.device->dev);
	if (IS_ERR(atk_data.class_dev)) {
		ret = PTR_ERR(atk_data.class_dev);
		goto bus_unreg;
	}
	
	ret = atk_create_files(&atk_data.device->dev);
	if (ret)
		goto remove_files;

	return ret;

remove_files:
	atk_remove_files(&atk_data.device->dev);
	hwmon_device_unregister(atk_data.class_dev);
bus_unreg:
	acpi_bus_unregister_driver(&atk_driver);

	return ret;
}

void atk_exit(void) {
	hwmon_device_unregister(atk_data.class_dev);

	atk_remove_files(&atk_data.device->dev);

	acpi_bus_unregister_driver(&atk_driver);


}

module_init(atk_init);
module_exit(atk_exit);

MODULE_LICENSE("GPL");



Luca
-- 
Dicono che  il cane sia  il miglior  amico dell'uomo. Secondo me  non e`
vero. Quanti dei vostri amici avete fatto castrare, recentemente?

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bodo Eggert]

On Wed, 7 Mar 2007, Jean Delvare wrote:
> On Tue, 6 Mar 2007 21:40:19 +0100 (CET), Bodo Eggert wrote:

> > 1) I asume port allocations or ACPI foreign port acces to be rare, so
> >    there would be little impact on (un)registering hardware. Off cause
> >    there are some long ACPI calls (like reading the battery?), in these
> >    cases it might be beneficial to release the global allocation lock
> >    after gaining exclusive access to the driver's port range. (This can
> >    only be done if the device driver is loaded.)
> 
> The problem is that the rarity of ACPI foreign port access doesn't
> matter much, as you have to do all the tests first to find out that it
> was a foreign port, and I suspect that these tests will cost more than
> taking the lock itself.

Testing for the internal ranges can be done by searching a list of about
five ranges, that should be not much slower than taking an uncontended
lock, and very much faster than waiting for smbus to finish it's current
task in case the lock is contended. I'm more concerned about driver
(un)loading being slowed down by ACPI.

[...]
> We may be able to workaround the inter-driver exclusion though, if we
> use a semaphore initialized to N, have ACPI take N, but other drivers
> take just one. This would let ACPI be exclusive with all the other
> drivers, but drivers themselves could otherwise run concurrently. That
> being said, we do not appear to have the required primitives to take
> more than 1 semaphore resource at once at the moment, so we'd need to
> do that first.

worst case I can imagine: IDE/graphic stalles while ACPI takes a second to
read the battery state.

-- 
Top 100 things you don't want the sysadmin to say:
82. Where's the DIR command?

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bodo Eggert]

On Wed, 7 Mar 2007, Jean Delvare wrote:
> On Tue, 6 Mar 2007 21:40:19 +0100 (CET), Bodo Eggert wrote:

> > 1) I asume port allocations or ACPI foreign port acces to be rare, so
> >    there would be little impact on (un)registering hardware. Off cause
> >    there are some long ACPI calls (like reading the battery?), in these
> >    cases it might be beneficial to release the global allocation lock
> >    after gaining exclusive access to the driver's port range. (This can
> >    only be done if the device driver is loaded.)
> 
> The problem is that the rarity of ACPI foreign port access doesn't
> matter much, as you have to do all the tests first to find out that it
> was a foreign port, and I suspect that these tests will cost more than
> taking the lock itself.

Testing for the internal ranges can be done by searching a list of about
five ranges, that should be not much slower than taking an uncontended
lock, and very much faster than waiting for smbus to finish it's current
task in case the lock is contended. I'm more concerned about driver
(un)loading being slowed down by ACPI.

[...]
> We may be able to workaround the inter-driver exclusion though, if we
> use a semaphore initialized to N, have ACPI take N, but other drivers
> take just one. This would let ACPI be exclusive with all the other
> drivers, but drivers themselves could otherwise run concurrently. That
> being said, we do not appear to have the required primitives to take
> more than 1 semaphore resource at once at the moment, so we'd need to
> do that first.

worst case I can imagine: IDE/graphic stalles while ACPI takes a second to
read the battery state.

-- 
Top 100 things you don't want the sysadmin to say:
82. Where's the DIR command?

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > > > 2) make ACPI take this lock whenever it touches ports not allocated by itself
> > > >    and release it on function return.
> > > 
> > > This is costly.
> > 
> > TANSTAAFL. You'll need to take some lock, and if you want port emulation 
> > or per-device-mutex, you'll have to pay the price.
> 
> True, but precisely, I wonder what will perform the best in practice, a
> single semaphore that is unconditionally taken by ACPI/AML, or your
> finer-grained approach and its many additional tests.

I'd say mutexes are not that bad. ... i/o accesses are expensive,
anyway, right?

> We may be able to workaround the inter-driver exclusion though, if we
> use a semaphore initialized to N, have ACPI take N, but other drivers
> take just one. This would let ACPI be exclusive with all the other
> drivers, but drivers themselves could otherwise run concurrently. That
> being said, we do not appear to have the required primitives to take
> more than 1 semaphore resource at once at the moment, so we'd need to
> do that first.

Hmm, nice hack. Yes, driver-driver serialization would be nasty.
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Pavel Machek]

Hi!

> > > > 2) make ACPI take this lock whenever it touches ports not allocated by itself
> > > >    and release it on function return.
> > > 
> > > This is costly.
> > 
> > TANSTAAFL. You'll need to take some lock, and if you want port emulation 
> > or per-device-mutex, you'll have to pay the price.
> 
> True, but precisely, I wonder what will perform the best in practice, a
> single semaphore that is unconditionally taken by ACPI/AML, or your
> finer-grained approach and its many additional tests.

I'd say mutexes are not that bad. ... i/o accesses are expensive,
anyway, right?

> We may be able to workaround the inter-driver exclusion though, if we
> use a semaphore initialized to N, have ACPI take N, but other drivers
> take just one. This would let ACPI be exclusive with all the other
> drivers, but drivers themselves could otherwise run concurrently. That
> being said, we do not appear to have the required primitives to take
> more than 1 semaphore resource at once at the moment, so we'd need to
> do that first.

Hmm, nice hack. Yes, driver-driver serialization would be nasty.
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Bodo,

On Tue, 6 Mar 2007 21:40:19 +0100 (CET), Bodo Eggert wrote:
> On Tue, 6 Mar 2007, Jean Delvare wrote:
> > On Mon, 05 Mar 2007 14:56:44 +0100, Bodo Eggert wrote:
> 
> > > 2) make ACPI take this lock whenever it touches ports not allocated by itself
> > >    and release it on function return.
> > 
> > This is costly.
> 
> TANSTAAFL. You'll need to take some lock, and if you want port emulation 
> or per-device-mutex, you'll have to pay the price.

True, but precisely, I wonder what will perform the best in practice, a
single semaphore that is unconditionally taken by ACPI/AML, or your
finer-grained approach and its many additional tests.

> 
> > > 3) Make ACPI callback the allocating device if it touches allocated ranges,
> > >    and on function return.
> > 
> > By callback, do you mean port forwarding as Rudolf Marek did, or more
> > simple "stop touching the device" and "you can touch the device again"
> > sort of signals? If the latter, this is no different from a mutex.
> 
> It may be the described port emulation callback (which is very nice and
> possibly desired for some devices), but if you add an "I am done tampering
> with your hardware" callback, the driver can also implement the mutex.
> 
> Some drivers will not understand all possible access patterns, they may 
> try to emulate first, and if they fail, fall back to taking the mutex and 
> passing the raw port access.

Good point.

> 
> > > 1 + 2 will replace allocating single ports and freeing them again (so ACPI won't
> > > prevent e.g. the display driver from loading just because it turned on the
> > > backlight) while preventing races with newly allocated ranges, and 3 allows
> > > coexistence with other drivers.
> > 
> > What benefit do you see compared to a lock taken by both AML and the
> > hardware monitoring drivers?
> 
> 1) I asume port allocations or ACPI foreign port acces to be rare, so
>    there would be little impact on (un)registering hardware. Off cause
>    there are some long ACPI calls (like reading the battery?), in these
>    cases it might be beneficial to release the global allocation lock
>    after gaining exclusive access to the driver's port range. (This can
>    only be done if the device driver is loaded.)

The problem is that the rarity of ACPI foreign port access doesn't
matter much, as you have to do all the tests first to find out that it
was a foreign port, and I suspect that these tests will cost more than
taking the lock itself.

> 2) If there is a global lock for accessing ACPI-tampered hardware, each
>    driver that *might* conflict with ACPI _must_ take it on each access.  
>    This would seialize unrelated sensor busses, IDE, graphics ...  (AFAI
>    can imagine.)  A callback allows using a per-(device|bus)-mutex instead.

True, this is the problem with having a single mutex: everything gets
serialized. I have no idea how often we are running AML code though, so
I don't know if this is a real problem in practice.

We may be able to workaround the inter-driver exclusion though, if we
use a semaphore initialized to N, have ACPI take N, but other drivers
take just one. This would let ACPI be exclusive with all the other
drivers, but drivers themselves could otherwise run concurrently. That
being said, we do not appear to have the required primitives to take
more than 1 semaphore resource at once at the moment, so we'd need to
do that first.

OTOH, I certainly hope that not that many drivers are affected on a
given system, so maybe serialization is acceptable. We'll need some
field experimentation to find out.

> Rudolph's implementation is close enough to imagine what would be done,
> the changes would be:
> 
> - Add a "done" call to resource_ops
> 
> in ACPI:
> - Check for own region first and do raw access there
> - get the lock before __request_region
> - s/__request/__check/
> 
> in driver:
> - If emulation fails, get the local lock and switch to passthrough mode
> - if in passthrough mode and done is called, release the lock

Yep, that's a bit complex, but should work, thanks for sketching this.

-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Bodo,

On Tue, 6 Mar 2007 21:40:19 +0100 (CET), Bodo Eggert wrote:
> On Tue, 6 Mar 2007, Jean Delvare wrote:
> > On Mon, 05 Mar 2007 14:56:44 +0100, Bodo Eggert wrote:
> 
> > > 2) make ACPI take this lock whenever it touches ports not allocated by itself
> > >    and release it on function return.
> > 
> > This is costly.
> 
> TANSTAAFL. You'll need to take some lock, and if you want port emulation 
> or per-device-mutex, you'll have to pay the price.

True, but precisely, I wonder what will perform the best in practice, a
single semaphore that is unconditionally taken by ACPI/AML, or your
finer-grained approach and its many additional tests.

> 
> > > 3) Make ACPI callback the allocating device if it touches allocated ranges,
> > >    and on function return.
> > 
> > By callback, do you mean port forwarding as Rudolf Marek did, or more
> > simple "stop touching the device" and "you can touch the device again"
> > sort of signals? If the latter, this is no different from a mutex.
> 
> It may be the described port emulation callback (which is very nice and
> possibly desired for some devices), but if you add an "I am done tampering
> with your hardware" callback, the driver can also implement the mutex.
> 
> Some drivers will not understand all possible access patterns, they may 
> try to emulate first, and if they fail, fall back to taking the mutex and 
> passing the raw port access.

Good point.

> 
> > > 1 + 2 will replace allocating single ports and freeing them again (so ACPI won't
> > > prevent e.g. the display driver from loading just because it turned on the
> > > backlight) while preventing races with newly allocated ranges, and 3 allows
> > > coexistence with other drivers.
> > 
> > What benefit do you see compared to a lock taken by both AML and the
> > hardware monitoring drivers?
> 
> 1) I asume port allocations or ACPI foreign port acces to be rare, so
>    there would be little impact on (un)registering hardware. Off cause
>    there are some long ACPI calls (like reading the battery?), in these
>    cases it might be beneficial to release the global allocation lock
>    after gaining exclusive access to the driver's port range. (This can
>    only be done if the device driver is loaded.)

The problem is that the rarity of ACPI foreign port access doesn't
matter much, as you have to do all the tests first to find out that it
was a foreign port, and I suspect that these tests will cost more than
taking the lock itself.

> 2) If there is a global lock for accessing ACPI-tampered hardware, each
>    driver that *might* conflict with ACPI _must_ take it on each access.  
>    This would seialize unrelated sensor busses, IDE, graphics ...  (AFAI
>    can imagine.)  A callback allows using a per-(device|bus)-mutex instead.

True, this is the problem with having a single mutex: everything gets
serialized. I have no idea how often we are running AML code though, so
I don't know if this is a real problem in practice.

We may be able to workaround the inter-driver exclusion though, if we
use a semaphore initialized to N, have ACPI take N, but other drivers
take just one. This would let ACPI be exclusive with all the other
drivers, but drivers themselves could otherwise run concurrently. That
being said, we do not appear to have the required primitives to take
more than 1 semaphore resource at once at the moment, so we'd need to
do that first.

OTOH, I certainly hope that not that many drivers are affected on a
given system, so maybe serialization is acceptable. We'll need some
field experimentation to find out.

> Rudolph's implementation is close enough to imagine what would be done,
> the changes would be:
> 
> - Add a "done" call to resource_ops
> 
> in ACPI:
> - Check for own region first and do raw access there
> - get the lock before __request_region
> - s/__request/__check/
> 
> in driver:
> - If emulation fails, get the local lock and switch to passthrough mode
> - if in passthrough mode and done is called, release the lock

Yep, that's a bit complex, but should work, thanks for sketching this.

-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bodo Eggert]

On Tue, 6 Mar 2007, Jean Delvare wrote:
> On Mon, 05 Mar 2007 14:56:44 +0100, Bodo Eggert wrote:

> > 2) make ACPI take this lock whenever it touches ports not allocated by itself
> >    and release it on function return.
> 
> This is costly.

TANSTAAFL. You'll need to take some lock, and if you want port emulation 
or per-device-mutex, you'll have to pay the price.

> > 3) Make ACPI callback the allocating device if it touches allocated ranges,
> >    and on function return.
> 
> By callback, do you mean port forwarding as Rudolf Marek did, or more
> simple "stop touching the device" and "you can touch the device again"
> sort of signals? If the latter, this is no different from a mutex.

It may be the described port emulation callback (which is very nice and
possibly desired for some devices), but if you add an "I am done tampering
with your hardware" callback, the driver can also implement the mutex.

Some drivers will not understand all possible access patterns, they may 
try to emulate first, and if they fail, fall back to taking the mutex and 
passing the raw port access.

> > 1 + 2 will replace allocating single ports and freeing them again (so ACPI won't
> > prevent e.g. the display driver from loading just because it turned on the
> > backlight) while preventing races with newly allocated ranges, and 3 allows
> > coexistence with other drivers.
> 
> What benefit do you see compared to a lock taken by both AML and the
> hardware monitoring drivers?

1) I asume port allocations or ACPI foreign port acces to be rare, so
   there would be little impact on (un)registering hardware. Off cause
   there are some long ACPI calls (like reading the battery?), in these
   cases it might be beneficial to release the global allocation lock
   after gaining exclusive access to the driver's port range. (This can
   only be done if the device driver is loaded.)

2) If there is a global lock for accessing ACPI-tampered hardware, each
   driver that *might* conflict with ACPI _must_ take it on each access.  
   This would seialize unrelated sensor busses, IDE, graphics ...  (AFAI
   can imagine.)  A callback allows using a per-(device|bus)-mutex instead.

> Care to submit a sample implementation?

I think somebody experienced with the internals should do this.

Rudolph's implementation is close enough to imagine what would be done,
the changes would be:

- Add a "done" call to resource_ops

in ACPI:
- Check for own region first and do raw access there
- get the lock before __request_region
- s/__request/__check/

in driver:
- If emulation fails, get the local lock and switch to passthrough mode
- if in passthrough mode and done is called, release the lock
-- 
Those who hesitate under fire usually do not end up KIA or WIA. 

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bodo Eggert]

On Tue, 6 Mar 2007, Jean Delvare wrote:
> On Mon, 05 Mar 2007 14:56:44 +0100, Bodo Eggert wrote:

> > 2) make ACPI take this lock whenever it touches ports not allocated by itself
> >    and release it on function return.
> 
> This is costly.

TANSTAAFL. You'll need to take some lock, and if you want port emulation 
or per-device-mutex, you'll have to pay the price.

> > 3) Make ACPI callback the allocating device if it touches allocated ranges,
> >    and on function return.
> 
> By callback, do you mean port forwarding as Rudolf Marek did, or more
> simple "stop touching the device" and "you can touch the device again"
> sort of signals? If the latter, this is no different from a mutex.

It may be the described port emulation callback (which is very nice and
possibly desired for some devices), but if you add an "I am done tampering
with your hardware" callback, the driver can also implement the mutex.

Some drivers will not understand all possible access patterns, they may 
try to emulate first, and if they fail, fall back to taking the mutex and 
passing the raw port access.

> > 1 + 2 will replace allocating single ports and freeing them again (so ACPI won't
> > prevent e.g. the display driver from loading just because it turned on the
> > backlight) while preventing races with newly allocated ranges, and 3 allows
> > coexistence with other drivers.
> 
> What benefit do you see compared to a lock taken by both AML and the
> hardware monitoring drivers?

1) I asume port allocations or ACPI foreign port acces to be rare, so
   there would be little impact on (un)registering hardware. Off cause
   there are some long ACPI calls (like reading the battery?), in these
   cases it might be beneficial to release the global allocation lock
   after gaining exclusive access to the driver's port range. (This can
   only be done if the device driver is loaded.)

2) If there is a global lock for accessing ACPI-tampered hardware, each
   driver that *might* conflict with ACPI _must_ take it on each access.  
   This would seialize unrelated sensor busses, IDE, graphics ...  (AFAI
   can imagine.)  A callback allows using a per-(device|bus)-mutex instead.

> Care to submit a sample implementation?

I think somebody experienced with the internals should do this.

Rudolph's implementation is close enough to imagine what would be done,
the changes would be:

- Add a "done" call to resource_ops

in ACPI:
- Check for own region first and do raw access there
- get the lock before __request_region
- s/__request/__check/

in driver:
- If emulation fails, get the local lock and switch to passthrough mode
- if in passthrough mode and done is called, release the lock
-- 
Those who hesitate under fire usually do not end up KIA or WIA. 

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Bodo,

On Mon, 05 Mar 2007 14:56:44 +0100, Bodo Eggert wrote:
> 1) Make a general resource allocation lock, if there is none.

Such a lock certainly exists, but I doubt it is public at the moment.

> 2) make ACPI take this lock whenever it touches ports not allocated by itself
>    and release it on function return.

This is costly.

> 3) Make ACPI callback the allocating device if it touches allocated ranges,
>    and on function return.

By callback, do you mean port forwarding as Rudolf Marek did, or more
simple "stop touching the device" and "you can touch the device again"
sort of signals? If the latter, this is no different from a mutex.

> 1 + 2 will replace allocating single ports and freeing them again (so ACPI won't
> prevent e.g. the display driver from loading just because it turned on the
> backlight) while preventing races with newly allocated ranges, and 3 allows
> coexistence with other drivers.

What benefit do you see compared to a lock taken by both AML and the
hardware monitoring drivers?

Care to submit a sample implementation?

Thanks,
-- 
Jean Delvare

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Jean Delvare]

Hi Bodo,

On Mon, 05 Mar 2007 14:56:44 +0100, Bodo Eggert wrote:
> 1) Make a general resource allocation lock, if there is none.

Such a lock certainly exists, but I doubt it is public at the moment.

> 2) make ACPI take this lock whenever it touches ports not allocated by itself
>    and release it on function return.

This is costly.

> 3) Make ACPI callback the allocating device if it touches allocated ranges,
>    and on function return.

By callback, do you mean port forwarding as Rudolf Marek did, or more
simple "stop touching the device" and "you can touch the device again"
sort of signals? If the latter, this is no different from a mutex.

> 1 + 2 will replace allocating single ports and freeing them again (so ACPI won't
> prevent e.g. the display driver from loading just because it turned on the
> backlight) while preventing races with newly allocated ranges, and 3 allows
> coexistence with other drivers.

What benefit do you see compared to a lock taken by both AML and the
hardware monitoring drivers?

Care to submit a sample implementation?

Thanks,
-- 
Jean Delvare

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bodo Eggert]

David Hubbard <david.c.hubbard@gmail.com> wrote:

> For I/O and memory that ACPI accesses and has not reserved, the AML
> interpreter could allocate at run-time.
> 
> I'm not sure how to implement exactly. For example, it would be bad to
> have a /proc/ioports that had a lot of single ports allocated, for
> example:
> 1000-107f : 0000:00:1f.0
>  1000-1000 : ACPI PM1a_EVT_BLK
>  1001-1001 : ACPI PM1a_EVT_BLK
>  1002-1002 : ACPI PM1a_EVT_BLK
>  1003-1003 : ACPI PM1a_EVT_BLK
> 
> Thus the AML interpreter would need to have some reasonable
> intelligence when allocating regions. Conflict resolution would also
> be more difficult, e.g. if a hwmon driver was loaded first and then
> acpi as a module, ACPI could not allocate the region. Maybe run-time
> allocating won't work.
> 
> And then, how would ACPI release a region after it has used it? The
> easiest method would be to never release anything used even once.
> 
> Thoughts?

1) Make a general resource allocation lock, if there is none.
2) make ACPI take this lock whenever it touches ports not allocated by itself
   and release it on function return.
3) Make ACPI callback the allocating device if it touches allocated ranges,
   and on function return.

1 + 2 will replace allocating single ports and freeing them again (so ACPI won't
prevent e.g. the display driver from loading just because it turned on the
backlight) while preventing races with newly allocated ranges, and 3 allows
coexistence with other drivers.
-- 
The problem with the easy way out is that it has already been mined. 

Friß, Spammer: TKF8@jQXoImLG.7eggert.dyndns.org
 lG@aOkf0g.7eggert.dyndns.org hffltUu@6au8pBwc.7eggert.dyndns.org
-
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bodo Eggert]

David Hubbard <david.c.hubbard@gmail.com> wrote:

> For I/O and memory that ACPI accesses and has not reserved, the AML
> interpreter could allocate at run-time.
> 
> I'm not sure how to implement exactly. For example, it would be bad to
> have a /proc/ioports that had a lot of single ports allocated, for
> example:
> 1000-107f : 0000:00:1f.0
>  1000-1000 : ACPI PM1a_EVT_BLK
>  1001-1001 : ACPI PM1a_EVT_BLK
>  1002-1002 : ACPI PM1a_EVT_BLK
>  1003-1003 : ACPI PM1a_EVT_BLK
> 
> Thus the AML interpreter would need to have some reasonable
> intelligence when allocating regions. Conflict resolution would also
> be more difficult, e.g. if a hwmon driver was loaded first and then
> acpi as a module, ACPI could not allocate the region. Maybe run-time
> allocating won't work.
> 
> And then, how would ACPI release a region after it has used it? The
> easiest method would be to never release anything used even once.
> 
> Thoughts?

1) Make a general resource allocation lock, if there is none.
2) make ACPI take this lock whenever it touches ports not allocated by itself
   and release it on function return.
3) Make ACPI callback the allocating device if it touches allocated ranges,
   and on function return.

1 + 2 will replace allocating single ports and freeing them again (so ACPI won't
prevent e.g. the display driver from loading just because it turned on the
backlight) while preventing races with newly allocated ranges, and 3 allows
coexistence with other drivers.
-- 
The problem with the easy way out is that it has already been mined. 

Friß, Spammer: TKF8@jQXoImLG.7eggert.dyndns.org
 lG@aOkf0g.7eggert.dyndns.org hffltUu@6au8pBwc.7eggert.dyndns.org

[lm-sensors] Could the k8temp driver be interfering with ACPI?

[Bodo Eggert]

David Hubbard <david.c.hubbard at gmail.com> wrote:

> For I/O and memory that ACPI accesses and has not reserved, the AML
> interpreter could allocate at run-time.
> 
> I'm not sure how to implement exactly. For example, it would be bad to
> have a /proc/ioports that had a lot of single ports allocated, for
> example:
> 1000-107f : 0000:00:1f.0
>  1000-1000 : ACPI PM1a_EVT_BLK
>  1001-1001 : ACPI PM1a_EVT_BLK
>  1002-1002 : ACPI PM1a_EVT_BLK
>  1003-1003 : ACPI PM1a_EVT_BLK
> 
> Thus the AML interpreter would need to have some reasonable
> intelligence when allocating regions. Conflict resolution would also
> be more difficult, e.g. if a hwmon driver was loaded first and then
> acpi as a module, ACPI could not allocate the region. Maybe run-time
> allocating won't work.
> 
> And then, how would ACPI release a region after it has used it? The
> easiest method would be to never release anything used even once.
> 
> Thoughts?

1) Make a general resource allocation lock, if there is none.
2) make ACPI take this lock whenever it touches ports not allocated by itself
   and release it on function return.
3) Make ACPI callback the allocating device if it touches allocated ranges,
   and on function return.

1 + 2 will replace allocating single ports and freeing them again (so ACPI won't
prevent e.g. the display driver from loading just because it turned on the
backlight) while preventing races with newly allocated ranges, and 3 allows
coexistence with other drivers.
-- 
The problem with the easy way out is that it has already been mined. 

Fri?, Spammer: TKF8 at jQXoImLG.7eggert.dyndns.org
 lG at aOkf0g.7eggert.dyndns.org hffltUu at 6au8pBwc.7eggert.dyndns.org