Re: [LinuxBIOS] some more tyan/s2881 questions

[Lu, Yinghai]

I have inited the hw sensors in LinuxBIOS for s2881. So you could use
lmsensor to check the FAN speed. The config file is on Tyan web.

For the Fan control, there should more some reg setting...to reduce the
FAN speed winbond and adm1027...

YH

[LinuxBIOS] some more tyan/s2881 questions

[Ward Vandewege]

On Thu, Apr 20, 2006 at 04:59:52PM -0700, Lu, Yinghai wrote:
> I have inited the hw sensors in LinuxBIOS for s2881. So you could use
> lmsensor to check the FAN speed. The config file is on Tyan web.

Yes; that's what I did today, it works well. I can even set the FAN speed.
Thanks for doing that!
> For the Fan control, there should more some reg setting...to reduce the
> FAN speed winbond and adm1027...

Any idea how hard that will be/how long it will take? So you're saying that
would activate the automatic fan speed control?

Thanks,
Ward.

[LinuxBIOS] s2881 fan control

[Ward Vandewege]

On Thu, Apr 20, 2006 at 04:59:52PM -0700, Lu, Yinghai wrote:
> I have inited the hw sensors in LinuxBIOS for s2881. So you could use
> lmsensor to check the FAN speed. The config file is on Tyan web.
>
> For the Fan control, there should more some reg setting...to reduce the
> FAN speed winbond and adm1027...

An update on this.

As I reported earlier, setting fan speed works fine from LinuxBIOS. But it
seems this is only the case until I unplug the power from the machine.

When the machine is newly powered in, I need to boot up the proprietary bios
first to be able to _set_ the fan speed.

Once I've done that, I can reboot into LinuxBIOS, and set the speed just fine
using

/sys/bus/i2c/drivers/lm85/0-002d/pwm1

for instance like this:

echo '60' > /sys/bus/i2c/drivers/lm85/0-002d/pwm1

See attached set_fan_speed_ok.txt for a typical output.

I can shut the machine down, do a cold boot into LinuxBIOS, and it still
works.

Until I unplug the power to the machine, that is.

After that, I can still read out everything (see attached
set_fan_speed_bad.txt) correctly, but now also _set_ the fan speed.

Suggestions?

Thanks,
Ward.

[LinuxBIOS] s2881 fan control

[Ward Vandewege]

On Fri, May 05, 2006 at 04:57:17PM -0400, Ward Vandewege wrote:
> On Thu, Apr 20, 2006 at 04:59:52PM -0700, Lu, Yinghai wrote:
> > I have inited the hw sensors in LinuxBIOS for s2881. So you could use
> > lmsensor to check the FAN speed. The config file is on Tyan web.
> >
> > For the Fan control, there should more some reg setting...to reduce the
> > FAN speed winbond and adm1027...
> I can shut the machine down, do a cold boot into LinuxBIOS, and it still
> works.
>
> Until I unplug the power to the machine, that is.
>
> After that, I can still read out everything (see attached
> set_fan_speed_bad.txt) correctly, but now also _set_ the fan speed.

I'm sorry this is confusing. After unplugging and replugging the power to the
machine, I need to boot the proprietary BIOS once to be able to set the fan
speed on any consecutive (cold) boot.

Ward.

[LinuxBIOS] s2881 fan control

[Ward Vandewege]

On Fri, May 05, 2006 at 04:57:17PM -0400, Ward Vandewege wrote:
> On Thu, Apr 20, 2006 at 04:59:52PM -0700, Lu, Yinghai wrote:
> > I have inited the hw sensors in LinuxBIOS for s2881. So you could use
> > lmsensor to check the FAN speed. The config file is on Tyan web.
> >
> > For the Fan control, there should more some reg setting...to reduce the
> > FAN speed winbond and adm1027...
> I can shut the machine down, do a cold boot into LinuxBIOS, and it still
> works.
>
> Until I unplug the power to the machine, that is.
>
> After that, I can still read out everything (see attached
> set_fan_speed_bad.txt) correctly, but now also _set_ the fan speed.

I'm sorry this is confusing. After unplugging and replugging the power to the
machine, I need to boot the proprietary BIOS once to be able to set the fan
speed on any consecutive (cold) boot.

Ward.

[LinuxBIOS] s2881 fan control

[Stefan Reinauer]

* Ward Vandewege  [060505 23:19]:
> > I can shut the machine down, do a cold boot into LinuxBIOS, and it still
> > works.
> >
> > Until I unplug the power to the machine, that is.
> >
> > After that, I can still read out everything (see attached
> > set_fan_speed_bad.txt) correctly, but now also _set_ the fan speed.
>
> I'm sorry this is confusing. After unplugging and replugging the power to the
> machine, I need to boot the proprietary BIOS once to be able to set the fan
> speed on any consecutive (cold) boot.

Can you do an lspci -xxx on the machine while running LinuxBIOS after a
cold boot

a) without factory bios running once before
b) with factory bios running once before

and diff them to see if there's something we can easily do. (ie. in the 
"ACPI bridge" device?)

Stefan

[LinuxBIOS] s2881 fan control

[Ward Vandewege]

On Sat, May 06, 2006 at 02:25:21PM +0200, Stefan Reinauer wrote:
> Ward Vandewege  [060505 23:19]:
> > > I can shut the machine down, do a cold boot into LinuxBIOS, and it still
> > > works.
> > >
> > > Until I unplug the power to the machine, that is.
> > >
> > > After that, I can still read out everything (see attached
> > > set_fan_speed_bad.txt) correctly, but now also _set_ the fan speed.
> >
> > I'm sorry this is confusing. After unplugging and replugging the power to the
> > machine, I need to boot the proprietary BIOS once to be able to set the fan
> > speed on any consecutive (cold) boot.
>
> Can you do an lspci -xxx on the machine while running LinuxBIOS after a
> cold boot
>
> a) without factory bios running once before
> b) with factory bios running once before
>
> and diff them to see if there's something we can easily do. (ie. in the 
> "ACPI bridge" device?)

OK; I've attached the two files. Only three bytes are different in the ACPI
bridge (in this comparison the first file is after proprietary boot, the
second without).

0000:01:04.3 Bridge: Advanced Micro Devices [AMD] AMD-8111 ACPI (rev 05)
 00: 22 10 6b 74 00 00 80 02 05 00 80 06 00 40 00 00
 10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 20: 00 00 00 00 00 00 00 00 00 00 00 00 f1 10 81 28
 30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 40: 80 f1 00 07 00 00 00 00 20 14 50 00 00 00 00 00
 50: 00 80 00 00 0f 00 00 00 01 20 00 00 00 00 00 00
 60: 00 00 80 06 13 00 00 00 00 00 00 00 00 00 00 00
 70: 00 00 00 00 00 00 00 00 00 00 00 00 f1 10 81 28
 80: 00 00 00 00 01 00 00 00 00 00 00 00 00 00 00 00
 90: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
-f0: 19 5f 3b 00 00 00 00 00 00 00 00 00 00 00 00 00
+f0: c3 ad 1e 00 00 00 00 00 00 00 00 00 00 00 00 00

But there are actually quite a few other differences throughout the file,
which surprises me a litte. Lots of differences in the various northbridge
entries, and even the network cards and the sata controller. Is this normal?

Thanks,
Ward.

[LinuxBIOS] s2881 fan control

[Ward Vandewege]

(and now _with_ attachments... sigh)

On Sat, May 06, 2006 at 02:25:21PM +0200, Stefan Reinauer wrote:
> Ward Vandewege  [060505 23:19]:
> > > I can shut the machine down, do a cold boot into LinuxBIOS, and it still
> > > works.
> > >
> > > Until I unplug the power to the machine, that is.
> > >
> > > After that, I can still read out everything (see attached
> > > set_fan_speed_bad.txt) correctly, but now also _set_ the fan speed.
> >
> > I'm sorry this is confusing. After unplugging and replugging the power to the
> > machine, I need to boot the proprietary BIOS once to be able to set the fan
> > speed on any consecutive (cold) boot.
>
> Can you do an lspci -xxx on the machine while running LinuxBIOS after a
> cold boot
>
> a) without factory bios running once before
> b) with factory bios running once before
>
> and diff them to see if there's something we can easily do. (ie. in the 
> "ACPI bridge" device?)

OK; I've attached the two files. Only three bytes are different in the ACPI
bridge (in this comparison the first file is after proprietary boot, the
second without). Both times the system was running LinuxBIOS.

0000:01:04.3 Bridge: Advanced Micro Devices [AMD] AMD-8111 ACPI (rev 05)
 00: 22 10 6b 74 00 00 80 02 05 00 80 06 00 40 00 00
 10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 20: 00 00 00 00 00 00 00 00 00 00 00 00 f1 10 81 28
 30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 40: 80 f1 00 07 00 00 00 00 20 14 50 00 00 00 00 00
 50: 00 80 00 00 0f 00 00 00 01 20 00 00 00 00 00 00
 60: 00 00 80 06 13 00 00 00 00 00 00 00 00 00 00 00
 70: 00 00 00 00 00 00 00 00 00 00 00 00 f1 10 81 28
 80: 00 00 00 00 01 00 00 00 00 00 00 00 00 00 00 00
 90: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 a0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 b0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 c0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 d0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 e0: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
-f0: 19 5f 3b 00 00 00 00 00 00 00 00 00 00 00 00 00
+f0: c3 ad 1e 00 00 00 00 00 00 00 00 00 00 00 00 00

But there are actually quite a few other differences throughout the file,
which surprises me a litte. Lots of differences in the various northbridge
entries, and even the network cards and the sata controller. Is this normal?

Thanks,
Ward.

[LinuxBIOS] s2881 fan control

[Ward Vandewege]

On Tue, May 09, 2006 at 03:05:07PM -0400, Ward Vandewege wrote:
... lspci output ...
> But there are actually quite a few other differences throughout the file,
> which surprises me a litte. Lots of differences in the various northbridge
> entries, and even the network cards and the sata controller. Is this normal?

A followup on this. First of all, I didn't realize that the PCI configuration
space is dynamic - things change all the time.

Secondly;
> 0000:01:04.3 Bridge: Advanced Micro Devices [AMD] AMD-8111 ACPI (rev 05)

... snip ...
> -f0: 19 5f 3b 00 00 00 00 00 00 00 00 00 00 00 00 00
> +f0: c3 ad 1e 00 00 00 00 00 00 00 00 00 00 00 00 00

As it turns out, those three bytes change constantly, so I guess they are not
significant.

Also, I missed one byte that IS consistently different after a boot with the
proprietary BIOS:

0000:01:04.3 Bridge: Advanced Micro Devices [AMD] AMD-8111 ACPI (rev 05)
...
-40: 80 f1 00 04 00 00 00 00 20 14 50 00 00 00 00 00
+40: 80 f1 00 07 00 00 00 00 20 14 50 00 00 00 00 00

How would I go about changing that byte during the LinuxBIOS boot? Do I need
to edit resourcemap.c in src/mainboard/tyan/s2881?

Thanks,
Ward.

[LinuxBIOS] s2881 fan control

[Stefan Reinauer]

* Ward Vandewege  [060511 22:45]:
> On Tue, May 09, 2006 at 03:05:07PM -0400, Ward Vandewege wrote:
> ... lspci output ...
> > But there are actually quite a few other differences throughout the file,
> > which surprises me a litte. Lots of differences in the various northbridge
> > entries, and even the network cards and the sata controller. Is this normal?
>
> A followup on this. First of all, I didn't realize that the PCI configuration
> space is dynamic - things change all the time.

Its probably some timer that counts there. This kind of thing almost
requires having the datasheet sitting in another window or printed out
next to you.
> Also, I missed one byte that IS consistently different after a boot with the
> proprietary BIOS:
>
> 0000:01:04.3 Bridge: Advanced Micro Devices [AMD] AMD-8111 ACPI (rev 05)
> ...
> -40: 80 f1 00 04 00 00 00 00 20 14 50 00 00 00 00 00
> +40: 80 f1 00 07 00 00 00 00 20 14 50 00 00 00 00 00

Try the attached patch

[LinuxBIOS] s2881 fan control

[Ward Vandewege]

On Thu, May 11, 2006 at 10:59:56PM +0200, Stefan Reinauer wrote:
> Its probably some timer that counts there. This kind of thing almost
> requires having the datasheet sitting in another window or printed out
> next to you.

Right.
> > Also, I missed one byte that IS consistently different after a boot with the
> > proprietary BIOS:
> >
> > 0000:01:04.3 Bridge: Advanced Micro Devices [AMD] AMD-8111 ACPI (rev 05)
> > ...
> > -40: 80 f1 00 04 00 00 00 00 20 14 50 00 00 00 00 00
> > +40: 80 f1 00 07 00 00 00 00 20 14 50 00 00 00 00 00
>
> Try the attached patch

That didn't help. 0x43 was still set to 04, and I could not set the fan
speed. There was a bracket missing from the patch, btw.

Suggestions?
Ward.

[LinuxBIOS] s2881 fan control

[Stefan Reinauer]

* Ward Vandewege  [060511 23:50]:
> > > 0000:01:04.3 Bridge: Advanced Micro Devices [AMD] AMD-8111 ACPI (rev 05)
> > > ...
> > > -40: 80 f1 00 04 00 00 00 00 20 14 50 00 00 00 00 00
> > > +40: 80 f1 00 07 00 00 00 00 20 14 50 00 00 00 00 00
> >
> > Try the attached patch
>
> That didn't help. 0x43 was still set to 04, and I could not set the fan
> speed. There was a bracket missing from the patch, btw.

Looking at the datasheet of the 8111 this was a blind shot in the wrong
direction.

The bits 0:2 in this register are:

PPSTATE. Previous power state. Read-only. This field holds the most
previous power state from which the system came into the FON state. This
field resides on the VDD_COREX plane. Here are the possible states:

PPSTATE Power state
0h Reserved
1h POS power on suspend
2h C2
3h C3
4h MOFF mechanical off
5h STR suspend to RAM
6h STD suspend to disk
7h SOFF soft off

So LinuxBIOS reports mechanical off while factory bios reports soft off.
Interesting.

But I found some stuff that is worth following. Fans are controlled by
the IO mapped power management registers (PMxx) of the 8111. They are
called System Management I/O Mapped Registers in the specs on pages 
218 et sqq.

The base of the PMxx registers can be read from DevB:3x58 (32bit)
bits 15:8. (ie. pmbase=pci_read_dword(acpi_dev, 0x58) & 0xff00;)
See the aruma acpi code for more information. 
Iirc PMbase is 0xDD00 per default. So PM12 would require inb/outb to
0xdd12.

Also, there's GPIOs connected with FAN control: (pg 244 et sqq)
GPIO9  PMC9 FANCON1 08h (FANCON1 output) 1Xb N/A
GPIO10 PMCA FANRPM  08h (FANRPM input)   1Xb direct

And on the SMBus you have the following registers:
Fan Control Register PMF8 (Sounds like this belongs to the PMxx
registers)

Ok, now we need to glue this all together so that it works.

Stefan

[LinuxBIOS] some more tyan/s2881 questions

[Ward Vandewege]

Hi YH,

Have you made any progress on this? Not being able to control the fans is the
last thing that is stopping us from deploying LinuxBIOS. I'm worried that the
fans will wear out because they come on at full speed after boot, and I can't
control them under LinuxBIOS (except if I boot the proprietary BIOS once, and
then don't unplug the power to the machine before booting in LinuxBIOS).

Any help would be _greatly_ appreciated...

Thanks,
Ward.

On Thu, Apr 20, 2006 at 04:59:52PM -0700, Lu, Yinghai wrote:
> I have inited the hw sensors in LinuxBIOS for s2881. So you could use
> lmsensor to check the FAN speed. The config file is on Tyan web.
>
> For the Fan control, there should more some reg setting...to reduce the
> FAN speed winbond and adm1027...
>
> YH
>
> !DSPAM:444820c0151711016910990!