[patch] ACPI support on M57SLI

[Harald Gutmann]

Hello,
the result of the past two/three days work/interest on coreboot is a big part 
of ACPI support on M57SLI which works on my revision (v2).

Soft-Power-Off works, and that's what i missed really hard in coreboot for 
m57sli.

The other parts of ACPI support should work (mostly) also fine. (except setting 
up MMCONFIG in acpi_fill_mcfg, and the powernow features which hopefully can be 
done with the recently commited patch "AMD boards to use new Power Now 
infrastructure" from Rudolf Marek).

I think i'll get the time this weekend or at the beginning of the next week to 
address those remaining problems.
In the mean time i hope that someone can test the code which i didn't 
originally write, but modified, to work at M57SLI.

For myself I own an M57SLI with hardware rev.2, maybe Ward can test if that 
code will work on v1 too.
One other "problem" with the actual code is, that most of the interesting 
parts are hardcoded (numbers where variables should be/could be).

The attached patch is quite complete, except the problems mentioned above and 
the copyright informations are missing in the patch.
This one is just for testing, and i'll try to get the remaining problems fixed, 
and i also will add copyright information when the work is finally done.

I think it's not the time to add this patch into svn because of the above 
mentioned missing parts. But it's a great time to test and enjoy this patch.

Kind regards,
Harald Gutmann

[Peter Stuge]

Harald Gutmann wrote:
> the result of the past two/three days work/interest on coreboot is
> a big part of ACPI support on M57SLI which works on my revision
> (v2).
>
> Soft-Power-Off works, and that's what i missed really hard in
> coreboot for m57sli.

Good work getting that going!

I don't know if we want to take in the factory DSDT though..

//Peter

[Harald Gutmann]

On Friday 06 March 2009 20:45:33 Peter Stuge wrote:
> Harald Gutmann wrote:
> > the result of the past two/three days work/interest on coreboot is
> > a big part of ACPI support on M57SLI which works on my revision
> > (v2).
> >
> > Soft-Power-Off works, and that's what i missed really hard in
> > coreboot for m57sli.
>
> Good work getting that going!
>
> I don't know if we want to take in the factory DSDT though..

As you've seen correctly it's the factory DSDT which i created using acpidump, 
acpixtract and iasl.
I don't know if it would be a license problem to take that in, but it works, 
and as told above this patch shouldn't be committed right now.
> //Peter

Harald

[ron minnich]

On Fri, Mar 6, 2009 at 11:45 AM, Peter Stuge  wrote:
> I don't know if we want to take in the factory DSDT though..

I did not read the patch. Factory DSDT is in it?

ron

[Harald Gutmann]

On Friday 06 March 2009 21:09:42 ron minnich wrote:
> On Fri, Mar 6, 2009 at 11:45 AM, Peter Stuge  wrote:
> > I don't know if we want to take in the factory DSDT though..
>
> I did not read the patch. Factory DSDT is in it?

yes, it's in it. created with acpidump and acpixtract.
the file, which factory bios gives with that tools was directly used in the 
patch.
like said before, i don't know if this is a licensing problem or not. and also 
like stated before, this patch is not suggested for commiting.
It works, and it shouldn't be much work to get the rest of the missing acpi 
features done. (until now, i didn't really care about dsdt.dsl, but i tried to 
get acpi_tables.c and fadt.c working.)
> ron

regards Harald

[Florentin Demetrescu]

Hi Harold,
Great work!
I will try to test your patch on my board this WE.
I will keep you informed about the results
 Florentin

Quoting Harald Gutmann :
> On Friday 06 March 2009 21:09:42 ron minnich wrote:
> > On Fri, Mar 6, 2009 at 11:45 AM, Peter Stuge  wrote:
> > > I don't know if we want to take in the factory DSDT though..
> >
> > I did not read the patch. Factory DSDT is in it?
>
> yes, it's in it. created with acpidump and acpixtract.
> the file, which factory bios gives with that tools was directly used in the
> patch.
> like said before, i don't know if this is a licensing problem or not. and
> also
> like stated before, this patch is not suggested for commiting.
> It works, and it shouldn't be much work to get the rest of the missing acpi
> features done. (until now, i didn't really care about dsdt.dsl, but i tried
> to
> get acpi_tables.c and fadt.c working.)
> > ron
>
> regards Harald

[Peter Stuge]

Harald Gutmann wrote:
> > > I don't know if we want to take in the factory DSDT though..
> >
> > I did not read the patch. Factory DSDT is in it?
>
> yes, it's in it. created with acpidump and acpixtract.
> the file, which factory bios gives with that tools was directly
> used in the patch.
> like said before, i don't know if this is a licensing problem or
> not.

We want to avoid the issue by not including factory parts without
express permission, though.
> and also like stated before, this patch is not suggested for
> commiting.

Yup. I'm looking forward to more test results in any case! :)

//Peter

[ron minnich]

On Fri, Mar 6, 2009 at 12:24 PM, Harald Gutmann  wrote:
> like said before, i don't know if this is a licensing problem or not. and also
> like stated before, this patch is not suggested for commiting.

Understand. Yes, we certainly don't want to put a copy of factory ACPI
in our SVN UNLESS: someone wants to contact the manufacturer and gets
permission.

If they do get permission, I suggest that a copy of the email granting
permission be dropped in the mainboard directory. I have had
experience with the following:
1. Vendor: please do this port.
2. We do the port.
(years pass)
(people move around)
3. Vendor: WHY IS OUR MAINBOARD IN YOUR SOURCE TREE? PLEASE REMOVE!
   (we didn't :-)

So you have to be careful once you get the ok ... either get ACPI with
a GPL-from-vendor note in it or leave a nice audit trail, preferably
with a copy of the email in comments.

AMD, as always, is one example of how to get this done correctly.

ron

[Harald Gutmann]

Hello once again,

as i love it that much to talk with myself, i was wondering while 
working/creating ACPI stuff, why HPTE support is turned off in mcp55_lpc.c with 
using #if 0.

Has maybe someone of you an idea why that was done? (HPTE is also in context 
with ACPI)

Here there is the code snippet i refer to:

#if 0
static void enable_hpet(struct device *dev)
{
        unsigned long hpet_address;

pci_write_config32(dev,0x44, 0xfed00001);
        hpet_address=pci_read_config32(dev,0x44)& 0xfffffffe;
        printk_debug("enabling HPET @0x%x\n", hpet_address);
}
#endif

Ideas?

Kind Regards,
Harald Gutmann

PS: This was not mentioned in the first post on the list, but HPTE support 
should be also included into ACPI to get it full working. I just missed it in 
the first post.

[Myles Watson]

> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]]
> On Behalf Of Harald Gutmann
> Sent: Friday, March 06, 2009 12:50 PM
> To: [e-mail]
> Subject: Re: [coreboot] [patch] ACPI support on M57SLI
>
> Hello once again,
>
> as i love it that much to talk with myself, i was wondering while
> working/creating ACPI stuff, why HPTE support is turned off in mcp55_lpc.c
> with
> using #if 0.
>
> Has maybe someone of you an idea why that was done? (HPTE is also in
> context
> with ACPI)

I don't know why HPET is disabled in mcp55.  It was also disabled in
ck804_lpc.c.  It worked fine when I enabled it.

Thanks,
Myles

[Harald Gutmann]

On Friday 06 March 2009 20:50:03 Myles Watson wrote:
> > -----Original Message-----
> > From: [e-mail]
> > [mailto:[e-mail]] On Behalf Of Harald Gutmann
> > Sent: Friday, March 06, 2009 12:50 PM
> > To: [e-mail]
> > Subject: Re: [coreboot] [patch] ACPI support on M57SLI
> >
> > Hello once again,
> >
> > as i love it that much to talk with myself, i was wondering while
> > working/creating ACPI stuff, why HPTE support is turned off in
> > mcp55_lpc.c with
> > using #if 0.
> >
> > Has maybe someone of you an idea why that was done? (HPTE is also in
> > context
> > with ACPI)
>
> I don't know why HPET is disabled in mcp55.  It was also disabled in
> ck804_lpc.c.  It worked fine when I enabled it.

Until now i didn't try to activate HPTE in mcp55 because HPTE was not 
necessary to get ACPI running, but i was wondering why it's disabled.
Thanks for sharing your experience on ck804, i'll just try to enable it.
> Thanks,
> Myles

Kind regards,
Harald

[Ward Vandewege]

Hi Harald,

Awesome! Great work.

I tested on an m57sli v1.

Boot log:

http://ward.vandewege.net/coreboot/m57sli/m57sli-hargut-acpi.log

ACPI is detected, etc, but soft power off does not work - the machine almost
powers off, but then immediately resets.

Thanks,
Ward.

[Harald Gutmann]

On Friday 06 March 2009 22:42:54 Ward Vandewege wrote:
> Hi Harald,

Hello Ward,

please try to post the following outputs/files:
acpidump from original bios.
acpidump from coreboot with my patches.

As mentioned some posts before all values of my patch are hardcoded, therefore 
it can easily be it won't work on v1.
With the informations of the acpidump files of original and coreboot for v1, it 
might be easier for me to create the code so that it's not hard-coded any 
more.

On irc you mentioned that you have both versions of m57sli, please be so 
kindly and test this patch on v2 too. (and if possible send acpidump from 
proprietary bios and coreboot too.)
> Awesome! Great work.
>
> I tested on an m57sli v1.
>
> Boot log:
>
> http://ward.vandewege.net/coreboot/m57sli/m57sli-hargut-acpi.log

thanks for the boot-log, but the problem for me to verify that boot-log is, 
that i can't check out my own boot-log atm, because if've no hardware to do 
that.
> ACPI is detected, etc, but soft power off does not work - the machine
> almost powers off, but then immediately resets.

Immediately resets, strange but it could maybe fixed easily with knowing the 
differences between your original bios acpi-tables and coreboot-acpi tables on 
hardware rev.1.
> Thanks,
> Ward.

Regards,
Harald