AMD S3: Experiences with ACPI 2.0 method _BFS (Back From Sleep)

[Paul Menzel]

Dear coreboot folks,

searching for ACPI in the output of `dmesg` I noticed the following
message.

$ uname -v
        #1 SMP Debian 3.2.35-2
        $ dmesg | grep -i acpi
        [    0.154063] ACPI: BIOS offers _BFS
        [    0.154067] ACPI: If "acpi.bfs=1" improves resume, please notify [e-mail]

Doing

$ git grep _GTA

reveals that it is only implemented in the DSDT of AMD based boards. As
resuming currently does not work on the ASRock E350M1 [1], I cannot test
that.

Was this ever tested by adding `acpi.bfs=1` to the Linux kernel command
line and what was the outcome? A test on AMD Persimmon would be very
interesting.

It looks like support for this has been removed in the meantime though

commit 3f6f49c7854c9294119437a82c5b35be78f9cea6
        Author: Len Brown 
        Date:   Thu Jul 26 20:08:54 2012 -0400

ACPI: delete _GTS/_BFS support

since Linux Git tag v3.6-rc2.

Thanks,

Paul

[1] http://www.coreboot.org/ASRock_E350M1
[2] https://patchwork.kernel.org/patch/1246671/

AMD S3: Experiences with ACPI 2.0 method _BFS (Back From Sleep)

[Paul Menzel]

Dear coreboot folks,

Am Montag, den 18.02.2013, 17:09 +0100 schrieb Paul Menzel:
> searching for ACPI in the output of `dmesg` I noticed the following
> message.
>     $ uname -v
>     #1 SMP Debian 3.2.35-2
>     $ dmesg | grep -i acpi
>     [    0.154063] ACPI: BIOS offers _BFS
>     [    0.154067] ACPI: If "acpi.bfs=1" improves resume, please notify [e-mail]
>
> Doing
>
> $ git grep _GTA
>
> reveals that it is only implemented in the DSDT of AMD based boards.

Looking at the actual “implementation” it looks like it is not
implemented at all (see for AMD Thatcher’s DSDT) [1]. ;-)

$ more src/mainboard/amd/parmer/dsdt.asl
        […]
        	/*
        	*	_BFS OEM Back From Sleep method
        	*
        	*	Entry:
        	*		Arg0=The value of the sleeping state S1=1, S2=2
        	*
        	*	Exit:
        	*		-none-
        	*/
        	Method(_BFS, 1) {
        		/* DBGO("\_BFS\n") */
        		/* DBGO("From S") */
        		/* DBGO(Arg0) */
        		/* DBGO(" to S0\n") */
        	}
        […]

Would a patch be useful to remove that from all of the AMD DSDTs as Len
Brown wrote they requested to remove that from the next ACPI
specification?

For the curious, the method _BFS is documented in

7.3.1 _BFS (Back From Sleep)

of the ACPI 4.0a specification [
> As resuming currently does not work on the ASRock E350M1 [1], I cannot
> test that.
>
> Was this ever tested by adding `acpi.bfs=1` to the Linux kernel command
> line and what was the outcome? A test on AMD Persimmon would be very
> interesting.
>
> It looks like support for this has been removed in the meantime though
>     commit 3f6f49c7854c9294119437a82c5b35be78f9cea6
>     Author: Len Brown <[e-mail]>
>     Date:   Thu Jul 26 20:08:54 2012 -0400
>
>         ACPI: delete _GTS/_BFS support
>
> since Linux Git tag v3.6-rc2.

Thanks,

Paul
> [1] http://www.coreboot.org/ASRock_E350M1
> [2] https://patchwork.kernel.org/patch/1246671/

[3] http://review.coreboot.org/gitweb?p=coreboot.git;a=blob;f=src/mainboard/amd/...
[4] http://www.acpi.info/spec40a.htm

AMD S3: Experiences with ACPI 2.0 method _BFS (Back From Sleep)

[Marc Jones]

On Mon, Feb 18, 2013 at 9:24 AM, Paul Menzel
 wrote:
> Dear coreboot folks,
>
> Am Montag, den 18.02.2013, 17:09 +0100 schrieb Paul Menzel:
> > searching for ACPI in the output of `dmesg` I noticed the following
> > message.
> >     $ uname -v
> >     #1 SMP Debian 3.2.35-2
> >     $ dmesg | grep -i acpi
> >     [    0.154063] ACPI: BIOS offers _BFS
> >     [    0.154067] ACPI: If "acpi.bfs=1" improves resume, please notify [e-mail]
> >
> > Doing
> >
> > $ git grep _GTA
> >
> > reveals that it is only implemented in the DSDT of AMD based boards.

The kernel sees the ACPI method, but doesn't do anything with it. If
you would like the kernel to use the method, pass acpi.bfs=1.
> Looking at the actual “implementation” it looks like it is not
> implemented at all (see for AMD Thatcher’s DSDT) [1]. ;-)

...
> > Was this ever tested by adding `acpi.bfs=1` to the Linux kernel command
> > line and what was the outcome? A test on AMD Persimmon would be very
> > interesting.

Why would this be interesting? As noted, the method is present, but
doesn't do anything. Some debug information could be enabled if the OS
called the function.
> > It looks like support for this has been removed in the meantime though
> >     commit 3f6f49c7854c9294119437a82c5b35be78f9cea6
> >     Author: Len Brown <[e-mail]>
> >     Date:   Thu Jul 26 20:08:54 2012 -0400
> >
> >         ACPI: delete _GTS/_BFS support
> >
> > since Linux Git tag v3.6-rc2.

The function is part of the spec up to 4.0 and Len recommends that it
be remove it in the future revisions.

It seems safe to remove it, but it isn't causing harm and is part of
the spec as implemented by coreboot.

Marc