[LinuxBIOS] [PATCH] enable fidvid on asus a8v-e se

[Rudolf Marek]

Hello,

Following patch (fidvid.patch) adds support for FID/VID changes messages. Upon 
incoming SMAF message from CPU (C3 or FID/VID change), the SB will assert SLP# 
which is connected to LDTSTOP_L on K8 CPUs. Question is for how long. Imho for 
100us. Which is more than plenty (2us required) I will try to justify this once 
I know what bios to set in SB.

Signed-off-by: Rudolf Marek

Second patch just gives an example what needs to be in DSDT to make it actually 
work. Usually this part is stored in secondary DSDT table (SSDT), which is build 
in runtime in BIOS. The values are computed for the CPU used in the system and 
cannon be static. I dont know how to do the computations yet.

Rudolf

[Uwe Hermann]

On Fri, Nov 09, 2007 at 11:03:07PM +0100, Rudolf Marek wrote:
> Hello,
>
> Following patch (fidvid.patch) adds support for FID/VID changes messages. 
> Upon incoming SMAF message from CPU (C3 or FID/VID change), the SB will 
> assert SLP# which is connected to LDTSTOP_L on K8 CPUs. Question is for how 
> long. Imho for 100us. Which is more than plenty (2us required) I will try 
> to justify this once I know what bios to set in SB.
>
> Signed-off-by: Rudolf Marek

Thanks, r2965.

I only committed the first patch for now; is the second one ready for
being committed too or is it work in progress? Does it work?
> Index: mainboard/asus/a8v-e_se/cache_as_ram_auto.c
>
> --- mainboard/asus/a8v-e_se/cache_as_ram_auto.c	(revision 2953)
> +++ mainboard/asus/a8v-e_se/cache_as_ram_auto.c	(working copy)
> @@ -309,6 +309,12 @@
>  #endif
>    init_timer();
>    ht_setup_chains_x(sysinfo);	// it will init sblnk and sbbusn, nodes, sbdn
>
> +#if K8_SET_FIDVID == 1
>    enable_fid_change();
>    init_fidvid_bsp(bsp_apicid);
>
> +#endif

I dropped the '#if' part, the whole file where those two functions are
defined is placed in a '#if K8_SET_FIDVID == 1'. Either way, your board
sets K8_SET_FIDVID to 1, so it doesn't really matter.

Uwe.

[Rudolf Marek]

> I only committed the first patch for now; is the second one ready for
> being committed too or is it work in progress? Does it work?

Yes this is OK. Second is just for illustration. The LB should construct the 
bytecode of the second patch in runtime, because the content must be changed 
when processor is changed.

This code in second patch should be placed as "external symbols" in SSDT table. 
Putting them into the DSDT will work, but user is forced to grab correct values 
specific for his/her processor. This patch is for Opteron 175 CPUID 00020f32
I dont know yet how to construct those tables at runtime, nor I'm not sure how 
to create valid data for all processors.
> I dropped the '#if' part, the whole file where those two functions are
> defined is placed in a '#if K8_SET_FIDVID == 1'. Either way, your board
> sets K8_SET_FIDVID to 1, so it doesn't really matter.

Ok. Thanks,

Rudolf