Re: [Libreboot-dev] set vram on gm45 to 352M

[Arthur Heymans]

Denis 'GNUtoo' Carikli <address@hidden> writes:

> Hi,
>
> Here I'll assume that the CMOS code works well or get fixed if it
> doesn't.
>
> On Thu, 19 May 2016 16:50:16 +0200
> Arthur Heymans <address@hidden> wrote:
>
>> there are 2 good options to fix this:
>> - make the flash script reset cmos
> * If the user is flashing libreboot but is running the proprietary boot
>   fimrware(BIOS or UEFI) the CMOS should be detected as corrupt in the
>   romstage. That can be used to set proper defaults in that case, like
>   in src/ec/lenovo/h8, where we have:
>   > if (get_option(&val, "wlan") != CB_SUCCESS)
>   >         val = 1;
>   > h8_wlan_enable(val);
Something similar is already done. I just made extra sure that no
 invalid number can be used. This is most likely not necessary but if the
 number is to high the system will likely not boot. I did this because I
 was confused by user reports with not working displays and thought that
 it was cmos. It was 352M mode that has problems with some dimm configurations.
I removed the 352M option.
> * If the user is flashing libreboot while running coreboot or
>   libreboot, you could do some checks after flashing:
>   For instance you could compare the layout of the image you flashed
>   with the one currently in use, to determine if the new options need to
>   be set to good default.
>   I'm not sure what state is the CMOS in such case. Would it be
>   considered corrupt? If so it would reset coreboot/libreboot's
>   defaults.
>
>
>> - make coreboot always load cmos.defaults (a .config option)
> Probably not a good idea.
>
Leah decided to do this. It is possible the change cmos.defaults using nvramtool
which is now documented. I think this solution is good for libreboot. It
makes sure sane defaults are used and documents how users can change
them if they so desire.
This is also how the current libreboot stable is configured.
>> and one ugly way:
>> hardcoding this.
> Instead, try to see if the CMOS handling works fine when the flash is
> corrupt.
>
> Denis.

Thanks for your comments.

-- 
Arthur Heymans