Re: [coreboot] Problem with the RTC

[Stefan Reinauer]

On 15.11.2008, at 11:47, Chris Lingard  wrote:
> coreboot-v2 has been giving the following error:
>
> PCI: 00:01.0 init
> set power on after power fail
> RTC Init
> Invalid CMOS LB checksum
> PNP: 002e.1 init
>
> I have put some debugging in rtc_checksum_valid of mc146818rtc.c to  
> output the variables, and got
>
> PU model AMD Athlon(tm) 64 X2 Dual Core Processor 6000+
> Setting up local apic... apic_id: 0x01 done.
> CPU #1 Initialized
> All AP CPUs stopped
> PCI: 00:18.0 init
> PCI: 00:01.0 init
> set power on after power fail
> RTC Init
> Sum f31c
> Old Sum f31c
> Sum cfe1
> Old Sum 2225
> Invalid CMOS LB checksum
> PNP: 002e.1 init
>
> This persists on reboot, and on cold start. I have no idea why the  
> second call to this procedure results in this, but think it is a  
> timing problem.

The first sum is the 'normal of bios checksum' while the second is an  
extended checksum.

You can correct it by changing a setting with nvramtool

Stefan
> This procedure is called twice from rtc_init, and I think that the  
> second call is wrong; so I changed the second one:
>    /* See if there is a LB CMOS checksum error */
>    /*  checksum_invalid = !rtc_checksum_valid(LB_CKS_RANGE_START,
>                    LB_CKS_RANGE_END,LB_CKS_LOC);  */
>    if(checksum_invalid)
>            printk_debug("Invalid CMOS LB checksum\n");
>            {
>            /* Make certain we have a valid checksum */
>            rtc_set_checksum(PC_CKS_RANGE_START,
>                    PC_CKS_RANGE_END,PC_CKS_LOC);
>            }
>
> checksum_invalid is already set, so the second call to  
> rtc_checksum_valid is commented out
>
> The check sum was being written back every time, right or wrong, so  
> I have put rtc_set_checksum within the failure bit.
>
> Chris Lingard