[LinuxBIOS] cmos.layout field meanings.

[Steve Isaacs]

For my design review I'm trying to document the meaning of the various
fields in CMOS RAM. I've done some grepping and found limited use of the
generated CMOS_VSTART_ and CMOS_VLEN_ definitions which are generated
from the board's cmos.layout.

Also, I've found a number of inconsistencies regarding checksum
calculation. For example CMOS_VSTART_check_sum is defined in cmos.layout
as 984 for some (most?) boards but the file pc80/mc146818rtc.h defines
the checksum range as bytes 16 (PC_CKS_RANGE_START) thru 45
(PC_CKS_RANGE_END) which is then used by the rtc_init to verify CMOS RAM
contents. Why wouldn't cmos.layout be used for both?

I'm sensing some bit position dependent coupling between execution in
CAR and execution from RAM using different definitions which may cause
problems if cmos.layout is changed for a board. It appears there may be
some additional coupling for bootloaders as well? (e.g. boot_first,
boot_second, etc.)

Is there a document I can read to help me understand this? Am I looking
at this wrong?

Thanks,

Steve

[ron minnich]

Steve, I am afraid few people really understand all the ins and outs
of the cmos bits, and most don't really use them. We used them
somewhat at LANL, but the real usage was by linux networx on their
cluster nodes. I would say we used maybe 2 bits at most :-), and LNXI
used just about all there were to use.

So, you are now the expert ... :-)

thanks

ron

[Steve Isaacs]

On Tue, 2007-12-18 at 17:05 -0800, ron minnich wrote:
> Steve, I am afraid few people really understand all the ins and outs
> of the cmos bits, and most don't really use them. We used them
> somewhat at LANL, but the real usage was by linux networx on their
> cluster nodes. I would say we used maybe 2 bits at most :-), and LNXI
> used just about all there were to use.
>
> So, you are now the expert ... :-)

ho-boy, blind leading the blind ... Ok, I'll accept the challenge. I'll
offer up any information I discover in the process.

The bits I'm most interested in at this time are the ones that payloads
such as FILO, etherboot may depend upon. Anybody have any clues?

Steve

[Peter Stuge]

On Wed, Dec 19, 2007 at 01:03:51PM -0800, Steve Isaacs wrote:
> The bits I'm most interested in at this time are the ones that
> payloads such as FILO, etherboot may depend upon. Anybody have
> any clues?

Sorry, best to just dig into the code. At least FILO is easy to grep.

//Peter