[coreboot] SST25VF016B (2MB) flash on m57sli (IT8716F).
rminnich at gmail.com
Fri Jan 18 08:38:45 CET 2008
On Jan 17, 2008 2:32 PM, Ronald Hoogenboom <ronald at zonnet.nl> wrote:
> Now I'm looking at how to make the rom_stream read the flash chip like
> in the over512k_read_chip.
> But I'm a bit stuck on the overal mechanics of the initialization
> process and how to get hold of the assigned io port for the SPI serial
> flash controller in the IT8716.
> This is supposed to be configured by the GPIO Config registers (LDN 07)
> at index 0x64 and 0x65, but in the pnp_dev_info array in
> it8716f/superio.c, there is nothing listed after IT8716F_GPIO, while at
> other indexes, there is. Does this mean that the superio.c file fails to
> configure this io port?
yes, it means that the port is not configured. There is a lot of TODO
in this file.
It may be time to fill it in.
> Suppose that this port IS configured correcly, what means should my
> rom_stream use to retrieve this port address?
I think you make a new stream type, let's called it bad_stream, that knows
how to handle the addresses. The bad_stream would know that when the
address is in a certain range, i.e. below the address (0-512KB), it
will use I/O, and when it is above, it can use memory access.
So in src/stream/Config.lb
you add something like this:
> Is there some doc about these pnp functions?
no, sorry. We are doing better in v3 ... we know the lack of docs is a problem.
> Suppose I've figured it all out and created my new rom_stream_over512k
> (or so) object, how do I mingle it into the Config.lb files, so it will
> automatically get chosen when a ROM_SIZE of over 512k is selected in the
> target configuration (AND the superio is an IT8716F).
I think you make those boards always use bad_stream (or whatever) in all cases,
and not try to get the config tool to pick different types, as in the
But, it might be possible to do this:
if ROM_SIZE > 512*1024
Something like this might work.
The it8761f is NOT the worst thing we have seen, it's just really, really bad.
So I guess we have to put in another kludge for badly design chipsets.
It's not the first time :-)
> In the targets/gigabyte/m57sli directory, there is a Config.lb.kernel
> file, which I used as a base for my build, but it only uses a 'fallback'
> and a 'failover' image. Why no 'normal' one? What is supposed to be the
> purpose of each image? (You see, I'm really missing some key parts of
> the big picture...)
you're not missing anything. The picture, which was once, years ago,
pretty clean, has gotten grafitti-tagged by many authors and the
necessities of strange hardware.
This will make no sense, but I'll try. Fallback is the image you can
ALWAYS use, even when normal does not work. Normal is an image you can
use unless something breaks, and you use fallback.
So, what's the one image you must always have? fallback. What's the
optional image? normal. So, when there is only going to be one image,
it's called fallback.
What people don't realize is that the images can have any name, so,
for instance, on single image ROMs, they should really not use the
name 'fallback' -- there is nothing sacred about the names. 'single'
failover -- that snuck in at some point and I think YH Lu understands why.
This is a bit messy as the guys who first put the fallback/normal in
never intended for bioses to be built WITHOUT fallback/normal. I
relied heavily on the fallback stuff at LANL but have since hardly
used it. Not all people think it is needed -- OLPC for example has
explicitly rejected a need for a fallback bios image. It's an open
question as to whether fallback bios images have real value for
Let me know how your new stream code goes as I am going to need it, I think.
More information about the coreboot