[U-Boot-Users] RE: u-boot patches for BUBINGA

[Bill Hargen]

> From: Bruce Boettjer [mailto:bruceb at momenco.com]
> Sent: Friday, December 12, 2003 6:51 PM
> To: 'Bill Hargen'; u-boot-users at lists.sourceforge.net
> Subject: RE: u-boot patches for BUBINGA
> 
> Just a quick note... The memory config on the 405EP Eval 
> board is still
> incorrect.  The Physical DIMM is 64-bit, but the memory bus 
> is 32-bit...
> therefore, only half of the memory gets used...
> 
> In spd_sdram.c, I added a line after 354 that divided 
> 'total_size' by two.
> This appears to do the init correctly.

My understanding of the Bubinga design is that the 64-bit wide
memory DIMM gets used as two banks of 32-bits each.  There is
some discussion of this in the manual for the eval board.  My
board has a DIMM marked as 128MB and u-boot is reporting the
same and using it as two 64MB banks.  However, the manual is
carefull to not promise exactly what memory DIMM is shipped
with the board, so your system could be different.  But still,
the basic design is to use a 64-bit DIMM as two 32-bit banks.
I'm assuming that the SDRAM SPD code is doing the proper math
to make the sizing work.  I didn't have to do a divide-by-two.
My only problem was that the code was using banks 0 and 2
instead of 0 and 1.  (Worked fine right up to the point where
u-boot jumps to its RAM copy at the top of the second bank!)

> BTW-Is anyone able to get a uImage built kernel to boot off a 
> 405EP Eval
> board?  I'd love to know what version... My kernel dies 
> pretty early...

I'm using the MontaVista 3.0 package.  They released a new LSP
for the Bubinga during the summer.  It didn't work because it
was assuming that the board information structure passed from
the boot loader was the format used by IBM's boot code.  u-boot
is different - similar stuff like memory size, but at different
structure offsets.  I had to change it to expect the u-boot
structure.  Everything mapped OK except for the "PLLOUTA" clock
speed that used used to compute serial port baud rate clocks.
I recomputed it based on another clock speed and some clock
divisor register settings.

The only strange thing was that I didn't see any code in the
MontaVista package to deal with the PPCBoot/u-boot structure.
I would have expected some of the PPC Linux ports to use this
structure.  Perhaps MontaVista has stripped out such stuff from
the LSP.  This also may be addressed in Wolfgang's u-boot manual,
and probably wouldn't have been an issue if I was using his ELDK.

--Bill