[U-Boot] ppc440 (440EPx sequoia) - inhibiting the cpu cache?

[U-Boot] ppc440 (440EPx sequoia) - inhibiting the cpu cache?

[Jonathan Barrow]

Hi,

I'm in the process of bringing up a custom board featuring AMCC's
440EPx processor.  I'm using the uboot-2008-10 source tree and
"sequoia" board for my starting configuration, also using ELDK v4.2.
for cross-compiler tools and such.

Although my board initializes and completes its memory tests
"relocate_code" fails.  I've noted in the FAQ and mailing lists the
sensitive nature of sdram timing and have spent considerable effort
making sure the ddr_xx settings are correct.  I thought my latest
adjustment would save the day,  I realized 'BSTLEN' should likely be
changed from 4 words to 8 words because my memory configuration is x32
not x64.  Unfortunately my latest u-boot.bin is still not happy.

To be really really certain my problem lies elsewhere, for the
purposes of hardware bringup  I was wondering if there was a
relatively easy way to inhibit memory bursts, by turning off the CPU
Cache.   I searched many of the the .h files and didnt' notice
anything along the lines of EN or DISABLE the cache,  only size
adjustments and associations to Chip Selects.  If I overlooked
something I'd would really appreciate a tip on strategy getting past
relocate_code.  This is what I have so far:

U-Boot 2008.10 (Jan  8 2009 - 14:05:28)

CPU:   AMCC PowerPC 440EPx Rev. A at 264 MHz (PLB=88, OPB=44, EBC=22 MHz)
      Security/Kasumi support
      Bootstrap Option H - Boot ROM Location I2C (Addr 0x52)
, PCI async ext clock used       32 kB I-Cache 32 kB D-Cache
Board: Raptor1K - Avtec Systems  PPC440EPx ACU, Rev. 0, PCI=33 MHz
I2C:   ready
DRAM:  128 MB
AB
reloc addr_sp: 7ea9f18 id: 7ea9f30  addr: 7fac000



Sincerely,
J.J. Barrow

[U-Boot] ppc440 (440EPx sequoia) - inhibiting the cpu cache?

[Stefan Roese]

Hi Jonathan,

On Thursday 08 January 2009, Jonathan Barrow wrote:
> I'm in the process of bringing up a custom board featuring AMCC's
> 440EPx processor.  I'm using the uboot-2008-10 source tree and
> "sequoia" board for my starting configuration, also using ELDK v4.2.
> for cross-compiler tools and such.
>
> Although my board initializes and completes its memory tests
> "relocate_code" fails.  I've noted in the FAQ and mailing lists the
> sensitive nature of sdram timing and have spent considerable effort
> making sure the ddr_xx settings are correct.  I thought my latest
> adjustment would save the day,  I realized 'BSTLEN' should likely be
> changed from 4 words to 8 words because my memory configuration is x32
> not x64.  Unfortunately my latest u-boot.bin is still not happy.
>
> To be really really certain my problem lies elsewhere, for the
> purposes of hardware bringup  I was wondering if there was a
> relatively easy way to inhibit memory bursts, by turning off the CPU
> Cache.   I searched many of the the .h files and didnt' notice
> anything along the lines of EN or DISABLE the cache,  only size
> adjustments and associations to Chip Selects.

Usually caches are disabled on 44x platforms in U-Boot. So you shouldn't see 
any burst access to SDRAM anyway.

> If I overlooked 
> something I'd would really appreciate a tip on strategy getting past
> relocate_code.  This is what I have so far:

Most likely still an SDRAM configuration problem. Did you check if SDRAM is 
working after initialization either using a debugger or by using some simple 
memory test (write value to SDRAM and read it back and print the result)? 
This is what I would recommend to do. Only if this works you can continue to 
run U-Boot from RAM.

Best regards,
Stefan

=====================================================================
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: +49-8142-66989-0 Fax: +49-8142-66989-80  Email: office at denx.de
=====================================================================