[coreboot] Gigabyte M57SLI - 4MB Chip - LAB doesn't boot like expected

Carl-Daniel Hailfinger c-d.hailfinger.devel.2006 at gmx.net
Sat May 24 14:22:43 CEST 2008


On 24.05.2008 13:00, Harald Gutmann wrote:
> Am Freitag, 23. Mai 2008 20:23:11 schrieb Carl-Daniel Hailfinger:
>   
>> Try this:
>> Report actual and wanted LZMA decompression scratchpad size:
>> Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006 at gmx.net>
>>     
>
> The patch worked, i had to set the decoder scratchpad size to 50335340, which 
> seems quite much to me, but there was no more error on "Uncompressing to 
> RAM".
>   

The maximum scratchpad size right now is ~512k. Anything above that will
corrupt memory.

> Acked-by: Harald Gutmann <harald.gutmann at gmx.net>
> To add that also into the coreboot v2 reprository, and not just in v3.
>
> But booting with LAB didn't work, the last lines which i get for now are:
> "rom_stream: 0xffc00000 - 0xfffe7fff
> Uncompressing to RAM 0x0100000"
>
> Maybe, as i set the scratchpad size that high, it takes really long to 
> decompress the image to the RAM. I waited about 2 minutes and nothing more 
> happened, i think that should be enaugh time to decompress 4MB into the RAM.
>
> Any idea what is wrong with my image, or what could be done to get debug 
> informations?
>   

The scratchpad has to fit in a special area below 1 MB. Considering
various other constraints, the maximum size is probably around 512k. A
bigger scratchpad will clobber interrupt vectors and wrap around to the
top of the 4 GB address space. The decompressor will then write to the
area occupied by ROM (writes discarded) and non-RAM areas (writes
discarded). That's a bad thing.

If a compressed payload needs more than 512k scratchpad, the payload is
most likely corrupt.


Regards,
Carl-Daniel




More information about the coreboot mailing list