Hello all,
With A LOT of empirical testing, I have gotten MILO to work from the
UDB/Multia flash (finally!).
There were a couple of snags along the way which I will summarize:
1. The SROM requires a 64 byte flash image header and will accept the
same header as the NoName machine.
2. The SROM requires that the flash ROM image be exactly a multiple of
256K and have a final byte which is the checksum of the all the preceding
bytes absent the header. The algorithm is:
unsigned char checksum;
checksum = flash_image[0];
for( ii = 1; ii < flash_image_size; ii++ )
{
left_rotate_1_bit( checksum );
checksum += flash_image[ii];
}
return checksum;
3. The SROM apparently expects the first 2 flash ROMs to be used as a
single entity ("device" srmflash). When I was trying to blast a <256K MILO
which worked on the failsafe floppy, the LED in the back would flash an
undocumented (that I could find anyway) error code. As soon as I used both
flash ROMs (even though the second was all zeros (except for the final
checksum byte)), it worked!
OK, I am now ready to follow the recipe as recommended by Eric:
>>Do you have a working milo bootloader? If so all of the hard lifting
>>has been done and you only need to rearrange and test the code.
>>
>>The boot order for alpha should be:
>>SROM (which does memory setup) -> (linuxBIOS code in flash rom) -> Linux
kernel.
>>
>>The linuxBIOS code will need to do the following things:
>>load palcode.
>>Setup some devices??
>>jump to linux.
>>
>>What you will end up with is pretty much milo on steroids. :)
So it's off to study the boot process to see where surgery is needed.
Regards,
Bill
