Dear Simon,
Am 26.08.2011 12:10, schrieb Simon Schwarz:
Dear Andreas,
<snip>
BTW while reading this patch I got another idea to solve problem 'how get we saved the boot information to<storage>'. The required information regardless of whether it is ATAGS or FDT is only a blob at some place in ram after the 'bootm x' commands used in 'Add savebp command'. Saving a blob from location X with size Y to location Z is easy and already implemented.
So the only required thing is to get the 'blob' prepared in RAM. In my opinion this could be a subcommand of bootm instead of a new command.
How about:
---8<--- # bootm savebp ...done boot information is @0x80000100 with size 0x100 # nandecc hw # nand erase ... # nand write 80000100 ... --->8---
This was actually my first implementation (although not with a switch): http://thread.gmane.org/gmane.comp.boot-loaders.u-boot/102669
It was criticised because saving the parameter images is not the responsibility of bootm - which is true. Although adding a switch would be a cleaner solution than my first implementation - would that be acceptable?
How about implementing it as bootm subcommand?
Another question, how often will one change the boot information for SPL booting? I guess most people will use the normal u-boot to test different configurations and start directly into the OS. When one has some satisfying configuration he will store this for SPL boot into flash regardless whether it is a separate command or a list of commands.
The current solution (having savebp command utilize do_bootm()) couples these two commands in a way which may lead to problems in future (but I don't know what others think about ... Wolfgang?).
I vote for a well documented guide or an example script to gather the boot information and store it for SPL booting (at least for this patch set). I know your bachelor thesis filing date is end of September and I would appreciate to have this functionality in before. The rest of this series seems quite good to me. Except for some minor changes to the arm-bootm prep command ... but read my answer to that patch.
best regards
Andreas Bießmann