[Gumstix-users] [overo] x-loader, bootrom, and various related questions

[Dustin Howett <dustin@ho...>]

I've been working on some modifications to x-load lately, and a few
questions have come to mind.

1. Why is there concern towards saving space in the X-Loader
partition? x-load is 21kb, approximately, and the entire partition is
about 512kb. That 512kb partition is loaded in its entirety (as far as
I know) into memory every boot - so, why not make use of all the space
that's available?
I recently re-added partition recognition support and abolished the
(in my opinion, rather limiting) sector-63 mandate on loading u-boot
form FAT on MMC - it cost me 3 kilobytes that were otherwise going to
waste. Why wasn't this included to begin with?
I'd also tried to replace FAT support with ext2 support in x-load, for
loading u-boot. This worked only once, strangely enough, and that was
in the middle of a string of failures. My bootloader-fu isn't so
great, so I'm willing to take the blame there :P

1a. I am reasonably certain that x-load is based on u-boot - possibly
a rather old u-boot, yes, but the family history appears to be
evident. Is this the case?
1b. Is Steve Sakoman's git the best source for u-boot and x-load for
this device? 'master' branches okay? u-boot 2010.03 failed to boot on
my device, but 2010.06 seems fine. I was stuck on 2009.11 for the
longest time . . .

2. I can only assume the sect. 63 limit is in the bootrom as well, for
loading MLO from a compliant SD card - Is the bootrom source available
I've dumped it from RAM and have been reverse-engineering it for a few
hours, but it would really be much easier if the source was available
:P
I can understand trying to keep code size small for the bootrom -
there seem to be about 32kb available for it - so something like
libdisk (if it was based on x-load, which I'm still not entirely
convinced it's not - or that it is - either way) would be a largely
unnecessary bloat to have the convenience of abnormally-aligned
partitions.

2a. Are there any other formats the RBL will accept for the SD card to
be in? It feels a little limiting to have to put MLO on a FAT
partition 63 sectors in and make it the first file on the disk.
I could always flash my x-load modifications to NAND, but I don't want
to wear it out prematurely. I can't test filesystem-level changes in
x-load-as-MLO if I have to conform to one specific filesystem ;)

Thanks for your time in looking over these questions!
-- Dustin L. Howett

Re: [Gumstix-users] TI Linux PSP on Overo

[Chris Whittenburg <whittenburg@gm...>]

/dev/video0 won't be there on boot, the tvp51xx has to be found first.

In your camera file, the last line defines the init function:
arch_initcall(something).  Does this function get called?  Did you
make changes to the Makefile there to add your board-overo-camera.c?

If so, then the next thing to check is: Did you use menuconfig to turn
on the proper CONFIG_VIDEO_TVP514X options?  You will need to do this
so that tvp514x-int.c gets compiled into your kernel.  Does
tvp514x_init() get called?

-chris

On Thu, Jul 8, 2010 at 4:33 PM, William von Achen <wvonachen@...> wrote:
> OK, I've finally gotten some more time to work on this ...
>
> I created a board-overo-camera.c file based on board-omap3evm-camera.c,
> changing very little other than removing the mux-related stuff and
> altering the definition of the decoder's reset pin so that it maps to
> the GPIO we tied it to. Also changed the i2c slave address of the
> decoder to 0x5d to match our board.
>
> However, I still don't see /dev/video0 populated on boot. I hooked a
> logic analyzer up to the i2c pins and the only traffic I see during boot
> is going to slave address 0x51, which seems to be an EEPROM on the Tobi
> board. After inserting a bunch of printk statements into the functions
> in board-overo-camera.c, I could see that the functions
> tvp5146_set_prv_data and tvp5146_power_set are never called, the latter
> of which looks like it does most of the setup.
>
> Am I correct in assuming that these functions are supposed to be called
> automatically by v4l2? Or do I need to do something in the init function
> to kick off the process? Currently, the init function only calls
> omap3evmdc_mdc_config and i2c_register_board_info, both successfully.
>
> Thanks,
> --
> William von Achen

[Gumstix-users] QT4-embedded problem with QtXMLPatterns

[dcraw101 <dcraw101@ya...>]

I recently have run into the following error on "bitbake qt4-embedded" for
the overo:
"QtXmlPatterns was requested, but it can't be built due to exceptions being
disabled."

I search the commit log for when QtXmlPatterns was added on around June 24.
I was able to get qt4-embedded working by checking out a slightly earlier
commit
on the Gumstix overo branch:
(d95b8c3c2801a42c85e76146e84f690e9670ece7)

Has anybody else had this problem with qt4-embedded?
-- 
View this message in context: http://old.nabble.com/QT4-embedded-problem-with-QtXMLPatterns-tp29113226p29113226.html
Sent from the Gumstix mailing list archive at Nabble.com.