OMAP4

Does anyone know how to obtain the source for the boot files used on the OMAP4 port?

Dir. SDFS::ARMiniX.$.!Boot.Loader Option 02 (Run) 
MLO          WR/     DOS       13:01:08 27-Feb-2012   23 kbytes
U-BOOT/BIN   WR/     DOS       15:06:48 31-Mar-2012  241 kbytes
preEnv/txt   WR/     Text      12:15:52 17-Apr-2012   29  bytes
uenv/txt     WR/     Text      07:25:48 17-Apr-2012  100  bytes

A matching source for the distributed binary is (I thnk) a pre-requisite for ROOL to distribute the whole SD card image for the Pandaboard. I cannot see how to persuade the source repository to give me the requisite files so that they can be put into a ‘Sources’ directory on the RISC OS side. Any ideas please?

Ask Chris Gransden? https://www.riscosopen.org/forum/forums/11/topics/2195

The post is about the binary files. That’s not the problem. I want the matching source.

I’m not sure exactly what you mean, from what I see you can use git to do that?
Getting the sources looks easy, what isn’t so easy is getting a working version for RISC OS (or that was my problem)

Elinux seems to show how to get the sources

If you’ve managed to get preEnv/txt uenv/txt to work with RISC OS I’d appreciate any documentation, as I couldn’t get the damned thing to work (in a hurry) I used boot/scr

I’ve done a tutorial for system disc

But I struggled to get the boot stuff working, I’m hoping to get back to that later – any help would be great!

I get the impression that the binaries in Chris Gransdens post are built by him. In that case, you can ask him for the corresponding source code.

If you want to fetch and/or build the source for MLO and u-boot.bin these instructions work on a recent version of Ubuntu.

Download and install an arm cross compiler.

The one I use is available here.

The following assumes the arm cross compiler is installed into your home folder,

git clone git://git.omapzoom.org/repo/x-loader.git
git checkout omap4_dev
git checkout 81fabcae04d463941457bcf9831e0402dce0d0ef

Apply this patch,

diff --git a/cpu/omap4/rom_public_dispatcher.S b/cpu/omap4/rom_public_dispatcher.S
index 36d9591..ea32802 100644
--- a/cpu/omap4/rom_public_dispatcher.S
+++ b/cpu/omap4/rom_public_dispatcher.S
@@ -29,6 +29,7 @@
  * @arg: Argument passed to ROM api (input in r1 will move to r0)
  */
 .globl omap_smc_rom
+.arch_extension sec
 omap_smc_rom:
 	push {r4-r11,lr}
 	mov	r12,	r0

export PATH=~/arm-2011.03/bin:/bin:/usr/bin
make ARCH=arm distclean
make ARCH=arm CROSS_COMPILE=arm-none-linux-gnueabi- omap4430panda_config
make ARCH=arm CROSS_COMPILE=arm-none-linux-gnueabi- ift

Up until a couple of days ago a precise version had to be checked otherwise RISC OS would hang on boot.

Ummmm…. The last commit for the omapzoom git repo was 2010-08-10 according to web git.
Is web git out of date?
I note this comment on the elinux site:
“As of 12/18/2011, the master branch tagged Linaro-u-boot-2011.12 produces a working MLO/U-Boot pair which will boot both Panda and Panda-ES boards. Previously, the Panda EA1 with rev ES2.0 silicon would not boot.”

Up until a couple of days ago

Tell me more. What I am trying to do is to get the source just so that the ridiculous licensing requirements can be complied with to allow a full SD card image for the Pandaboard to be distributed. It is not necessary for any instructions on how to build the binary as no-one will do that anyway. Probably no-one will ever read the sources either so it doesn’t matter how unusable they are. I will never know whether the source is the correct one anyway as I will have no way of checking so I’d rather get the necessary text file containing the source from someone who knows what they are doing! It can then be put into $.Sources.Boot and forgotten forever.

So no help with this then? Anyway this will become moot if no more Pandaboards are going to be made.

Chris: I emailed you a link to a single file copy of the sources a couple of weeks!

Also please do not continue repeating the myth that no Pandaboards are being made. Anything Farnell stop stocking they list as discontinued. All the main PandaBoard distributors are still taking orders. e.g. Tigal are saying:
“Due to short supply of components, delivery lead times for the Pandaboard are currently longer than usual. Orders are processed and shipped at a first-come-first-serve basis.”

Also please do not continue repeating the myth that no Pandaboards are being made.

Didn’t know it was a myth. Glad to hear that things are OK.

I emailed you a link to a single file copy of the sources a couple of weeks ago

Many thanks – I have found this. Sorry I have been rather busy lately.

My suggestion therefore is that an SD card image for the Pandaboard is hosted by ROOL that contains HardDisc4 and the necessary MLO and u-boot as well as the uboot and MLO sources inside a directory ‘Sources’ (some 71Mbytes – to comply with the licensing stuff). This can then be alongside the rom image download.

Is there anything I can do to help make this happen please?

Bump! I have produced an SD card image (2G image zips down to 96Mbytes – it has to include 70Mbytes of sources files which is why it is so big) for the Panda using the MLO and uboot and their sources plus HardDisc4. ROOL would need to get Chris Grandsen’s permission to host it (as it has his ReadMe files etc. included) and they would need to decide how much extra stuff (if any) can be bundled in the HD4 image. Over to ROOL.

Good news…

Peharps the time to consider that Pi and Panda ports are stable (because they are :) )?

Maybe it’s time for RISC OS 5.22?

There a few things that need fixing first, like in Edit reverting CTRL K or adding standard global clip board support.

I’ve done a tutorial for system disc

The link to http://www.paymentlabs.com/sites/default/files/systemdisc/pandaboot.zip seems to be broken – not found.

Also, if I read it right, the paymentlabs tutorial seems to suggest I’ll need a BOOT/SCR file for the Pandaboard ES whereas the Piccallo System Disc tutorial doesn’t mention this for the Panda – which is right?

I’ll need a BOOT/SCR file for the Pandaboard ES

Well, I’ve certainly got a BOOT/SCR file on my PandaRO.

Thanks Chris. Is the PandaRO set-up any different from a bare Pandaboard? If I need it, where would I get a BOOT/SCR file from? There isn’t one in Chris Gransden’s pandaboot.zip and I’ve not been able to find a specific RISC OS one doing an internet search. BTW I’ve already got MLO, PREENV/TXT, RISCOS, U-BOOT/BIN and UENV/TXT (hopefully my new Pandaboard ES will arrive tomorrow or Friday and I want to make sure I’m all set up for it).

Just came across this in the ROOL Beginners FAQ:

“Now you need to copy the ‘firmware’ files into !Boot.Loader …
… for the Pandaboard and Pandaboard ES they are:
MLO, PREENV/TXT, RISCOS, U-BOOT/BIN, UENV/TXT”

Is it a case of either using BOOT/SCR or PREENV/TXT and UENV/TXT perhaps? (as it will be obvious by now, I really haven’t a clue what all these files do!!)

Is it a case of either using BOOT/SCR or PREENV/TXT and UENV/TXT perhaps?

Looks like it – I just have BOOT/SCR, MLO, and U-BOOT-BIN

Does anyone have a working u-boot+MLO for a rev A6 Pandaboard? Despite being at least a year or two old (Oct 2012 judging by the date it was added to the omappedia wiki), support for this board revision seems to be very patchy – Chris Gransden’s prebuilt copies don’t work, TI’s latest sources at omap-zoom don’t work, the official validation image doesn’t work, the latest Android image doesn’t work, etc. The only versions I’ve found which do work are the ArchLinux builds from here. With those Linux will boot fine, but RISC OS is incredibly unstable. As far as I can tell the problems are threefold:

1. The rev A6 pandaboard uses different memory which requires different timings (+ chip select setups?). This seems to be the key fix from the ArchLinux sources (or here in the mainline u-boot sources).
2. The ArchLinux u-boot/MLO leaves a lot of the subsystems disabled – e.g. the display subsystem. This was causing a crash initialising OMAPVideo, although it’s only about 10% of the time that the OS will get that far. Often it would hang a lot sooner, at various points during the ROM init (although those points are oddly repeatable when resetting the board). I’ve had it crash during the initial ROM relocation (which actually triggers a register dump from u-boot + reboot), I’ve had it hang attempting IIC operations (IIC clocks probably disabled), and I’ve had plenty of odd data/prefetch aborts indicating that something is unhappy (including one instance of the kernel claiming that UtilityModule isn’t 32bit compatible!)
3. Newer U-boot+MLO combinations are what are known as SPL versions – rather than having two separate source trees the one tree is used to build both u-boot + x-loader (x-loader was originally just a stripped down u-boot). But the key thing here is that an SPL x-loader expects u-boot to be supplied as a .img file (as produced by mkimage), not a raw .bin. Running mkimage on a .bin u-boot to produce a .img appears to produce a version which will work with an SPL MLO, but the reverse of stripping the header off of an .img doesn’t seem to produce a u-boot.bin that will work with a non-SPL MLO (presumably due to other changes which have been made, e.g. who turns on which clocks)

After grabbing the x-loader sources that Chris points to and applying the rev A6 memory timings fix, that produces an MLO which works. However it doesn’t work when combined with the ArchLinux u-boot (see above re. mixing SPL & non-SPL), and it doesn’t work with Chris Gransden’s u-boot (hangs just after u-boot prints the I2C init message).

However if I grab the latest TI u-boot and hack out some of the broken stuff (initially I was trying to combine it with the ArchLinux MLO, and it was claiming the user button was pressed and was attempting to fastboot an Android image) then that produces a version of u-boot which will successfully boot RISC OS seemingly 100% of the time. However as it’s still slightly misconfigured I have to enter the fatload etc. commands manually, which is less than ideal. It also reports there being 0KB DRAM, which isn’t reassuring (perhaps a clue as to why Chris’s one hangs, if u-boot is having trouble detecting the RAM)

I guess I should try building an SPL u-boot+MLO from the latest mainline u-boot sources and see what happens there; perhaps it’s just the ArchLinux u-boot which has been streamlined to the point that it breaks RISC OS? And if the mainline version does suffer from the same problems as the ArchLinux version, we should probably start working out exactly what’s changed so that we can fix the OS to initialise things for itself.

I wonder if the latest OMAP3 versions suffer from the same problems.

Lots of "However"s, there.

The revision A6 Pandaboard is the latest (which uses different memory) from (around) October 2012 so I guess we need to support it… The Pandaboard ES revision B3 (again the latest) also uses this memory – does anyone know whether everything works on that as well please? I can’t help – my Pandboard ES is revision B1.

There is a revision B3 patched version here although it doesn’t boot into RISC OS on a revision B1. It hangs at various points.

The Pandaboard ES revision B3 works with RISC OS. Willi has made the necessary changes.
Only the right uboot-files are needed for each board.