Venting Machine

Joggler Wiki has moved...

2010-11-03T11:06:00.005Z

jogglerwiki.info was setup by Daniel Case earlier this year, and lots of people started to add stuff to it, as it looked like a good alternative to the rather dubiously ran forum with a similar name...
It turns out that this summer the site disappeared, and we realised that Daniel Case hadn't paid the hosting -- so we quickly contacted the hosting company directly and shelled out some money to get it back online, quick.. Daniel Case was nowhere to be found at that point.
Some enterprising users also decide to "dump" the wiki and make a backup of it, using mediawiki's prefered method... That was a wise move.

Some time later, Daniel Case tried to sell the wiki in various places, including the ubuntu-uk mailing list then appeared to change his mind somehow... The interesting thing is that he "worked out" he could continue hosting it... after WE paid for it, no wonder.

Recently, the site looked even stranger than usual (it had been overran by spammer due to inept admin), and a bit more digging revealed that the site had been sold by Daniel Case to someone in russia, for $150, including username, emails and password hashes...

Now, thats a friendly thing to do !

Luckily, the users who had a backup installed a new mediawiki, and another user bought the other domain names jogglerwiki.com and related, and the wiki has now been reinstated fully at:
jogglerwiki.com

As for Daniel Case, he appeared to be trying to get involved with ubuntu's community as a "trusted" member.. Well lets hope he doesn't offer to host anything for them!

New: libejson, a json++ parser on steroids

2010-09-21T23:44:00.005+01:00

Here is a piece of code I had in my drawers for a couple of years in various shape or form, I dusted it off, added comments here and there, slapped the GPL on it and released it to the unsuspecting public. This version's parser uses ragel. If you don't know about ragel, you should.

This is a parser for an extension of JSON. It has a few remaining issues, like adding security bound checking, humane error reporting and so on, but it's fully functional. I defined that format originally to replace something I had been using for many, many years : Apple "plist" XML subset. Great, but verbose, verbose and fatly verbose.

libejson is a library designed to parse an JSON text (http://json.org/), as well as EJSON, an "extended" version of JSON that I adapted to be more friendly for settings files, storage, and other uses like human interface description files.
libejson uses a simple "driver" (collection of callbacks) system, it parses the input and repeatedly calls your callbacks; your code is responsible for storing, allocating and generally "do stuff" with the parsed data. Think "expat" library for parsing XML.

"Extended" JSON implements these extensions:

EJSON makes quoting node identifiers optional, if they don't contain spaces.
Normal JSON forces you to use syntax like '{ "hello" : "world" }' in EJSON, you can use '{ hello : "world" }'. This makes it easier to read/edit files by hand.

EJSON nodes can have flags/attributes
You can specify a list of words as a node "attribute". This allows flags to be passed to the host program for this particular node. It's especially important for files that describe human interface, or to allow easier debugging in some cases.
Syntax Example: { node_label (visible, translated) : "Bonjour" }

EJSON explicitly allows trailing comma in lists
JSON does not allow trailing comma in lists, which makes code that generate JSON more complicated for no good reason, and also makes "patching" JSON files more difficult. EJSON allows trailing commas in objects and arrays.
Syntax Example: [0,1,2,]

EJSON allows binary data values

You can have arbitrary binary data in nodes, encoded in base64. This is often very useful to store small pieces of binary, small images, or other things embedded in a bigger file. The base64 values are delimited by % (percent) characters.
Syntax Example: { node : % WW91IGNhbiBlaXRoZXIgdXBsb2FkIGEgZmlsZ== % }

EJSON allows hexadecimal integer constants

hex constants are essential for files edited by hand.
Syntax Example: { node : 0xdeadbeef }

Example:

{
 device : [
 {
  name :  "Keyboard",
  label (visible) : "105 keys keyboard",
  flags : 0xcafef00d,
  icon : %
   RGVjb2RlIGJhc2U2NCBzdHJpbmdzIChiYXNlNjQgc3RyaW5nIGxvb2tzIGxpa2UgWVRNME5ab21J
   ekkyT1RzbUl6TTBOVHVlWVE9PSkNCkRlY29kZSBhIGJhc2U2NCBlbmNvZGVkIGZpbGUgKGZvciBl
   eGFtcGxlIElDTyBmaWxlcyBvciBmaWxlcyBmcm9tIE1JTUUgbWVzc2FnZSkNCkNvbnZlcnQgc291
   cmNlIHRleHQgZGF0YSBmcm9tIHNldmVyYWwgY29kZSBwYQ== %
 },
 ]
}

mini6410 announced by FriendlyARM

2010-07-02T12:58:00.005+01:00

Something that has been a rumor for some time, now it's real : a mini6410 with a beefy ARM11, at a pretty amazing price point!

The feature page is in Chinese, but you can read the labels on the picture to get an idea of what it packs.

Roughtly:

ARM11 (s3c6410) 533-667 Mhz. That SoC has a LOT of cool stuff on it, including 2D/3D acceleration, floating point etc. This is the same range of Samsung SoC that was in the iPhone up to the 3GS.
128MB of SDRAM with a 32 bits wide bus. Now that is going to be handy.
256MB+ NAND as usual.
USB 2 device, at last!
The board page notes USB 1.1 for host -- thats a bit of a surprise, I hope it's wrong...
Composite video out
And IR receiver on the board, here comes remote control !
Also a new 4.3" LCD with 480x272 -- I have a suspicion in might be the "same" as the Sony PSP, thats an excellent screen.

Other than that it has all the same as the mini2440, like SD card, serials, and exposed headers for a lot of stuff.

The layout is just 1cm bigger than the mini2440, and they made a nice effort of installing the push buttons on the side of the board, so they can be used (sic)...

Despite the small difference in frequency with the s3c2440, it is a different beast of a CPU, it has a large data/code cache and is most probably at the very least twice as fast !

Now the price looks scary. The webpage quotes 850 RBM, which is barely $130 with a 4.3" screen! The last dev boards I've seen with a s3c6410 it was more like $800+ !

Can't wait to get one !

Howto: Mod a joggler with a heatsink

2010-07-01T19:00:00.006+01:00

Jogglers are very nice, but the hardware has one drawback : the case was pretty badly done and the CPU/chipset combo lead to the CPU being throttled down to 800Mhz very easily.

This mod adds a passive heatsink on the chipset, and completely eliminates the throttling.

Another heatsink mounted using the method in the video. This one is MALICO - MBH375002-33P/2.6, 35X35X25MM it's quite a bit bigger that the one in the video, but works even better.

O2 Joggler, get one while you can

2010-05-22T12:46:00.004+01:00

If you're in the UK, you probably heard of the O2 Joggler, aka OpenPeak hardware. It's a rather gorgeous 7" desktop frame, with a glass front and a capacitive touch screen.

Well the fact is, they are on sale at £50, and it's a massive bargain if you do embedded projects.

It packs:

1.3Ghz Intel Atom (Poulsbo) with floating point SSE2,3 etc, 2 threads.
800x480 32bpp screen
512MB of RAM
Wifi & gigabit ethernet
USB Host, audio out
Light sensor, and dim-able backlight.

Also, it behaves like a EFI PC, you can boot it straight out of a USB stick if you like, and there's space inside to squeeze in a memory stick or equivalent. You can boot debian, ubuntu or whatever distro you like...

In other words, it's a steal for that amount of money..

More info on http://jogglerwiki.info/

New NOR flash support in u-boot

2010-04-21T07:41:00.003+01:00

Pushed a new patchset to u-boot git that allow it to boot from NOR flash. I ripped apart the old NOR code that never actually worked, and now use the CFI (Common Flash Interface) to locate and use the flash chip.
There are gotchas tho:

The NOR is only visible and usable when the "boot switch" on the mini2440 is on "NOR" -- when on "NAND" the NOR flash address space is not accessible in hardware (that I have found anyway, could it be mirror mapped somewhere ? Steppingstone is…).
You can save the environment in NOR too (edit the config.h file), but again, if you were to boot from NAND it won't be available.
There is a new u-boot command "flinit" that allows the user to re-probe the NOR… So you can boot from NAND, throw the switch, and probe the NOR that had now become available.

Now the good news is that you can wipe out vivi without a single remorse and install a u-boot that you have loaded in RAM by:

MINI2440# protect off all
MINI2440# erase 0 3ffff
MINI2440# cp.b 32000000 0 3ffff

Experiment, but remember I'm not responsible if you brick your board -- if you screw up somehow and you don't have a way to boot from NAND already there, you will have to debrick the board with JTAG...

The 256MB & 1GB Nand boards issues of date

2010-03-20T11:00:00.002Z

The World Famous MiniBringup on the wiki is still working for mostly everyone, but there are a few notes for the newer boards.

The 1GB and 256MB NAND boards takes ages to scrub the nand and create the back block table. Be patient. The 1G is like 10 minutes where you will think the board is hung... Be really patient!
The 256MB board seems to have a problem with flashing u-boot itself; apparently 'vivi' is screwing up the environment for u-boot to do it's job, and the board just hangs on reboot. I don't have a fix for that yet. Strangely, the 1GB one works..

New: Google Group / mailing list

2010-03-20T10:46:00.006Z

I've created a new Google Group mailing list for mini2440-linux discussions; it turns out I just can't follow 5 forums, 10 blogs discussions, IRC and Google Code comments page all at once !

Send your friends there, and do help each others -- I tend to reply only to problems that interest me (ie kernely bootloader-y ones) !

http://groups.google.com/group/mini2440

Subscribe to mini2440

Email:

New kernel 2.6.32.7, OpenEmbedded

2010-02-04T21:20:00.002Z

Due to problems getting 2.6.33-rc working (new touchscreen support is borken), I reactivated the .32 branch, re-merged, re-tested and pushed it into the git tree as the mini2440-stable branch.

I also updated OpenEmbedded/Angstrom stable/2009 and rebased the MINI2440 support on top of that (as well as a dozen or so more recent packages)

All seems well, enjoy the new goodies!

New u-boot, to start with

2010-02-02T22:57:00.003Z

A new u-boot is in the git tree. It contains two main changes:

It will recognize larger NAND sizes, it's been tested on 256MB and 1GB NAND mini's. I still don't understand why anyone would need that much, when there is a SD card slot and a USB port but there we go, FriendlyARM makes them, and I support them !
Second patch was made blind. I /heard/ about the /possibility/ of having 128MB SDRAM on future minis, and I made a patch that probes the memory, and try to detect and adapt to 128MB. Might also crash miserably, we'll know in due course!
Another more minor change are slightly tighter timings on the SDRAM, and enabling the burst mode of the controller. No idea if it makes any speed differences, but it sounds cool.

Otherwise, you might wonder about the kernel(s) status. Well, the 2.6.31 is up to date of course, there is a "stable" 2.6.32.X in git too, but I personally had a few problems with it which is why I never announced it here.

The 2.6.33 is well under way, and looks better and more stable that the .32 ever was, so as long as it goes on well we'll have a new stable branch up and running soon!

Open-sourced my AVR simulator

2009-11-24T22:53:00.004Z

Last year or so I wrote (most) of an AVR simulator. It can run quite a lot of code for a few AVR devices (mega48/88/168, mega644 etc), with timers, eeprom, interupts and so on. I now dusted it up, slapped the GPL on it and opensourced it!

I use avg-gcc own headers to help define emulated devices, so it's reasonably easy to add new ones. It still needs a lot of work, but it's quite a bit more lightweight and manageable than the existing, old ones.

http://gitorious.org/simavr

New kernel v2.6.32-rc8, with T35 and new 5.6" TFT

2009-11-24T22:42:00.004Z

Added support for these two screens. the T35 is the new QVGA that is now shipped as the default screen with most mini2440, and the 5.6" is a 640x480 - a bit rare - but is quite a nice format (apart from the lack of bevel).

Remember to switch the screen power jumper to 3.3V on the mini to use the 5.6" screen...

Also remember that to use the timings for the new 3.5" screen, the u-boot kernel boot parameter needs to be changed from the default mini2440=0tb to mini2440=3tb.

Thanks to esky-sh for sending me the 5.6" screen setup !

New kernel 2.6.31.6, v2.6.32-rc8, OpenEmbedded

2009-11-22T22:31:00.002Z

I've had a little time to update the kernels, and make the first official 2.6.32 one too. It all seems to be working.

I updated the kernel config to get ubifs modules built, so hopefully we can get that working on the bigger NAND instead of jffs2.

OpenEmbedded was also updated to latest stable/2009, with new kernel recipes.

New kernel 2.6.31.1, v2.6.30.8, OpenEmbedded, u-boot

2009-09-26T11:34:00.003+01:00

Lots of changes in the trees. I wanted to make sure we were working reasonably well on the mini2440-128 before pushing anything.

Anyway, here are the goodies -- all in the relevant git repositories:

New u-boot; dynamic probe for nand 64 & 128. No other major changes. USB "device" is still borken, I'm not sure why!
New kernel 2.6.30.8, remember that a problem was found related to low memory conditions with that kernel -- I haven't had time to backport the fix, I probably won't -- so you are well encouraged to change to ...
New kernel 2.6.31.1 -- this is the new mini2440-stable branch. It was merged some time ago, but again kept under wraps to make sure it would work with a mini 128MB.
New Angstrom/OpenEmbedded. Merged the stable branch, and updated the recipes to pick up the new sus-mentioned goodies.

New mini2440 hardware, with 128MB Nand

2009-09-26T10:59:00.003+01:00

A new version of the mini2440 has surfaced, with a 128M nand chip instead of the 64MB one. It appears the 64MB one is "old technology" using 512 bytes blocks and 16K erase blocks. This was the only nand format supported by the s3c2410, cheap, and thats probably why it ended up on the mini.

The new 128M nand uses 2K blocks and 128K erase blocks, this creates problems:

u-boot couldn't start from NAND. Very early in u-boot it has "tight" compact code that loads itself into memory before starting "for real". I had to patch that to autodetect the NAND chipset and adapts the reading mechanism accordingly.
yaffs is borken. This u-boot has yaffs 1 in, and that was never indended for NAND bigger than that. yaffs2 exists, but it isn't in the kernel and has a bit of a unclear history, tools and such. So no more yaffs.
jffs2 images done for the mini-64 will NOT work on the mini-128, and vice-versa. Unfortunately there is no way around it appart from generating a different image binary.
There is no way to flash 128MB of flash in u-boot, since one need to load the image in memory first. So AFAIK there is no way to flash the "root" partition of 122MB easily. The option that will work tho is to split that "rootfs" partition into two. More discussion needed on this one...

On the good side of things, a new u-boot is now available in the git tree that allows booting, regardless of the NAND chip on the board. The kernel works perfectly well after that, no differences.

New kernel 2.6.31-rc7, v2.6.30.5 and OpenEmbedded

2009-08-22T12:14:00.003+01:00

Finaly a stable version! It turns out .31 is a lot more heavy handed at purging init data at boot time, and it triggers a problem that was potentialy there all along. So now that's fixed.

Therefore I pushed:

v2.6.31-rc7 into the mini2440-dev branch
v2.6.30.5 into the mini2440-stable branch
A new OpenEmbedded, merged with stable/2009

Kernel 2.6.31 borkenness

2009-08-20T21:21:00.002+01:00

I merged the codebase back into linus's tree on 2.6.31-rc6, and the resulting kernel crashes in some new and creative way in GPIO access. Unfortunately, that bit of code is maintained by our friend Ben Dooks who's on holidays... Hmm..

So, stay clear from -rc6, and wait for the green light!

The good news is that I also merged the "stable" 2.6.30.5 kernel and that one works as expected ;-)

New mini2440 kernel 2.6.31-rc4 & 2.6.30.3

2009-07-30T09:02:00.003+01:00

Updated both the dev and stable branches on git. I haven't updated OE/angstrom as it seems borken upstream (happends often).. We'll try again next week!

Not one, not two, but THREE new kernels!

2009-07-09T08:15:00.002+01:00

After a fructuous "merge evenning" here are 3 new kernels for the mini2440!

2.6.19.6, on git head mini2440-stable-v2.6.29
2.6.30.1, on git head mini2440-stable-v2.6.30 and mini2440-stable.

This one is the new official "stable" kernel, however I do very limited testing appart from "it compiles, it boots, lets ship it" -- these kernels only get official linux security/stability fixes.

And last, but not least:

2.6.31-rc2 on git head mini2440-dev.

It merges the new official mini2440 support with touchscreen support (not present in the official linux) and a few other patches we need. This kernel seems quite solid so far..

I've also updated the OpenEnbedded tree to merge the upstream changes, and to bump the existing kernel recipes as well as add a new one for for .31.

mini2440 patch is in official kernel!

2009-06-24T08:52:00.003+01:00

At last, mini2440 support has been merged into mainstream linux for 2.6.31. Of course, not /all/ of it is there, we'll still need a separate tree for touchscreen support and related; but the core machine support is there, and it's already quite a milestone!

Now waiting for linus to tag 2.6.31-rc1 before re-merging it all together into a branch here...

RELAY2440 week!

2009-06-15T10:14:00.005+01:00

At last, I'm getting the PCB from china's pcbcart.com tomorrow, so as a celebration I made a rendering of the board using eagle3d -- very impressive piece of (free) code, eagle3d !

Official mini2440 2.6.30 released!

2009-06-10T09:31:00.004+01:00

Here we are, mainstream v2.6.30 is already out so the mini gets a new kernel too.

I took this opportunity to rename a few branches:

dev-mini2440-stable is now mini2440-stable-v2.6.29
mini2440-dev is now mini2440-stable-v2.6.30 and mini2440-stable

This makes the OpenEmbedded/Angstrom recipes a bit clearer. The 2.6.30 branch will now get the security/stability updates, and the new -dev branch with .31 will appear as soon as Linus tags one.

There is still a chance the mini2440 patch will appear in the "official" mainstream kernel .31, but apparently there is already a huge amount of stuff in the queue for the arm architecture, so it's not guaranteed...

Make sure you git pull the OE tree too, to get the new kernel recipes..

New Kernel 2.6.30-rc8, qemu and OE/Angstrom

2009-06-03T11:36:00.004+01:00

The mini2440-dev branch was updated to 2.6.30-rc8 - this is just a rebase, no direct functional changes; I'm still waiting for the patches to filter upstream for .31 time...

I also updated qemu & openembedded some time ago, so make sure to pull from these trees too!

New Kernel(s) day again - 2.6.30-rc6 & 2.6.29.4

2009-05-20T12:00:00.002+01:00

Reworked a bit the mini2440-dev branch, and rebased it on the new v2.6.30-rc6 from upstream. No functional change in the mini2440 support itself, just various cleanups and documentation in view of getting the patchset merged into the upstream kernel itself.

Also merged-in 2.6.29.4 into the dev-mini2440-stable branch.

It's all in the git tree!

Kernel 2.6.30-rc5 is the hot cake!

2009-05-13T19:20:00.004+01:00

I've finaly pushed my "clean" development branch to the git repository, containing the same sort of support as the -stable one.

The dev-mini2440-stable branch will now only be maintained to get stability & security updates from upstream, anything 'new' will appear only in the mini2440-dev branch.

The new dev branch got yet another new backlight subsystem, and various other cleanups.

Kernel 2.6.29.3 is up, updated OpenEmbedded & qemu

2009-05-11T09:08:00.003+01:00

Added the latest "stable" patch to the kernel git -stable branch. I'm still working on the dev branch to make it upstream-compatible..

So I also updated OpenEmbedded, to bump the kernel version. I also removed DirectFB 1.3. Somehow their maintainer removed the 1.3 archive -that was working- and replaced it with a 1.3.1 that doesn't even compile. So it's just a good idea to find a way not to need DirectFB. I managed to bring the memory footprint of my own app from 20+MB to 6MB by replacing directfb with just fbdev & tslib...

qemu also got a small bump, to bring it in line with the "trunk" of qemu. No functional changes I noticed, it still works just fine.

mini2440 relay board

2009-05-08T13:00:00.003+01:00

Designed this board to go alongside a mini2440. Takes 3.3 or 5V logic and drives 8 relays. One can also power it from an external source if needed! This is board #2 of a 2 board system I'm making for the mini2440, adding all kind of gizmos to the base board...

OT: Cancelled my pair.com account after 15 years

2009-05-01T22:48:00.002+01:00

It's really bizarre, I used pair.com as my mail relay for so long I can't remember the days I was'nt using them. Remember, the days WAY before the bubble, when your pizza delivery man didn't know internet existed. When javascript wasn't used, flash didn't exist and what was cool was the MIDI plugin in netscape. Well maybe most of you reading this don't remember at all, it's fine :-)

Anyway, the problem of having had a domain for so long is that you get spam. You get millions of spam. In fact, about 3 millions spam a month, by my last count. Yes, 3 real millions, so stop bitching about your mailbox, hon.

pair.com kept using their 1994 qmail install, which is slightly outdated by now. they just can't spool mail at that rate, you need to know about spam at the SMTP transaction level. They don't. They did various wrong move recently, forcing me to find workaround for their stupidity (stupidity at that rate of spam is 100MB in an hour, to download later...).

So now they are gone. Mail bounces of a VM at bytemark.co.uk and that barely raise the load average above zero. Progress!

Kernel 2.6.30-rc2, and OpenEmbedded updated

2009-04-20T17:35:00.004+01:00

The -dev branch of the kernel was updated to 2.9.30-rc2, it's quite a bit better than rc1 and appears to be fully functional. It's all in the git tree !

The -stable branch also had a minor update, I gave up on having variable backlight, it seems that despite my high hope that that PWM pin was hooked up to something, it isn't. So instead I linked the backlight on/off pin to the generic "blank" mode, so the screen more or less powers down entirely when it blanks out due to inactivity.

Also updated OpenEmbedded git, with an update from the stable/2009 branch from upstream, and a string of patches for libraries & tools that I needed or were either borken, or missing.

A CD case & some glue

2009-04-17T11:19:00.006+01:00

How to make a "case" for a development board is always a problem. Most especialy if you still want to poke at the innards of course.

Here, the mini2440 is so small if fits nicely into a hacked CD case. A bit of cutting (I used parts of the CD inside to make the "legs"), a spot of glue and some of God's Own Blue-Tack(tm) and thats all you need for an almost presentation friendly case in 10 minutes!

Well, it's a crystal case, don't mess with it, it WILL shatter if you look at it...

At the back is the old LCD I use for the soekris box, underneath. Indestructible home router, that one. Notice my creativity in making LCD stands hasn't started today, yes, the LCD at the back has been held for 7 years or so by a pair of clips.

Now, look at the lurvely legs:

New -dev branch for 2.6.30 kernel

2009-04-08T11:18:00.002+01:00

I merged our kernel patch and the latest 2.6.30-rc1 , and called that branch dev-mini2440-dev (funnily enough!) into our git-tree.

All seem to be working as usual, but please report anything strange...

New openembedded/angstrom distro

2009-04-05T15:01:00.003+01:00

Finaly updated the OE tree, since they have decided to have a "stable" branch we can rely on. So the OE git tree has been updated rather extensively, and the base distro works fine. I also created a "mini2440-image" task for bitbake, with a few shells tools on top of the basic "console-image".

The distro doesn't have any UI (no Qt, Opie and others) but can be remotely updated via "okpg". Note that this distro is a base to play/test with, it doesn't preclude you from having to "bitbake" your own packages if you want your own software or use the cross-toolchain.

I also uploaded binaries for this image to the Google Code Page. Pick the one you like and experiment. That should get you something to do while the bitbake compiles!

2.6.29 networking bug

2009-03-31T11:04:00.004+01:00

The stock 2.6.29 contains a bug that was killing networking on the mini. Thanks to Daniel who raised the issue and pointed to the fix in linus's tree.

The dev-mini2440-stable branch has been updated with that fix, make sure to "git pull" it..

Official mini2440 2.6.29 released! (etc)

2009-03-24T09:10:00.005Z

Following right up on Linus's release of the official 2.6.29, the git for mini2440 kernel has been updated to match.

I also changed the default configuration to remove the PREMPT support, as I find a rather bad reproducible bug in some case when running out of memory. Make sure to "make mini2440_defconfig".

The u-boot also has been updated, with support for usb storage (ie, sticks) and prelim support for UDC (nand update from a linux host via udc-tools).

qemu also has been updated, so it can load a u-boot directly from "nand" and thus boot straight without any extra u-boot.bin or uImage needed, if your "nand" file is configured properly...

New kernel & qemu

2009-03-12T16:39:00.003Z

Small update to the kernel:

It now try to use u-boot bad block table in NAND, and skip the full NAND scan. You can create a new bad block table in u-boot with "nand createbbt" (note this will wipe the nand, read the readme, and be ready to reinstall a copy of u-boot itself!)
Pumped the i2c bus frequency to 400khz, since the eeprom is happy at that speed.
eeprom now has device access into /sys/..., you can poke each of the 4 256 bytes pages individualy.
Tip of the day for a faster boot: add "quiet" to the kernel bootargs..

And Qemu:

The big change is the full support of NAND in linux, including erase/write, and also jffs2!
As a consequence, you can boot a NAND image straight if it has a jffs2, including the Opie one.

New qemu bits

2009-03-10T00:25:00.002Z

I've added support for the (emulated) 24c08 i2c eeprom in the qemu emulation of the mini2440. It seems to be working great, in u-boot (try 'iprobe') and linux (try i2cprobe etc, from debian i2c-tools package).

I also changed the startup script to start using a "snapshot" disk, where qemu can save the state of the VM (via the control console). I haven't managed to get it to really work, yet, but close !

Nice "getting started"

2009-03-08T22:16:00.003Z

"hari" from the LinuxMCE project has written a nice "howto" on how to get started, build u-boot, a kernel, and flash the lot.

It goes into the details of what to do if you hit bad NAND blocks too..

All of that was done interactively on IRC too !

2009-03-07T15:07:00.002Z

Emdebian "grip" on mini2440

2009-03-06T18:28:00.004Z

I have prepared a howto, and a filesystem for using the latest Emdebian "grip" on the mini2440.

The howto is there, and the resulting mini-debian filesystem is there .

At last a system where nobody has to fight with the source based disasters^H^H^H^H^Htros

Updated qemu

2009-03-05T13:57:00.002Z

Small update to qemu, made the ethernet driver 'saveable' in a snapshot, and other minor changes.

Also made the mini2440 memory amount fixed to 64MB, that removes the need to pass -m to qemu.

Everything is in dev-mini2440-stable in the git tree.

Kernel updated to 2.6.29-rc7

2009-03-05T13:22:00.003Z

I updated the kernel dev-mini2440-stable branch in the git tree to the latest tagged by Linus. I also made a small change to the touchscreen handling of the mini2440. The current mode will work with X11/GPE and other window managers; the previous only seem to work with Opie.

qemu, it does work

2009-03-02T13:27:00.002Z

Here is a screenshot of qemu booting with LCD working. It's not a joke it really works ! And qemu is also faster than the hardware, especialy via NFS..

Linux Support Status

2009-03-02T13:00:00.003Z

So, what works ? Well, I decided to re-write the original software package more or less from scratch since what came with it was dreadful. Outdated 2.6.13 kernel, outdated u-boot, and hot less than /four/ toolchains to build stuff.
All that is changed:

Kernel 2.6.29, direct from linus's git tree.
All works, including audio, touchscreen, LCD of various sizes, SD card etc
u-boot 1.3.2
This was branched of openmoko's tree. somehow the main u-boot tree doesn't have any support for Samsung SoC, it seems to have seriously forked there and the two have evolved considerably in their own directions.
For now, this u-boot works fine, inckuding MMC, network and all. It does not use harware ECC for the NAND because I still keep the old 'vivi' backup bootloader in the NOR flash, and vivi messes up the ECC.
openembedded
Well, that will need a post on it's own. Thats worth a rant !
qemu
I managed to get qemu working at about 90% of emulation, it's now rather functional, where u-boot, kernel start, LCD works, usb works, network works. There are still problems with the touchscreen and the NAND flash, but it's well worth the trouble ! Grrest fun.

MINI2440

2009-03-02T12:49:00.003Z

So, this is the main board, it comes also with a 3.5" or 7" TFT, but this is the guts of it. The board can be bought for really cheap -- like <$150 with a 3.5TFT with touchscreen !

Samsung S3C2440 (405Mhz)
64 MB SDRAM (100Mhz)
2 MB NOR Flash
64 MB NAND Flash
I2C EEprom
Audio
100BaseT Ethernet (DM9000E)
USB host & device
3 UARTs
Buzzer
EBI Connector
Camera Connector
GPIO Connector

Reactivation!

2009-03-02T12:32:00.004Z

Well, that was funny, seems I'm not a dedicated blogger it seems! 4 years after that first post!

Anyway, I'm finaly activating this blog because I'm maintaining a kernel, u-boot and q-emu tree for the mini2440 S3C2440 development board.
And I'm tired of posting to various small, private forums about it so at least here I have control.

The sidebar had RSS for the kernel & qemu tree, but the repositories are also available for u-boot and OpenEmbedded.

Oh well

2005-03-08T22:34:00.002Z

I think I created a blog years ago -probably even on blogger.com- and posted twice maybe. Had to try.

Maybe now is a better time, we'll see...