Mmc Soldered. Driver Almost Working Options

[guylhem]

Hello

I simply copy/paste a module compilation command line. You can try to put the latest mmc.c you can find on the page in the kernel directory where the old mmc.c is. Alternatively, I can send you a binary if you want to test.

PS: IIRC you're into hardware mods too. May I ask you to give me a hard in flash upgrading of the simpad?

Guylhem

[sash]

PS: IIRC you're into hardware mods too. May I ask you to give me a hard in flash upgrading of the simpad?

Guylhem

That's right, I upgraded my T-sinuspad Flash.

In short:
I soldered a second 16mb Flash with soldering paste and an hot-air-pen on the free pads.

But no idea how to upgrade the flash more than 32 mb.
I was not able to find any flash chips i could pay
And I also dont know how often you can desolder (is it correct?) and solder again,
until the pads loose their adhesive!?

[guylhem]

Hello

I found some nor flash chips references that should be compatible, but I couldn't pruchase them.

If you are interested in attempting to sodler them, I will do my best to find them. I don't think soldering/resoldering should be a problem, as long as you don't damage anything

Guylhem

[sash]

At the moment I try to get my 512mb sd card to work.
I think a half gigabyte additional storage should be enough

The problem is that with your image the mmc driver ist working, but I can not load the needed hostap
version for my Sundisk Connect+.

[fredyd]

Can the simpad MMC module be considered stable enough to be included in the mainstream kernel (2.4.27) ?
I'm wondering how solid it is, considering the comments at the top of the code...

[guylhem]

fredyd, it's not stable enough - it needs power management support (if you suspend with a mounted partition, there's corruption...) will do that sometime

sash, did you succeed in having both the hard disk part and wifi part at the same time? I could under wince, I can't on linux. IIRC it's due to not applying some voltage, ie a driver problem.

Guylhem

[sash]

@Guylhem: You were right - it's a driver problem.
I think first it was the actual pcmcia-cs package which does not support binding two
modules/driver to the same slot. And second nobody definitivly knows which initialising
command(s) are used to tell the connect+ to run in wifi+storage mode

Thus I soldered the sc-card to extend the memory.


Can you help me a little bit further to compile the module for an actual kernel?

I tried the following with the kernel from oe-tree:
1. modified the kernel-patch to enable mmc support (CONFIG_MMC=y)
2. replaced the linux/drivers/mmc/mmc_core.c with your mmc driver

If I compile the kernel using bitbake, a mmcdrivers.o and mmc_core.o is generated.
But both of them produce no output when I load them. insmod reports something from tainted kernel module.
IIRC it is only a warning and the module should function anyhow!?
lsmod reports that the module is loaded, but there is no dev/mmc directory.

Any ideas?


btw: The power management support could be released by the apm scripts.
I did the same with wpa_supplicant on my laptop: Unloading for suspend and reloading on wakeup.

[albertr]

Well, I'm not Guylhem, but here's how I've compiled it inside the kernel tree - just put Guylhem's mmc.c file into linux/drivers/block directory and added the following:

"tristate 'MMC SPI driver' CONFIG_BLK_DEV_MMC"
to linux/drivers/block/Config.in

and
"obj-$(CONFIG_BLK_DEV_MMC) += mmc.o"
to linux/drivers/block/Makefile

`make menuconfig` then choose "MMC SPI driver" to compile as a module.
That should do it.
Don't enable MMC support that exists in the latest kernels, it has nothing to do with this driver.

As far as power management goes, I think Guylhem is right. The driver needs hooks to apm to properly re-init the card after loss of power. I have my card powered up all the time, so I don't have this problem, but power drain even in idle mode seems to be significant, so I'm planning to re-wire it to some controlled 3.3V domain. Removing/Reloading the driver on apm events is not an option with storage devices IMHO, since it would invalidate its mount state. I'm not sure what would happen if the card is used as a swap device/file in this case...
-albertr

[sash]

@albertr: It works - thanks

I have made a kernel path for actual opensimpad kernel from openembedded repository.
If guylhem (or somebody else) would load it up to his webspace anybody could
compile the driver and make an ipk easily by replacing the original patch before
compiling the kernel.

I think you're right - it's not a good idea to power down/unmount the card with an
opened file.
Do you know how much current your card needs during standby?

Maybe I also let my card always powered on until the power management in implemented.

[albertr]

I didn't measure it yet - still didn't get around to open my Simpad again. But I left it suspended for a week with a fully charged battery and it drained it completely before I tried to resume it.
-albertr

[guylhem]

sash, I'd be happy to host the patch, but IMHO it should be send to Frederic Devernay who host images on his website.

BTW, what about restarting more officially the opensimpad project?

I'd like to see my kernel in the opensimpad project, until we or someone else finds some time to work on 2.6

Guylhem

[sash]

BTW, what about restarting more officially the opensimpad project?

The opensimpad project was integrated into the openembedded tree.
Or what did you mean with "restarting the opensimpad project"?

So I thing we only have to contact one of the oe guys to integrate the patch to the oe-tree.
Maybe we could help them to maintain the simpad part of the project!

My problem is that I am not close with irc where most of the discussions take place.
But I'd help as good as possible.

[sirtet]

about restarting opensimpad
see the new topic

restart opensimpad

thomas

[fredyd]

Sash sent me the MMC driver, and after a few fixes (kernel compile warnings) I commited it to OpenEmbedded (it's a module after all, so it won't bother people who don't use it).
It's also easier to make images for the t-sinuspad or the 128Mb simpad from OE (see the instructions at the beginning of opensimpad-64+0_2.4.27-vrs1-pxa1-jpm1.bb)

The new images with MMC support will be ready quite soon (tomorrow?).

I wish someone could work on sound support now... Guylhem, won't you need it for your GSM module?

Fred

sash, I'd be happy to host the patch, but IMHO it should be send to Frederic Devernay who host images on his website.

BTW, what about restarting more officially the opensimpad project?

I'd like to see my kernel in the opensimpad project, until we or someone else finds some time to work on 2.6

Guylhem

[fredyd]

From the OE commit log:

opensimpad: MMC support + better support for SinusPad and 128M SIMpad
- There's now a config option to build the kernel for 128Mb SIMpad
(as there was one for the SinusPad)
- The .bb files automatically turn on these options depending on the
total memory available.
- The MMC driver was made by Clement Ballabriga <clement@asso.ups-tlse.fr>
and Guylhem Aznar <simpad@externe.net>.
Adding an MMC card (internal or external) to the SIMpad is described
on these pages:
http://externe.net/zaurus/simpad-bluetooth/
http://www.iral.com/~albertr/linux/simpad/mmc/