NAND Testing - 05-25 Update: New LK, Recovery.img, Kernel Updates through Recovery

[natemcnutty]

Originally Posted by Lmiller1708

Attached is the Log from tonight. (Had to zip it)
Same issues as before... Couple of them should be easy to solve.
But currently the modules are not getting added.
I'm going to strip down the system as much as possible to save ~15 MB.
Just to get it to load the modules and hopefully get a clean boot.

I also attached my INIT from the initrd. I left most of the stuff from the Vogue in there... Just incase...

Update: After some reading HERE, I decided to change my Command Line code to this:

mtdparts=msm_nand:0x10000000@0x0(nand)ro,0x01500000@0x04000000(root),0x07BA0000@0x05500000(system),0x06f60000@0x0CFA0000(userdata)

This gave me a larger system while keeping the data partition the same. This them allowed me to load in the modules.
I also added the "ro" at the end of the nand partition, not too sure what this does but devices with similar mtdparts cmdline have this. Basically everything but the Vogue.

Also I still can't get my System nor my Data to stick after a reboot. I thought it was just my Data but now it seems to be both. Reading back in the irc logs i see wozzer had this same issue. Might be something with the mtdparts cmdline.

Hehe, looks like you had a similar idea as I did. I extended the size of the system partition by 30MB in my tinboot and was finally able to install the modules. It gets further than I did before, but still have some refactoring to do.

Also of interest for the mtdparts commands, I believe the ro option means it is treated as read-only, and if we place a minus before the data partition, it will use the remaining available flash for data. Here is the way I have mine right now:

mtdparts=msm_nand:0x10000000@0x0(nand),0x01500000@0x04000000(root),0x076C0000@0x05540000(system),-0x06f60000@0x0cc00000(userdata)

Originally Posted by [ACL]

nahh wozzer never got this far.. he didnt have the driver remember ..lol. nor was yaffs even enabled on the kernel. You issue i think may be android. Once android boots it goes and tries to remount everything. you need to comment those lines out in the rc file and experiment. Read my post below.

Edit: oops so it only happens after reboot? ok thats something else then. odd.. you may need to check the dmesg to see if there are errors writing to nand. Or go over your script and make sure you arent reinstalling each time.. you could be rewiping nand

also how did you handle the directory structure for the rootfs. Your init looks like it just bypassed the whole thing. did you just merge it with the initrd?

I bypassed rootfs entirely by merging stuff from rootfs into initrd. Unfortunately, I'm still ironing out some kinks that were caused by doing that. This is the way that the Vogue currently does it.

Also, I was looking back through IRC logs, and it looks like Wozzer was having issues with data staying on NAND as well.

IRC logs

[Lmiller1708]

Originally Posted by natemcnutty

Hehe, looks like you had a similar idea as I did. I extended the size of the system partition by 30MB in my tinboot and was finally able to install the modules. It gets further than I did before, but still have some refactoring to do.

Also of interest for the mtdparts commands, I believe the ro option means it is treated as read-only, and if we place a minus before the data partition, it will use the remaining available flash for data. Here is the way I have mine right now:

mtdparts=msm_nand:0x10000000@0x0(nand),0x01500000@0x04000000(root),0x076C0000@0x05540000(system),-0x06f60000@0x0cc00000(userdata)

I bypassed rootfs entirely by merging stuff from rootfs into initrd. Unfortunately, I'm still ironing out some kinks that were caused by doing that. This is the way that the Vogue currently does it.

Also, I was looking back through IRC logs, and it looks like Wozzer was having issues with data staying on NAND as well.

IRC logs

Great. Yeah I did notice that "-" thing... But I never tried it out!
Also it does look like the system offset should be 0x05500000 instead of 0x05540000 but this is just how my math worked out... Recheck it for me.
Are you able to keep your data after a reboot?

@ACL
Yes I was doing the same as Nate. But I don't see any reason why you can't mount the rfs to the root mtd1?

[natemcnutty]

Originally Posted by Lmiller1708

Great. Yeah I did notice that "-" thing... But I never tried it out!
Also it does look like the system offset should be 0x05500000 instead of 0x05540000 but this is just how my math worked out... Recheck it for me.
Are you able to keep your data after a reboot?

@ACL
Yes I was doing the same as Nate. But I don't see any reason why you can't mount the rfs to the root mtd1?

I'll have to mess around with it more tomorrow. Just a heads up, I think you might be missing a few things in your init. Here is what I have right now (though I'm still stuck in the boot process):

http://dl.dropbox.com/u/11790134/init

[Lmiller1708]

Originally Posted by natemcnutty

I'll have to mess around with it more tomorrow. Just a heads up, I think you might be missing a few things in your init. Here is what I have right now (though I'm still stuck in the boot process):

http://dl.dropbox.com/u/11790134/init

Thanks Nate. I will look at it a bit more.

To get it to boot comment out this line:
fail "from sourceforge"
It's in the Fix module section about half way down.
Not even sure we need that anymore.

Also I still haven't been able to keep my system/data after a reset.
But I'm able to go back to the installer and reload everything just fine.
It's almost like everything is getting written to memory... Just like what was happening to wozzer.

I'm starting to think I'm not building correctly... I might try re-downloading the linux-msm-rhod-nand build. Do you guys recommend something else?

[[ACL]]

ok i seem to have recreated the issue. Actually somewhere between my pushes i broke something. Right after the installation happens i notice that the unmount fails. Not sure why but we must be able to unmount and remount /system as ro. I think this is done so a checkpoint is made.

So i'm still playing with it a bit more. may need to revert code or figure out where i went off.

[Lmiller1708]

Originally Posted by [ACL]

ok i seem to have recreated the issue. Actually somewhere between my pushes i broke something. Right after the installation happens i notice that the unmount fails. Not sure why but we must be able to unmount and remount /system as ro. I think this is done so a checkpoint is made.

So i'm still playing with it a bit more. may need to revert code or figure out where i went off.

That's good to hear! At least I'm not doing anything wrong!

P.S. Coding + Flashing + Beer + Superbowl = A Great night!
Go Pack! Coming from a Vikings Fan that's hard to say... But hey I live in Wisconsin... So I guess I should!

[Lmiller1708]

Well ported over more items from the rootfs... It acts better but now I'm stuck on getting Android to initialize.

It seems like it unmounted the system.
Here is what I get:
save exit: isCheckpointed 1

Then it will just sit there... Any Ideas?

[natemcnutty]

Originally Posted by Lmiller1708

Well ported over more items from the rootfs... It acts better but now I'm stuck on getting Android to initialize.

It seems like it unmounted the system.
Here is what I get:
save exit: isCheckpointed 1

Then it will just sit there... Any Ideas?

Have you tried copying the init.android from rootfs over the one in initrd yet? I'm heading in to work in a few minutes, and hopefully I'll have some more time to work on this. Also, thanks for the mention of commenting out the modules part. I did get it to work without having to comment that out previously, so that's why I was continuing to leave it in.

I was really hoping to leave it similar to Vogue so that we could port over any changes, but at this point, I'm feeling like stripping everything out that we don't need for NAND booting and then do our own updates from there (like ACL has been doing with the installer script). I have a version of both on my machine, and I think I'm going to start focusing on the NAND only one instead of trying to keep all the other stuff in it.

[natemcnutty]

Well, it's official, we have full NAND booting FRX04. Unfortunately, there are still a few things that aren't quite working in initrd which cause FC's shortly after boot, and I can't seem to get service. Gonna try to clean this up some more tonight and hopefully have something for everyone to test out tomorrow.

[Lmiller1708]

Originally Posted by natemcnutty

Have you tried copying the init.android from rootfs over the one in initrd yet? I'm heading in to work in a few minutes, and hopefully I'll have some more time to work on this. Also, thanks for the mention of commenting out the modules part. I did get it to work without having to comment that out previously, so that's why I was continuing to leave it in.

I was really hoping to leave it similar to Vogue so that we could port over any changes, but at this point, I'm feeling like stripping everything out that we don't need for NAND booting and then do our own updates from there (like ACL has been doing with the installer script). I have a version of both on my machine, and I think I'm going to start focusing on the NAND only one instead of trying to keep all the other stuff in it.

I do agree we should keep it similar to the Vogue.
I do like there install system however and think we should keep it.

But!!

I was able to get NAND to boot!
Called my voicemail.
Sent and recived a SMS.
LCDDenisty looks like it is set to 210.

Still have some more work to do and clean up the code... BUT!!!



WE HAVE NAND!