Database Develop

[Q] HOWTO for building a custom ROM?

I consider myself a smart guy. I am also pretty tenacious when it comes to problem solving.
While my goal is simple, the work required to achieve it may be daunting.
I am sick and tired of AT&T's bloatware, branding and other nonsense and I want to create a ROM with the following:
* Stock Android 2.2 OS + Samsung drivers
* Minimal set of required applications and services required to run the OS plus the following add-ons:
1) ROM Manager
2) Titanium Backup
3) Root
Ideally I'd get the 2.2 source from Google, build it, add the Samsung drivers, and get the filesystem layout. Then I would be able to add in the apks as necessary.
Perhaps I am biting off more than I can chew at this point, but I don't know. I've searched for articles on building custom ROMs but most are pretty old and outdated.
Am I crazy here? Is this doable?

gdanko said:
I consider myself a smart guy. I am also pretty tenacious when it comes to problem solving.
While my goal is simple, the work required to achieve it may be daunting.
I am sick and tired of AT&T's bloatware, branding and other nonsense and I want to create a ROM with the following:
* Stock Android 2.2 OS + Samsung drivers
* Minimal set of required applications and services required to run the OS plus the following add-ons:
1) ROM Manager
2) Titanium Backup
3) Root
Ideally I'd get the 2.2 source from Google, build it, add the Samsung drivers, and get the filesystem layout. Then I would be able to add in the apks as necessary.
Perhaps I am biting off more than I can chew at this point, but I don't know. I've searched for articles on building custom ROMs but most are pretty old and outdated.
Am I crazy here? Is this doable?
Click to expand...
Click to collapse
You are not crazy, you can do it. And please write an uptodate tutorial about building custom ROMs.

Well I need to find a basic HOWTO for building ROMs that isnt > 1 year old. I am hunting for something reliable.

I found a good HOWTO here but it's for the Desire.
http://forum.xda-developers.com/showthread.php?t=709105
There is reference to "vendor files" which appear to be device-specific files for the build. Do such animals exist for the Captivate?
That aside, I was able to build a generic Froyo and I have the Captivate kernel source downloaded from Samsung. Once I can make this build image Captivate-specific I can tie it all together with the kernel.

The vendor files are available at Samsung, too, but the files appear to be 2.1-specific as referenced in this text file:
1. Get android open source.
: version info - Android eclair 2.1 (android-2.1_r2)
( Download site : http://source.android.com )
2. Overwrite modules that you want to build.
3. Add the following lines at the end of build/target/board/generic/BoardConfig.mk
BOARD_HAVE_BLUETOOTH := true
BT_USE_BTL_IF := true
BT_ALT_STACK := true
BRCM_BTL_INCLUDE_A2DP := true
BRCM_BT_USE_BTL_IF := true
4. make update-api
5. make

gdanko said:
The vendor files are available at Samsung, too, but the files appear to be 2.1-specific as referenced in this text file:
1. Get android open source.
: version info - Android eclair 2.1 (android-2.1_r2)
( Download site : http://source.android.com )
2. Overwrite modules that you want to build.
3. Add the following lines at the end of build/target/board/generic/BoardConfig.mk
BOARD_HAVE_BLUETOOTH := true
BT_USE_BTL_IF := true
BT_ALT_STACK := true
BRCM_BTL_INCLUDE_A2DP := true
BRCM_BT_USE_BTL_IF := true
4. make update-api
5. make
Click to expand...
Click to collapse
that would probably be because Samsung has not officially released 2.2 or the source code for it.

Pirateghost said:
that would probably be because Samsung has not officially released 2.2 or the source code for it.
Click to expand...
Click to collapse
Yep, although looks like you're making way more headway than me in the custom ROM front. I'd love a good generic walk though. There is nothing I've found other than either very technical, or very vague information.

Once I have the process down I will write up a *very* detailed walk through. Just help me get to that point.
I do have a question.
Building the ROM and building the kernel are independent of one another, right? I don't need to replace the kernel as well, do I? Looking at my Nandroid backups I have the following files:
cache.img
data.img
datadata.img
system.img
nandroid.md5
It is my understanding that the kernel is called zImage. So with that said, if I am able to build the aforementioned .img files I should be able to build the ROM, right?
Thoughts?
Furthermore, since Samsung hasn't released the 2.2 sources yet I will just use 2.1 as a test bed. As stated previously, my goal is to create a 100% plain vanilla ROM for the Captivate and I think with all of you to support me I can do that. From there we can add more useful things if need be. But for now, I want to make it stock.

I had jdk 1.6 and had previously just commented out the java version check as outlined in another HOWTO. But I opted to just grab the 1.5 jdk from Sun (Oracle)
make update-api looks good:
[email protected]:~/android_build/mydroid$ make update-api
============================================
PLATFORM_VERSION_CODENAME=REL
PLATFORM_VERSION=2.1-update1
TARGET_PRODUCT=generic
TARGET_BUILD_VARIANT=eng
TARGET_SIMULATOR=
TARGET_BUILD_TYPE=release
TARGET_ARCH=arm
HOST_ARCH=x86
HOST_OS=linux
HOST_BUILD_TYPE=release
BUILD_ID=ECLAIR
============================================
However, I am told newer versions of gcc are problematic with Android so I want to find gcc3.4 which I believe is the best version to build Android. Assuming this builds okay and I think it will, how do I get the apks into the .img files?

After merging Samsung's files into the Eclair source the build is complete!
[email protected]:~/android_build/mydroid/out/target/product/generic$ ls -l *.img
-rw-r--r-- 1 gdanko gdanko 159163 2010-10-13 08:42 ramdisk.img
-rw------- 1 gdanko gdanko 57541440 2010-10-13 08:42 system.img
-rw------- 1 gdanko gdanko 2112 2010-10-13 08:42 userdata.img
Now, my Nandroid backups have the following files:
bash-3.2$ ls -l
-rw-rwxr-x system sdcard_rw 268821696 2010-10-04 15:01 system.img
-rw-rwxr-x system sdcard_rw 162976704 2010-10-04 15:02 data.img
-rw-rwxr-x system sdcard_rw 1203840 2010-10-04 15:02 datadata.img
-rw-rwxr-x system sdcard_rw 16896 2010-10-04 15:02 cache.img
-rw-rwxr-x system sdcard_rw 179 2010-10-04 15:03 nandroid.md5
I am noticing that a stock AT&T Captivate system.img is considerably larger than my system.img. Is this okay?
Also, I am missing data.img, datadata.img, and cache.img in my build. Any idea why?
So... assuming this IS successul, what do I need to do next to use this ROM? I will obviously first test it on the emulator. Furthermore, I want to verify if I can just leave the stock Captivate kernel in place.
Any thoughts?

i would think it would be fine to leave the captivate kernel in place. i am curious to see where this goes.
by the way, check out the dev section. i9000 froyo source went up this morning
http://forum.xda-developers.com/showthread.php?t=807423

I am trying to get the emulator set up to see if my ROM image works. I am confident it will. I wont build a 2.2 ROM until I have a 2.2 kernel. There is no config on the device so I don't know the kernel settings. Too risky for me.

This is pretty interesting. I'll be subscribing to this thread for updates. It would be cool if someone actually made some progress on an AOSP 2.2 rom for us.

I am pretty confident my 2.1 ROM will work. If it does, the 2.2 will be a slam dunk provided Samsung gives us the SGH-I897 files.

gdanko said:
[email protected]:~/android_build/mydroid/out/target/product/generic$ ls -l *.img
-rw-r--r-- 1 gdanko gdanko 159163 2010-10-13 08:42 ramdisk.img
-rw------- 1 gdanko gdanko 57541440 2010-10-13 08:42 system.img
-rw------- 1 gdanko gdanko 2112 2010-10-13 08:42 userdata.img
Now, my Nandroid backups have the following files:
bash-3.2$ ls -l
-rw-rwxr-x system sdcard_rw 268821696 2010-10-04 15:01 system.img
-rw-rwxr-x system sdcard_rw 162976704 2010-10-04 15:02 data.img
-rw-rwxr-x system sdcard_rw 1203840 2010-10-04 15:02 datadata.img
-rw-rwxr-x system sdcard_rw 16896 2010-10-04 15:02 cache.img
-rw-rwxr-x system sdcard_rw 179 2010-10-04 15:03 nandroid.md5
Click to expand...
Click to collapse
Talking to a guy here at work I found out:
* data.img is /data
* datadata.img is /data/data
This is not created when I build my ROM. Is there an option to build data.img? I believe /data is needed in order to boot.

Here are the first images of my ROM running on the emulator. Only real problem is that calendar crashes. I will any ideas why??? :/
Enclosing logcat output. It looks like the calendar is trying to connect to google and cannot since the apis are missing. I don't think it's an issue with the ROM itself.

I bit the bullet and installed the ROM to my device to see what it'd do. I of course created a Nandroid backup first.
The ROM did install successfully via this method:
1) From my Linux box, md5sum system.img > nandroid.md5
2) mkdir /mnt/captivate/clockworkmod/backup/Test
3) cp system.img /mnt/captivate/clockworkmod/backup/Test/
4) cp nandroid.md5 /mnt/captivate/clockworkmod/backup/Test/
5) Run CWM and Backup Current ROM
6) Run CWM and boot into recovery
7) From recovery mode, factory reset the device
8) From recovery mode, restore "Test"
9) From recovery mode, reboot!
The ROM actually booted up with the following issues:
1) No internet access. Most likely because I have not installed the radio's software? How and where do I get it?
2) The screen flickers like crazy. I am not sure why. In the Samsung directory there is an OpenGL archive. I should see if I can use that to fix screen issues.
3) Calendar crashes, this happens in the emulator too. adb logcat shows the calendar is trying to reach Google but the APIs are not installed.
4) The "Product" in Settings > About is "generic". How can I make that "SAMSUNG-SGH-I897"? Just change the environment variable?
I am looking forward to your replies! I think with the help of the community I can rid the Captivate (and other devices) of the tyranny that is crapware!

gdanko did you give up on this or move to something better I would really like to see where this went and what happen to your howto.

Seems like this thread is dead unfort - proabably either - got stuck with RL crap or he bought a different phone

[MOD] Full Ubuntu on the Atrix (now fully automated: 4.1.26/4.1.52)

----- Announcements -----
Closed in favour of this thread.
As noted in the poll, interest is high enough in a union filesystem that it will be the next thing investigated. Unfortunately, anybody who wants to move from any version 1.x or earlier of this script will probably need to re-install everything for version 2.x, as the way the target filesystem is designed is going to change dramatically. Sorry.
There's a typo that jdkramar found, but I expect that most of you won't hit it (unless you've modified your /etc/sudoers), and those that will know enough to fix the script.
----- Your regularly scheduled post below. -----
For those users who have requested a full Linux on their Android device, I now present a relatively easily upgradable Ubuntu on the Motorola Atrix. It's not perfect, but it's surprisingly good.
There are a number of problems we have with the webtop environment that we would like to address in order to have "proper" Ubuntu, including (additional explanation below about each of these points):
The restrictiveness of the environment Motorola's set up (easy to bypass).
A lack of disk space to do anything (only having ~80 MB free really hurts).
An unwillingness to create a third Linux-based environment.
A non-functional apt/aptitude (easy to fix).
Note: This is different than the "webtop over HDMI sans dock" effort. If you're looking for that, please look at this other thread instead. Although unrelated, they shouldn't conflict with each other.
Caveats:
You will be hacking your device. The base script that modifies your device has been reasonably well tested and operates with a decent level of paranoia, so it is highly unlikely that the script will break anything. However, any software you install after you have access to a full Ubuntu presents a very real chance that you will either soft-brick your device or get it into an infinite reboot loop, particularly if you don't know what you're doing. Having a decent knowledge of Unix/Linux is recommended if you wish to proceed. You take full responsibility for what may happen to your device if you execute this script.
You'll need a rooted Atrix in order to do this*, although I doubt anyone's surprised about that. The attached setup script takes care of the steps in post #4, but you should note a few things:
Before you execute the script:
In response to the request that threads indicate whether or not this will work on any Motorola Atrix, it should. If you'd like verification, send me the output of "/usr/bin/dpkg-query -l" on your Atrix's unmodified Ubuntu, and I can double-check. So far, this is verified to work on:
AT&T (me! )
Bell
The script will create a 1 GB filesystem file in /data, so you'll need to have at least that much free space there.
Before running the install script, you'll need to have seven or less apps in the Media area. You can check this by going to Settings → Applications → Manage applications, then checking the Media area tab. The number of apps there will need to be seven or less. If you have more than that, temporarily uninstall apps or move them back to the phone (you can move them back after the script runs and reboots).
While you execute the script:
When the script asks question, it offers reasonably "sane" options by default (although it does try to be safe).
Resetting a filesystem file means that it will use the file that's already there, but set it back to match your original /osh partition. It's generally quite a bit faster than deleting it and recreating it, but deleting it is sometimes the right decision (like if you want to change its size).
The script asks about your MAC (mandatory access control) files because it can't be sure that you haven't altered your original files to your taste. If you have no idea what that sentence just said, pick either the very permissive or somewhat permissive MAC configuration files (the former should cause you fewer headaches).
If you haven't altered your AWN configuration (the tray at the bottom), I suggest you install the modified app launcher configuration (which is the default). If you have altered the configuration, the script won't ask, assuming that you'd like to keep your current one.
Since the setup script downloads Ubuntu packages on the fly (it made more sense than trying to have a giant archive with all of the packages embedded in it), the quality of your connection may result in the script dying partway through. If this happens, you should just be able to restart the script; it'll start again from the beginning, but nothing bad should happen as a result. If enough people report problems with downloading packages, I'll look into a workaround.
After you execute the script:
I've seen a couple of instances where on the first reboot to the alternate /osh partition where MotoBlur thinks that the SIM card has changed. Another reboot fixes this.
For those users who have used a previous version of the script, an upgrade script(s) are included to bring you up to the current level of what's automated.
For those users who have used a previous version of the script and made changes after that, the upgrade script(s) should be able to handle those changes gracefully.
If you want to uninstall:
Using adb with root access:
adb shell
su
cd /system/bin
mv mountosh mountosh.new
mv mountosh.orig mountosh
cd /data
rm ubuntu.disk
cd /home/adas/.gconf/apps/avant-window-manager
rm -r window_navigator
reboot
Once installation is complete, you can start playing with synaptic to install packages. You may need to be careful upgrading any of the -mot/~mot versioned packages, as that can break functionality. I'm still compiling a list of which packages can be upgraded versus which can be left alone (listed below).
Here's a brief runthrough of the type of operations you can do afterwards. Upon rebooting, the webtop screen now looks like this (note the altered set of icons in the tray):
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
Running synaptic brings up a list of available packages:
If we're looking for a decent image viewer, eog should do the trick:
Once we install it, Nautilus (the file manager) now has an interesting option in the menu for pictures, Open with "Image Viewer":
Selecting that brings up what you would expect (moved off to the side so that it doesn't take up the entire desktop):
I haven't yet tested upgrading to Ubuntu 9.10 yet (let alone Ubuntu 10.x), but everything else looks to work fine, with the usual caveats. Further updates to come as they're available!
Changelog:
1.0.6: "By default, Wget will assume a value of 10 seconds." my foot!
1.0.5: More fixes:
Having a space in the directory structure should no longer be disruptive to the script's behaviour.
Questions are now case-insensitive.
More tweaks to the somewhat permissive TOMOYO configuration files.
If the LXTerminal binary has been deleted (as appears to be the case on Bell), it is now re-installed.
The built-in package tester is now more resilient. It supports 1.4.26 and 1.4.52 properly.
The script now asks whether the dock should just be blown away with the replacement, rather than trying to make assumptions.
1.0.4: Quite a few fixes:
Rename upgrade scripts, so that people get less confused (hopefully!).
Tweak the check for whether it's already running from the filesystem file, since the earlier check didn't actually work (doh!).
/osh/data doesn't exist by default, so have the script stop assuming that.
Tweak the pulseaudio re-install so that it's a bit more reliable.
The expected list of packages to manipulate doesn't work for 4.1.26. Set it to the 4.1.26 numbers for now, and re-factor for 4.1.52 with the next revision.
Reroute /bin/ps' stderr to /dev/null so that it doesn't pollute stdout.
The set of unmount instructions at the end need to be split up, since you can rightfully get to the end while skipping some of the mount instructions.
Prior to attempting to alter /system, ensure that it's mounted read/write.
1.0.3: Not everybody runs batch files from the command line, so add a "pause" at the end so that users can see what happened.
1.0.2: Apparently, relying on the score that aptitude returns as the check for whether or not it's okay to auto-fix things is too unreliable. So, instead, opt for the (somewhat riskier, but should be reasonable) check of the number of packages to remove/install/upgrade/downgrade. The check can be made package-specific if need be, but I'd rather have a script that I can re-use later for other upgrades if I need it.
1.0.1: If the package management auto-fix doesn't go through, it's not likely that the script will be able to install gksu or synaptic either, so those steps need to be fixed.
1.0: The "very permissive" MAC option was broken. That's now been fixed, along with completing the automation of the entire process.
0.7.2: Added a check for having a free loop device, and also re-added a "very permissive" MAC option.
0.7.1: Removing /sbin/tomoyo-init appears to cause the X environment not to load at all, so disallow that option for now.
0.7: In addition to making it slightly more user-friendly (by adding questions for when the script isn't sure how to handle a situation), it now handles through to the initial dpkg installation.
0.5.2: Dump rsync's output to /tmp/rsync.out since it takes a really long time, allowing for users to tail the output if they know how. Also, run adb kill-server at the end of the script so that the adb daemon doesn't continue to run (which makes it really annoying to try and delete the directory).
0.5.1: 0.5 had a bug where it tried to check for a return from psneuter, which kills the adb connection (so no return value could be obtained). Instead, use whoami to verify whether or not psneuter succeeded in running.
0.5: The attached script should handle up through to the rsync phase automatically. There's a considerable amount of error checking, so it should be safe to use (I've uploaded a version of the script that should take you as far as the mountosh swapping, which means that you'll now be using a different Ubuntu partition than the default).
* This is a technicality, since the script hacks your device to be able to run commands over ADB as root.
----- Donation notes -----
If you want to donate, rather than to me, why not donate to the Japanese earthquake/tsunami relief effort instead? Here are a couple of (non-attributed) pointers if you don't know where else to look:
International Federation of Red Cross and Red Crescent Societies: no minimum; USD, CHF, EUR; Visa/MC
American Red Cross: $10 minimum (ouch); USD; Visa/MC/AmEx/Discover/Amazon
Canadian Red Cross: no minimum (?); CAD; Visa/MC/AmEx/PayPal

In regards to the above list, a bit of explanation:
The restrictiveness of the environment Motorola's set up (easy to bypass).
As shown in the above steps, renaming and/or removing /sbin/tomoyo-init is all it takes to disable TOMOYO Linux. Leaving it in place isn't necessarily a bad idea, since it means that any of the standard entry points into an Atrix are reasonably locked down. The TOMOYO configuration file that I'll shortly be attaching leaves certain executables completely unlocked (those that need to be able to run anything), but finding out what's hitting up against the limits is a simple matter of setting everything to run_level 2 and checking dmesg.
A lack of disk space to do anything (only having ~80 MB free really hurts).
As I note above, 1 GB isn't much either. On the other hand, what I'm going to look into when I have that mythical thing called free time is to use my internal contacts to get a copy of the kernel source code and then see if I can build myself a FUSE module. At that point, I should be able to pull off a union mount, which should help dramatically.
We haven't figured out how to repartition the Atrix's partition scheme, so we don't have much flexibility on making the existing partition larger. Creating a filesystem file in the Internal Storage would be nice, but a) that partition (p18) isn't available when mountosh runs, and b) it'd make it difficult, if not impossible, to cleanly USB mount the partition. Creating a partition on the SD card would be nice, but a) mmcblk1 isn't available when mountosh runs either, and b) there would be similar constraints if a user ever wanted to pull the SD card.
An unwillingness to create a third Linux-based environment.
I respect what the people who are trying to create a "clean" chrooted environment are trying to do, but it feels to me that there's the whole "throwing the baby out with the bathwater" aspect here, since there really isn't that much more to do beyond what Motorola's provided. Besides of which, some of what Motorola has done with their environment isn't possible to duplicate without taking the files (like the aiw (Android In Window) package). So I would prefer to take the approach of taking the chains off the existing system.
A non-functional apt/aptitude (easy to fix).
Not much to say here, right?
The script builds a larger disk using /data as its home. The primary advantage is that we have access to it at the right point during boot. The primary disadvantage is that we don't have anywhere as much as we'd like to have (since /data is 2 GB total). But, you work with what you've got!

Known package issues:
Be careful upgrading any of the -mot/~mot packages, as that can break functionality. I'm still compiling a list of which packages can be upgraded versus which can be left alone.
Can be upgraded with loss of functionality:
libnautilus-extension1-1:2.26.2-0ubuntu1-mot1
nautilus-1:2.26.2-0ubuntu1-mot1
nautilus-data-1:2.26.2-0ubuntu1-mot1
Upgrading these packages plus at least one additional package I've not yet fully identified breaks viewing mountable storage and the ability to unmount it.
xserver-xorg-core-2:1.6.0-0ubuntu14
Using the stock xserver-xorg-core 2:1.6.0-0ubuntu14 that's already installed without recovering /usr/bin/Xorg appears to lead to a loss of the status bar at the top. This particular issue is now handled by the script.
Cannot be upgraded:
gtk2-engines-1:2.18.1-0ubuntu1~mot1
This breaks aiw (Android In Window) so that there's no frame around the window and it can no longer be manipulated in any way.
xscreensaver-5.10-6-motorola1?
xscreensaver-data-5.10-6-motorola1?
xscreensaver-data-extra-5.10-6-motorola1?
This will likely break displaying aiw (Android In Window) as the unlocking mechanism for the screensaver. Still needs to be tested.

Archived notes:
The below steps are performed by the script in the first post, but in case you really wanted to know what's going on behind the scenes....
----- The setup script takes care of steps starting here. -----
From here on until noted otherwise, all commands are assumed to be run as root (so you either are root, or you're calling every command via sudo).
First, we should make sure that there's enough free space on the device:
/bin/df -h /data
There should be at least 1.0G under the Used column. If not, you won't have enough to create a decent disk. If so, then you can keep going:
/bin/dd if=/dev/zero of=/data/ubuntu.disk bs=1024 count=1048576
/sbin/losetup /dev/block/loop7 /data/ubuntu.disk
/sbin/mkfs -t ext3 -m 1 -b 2048 /dev/block/loop7
mkdir /tmp/osh
/bin/mount -t ext3 /dev/block/loop7 /tmp/osh
At this point, we've created a 1 GB disk file (1,024×1,024=1,048,576), formatted it as ext3, and mounted it in /tmp/osh. The next step is that we need to grab a copy of rsync so that we can perform our copy. I'll assume that rync is in /mnt/sdcard-ext for now:
mkdir /tmp/deb
/usr/bin/dpkg-deb -x /mnt/sdcard-ext/rsync* /tmp/deb
/tmp/deb/usr/bin/rsync -avx /osh/ /tmp/osh/
And now we have a duplicate of our /osh partition, but with more space this time (1 GB instead of 756 MB, which isn't great, but is a hell of a lot better). And, we know how to intercept the point in init.rc where /osh is mounted so that we can redirect it. Put the following into a file named mountosh.new, then copy it to /mnt/sdcard-ext. Here's the file:
Code:
#!/system/bin/sh
# Run mountosh.orig
/system/bin/mountosh.orig "[email protected]"
# Then, mount the filesystem file over the existing /osh
# partition.
/sbin/losetup /dev/block/loop7 /data/ubuntu.disk
/system/bin/mount -t ext3 /dev/block/loop7 /osh
After that:
mv /system/bin/mountosh /system/bin/mountosh.orig
cp /mnt/sdcard-ext/mountosh.new /system/bin/mountosh
chmod 0755 /system/bin/mountosh
chown 0 /system/bin/mountosh
chgrp 2000 /system/bin/mountosh
You can now reboot your device, and you should now boot into the new partition we've just created.
----- The 0.5 version of the setup script performs up through here. -----
Here, an interesting question pops up: do you want mandatory access control (MAC) in place? In my case, I don't have a problem with it, so I can provide updated TOMOYO configuration files that reflect that. If you would prefer to disable it completely, run the following commands:
rm osh/etc/tomoyo/exception_policy.conf
touch osh/etc/tomoyo/exception_policy.conf
rm osh/etc/tomoyo/domain_policy.conf
touch osh/etc/tomoyo/domain_policy.conf
and then reboot your device again. This configures TOMOYO so that it monitors nothing.
Next, we go through and install a series of packages which are either loaded in a broken state (because Motorola force-installed conflicting packages afterward) or packages which are expected to be present. Some of these packages have specific paths listed afterward; if there are, then those files need to be backed up before package reinstallation, then restored afterward. This is important.
coreutils
cpio
dbus
/etc/init.d/dbus
dbus-x11
/etc/X11/Xsession.d/75dbus_dbus-launch
dhcp3-client
findutils
gpgv
pulseaudio
/etc/pulse/daemon.conf
/etc/pulse/default.pa
udev
/etc/init.d/udev
xserver-xorg-core
/usr/bin/Xorg
You'll need to install each package with:
dpkg -i --root=/osh --force-overwrite <package>
At this point, we can now update the list of APT sources so that we can start querying the public Ubuntu depots. Edit your /etc/apt/sources.list to have these entries:
Code:
deb http://ports.ubuntu.com jaunty main universe multiverse restricted
deb http://ports.ubuntu.com jaunty-security main universe multiverse restricted
deb http://ports.ubuntu.com jaunty-updates main universe multiverse restricted
I would also recommend that you add this line to the bottom of your /etc/apt/apt.conf.d/05aptitude file, since the reality of the situation is that we still don't have much space (it'll turn off auto-installing packages that aren't necessary but are recommended):
Code:
Apt::Install-Recommends "false";
At this point, you should be able to run the following with no problems:
apt-get update
----- The 0.7 version of the setup script performs up through here. -----
If this succeeds, we can move on to running aptitude:
aptitude
It will complain that a number of package installations are broken. This is expected, as that's how Motorola built out the distribution. The current script executes the "default" solution, which at the time of writing is four uninstallations, one downgrade, and ten installs. Also make sure that no "unnecessary" packages are uninstalled, since some of them are actually necessary.
We can then install gksu and aptitude so that we have graphical access to the package repositories from aptitude.
----- The 1.0 version of the setup script performs up through here. -----

You my friend are incredibly good. This is insane
Edit: removed huge quote...

Might be a good idea to not quote the entire massive post.
Looking forward to seeing where this goes... How well does it run?

This is great. Can't wait to try it monday. Keep up the good work.
Sent from my MB860 using XDA Premium App

how does this perform vs the included webtop mode?

CC Lemon said:
Looking forward to seeing where this goes... How well does it run?
Click to expand...
Click to collapse
lasersocks said:
how does this perform vs the included webtop mode?
Click to expand...
Click to collapse
It is the included webtop mode - it's just a matter of pulling off some of the restrictions that Motorola put on it. I should probably tweak it just a bit more to where I'm happy with it, and then I'll be able to start making suggestions on what to install. One of the things that people would probably want most is synaptic (graphical package manager), for example, and I should just have a script that installs it for people.

if you get this working, can you make a video please? would be nice to see how it is.

pure genius

can u post a video about your work ? =)

Very good work! You should join the irc sometime.
freenode
#moto-atrix

Now with more scripting!
I've added a version 0.5 of a setup script that automates some of what happens (I've denoted how far in the process it performs right now). It should print out a user-friendly version of what it's doing, in addition to what it's failing on if it fails. Appropriate notes added in the first post as well.

Sogarth said:
Now with more scripting!
I've added a version 0.5 of a setup script that automates some of what happens (I've denoted how far in the process it performs right now). It should print out a user-friendly version of what it's doing, in addition to what it's failing on if it fails. Appropriate notes added in the first post as well.
Click to expand...
Click to collapse
I'll try again. Got up until mounting new partition. Created mountosh.new copied over to phone rebooted. Didn't mount. Didn't have anything in /sbin dir. Like losetup.
Back to .sbf now. Going to try script and give it another go.

I cant wait for my Atrix I'm getting more and more excited every day seeing whats happening here
Thanks a lot for this, especially that I'm not very advanced linux user
Does the usb mice and keyboard work properly? Or other usb-stuff?

dicksteele said:
I'll try again. Got up until mounting new partition. Created mountosh.new copied over to phone rebooted. Didn't mount. Didn't have anything in /sbin dir. Like losetup.
Back to .sbf now. Going to try script and give it another go.
Click to expand...
Click to collapse
Hmm... that's really, really strange.
Edit ubuntu.bad and comment out the reboot line, then. Should just be a matter of adding rem at the beginning of that line.
Also, you shouldn't have to use the sbf. Using the soft brick recovery instructions should be enough, since all you would need to do would be to rename mountosh to mountosh.new, then rename mountosh.orig back to mountosh to get the original state back.

Sogarth said:
Now with more scripting!
I've added a version 0.5 of a setup script that automates some of what happens (I've denoted how far in the process it performs right now). It should print out a user-friendly version of what it's doing, in addition to what it's failing on if it fails. Appropriate notes added in the first post as well.
Click to expand...
Click to collapse
Shouldn't for /f "tokens=*" %%l in ('%~dp0adb.exe shell "chmod 6755 /tmp/psneuter > /dev/null 2>&1 && echo PASS"') do set retval=%%l
be chmod 0755 ? Getting error can't execute psneuter. First I thought it was because I already had one in tmp from AROOT. Trying again now.

Sogarth said:
Hmm... that's really, really strange.
Edit ubuntu.bad and comment out the reboot line, then. Should just be a matter of adding rem at the beginning of that line.
Also, you shouldn't have to use the sbf. Using the soft brick recovery instructions should be enough, since all you would need to do would be to rename mountosh to mountosh.new, then rename mountosh.orig back to mountosh to get the original state back.
Click to expand...
Click to collapse
I was running Gingerblur. Wanted to start fresh. And it wasn't from running batch file. It was before you creating batch. Tried from first post instructions. No biggie. I'm having fun !!!

dicksteele said:
Shouldn't for /f "tokens=*" %%l in ('%~dp0adb.exe shell "chmod 6755 /tmp/psneuter > /dev/null 2>&1 && echo PASS"') do set retval=%%l
be chmod 0755 ? Getting error can't execute psneuter. First I thought it was because I already had one in tmp from AROOT. Trying again now.
Click to expand...
Click to collapse
No, it's actually chmod 6755 since it's setting u+s. What's a bug in there (and I'm about to upload a fixed version) is that /tmp/psneuter actually kills the connection immediately, so it can never return "PASS". I added in a user check afterward instead.
A fixed 0.5.1 uploaded.

[DEV-ONLY][RECOVERY]CWM 6.0.1.x Using Pseudo File System

During my time hacking on android I've discovered some nice easter eggs deep in the android platform. One such easter egg is the mounting of ext4 images directly in the init.rc script. This is a feature I have never seen used by any oems and only by one custom rom [ EDIT: and by letama in his Sony Xperia Boot Manager ]! looking at the git logs this functionality has been present since September 2010 [ commit 49b8124a1759cb8b27e0c21a1a5a54b8a81bdb19 ]. What this effectively gives us is the ability to overlay a pseudo partition layout over the top over the existing layout, thus avoiding any "Danger" of accidental bricking the device by reformatting the SDCard. This is very similar to the way archos mount the stock file system and a variation/extension on the existing methods we use for the SDE Roms.
Although the explanation assumes the use of the SD models it should be fairly straightforward to apply the the HDD models.
THE METHOD:
PART 1 - Prepare a recovery ext4 image file
1. Build CWM6 from the CM10 source.
2. Modify The Recovery's init.rc file to look something similar to this
Code:
on early-init
start ueventd
on init
export PATH /sbin
export ANDROID_ROOT /system
export ANDROID_DATA /data
export EXTERNAL_STORAGE /sdcard
symlink /system/etc /etc
mkdir /boot
mkdir /sdcard
mkdir /system
mkdir /data
mkdir /cache
mount /tmp /tmp tmpfs
mkdir /partitions 0771 system system
mount ext4 /dev/block/mmcblk0p4 /partitions
# Mount /system rw first to give the filesystem a chance to save a checkpoint
mount ext4 [email protected]/partitions/CAC /cache nosuid nodev
mount ext4 [email protected]/partitions/DATA /data nosuid nodev
mount ext4 [email protected]/partitions/SYS /system
mount ext4 [email protected]/partitions/SDCARD /sdcard nosuid nodev
mount ext4 [email protected]/partitions/BOOT /boot
on boot
ifup lo
hostname localhost
domainname localdomain
class_start default
service ueventd /sbin/ueventd
critical
service recovery /sbin/recovery
service adbd /sbin/adbd recovery
disabled
# Always start adbd on userdebug and eng builds
# In recovery, always run adbd as root.
on property:ro.debuggable=1
write /sys/class/android_usb/android0/enable 0
write /sys/class/android_usb/android0/idVendor 18D1
write /sys/class/android_usb/android0/idProduct D001
write /sys/class/android_usb/android0/functions adb
#write /sys/class/android_usb/android0/enable 1
write /sys/class/android_usb/android0/iManufacturer $ro.product.manufacturer
write /sys/class/android_usb/android0/iProduct $ro.product.model
write /sys/class/android_usb/android0/iSerial A101S_REC
#start adbd
setprop service.adb.root 1
# Restart adbd so it can run as root
on property:service.adb.root=1
write /sys/class/android_usb/android0/enable 0
restart adbd
write /sys/class/android_usb/android0/enable 1
3. Modify the etc/recovery.fstab to look like this
Code:
# mount point fstype device
/cache ext4 /dev/block/loop1
/data ext4 /dev/block/loop2
/system ext4 /dev/block/loop3
/sdcard ext4 /dev/block/loop4
4. Creating an empty ext4 image file name REC and mount it on your pc. [ 5MB should do it ]
5. Copy the contents of the built recovery/root directory to the root of your mounted image.
6. chmod init.rc , default.prop and ueventd.rc to 644 ( rw-r-r- )
7. umount the ext4 image and push it to the root of you data partition
That's stage 1 complete. Part 2 Will Follow Shortly.....
Part 2 - Make a dual boot initramfs.cpio.lzo
1. Change the name of the /data directory to /bootdata by modifying the etc/mountpoints file in the initramfs.cpio.lzo. This stops CWM getting confused when trying to un/mount the data partition
Code:
mount_name mount_dev mount_point mount_fs mount_opts volume_name error_code custom_opt
rawfs /dev/mmcblk0p1 /mnt/rawfs rawfs none 150
system /dev/mmcblk0p2 /mnt/system ext4 rw,noatime,noexec system 152
bootdata /dev/mmcblk0p4 /bootdata ext4 rw,noatime,noexec bootdata 154 crypt_compat
storage /dev/mmcblk1p1 /mnt/storage ext4 rw,noatime #storage_name# 155
storage_A80S /bootdata/media /mnt/storage bind bootdata none 155
storage_A101S /bootdata/media /mnt/storage bind bootdata none 155
storage_A101XS /bootdata/media /mnt/storage bind bootdata none 155
storage_LUDO /bootdata/media /mnt/storage bind bootdata none 155
storage_A80H /dev/hdd1 /mnt/storage ext4 rw,noatime #storage_name# 155
storage_A101H /dev/hdd1 /mnt/storage ext4 rw,noatime #storage_name# 155
usbhost_ehci /dev/storage_ehci1 /mnt/usbhost_ehci vfat rw,noatime,utf8,shortname=mixed none 156
usbhost_otg /dev/storage_otg1 /mnt/usbhost_otg vfat rw,noatime,utf8,shortname=mixed none 156
rfsext4 /dev/loop0 /new-root ext4 rw,noatime none 157
rfsext3 /dev/loop0 /new-root ext3 rw,noatime none 157
rootfs /dev/loop0 /new-root squashfs ro,cts_compat none 157
ramdisk /tmp/ramdisk /ramdisk vfat loop,rw,utf8,shortname=mixed #ramdisk_name# 158 ramdisk,ramdisk_size=256
2. Using sirduke989 dmenu initramfs you can modify the init script in the initramfs to mount /bootdata instead of /data and also add /bootdata/REC and /bootdata/BOOT to the list
of known locations , I see this a temporary measure as there are a number of other ways to enable dual ( Triple?!? ) booting
3. Flash the modified initramfs and your choice of kernel using either the recovery menu or kd_flasher, I used the 3.0.21 kernel extracted from the 4.0.24 aos file.
You should now be able to boot into CWM Recovery!
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
Clearly I'm a developer not a photographer!!
Part 3 - Create rest of the "partition" images.
You should have a /partitions directory in you device root, This is what is normally mounted as your /data ( /dev/block/mmcblk0p4) and contains normal android user data e.g installed app settings databases etc. This is where I've created the reset of my Partitions which are just more ext4 images files. I did this using "dd if=/dev/zero ...." and "mke2fs -text4 ...." on the device through adb whilst booted into CWM. This saved time in pushing large empty ext4 files from my pc.
I called my image CAC ( cache ) DATA ( data ) SYS ( system ) SDCARD ( sdcard ) BOOT ( boot ) you can obviously call them what you like and place them anywhere as long as you match up the image names with those in init.rc and make sure the loop numbers are correct in the etc/recovery.fstab everything should be fine.
You can play around with the files sizes, I have an 8gb my current file sizes at the moment are
BOOT = 25MB
CAC = 500MB
DATA = 3GB
SYS = 500MB
SDCARD = 2GB
The sdcard mount point is probably worth pointing at an external sd if you have one available. I have a 32GB Class 10 that I'll probably set up.
After you've setup your psuedo partitions you should then be able to reboot into recovery, if you've done things correctly you mount output should contain the following
Code:
/dev/block/mmcblk0p4 on /partitions type ext4 (rw,relatime,user_xattr,barrier=1,data=ordered)
/dev/block/loop1 on /cache type ext4 (rw,nosuid,nodev,relatime,user_xattr,barrier=1,data=ordered)
/dev/block/loop2 on /data type ext4 (rw,nosuid,nodev,relatime,user_xattr,barrier=1,data=ordered)
/dev/block/loop3 on /system type ext4 (rw,relatime,user_xattr,barrier=1,data=ordered)
/dev/block/loop4 on /sdcard type ext4 (rw,nosuid,nodev,relatime,user_xattr,barrier=1,data=ordered)
/dev/block/loop5 on /boot type ext4 (rw,relatime,user_xattr,barrier=1,data=ordered)
Everything seems to function correctly, I have successful done a backup and restore of my system partition. I have also applied CWM-SuperSU.zip through install zip from sdcard. Mounting and Remounting works although I'm not sure if Mount USB Storage works yet, I didn't on linux and I've not tested on windows and finally wiping and formating was also successful.
Part 4 - Notes on setting up rom images.
Now you may of already realized normal archos images don't come as separate the /boot and /system images so work is require to split them up.
Also if you want to split the /system from the reset of a archos image your boot partition will need to be about 50MB as archos have they /bin /lib /usr directories which contains binary files that use /lib/libuClibc-*.so as it's libc which brings there root filesystem in at around 38MB.
There is a very strong case for ditching these binaries especially when using AOSP/CM based roms. My intial tests show this is possible.
Just like the recovery init.rc Similar changes have to be made to the roms init.rc
Moving Forward:
of course, there's a lot to do but I wanted to at least get this initial information out there for people to consider. I'm currently booting a Linaro 4.1.1 rom using the split partitions. I have also been working on better booting methods which is why I haven't given any details re the initramfs init script but It's fairly straight forward to adjust and adapt. I'll write up more details soon!

More Research!
As I mentioned, I've been further looking into different booting methods and I think I'm approaching what could be a workable solution that will make the Gen9 more like standard android devices
Here's some more of my findings
1. It turns out that we can dump the existing initramfs.cpio.lzo and we can use a standard android ramdisk layout as the android init will load instead of the init script that is currently being used, this also removes the need for switch root and other nonsense that archos have in there. There was one gotcha when had me stumped for about ten minutes, I needed to add "write /sys/class/leds/lcd-backlight/brightness 75" to the init.rc to turn the screen on.
2. It's possible to stop android either using adb shell stop or stopping each service zygote etc, and start CWM while android is booted. It's probably also feasible the manage booting between recovery and android using the persist properties system which should make switching between the 2 fairly easy to control without much tweaking to any binaries. Looking at other devices, namely samsung, they seem to do something similar with recovery being in the same boot.img as the standard files, they simply load a recovery.rc instead of the main init.rc, this might mean that we have to patch CWM to load the correct init.rc I've not looked at the code properly yet but It's not going to be an issue anyway as all the code is fully available, You've gotta love open source.
3. By mounting /dev/block/mmcblk0p1 to /mnt/rawfs we are still able to use abcbox, reboot_into writes to the params file in the partition to control boot switching, so we can maintain booting into sde while leaving the stock android partition in place. I was unable to get any immediate joy from kd_flasher, that maybe because we currently have the ramdisk we want to overwrite mounted as the rootfs. Again I can't imagine it being too difficult to jig this, It can probably be worked out by looking at the current recovery ramdisk scripts should kd_flasher style functionality be required at any point.
4. Most of the binaries that rely on uClibc can be recompiled against bionic without any issue, usb_modeswitch for example. If there are any closed source ones, then the dynamic linker ld-uclibc or whatever is called, ultimately symlinks back to uClibc and we can just grab the one file and place it in the /lib directory. I tested /usr/bin/lsdvd in this way and It seemed to work fine.
I've got all this going on while still leaving a stock android fully intact, which is a great fallback Just in case.... Keeping these modifications at a safe level is one of the primary goals to enable much wider use
I'll put together some examples within the next couple of days to demostrate what I'm talking about here.
I've got a Linaro 4.1.1 ( JRO03R ) which has working powervr drivers with a 3.0.21 kernel, although that's about all that's working on it at the minute.
It's more a proof of concept than anything else, The kernel would need recompiling to add tracefs functionality which is required by jellybean but using the same magic should leave the powervr drivers functioning still, If anyone's interested I can stick that up, I've foolishly deleted ( misplaced/can't remember ) the device files I used to build this.... Too many android source trees and not using git properly leads to school boy errors.
I'm currently working on an omapzoom 4.1.2 tree using the blaze_tablet device as a base, I think this may yield the best results for the archos.
I suppose one other thing to do is the fix up a stock rom to use these methods and give it CWM, that should be pretty simple to do. Although ICS is ooold and I'm really not a fan of some of archos' methods e.g booting 4 different devices off one firmware. Although to their credit they do demostrate just what possible with deviating android from it's normal standard structures.
Hopefully this has whetted your appetites, I'm pretty excited about what's possible here as I feel it brings these archos devices in line with most others.

Me Again!
Just a cheeky little update, I been trying the figure out the best approach to handle switching between android and recovery mode. In effect I kind of wanted to create a Stage 4 bootloader! because you can never have too many bootloaders LOL I certainly wanted to do a "proper" job on it and try to avoid changes to the android platform code.
While to doing research into this I found this patch to the linux kernel which the android team submitted for review, Reading the mailing list thread I don't think it's been accepted yet! It's true what they say about the Kernel Mailing List, You need to bring your A game and be sure of what your doing..
Anyways the patch add a boot-control-block driver to kernel which check for a boot flag, which is exactly what I need to make booting into alternative configuration nice and simple. I suppose it wouldn't be too difficult to chuck in support for the fastboot protocol on one of those configurations. So a CWM Shouldn't be too far off now!!!
As a little treat I've attached a recovery based ram disk if anyone what's to play, just flash it with you favourite kernel on to your sde partition. Then You can boot into recovery and set your self up a pseudo partition image layout through adb. You won't be able to be into android, obviously until you put your old initramfs back.......
This is totally unsupported.
Click to expand...
Click to collapse
I'm just chucking up for those who want to get a feel for to do so. If your uncomfortable playing around in this area then stand well back, It's not prime time yet!!!!
However Feel free to ask questions of a technical bent but If you can't get it to boot then tough luck I'm afraid for now! :laugh:
You shouldn't be able to do any damage with this be I wouldn't go selecting wipe/format etc until you've got some partition images sorted.
I've add abcbox to sbin and symlink reboot_into. It does not seem to fully reboot but It will set the boot flag which you then follow with a call to reboot, That will reboot back into CWM (sde).
Onward
EDIT: Here's the Init.rc and etc/recovery.fstab that It attempts to use.
Code:
on early-init
start ueventd
on init
export PATH /sbin
export ANDROID_ROOT /system
export ANDROID_DATA /data
export EXTERNAL_STORAGE /sdcard
symlink /system/etc /etc
mkdir /boot
mkdir /sdcard
mkdir /system
mkdir /data
mkdir /cache
mkdir /mnt
mkdir /mnt/rawfs
mount /tmp /tmp tmpfs
mkdir /partitions 0771 system system
mount ext4 /dev/block/mmcblk0p4 /partitions
mount rawfs /dev/block/mmcblk0p1 /mnt/rawfs
# Mount /system rw first to give the filesystem a chance to save a checkpoint
mount ext4 [email protected]/partitions/CAC /cache nosuid nodev
mount ext4 [email protected]/partitions/DATA /data nosuid nodev
mount ext4 [email protected]/partitions/SYS /system
mount ext4 [email protected]/partitions/SDCARD /sdcard nosuid nodev
mount ext4 [email protected]/partitions/BOOT /boot
on boot
ifup lo
hostname localhost
domainname localdomain
class_start default
service ueventd /sbin/ueventd
critical
service recovery /sbin/recovery
service adbd /sbin/adbd recovery
disabled
# Always start adbd on userdebug and eng builds
# In recovery, always run adbd as root.
on property:ro.debuggable=1
write /sys/class/android_usb/android0/enable 0
write /sys/class/android_usb/android0/idVendor 18D1
write /sys/class/android_usb/android0/idProduct D001
write /sys/class/android_usb/android0/functions adb
#write /sys/class/android_usb/android0/enable 1
write /sys/class/android_usb/android0/iManufacturer $ro.product.manufacturer
write /sys/class/android_usb/android0/iProduct $ro.product.model
write /sys/class/android_usb/android0/iSerial A101S_REC
write /sys/class/leds/lcd-backlight/brightness 75
#start adbd
setprop service.adb.root 1
# Restart adbd so it can run as root
on property:service.adb.root=1
write /sys/class/android_usb/android0/enable 0
restart adbd
write /sys/class/android_usb/android0/enable 1
Recovery.fstab
Code:
# mount point fstype device
/cache ext4 /dev/block/loop0
/data ext4 /dev/block/loop1
/system ext4 /dev/block/loop2
/sdcard ext4 /dev/block/loop3

Any words about hdd versions?

DragosP2010 said:
Any words about hdd versions?
Click to expand...
Click to collapse
I all depends how you want the structure it, What it would change is the mount point of the paritions directory. After that. everything is loop mounted and sitting on top of the existing structure.
Code:
mount ext4 /dev/block/mmcblk0p4 /partitions

Great stuff!!!!
Hey trevd,
that's fantastic... i will definitely try this CWM in a few days with my custom kernel and bootloader (mountpoint will need some tweaks as well ).
I'm very busy these days, so i gess i'll leave some longer statements to the recent developments in a few days.
Just in short... it's very pleasant to see all these open developments popping up and i really, really appreciate this kind of hacking!
Keep on your great work... you rock!!
Cheers,
scholbert

Hi Trevd,
Nice job!
I've been using the same kind of trick on my Xperia S boot manager project, recovery on loops and mount [email protected] in inits. You may want to take a look at what I did there (see my sig), it may have some use for your project.
Basically, what I do is storing multiple kernels+cpios on the regular kernel partition. I use one (trimmed down to maximize space) to handle the boot logic and cwm, and I have enough space to handle two "regular" kernels. I handle kernel switch just before they load with a small assembly loader. It works very nicely on Xperia, and it's very nice to be able to dual boot with isolated cwms. I can't remember maximum size on a g9 kernel rawfs file, but I think you could have at least enough space to have two kernels to isolate recovery.

Hey letama,
nice to read you :highfive:
letama said:
I've been using the same kind of trick on my Xperia S boot manager project, recovery on loops and mount [email protected] in inits. You may want to take a look at what I did there (see my sig), it may have some use for your project.
Click to expand...
Click to collapse
Cool stuff again...
letama said:
Basically, what I do is storing multiple kernels+cpios on the regular kernel partition. I use one (trimmed down to maximize space) to handle the boot logic and cwm, and I have enough space to handle two "regular" kernels. I handle kernel switch just before they load with a small assembly loader. It works very nicely on Xperia, and it's very nice to be able to dual boot with isolated cwms. I can't remember maximum size on a g9 kernel rawfs file, but I think you could have at least enough space to have two kernels to isolate recovery.
Click to expand...
Click to collapse
AFAIK, we got around ~30MBytes in the raws partition on our tablets so would be possible to put some more kernel+cpios here easily.
Anyway, made some experiments with my latest u-boot port for the tablet this weekend.
I was able to bring up my A80S completely from MicroSD and boot into CWM by using uImage and uInitramfs (based on trevd's CWM image).
There's also some lowlevel multiboot implemented now by using the volume keys... but i know that i use a very special setup, so this is more a proof of concept and not a suitable environment for the average user
Cheers,
scholbert

Thanks Letama + Scholbert
I'll look at all this stuff this week....As a aside, I've played around with mmcblk0p3 and given myself an mmcblk0p5 / 6 of 4MB each. I found parted to be pretty useful (read:safe) for this..... I'm dubious about playing around with the rawfs too much at this point, mainly because I don't understand it fully, yet!.
Have you guy seen this https://github.com/swetland/omap4boot ( I think it's along the line of what Scholbert been working with/on )
Like I've mentioned to ultimate goal is a solution that is "safe" for the average user and also leaves the rest of the tablet in-tact, it maybe a lofty goal but worth a shot. :good:
Thanks for the input guys!

scholbert said:
AFAIK, we got around ~30MBytes in the raws partition on our tablets so would be possible to put some more kernel+cpios here easily.
Click to expand...
Click to collapse
30 for init kernel ? That's plenty indeed! Cool!
Anyway, made some experiments with my latest u-boot port for the tablet this weekend.
I was able to bring up my A80S completely from MicroSD and boot into CWM by using uImage and uInitramfs (based on trevd's CWM image).
There's also some lowlevel multiboot implemented now by using the volume keys... but i know that i use a very special setup, so this is more a proof of concept and not a suitable environment for the average user
Click to expand...
Click to collapse
Nice job! Too bad that Archos doesn't do this on their board with a small internal switch. Would be cool for people like me with big fingers and poor soldering skills
trevd said:
I'll look at all this stuff this week....As a aside, I've played around with mmcblk0p3 and given myself an mmcblk0p5 / 6 of 4MB each. I found parted to be pretty useful (read:safe) for this..... I'm dubious about playing around with the rawfs too much at this point, mainly because I don't understand it fully, yet!.
Click to expand...
Click to collapse
Hummm... You're going to end with a full re-partitioning scheme with system and data if you continue this way . Just be careful when you recreate partitions to let the empty space at the beginning of the disk untouched, instant brick ahead if you go there...
Just in case, here is what I did with my repartition script (do you have it ?): delete p3, delete p4, re-create p3 (careful with start point, leave the hole! it should start just after p2) as extended partition, big enough to hold the new partitions, recreate p4 (same thing about the hole, it should start after p3) with what remains and then you can create p5,p6,p7 with the size you want inside p3.
Last advice: rawfs, don't touch it .
Anyway, the good thing with what I did on Xperia S is that you don't mess with rawfs and re-partition, it's just like flashing a very big SDE kernel from recovery with unmodfified sde firmware, that's all. If I find some time, I'll take a look to see if we can do the same thing here.

letama said:
Just in case, here is what I did with my repartition script (do you have it ?): delete p3, delete p4, re-create p3 (careful with start point, leave the hole! it should start just after p2) as extended partition, big enough to hold the new partitions, recreate p4 (same thing about the hole, it should start after p3) with what remains and then you can create p5,p6,p7 with the size you want inside p3.
Click to expand...
Click to collapse
Hi
Yes I have read your previous threads on the subject which provided alot of the inspiration for the work currently at hand, It is also why I am being ultra careful around the partitions
I think maybe I'm just being too clever trying too pull everything back a step into the intramfs when we can just do the old switch root method method.... It's a little messy on the inside but it will get the job done!

Well, I don't like much the switch root too, it's not a very "Android way" of doing things and make some apps not very happy with it, but yes, it will get the job done, one root for recovery, one root for firmware. And Archos stock would be difficult without switch root, they did put far too much stuff outside of system.

letama said:
Well, I don't like much the switch root too, it's not a very "Android way" of doing things and make some apps not very happy with it, but yes, it will get the job done, one root for recovery, one root for firmware. And Archos stock would be difficult without switch root, they did put far too much stuff outside of system.
Click to expand...
Click to collapse
I'm very much for the "Android Way" I believe the archos stock roms can be re-jigged as the stuff outside of the system is not required by the system, this is all stuff that is a result of the BuildRoot build system and has a dependency on uClibc.
I'm going to try and get something usable this week, can I store additional files in the rawfs partition without running into trouble?

trevd said:
I'm very much for the "Android Way" I believe the archos stock roms can be re-jigged as the stuff outside of the system is not required by the system, this is all stuff that is a result of the BuildRoot build system and has a dependency on uClibc.
Click to expand...
Click to collapse
Well, it's required. It's used inside Android, it handles audio, wifi, codecs, smb...
I'm going to try and get something usable this week, can I store additional files in the rawfs partition without running into trouble?
Click to expand...
Click to collapse
In rawfs or initramfs ? I wouldn't add any file in rawfs, it would be difficult to do and I don't know how would behave the bootloader if it sees new files there. Initramfs you're free to do whatever you want until you reach maximum size of kernel+initramfs.

trevd said:
Have you guy seen this https://github.com/swetland/omap4boot ( I think it's along the line of what Scholbert been working with/on )
Click to expand...
Click to collapse
Well kind of... while i try to stick with the MicroSD our fellow vincencb follows the omap4boot path.
He already made a port of barebox bootloader to work with this tool and pushed it to the repos.
This way you may put anything you like on the tablet's RAM by using MicroUSB for communication and file transfer.
My way is more to get a full featured u-boot and put it into a state, where it might replace stock loader.
Last step is to put it in internal eMMC... so this is also research and development for now.
trevd said:
Like I've mentioned to ultimate goal is a solution that is "safe" for the average user and also leaves the rest of the tablet in-tact, it maybe a lofty goal but worth a shot. :good:
Click to expand...
Click to collapse
Yepp that sounds like a reasonable approach.
letama said:
30 for init kernel ? That's plenty indeed! Cool!
Click to expand...
Click to collapse
To be even more precisely, there are 32512*1K blocks for the rawfs partition.
On my device there's ~12MB left...
letama said:
Nice job! Too bad that Archos doesn't do this on their board with a small internal switch. Would be cool for people like me with big fingers and poor soldering skills
Click to expand...
Click to collapse
Yeah, a real switch would be nice indeed... there's some unused testpoints giving us additional GPIO
Need to solder though...
trevd said:
I'm very much for the "Android Way" I believe the archos stock roms can be re-jigged as the stuff outside of the system is not required by the system, this is all stuff that is a result of the BuildRoot build system and has a dependency on uClibc.
Click to expand...
Click to collapse
Some stuff outside of system is quite useful and gives us something like a minimal linux ecosystem.
Very useful at console level... some tools seem to be used by the Android system as well.
trevd said:
I'm going to try and get something usable this week, can I store additional files in the rawfs partition without running into trouble?
Click to expand...
Click to collapse
Mmmh, letama maybe right with being very careful with this part of internal storage. If it get's corrupt you'll risk a brick (could be restored though by using external boot mechanism).
Anyway the best would be to mount it RW and use the kernel driver to access it... the unknown part is still the bootcode.
There's some kind of allocation table at the beginning of rawfs partition. It is yet unknown how bootcode behaves with an additional entry
Anyway, this is a real nice project and i would really appreciate to see it pushing forward.
Take your time trevd, and again thanks a lot for contribution!!
Have a nice day,
scholbert

Not to put this down in any way, but wouldn't TWRP be better for the G9? It has a full touch tablet UI, which is better than CWM's
Sent from my Galaxy Nexus using Tapatalk 2

Quinny899 said:
Not to put this down in any way, but wouldn't TWRP be better for the G9? It has a full touch tablet UI, which is better than CWM's
Sent from my Galaxy Nexus using Tapatalk 2
Click to expand...
Click to collapse
Hi Quinny
There's nothing to stop us using whatever recovery we like.... they all work the same way ( I think ) , i.e the code is compiled into a recovery binary. Unfortunately my touch screen stopped working long ago so I wouldn't really benefit

scholbert said:
To be even more precisely, there are 32512*1K blocks for the rawfs partition.
On my device there's ~12MB left...
Click to expand...
Click to collapse
Ah, yes, but there is space reserved for each file there, what I have to figure is how much is reserved for custom file (sde kernel). I don't want to have to shift the files located after custom to make room for sde kernel , it would defeat the "no-fuss/no-risk" of the method.
Anyway the best would be to mount it RW and use the kernel driver to access it... the unknown part is still the bootcode.
There's some kind of allocation table at the beginning of rawfs partition. It is yet unknown how bootcode behaves with an additional entry
Click to expand...
Click to collapse
Definitely. Re-partitioning is much safer if space is needed.

trevd said:
.... they all work the same way ( I think ) , i.e the code is compiled into a recovery binary.
Click to expand...
Click to collapse
This is a good keyword: The recovery binary itself!
Most tools inside the recovery are simply linked to busybox, which itself is a link to recovery executable.
In other words, we have some code responsable for the menu and framebuffer stuff and we have busybox.
The strings command gave me version 1.2.0. Now my question...
How to configure this part of code?
I'd like to enhance the busybox part.
Could you please provide a little to howto for a compiler run?
Will i need all that Android stuff installed...
I you have any clue, please point me in the right direction.
Lazy,
scholbert

scholbert said:
How to configure this part of code?
I'd like to enhance the busybox part.
Could you please provide a little to howto for a compiler run?
Will i need all that Android stuff installed...
Click to expand...
Click to collapse
Yes, you need full android repo. Android Build system is messy and tightly coupled, busybox is compiled with bionic (android libc), recovery is built on top of it with few android libraries links. Isolating all this "mess" would be difficult. Except disk space required, there is no big deal in getting full android repo.
I'd suggest to take a look at this, you should do the "Prepare the Build Environment" section.
I don't know how trevd built his recovery, but what I did is create a gen9 device to get proper configuration for recovery (frame buffer config, ...). You can get mine if you want, it's a little outdated (latest cwm doesn't need a specific gfx anymore, the custom one I used has been moved to upstream for instance), but it should give you a base.
To do that, you have to create an "archos" directory in cm9/device directory, then from inside it do:
Code:
git clone git://gitorious.org/archos-ics/device-g9.git gen9
Then, you need to setup the build env once. From cm9 root, you have to do that:
Code:
. build/envsetup.sh
(it setups android build environment)
then
Code:
lunch
then choose full_gen9-eng.
(it selects device target for build)
You should have something like that:
Code:
============================================
PLATFORM_VERSION_CODENAME=REL
PLATFORM_VERSION=4.0.4
TARGET_PRODUCT=full_gen9
TARGET_BUILD_VARIANT=eng
TARGET_BUILD_TYPE=release
TARGET_BUILD_APPS=
TARGET_ARCH=arm
TARGET_ARCH_VARIANT=armv7-a-neon
HOST_ARCH=x86
HOST_OS=linux
HOST_BUILD_TYPE=release
BUILD_ID=IMM76L
============================================
from this point, you can compile:
make -j4 recoveryimage (change j4 with the number of parallel build you want depending on your cpu).
whole recovery fs should be in out/target/product/gen9/recovery/root, with recovery command in sbin
Last, how do you want to extend it? If you want to add custom commands, take a look at cwm bootable/recovery/extendedcommands.c, it may be easier to add stuff there than in busybox.

[Q] encryption, ext2/4, and "filesystem too large to mount safely" error fix?

[Q] encryption, ext2/4, and "filesystem too large to mount safely" error fix?
Hello again,
With my Droid 3 (thanks to Minimoto) again working great, I've turned my attention back to encrypting some of my data on the external SD card. I had used "LUKS Manager" some time ago so that seemed the logical place to start.
Okay, the short *short* version is: What does the error message (from dmesg output) "EXT4-fs (dm-4): filesystem too large to mount safely on this system" mean and how do I fix it or work around it?
The details:
I have a 32 GB SD card with a normal ~20 GB FAT partition that Android sees and uses just fine. On my laptop I created an ext4 file system on the second partition (the remaining ~12 GB of space). Android does not see or use this, but from a terminal I *can* mount it without problems. I chose to mount /dev/block/mmcblk0p2 (external SD partition 2) at /mnt/sdcard-ext-p2. Just FYI, but I wrote a short script and put it at /etc/init.d/03mount_sdpart2 to mount this second partition at the correct mount point and it works fine. Even after a reboot the script runs without problems and now I always have my new ext4 file system present.
The reason for creating the ext4 file system is because FAT does not support files larger than 4 GB and therefore my encrypted volume cannot be larger than 4 GB *if* I create that volume on a FAT file system. With the new ext4 file system, now using the LUKS Manager app, I created a new volume at the above mount point with a size of approximately 11.25 GB. This worked fine, too. The last step, however, actually mounting this encrypted volume, keeps failing. It was supposed to be mounted at /mnt/sdcard/LUKS. Unfortunately, LUKS Manager did not produce much in the way of information about why it had failed.
A quick note about file sizes: If I use an app like "ES File Explorer" and go to /mnt/sdcard-ext-p2 where the large volume file exists, ES says the volume file has a size of -805306366 bytes, obviously wrong. Fortunately, if I use the terminal and look at the file size with "ls -l" it has the correct size. Furthermore, after I opened the volume below using lm.cryptsetup I can use its "status" command and after converting the values from sectors (each sector is 512 bytes) it also displays the correct size of the encrypted volume.
So, back to the terminal where I decided to do the steps manually and see where it failed. For the most part, I used different device and directory names than what LUKS Manager would have used, mostly to be sure I wouldn't accidentally conflict with another process/app. The 'lm.cryptsetup' binary is provided by LUKS Manager and placed in /system/bin. I have the version of busybox that, I assume, ships with Minimoto which is busybox v1.20.0. This is where the 'losetup' I am using comes from.
Concerning loopback devices: LUKS Manager defaults to using/creating high numbered loopback devices, 300 for the first volume created. On Android the loop devices are created in /dev/block, but some tools don't seem to look there. In particular, lm.cryptsetup complains about not being able to find a free loopback device and 'losetup -f' (which displays the next free/available loopback device) always refers to devices in /dev regardless of where they really are. Making symlinks such as "ln -s /dev/block/loop0 /dev/loop0" for each of the currently present loop devs (0 through 7, and 300) fixes both of these problems. That said, whatever manner LUKS Manager uses to execute the underlying tools works properly with the loop devs being located in /dev/block. This can be verified by using "lm.cryptsetup status <crypt_dev>" and the status output shows that it is correctly attached to, for example, /dev/block/loop300. Making these symlinks is more of a convenience/fix for when one is working directly on the command line.
Where, before, 'losetup -f' indicated that /dev/loop0 was free (even though it did not exist), after making the symlinks the same command indicates that /dev/loop4 is the next free loop. On my phone, at that moment in time, that was the correct answer. loop4 also makes sense because I have four apps that I have "moved to SD" and for each app that you do this one loopback device is used (so loop0 - loop3 were used by apps "moved" from main internal storage). Creating these symlinks also made lm.cryptsetup stop complaining/erroring about not being able to find a free loop dev. Finally, running "losetup" without arguments is supposed to list the used loop devs and what is using them. Before making the symlinks is produced no output, and now, after the symlinks, it displays:
Code:
/dev/loop0: 0 /mnt/secure/asec/blah_appA.asec
/dev/loop1: 0 /mnt/secure/asec/blah_appB.asec
/dev/loop2: 0 /mnt/secure/asec/blah_appC.asec
/dev/loop3: 0 /mnt/secure/asec/blah_appD.asec
/dev/loop5: 0 /ss/safestrap/rom-slot2/cache.img
/dev/loop6: 0 /ss/safestrap/rom-slot2/userdata.img
/dev/loop7: 0 /ss/safestrap/rom-slot2/system.img
As you can see, loop4 is missing/available. Also, rom-slot2 is correct as it is where I opted to install Minimoto, rom-slot1 currently containing CM 10.1.
With the loop device issues taken care of, the steps I performed are as follows:
Code:
$ su
# lm.cryptsetup luksOpen /mnt/sdcard-ext-p2/MyCrypto.vol MyCrypto
<type in passphrase>
# ls /dev/mapper
MyCrypto control
# mkdir /mnt/sdcard/luks-tst
# mount /dev/mapper/MyCrypto /mnt/sdcard/luks-tst
mount: mounting /dev/mapper/MyCrypto of /mnt/sdcard/luks-tst failed: File too large
# dmesg | tail -2
[20665.748504] EXT4-fs (dm-4): filesystem too large to mount safely on this system
[20665.750732] EXT4-fs (dm-4): filesystem too large to mount safely on this system
You can see from the above block my commands and the output, especially errors, that followed. Clearly, the file is "too large"... but how? The whole point of this extra partition and ext4 file system stuff was specifically to get around the FAT 4GB file size limitation. Unfortunately, while these errors tell me that something is too large, what *part* is too large? Is the 11.25 GB volume *container* too large, or the ext2/4 file system that exists inside the volume? And if either is too large, what is the maximum size I can make them? I did try adding "-t ext2" and "-t ext4" to the mount command, but neither one changed the outcome nor did they change the messages that were output.
The 'mount' binary (like most others) is provided by busybox, so it could possibly be part of the problem. However, the last two errors above come from the kernel log (via dmesg) which means that at least part of the issue is the kernel, maybe the file system modules. I also checked the logcat output, just in case, but it did not contain anything related or useful. Minimoto 1.7 is using kernel version 2.6.35.7 and perhaps it has some maximum size issues I am unaware of. With the exception of my Droid 3, I haven't used a 2.6.x Linux kernel in a very long time.
I've searched around here as well as the fairly small LUKS Manager message board and the Net at large, but I haven't been able to find the answers I'm looking for. Any ideas as to what I might have done wrong, or something I haven't done but should? I'm not sure how to proceed. Just to be perfectly safe, I did try rebooting but it made no difference.
--John Gruenenfelder

Re: [Q] encryption, ext2/4, and "filesystem too large to mount safely" error fix?
Could it be that your mount point is within the FAT fs, what about creating a new mount point at say /mnt/LUKS
Sent from my XT860 using xda premium

Re: [Q] encryption, ext2/4, and "filesystem too large to mount safely" error fix?
I tried using /mnt/LUKS (instead of the previous /mnt/sdcard/LUKS) as the mount point, but nothing changed. The "filesystem is too large..." messages still appeared.
I don't know why there are two such messages separated by about 2/1000th of a second in the dmesg output even though I issued just one command. It was the same way in my first post.
If I recall, the original reason for putting the mount point under /mnt/sdcard was so that most apps could see/use the new area without having any extra knowledge.
--Sent from my DROID3 using xda app-developers app

Fire TV 2 - Ubuntu (Headless) Install Guide

This guide is intend to help you with "installing" Ubuntu 14.04 (12.04 also works) on the Amazon Fire TV 2 after @rbox recovery has been setup. Only headless mode is possible, similar to Ubuntu Server, but it still makes a nice little ARMv8 development box. Starting X.org or running systemd based Linux distributions will likely never be possible due to features missing from the Amazon kernel. Creative use of the framebuffer is possible if desired, maybe eventually a terminal emulator could be started. As long as you don't mount and modify mmcblk0pX there should be no possible way to mess up Android or brick the device. It's 100% reversible by just removing the SD card. You accept all responsibility for what you do with this work should something go wrong and the device becomes inoperable. With disclaimers and precursor knowledge out of the way let's get started.
To follow this guide you will need:
A micro SD card (2 GB+ recommended)
A Linux system
To login into Ubuntu you will need either:
A 1.8 V TTY USB serial device connected to the UART
A pair of USB serial devices and a null modem cable
I actually used a pair of Xbee's for testing the ttyUSB0 stuff, so hence a pair of FTDI chips would also work.
Preparing the SD Card
To get started you need to first partition the micro SD card:
Type = MBR
Part 1 = 100 MB, Fat32 (vfat)
Part 2 = Remainder, Ext4
Extract the attached zip file to the root of the first partition (extracted filename must be "ramdisk-recovery.cpio.lzma"). This is an alternative initramfs that simply uses busybox to clean up from the partial Android boot and prepare the filesystem for regular Linux. Extract an Ubuntu core root filesystem archive, ubuntu-core-14.04.4-core-arm64.tar.gz, to the root of the second partition as the root user (to preserve ownership/permissions). Make sure you sync or eject the device when done with this work so the data gets flushed to the SD card.
Now we need to make a few changes to the root filesystem to avoid usability issues and allow logins.
Replace /etc/fstab with the following contents to correct some mount options. This "disables" SELinux which fixes dpkg errors and some other login annoyances.
Code:
/dev/mmcblk1p2 / ext4 defaults,relatime 0 0
selinuxfs /sys/fs/selinux selinuxfs ro,relatime 0 0
Replace /etc/init/console.conf with the following contents to allow logins from the UART. Once the root password has been set (root is disabled by default) you can remove "-a root" if desired.
Code:
# console - getty
#
# This service maintains a getty on console from the point the system is
# started until it is shut down again.
start on stopped rc RUNLEVEL=[2345]
stop on runlevel [!2345]
respawn
exec /sbin/getty -s -a root console
Create /etc/init/ttyUSB0.conf with the following contents to allow logins from an attached USB serial device. This should help people who don't want to take apart their device to solder wires onto the UART test points. SSH would of course be an alternative but it's not installed by default in Ubuntu core and this guide is about the building blocks not providing pre-made images (yet). Since udev doesn't work due to devtmpfs not being enabled in the kernel you will need to attach the USB serial device before booting for this to work. As before you can remove "-a root" later if desired once the root password is set. Also you should change the baud rate if needed.
Code:
start on (tty-device-added ttyUSB0)
stop on (runlevel [!2345] or tty-device-removed ttyUSB0)
respawn
exec /sbin/getty -L -a root 115200 ttyUSB0 vt102
Preparing the Fire TV
Until the search order for the initramfs file is changed by @rbox you will need to rename the initramfs on the system partition so it will continue to search for one on the SD card or USB stick. You need to connect to the device using adb either over USB or the network to execute the following commands.
Code:
adb$ su
adb# mount -o remount,rw /system
adb# mv /system/recovery/ramdisk-recovery.cpio.lzma /system/recovery/ramdisk-recovery.cpio.lzma.bak
adb# mount -o remount,ro /system
Right now this prevents "su" from working, which should be fixed by @rbox in due time. To get "su" working again you should extract the original recovery initramfs file to a USB stick and boot the device with that USB stick inserted instead of the previously created SD card. Then to restore "su" you can repeat the above steps just swapping the order of the files in the "mv" command.
Booting Ubuntu
After connecting your serial device of choice simply insert the SD card and power on the device. It's that easy! With luck you should get a shell prompt that is already logged in as root. It's a good idea to set the root password before going much further. The device isn't too useful without networking, so you can install more packages. To solve that connect an ethernet cable (since it's simpler) and type "dhclient eth0" to get online. At this point you can install openssh-server using apt-get or do anything else you'd normally do on an Ubuntu VM or headless Ubuntu system. I'm interested in hearing what people plan to do with a more-or-less high-end ARM development system.

Tips and Tricks
NOTE: These changes, unless otherwise noted, are performed while logged into the target Ubuntu system.
Setting the Hostname
You can change the hostname using the following command:
Code:
echo sloane > /etc/hostname
You should also create a simple /etc/hosts file that matches the chosen hostname.
Code:
127.0.0.1 localhost
127.0.1.1 sloane
Enable Ethernet at Boot
Create the file /etc/network/interfaces.d/eth0 with the following contents:
Code:
auto eth0
iface eth0 inet dhcp
Allow Users Network Access
Since we are stuck running an Android kernel you need to create the following group and add users who need network access (such as ping) to this special group.
Code:
groupadd -g 3003 aid_inet
usermod -G aid_inet -a root
usermod -G aid_inet -a <username>
Removing Failed Services
There are a few services that fail to start due to hardware limitations. We should just prevent them from starting in the first place. We have no VT support enabled in the kernel (boo) so we can just remove the ttyX login prompt services. Also the console setup doesn't work since our console is a serial device not a virtual terminal or other "graphical" type terminal emulator.
Code:
rm /etc/init/tty?.conf
echo manual > /etc/init/console-font.override
echo manual > /etc/init/console-setup.override
Fix /dev Hotplug
As stated before udev doesn't work due to missing kernel features. The busybox applet mdev is a simple replacement for most users. After installing the "busybox-static" package run the following command:
Code:
ln -s /bin/busybox /sbin/mdev
Now add the following line to /etc/rc.local before "exit 0".
Code:
echo /sbin/mdev > /proc/sys/kernel/hotplug
Pre-installing SSH
See: http://forum.xda-developers.com/showpost.php?p=65595013&postcount=13 (thanks @segfault1978)

Thanks a lot, that was exactly the thing I was searching for. Since before today the Raspi3 came out, this box is the cheapest ARMv8 development machine available. With your instruction I was able to login via SSH and install all required software for my development environment. No GUI needed for that, I'm doing all remotely via SSH. Again, thank you!

segfault1978 said:
Thanks a lot, that was exactly the thing I was searching for. Since before today the Raspi3 came out, this box is the cheapest ARMv8 development machine available. With your instruction I was able to login via SSH and install all required software for my development environment. No GUI needed for that, I'm doing all remotely via SSH. Again, thank you!
Click to expand...
Click to collapse
Awesome to hear that it worked for you. Just curious if you went the USB serial route or soldered to the UART pins.

There is also the Dragonboard 410c which is a quad core A53 but has a bit more than the raspberry pi. The price is higher though but it has been out probably a year or so. Just FYI. The raspberry pi 3 is a good deal.

zeroepoch said:
Awesome to hear that it worked for you. Just curious if you went the USB serial route or soldered to the UART pins.
Click to expand...
Click to collapse
None of these methods (since I was in my weekend and all cables and adapters reside in my office)
I placed all .deb-files for openssh-server including all requiremens onto the microSD card, and placed a call "dpkg -i /*.deb" with logging options in /etc/rc.local. I also configured network by mounting the sd card, editing /etc/network/interfaces, and last changed /etc/shadow to have a valid root account for login. It took my some try-and-error loops, but finally it worked as expected. Call me crazy, but I succeeded without any hardware.
---------- Post added at 09:09 PM ---------- Previous post was at 09:03 PM ----------
zeroepoch said:
There is also the Dragonboard 410c which is a quad core A53 but has a bit more than the raspberry pi. The price is higher though but it has been out probably a year or so. Just FYI. The raspberry pi 3 is a good deal.
Click to expand...
Click to collapse
Thank you for the hint, I'll have a look for the availability of this board in germany.

I'm facing a memory problem, resulting in a reboot of the device when all RAM is being used. My compile session takes more than 1.x GB of RAM for the quite complex compilation of all required packages. I can reproduce the situation where all memory is consumed and the device instantly reboots when hitting "no memory left" situation. Since "swapon" is not supported by the kernel (really?): is there any way to enable swap functionality, i.e. via a kernel module? How to overcome this situation where more memory is needed?

segfault1978 said:
None of these methods (since I was in my weekend and all cables and adapters reside in my office)
I placed all .deb-files for openssh-server including all requiremens onto the microSD card, and placed a call "dpkg -i /*.deb" with logging options in /etc/rc.local. I also configured network by mounting the sd card, editing /etc/network/interfaces, and last changed /etc/shadow to have a valid root account for login. It took my some try-and-error loops, but finally it worked as expected. Call me crazy, but I succeeded without any hardware.
Click to expand...
Click to collapse
That is pretty crazy, but since you knew the changes required it worked Not everyone I expected to have such experience. I figured someone might even try to do a qemu chroot or debbootstrap to preinstall openssh. Multiple ways to solve the same problem I guess.
segfault1978 said:
I'm facing a memory problem, resulting in a reboot of the device when all RAM is being used. My compile session takes more than 1.x GB of RAM for the quite complex compilation of all required packages. I can reproduce the situation where all memory is consumed and the device instantly reboots when hitting "no memory left" situation. Since "swapon" is not supported by the kernel (really?): is there any way to enable swap functionality, i.e. via a kernel module? How to overcome this situation where more memory is needed?
Click to expand...
Click to collapse
Looking at the default kernel config from the source code drop from Amazon I see:
Code:
# CONFIG_SWAP is not set
Swap can not be compiled as a module. Even if you chose to use a USB stick or something as the swap device It wouldn't work. Given that we can't change the kernel we can't try stuff like zram or zswap either. The only other suggestion I might have is if you're using "-j4" or something while compiling just remove that so it does a single threaded compile. I'm sure you already tried that. Beyond that you could look at using the Linaro AArch64 toolchain and cross compile. Since we're running Ubuntu you shouldn't need to worry about static binaries.

zeroepoch said:
Looking at the default kernel config from the source code drop from Amazon I see:
Code:
# CONFIG_SWAP is not set
Swap can not be compiled as a module. Even if you chose to use a USB stick or something as the swap device It wouldn't work. Given that we can't change the kernel we can't try stuff like zram or zswap either. The only other suggestion I might have is if you're using "-j4" or something while compiling just remove that so it does a single threaded compile. I'm sure you already tried that. Beyond that you could look at using the Linaro AArch64 toolchain and cross compile. Since we're running Ubuntu you shouldn't need to worry about static binaries.
Click to expand...
Click to collapse
Unfortunately, I'm not compiling with parallel processes (I'm compilig Icinga2 for arm64), I'm running
Code:
dpkg-buildpackage -us -uc
within the source package. One single cpp call consumes so much memory (which is crazy in my eyes, never seen such a big compiler process until today), so I'll investigate the option of cross compiling and afterwards creating the deb file outside of the machine.
Code:
cd /root/icinga2-2.4.3/obj-aarch64-linux-gnu/lib/base && /usr/bin/aarch64-linux-gnu-g++ -DI2_BASE_BUILD -Doverride="" -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Werror=format-security -D_FORTIFY_SOURCE=2 -g -pthread -std=c++11 -Wno-inconsistent-missing-override -fPIC -I/root/icinga2-2.4.3 -I/root/icinga2-2.4.3/lib -I/root/icinga2-2.4.3/obj-aarch64-linux-gnu -I/root/icinga2-2.4.3/obj-aarch64-linux-gnu/lib -I/root/icinga2-2.4.3/third-party/execvpe -I/root/icinga2-2.4.3/third-party/mmatch -I/root/icinga2-2.4.3/third-party/socketpair -o CMakeFiles/base.dir/base_unity.cpp.o -c /root/icinga2-2.4.3/obj-aarch64-linux-gnu/lib/base/base_unity.cpp
I'm a novice in android devices: What would be required to use a custom kernel? A hacked boot loader, which is not available for the AFTV2?

segfault1978 said:
I'm a novice in android devices: What would be required to use a custom kernel? A hacked boot loader, which is not available for the AFTV2?
Click to expand...
Click to collapse
Yep... we need to be able to use fastboot to boot an unsigned kernel and initramfs (boot.img). I tried at one point to overwrite the boot partition with own image and it failed to boot. Since I had the preloader stuff worked out already I was able to restore the original boot image and get it working again.
On a side note, if you don't mind could you post the list of packages needed to install SSH server from rc.local? Others might find that useful. To get around the unset password issue you could have also saved a public key in /root/.ssh/authorized_keys which would also avoid you needing to change /etc/ssh/sshd_config to allow password logins as root.

segfault1978 said:
within the source package. One single cpp call consumes so much memory (which is crazy in my eyes, never seen such a big compiler process until today), so I'll investigate the option of cross compiling and afterwards creating the deb file outside of the machine.
Code:
cd /root/icinga2-2.4.3/obj-aarch64-linux-gnu/lib/base && /usr/bin/aarch64-linux-gnu-g++ -DI2_BASE_BUILD -Doverride="" -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Werror=format-security -D_FORTIFY_SOURCE=2 -g -pthread -std=c++11 -Wno-inconsistent-missing-override -fPIC -I/root/icinga2-2.4.3 -I/root/icinga2-2.4.3/lib -I/root/icinga2-2.4.3/obj-aarch64-linux-gnu -I/root/icinga2-2.4.3/obj-aarch64-linux-gnu/lib -I/root/icinga2-2.4.3/third-party/execvpe -I/root/icinga2-2.4.3/third-party/mmatch -I/root/icinga2-2.4.3/third-party/socketpair -o CMakeFiles/base.dir/base_unity.cpp.o -c /root/icinga2-2.4.3/obj-aarch64-linux-gnu/lib/base/base_unity.cpp
Click to expand...
Click to collapse
I see you are not using -pipe which is good, but a quick search suggested something that might not be simple since this package has it's own build system but changing from -O2 to -O1 might help.

I noticed that Debian has the package available for the same version and already built for arm64.
https://packages.debian.org/sid/icinga2
Maybe you already tried that. You could try going back to jessie which is probably an older version but I think most jessie packages work with Ubuntu 14.04.

zeroepoch said:
Yep... we need to be able to use fastboot to boot an unsigned kernel and initramfs (boot.img). I tried at one point to overwrite the boot partition with own image and it failed to boot. Since I had the preloader stuff worked out already I was able to restore the original boot image and get it working again.
On a side note, if you don't mind could you post the list of packages needed to install SSH server from rc.local? Others might find that useful. To get around the unset password issue you could have also saved a public key in /root/.ssh/authorized_keys which would also avoid you needing to change /etc/ssh/sshd_config to allow password logins as root.
Click to expand...
Click to collapse
This is the list of packages I manually downloaded for ARM64 (unfortunately I used Debian packages which worked first, but leads to a hell situation afterwards when dealing with other dependencies; be sure to use Ubuntu packages in order to avoid problems afterwards):
Code:
busybox_1.22.0
libedit2_3.1
libgssapi-krb5
libk5crypto3
libkeyutils1
libkrb5
libkrb5support0
libwrap0
openssh-client
openssh-server
openssh-sftp-server
This is the piece of calls in /etc/rc.local, right before the exit:
Code:
dpkg --force-all -i /*.deb > /install.log 2>/install.err
echo $? >> /install.log
echo "installation finished" >> /install.log
It took about 1-2 minutes before SSH started to work automatically, you can mount the SD card afterwards in another system in order to check the written logfiles.

Here are some notes for getting wireless working. In addition to the normal OS steps (installing wpasupplicant or wireless-tools and editing /etc/network/interfaces or using wicd) you will need some firmware files from /system (/dev/mmcblk0p13).
Code:
mount -o ro /dev/mmcblk0p13 /mnt
cp -r /mnt/etc/firmware/ /lib/
cp -r /mnt/etc/Wireless /etc/
umount /mnt

segfault1978 said:
None of these methods (since I was in my weekend and all cables and adapters reside in my office)
I placed all .deb-files for openssh-server including all requiremens onto the microSD card, and placed a call "dpkg -i /*.deb" with logging options in /etc/rc.local. I also configured network by mounting the sd card, editing /etc/network/interfaces, and last changed /etc/shadow to have a valid root account for login. It took my some try-and-error loops, but finally it worked as expected. Call me crazy, but I succeeded without any hardware.
Thank you for the the fire tv guide.
Click to expand...
Click to collapse

So, what's the verdict on this? I want to use my firetv 2 as an emby server. Is it worth it trying to get Ubuntu to run?
Sent from my Mi A1 using Tapatalk

mrchrister said:
So, what's the verdict on this? I want to use my firetv 2 as an emby server. Is it worth it trying to get Ubuntu to run?
Click to expand...
Click to collapse
If it has arm64 packages and headless you can give it a try. If you don't like it you can just revert the ramdisk and go back to Android.

Ok sweet, thanks for the reply
Sent from my Mi A1 using Tapatalk

Just curious if anyone's tried running Plex server on this?
I've been looking for a better solution without shelling out $500+ for a dedicated NAS. AFTV is tiny so I could hardwire it and hide it away.
Thanks

I got the same idea. Right now the firetv is still being used as a media streamer but I'm thinking of doing this soon