[DEV] "Root easily your Gen8 device" (developers only thread!) - Gen8 Android Development

This thread is for discussing features and improvements of this rooting method
For questions and problems read the [HOWTO] thread
Project site online: archos-gen8-sde-rooting

---

wdl1908 said:
chulri what about creating a 1Gb file on the /mnt/storage and formatting that as ext3 copying all the original /data to it and then mounting that with a loop interface on /data.
Click to expand...
Click to collapse
That's what I tried previously (before the /data thing), but I had no luck and it's a big issue because android tries to unmount /mnt/storage when you connect your archos device to the computer but that's not possible because the lock of the mounted rw-file makes umounting of /mnt/storage impossible and I have to mount the rw-file before /mnt/storage gets mounted, that's another issue which must be resolved.

---

chulri said:
That's what I tried previously (before the /data thing), but I had no luck and it's a big issue because android tries to unmount /mnt/storage when you connect your archos device to the computer but that's not possible because the lock of the mounted rw-file makes umounting of /mnt/storage impossible and I have to mount the rw-file before /mnt/storage gets mounted, that's another issue which must be resolved.
Click to expand...
Click to collapse
Yeah I came to the same conclusion as you, that what I suggested would bork the usb mount option to the PC.
Another thing I realized is that the official firmware upgrades could probably update files on the data partition. So moving the whole partition is not an option as that would break the upgrade process.
I've been looking at splitting the storage partition in several parts with parted I found an arm binary at http://plugapps.com/arm/ maybe these can be included in the initramfs.
I've also been analyzing my data partition
Code:
# du -s /data/*
112003 app
70503 dalvik-cache
40084 data
4622 test
The test directory is the place where the google market is installed via arctools or gappsinstaller.
So if it's possible to split the storage partition in several part we could move these dirs to it's own partition. This would not be optimal a good solution would be to move the complete data partition over but this needs a bit of thinking how to handle upgrades.

---

wdl1908 said:
So if it's possible to split the storage partition in several part we could move these dirs to it's own partition. This would not be optimal a good solution would be to move the complete data partition over but this needs a bit of thinking how to handle upgrades.
Click to expand...
Click to collapse
We could shrink the internal storage and append partition(s) after it. I'll give it a try, as long as my usb port is broken I have more time to focus on this here

---

chulri said:
We could shrink the internal storage and append partition(s) after it. I'll give it a try, as long as my usb port is broken I have more time to focus on this here
Click to expand...
Click to collapse
I've been trying to cross compile e2fsprogs and parted but I can't seem to get it.
Code:
e2fsprogs-1.41.14$ cross ./configure --host=arm-linux-uclibcgnueabi --build=i686-linux
Completes without errors but the make does not complete.
Code:
gen_uuid.c:(.text+0x418): undefined reference to `__aeabi_read_tp'
../../lib/libuuid.a(gen_uuid.o):gen_uuid.c:(.text+0x788): more undefined references to `__aeabi_read_tp' follow
I've tried to use the pre-compiled packages but it seems they don't work or i'm missing something.

---

fdisk is already included in initramfs thus no need for a parted binary.
edit: but to minimize data loss we need a resize2fs binary to resize the fat/ext3 partition
mkfs.ext3 (for the rw partition) and fsck are included too in the initramfs by archos

---

chulri said:
fdisk is already included in initramfs thus no need for a parted binary.
edit: but to minimize data loss we need a resize2fs binary to resize the fat/ext3 partition
mkfs.ext3 (for the rw partition) and fsck are included too in the initramfs by archos
Click to expand...
Click to collapse
Yep resize2fs is part of e2fsprogs. I've been working on the packages in the buildroot there seems to be a lot of errors but I finally succeeded in building the e2fsprogs package. but riseze2fs is not included. I need to check the config for that package maybe there is an option missing.

---

To get the buildroot working properly you need to copy the file
Code:
cp local/g8_arm/g8_arm.config .config
remove the line
Code:
package/apdf/Config.in
from .config.cmd
remove the line
Code:
depends on BR2_EXT_UCLIBC_VERSION_0_9_30_1
from toolchain/uClibc/Config.in
then in the buildroot directory execute
Code:
make menuconfig
enable the e2fsprogs in Package selection -> Harware handling
also you have to remove the --disable-resizer from the e2fsprogs.mk file else the resizer is not build.
I can't believe this build package from archos is very up-to-date it seems very strange that all these bugs are in there how where they ever building a good firmware.

---

but that builds dynamic linked binaries, doesn't it? but we need a static build, don't we?

---

chulri said:
but that builds dynamic linked binaries, doesn't it? but we need a static build, don't we?
Click to expand...
Click to collapse
Yeah stupid me it needs to run in the initramfs and that does not contain any libraries. Let me check to see if it's possible to link it statically.

---

and because it has to be statically linked, maybe we better just take the newest e2fsprogs (btw.: does this support FAT resizing?!) and compile it without that buildroot stuff (except for the toolchain of course, we need that to crosscompile)

---

chulri said:
and because it has to be statically linked, maybe we better just take the newest e2fsprogs (btw.: does this support FAT resizing?!) and compile it without that buildroot stuff (except for the toolchain of course, we need that to crosscompile)
Click to expand...
Click to collapse
Well I tried that and failed. That's why I wanted to do it in the buildroot. I'll try again later need to create a clean environment and do some diffs after I fixed all the stuff that's wrong.

---

wdl1908 said:
Well I tried that and failed. That's why I wanted to do it in the buildroot. I'll try again later need to create a clean environment and do some diffs after I fixed all the stuff that's wrong.
Click to expand...
Click to collapse
I think I got it.
Add
Code:
export BOARD=g8_arm
To your .bashrc then in the buildroot directory do a make it will take a while as it needs to build everything. This is just a precaution as i think the statically linked resize2fs needs some linking with uclib libraries.
After that is finished do a
Code:
make e2fsprogs LDFLAGS=-static
in the buildroot directory. You should find the resize2fs binary in the directory buildroot/project_build_arm/uclibc/root/sbin
Code:
$ file resize2fs
resize2fs: ELF 32-bit LSB executable, ARM, version 1 (SYSV), statically linked, not stripped

---

resize2fs does not resize vfat so we probably need parted and some extra utils

---

How to compile parted with buildroot.
I found the attached files on some forum
e3fsprogs.mk is a replacement for the existing file.
Config.in parted.mk and parted-001-ui.cast.patch need to be placed in the directory buildroot/package/parted
then execute the following commands
Code:
make e2fsprogs LDFLAGS=-static
make e2fsprogs-libs
make parted LDFLAGS=-static
you can find the statically linked parted in buildroot/build_arm/parted-2.3/parted
and this is what i tried.
Code:
# [B]parted /dev/block/mmcblk1[/B]
GNU Parted 2.3
Using /dev/block/mmcblk1
Welcome to GNU Parted! Type 'help' to view a list of commands.
(parted) [B]print[/B]
print
Model: MMC MMC08G (sd/mmc)
Disk /dev/block/mmcblk1: 7466MB
Sector size (logical/physical): 512B/512B
Partition Table: msdos
Number Start End Size Type File system Flags
1 8192B 7466MB 7466MB primary fat32 lba
(parted) [B]resize[/B]
resize
WARNING: you are attempting to use ./parted to operate on (resize) a file system.
./parted's file system manipulation code is not as robust as what you'll find in
dedicated, file-system-specific packages like e2fsprogs. We recommend
you use ./parted only to manipulate partition tables, whenever possible.
Support for performing most operations on most types of file systems
will be removed in an upcoming release.
Partition number? [B]1[/B]
1
Start? [8192B]?
End? [7466MB]? [B]6466MB[/B]
6466MB
(parted) [B]check[/B]
check
WARNING: you are attempting to use ./parted to operate on (check) a file system.
./parted's file system manipulation code is not as robust as what you'll find in
dedicated, file-system-specific packages like e2fsprogs. We recommend
you use ./parted only to manipulate partition tables, whenever possible.
Support for performing most operations on most types of file systems
will be removed in an upcoming release.
Partition number? [B]1[/B]
1
(parted) [B]quit[/B]
quit
Information: You may need to update /etc/fstab.
#[B]fdisk /dev/block/mmcblk1[/B]
The number of cylinders for this disk is set to 227840.
There is nothing wrong with that, but this is larger than 1024,
and could in certain setups cause problems with:
1) software that runs at boot time (e.g., old versions of LILO)
2) booting and partitioning software from other OSs
(e.g., DOS FDISK, OS/2 FDISK)
Command (m for help): [B]p[/B]
Disk /dev/block/mmcblk1: 7465 MB, 7465861120 bytes
4 heads, 16 sectors/track, 227840 cylinders
Units = cylinders of 64 * 512 = 32768 bytes
Device Boot Start End Blocks Id System
/dev/block/mmcblk1p1 1 197327 6314445+ c Win95 FAT32 (LBA)
Command (m for help): [B]n[/B]
Command action
e extended
p primary partition (1-4)
p
Partition number (1-4): [B]2[/B]
First cylinder (197327-227840, default 197327): Using default value 197327
Last cylinder or +size or +sizeM or +sizeK (197327-227840, default 227840): Using default value 227840
Command (m for help): [B]p[/B]
Disk /dev/block/mmcblk1: 7465 MB, 7465861120 bytes
4 heads, 16 sectors/track, 227840 cylinders
Units = cylinders of 64 * 512 = 32768 bytes
Device Boot Start End Blocks Id System
/dev/block/mmcblk1p1 1 197327 6314445+ c Win95 FAT32 (LBA)
/dev/block/mmcblk1p2 197327 227840 976426+ 83 Linux
Command (m for help): [B]w[/B]
The partition table has been altered!
Calling ioctl() to re-read partition table
# [B]mkfs.ext3 -v -I 128 /dev/block/mmcblk1p2[/B]
mke2fs 1.40.9 (27-Apr-2008)
Filesystem label=
OS type: Linux
Block size=4096 (log=2)
Fragment size=4096 (log=2)
61184 inodes, 244106 blocks
12205 blocks (5.00%) reserved for the super user
First data block=0
Maximum filesystem blocks=251658240
8 block groups
32768 blocks per group, 32768 fragments per group
7648 inodes per group
Superblock backups stored on blocks:
32768, 98304, 163840, 229376
Writing inode tables: done
Creating journal (4096 blocks): done
Writing superblocks and filesystem accounting information: done
This filesystem will be automatically checked every 28 mounts or
180 days, whichever comes first. Use tune2fs -c or -i to override.
# [B]tune2fs.static -c -1 -i 0 -m 1 /dev/block/mmcblk1p2[/B]
tune2fs 1.40.9 (27-Apr-2008)
Setting maximal mount count to -1
Setting interval between checks to 0 seconds
Setting reserved blocks percentage to 1% (2441 blocks)
I leave the scripting to you but with these utils it should work perfectly to resize the partition and create the second partition.
I copied the mkfs.ext3, fdisk and tune2fs.static from the recovery initramfs

---

I started hacking around and I came to the conclusion that it would be better to change the initramfs to mount /data from mmcblk1p2 if that partition exists and not if it's not existing and move the whole partition resizing, partition creating, partition deleting (if you wan't to revert without dataloss) and again resizing into an app. so the user has more control over what he's doing and see's if something is failing and not just get's a bootloop or some fancy log file.
edit: /data is mounted by /init.rc script, all of the above can be done by an app, incl. modifying /init.rc script. no need for any special initramfs, yay!
everybody who has +rw rooting installed will be able to use that app. I'm starting development...
edit2: project page online: http://code.google.com/p/archos-gen8-sde-rooting/
stay tuned

---

First test app: http://code.google.com/p/archos-gen8-sde-rooting/downloads/detail?name=AppDataResizer_v0.1.apk
Release notes:
initial test version v0.1:
- parted binary added
- test button lists partitions of mmcblk1 device
Click to expand...
Click to collapse
note: 250 GB version of the A70 is currently not supported.

---

chulri said:
First test app: http://code.google.com/p/archos-gen8-sde-rooting/downloads/detail?name=AppDataResizer_v0.1.apk
Release notes:
note: 250 GB version of the A70 is currently not supported.
Click to expand...
Click to collapse
Nice. Yep I was thinking about the transition from standard to custom and also came to the conclusion it had to be done outside the boot process scripts.
Edit: Should this app be installable via the usual way or should it be a system app? (Copied to /system/app)
What I was thinking was split the process in 3 steps.
Step1: Resize storage partition, Add new-data partition and format.
Step2: Copy existing /data to /new-data
Step3: Enable/Disable new-data
Maybe a step2a: To run after upgrade of firmware to check things that have changed.
The step1 requires a reboot as the partitioning should be done in the initramfs if you do that when apps are running you're going to have a hell of a time getting the storage partition unmounted (I know I had the problem when testing the parted binary)
Step2 can be done without any problem when storage is mounted and Step3 requires a reboot after the init.rc is changed.

---

wdl1908 said:
Nice. Yep I was thinking about the transition from standard to custom and also came to the conclusion it had to be done outside the boot process scripts.
Edit: Should this app be installable via the usual way or should it be a system app? (Copied to /system/app)
Click to expand...
Click to collapse
usual way (download and install) (or maybe I include it in the initramfs (like the Superuser.apk) and copy it to /system/app, but I don't like modifying initramfs any further, no need for 100 different versions )
wdl1908 said:
What I was thinking was split the process in 3 steps.
Step1: Resize storage partition, Add new-data partition and format.
Step2: Copy existing /data to /new-data
Step3: Enable/Disable new-data
Maybe a step2a: To run after upgrade of firmware to check things that have changed.
Click to expand...
Click to collapse
great, that were my plans too.
wdl1908 said:
The step1 requires a reboot as the partitioning should be done in the initramfs if you do that when apps are running you're going to have a hell of a time getting the storage partition unmounted (I know I had the problem when testing the parted binary)
Click to expand...
Click to collapse
No API to unmount /sdcard/? I think I got one: IMountService it's not a public API but android.os.FileUtils isn't public either and it's working great. I think IMountService is the API that the popup, which pops up when you connect your android device to the computer, uses. I think, no need to worry because android handles everything pretty well when you connect your device to the computer, isn't it?
wdl1908 said:
Step2 can be done without any problem when storage is mounted and Step3 requires a reboot after the init.rc is changed.
Click to expand...
Click to collapse
agreed

---

chulri said:
usual way (download and install) (or maybe I include it in the initramfs (like the Superuser.apk) and copy it to /system/app, but I don't like modifying initramfs any further, no need for 100 different versions )
Click to expand...
Click to collapse
I tried to install it but it won't install. logcat gives something like
Pckage chrulri.gen8.AppDataResizer has no certificated at entry res/layout/main.xml
Yep I agree no need for different versions of the initramfs
chulri said:
No API to unmount /sdcard/?
Click to expand...
Click to collapse
The problem is not only the mount of /mnt/storage if the user has used move2sd there are a lot of other mounts present that also uses that partition.
maybe the API will do. Needs to be tested.
One other remark. Let the AppDataResizer check for the unionfs directory so that you can be sure you'r running on the correct initramfs.