FORUMS

[Kernel] -fox Vanilla (ext4) for XDAndroid devices

273 posts
Thanks Meter: 97
 
By Starfox, Senior Member on 16th August 2011, 10:40 PM
Post Reply Email Thread
This kernel is based upon the current git tree for 2.6.27 with minor tweaks included, such as ext4. This is suitable for all devices and should work as well as the autobuild. Please note this isn't an autobuild, so there might be a slight delay between a commit and an update to this thread. Also I do not intend to deviate much from the autobuild other than the changes mentioned below, so if you really need a feature, please include a compelling reason why it should be done.

You will need a new haret.exe as well as an additional line in startup.txt to boot this kernel. Instructions for that can be found at http://forum.xda-developers.com/show...08&postcount=1.

Commit 1348 (8/20): http://db.tt/UnLHCuR

Older version: Commit 1346 (8/16): http://db.tt/RhUlSDd

Changes:
Support for ext4 (You must use a rootfs dated 8/16 or later, and have the option ext4dev in your startup.txt cmdline)
Deadline I/O scheduler as default
Sleepmode 1 set by default (No longer needs pm.sleepmode=1 in your startup.txt cmdline)
Ondemand CPU governor minfreq lowered to 112MHz
Powersave and Conservative CPU governor added
Netfilter Redirect added as module
RNDIS set as default for tethering
Swap notifier patch for compcache included and enabled

Most of these changes are in .config, and can be verified by looking at /proc/config.gz in your running kernel. The swap notifier patch allows compcache to skip certain operations, which should improve performance if you do use ramzswap. Ext4 is still considered experimental in the .27 tree, but has been updated with Ted T'so's backport patches in his ext4 git.

Please again note the following on ext4 support:
1) You must use a rootfs dated 8/16 or later for the proper support
2) You must use my kernel, as it is the only one with the required support compiled.
3) You must have the cmdline option "ext4dev" inside your startup.txt. Otherwise it will still continue on using ext2 data.img.
4) Using ext4 will create a brand new file called data.ext4. This is intentional, as I did not want to have the same file (data.img) with different format to confuse the matter further. This also means that none of your existing app/setting will be preserved. Be sure to back up your data & app with something like Titanium Backup, and have at least 500MB free on your SD card so that the system can create a new store.

GPL availability notice:
kernel source as available from http://gitorious.org/linux-on-qualco...htc-msm-2.6.27
compcache module+patch from http://code.google.com/p/compcache/d...e-0.5.4.tar.gz
ti wlan git clone'ed from git://android.git.kernel.org/platform/system/wlan/ti.git
.config used to compile available from within /proc/config.gz

-- Starfox
The Following 6 Users Say Thank You to Starfox For This Useful Post: [ View ] Gift Starfox Ad-Free
17th August 2011, 01:50 AM |#2  
Senior Member
Thanks Meter: 45
 
More
Just loaded this one up and found something very odd... my ts-calibration was all off... Booted back into WM and deleted and re-calibrated... Still seeing the same issue...

Not sure why a module would do that but just thought I would let you know...

I went back to your 8/15 and 8/16 test kernel and could not even use the slide lock... This one is really wierd... Going to delete my data.ext4 and see if I can resolve that way...

Thanks for all the hard work you've been putting in to come up with these kernels...

Edit: Just saw the update on this page... Thanks...
17th August 2011, 03:45 AM |#3  
OP Senior Member
Thanks Meter: 97
 
Donate to Me
More
Sorry about having to pull the file, I've recompiled the whole thing using gcc-4.4 and the touchscreen seems normal again. For those who may have downloaded the first file, please download it again, even if you don't have issues. The correct filesize for the new version is 2460710 bytes.

For the technical-mined out there, the first attempt was compiled using arm-linux-gnueabi-gcc 4.5.2-8 under Ubuntu 11.04 amd64. Certain versions of compiler may introduce bugs where it might over-optimize certain things, and to complicate things, we are trying to make it run under a different architecture. That can introduce unforseen bugs due to endianess, data size differences, etc. that would otherwise not happen. Cross-compiling is a voodoo art at best, and the kernel source tree expands to ~900MB including all the intermediate files produced during the build process.

If you have a Ubuntu 11.04 machine and would like to help me determine if it was actually a cross-compile bug, send me a PM and I'll try to walk you through setting up the build environment and compiling the kernel. Since the bug popped up with the touchscreen, it will be relatively easy to confirm whether the compiler was responsible.

-- Starfox
The Following User Says Thank You to Starfox For This Useful Post: [ View ] Gift Starfox Ad-Free
18th August 2011, 07:06 AM |#4  
Junior Member
Thanks Meter: 0
 
More
Hmm - I must be missing something.

It says:
ext2fs_check_if_mount: No such file or directory while determining whether /dev/block/loop0 is mounted.
EXT3-fs (loop0): error: couldn't mount because of unsupported optional features (240)
EXT2-fs (loop0): error: couldn't mount because of unsupported optional features (240)
MILFS: Can't find milfs on dev loop0
mount: mounting /dev/block/loop0 on /dev failed: Invalid argument
Failed
Failed to mount /data (/sdcard/)

Anything I'm missing - I've triple checked all the instructions.

* RHOD100 UK
* Running on SD Card with new Haret and amended Startup.txt
18th August 2011, 07:22 AM |#5  
OP Senior Member
Thanks Meter: 97
 
Donate to Me
More
What exactly is in your startup.txt cmdline, and which rootfs are you using?

-- Starfox
18th August 2011, 04:49 PM |#6  
Junior Member
Thanks Meter: 0
 
More
Right. So, I went back to the Commit 1346 (8/16) and that went away.

Now I'm getting "ignored duplicate definition of service" and the "request_suspend_state" loops... it might be due to the new Class 10 SD that I'm using for this testing - Going to find an old Class 2 and try again.

FYI - the startup.txt command line is
set cmdline "lcd.density=240 msmvkeyb_toggle=off gsensor_axis=2,1,3 force_cdma=0 msmsdcc_1bit msmsdcc_fmax=14000000 msmsdcc_nopwrsave pm.sleep_mode=1 hw3d.force=1 acpuclock.oc_freq_khz=716800 ext4dev physkeyboard=rhod100_uk-b4HOME"

rootfs is rootfs-20110816-7e04198.zip
18th August 2011, 08:55 PM |#7  
Senior Member
Thanks Meter: 45
 
More
StarFox,
Loading this kernel now to test it out... I've been running your 8/16 test kernel for over two days with only the wifi being a problem. The LEDs have working great!

I'll let you know how this kernel goes... Hopefully better than the Cross-Compiled on I grabbed... LOL
The Following User Says Thank You to mgross029 For This Useful Post: [ View ] Gift mgross029 Ad-Free
19th August 2011, 01:04 PM |#8  
Senior Member
Thanks Meter: 45
 
More
StarFox,
Ran with this kernel through yesterday and into this morning... It all seems stable... However, I did notice longer than normal GPS signal acquisition times. When I was running the 1340 kernel I noticed this too. When I went to your Test 8/16 kernel I was getting a lock much quicker. Going to go back to the 8/16 test kernel to verify this.

Thanks...
20th August 2011, 04:26 PM |#9  
OP Senior Member
Thanks Meter: 97
 
Donate to Me
More
Quote:
Originally Posted by bughunt

Now I'm getting "ignored duplicate definition of service" and the "request_suspend_state" loops... it might be due to the new Class 10 SD that I'm using for this testing - Going to find an old Class 2 and try again.

FYI - the startup.txt command line is
set cmdline "lcd.density=240 msmvkeyb_toggle=off gsensor_axis=2,1,3 force_cdma=0 msmsdcc_1bit msmsdcc_fmax=14000000 msmsdcc_nopwrsave pm.sleep_mode=1 hw3d.force=1 acpuclock.oc_freq_khz=716800 ext4dev physkeyboard=rhod100_uk-b4HOME"

Okay, first, you're running a class 10 card with 1-bit and 14Mhz freq. Take these ones out: force_cdma=0 msmsdcc_1bit msmsdcc_fmax=14000000 msmsdcc_nopwrsave hw3d.force=1

You can take this one out if running my kernel:
pm.sleep_mode=1

How old is this startup.txt? These are values that have not been used for a while.

-- Starfox
The Following User Says Thank You to Starfox For This Useful Post: [ View ] Gift Starfox Ad-Free
20th August 2011, 06:14 PM |#10  
OP Senior Member
Thanks Meter: 97
 
Donate to Me
More
Update to build 1348, the new liblights will require you to have it bindmounted.

I'd appreciate if people would post different machines this has successfully worked on. I know it works on Rhodium, looking for Diamond/Raphael/users. Thanks.

-- Starfox
22nd August 2011, 07:22 PM |#11  
Junior Member
Thanks Meter: 0
 
More
Thumbs up
Quote:
Originally Posted by Starfox

Okay, first, you're running a class 10 card with 1-bit and 14Mhz freq. Take these ones out: force_cdma=0 msmsdcc_1bit msmsdcc_fmax=14000000 msmsdcc_nopwrsave hw3d.force=1

You can take this one out if running my kernel:
pm.sleep_mode=1

How old is this startup.txt? These are values that have not been used for a while.

-- Starfox

Right - that's fixed. It now boots up and seems a bit more snappy than before.
Post Reply Subscribe to Thread

Tags
ext4, fox, vanilla, xdandroid

Guest Quick Reply (no urls or BBcode)
Message:
Previous Thread Next Thread
Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes