[KERNEL] [Y300/G510] Stock Huawei

[Guest porozex]

yeah tried them too and didnt see any difference. mostly is a matter of ram if it gets full everything lags phone,games,apps etc.. there is nothing to do even if i raise the minfree values, yes it helps but not always heavy apps will still lag

[Guest]

A little bit off-topic ... but here are my 2 cents.

 

Looking at the source code

	inline QCCompositionType::QCCompositionType()
	{
	 char property[PROPERTY_VALUE_MAX];
	 mCompositionType = 0;
	 if (property_get("debug.sf.hw", property, NULL) > 0) {
	 if(atoi(property) == 0) {
	 mCompositionType = COMPOSITION_TYPE_CPU;
	 } else { //debug.sf.hw = 1
	 property_get("debug.composition.type", property, NULL);
	 if (property == NULL) {
	 mCompositionType = COMPOSITION_TYPE_GPU;
	 } else if ((strncmp(property, "mdp", 3)) == 0) {
	 mCompositionType = COMPOSITION_TYPE_MDP;
	 } else if ((strncmp(property, "c2d", 3)) == 0) {
	 mCompositionType = COMPOSITION_TYPE_C2D;
	 } else if ((strncmp(property, "dyn", 3)) == 0) {
	#ifdef USE_MDP3
	 mCompositionType = COMPOSITION_TYPE_DYN | COMPOSITION_TYPE_MDP;
	#else
	 mCompositionType = COMPOSITION_TYPE_DYN | COMPOSITION_TYPE_C2D;
	#endif
	 } else {
	 mCompositionType = COMPOSITION_TYPE_GPU;
	 }
	 }
	 } else { //debug.sf.hw is not set. Use cpu composition
	 mCompositionType = COMPOSITION_TYPE_CPU;
	 }
	
	}
https://android.googlesource.com/platform/hardware/qcom/display/+/jb-mr1.1-dev-plus-aosp/libqdutils/comptype.h

...there are 2 things to notice. The value for composition type only gets used when debug.sf.hw is set. And the typedef of "DYN" seems to make dynamically use of either C2D or MDP at the same time. Which one gets chosen depends on how the lib are being compiled.

 

[Guest ZolaIII]

3rd thing for you to notice that C2D is invalid as it's not present in driver.

[Guest Syssx420]

About those graphical glitches with new KGSL.

I have an idea. I've found that Huawei's kernel sometimes ignores(?) debug.sf.hw line and instead of using the Adreno libs in /system/lib/egl it uses libgles_android.so and eglsubAndroid.so (they're probably used for software rendering or what). Did you tried to delete this line "0 0 android" from /system/lib/egl.cfg and just to be sure also deleting those two files?

With those two files I've sometimes had like 6200+ in Antutu using your kernel and Heros's AOSP MOD rom and sometimes I had 7800+ even 8020 (Everything after fresh install with debug.sf.hw line enabled). Scores were changing with some restarts. It was really random like 6 restart with 7800+ and then two restart with only 6000+.

Maybe it glitches because phone is not using Adreno libs?

Also I'm a total noob over this I'm just trying to help.

 

edit: Forgot to say that after deleting those two files and modifying egl.cfg Antutu score is stable.

Edited July 5, 2014 by Syssx420

[Guest juanpablocastillo]

moddingg33k: Any progress in integrating BFS scheduler in the kernel?

[Guest]

Tried it 3-4 times, no success. Other people failed too at it already. Look at this (last 2-3 comments on that page): http://ck-hack.blogspot.de/2012/07/bfs-424-linux-34-ck3.html

 

Unfortunately I got no clue how to debug the kernel if it panics before starting the init phase, so I can't track down the issue and fix it

[Guest juanpablocastillo]

There are a couple of posts on XDA that talk about it:

http://forum.xda-developers.com/showthread.php?t=1502037

But is old (of 2.6.x kernel era).

Here is other incorporated in a kernel for the Galaxy W (kernel 3.4.66 for more precision):

http://forum.xda-developers.com/showthread.php?t=2267671

[Guest]

Thx for those links. The issues isn't that the kernel wouldn't compile. It compiles fine without any errors. The problem is that the device can't boot with integrated and enabled BFS

[Guest]

I've updated the kernel: http://www66.zippyshare.com/v/526858/file.html
 
Changes:

• revert possible problematic commits
• upgraded kernel to 3.4.98 (seems to cause some minor regressions which I can hopefully fix in upcoming releases )
• updated:
  • KGSL drivers (only partially)
  • MSM video drivers (for LCD display)

Don't expect any FPS boost. The most you might get is a more stable UI/GPU experience.

Edited July 13, 2014 by Guest

[Guest Domino2115]

Phone sometime cant wake up from deepsleep, phone freeze and i have to pull out battery.

[Guest porozex]

Phone sometime cant wake up from deepsleep, phone freeze and i have to pull out battery.

i can confirm that same problem on h3ros aosp mod

[Guest]

On which ROMs do those problems happen? I did pull updated ramdisk files from the B200 kernel and merged them into my's.

[Guest porozex]

On which ROMs do those problems happen? I did pull updated ramdisk files from the B200 kernel and merged them into my's.

aosp mod is based on b200

[Guest]

@porozex: So you are currently using H3ROS's ROM, right? Does the same problem happen with the stock B200 kernel?
 
I've updated the uploaded file. Only the content of the ramdisk got restored to a previous state: http://www57.zippyshare.com/v/18200987/file.html -> http://www66.zippyshare.com/v/526858/file.html

Edited July 12, 2014 by Guest

[Guest porozex]

@porozex: So you are currently using H3ROS's ROM, right? Does the same problem happen with the stock B200 kernel?

 

I've updated the uploaded file. Only the content of the ramdisk got restored to a previous state: http://www57.zippyshare.com/v/18200987/file.html

as i remember when i was using it, it never happend

[Guest Domino2115]

On which ROMs do those problems happen? I did pull updated ramdisk files from the B200 kernel and merged them into my's.

B171 o2

[Guest]

@Domino: w000t That's a very special version of an german mobile carrier company. Did you use newer ROMs, too? Or is it that you got a special reason using exactly this version? :)

[Guest]

Phone sometime cant wake up from deepsleep, phone freeze and i have to pull out battery.

 

Errrrm ... just happened to me as well right now (after those other 100 wakeups before worked all of the time).

 

Is it worth to track and fix this issue? Or isn't there any change at all compared to older kernel builds?

 

Fix attempt: http://www66.zippyshare.com/v/526858/file.html

Edited July 12, 2014 by Guest

[Guest Domino2115]

It happend just with today build. Okay downloading thanks

[Guest]

I fear I will stop kernel development for this device. It's not worth the effort any longer.

[Guest Domino2115]

I fear I will stop kernel development for this device. It's not worth the effort any longer.

nooooo.. :/

[Guest H3ROS]

The problem is with Huawei and their kernel source. It's butchered, plus the source that they provide doesn't match up with what they distribute in their updates. You can tell because the config file contains things that aren't found in the 2013 source.

 

The B200 ramdisk shouldn't be the problem since my ROM uses the B200 kernel. So one of the commits has caused the deep sleep issue, but which one? I'd look at any sleep, suspend or power related commits made over the last 2 days. At a glance I can see that there's a few.

 

It seems that things become the most unstable once the kernel has been updated, which probably happens because of the changes Huawei has made. So maybe keep things at the stock version and simply backport any features and performance improvements?

Edited July 12, 2014 by H3ROS

[Guest]

The current commit history is quite short. Next to implementing some additional schedulers it only contains some driver updates and some small fixes for GCC and other stuff.

 

The sleep issue got probably caused by the updates of LCD drivers. Seems like one of those commits related caused an race/deadlock situation.

 

 

Now, being honest to myself: A stock Huawei kernel performs much better and there isn't anything being improved by this custom kernel what an average user could profit from.

[Guest H3ROS]

zRAM, SWAP and frandom are all useful though since it's a low RAM device. So the question there is can you manage to compile modules that the stock kernel can load? I got close with the vermagic for frandom, but then it gave me a module_layout disagreement error. That might be fixable though if someone was to compile it with the Google 4.6.x toolchain and the same kernel config as B200.

 

Someone suggests hex editing here too: http://stackoverflow.com/questions/2720177/module-layout-version-incompatibility

Edited July 12, 2014 by H3ROS

[Guest]

@H3ROS: on this post http://www.modaco.com/topic/372447-kernel-y300g510-stock-huawei/page-28#entry2218617 it seems like you figured out all that stuff on your own?

 

Hacking modules to get them working on already compiled kernels isn't something I'm hot on spending time on. It's possible, no question.

 

The ZRAM version of my current kernel build would not work on a stock kernel I guess.

 

Frandom however is a module which should be working fine.

 

SWAP is something you have to enable in the kernel config during you build a kernel. It's nothing you can compile as module.