[KERNEL] [Y300/G510] Stock Huawei

[Guest ZolaIII]

I've installed lastest experimental but phone dont even boot it says'optimalizing apps" and when its done phone reboot

If you didn't get it on the last nights nightly the WiFi module is not working you left it no when you flashed the kernel so its casing constant reboots when trying to initialized.

[Guest AssaSsiNMiLeS]

I'm using your Linaro 4.8.3 version from yesterday before you reverted the LMK changes and it is still working perfectly. I guess that the LMK changes weren't responsible for any problems and instead it were my apps which caused that. Now I'm gonna try your new 4.9 build.

Edited June 12, 2014 by AssaSsiNMiLeS

[Guest Domino2115]

I'm using your Linaro 4.8.3 version from yesterday before you reverted the LMK changes and it is still working perfectly. I guess that the LMK changes weren't responsible for any problems and instead it were my apps which caused that. Now I'm gonna try your new 4.9 build.

I had same apps like with kernel from 31th of june no problems then I installed kernel with lmk and oom patches it start freezes,laggs and after that i installed again kernel from 31th of june no problems.

[Guest]

GCC 4.6: http://www58.zippyshare.com/v/49859364/file.html

 

GCC 4.8: http://www3.zippyshare.com/v/33709362/file.html

 

GCC 4.9: http://www42.zippyshare.com/v/41874288/file.html

 

I think Cexstel's kernel is also compiled with Linaro 4.6.4 if he didn't change from it.

 

From my experience GCC versions higher than 4.6+ improve UI responsiveness while invoking the risk of potential untraceable issues on older kernel versions. And yes, 3.4.y is old

Edited June 12, 2014 by Guest

[Guest juanpablocastillo]

I don't understand. So, you can't merge the CAF 3.4 kernel or you can not do as you wanted? Sorry if annoying.

I put this here in case you did not see. Sorry if upset again.

Edited June 12, 2014 by juanpablocastillo

[Guest]

I saw that and I promised to answer it:

 

While working on many different things on the current kernel I noticed, that the MSM-3.4 kernel (even of 2013) is out of sync with Linux's mainline. They add a lot of updates to become closer again, but are still lagging some mainlined ARMv7 patches for example. That's fixed in the MSM-3.7 kernel, because they took a fresh snapshot of the mainline sources and integrated their CodeAurora specific changes into it. Unluckily this version doesn't get maintained anymore and there might also be problems with existing older drivers not running on this version.

 

Using the complete CAF MSM-3.4 snapshot might also not be the best option, since the support for older hardware got dropped. Our MSM7627a board is one of those examples. The same goes for our MSM8225 CPU. I'm currently unsure which way might be the best to go. The only thing which is very clear: it's not worth the time to completely integrate commit per commit into huawei's published source code. That would take too long.

 

Like experienced with the last 2-3 experimental builds another issue became clear: the more that get's updated the more problems occur suddenly.

 

I'm up to any suggestions.

[Guest]

For anyone who didn't try it yet and want's to test it, here's a re-packed kernel of chil360 which is getting used in his SlimKat ROM

 

http://www13.zippyshare.com/v/17579740/file.html

Edited June 12, 2014 by Guest

[Guest juanpablocastillo]

Ok, thanks for clarify that.

[Guest]

I'll start an attempt whether an clean 3.4.93 linux kernel is able to run on our device and only merging necessary things from CodeAurora which are MSM specific.

[Guest AssaSsiNMiLeS]

I think as a next step you can implement ROW scheduler and update the FM drivers as you already wanted to do. Then you could maybe give the LKM another try and then you still can decide if our device would benefit of some cherry-picks from 3.4 CAF.

 

Edit: Sorry, didn't read your post early enough :D

Edited June 12, 2014 by AssaSsiNMiLeS

[Guest Chefo93]

I saw that and I promised to answer it:

 

While working on many different things on the current kernel I noticed, that the MSM-3.4 kernel (even of 2013) is out of sync with Linux's mainline. They add a lot of updates to become closer again, but are still lagging some mainlined ARMv7 patches for example. That's fixed in the MSM-3.7 kernel, because they took a fresh snapshot of the mainline sources and integrated their CodeAurora specific changes into it. Unluckily this version doesn't get maintained anymore and there might also be problems with existing older drivers not running on this version.

 

Using the complete CAF MSM-3.4 snapshot might also not be the best option, since the support for older hardware got dropped. Our MSM7627a board is one of those examples. The same goes for our MSM8225 CPU. I'm currently unsure which way might be the best to go. The only thing which is very clear: it's not worth the time to completely integrate commit per commit into huawei's published source code. That would take too long.

 

Like experienced with the last 2-3 experimental builds another issue became clear: the more that get's updated the more problems occur suddenly.

 

I'm up to any suggestions.

About the part where you say that it is somewhat possible to use MSM-3.7 kernel... If you have the time and patience, and if that version is already stable, why not give it a shot? You seem to be quite skilled judging by what you've shown so far! :) I understand the problems that could occur with the drivers, but is the lack of maintenance for msm-3.7 that bad? 

[Guest]

Hm...the maintenance thingy shouldn't be a too big problem. Since the MSM-3.7 is much closer to the original Linux it's in fact even much easier to apply commits cherry-picked from kernel.org.

 

The real problem I fear or better said one thing I'm definitely sure about: I needed to find a solution for the WiFi driver as it won't work anymore. I did several tests implementing kernel-core stuff from 3.7 and all of the time the WiFi driver refused to work.

 

Since getting some foreign kernel sources working is a lot of work I need to decided first which way to go, before wasting tons of hours and days for "nothing" at the end. It's not easy to make a smart decision. Of course it would maybe worth trying to get the MSM-3.7 kernel working. At least this seems to be the most promising solution so far.

 

On the build of 31.05. I probably achieved something amazing and thought I could even further advance from there on. Unfortunately the last days showed up a lot of troubles and made clear that there's too much work involved for a single person.

 

That's why I really need to find another way than simply adding commits to Huawei's kernel sources.

[Guest luca020400]

Why 3.7 ? In  https://android.googlesource.com/kernel/common/ there isn't 3.7 brach kernel but only 3.4 and 3.10

[Guest]

I'll see if it's possible to get an official Linux 3.4.93 kernel working first. This is the most maintained and most driver compatible version so far for an ARM-MSM platform.

 

Just downloaded the sources and try to get it compiled and running.

Edited June 12, 2014 by Guest

[Guest]

Why 3.7 ? In  https://android.googlesource.com/kernel/common/ there isn't 3.7 brach kernel but only 3.4 and 3.10

 

Google isn't the place to look for MSM kernel sources. CodeAurora is taking care of that and they got an MSM-3.4 and MSM-3.7 branch: https://www.codeaurora.org/cgit/quic/la/kernel/msm/tree/?h=msm-3.7

[Guest luca020400]

Ok I haven't seen the https://www.codeaurora.org

Sorry

[Guest]

You don't have to be sorry m8

 

Manufactures might have access to other platforms, too. But the only public sources I'm aware of are http://www.kernel.org and http://www.codeaurora.org

[Guest fonz93]

For what i tried in the past, chil360 kernel had not the same smoothness that your kernel has, maybe because is made with gcc 4.7, also there was a bug but i don t remember wich bug was

Edited June 12, 2014 by fonz93

[Guest luca020400]

WiFi thether reboot the phone

[Guest]

WiFi thether reboot the phone

 

You mean in the "fixed" version? I always use to check WiFi and WiFi-tethering. But sometimes I just don't when there is no reason why it should fail

[Guest luca020400]

No in the old version

I will try fixed version

 

EDIT : wifi tether doesn't work I got an error

Edited June 12, 2014 by luca020400

[Guest AssaSsiNMiLeS]

Your 4.9 build from post #317 works perfectly fine for me. UI is very smooth and all those lags/ freezes have gone away.

[Guest]

GCC 4.9 seems to have positive effects on the compiled kernel code. Thx for feedback!

[Guest AssaSsiNMiLeS]

I'm just curious: A few weeks ago you talked about testing DoubleTap2Wake or something. Will you try this once or do you think the battery impact is too heavy?

[Guest]

Let me finish my little experiment(s). If I fail I'll try to restore the performance of the current experimental build. After that I'm up again for such things :)