[Q] weird kernel function - Android Q&A, Help & Troubleshooting

As title suggests, there is a wierd (and Very annoying, since it forces me to use a while loop in a script to stick back to interactive) function in my kernel, which forces the CPU governor into ondemand every time screen-on/off occurs
Phone: Acer Liquid Glow E330
Kernel: all, this thing applies to stock, vache's, and lupohirp's kernel
PrintK says something about a str_governor variable (set to ondemand) which by circumstances gotta be the default governor
Any clues? Ideas?
Sent from my scytheaxe

SuicideFlasher said:
As title suggests, there is a wierd (and Very annoying, since it forces me to use a while loop in a script to stick back to interactive) function in my kernel, which forces the CPU governor into ondemand every time screen-on/off occurs
Phone: Acer Liquid Glow E330
Kernel: all, this thing applies to stock, vache's, and lupohirp's kernel
PrintK says something about a str_governor variable (set to ondemand) which by circumstances gotta be the default governor
Any clues? Ideas?
Sent from my scytheaxe
Click to expand...
Click to collapse
I never encounter such issue before, maybe you can change the kernal.

angelamar90 said:
I never encounter such issue before, maybe you can change the kernal.
Click to expand...
Click to collapse
Already tried vache's kernel and lupo's kernel, also tried to mod the stock boot.img to no avail. The issue is not in the ramdisk
Sent from my scytheaxe

Related

Latest stable 2.6.32 kernel zimage

Google just moved out the 2.6.32 kernel from the experimental git repo to the msm repo. This means that google thinks they are stable now.
There are currently two 32 branches:
android-msm-2.6.32-nexusonec
http://android.git.kernel.org/?p=kernel/msm.git;a=shortlog;h=refs/heads/android-msm-2.6.32-nexusonec
android-msm-2.6.32
http://android.git.kernel.org/?p=kernel/msm.git;a=shortlog;h=refs/heads/android-msm-2.6.32
I've compiled both and I'm finding the nexusonec branch more stable than the 32 branch. The 32 branch causes issues with the gallery and camera app not recognizing the sd card. Everything seems to work fine in the nexusonec branch (sdcard, wifi, BT, camera, etc). Speed is definitely faster than 29.
For those who are interesting in flashing the 32 zimage, they can d/l it here:
http://drop.io/xdp6104#
Right now the kernel is just a stock config, meaning no tethering or anything else. Only change I've made from stock is to use the deadline I/O scheduler. I'm sure cyanogen will add those other goodies when he releases a new version.
Very nice: cant wait for a Enomther rom with this new Kernel
nice....
ive compiled a nice cyanogen kernel 2.6.32 already
just need to get the wifi module compiled correctly and we are gtg!
Does this mean that they've fixed some of the power saving regressions people were seeing with 2.6.32 before?
Installed on my Nexus with enomther 1.61: works nicely
Tot mem 395448K, Free 265916K
Does this still have the charging issue that blocks the battery to charge when it goes under 90%...???
eug89 said:
Does this still have the charging issue that blocks the battery to charge when it goes under 90%...???
Click to expand...
Click to collapse
86% now and charging..
eug89 said:
Does this still have the charging issue that blocks the battery to charge when it goes under 90%...???
Click to expand...
Click to collapse
It wasn't an issue per-se it was a 'feature'
The problem was caused by downgrading to older kernels after using 2.6.32. Certain hardware states that were set by the new kernel telling the battery not to charge weren't being un-set by the new kernel, and the older kernel didn't even know about them. They patched the newer kernel pretty quickly after getting feedback so that it removes the hardware state when shutting down, I believe. So, yes, it's 'fixed'
FettsVett said:
It wasn't an issue per-se it was a 'feature'
The problem was caused by downgrading to older kernels after using 2.6.32. Certain hardware states that were set by the new kernel telling the battery not to charge weren't being un-set by the new kernel, and the older kernel didn't even know about them. They patched the newer kernel pretty quickly after getting feedback so that it removes the hardware state when shutting down, I believe. So, yes, it's 'fixed'
Click to expand...
Click to collapse
Thanks, i called it issue because i didn't know how to call it but i knew that it was meant to be a sort of new feature of the 2..6.32 kernel and I wanted to write above 90% and not under 90%... did a bit of a mess
woahhh its way faster
thanks dwang
will do some thorough testing and post back soon.
edit: Hmmm having some trouble with fastboot... unable to push wifi via adb?
anyone else encountering this?
Are you trying to push the ko module on fastboot mode?
This kernel is FAST and SNAPPY! HIGHLY RECOMMENDED!
boqonx said:
Are you trying to push the ko module on fastboot mode?
This kernel is FAST and SNAPPY! HIGHLY RECOMMENDED!
Click to expand...
Click to collapse
yeah. iirc thats where you do it right?
Let the OS run, then push through adb. Can't do it when it's on recovery either because the /system/lib is not mounted. Wifi should work even without a restart.
yep just did that
i think i was just confused on where to use adb
it still works so all good. this kernel is zippy.
Damn right This thing flies now! Scrolling through messages is MUCH smoother. Seen a lot of tearing when scrolling on CM's kernel.
I released my undervolted version of CM's in the other thread. He changed the scheduler from anticipatory to deadline as well.
Will try! Had any reports on instabilities yet?
Dwang: great job on the deadline scheduler suggestion. It really does make the whole experience smoother.
Can someone write the exact adb command to push it?
Code:
adb remount
adb push ./bcm4329.ko /system/lib/modules
himmatwala said:
Can someone write the exact adb command to push it?
Click to expand...
Click to collapse

[Q] non OC Kernel for CM7

Is there a Kernel that is not OC for CM7?
it does not come oc'd at stock.
912MHZ is the default max scaling clock.
kernel does not OC unless set by user after install. Correct?
Correct.
1ghz max is default with latest kernel.
i am getting piss poor result on my tablet. after nvflash then cm7 is getting 13xx and 15xx and most of the other time it would give me "service called failed" even tho it is connected. is it possible nvflash did not remove the kernel? what would i need to flash to bring it back to "normal"?
Flash the latest kernel. This corrects a bug with cpu1 not hotplugging properly. Not contained in the latest nightly as of yet.

[Q] CPU not Stepping Down - Warm Phone

Hi.
I'm facing a "hard to solve" issue with my Atrix 4G...
It is a non-ATT (international). Using RSD Lite, I flashed the Retail SBF and it worked OK, battery life was "normal" and the phone was cold most of time.
I then unlocked the bootloader flashing the IHOP SBF then using moto_fastboot method. Installed Clockworkmod Recovery and flashed the HomeBase rom with the updates.
What happen is: The phone starts to get very warm and the battery is drained very quickly (couple of hours).
Using the SetCPU, I could verify that the CPU does not scale down automatically. It's always at 1300Mhz, it doesn't matter if the phone is idle or not.
I can reduce the CPU speed using the SetCPU but the processor is with the two cores active all the time and it never reduces the speed below the limit set in SetCPU.
I thought it may be a problem with my phone, so I flashed the retail SBF again and I could verify that, bakc with the retail rom, CPU is scaling down correctly. It goes up to 1000Mhz when requested but stays ~300Mhz when phone is idle.
I did the BL unlock/motofastboot/clockwordmod/ procedure again and tried with different roms (Aura, Alien, RedPill, etc) always cleaning up the Cache/UserData/Dalvik and using the faux 1.3Ghz unified kernel but got no result, on every 2.3.4 rom I flash, the CPU is always at the maximum speed set by the kernel or limited by the SetCPU.
I've found a similar problem on this thread: http://forum.xda-developers.com/showthread.php?t=1190634. and did the same investigation but I was not able to identify any aplication that is making the CPU usage to be constantly high (even tried killing all Apps, not result).
Any ideas what I can do/try to solve this issue?
Thanks a lot.
That happened to me too!! I just reflash the rom wiping everything and the cpu started acting normal
Try to use fastboot to wipe if ur using CWM
Sent from my CM7 Motorola Atrix 4G
For me turning off the phone and back on worked, if not for you then try wiping the usual stuff before reflashing a rom or back up
Feom what I've seen, this is often caused by the webtop. Most ROMs don't touch the webtop, and it seems for some people, this causes problems. I fixed my issue by installing a ROM that (I believe) wiped the webtop. That ROM was deprecated a long time ago so I don't know if it's still available. I'd take a look at the Developer forum. There are some webtop files you can download that reportedly fix the issue. You might also try using fastboot to erase the webtop.
Sent from my MB860 using XDA App
@vibrantuser
Thanks. Yes, I tried flashing and re-flashing couple of times, no luck whatsoever...
@Ilkinansr92
Thanks for the tip. Will try that and see how it goes.
@ronii1123
Will try that this weekend I'll report about progress.
@frelnik
Thanks for the tip. It makes sense as the retail SBF re-flashed the Webtop area and all roms I tried doesn't touch it. I'll try to get a webtop image and flash it using the fastboot or the CWM method. If you can point a link to any that may work, I'll really appreciate it.
Thanks a lot.
I installed the ROM from here to fix my issues. Every ROM thereafter was fine:
http://briefmobile.com/download-android-2-3-4-gingerbread-for-motorola-atrix-4g
I believe I've heard some people had success with the webtop CWM from here:
http://forum.xda-developers.com/showthread.php?t=1169457
I do wonder if erasing the webtop would be enough. You could try
fastboot erase webtop
faux123 said:
Just a quick Info dump:
Just going through the code and realized why SetCPU failed to keep the clock frequencies set properly. SetCPU has NOT been designed to work with dual core properly. It only sets the scaling_max_frequency for cpu0 (master CPU) and not for cpu1 (slave). When under load, cpu1 gets awaken by the system, but it's frequencies are still the old 1.3 GHz and since Tegra2 is a synchronous SMP design, the scaler would use the highest frequency again (this case 1.3 GHz). This is why after playing games, or apps that needed the 2nd cpu would cause the system clock to revert to original default 1.3 GHz even when you set it via SetCPU to something lower.
This is not easy to fix, when SMP is involved. I have tried a few hax already and still unable to tame the 2 headed beast. I will continue to work on it or find an alternative solution to SetCPU's issues
Click to expand...
Click to collapse
quoted from Faux himself, there's a bug in his kernel that prevents us from underclocking it. should be fixed in the next release.
Alcapone263 said:
quoted from Faux himself, there's a bug in his kernel that prevents us from underclocking it. should be fixed in the next release.
Click to expand...
Click to collapse
That's a separate problem, I think. My understanding is that the OP's issue is the CPU is always running at the maximum speed. It never clocks down to the slower speeds when idle. The issue with Faux's kernel and SetCPU is that no matter how you set it, when stressed, the CPU runs at the maximum speed the kernel is capable of, not the speed you set in the app. The CPU does clock down when idle, but will ramp up to 1.3 GHz even if you set it to only 1 GHz in SetCPU. I've experienced both issues.
You have to set it at 800 mhz for it to stick. For some reason setcpu only recognizes under clocking under 1 GHz, otherwise it presumes that it's running at the 1 GHz stock speed without registering the overclock, therefore thinking the stock is 1.3 GHz! I'm not sure but I would have a guess and say that it's something to do with the unchangeable governors? Hope this makes sense lol
Sent from my OC Atrix running homebase 4.7, simplicity themed using xda premium
[SOLVED] CPU not Stepping Down - Warm Phone
frelnik said:
I installed the ROM from here to fix my issues. Every ROM thereafter was fine:
http://briefmobile.com/download-android-2-3-4-gingerbread-for-motorola-atrix-4g
I believe I've heard some people had success with the webtop CWM from here:
http://forum.xda-developers.com/showthread.php?t=1169457
I do wonder if erasing the webtop would be enough. You could try
fastboot erase webtop
Click to expand...
Click to collapse
Hey freinik.
I cleaned up the webtop and reflashed the Home Base rom I was using.
Perfect!
CPU is now scaling correctly. Your suggestions was exactly what was needed.
After cleaning the webtop (using the fastboot erase webtop) the issue was solved.
Thanks a lot.
Think in having the same issue with my international atrix alien rom.
Can you provide clear instructions on how to remove webtop and what will it affect if any by removing it?
Hope you can help.
Sent from my mobile
@ultim
Hi, Thanks for asking.
I erased my webtop using the moto fastboot command line, you can get it here:
http://forum.xda-developers.com/showthread.php?t=1138092&highlight=fastboot
After installing, reboot you phone and enter the "fastboot" mode (holding volume down until the "fastboot" text appears, then press volume up.
Once on fastboot mode, connect your phone to the USB port and run:
moto_fastboot erase webtop
It should erase your webtop partition.
It normaly helps to clean the cache/data, you can do that by running:
moto_fastboot -w
You can then, reboot the phone.
moto_fastboot reboot
If you're using Linux, moto_fastboot should be run as superuser, you can do that by using sudo, ie. sudo moto_fastboot reboot.
Hope this helps.
@Nyx_ Thank you for the instructions. My battery life is so much better now. Before application 'X' (i assume its webtop) would use about a third of my battery all the time.
Thanks for this. I recently discovered the same thing was happening to my phone and deduced that the webtop was causing this. I was using faux's newest kernel. I then related the stock kennel and everything was file after that, no cleaning of the webtop was needed.
Sent from my MB860 using xda premium

[Q] New kernels and BLN

Ok I tried some new roms and started noticing a pattern in that BLN was behaving erratic.
By erratic I mean this.
If I set the LED's to stay on, they will turn on but then raondomly go off for a second or so and then come back on again, there was no pattern except it took longer to happen after a phone reboot and when the usb plugged in from pc.
If I set them to blink and I set the off period for longer than 4 seconds, then after they go off for the first or send time then after just 1 second they come back on again but only if I set the off to more than 4 seconds, weird.
In these cases was testing on freshly installed roms with wiped user data.
One of the roms was based on stock which made this even stranger as BLN works 100% perfect on stock rom with cf-root, the rom was based on cf-root b84 so I chanced it to downgrade it to cf-root b82 and bam BLN fixed.
So the issue is down to a fubar kernel.
Now CM's stock rom doesnt support BLN, so on the CM roms I was using the blackhawk rom which has the exact same symptons as cf-root b84, seems only cf-root b82 works properly.
Here is the changelog for cf-root b83 and b84.
CF-Root-S5830-v3.7-b84.zip (12.55 MB)
BUGFIX: Invisible boot animation.
Install CF-Root-S5830-v3.7-b84.zip via "CWM recovery" first and then install b84-fixed.zip via "CWM Manager - Flash kernel"!
CF-Root-S5830-v3.7-b83.zip (12.38 MB)
- Untested since on CyanogenMod 7.2 firmware for now
- Using b82 recovery image
- CPU and system bus overclock enabled
- GPU driver backported from 2.6.35-11 kernel source
- interactive, interactiveX, smart*ss, and smart*ssV2 governor
- Samsung on the fly governor setting disabled
- WiFi driver taken from DXKPD
Click to expand...
Click to collapse
I dont have b83 but am about to flash mythICS which is based on b83, my gut guess is the bug was introduced in b83 as b84 only has a tiny changelog.
What I am asking is if someone can either do me a kernel for CM based on the above information (so dont include new GP driver, dont disable on the fly governor, dnt add new wifi driver) and see if the result is a fixed kernel, or if anyone has instructions on how to build my own kernel based on CM's sources.
I dont really need a new kernel for stock based roms as I can just downgrade to b82 cf-root to fix them.
Also is this better in dev section or here?
confirmed b83 is indeed borked and b82 also fixed on mythICS.
So I need a b82 for CM

Governor control

How to get GOVERNOR CONTROL settings to stick with stock kernel?
I have set specific parameters (sampling_rate, up_treshold, hotplug_freq, etc) for each CPU core. But, it last only for couple minutes or until i lock my device. I don't know why but it's reverting back to default settings after i unlock it.
I know i can flash custom kernel, but i want to make it work with stock.
Any help 'bout this?
ROM used: Resurrection Remix 5.8.3
Not possible. You MUST have a custom kernel to change any aspects. Although you might think you are making changes until it reboots, it is a placebo. Stock kernels are set for life and cannot be changed.
Beamed in by telepathy.

Categories

Resources