[Kernel][CWM]Fecality I9100 3/7/2012 (GPU Voltage Fix)

af974 · Feb 27, 2012

26/02, the last one.

khartaras said:
Weird... Are you running the 2/26 build? Or a different one?

khartaras · Feb 27, 2012

Entropy512 said:
The problem was intermittent - some people (including myself) never experienced it, others had it occasionally, one guy had it happen all the time.

It also affected the ability to touch the screen during phone calls for him.

Strange. It doesn't happen EVERY time, but it does happen very frequently. Oh well, since you have the problem pinned down I guess it's not really a problem any more.

sachs · Feb 27, 2012

La4 ROM....lost smooth, bad touch.

Sent from my GT-I9100 using Tapatalk

Entropy512 · Feb 27, 2012

khartaras said:
Strange. It doesn't happen EVERY time, but it does happen very frequently. Oh well, since you have the problem pinned down I guess it's not really a problem any more.

We'll see - I THINK I know what it is but I need time to look at the source.

This could be a bit difficult to get working properly, because it is I9100-specific and getting the right combo of #ifdefs could be difficult. Basically - some of the CONFIG_TARGET_LOCALE_NAATT ifdefs may be causing code that should be common to I9100 and I777 to be compiled for I777 only.

Entropy512 · Feb 27, 2012

sachs said:
La4 ROM....lost smooth, bad touch.

Sent from my GT-I9100 using Tapatalk

No clue what you're saying here... At least one of your complaints is addressed in the FAQ if you're saying what I think you're saying.

teiglin · Feb 27, 2012

F3nr1s said:
I dont have any probleme with the circle lockscreen too . i test with clear ram and i have the circle all times

Just to be clear, the issue isn't related to the circle lockscreen, or to any particular lockscreen. The problem is that the phone is slow to become responsive after waking from deep sleep, so it takes a five seconds or so before you can actually unlock the phone.

Entropy, let me know if there are any values or logs I can get you that would help.

Entropy512 · Feb 27, 2012

teiglin said:
Just to be clear, the issue isn't related to the circle lockscreen, or to any particular lockscreen. The problem is that the phone is slow to become responsive after waking from deep sleep, so it takes a five seconds or so before you can actually unlock the phone.

Entropy, let me know if there are any values or logs I can get you that would help.

I'm going to try and put out a test kernel tonight - given my history with this issue, I think I know the solution and unfortunately this one is next to impossible to get good logs for. And yes - this affected I777 users long before circle lockscreen existed. It happened when I tried to pull in the I9100 update3 touchscreen drivers - I eventually reverted those back to I777. However, I left the I777 ifdefs in there - so when building for I9100, the I9100 code remains.

What I can't figure out is why I9100 users aren't complaining about it left and right with any kernel that uses the I9100 update3 base. (Except Siyah, which uses i777 touchscreen drivers.)

teiglin · Feb 27, 2012

Entropy512 said:
I'm going to try and put out a test kernel tonight - given my history with this issue, I think I know the solution and unfortunately this one is next to impossible to get good logs for. And yes - this affected I777 users long before circle lockscreen existed. It happened when I tried to pull in the I9100 update3 touchscreen drivers - I eventually reverted those back to I777. However, I left the I777 ifdefs in there - so when building for I9100, the I9100 code remains.

What I can't figure out is why I9100 users aren't complaining about it left and right with any kernel that uses the I9100 update3 base. (Except Siyah, which uses i777 touchscreen drivers.)

Cool, I look forward to the test kernel, though I have it in relatively low volume (only four or five times since I flashed the kernel yesterday), so I'm probably not the best test case.

I'm trying to remember when I first started experiencing it. I don't think I've had it happen on pongster's kernel recently, but for a while I was bouncing between his kernel and siyah pretty frequently, just screwing around mostly, so what happened on which kernel is pretty muddled in my head going back more than a week or two. I'm pretty sure it was widely-complained-about for a while, so I'm guessing people besides GM must have also backported the I777 drivers (or found another fix?). I'll have to poke around git tonight to see.

Entropy512 · Feb 27, 2012

teiglin said:
Cool, I look forward to the test kernel, though I have it in relatively low volume (only four or five times since I flashed the kernel yesterday), so I'm probably not the best test case.

I'm trying to remember when I first started experiencing it. I don't think I've had it happen on pongster's kernel recently, but for a while I was bouncing between his kernel and siyah pretty frequently, just screwing around mostly, so what happened on which kernel is pretty muddled in my head going back more than a week or two. I'm pretty sure it was widely-complained-about for a while, so I'm guessing people besides GM must have also backported the I777 drivers (or found another fix?). I'll have to poke around git tonight to see.

I know GM definitely reverted back to the I777 drivers - His revert was why I reverted my source base. However, since I built for I777 target, my #ifdefs may be slightly different from his - i.e. I think his may enable the I777 code at all times, while mine only enables it when compiling for I777 (since until this week, I only compiled for I777 except for a test kernel I dropped in the VR thread.)

I'm home now, looking at the code.

Edit: Weird, GM left all of the #ifdefs checking for CONFIG_TARGET_LOCALE_NAATT in mxt224_u1.c there - so if you're not having the issue on SiyahKernel, it's not that file. Hmm.

Edit: Same for all CONFIG_TARGET_LOCALE_NAATT items in arch/arm/mach-s5pv310/mach-c1.c

I am wondering if maybe the I9100 community has other workarounds they use for touchscreen issues that just plain don't exist on the I777... Looking through all other touchscreen related commits in Siyah.

Right now the main difference I can find is a difference in the default touch threshold.

Try putting this into an init.d script to see if it changes anything:

Code:

#!/sbin/sh
log "Increasing touchscreen sensitivity"
echo "50" > /sys/devices/virtual/sec/sec_touchscreen/tsp_threshold

teiglin · Feb 27, 2012

Entropy512 said:
Right now the main difference I can find is a difference in the default touch threshold.

Try putting this into an init.d script to see if it changes anything:

Code:

#!/sbin/sh log "Increasing touchscreen sensitivity" echo "50" > /sys/devices/virtual/sec/sec_touchscreen/tsp_threshold

Okay, I'll set that and give it a shot (can I just set it or do I actually need to reboot?). You'll have to give me a bit since it doesn't always happen, plus I have to wait long enough to be sure the phone is in deep sleep. For the record, that value was 40 before I changed it.

Since you mention touchscreen sensitivity, I feel I should mention that I use vitalij's TouchScreenTune to shorten blen, though his app doesn't allow setting tsp_threshold. Could the two be related?

Edit: Tried twice now after putting airplane mode on and waiting a couple minutes, didn't happen either time with that tweak. I'll keep it there and let you know, though, since twice is hardly conclusive given the fairly low frequency with which I normally hit this.

Entropy512 · Feb 28, 2012

40? odd... It should default to 70 unless you had another init.d script change it.

And it can be changed on the fly.

That is something interesting - Over here in I9100-land there's all sorts of touchscreen fixes and tweaks. In I777 land, the only time people have ever complained about their touchscreen is when running kernels based on I9100 source.

teiglin · Feb 28, 2012

Entropy512 said:
40? odd... It should default to 70 unless you had another init.d script change it.

And it can be changed on the fly.

That is something interesting - Over here in I9100-land there's all sorts of touchscreen fixes and tweaks. In I777 land, the only time people have ever complained about their touchscreen is when running kernels based on I9100 source.

Hah, figures. I'll need to poke around my other tweaks--something is definitely setting it to 40, since my changes aren't sticking. I did hit the bug again, by the way, but I went back and looked at the tsp_threshold, and it was 40, despite the fact that I had manually changed it to 50 before turning the screen off.

Edit: It's not TouchScreenTune, and tsp_threshold doesn't even appear in any of my init.d scripts aside from the one I just created. Nevertheless, every time I turn my screen on, the value is set to 40. I have no idea what could be causing this--what is executed when you turn the screen on?

Entropy512 · Feb 28, 2012

I'm less sure that it's the threshold then, since something is overriding it anyway, it doesn't matter.

I'm wondering if one of the tweak apps you're using is somehow interfering...

teiglin · Feb 28, 2012

Entropy512 said:
I'm less sure that it's the threshold then, since something is overriding it anyway, it doesn't matter.

I'm wondering if one of the tweak apps you're using is somehow interfering...

Well, that's why I mentioned TouchScreenTune, as it's the only tweak that I use that (afaik) is doing anything when you turn the screen on. I've turned it off for now, and I will let you know if that removes the issue. My init.d stuff is pretty minimal, no superscripts or anything like that--just the stuff that came with CheckROM (loading cifs, sd readahead, changing nice of kswapd0), plus a tweak for my conservative thresholds.

I poked around a bit more, and it seems tsp_thresholds are set automatically, 40 or 70 when charging, which is consistent with the behavior I'm seeing.

cheewaca · Feb 28, 2012

Thank you for share your kernel, testing it now.

Entropy512 · Feb 28, 2012

teiglin said:
Well, that's why I mentioned TouchScreenTune, as it's the only tweak that I use that (afaik) is doing anything when you turn the screen on. I've turned it off for now, and I will let you know if that removes the issue. My init.d stuff is pretty minimal, no superscripts or anything like that--just the stuff that came with CheckROM (loading cifs, sd readahead, changing nice of kswapd0), plus a tweak for my conservative thresholds.

I poked around a bit more, and it seems tsp_thresholds are set automatically, 40 or 70 when charging, which is consistent with the behavior I'm seeing.

Hmm, good catch there.

I guess the question is whether khartras is also using touchscreentune - if he is too, then it may cause problems when combined with the I777 touch driver - but then why doesn't Siyah also have the same issues (as it seems to have the same code as far as I can tell, with the exception of possibly changing some defaults. I need to take a second look.)

Edit: I think I found a possible culprit, need to look at SiyahKernel - See https://github.com/Entropy512/linux...r/drivers/input/touchscreen/mxt224_u1.c#L3250

On I777, the suspend handler gets called later and the resume handler gets called earlier.

Well, I've found plenty of possible culprits - just no explanation why you didn't have issues in Siyah...

teiglin · Feb 28, 2012

Entropy512 said:
Hmm, good catch there.

I guess the question is whether khartras is also using touchscreentune - if he is too, then it may cause problems when combined with the I777 touch driver - but then why doesn't Siyah also have the same issues (as it seems to have the same code as far as I can tell, with the exception of possibly changing some defaults. I need to take a second look.)

Edit: I think I found a possible culprit, need to look at SiyahKernel - See https://github.com/Entropy512/linux...r/drivers/input/touchscreen/mxt224_u1.c#L3250

On I777, the suspend handler gets called later and the resume handler gets called earlier.

Thanks for your patience on this. Hopefully khartras will weigh in whether he is using TouchScreenTune--I haven't hit the bug since turning it off (including turning the screen on a few times for no reason but to check).

I don't totally understand what the data->early_suspend.level means--it seems like you're saying that setting it to the higher value (of DISABLE_FB+1 instead of BLANK_SCREEN+1) changes when the suspend and resume handlers are called? And that timing may be crossing wires with whatever TouchScreenTune is doing and causing the short hang when waking up/turning the screen on?

Edit: yeah, I'm not sure at all. Missed your edit while I was blindly wading through the various touchscreen edits. I guess it's not surprising that it could be a lot of things.

Entropy512 · Feb 28, 2012

Yeah. During suspend, the handlers are called in ascending order. During resume - descending order. So on I777, the earlysuspend handler is called later and the lateresume handler is called earlier.

Early suspend and late resume are screen off/on - deep sleep is regular suspend/resume.

As to CWM Manager compatibility - After poking at cfroot's initramfs, my head hurts. Do any other custom kernels support it?

teiglin · Feb 28, 2012

Entropy512 said:
Yeah. During suspend, the handlers are called in ascending order. During resume - descending order. So on I777, the earlysuspend handler is called later and the lateresume handler is called earlier.

Early suspend and late resume are screen off/on - deep sleep is regular suspend/resume.

As to CWM Manager compatibility - After poking at cfroot's initramfs, my head hurts. Do any other custom kernels support it?

I see, thanks for the explanation.

I thought Siyah supports CWM Manager, and I'm pretty sure RedPill does too. I haven't used the app except to remove yellow triangles, though, so I can't say I've tested it at all.

Also FWIW (not very much), another hour and a half without hitting the wakeup hang with TouchScreenTune off, but I haven't used my phone much.

antt00 · Feb 28, 2012

Entropy512 said:
Yeah. During suspend, the handlers are called in ascending order. During resume - descending order. So on I777, the earlysuspend handler is called later and the lateresume handler is called earlier.

Early suspend and late resume are screen off/on - deep sleep is regular suspend/resume.

As to CWM Manager compatibility - After poking at cfroot's initramfs, my head hurts. Do any other custom kernels support it?

Redpill and siyah support chainfire's CWM manager app

[Kernel][CWM]Fecality I9100 3/7/2012 (GPU Voltage Fix)

Retired Recognized Developer / Retired Forum Mod

Senior Member

Senior Member

Senior Recognized Developer

Senior Recognized Developer

Senior Member

Senior Recognized Developer

Senior Member

Senior Recognized Developer

Senior Member

Senior Recognized Developer

Senior Member

Senior Recognized Developer

Senior Member

Senior Member

Senior Recognized Developer

Senior Member

Senior Recognized Developer

Senior Member

Senior Member

Similar threads

Top Liked Posts