[ROM] [LOLLIPOP] [LMY48G] [Synman's GwT ROM for Tilapia] [carcosa] [2015/07/03]

Xenon.Kay

Member
Feb 9, 2014
7
1
0
Is there a simple way to switch between the AOSP and carcosa kernels built for GwT? The 3G radio on my N7 is exhibiting some odd behavior on the carcosa version and I'd like to be able to test whether AOSP does the same thing under the same conditions.

One more battery life tip. I did have some battery drain and slow charging associated with Google Play Services. The AnalyticsService was running 100% of the time; I disabled it with the DisableService app (use with caution), and my battery life and slow charging improved markedly.
I've done this on my N7 but AnalyticsService just restarts itself after a while, despite having been disabled by Disable Service. Seems like a more aggressive approach involving Xposed might be necessary to make the service stay down.
 

fishwithlungs

Member
Dec 10, 2013
22
6
0
Sound quality from built in speakers

@Synman - great work,the ROM is really smooth and fast.
One thing I've noticed,the built in speakers do have a problem,the buzzing and distortion makes it impossible to turn the volume more than halfway up,headphones are working fine.
No such problem on 4.4.4.
Am I the only one with this problem,or is there a fix for built in audio?
 
Last edited:

jfmcbrayer

Senior Member
May 9, 2011
211
110
0
Is there a simple way to switch between the AOSP and carcosa kernels built for GwT? The 3G radio on my N7 is exhibiting some odd behavior on the carcosa version and I'd like to be able to test whether AOSP does the same thing under the same conditions.
I'd install the AOSP version of the ROM, then back up your not partition, and install carcosa kernel from the zip file on the thread in my sig. Alternatively, to install the AOSP kernel over the carcosa kernel, extract the boot.img from the ROM zip and flash it with Flashify.

I've done this on my N7 but AnalyticsService just restarts itself after a while, despite having been disabled by Disable Service. Seems like a more aggressive approach involving Xposed might be necessary to make the service stay down.
I've noticed this, too.
 
  • Like
Reactions: Xenon.Kay

Xenon.Kay

Member
Feb 9, 2014
7
1
0
Thanks for that. I'll test for a bit and see if the behavior I've experienced persists under different kernels. I'll document what I've been seeing once I'm sure I know myself.
 
  • Like
Reactions: loco72

jfmcbrayer

Senior Member
May 9, 2011
211
110
0
Thanks for that. I'll test for a bit and see if the behavior I've experienced persists under different kernels. I'll document what I've been seeing once I'm sure I know myself.
I'll be interested in seeing your results. If there is a difference, I'll have another test for you.
 

Xenon.Kay

Member
Feb 9, 2014
7
1
0
Okay, I'm putting the AOSP kernel through the same routine and with only this change there are already some noticeable differences.

Both kernels:
The tablet is connected to wifi in a location overnight where it doesn't get cell reception (a good 8 hours or so typically). When later taken away from the wifi signal and into an area with reception, it does not roll over to mobile data gracefully. The cell radio indicator shows a good signal but with the exclamation mark indicating that mobile data is not available. Turning airplane mode on for a few seconds and then turning it off again to cycle the radio while already within cell reception causes it to correct itself.

Carcosa only:
After the tablet has been asleep for a while (maybe half-hour to hour range), it loses mobile data and does not recover after wakeup even though it's under good reception the whole time. The radio icon indicates a good signal but mobile data is not available, similar to the above behavior. It will repeat this all day long, every time it sleeps long enough. Cycling airplane mode usually allows it to recover, but not always. Which leads to the strange bit:

Sometimes when the mobile data is down and airplane mode is switched on, the radio instead stays on and will not turn off, as though it were stuck. It would have airplane mode on yet the radio would stay up and indicate a signal. It would even still visibly switch between HSPA and 3G according to tower availability. Once this behavior occurs it persists indefinitely until the tablet is rebooted.
 

jfmcbrayer

Senior Member
May 9, 2011
211
110
0
Good, thanks!

Now try this. On carcosa kernel, do the following as root:

echo 20 > /sys/module/baseband_xmm_power/parameters/bb_timeout

See if this affects the connection loss you're seeing. This will re-introduce the baseband_xmm_power wakelock, but with a 2 second timeout.
 

Xenon.Kay

Member
Feb 9, 2014
7
1
0
Cool, I've made a note of that and added it to-do list.

I still need to test the AOSP kernel a bit more though because I'm not convinced it's unaffected by the stuck radio behavior. It could be it just hasn't cropped up yet since it only drops the mobile data connection once a day. And presumably the reason for that is the wakelock keeping the radio up all day while it has reception.

edit:
I haven't been able to duplicate the radio sticking in airplane mode under AOSP so far and this test may take quite a while, so I've gone ahead and switched back to the carcosa kernel with the suggested timeout applied and will be monitoring its behavior.

update:
OK, with the carcosa kernel and bb_timeout set to 20 the internet connection failed to reestablish the first time the tablet was woken up after screen off for about one hour. The radio indicator shows a good 3G/HSPA signal but there's also the exclamation point and a lack of internet connectivity. Airplane mode on and off to cycle the radio restores internet connectivity. I'll be testing this some more to see if the frequency of this behavior has changed with the new timeout parameter. The radio hasn't gotten stuck "on" yet but the sample size for this test is still a bit small.
 
Last edited:

Xenon.Kay

Member
Feb 9, 2014
7
1
0
After testing carcosa with the suggested bb_timeout parameter for a couple more days it seems like it happens less often than without that change. Maybe 3-4 times per day now, as opposed to upwards of 10 before enabling the 2sec wakelock timeout.

This is still more often than on stock though, which would exhibit the behavior at most once per day after being asleep on the charger for approx. eight hours.

The tablet also did something different this morning: It crashed and rebooted after being taken off the charger. The screen came on as normal when the charger was disconnected, but a few seconds later it suddenly went blank. A few seconds after that it displayed the Google logo and started booting back up. This was with a full battery and still on the carcosa kernel with bb_timeout set to 20.

The only apps on the tablet consist of what comes with the GwT ROM, CM12 GAPPS package, SuperSU, K-9 Mail, Netflix, Terminal Emulator, VX ConnectBot, Hacker's Keyboard, DisableService and Flashify.
 

jfmcbrayer

Senior Member
May 9, 2011
211
110
0
You can keep tweaking the wakelock timeout, of course. Use any value you want, in tenths of a second. The trade-off is, of course, that eventually the battery usage due to the wakelock will also be the same as stock.

The crash sounds like it was just random, but if you had a last_kmsg, I could look at it.
 

Xenon.Kay

Member
Feb 9, 2014
7
1
0
The timeout does seem to have some kind of effect, so I'll keep playing with it. There's a long way to go before I reach the stock timeout.

I didn't think to check /proc/last_kmsg at the time, but I'll keep it in mind for next time. I think that was probably the first time I've ever seen this tablet crash during normal operation though, so I'd be surprised if it happens again.
 

spideytan

Member
Oct 26, 2007
15
1
0
With a stock rom, I think you will have to use an app: StickMount by ChainFire works for me. Honestly, I have not used it on Lollipop, yet.
Have installed StickMount and problem persists. Hope this problem will be resolved soon, and I guess its really Lollipop issues with OTG. Cos this problem never exists before in official KitKat and Omni Roms running 4.4.4
 

paolocorpo

Senior Member
Jan 9, 2013
1,783
378
113
Sassoferrato
  • Like
Reactions: bix66

jfmcbrayer

Senior Member
May 9, 2011
211
110
0
@jfmcbrayer : Have you already shared tour last updated kernel ?
I have not, but I will when I get some time. The updates are not actually that significant, IMO. They add support for hardware acceleration of in-kernel encryption routines, which might be helpful to people with device encryption. On the other hand, the stock kernel then disables that support, so maybe it's not fully baked.

EDIT: It's now updated.
 
Last edited:
  • Like
Reactions: bix66