[Kernel][GPL] msm_hsic_host wakelock fix (Now with WiFi notification fix!)

chrone · Feb 15, 2013

quick impression on one day mild usage of n4 4.2.2 battery life from 100% to 1% (6.6%/hr) on stock kernel:

battery life 14.9 hours
screen time 2.5 hours
3G most of the time, sometimes WiFi and 2G
autosync on
msm_hsic_host 1256 wakelocks, 10% ~ 2 hours wakelock

update:
uploaded bbs log and some screenshots.

thracemerin · Feb 15, 2013

elmalote said:
Thracemerin, does your experimental build from the 14th contain color adjustment, like your other kernels?

Not at the moment, it's just Google sources in v1, I will be looking at adding other stuff, I just wanted to make sure the new sources were solid before installing anything into them that may cause unforseen issues.

floepie said:
Just a quick question, boss. Isn't the stock kernel simply AOSP + proprietary drivers? So, what would be the difference? Or, is it simply a matter of who compiled the source, whether it be Google (stock OTA) or by the tools floating about in dev community land?

Yeah, the stock AOSP stock kernel and the OTA kernel should be the same other than who compiled it, since I'm using the Google toolchain as well there should be no difference. Digging around in the new sources I noticed that Google didn't push all the changes I made in my previous 4.2.1 versions, so I wanted to see if there was some other change that they made to fix the wake locks before I went ahead and added in the patches again.

thracemerin · Feb 15, 2013

Ok, New version:

Back in Post 2

I re-added some of the HSIC patches from the 4.2.1 version and this may or may not have re-introduced the 3G data drop/lock up issue, I want to see if it did or not before I continue.

I would like people to run the version from the 14th first (for a full day and post the logs I asked for before) to get a baseline on the wake lock before trying the 15th version.

If you get a data drop, lock up or any other unusual behavior on the 15th version, please let me know. Logcats, last_kmsg, etc would be helpful as well if you can.

sotong · Feb 16, 2013

Hi,

I have not been able to test V1 on wakelocks as my usage pattern the past day was for heavy gaming. However, i do notice that the wi-fi data dropouts i used to have (delayed notifications) have gone. I still noticed it using the stock 4.2.2 kernel, so i find this very odd. I will try re-flashing the stock kernel and giving it another go. I noticed this delayed notification also with other custom kernels based on 4.2.2.

thracemerin · Feb 16, 2013

sotong said:
Hi,

I have not been able to test V1 on wakelocks as my usage pattern the past day was for heavy gaming. However, i do notice that the wi-fi data dropouts i used to have (delayed notifications) have gone. I still noticed it using the stock 4.2.2 kernel, so i find this very odd. I will try re-flashing the stock kernel and giving it another go. I noticed this delayed notification also with other custom kernels based on 4.2.2.

So the delayed notification is gone? That's a good sign. I only ever had it intermittently on the 4.2.1 kernel, so it may just be random good luck that you don't have it anymore and it may pop up again, I don't see anywhere that Google fixed this, but it could have been hidden in something else.

sotong · Feb 16, 2013

thracemerin said:
So the delayed notification is gone? That's a good sign. I only ever had it intermittently on the 4.2.1 kernel, so it may just be random good luck that you don't have it anymore and it may pop up again, I don't see anywhere that Google fixed this, but it could have been hidden in something else.

That is correct. I just did further testing again:

1. Flash back to stock 4.2.2 kernel. Wait 5 mins. Sent gmail to my phone. Notification did not come in until i turn the screen on.
2. Flash to your V1 kernel. Wait 5 mins. Sent gmail again, and i got the notification almost immediately.

I am also puzzled as to why this is happening. But i am not complaining! Will test your V2 now. Thanks for your efforts

thracemerin · Feb 16, 2013

sotong said:
That is correct. I just did further testing again:

1. Flash back to stock 4.2.2 kernel. Wait 5 mins. Sent gmail to my phone. Notification did not come in until i turn the screen on.
2. Flash to your V1 kernel. Wait 5 mins. Sent gmail again, and i got the notification almost immediately.

I am also puzzled as to why this is happening. But i am not complaining! Will test your V2 now. Thanks for your efforts

Interesting...as far as I know V1 should be exactly identical to stock, so I can't explain why that would be, but I guess it's a happy side effect.

sotong · Feb 16, 2013

thracemerin said:
Interesting...as far as I know V1 should be exactly identical to stock, so I can't explain why that would be, but I guess it's a happy side effect.

Hi, just a quick update again. V2 seems to also have delayed notifications. Went back to V1, same test, and no delay.

thracemerin · Feb 16, 2013

sotong said:
Hi, just a quick update again. V2 seems to also have delayed notifications. Went back to V1, same test, and no delay.

I can't explain that, V1 and Stock are identical, V2 has nothing related to WiFi changed. So unless it's random coincidence that you aren't having problems with V1, I'm not sure what else it could be, I'll look into it further.

Edit: I tried hard to reproduce this on my local version with is essentially V2 and I can't, so either it's something I added in V2 and changed back in the local version (not likely) or it's something to do with your setup.

sotong · Feb 16, 2013

thracemerin said:
I can't explain that, V1 and Stock are identical, V2 has nothing related to WiFi changed. So unless it's random coincidence that you aren't having problems with V1, I'm not sure what else it could be, I'll look into it further.

Edit: I tried hard to reproduce this on my local version with is essentially V2 and I can't, so either it's something I added in V2 and changed back in the local version (not likely) or it's something to do with your setup.

I agree with you, it must be my set up because it does not make sense also. Could be a random coincidence. Is there a quicker way to test this (i read about pinging etc) instead of waiting several minutes when the phone is in standby?

Edit: You are correct. Just sent a gmail to my phone (on V1) and nothing came in

There goes my short-lived joy! Or gmail is having issues also.

thracemerin · Feb 16, 2013

sotong said:
I agree with you, it must be my set up because it does not make sense also. Could be a random coincidence. Is there a quicker way to test this (i read about pinging etc) instead of waiting several minutes when the phone is in standby?

Edit: You are correct. Just sent a gmail to my phone (on V1) and nothing came in There goes my short-lived joy! Or gmail is having issues also.

I guess that they more or less confirmed in this thread: http://xdaforums.com/showthread.php?t=2072930 that Google did not fix the WiFi notification issue, so I don't think it's you, blame Google.

thracemerin · Feb 16, 2013

New build in post 2 (see there for changelog)

I added a couple more CAF patches and used the data fix that Faux123 used in his enhanced stock kernel instead of the one I used previously, I want to see what happens with this as it relates to the wakelock and data use to see how it goes, if it doesn't work I'll install my previous solution.

Happy Flashing.

chrone · Feb 17, 2013

thracemerin said:
New build in post 2 (see there for changelog)

I added a couple more CAF patches and used the data fix that Faux123 used in his enhanced stock kernel instead of the one I used previously, I want to see what happens with this as it relates to the wakelock and data use to see how it goes, if it doesn't work I'll install my previous solution.

Happy Flashing.

i'm testing your v3 right now, will post the log tomorrow now after full day use.
no need to test v1 since it's the same as stock and i've posted the bbs log above, right?

thracemerin · Feb 17, 2013

chrone said:
i'm testing your v3 right now, will post the log tomorrow now after full day use.
no need to test v1 since it's the same as stock and i've posted the bbs log above, right?

Yeah v1 is the same as stock.

floepie · Feb 17, 2013

sotong said:
That is correct. I just did further testing again:

1. Flash back to stock 4.2.2 kernel. Wait 5 mins. Sent gmail to my phone. Notification did not come in until i turn the screen on.
2. Flash to your V1 kernel. Wait 5 mins. Sent gmail again, and i got the notification almost immediately.

I am also puzzled as to why this is happening. But i am not complaining! Will test your V2 now. Thanks for your efforts

Sending yourself a gmail turns out to be a very poor test. Invariably, if I send myself an email, notification delays are very random, and this has been confirmed by others on this board. However, getting emails from 3rd parties is much more reliable for delay testing. It may have something to do with "loopback" settings on the router. In any case, if you have a router which flushes its ARP cache every couple minutes, you will get delays (up to 15 minutes) because Google hasn't updated the wifi driver to offload ARP broadcast requests. They decided battery life was more important, so they chose to drop requests entirely when the screen is off.

Fortunately, if you have a dd-wrt router, you can set a static IP address from the router to get around this issue.

thracemerin · Feb 17, 2013

floepie said:
Sending yourself a gmail turns out to be a very poor test. Invariably, if I send myself an email, notification delays are very random, and this has been confirmed by others on this board. However, getting emails from 3rd parties is much more reliable for delay testing. It may have something to do with "loopback" settings on the router.

True, but the WiFi issue is definitely not fixed, I tested it using the ping method. If I used a PC that hadn't previously connected to the phone while the screen was off, the pings timed out, as soon as I turned the screen on they'd work until the ARP timed out, then it would go back to time out.

jakejm79 · Feb 17, 2013

Theracemerin, having upgraded to 4.2.2 (CM10.1 2/16) and using you v3 fix, my wakelock performance isn't as good (about 30 mins over 10 hours of sleep with wifi on) when compared to the older v2 (for 4.2.1 built 2/4) version I was using before. Hopefully you can add the rest of your fixes you had for the 4.2.1 kernel to the 4.2.2 kernel.

thracemerin · Feb 17, 2013

jakejm79 said:
Theracemerin, having upgraded to 4.2.2 (CM10.1 2/16) and using you v3 fix, my wakelock performance isn't as good (about 30 mins over 10 hours of sleep with wifi on) when compared to the older v2 (for 4.2.1 built 2/4) version I was using before. Hopefully you can add the rest of your fixes you had for the 4.2.1 kernel to the 4.2.2 kernel.

You're absolutely right, it's not. Faux123s solution to the data problem was to decrease the USB bus suspend time by 20% from the original, my solution from 4.2.1 was to use the CAF patch that set it to 0 as soon as the data stopped, this caused a problem if the data stream stopped momentarily it would get permanently stuck. I added a few additional patches to make sure that this wouldn't happen if it were just a momentary stop in the stream, so Faux123s solution reduces the wakelock time, but not to the same degree as mine, his patch is safer from a data drop standpoint though.

Edit: The new HEAD in my git has all my old patches from 4.2.1, I'm running a few tests on it to make sure we're good to go before pushing a test release.

jakejm79 · Feb 17, 2013

OK, that makes sense, I never really suffered from the data drop issue, maybe we could get a little more aggressive than just 20% (assuming you are saying it was reduced by 20% i.e. from 10 to 8 and not reduced to 20% i.e. from 10 to 2)

EDIT: OK sounds good, I will sit patiently.

thracemerin · Feb 17, 2013

jakejm79 said:
OK, that makes sense, I never really suffered from the data drop issue, maybe we could get a little more aggressive than just 20% (assuming you are saying it was reduced by 20% i.e. from 10 to 8 and not reduced to 20% i.e. from 10 to 2)

Yeah, 20% from original (ie 200->160).

See here: https://github.com/thracemerin/Mako/commit/e5b918aa41059e05ee3851fd117d20dd761e8526

I know it looks like I raised it from 0 to 160, but there was a previous patch that set it to 0 from 200 here: https://github.com/thracemerin/Mako/commit/56d96a8b55adc6ed005f35f06ef705ea8c812c3c

[Kernel][GPL] msm_hsic_host wakelock fix (Now with WiFi notification fix!)

Senior Member

Attachments

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Senior Member

Similar threads

Top Liked Posts