Post Reply

PowerSaverKernel - Usb battery drain bugs solved [UPDATED 2011/08/07]

OP farmatito

27th March 2011, 10:25 PM   |  #1  
OP Senior Member
Thanks Meter: 125
 
544 posts
Join Date:Joined: Apr 2009
In this thread you will find a set a patches I've applied to pershoot's
kernel2635 source (https://github.com/pershoot/kernel-2635.git) in a
attempt to compile my own kernels and to fix the long standing and not cared of usb
battery drain bug (able to eat your battery in less than 2hrs).
A (outdated) list of the applied patches is:
  • 001-fix_msm_clock_warning.patch
  • 002_fix_msm_nand_warning.patch
  • 003-2.6.35.13-sched-bfs-404.patch this is by Con Kolivas http://ck.kolivas.org/patches/bfs/2.6.35/
  • 004-bfs_fix.patch
  • 006-msm_evlog_fix.patch
  • 007-hw3d_refdata.fix
  • 008-htc_pwrsink_refdata.patch
  • 009-board_trout_sapphire_fix.patch
  • 010-eventpoll_fix.patch
  • 011-dm-crypt_fix.patch
  • 012-mt9t013_fix.patch
  • 013-msm_smd_fix.patch
  • 014-pm_fix.patch
  • 016-mddi_refdata_fix.patch
  • 017-lowmemorykiller_fix.patch
  • 022-adb_2.6.29_forward_porting.patch
  • 023-htc_battery.patch
  • 024-msm72k_udc.patch
  • 025-patch-2.6.35.11-12
  • 026-BFQ-v2.patch
  • 027-BFQ-v2-r1.patch
  • 028-smd_rpcrouter.patch
  • 029-audpp_missed_DMA.patch
  • 030-msm_serial_hs.patch
  • 031-irq_pm.patch
  • 032-smd_tty.patch
  • 033-linux-2.6.38.y.git-524429c31b486c05449666b94613f59f729c0a84.patch
  • 034-10mb_32A.patch (MT1.2/Old 32a memory map)
  • 035-9aa258a27b6183e8e80241e4820883eaed23139f.patch (Magic Plus (32a) extra 10MB ram)
  • 036-kmalloc_b0937124b376946a5c93504f34ac932dffb23e3e.p atch
  • 037-2.6.36_lowmemorykiller+fudgeswap.patch
  • 038-minfree_stats-bb4a7baa64d9a56124e1f13510e9688aeed0cc10.patch
  • 039-msm72k_udc_fix_msm.git-1aa53ebd36088a60fcf4a498f763358598a477b9.patch
  • 040-synaptics_warning_fix.patch
  • 041-BFS-401-2388fb7aee855019c4a7560efb54cc7550a09743.patch
  • 042-BFS-rr_interval_335f482a0274a640ae6ab1a0036a3b263de575 b6.patch
  • 043-i2c_msm_a146e17dd81a2e4eaf67928c945038f521409384.p atch
  • 044-BFS-1000Hz.patch
  • 045-patch-2.6.35.12-13
  • 046-zram.patch
  • 047-vzalloc_for_zram_linux-2.6.38.y.git-e1ca7788dec6773b1a2bce51b7141948f2b8bccf.patch
  • 049-pmem_c21c461ce3e32d26abebe6243039824839748dc5.patc h
  • 050-gingerbread_gadget_update.patch
  • 051-adb_2.6.29_forward_porting.patch
  • 052-msm72k_udc.patch
  • 053_msm72k_udc_warning_fix.patch
  • 054-ae9bd4d574b7b8480aee162273d58d0c3840510a.patch
  • 055-ARM-zImage.patch
  • 056-wifi-3e6f7025178699b354e0e9fd94331f47ebb82763.patch
  • 057-arm-mm-cache-ecce0382ee54a765abcdc982e01a5fd30099f52d.patch
  • 058-autogroup patches
  • 059-patch-2.6.35.13-14
  • 060-cpufreq_stat_fix.patch
  • configs for 2708, ebi0 and ebi1
Please refer to the attached patches.zip files for the latest
patch series.

Main features of this kernel are:
  • BFS (in the BFS versions)
  • BFQv2r1
  • Overclocking
  • various fixes and minor enhancements
  • autogroup patches (in the autogroup versions)
BUG 1 (adb):
Patches 1-2, 4-17 are rather trivial stuff as they just silence some warnings.
Patch 22 is the one that seems to fix or reduce the "usb battery drain bug"
in the sense that after applying this patch I was not able to trigger the bug
again and I've tried several times the last days. The same result was reported
by a few courageous testers that risked to install one of my precompiled kernels
and whom I would like to thank for their collaborative effort.
The steps used to trigger the bug were:
  1. plug in usb
  2. start adb session
  3. remove usb plug
  4. check if a sticky android os entry is in the battery stats (a sure sign of incipient drain)
The patch reverts f_adb.c to be more similar to how it was in kernel 2.6.29-msm.
Similar code could also be found in HTC's 2.6.35 sources for the Incredible S
at http://dl4.htc.com/RomCode/Source_an...89aa373.tar.gz
so it could also be a alternative to use this code base for future G1 kernels.

Attached you'll find also the config I used to build the kernels posted
in the ezGingerbread thread and in Enomther's thread.

I hope this patch fixes this bug also for you as it did for me
making me a happy G1 user again.

I would thank to all who collaborated to this effort with facts
or with their wise words and advices.

BUG 2 (drain after reboot with usb connected):
After studying more and more bug N2 and thanks to the comments of the posters who sent me in the right direction
I've made a good progress. It is very early to call it a conclusive solution yet, but I was able to reboot my phone
with usb connected and after 3 hrs it is fully charged.
What did the trick was to disable:

CONFIG_SERIAL_MSM

in the kernel. I don't know what it was there for but at least
on my G1 I've tested all basic functions:

  • phone call
  • sms
  • data
  • wifi
  • BT
  • adb
  • music
  • GPS

and every thing seems to work. More test are needed tough,
there are also some minor leftovers to fix (more on this
in the next days).
For the bold a new kernel to test (tested on enomther's rom,
so I don't know if it will work on other roms).

You'll also find a VERY EXPERIMENTAL
anykernel/autokernel flashable zip archive with
EBI0/EBI1/2708+ kernels. Use it AT YOUR OWN RISK
and kindly report success or failure while flashing.
I cannot test it on all device/ROM combinations
so it may or may not work for you.
MAKE A BACKUP BEFORE USING IT so if something
goes wrong you can restore your previous rom.
I can give you very limited support due to missing
free time nonetheless patches are welcome!

Autokernel with BFS v0.401 and 1000Hz Kernel timer resolution for Froyo (discontinued):
auto-boot-2.6.35.13-farmatito-pershoot-ds-BFS+-20110428.zip
http://www.multiupload.com/8CKICMRK6B
MD5SUM 999770a413cbe93be75e00b1529f2300

Autokernel with BFS v0.404 for Gingerbread (experimental):
auto-boot-2.6.35.13-farmatito-pershoot-ds-BFS+-20110518.zip
http://www.multiupload.com/YRXWMK1AY3
MD5SUM c1373c915c8bfe3ebb470e839756a735

Standard Autokernel for Froyo (discontinued):

auto-boot-2.6.35.13-farmatito-pershoot-ds+-20110428.zip
http://www.multiupload.com/LLMKG4TRTO
MD5SUM ee3f5c53b4871076f123491c9dc388e2

Standard Autokernel for Gingerbread (stable):
auto-boot-2.6.35.13-farmatito-pershoot-ds+-20110505.zip
http://www.multiupload.com/0SJ77OM4HA
MD5SUM61a8bb913ca43a355e1eef3152b008db

Standard Autokernel for Gingerbread
with autogroup patches and memory compaction (experimental):
auto-boot-2.6.35.14-farmatito-pershoot-ds+-20110802.zip
http://www.multiupload.com/3NJ7Q0Y4TN
MD5SUM 27536ef42348dd92bb006cd8511df5d9

This kernels are unsupported they worked for me and I hope
they work for you, feel free to report problems and I will try to
fix them depending on my free time and skills.

USE AT YOUR OWN RISK. NO WARRANTY.

Enjoy


Changelog
v 20110425:
  • fixes to the installer automagic.
v 20110428:
  • BFS v0.401 + 1000Hz kernel timer resolution (BFS-version only) + rr_interval tuning.
  • i2c_msm HTC update
  • updated to 2.6.35.13
V 20110505:
  • revert kernel timer resolution to 100Hz
  • revert rr_interval tuning (BFS version)
  • update android gadget to gingerbread version.
V 20110518:
  • BFS v0.404 in the BFS versions
V 20110612
  • Compensate for rounding on odd-frequency clocksources
  • zImage fixes
  • Enable Memory Compaction
  • wifi - Don't make disabled irqs wake the device
  • cache-l2x0: Correct l2x0 initialization
  • autogroup patches
V 20110802
  • update to 2.6.35.14 (autogroup version only)
Last edited by farmatito; 8th August 2011 at 05:07 AM.
The Following 40 Users Say Thank You to farmatito For This Useful Post: [ View ]
28th March 2011, 03:48 AM   |  #2  
Senior Member
Thanks Meter: 16
 
200 posts
Join Date:Joined: May 2010
I'm glad to see someone working to fix the USB bug, I hope it will be squashed in pershoots official builds eventually as well
28th March 2011, 08:03 AM   |  #3  
Senior Member
Flag Johannesburg
Thanks Meter: 32
 
346 posts
Join Date:Joined: Apr 2007
Good work! Will try it out.

A while back everything on my screen froze (time and battery level not updated) but otherwise the phone worked 100%, however the battery lasted 2 days and 20 hours (68hrs with wifi on, 3G on, 614 overclock, browsing and gmail, etc). Weird
28th March 2011, 05:01 PM   |  #4  
dilwaladoctor's Avatar
Senior Member
Flag Midwest
Thanks Meter: 20
 
175 posts
Join Date:Joined: Nov 2010
More
Forgive me if my question feels like a noob but exactly how should we apply this zip file.....should I flash it?.... OR....it has to be installed through adb (or terminal emulator?)
I read the OP but couldnt make about the methos of installing it....(kinda stupid of me I know...
Is it OK to use your kernel instead of this patch?....do they have same effect?
Last edited by dilwaladoctor; 28th March 2011 at 05:06 PM.
28th March 2011, 06:12 PM   |  #5  
OP Senior Member
Thanks Meter: 125
 
544 posts
Join Date:Joined: Apr 2009
Quote:
Originally Posted by dilwaladoctor

Forgive me if my question feels like a noob but exactly how should we apply this zip file.....should I flash it?.... OR....it has to be installed through adb (or terminal emulator?)
I read the OP but couldnt make about the methos of installing it....(kinda stupid of me I know...
Is it OK to use your kernel instead of this patch?....do they have same effect?

The patches in the zip need to be applied to the kernel sources, built and then
a anykernel zip installer must be created. Then you can flash it.
You can try also a precompiled kernel if it is compatible with your rom.
So best is to take a nandroid backup in case things go bad.
I suggest you to wait that it is included by the devs of the rom you use
if they see it fit.
Last edited by farmatito; 28th March 2011 at 10:44 PM.
The Following User Says Thank You to farmatito For This Useful Post: [ View ]
28th March 2011, 11:05 PM   |  #6  
Senior Member
London
Thanks Meter: 108
 
1,058 posts
Join Date:Joined: Aug 2009
More
Well done for keeping at it
30th March 2011, 02:07 PM   |  #7  
ezterry's Avatar
Retired Recognized Developer
Flag Asheville, NC
Thanks Meter: 1,000
 
1,823 posts
Join Date:Joined: Jan 2010
Donate to Me
More
My testing is as follows

Using the ezGingerbread binary on a clean ezGingerbread install with gapps+facebook but no wifi or sim:

The USB connection in use is just a wire to my macbook.

0) battery validation: boot system from battery only and leave for extended time.. I was at 83% after 13h.. this indicates no particular extra drain thus this test passed

1) basic test: boot while phone charging on macbook, after system is fully loaded unplug from USB, replug, unplug again.. this case the phone was drained in under 13h thus indicates the USB drain is present test failed

2) adv. test: boot phone on batter such as test 0, after test 0 drained to say 79% charge with usb slightly, the unplug usb again and leave.. hours later indicated no significant drain test passed.

3) adv. test 2: now at 60% after test2 plug into usb reboot and unplug after fully loaded.. such as in test1 drain quickly continued test failed.

Thus we seem to have a form of USB drain that is not in the 2.6.34.* kernels, however it may be possible on this kernel to avoid it by booting the phone on its own power (I did not test booting the phone on the A/C adapter, least not yet and more testing is needed to be sure there is no other oddities)
Last edited by ezterry; 30th March 2011 at 03:15 PM.
The Following User Says Thank You to ezterry For This Useful Post: [ View ]
30th March 2011, 03:44 PM   |  #8  
OP Senior Member
Thanks Meter: 125
 
544 posts
Join Date:Joined: Apr 2009
@ezterry :

so if understand it correctly it is booting/rebooting with
usb plugged in that still fails?

Do you see the same stuck "android os" entry in battery stats?
If not, than maybe it is a different bug, so more patching is needed.

UPDATE:

test:

plug in usb
reboot
wait a little while phone is charging
unplug usb
go to battery stats:
there you'll see "android os" entry
hit a few times refresh
"android os" %value decreases and after a few more refresh hits it disappears.

I think that if there is battery drain in this case it is not the same bug.
Will check the phone later to see how it behaves:
  • 16.59 84%
  • 18.18 77%
  • 20.42 68%
16 % / 4hrs = 4% /1hrs (in idle state)

Looks not so bad to me... but definitely would not last
48+ hours as my battery usually lasts.
In my opinion this is not the same bug tough, because
what I and other users were experiencing was a very
fast battery drain after disconnecting from a adb
session. The latter bug is indeed fixed.
Last edited by farmatito; 31st March 2011 at 08:55 PM.
1st April 2011, 06:56 AM   |  #9  
OP Senior Member
Thanks Meter: 125
 
544 posts
Join Date:Joined: Apr 2009
After more testing with this bug N 2 triggered the power
consumption with my extended battery is stable at about
4%/hr. I've looked at the most likely candidates

htc_battery.c
msm72k_udc.c

but I've not spotted differences between 2.6.34.8 and 2.6.35.11
that could explain the different behaviour.
Hints and help is welcome.
1st April 2011, 08:58 PM   |  #10  
Member
Thanks Meter: 0
 
60 posts
Join Date:Joined: Apr 2009
Can some one explain that USB drain problem?

I already had change my battery but it hasn't make any difference...
My G1 can only work for 4h and them the battery is over.

That patch can fix it?
How do I flash it?

Post Reply Subscribe to Thread

Tags
usb battery drain bug
Previous Thread Next Thread
Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes