Closed GoogleCodeExporter closed 9 years ago
Now that is interesting. Give me some time to chew on it.
Original comment by showp1984
on 1 Oct 2012 at 11:40
You do undervolt quite heavily, this could have been a cause for this
exception. (it's not a bug, it's suppose to error out on bad pages)
Normal voltages:
800, 825, 850, 875, 900, 912, 975, 1000, 1025, 1050, 1075, 1100, 1125, 1150,
1175, 1200, 1212, 1237
Your voltages:
700, 725, 750, 775, 800, 812, 875, 900, 925, 950, 975, 1000, 1025, 1050,
1075, 1100, 1112, 1137
So that's -100mv on all frequencies.
Also you do overclock to 1.5Ghz. (that's +100Mhz on each core)
So you are pushing your cpu harder and taking power away. That just screams for
problems.
Considering that this bug occured right after your init.d script, or whatever
you use to undervolt, set the new values it is a safe bet to say that it is
indeed is responsible for this.
Undervolt set @ 64.643417.
All cpus online @ 82.761217 (now juice is required)
BUG_ON triggered @ 85.044471
You do see the coherence? :)
Try to get this bug without oc/undervolting and report back.
Original comment by showp1984
on 2 Oct 2012 at 1:23
Okay, that was interesting. I was using the ncx kernel before with -100mv:
http://forum.xda-developers.com/showthread.php?t=1873615
Then made the update to nightly 20121001 with included boot.img, i started up
everything and then changed to this kernel.
System tuner must have memorized the voltages.
About overclock. CM says 1500mhz default under performance, this is what i am
using.
The phone has been running about 14 hours since i created this bug. No problem
at all but i noticed that it is not undervolted anymore. It must have reset
itself somehow.
Original comment by Lindahl....@gmail.com
on 2 Oct 2012 at 2:22
1.5 Ghz is only stock if one core runs at this speed. The kernel would
downclock all cores to 1.4Ghz as long as more than one would be online (which
is nearly always the case).
So I removed this freq reduction. Bricked will always clock all cores @ max,
regardless of the number of cores online. This will effectively overclock your
CPU when more than one cpu is online. Worst case scenario: All 4 are online
(+100Mhz on all 4 cores).
If you have Variant 0 there comes an additional variable into undervolting. To
include 1600 Mhz for Variant 0 phones I had to move 1500Mhz one voltage step
back (so it is now -25mV undervolted). If you now apply -100 on all frequencies
you will undervolt 1500Mhz with -125mV.
Afaik ncx doesn't do any of this.
Original comment by showp1984
on 2 Oct 2012 at 2:30
I don´t know if this is rom related or kernel related. I made i call now but
the screen refused to light up when moving it from my cheek, s2w or volume
down. A reboot with vol + power was needed.
It started just fine but before i could tap my pincode, the screen goes black
again. It woke up when i put the usb plug in it, nothing else helped.
Hope you can see something in the files.
Original comment by Lindahl....@gmail.com
on 3 Oct 2012 at 9:17
Attachments:
No wonder, you seem to have very aggressive lowmem settings applied. Look at
the kmsg, there are only messages of killing processes in there, and that whole
kmsg is only 4 seconds long!
So your phone effectively kills any process as soon as it started.
Revert to default lowmem settings and everything should be fine.
Original comment by showp1984
on 4 Oct 2012 at 6:11
What could cause that? As far as i know, i have not changed anything. Has this
values in build.prop something to do with it?
dalvik.vm.heapstartsize=8m
dalvik.vm.heapgrowthlimit=64m
dalvik.vm.heapsize=256m
dalvik.vm.heaputilization=0.25
dalvik.vm.heapidealfree=8388608
dalvik.vm.heapconcurrentstart=2097152
Original comment by Lindahl....@gmail.com
on 4 Oct 2012 at 10:34
I have deleted system tuner pro now to see if this solves the problem.
Original comment by Lindahl....@gmail.com
on 4 Oct 2012 at 10:35
Original issue reported on code.google.com by
Lindahl....@gmail.com
on 1 Oct 2012 at 11:19Attachments: