maemo.org - Talk - [Announce] kernel-power stable v49 in Extras-Testing

@maacruz - lol. But TBH it really does something, as soon after I tried it, I was unable to pick up a phone call because phone-ui was unresponsive for the first time since PR 1.0. AIUI it removes all cgroup assignments made by Nokia, thus giving fair chance to everyone to lock the UI :D :D :D. I am not much in cgroups so I may be wrong.

While really offtopic here, I think you should re-post your findings on "magical patch" thread. I don't think we should tolerate stupidity and ignorance here (on TMO).

freemangordon

2011-11-17 18:28

Re: [Announce] kernel-power unstable v49

Quote:

Originally Posted by vi_ (Post 1125065)

Take the battery out for 30 seconds. Then have a look.

I have been away for a while am well horny about the inclusion of compcache&hd recording. Is there a wiki entry that details the instructions required to get all this cool **** up to date?

OR

Should I just wait for KP49 to hit the repos and take it form there?

KP hopefully should be in devel in a matter of days, Harmattan DSP diver and compcache included. For HD you still have to follow the instructions on "720p" thread. Unfortunately seems from today Nokia has withdrawn firmware image that was used from their server, so the script should be updated for the new firmware.

maacruz

2011-11-17 19:49

Re: [Announce] kernel-power unstable v49

Quote:

Originally Posted by freemangordon (Post 1125114)

LOL, YOU ARE ACTUALLY RIGHT!! :D :D :D :D
I hadn't realized it, but actually nokia has setup cgroups in the N900, just it doesn't show in the usual places (/dev/cgroups or /etc/mtab)
But...

Code:

# cat /proc/mounts|grep cpu

cgroup /syspart cgroup rw,freezer,memory,cpu 0 0

It actually does something!!!
Well, OK, not what you wanted though, it completely screws the careful setup done by nokia, which actually has Desktop and Application cgroup categories :D :D :D
Oh my!
http://fc09.deviantart.net/fs42/f/20...andForLife.jpg

carbonjha

2011-11-17 20:19

Re: [Announce] kernel-power unstable v49

@pali

I found a script in http://atrey.karlin.mff.cuni.cz/~pal...ock/underclock

Code:

#!/bin/sh

if [ "`dbus-send --system --print-reply --dest=com.nokia.csd.Call /com/nokia/csd/call/1 com.nokia.csd.Call.Instance.GetStatus | tail -1 | sed 's/.* //'`" != "0" ]; then exit 0; fi

echo > /sys/devices/system/cpu/cpu0/cpufreq/ondemand/avoid_frequencies

what this script do?? would you please explain??

karam

2011-11-18 00:02

Re: [Announce] kernel-power unstable v49

@marcuuz

http://imageshack.us/photo/my-images...1111801493.png

http://imageshack.us/photo/my-images...1111801500.png

perhaps you can check the postinst script before
then the *funny* things will make sense then

and .profile launches .bashrc which will make hildon-desktop get cgrouped

mooninite

2011-11-18 02:58

Re: [Announce] kernel-power unstable v49

The latest kernel (freemangordon's build) seems to be more stable. I've been running it for 5 days now with only one reboot. Prior to this the KP kernel only lasted 24 to 48 hours before it would reboot.

I'm using the default settings file, so 250 to 805mhz with SR on.

Thanks.

Estel

2011-11-18 03:48

Re: [Announce] kernel-power unstable v49

Lol, that's improvement ;)

Just for recording sake - reboots out of blue sky aren't sign of anything good. You should not experience this. I haven't had random reboot since ancient times, when dinosaurs were hunting Australopithecus using speedpatches. So, basically, not in last half of a year. You should check what is causing this.

/Estel

// Edit

I'm getting strange problem with latest kp49, or I've missed something. I can't seem to blacklist bq27x00_batter.ko - no matter what I try. Of course, in my /etc/modprobe.d/blacklist (and in /etc/modprobe.d/kernel-power.blacklist) I got:

Code:

blacklist bq27x00_battery

I'm sure it doesn't contain any typos. "bq27x00_battery" is my second blacklisted module on the list, first one is "mousedev". I haven't had such problems in kp48.

Currently, I workaround'ed this by putting:

Code:

modprobe -r bq27x00_battery

...into my startup script in
/etc/event.d/
, but it is suboptimal solution.

Any hints?

freemangordon

2011-11-18 07:06

Re: [Announce] kernel-power unstable v49

Quote:

Originally Posted by mooninite (Post 1125303)

Default profile has limits 250-600. And you should really check what is the reason for reboot, it could be a problem with KP or with something else on your system.

pali	2011-11-18 07:10

Re: [Announce] kernel-power unstable v49

Quote:

Originally Posted by carbonjha (Post 1125167)

@pali

I found a script in http://atrey.karlin.mff.cuni.cz/~pal...ock/underclock

Code:

#!/bin/sh

if [ "`dbus-send --system --print-reply --dest=com.nokia.csd.Call /com/nokia/csd/call/1 com.nokia.csd.Call.Instance.GetStatus | tail -1 | sed 's/.* //'`" != "0" ]; then exit 0; fi

echo > /sys/devices/system/cpu/cpu0/cpufreq/ondemand/avoid_frequencies

what this script do?? would you please explain??

full script is:

Code:

#!/bin/sh

if [ "`dbus-send --system --print-reply --dest=com.nokia.csd.Call /com/nokia/csd/call/1 com.nokia.csd.Call.Instance.GetStatus | tail -1 | sed 's/.* //'`" != "0" ]; then exit 0; fi

echo > /sys/devices/system/cpu/cpu0/cpufreq/ondemand/avoid_frequencies

renice 1 `cat /syspart/applications/standby/background/tasks`

renice 1 `pgrep modest`

/usr/sbin/kernel-config limits 125 500

that script set freq. limit from 125 to 500, renice background tasks and modest only when there is no active voice/video call.

pali	2011-11-18 07:13

Re: [Announce] kernel-power unstable v49

Quote:

Originally Posted by Estel (Post 1125306)

Code:

blacklist bq27x00_battery

Code:

modprobe -r bq27x00_battery

...into my startup script in
/etc/event.d/
, but it is suboptimal solution.

Any hints?

Check status of all files in /etc/modprobe.d/ Also check file /etc/modules (all modules in this fill will be loaded, ignoring blacklist). You can also check dmesg output and find when was bq module loaded. Try also finding bq27x00_battery string in all files in /etc/: grep "bq27x00_battery" -R /etc/

impeham

2011-11-18 08:27

Re: [Announce] kernel-power unstable v49

After installing KP49, I switch the device off and then on and then immediately get this message:
Can't flash kernel, required files not found
Guru meditation #00000025.62017712
Hit any key to panic

And then I have to reboot to backupmenu and restore to the previous state (with kp48)

i installed kp49 with (all files are with version of 16-11):
dpkg -i kernel-power_2.6.28-10power49_armel.deb
dpkg -i kernel-power-modules_2.6.28-10power49_armel.deb
dpkg -i kernel-power-flasher_2.6.28-10power49_armel.deb
and rebooted.

pali	2011-11-18 08:54

Re: [Announce] kernel-power unstable v49

@impeham:
if you want multiboot support install -bootimg package.

freemangordon

2011-11-18 09:07

Re: [Announce] kernel-power unstable v49

@pali - KP seems stable enough to be put in devel repo, what do you think?

pali	2011-11-18 09:17

Re: [Announce] kernel-power unstable v49

Ok, I look at maemo package interface bug and try to push dsc package which has package interface now...

disappear

2011-11-18 12:32

Re: [Announce] kernel-power unstable v49

Hi today i dicided to install kp49,but i have a problem after type:
/usr/sbin/kernel-config show
the result was:
/usr/sbin/kernel-config: shift: line 352: can't shift that many

any ideas
note:i have installed: power,modules,headers,bootimg and flasher

pali	2011-11-18 12:34

Re: [Announce] kernel-power unstable v49

you forgot install new -settings package

disappear

2011-11-18 12:42

Re: [Announce] kernel-power unstable v49

Quote:

Originally Posted by pali
you forgot install new -settings package

Thanks for fast answer

Estel

2011-11-18 17:25

Re: [Announce] kernel-power unstable v49

Quote:

Originally Posted by pali (Post 1125355)

Bingo! It as entry in /etc/modules. I have no idea, how it appeared there - I definitely haven't put it inside, and I haven't had problems with kp48 or kernel-bfs. Anyway, it's working fine now.

---

I've tested latest kp49 with 500-900 mhz and sr_vdd1 *on*, and it's rock stable. I've stressed it with situations requiring many fast changes of used frequency, and by ones resulting in max load @900, + high power draw from battery too (700-800 mA total). Everything worked fine, without any glitches.

Now, few questions:

1. Should we also use vdd2, or is it not properly calculated yet (I remember freemangordon's post, about possible need for recalculating it too)?

2. What about situations, where CPU usage isn't high, but DSP is? Particularly, I mean 3D programs, using OpenGLES. For example, while I play Neverball, CPU is around 70% load @ 500 mhz, but - AIUI - DSP is calculating graphic, so any "slowness" situations, are caused by under-clocked (@ 500 mhz CPU) DSP.

In such situations, could we benefit from overclocking DSP to high number @ 500 mhz CPU (then, after finished with such programs, revert it back to regular profile), or I missunderstood something?

Thanks for this awesome work.

/Estel

// Edit

It seems, that vdd2 calculates 44 as value for every frequency. Doesn't sound correct, yes? It seems to stay in line with what freemangordon said about original mechanics - value calculated to work with 600, so too high for 125/250, and too low for 900.

On the other hand, vdd1 calculates all values perfectly - from 20 @ 125, to 55 @ 900.

freemangordon

2011-11-18 18:46

Re: [Announce] kernel-power unstable v49

@Estel - while calculated values for vdd2 are not optimal, it is still better to set SR for vdd2 on too, voltages will be lower than without SR.

Re DSP - DSP driver clocks the chip according to current load and maximum allowed frequency.

The logic is (from DSP side):

What my is current load? If it is above high threshold I need to call the diver to increase my frequency otherwise if it is bellow low threshold I need to call the driver to lower my frequency.

And from driver side(driven by event):
DSP wants a frequency change. OK could we go any further(up or down). If no, do nothing, otherwise check which is next DSP frequency(setup in profile file). Now, try to find valid DSP frequency which is equal or above that. Setup MPU to OPP for that frequency. Notify DSP for frequency(OPP) change.

So, in short, DSP driver rises and lowers DSP(and MPU accordongly) frequency when needed, you don't need to play with them. And when DSP is on MPU frequency never goes bellow 500.

Valid DSP frequencies are:
90,180,360,400,430, (450?), 480,500,520,540,560,580,600

Not 100% sure for the above and too lazy now to open dspbridge.c but close enough.

ammyt

2011-11-18 19:22

Re: [Announce] kernel-power unstable v49

Quote:

Originally Posted by freemangordon (Post 1125630)

...So, in short, DSP driver rises and lowers DSP(and MPU accordongly) frequency when needed, you don't need to play with them. And when DSP is on MPU frequency never goes bellow 500.
...

Doesn't that mean that I should set the dsp @600 in every frequency step starting from 500 in order to record 720p?

Estel

2011-11-18 20:55

Re: [Announce] kernel-power unstable v49

Quote:

Originally Posted by freemangordon (Post 1125630)

@Estel - while calculated values for vdd2 are not optimal, it is still better to set SR for vdd2 on too, voltages will be lower than without SR.

That what I'm afraid of ;) I'm worried that, for upper limit 900 mhz (CPU) vdd2 may result in too low frequency... Sorry for bugging You with questions, but I'm no expert in regards of SR, so I understand it as working similar to how vdd1 was miss-calculated - too low for stable 850/900, too-high for power-saving on 125/250. But it seems, that I'm wrong here?

As for ammyt question, know You're fighting with fire ideas of putting DSP limit to 600 on low-frequency, but for non-experts, he seems to have rationale in his logic. If DSP (via driver) adjust frequency, basing decision on load, what is wrong with allowing it to go 'high' ( ;) ), when needed?

Of course I'm not advocating doing so - from Your posts it seems a bad practice, I'm just asking about rationale.

Thanks in advance for explaining it even further.

/Estel

pali	2011-11-18 21:57

Re: [Announce] kernel-power unstable v49

I pushed kernel-power v49 and kernel-power-settings v13 into extras-devel. Wait until autobuilder compile both packages... and maybe maemo package interface will show it. I pushed original kernel tarball from SDK repository

woody14619

2011-11-18 23:07

Re: [Announce] kernel-power unstable v49

Quote:

Originally Posted by strange1712 (Post 1124700)

Desktop was unresponsive and laggy, checked dmesg and it was full of

Code:

HWRecoveryResetSGX: SGX Hardware Recovery triggered

Every interaction triggered one more line...
One more reboot fixed it...
I don't know what was that, it had never happened to my N900 before...

I've had that happen since PR1.0, with stock kernel. There's a small bug in the SGX driver in the kernel that in just the right scenario, will cause it to spin. In PR1.0 and stock kernel it can be triggered by having a couple of apps constantly updating their display, even when in the background. (Modrana and a couple widgets used to do this, which is how I found out about it.)

There are patches to the SGX driver upstream, but they haven't been back-ported to this kernel. I know Titan was looking into doing so, and some of it may have been ported back since you still had enough interface to reset the device. In stock and Titan, when sgx driver went south the display froze and you were done... if you were lucky a long power press would shut you down.

strange1712

2011-11-19 01:37

Re: [Announce] kernel-power unstable v49

Yes, indeed it was "painly usable", but usable, so I could read that output from dmesg. Thanks for getting light on this...

stlpaul

2011-11-19 04:26

Re: [Announce] kernel-power unstable v49

Quote:

Originally Posted by pali (Post 1125706)

Thanks! I am using it now. So far, so good. :)

Minhaz

2011-11-19 04:36

Re: [Announce] kernel-power unstable v49

WoW great news :D

knobby

2011-11-19 05:32

Re: [Announce] kernel-power unstable v49

now thats what i call perfection...

thanks a lot freemangordon....u are a genius...!

using 805 freq @ 540 dsp....
not a single reboot...
all videos play fine...
video recording is awesome...

and battery is lasting about 8-10 hours under heavy usage.......

thanks... :)

nkirk

2011-11-19 06:43

Re: [Announce] kernel-power unstable v49

@freemangordon,

I currently use the default unmodified frequency table, max clock 600(when not recording HD), but when playing back videos (800x reso 4Mbps) the video stops after some time. this is caused by too low voltage at certain frequency eg @600MHZ.

Is there any way to set a custom voltage to SR desired values ?, mine sits @ "38", as i can recall i required about 40 to achieve issue-less payback.
Or is it something to do with SRvdd2 not calibrating correct values ?

Thanks in advance.

nkirk.

freemangordon

2011-11-19 08:21

Re: [Announce] kernel-power unstable v49

@Estel - VDD2 has nothing to do with MPU, it is interconnect bus voltage. And has only 3 OPPs. Damn, just enable it.

@all - the reason why I strongly recommend to not overclock DSP bellow MPU@720 when SR enabled is the nkirk's post - voltages are just not enough. As simple as that.

nkirk

2011-11-19 08:54

Re: [Announce] kernel-power unstable v49

@freemangordon,

tnx for the quick reply, The default profile mpu 600mhz, dsp 430 mhz is what i'm having trouble with, the SR voltage value seems to be a little too low for the dsp to handle the stock clock.

p:s : Video played back was SD.

Thanks in advance.

nkirk

freemangordon

2011-11-19 08:59

Re: [Announce] kernel-power unstable v49

@nkirk - no, there is no such setting. How did you check that it is 600 that is problematic? And BTW could you post your vdd1_efuse values?

nkirk

2011-11-19 10:00

Re: [Announce] kernel-power unstable v49

vdd1_efuse

0089fad6
0099ff86
00aaf5a2
00aada94
00aac488
00aba4eb
00ab92d5
00ab88c8
00bbfdbc

It could be either 500/550/600 voltages causing it, im not able to exactly reproduce it now even with locking of frequencies, but it happened once just now with stranded on-demand scaler.(mpu 125-600 mhz), A/p my observation this happens often when the battery level is less than half/ when the device has been used for a few hours. (now the bat is at full capacity/ fresh boot.)

but the issue is not there when higher voltages are used with the same profiles.
Will update you with the findings when possible.

Tnx.
nkirk

pali	2011-11-19 12:27

Re: [Announce] kernel-power unstable v49

Maemo package interface again did not import new kernel-power packages :( I think this is really bug. I pushed correct dsc and tarball of v48 and v49!

pali	2011-11-19 12:37

Re: [Announce] kernel-power stable v49 in Extras-Devel

but kernel-power, kernel-power-modules, kernel-power-flasher, ... still missing.

And kernel-power-source has Size: 0 bytes and Promote button is not visible...

freemangordon

2011-11-19 12:40

Re: [Announce] kernel-power stable v49 in Extras-Devel

At leas debug symbols are there :D
http://maemo.org/packages/package_in....28-10power49/

BTW since when MohammadAG is KP maintainer?

Brian_7

2011-11-19 12:45

Re: [Announce] kernel-power stable v49 in Extras-Devel

Hello. I have the speed patch wich set up the default maximum speed to 805 MHZ. If i install KP V49 wich will enable smart reflex, will i get a reboot error?. Because a long time ago i got a reboot error every time that i overclock to 750 MHZ with smart reflex enable.

freemangordon

2011-11-19 12:47

Re: [Announce] kernel-power stable v49 in Extras-Devel

Pali, why don't you try to promote settings package?

EDIT:
Oops, sorry, you've already done that

pali	2011-11-19 12:51

Re: [Announce] kernel-power stable v49 in Extras-Devel

I did not promoted -seetings package yet, becuase I can't. Maemo package interface does not see kernel-power-flasher so I can't promote settings package:

http://maemo.org/packages/package_in...settings/0.13/

mr_jrt

2011-11-19 12:56

Re: [Announce] kernel-power stable v49 in Extras-Devel

...it does sometimes take a while to appear. Give it an hour or so. ...I've no idea why.

All times are GMT. The time now is 04:14.

Page 53 of 75

Show 40 post(s) from this thread on one page

vBulletin® Version 3.8.8