Active Topics
-
Is there a section to talk about Java ME? Apps, etc. (0)
to General by Kalatti - 1 day, 18 hrs ago -
Extra softwares in Sailfish using CLI, repositories, etc (117)
to SailfishOS by nieldk - 3 days, 15 hrs ago -
Firefox with Leste (6)
to Maemo 7 / Leste by teroyk - 4 days, 12 hrs ago - more...
|
2010-11-25
, 00:03
|
|
Posts: 7 |
Thanked: 5 times |
Joined on Aug 2010
|
#312
|
Hi Titan,
I have been playing with LVM for the n900, and it works quite well. I've successfully setup dmcrypt for /home and swap. The device boots asks for a passphrase, and feels just as fast with AES-128-CBC.
I would like to extend this to RAID-0 the onboard flash with my class-10 uSD card and see if swap performance increases.
Are you able to build the md/raid0.ko module for the latest power-kernel so I can test this?
Alternatively, can you point me in the direction for building my own modules for Kernel_Power?
Thanks!
-Eric
I have been playing with LVM for the n900, and it works quite well. I've successfully setup dmcrypt for /home and swap. The device boots asks for a passphrase, and feels just as fast with AES-128-CBC.
I would like to extend this to RAID-0 the onboard flash with my class-10 uSD card and see if swap performance increases.
Are you able to build the md/raid0.ko module for the latest power-kernel so I can test this?
Alternatively, can you point me in the direction for building my own modules for Kernel_Power?
Thanks!
-Eric
| The Following User Says Thank You to ewheelerinc For This Useful Post: | ||
|
|
2010-11-25
, 06:58
|
|
Posts: 50 |
Thanked: 444 times |
Joined on Apr 2010
@ Austria
|
#313
|
Hi Eric,
It is awesome someone tries that advanced setups.
I think it wouldn't be neccesary to use raid0 for swap partitions. You could simply add a second swap partition and as far as I know the kernel automatically uses both for best performance.
I haven't done this myself so far, so maybe you could also include this in your performance test.
Regards,
lxp
It is awesome someone tries that advanced setups.
I think it wouldn't be neccesary to use raid0 for swap partitions. You could simply add a second swap partition and as far as I know the kernel automatically uses both for best performance.
I haven't done this myself so far, so maybe you could also include this in your performance test.
Regards,
lxp
|
|
2010-11-25
, 07:52
|
|
Posts: 7 |
Thanked: 5 times |
Joined on Aug 2010
|
#314
|
Originally Posted by lxp
I know it will do some amount of swap balancing, but I don't know if it distributes swap writes across the same swap priority.
Hi Eric,
It is awesome someone tries that advanced setups.
I think it wouldn't be neccesary to use raid0 for swap partitions. You could simply add a second swap partition and as far as I know the kernel automatically uses both for best performance.
I haven't done this myself so far, so maybe you could also include this in your performance test.
Regards,
lxp
Does anyone know how Linux balances swap-outs across multiple swap devices?
For example, if Linux swaps to every-other-swap-device at page granularity, that may be suboptimal with 4k pages. The erase block size of flash is probably somewhere between 16k and 128k. Ideally we want the RAID0 stripe size to be exactly one erase-block size to minimize write amplification and write to the other block-device while the first is being written to (assuming they are asynchronous, which they should if the SD interfaces use DMA).
Does anyone know the erase-block size of the N900's onboard 32GB flash?
-Eric
|
|
2011-01-01
, 06:51
|
|
Posts: 73 |
Thanked: 47 times |
Joined on Aug 2010
|
#315
|
Originally Posted by egoshin
Happy New Year!
In addition I recommend you to run this commands and test it again:
echo "1" > /sys/devices/system/cpu/cpu0/cpufreq/ondemand/ignore_nice_load
renice 1 `cat /syspart/applications/standby/background/tasks`
renice 1 `ps | grep modest | cut -c1-5`
It should save you a lot of battery over standby. I have increase more then 50%.
(Note: modest renice would work until you call e-mail GUI, after that it reverts back. I am working on it)
Sorry to bring this out again. I tried your suggestion but I failed on this:
Code:
renice 1 `ps | grep modest | cut -c1-5`
Code:
renice: setpriority: No such process
Any suggestions or any further development on improving N900 battery life after PR1.3 and kernel power42, power46?
Also, beside email GUI problem, is this solution permanent? I mean do I have to do this after every reboot?
Last edited by Duy2anh; 2011-01-01 at 06:56.
|
|
2011-01-06
, 15:30
|
|
Posts: 468 |
Thanked: 610 times |
Joined on Jun 2006
|
#316
|
How do I compile a kernel-power kernel myself?
I followed guides on compiling kernel in scratchbox using information:
http://wiki.maemo.org/Documentation/...o_Kernel_Guide
and
http://www.natisbad.org/N900/n900-custom-kernel.html
the default nokia kernel seems to compile but the power kernel doesn't. As suggested here, I tried the extra step in copying the default config:
http://psfreedom.com/wiki/Nokia_N900...9_maemo_kernel
But when I do "make EXTRAVERSION=.10power46" I stil get an error:
If I only compile modules, that seems to work OK, but compiling the entire kernel doesn't work.
Can anybody help me?
I followed guides on compiling kernel in scratchbox using information:
http://wiki.maemo.org/Documentation/...o_Kernel_Guide
and
http://www.natisbad.org/N900/n900-custom-kernel.html
the default nokia kernel seems to compile but the power kernel doesn't. As suggested here, I tried the extra step in copying the default config:
http://psfreedom.com/wiki/Nokia_N900...9_maemo_kernel
Code:
cp debian/rx51power_defconfig arch/arm/configs/
Code:
In file included from arch/arm/mach-omap2/pm.c:37: arch/arm/plat-omap/include/mach/pm.h:111:2: error: #error "Power management for this processor not implemented yet" make[1]: *** [arch/arm/mach-omap2/pm.o] Error 1 make: *** [arch/arm/mach-omap2] Error 2
Can anybody help me?
|
|
2011-01-16
, 19:50
|
|
Posts: 18 |
Thanked: 72 times |
Joined on Sep 2008
|
#317
|
The EARLY_SSH functionality is currently broken on my N900. The problem is that something early in the boot sequence loads g_file_storage, which causes problems when /etc/event.d/kernel-power tries to load g_nokia. Adding a modprobe -r g_file_storage to the startup script just before the modprobe g_nokia made it work perfectly. Can you fix this in v47?
|
|
2011-01-18
, 10:50
|
|
Posts: 155 |
Thanked: 61 times |
Joined on Nov 2009
|
#318
|
Hi Titan
in case you couldnt guess... I still am very grateful for your work with the kernel and including so many useful patches and modules.
So I have one more request - any chance you could include the compcache / zram patches? I want to try using zram on the N900 - I think it could make quite a difference to avoid going to flash for swap.
I know it's already been requested - but
0.7 is even better - no need for a user mode util because it uses /sys/block/zramX/ files for config
and also - the zram module is still useful even without the swap_notify patch - that just makes it more efficient.
Last edited by phedders; 2011-01-18 at 10:58.
in case you couldnt guess... I still am very grateful for your work with the kernel and including so many useful patches and modules.
So I have one more request - any chance you could include the compcache / zram patches? I want to try using zram on the N900 - I think it could make quite a difference to avoid going to flash for swap.
I know it's already been requested - but
0.7 is even better - no need for a user mode util because it uses /sys/block/zramX/ files for config
and also - the zram module is still useful even without the swap_notify patch - that just makes it more efficient.
Last edited by phedders; 2011-01-18 at 10:58.
| The Following User Says Thank You to phedders For This Useful Post: | ||
|
|
2011-01-19
, 06:05
|
|
Posts: 455 |
Thanked: 278 times |
Joined on Dec 2009
@ Oregon, USA
|
#319
|
Originally Posted by phedders
I don't get it, why would you use a ramdisk for swap? The whole reason the kernel even uses swap is to take inactive pages out of memory so there is room for active pages. By creating a ramdisk, you reduce the amount of RAM you can use for these active pages. Maybe to reduce wear on the flash device? I think this would ultimately reduce performance since your usable RAM size will be significantly reduced?
Hi Titan
in case you couldnt guess... I still am very grateful for your work with the kernel and including so many useful patches and modules.
So I have one more request - any chance you could include the compcache / zram patches? I want to try using zram on the N900 - I think it could make quite a difference to avoid going to flash for swap.
I know it's already been requested - but
0.7 is even better - no need for a user mode util because it uses /sys/block/zramX/ files for config
and also - the zram module is still useful even without the swap_notify patch - that just makes it more efficient.
|
|
2011-01-19
, 06:13
|
|
Posts: 455 |
Thanked: 278 times |
Joined on Dec 2009
@ Oregon, USA
|
#320
|
Originally Posted by phedders
I don't get it, why would you use a ramdisk for swap? The whole reason the kernel even uses swap is to take inactive pages out of memory so there is room for active pages. By creating a ramdisk, you reduce the amount of RAM you can use for these active pages. Maybe to reduce wear on the flash device? I think this would ultimately reduce performance since your usable RAM size will be significantly reduced?
Hi Titan
in case you couldnt guess... I still am very grateful for your work with the kernel and including so many useful patches and modules.
So I have one more request - any chance you could include the compcache / zram patches? I want to try using zram on the N900 - I think it could make quite a difference to avoid going to flash for swap.
I know it's already been requested - but
0.7 is even better - no need for a user mode util because it uses /sys/block/zramX/ files for config
and also - the zram module is still useful even without the swap_notify patch - that just makes it more efficient.
| The Following User Says Thank You to craftyguy For This Useful Post: | ||
I search it in the catalogues but nothing find ...
(n900 _Switzerland)