Pandora Kernels For The Pandora


chuckr

Still Fresh
Joined
Mar 30, 2009
Messages
93
Age
72
Location
Calverton Maryland
I need to know who's organizing development of kernels for the Pandora, if there's a tarball of the sources (and all of the available, if not selected, options) so I could maybe build me a kernel with additional options. Specifically, right now it seems that /sbin/mount in the currently distributed rootfs doesn';t recognize -text4, but I want to use ext4 in my bootable flash image (the main dir, not the little boot partition). There is already what appears to be a working mkfs.ext4, but if it can't be mounted, it's not of any help yet, and I *think* it's going to need a new kernel option, which means I'm going to want the pre-existing config file (the one last used), and to rebuild (I think) /sbin/mount. I don't know who to talk to about this, but it won't have been the first kernel I've built, and adding only 1 featue here doesn't sound like a huge job. I'm giong about here making myself a cross-compiling environment on my big host, and hopefully I can get URLs for the stuff I need from whomever reads this plea.
 
Everything is on http://git.openpandora.org
See http://pandorawiki.org/Kernel_build_instructions .

Pandora currently uses 2.6.27 kernel which doesn't have ext4 (it does however have ext4dev, but it's not included in firmware).
 
Is there somehow a rough timeframe when we can expect a new Kernel for Pandora? Mainly because all of this OMAP/ARM releated "HW Acceleration 'n powersave-function" stuff. :)
 
fusion_power said:
Is there somehow a rough timeframe when we can expect a new Kernel for Pandora? Mainly because all of this OMAP/ARM releated "HW Acceleration 'n powersave-function" stuff. :)
And have some sweet usb DVB-T tuner while at it :D
The driver I need for the USB got included in the kernel 2.6.30
 
Last edited by a moderator:
Building the kernel native on the pandora is easy enough, just start with
Code:
export CROSS_COMPILE=""
if I remember right. (maybe there is a better way, but this seems to do the trick)
Takes a couple of hours @ 800 MHz.
 
notaz said:
Everything is on http://git.openpandora.org
See http://pandorawiki.org/Kernel_build_instructions .

Pandora currently uses 2.6.27 kernel which doesn't have ext4 (it does however have ext4dev, but it's not included in firmware).
Looks like these source codes are not identical to the sources of the default binaries.
I compiled it using the wiki instructions - wifi is not working, there is no omaplfb module.
 
Last edited by a moderator:
Vitel said:
notaz said:
Everything is on http://git.openpandora.org
See http://pandorawiki.org/Kernel_build_instructions .

Pandora currently uses 2.6.27 kernel which doesn't have ext4 (it does however have ext4dev, but it's not included in firmware).
Looks like these source codes are not identical to the sources of the default binaries.
I compiled it using the wiki instructions - wifi is not working, there is no omaplfb module.

That does sound like you are missing the kernel modules somehow?

Our image (OE) is built from that GIT and WiFi as well as omaplfb works there.

I don't know much about devving, but you are definately missing the latest kernel modules.
You can download them compiled at http://openpandora.org/firmware/images/modules-2.6.27-pandora+r20+gite770b4b956de8aee25c0c54ffdd3d2a4e62347ea-r20-omap3-pandora.tgz
 
Last edited:
Vitel said:
Looks like these source codes are not identical to the sources of the default binaries.
I compiled it using the wiki instructions - wifi is not working, there is no omaplfb module.
Wifi and SGX modules are built separately from the kernel by OE. Easiest way to get identical kernel and modules to what is in firmware is to use OE.

Spirit said:
I'd sacrifice a goat if someone shares a tickless kernel.
Do you really think it such a big deal? If you base this on urjaman's results, then know he also did OPP tweaks, used lowest brigtness and I'm not sure he had the display on when he measured his power usage.
 
Last edited by a moderator:
notaz said:
Wifi and SGX modules are built separately from the kernel by OE. Easiest way to get identical kernel and modules to what is in firmware is to use OE.
Looks like the wifi recipes are located in openpandora.oe/recipes/pandora-system/
Where are the recipes for SGX modules?
 
Last edited by a moderator:
Vitel said:
Where are the recipes for SGX modules?
http://cgit.openembedded.org/cgit.cgi/openembedded/tree/recipes/powervr-drivers
 
Last edited by a moderator:
notaz said:
Do you really think it such a big deal? If you base this on urjaman's results, then know he also did OPP tweaks, used lowest brigtness and I'm not sure he had the display on when he measured his power usage.
My tickless system uses less power in idle (no bluetooth, no wifi, no networkmanager, just XFCE and a terminal) with display on (backlight minimum) @500Mhz (OPP3, normal) (~190mA IIRC) than the current "ticky" system with backlight off (I think in the lowest power state i could find back then, though I think the LCD was still on), that was about 220mA (these measured with "sleep 30s; cat /sys/class/power_supply/bq27500-0/current_now" + lid close if with backlight off + wait ~40s + open lid and look at results; just a simple cat for the display on measurement)

The biggest :eek: moment of the tickless system was that it really made the processor speed in idle quite irrelevant. Some difference with OPP still, but (with backlight and display off, idle, tickless (and i have preemptive on in the kernel also, i felt experimental)) it was 124mA for 125mhz at OPP1 vs 130mA for 500Mhz (at OPP3). Without the tickless the mhz change resulted in much bigger changes in power usage. But of course if you're using the tickless system, it will be draining propably (when running with >500 (sometimes over 1000) wakeups/sec == me browsing the net (with lots of tabs) with firefox) near the same as a system without tickless (that had something like 940 wakeups/sec in idle vs less than 50 wakeups/sec with tickless).
 
Last edited by a moderator:
Alerino said:
ED said he was testing a new kernel, but still facing lots of instabilities
i'd say we may have a new kernel in two months.
I can imagine, that as longer it takts till the new kernel is in the Pandora, as more SW is made to run for the old Kernel and as harder it will be to make all the "old" Stuff run propper onto the New Kernel. ^^"

@urjaman: will your system make it into the official builds? :)
 
Last edited by a moderator:
fusion_power said:
Alerino said:
ED said he was testing a new kernel, but still facing lots of instabilities
i'd say we may have a new kernel in two months.
I can imagine, that as longer it takts till the new kernel is in the Pandora, as more SW is made to run for the old Kernel and as harder it will be to make all the "old" Stuff run propper onto the New Kernel. ^^"

@urjaman: will your system make it into the official builds? :)

Erm, nope :)
Normal software should not be affected by the kernel, only the modules and drivers.

So the old games will still work with a new kernel.
 
Last edited:
Back
Top