I've applied the new Nokia kernel patch (PR 1.3) and the latest BFS patch (v357) to the power kernel. I was waiting for Titan to add the PR 1.3 kernel patch to his power kernel, but as it currently isn't released, I thought I'd give it a shot myself.

I've had to remove the CFS code from Nokia's patch as it was conflicting with the BFS patch, but that shouldn't matter as CFS is replaced by BFS anyway. Furthermore, the power kernel kexec and led brightness patch were conflicting with the PR 1.3 patch, which AFAIK provides the same functionality, so I've excluded these two patches from the build.

This kernel is NOT multiboot compatible and got some other package issues, see post http://forums.internettablettalk.com...4&postcount=51.

IPv6, reiserfs and XFS support are removed and the deadline + anticipatory schedulers are added to the kernel.

Edit: new releases are to be found on the garage page: https://garage.maemo.org/projects/kernel-bfs/
Some older versions can be downloaded from this post

And here are the patches, if anyone want to try them out theirselves.

nokia-20103103+0m5.orig.diff -> Nokia's PR 1.3 patch
strip_patch_for_bfs_kernel.diff -> Has to be applied against Nokia's PR 1.3 patch. This patch removes the CFS updates, and some conflicting/overlapping features to be able to build.
add_patches_and_config.diff -> Include the new patches in the build + change the kernel configuration a bit.
bfs-350-to-357.patch -> This one is obvious

You'll have to manually copy bfs-350-to-357.patch and the nokia patch (after patching it with the strip patch) to the patches directory of the kernel (kernel-2.6.28/debian/patches)

These patches are meant to be used with a GIT clone of the kernel-bfs project.

Edit: to avoid confusion: this post has nothing to do with the post below this one. The patches provided in the post below are to be applied (like displayed in the pastebin link) against a clean git clone, not against the work in this, or the previous post. The BFS patch, strip CFS patch and Titans PR 1.3 patch are all included in the patches from the post below

Last night I've updated kernel-bfs to reflect the changes from kernel-power 2.6.28-maemo42 (which include PR 1.3 kernel patch, RDS support for FMRX driver, removed 500MHz lock while on charger and more)

Update 12-11-2010: added patch + patched tree which reflects kernel-power 2.6.28-maemo45
Maemo45 bumps the kernel-bfs version to -bfs4. Make sure your battery has plenty of juice before using this, there is a thread about boot problems with maemo45 (and reflashing with low battery apparently doesn't work). The compiled debs below are still maemo42 equivalents. Kernel-power 2.6.28-maemo45 brings U-boot and experimental USB hostmode; Unless you explicitly need those features, you want to use the maemo42 equivalent. Also, read this.

Compiled debs:
http://members.upc.nl/m.groenen60/ke...maemo42.tar.gz

Modified ('pre-patched') git clone of kernel-bfs
Kernel-power 2.6.28-maemo42 equivalent:http://members.upc.nl/m.groenen60/ke...-source.tar.gz
Kernel-power 2.6.28-maemo45 equivalent: http://members.upc.nl/m.groenen60/ke...-source.tar.gz

Patches for a clean git clone of kernel-bfs
(git clone -> maemo41) http://members.upc.nl/m.groenen60/ke...1_bfs357.patch
(maemo41 -> maemo 42) http://members.upc.nl/m.groenen60/ke.../maemo42.patch
(maemo42 -> maemo 45) http://members.upc.nl/m.groenen60/ke.../maemo45.patch
Also see: http://pastebin.com/FFHFw6Ac

Patched kernel-power-settings to depend on kernel-bfs instead of kernel-power
http://members.upc.nl/m.groenen60/ke...0.11_armel.deb

Of course, main credits go to coreyoconner for porting BFS and setting up the kernel-bfs project I'll contact him to see if he agrees with the patches and if so, maybe they can be pushed to the online git tree.

The compiled kernel name is 2.6.28-bfs3, following the name scheme suggested by the kernel-bfs tree.
Build instructions are to be found at http://members.upc.nl/m.groenen60/howtobuild.txt. You'll have to copy the debian folder to build the modified tree as described in the text document.

Edit: new releases are to be found on the garage page: https://garage.maemo.org/projects/kernel-bfs/

If anyone else is using this and needs the fcam drivers for it, with the new headers, here they are

Hi,

Could you please post step-by-step instructions to install this kernel, if you're starting with kernel power .45?

At the end I would like to have kernel-bfs with the same functions as kernel power .45.

silbah

what is the difference between this kernel and power45?

Just run "Deinstall Kernel-Power" from the Maemo application menu to uninstall kernel-power. To install kernel-bfs: download the .tar.gz containing compiled deb files, extract it and install the .deb. files.
Command to extract .tar.gz: tar -xzvf kernel-bfs-maemoXX.tar.gz
Command to install .deb files: (as root) dpkg -i XXX.deb
To install all debs at once: (as root) dpkg -i *.deb

In my previous post I've only provided a pre-compiled kernel-bfs equivalent of kernel-power 2.6.28-maemo42 because most users don't want maemo45. Here's a link to the pre-compiled equivalent of maemo45 as you've requested it:
http://members.upc.nl/m.groenen60/ke...maemo45.tar.gz
Please only use this if you explicitly need U-boot and/or hostmode, otherwise stay with the more stable maemo42 (as per Titan's suggestion)

The BFS scheduler . Google it to learn more about it. A good starting point is the FAQ made by the author of BFS (Con Kolivas): http://ck.kolivas.org/patches/bfs/bfs-faq.txt
There's also a Wikipedia page of it, but the link would probably be censored, so you'll have to google it yourself

did anyone got any improvements after installing this?.....i read on wikipedia that there is no improvement using this in the last android...why would be different for maemo?

Android != Maemo

Android is way more polished, smoother if you like, while Maemo behaves/is more like the typical desktop Linux OS. This makes it a hell of a great OS, but on a resource-limited device, it can get sluggish. This is where BFS jumps in.

BFS is designed with general Linux 'desktop' workloads in mind. It doesn't scale well with workstations/servers which have many cores. It is less complex and has less overhead than CFS. BFS doesn't focus on throughput, but rather on a more responsive desktop.

That's the reason I've been using it. User experience is really hard to measure, while static benchmarks generally don't represent the typical desktop Linux user's workload. Therefore, it is hard to prove the benefits of using BFS. I suggest you to try it out and judge for yourself.

i will give it a try.....