kernel 3.1.1

[critter]

hey guys, I know Tex is busy right now with the 64 bit spin so if anybody wants to try it
kernel 3.1.1 compiles 'out of the box' on pclos kde4 2011

Source is here  http://kernel.org/

make menuconfig (I didn't change anything here - yet)
make
make modules
make modules_install
make install

That's all. It'll take a while but seems quick and stable on this machine.
Enjoy, I'm off to slim it down and see if I get any performance gains.

[muungwana]

whats missing in the distro's kernels that make you go out and build your own?

[glamdring]

whats missing in the distro's kernels that make you go out and build your own?

Added support for some hardware I have, granted with some hacking I can get it to work on the current one in repos. I didn't update my main system, I can wait the little bit to let Texstar put it into repo.

[gseaman]

I use a custom 3.1 kernel with bfs and bfq. The advantage for me is improvements in the 3d speed of the intel video driver and the bfq io scheduler. I start with the existing PCLinuxOS .config file.

Galen

[critter]

The kernels in the repositories are fine but the latest , non-repo kernels, do support some leading edge hardware and possibly have other benefits. Con Kolivas BFS patch made a difference and I am sure that the kernel devs will have made some adjustments to the code to take full advantage of the scheduler improvements it implemented.

Apart from that I did it ...
to go beyond the frontiers where man has not been before.
because it is there
because I can
for the heck of it
because this is the advanced users section where people experiment with things to, hopefully, learn about new features that are becoming available and perhaps make those features more easily and sooner available to those less experienced users.

If none of those reasons appeal to you then please ignore the post. It was meant only as information for those users who may find a benefit from a later kernel or, like me, just like messing with things .

[Just17]

Of those of you who have run your own build of newer kernel, did anything not work as expected?

Thanks 

[AS]

Of those of you who have run your own build of newer kernel, did anything not work as expected?

Thanks 

the last time I tried (3.0.x) I was experiencing some build issue around dkms nvidia, I think this is now solved by recent nvidia releases, another issue while running 3.0.x (using nv) was related to net_applet displaying the X on a red background but connection was active ... it was a few months ago ...

AS

[gseaman]

Of those of you who have run your own build of newer kernel, did anything not work as expected?

Thanks 

dkms-fglrx doesn't work for me (radeon hd4200). My intel machines work better than ever, though. Also net_applet, but I think this is net_applets poor design, because it also doesn't work if I speed up booting with the standard kernel using e4rat. The main problem with net_applet is the status of the connection is reported incorrectly. The link to draknetcenter works.

Galen

[Just17]

Thanks for the reports 

[kjpetrie]

The only place I use non-pclos kernels is on my laptop where I need some KMS and powers saving features not compiled in Tex's kernel. With 2.6.28 kernels compiling these in causes X to hang on shutdown (at least I think it's X) leaving the machine dead to everything except REISUB.

I find the 2.6.36 (on 32 and 64-bit) and 3.0.4 (on 64-bit) work well without this problem, though the netapplet also doesn't work as intended on the latter. It connects but doesn't display the fact. That might be more do do with the 64-bit though, so I'll worry about that when it goes live.

However, what doesn't work is remastering, either because the script needs tweaking or because the compressed file systems rely on patches.

[jfdesign]

is fast on my intel machine too. all devices were recognized and works as i expected.
but as kjpetrie said, remaster won't work.

Also net_applet, but I think this is net_applets poor design, because it also doesn't work if I speed up booting with the standard kernel using e4rat. The main problem with net_applet is the status of the connection is reported incorrectly. The link to draknetcenter works.

here is the fix :

--- lib/network/tools.pm.orig   2011-11-01 22:13:45.588528869 +0700
+++ lib/network/tools.pm   2011-11-01 22:19:30.689503038 +0700
@@ -263,7 +263,12 @@
 
 sub get_routes() {
     my %routes;
-    foreach (cat_("/proc/net/route")) {
+    # jfdesign - fix reading routes on both older and newer kernels (2.6.39+)
+    #foreach (cat_("/proc/net/route")) {
+    my @routes = cat_("/proc/net/route");
+    require bootloader;
+    @routes = reverse(@routes) if bootloader::cmp_kernel_versions(c::kernel_version(), "2.6.39") >= 0;
+    foreach (@routes) {
    if (/^(\S+)\s+([0-9A-F]+)\s+([0-9A-F]+)\s+[0-9A-F]+\s+\d+\s+\d+\s+(\d+)\s+([0-9A-F]+)/) {
        if (defined $3) { $routes{$1}{gateway} = hex($3) ? host_hex_to_dotted($3) : $routes{$1}{network} }
        if (defined $2) { $routes{$1}{network} = host_hex_to_dotted($2) }

i name it net_applet-routes-fix.diff and place it as Patch12 on spec file
Patch12: net_applet-routes-fix.diff
%patch12 -p0

it must works both on <2.6.39 and 3.0+, but you may check yourself with 2.6.38.8 series since i won't go back. the code isn't mine and all credits goes to tv@...

[gseaman]

I could test and rebuild the package, but who is tv@... ? If the problem is solved, I would like to give appropriate credit in the spec file.

Galen

[jfdesign]

I could test and rebuild the package, but who is tv@... ? If the problem is solved, I would like to give appropriate credit in the spec file.

Galen

pm