The Official HWF Kernel Repo

Now I am sure I am doing something wrong!
I am sorry AT, all your efforts are waisted on me.

I installed the new headers then the new image (bothe for the Pentium-M) and then rebooted.
While installing the linux-image-blabla I get the following warning:

However, the installation finishes without any other issues.
Then on reboot I get a message about "usb 1-1: somthing-o-rather error -71"

Any ideas? WTF am I doing?

 

The error about the symlinks is normal, it has something to do with the kernel packager utility and seems to randomly affect some systems and not others. The fix is also easy, just make sure you're running the custom kernel, then do the following (as root):

Code:

ln -s /usr/src/linux-headers-`uname -r` /lib/modules/`uname -r`/source
ln -s /usr/src/linux-headers-`uname -r` /lib/modules/`uname -r`/build

It's best to copy and paste my commands exactly since they should work on any Debian system, including Ubuntu.

As for the USB error, does the system still boot OK, or does it hang on that error? It could be a normal thing for your hardware, I really can't say. But if you're using proprietary Nvidia or ATI drivers, it's normal for the system not to boot to a graphical display manager after installing a custom kernel. That's when you install the drivers the vendor way.

P.S. - The new Gutsy kernels are just finishing uploading as I type this.

 

Hi AT, I admire your patience....

No, the system hangs. The error message displays right after the kernel is decompressed. Then the boot hangs.

Which brings me to my next question:

I am assuming you are referring to the HWF kernel... but that is not possible as I cannot seem to boot into it....
Unless I am wrong and then I can run the 'fix' before I re-attempt to install the new HWF kernel.

Thanks AT....

 

Hey saba, the problem you're seeing with the symlinks is not a serious issue, and it certainly wouldn't prevent the system from booting. As such, it should be treated as a totally separate issue.

As for your system not booting, sounds like a laptop thing. Try appending the following to the end of your boot string:

Code:

noapic

...does it boot then? If not, please at least include the full error message in its original form so we can look into it.

-AT

 

Right, so -

Before adding the noapic, I am getting the following message before the system hangs:

After adding the noapic, I am getting the following two messages before the system hangs:

 

Yup, this problem is actually unrelated to the custom kernel, it's most likely a glitchy BIOS that's manifesting as USB issues. You don't happen to have an HP laptop, do you?

The root cause is probably an IRQ steering problem. You might try booting with the pci=irqpoll option, which may help.

 

Hi AT

No, I have a Dell D600 (Latitude) in which I have not got any control over the BIOS... so I guess I am stuck with that.

I will try your latest suggestion and see where I get to.

[ot]I succesfully installed and running VirtualBox, running Win XP, including access to WWW. Bye Windows! And thanks for everything!
I love KT *sniff*[/ot]

 

The Lenny builds of the new 2.6.23.17 kernel finished first, and are ready for download. As soon as the Gutsy builds are finished I'll post to let you know.

2.6.23.17 changelog



EDIT: The Ubuntu kernels took forever to build for some reason, but they're up now!

Index of /repo/kernels

 

A little update...

A little update...

It doesn't surprise me when people mention how much faster Linux is when running one of our custom kernels. That I expect. What's amazing to me is that I've had many emails and IM's telling me how much more stable Ubuntu Gutsy runs with our custom kernel in place. I'm not sure exactly why, except that I haven't included any Ubuntu-specific patches like the stock Ubuntu kernel does. In any case, thanks for the kind feedback and I hope you can get some good use out of our kernel builds.

-AT

 

personally, i'm not surprised, as it is great, my only issue is a freeze up every now & then, but, i know what that is, basically case is full of dust & crap and with the grx card mounted horizontally (midi tower case) so, my grx card is overheating because it hasn't being cleaned since i got it (about 3 years ago) except obviously an upgrade (mobo, CPU, GPU & RAM)

if i remember, i'll ask my carer to clean it out

Thankies AT you da man

 

Kernel 2.6.25 was released today, and with it some exciting new features which will allow us to push our machines even further under heavy load. As of right now, some of the newer restrictions which prevent export of gpl-only symbols to non-gpl binary blobs will break things like nvidia, virtualbox, ndiswrapper, etc. I'm experimenting with a little 3rd-party patching of my own, and if all goes well I'll have a new kernel build within the next few days.

-AT

 

Hey AT

I am not sure if this at all possible but if/when you prepare 2.6.25 is it possible to include/keep the bcm43xx module rather than his younger and distructive b43?
Or perhaps have both so users like me have the option of which one to load?

Thanks!

 

What's the problem with b43? I didn't realise the driver changed, but after a bit of tinkering it works fine, much better than bcm43xx actually.

 

I could be wrong (as I always say, I am not experienced enough to go into fine details) but with the b43 I keep getting system freezes. I would be merilly surfing and all of a sudden the system will stutter and then freeze completely. What I noticed is that before the freeze, network manager would loose connection.

The bottom line is that this behaviour never occurs when using the bcm43xx.
I cannot seem to find a good explenation/solution for a novice.

I was just glad to be able to use the HWF kernel which allows me to use bcm43xx....

I would love hearing about the tinkering you are referring to...

 

I'm not sure if it's a problem with the b43 driver vs the bcm43xx driver per se, or if it's just the way I was building the bcm43xx. I have actually compiled the bcm43xx driver to use mmio instead of pio memory access (pio is the default). I imagine the same settings exist for the b43 driver; you might try compiling it to use mmio and see if it's the memory tweak that helps or the old driver itself.

 

Thanks AT. I normally install 4b3/bcm43xx-fwcutter and it does the job. Do you know a guide somewhere I can easily follow do compile the b43 driver and install it? Or have I got it wrong?

 

This is down for now, until I can get my build environment back up to scratch. However, when I do start building kernels again, I will exclude Ubuntu and build exclusively for Debian distros. The main reason is that Ubuntu users are typically less technical and thus unwilling to do things manually, so rolling custom kernels for that distro is counter-intuitive. On the other hand, there's nothing stopping adventurous Ubuntu users from using kernels built for Lenny, which is genetically pretty close to Hardy and Intrepid. Also, I plan to build the latest 2 supported kernel trees; the bleeding edge, and the runner up.

 

good, i like debi, although i being using gutsy cos it was easier & i was using RC2 when my HDD went to the big computer shop in the sky, so, i'll need to download a new ISO when i sort this security thing out

wait, thinking about it i can invoke & download a secure debi ISO using CLI :doh:

 

If you haven't noticed, the posts in this thread is obsolete, though I've updated the first two posts in this thread to reflect the changes. I'm building for 10 different x86 architectures now, and the configuration is much closer to the Debian baseline, so compatibility should be good. The new kernel wiki is up over here:

HWF Custom Kernel Wiki

Should you experience any difficulties with the current kernel builds please feel free to post in this thread.

Best regards,
-AT

 

Added a new kernel, namely 2.6.30. There are some pretty substantial IO performance improvements, new filesystem additions and improvements to existing ones, some new security mechanisms, many new drivers, some performance gains and also a few regressions. Watching the GIT update logs, it looks like the next release will fix some of these and also make it easier to benchmark for performance regressions in the future. However, for now it may be a step backwards depending on what you're doing with your rig. For this reason, 2.6.29.4 is still up and available for download.

• Ultra-geeky official changelog here.
• User-digestible changelog over here.
• Download the HWF custom kernel over here.