Having trouble booting the 5.4 kernel on Ubuntu Jammy 22.04 about linux-kernel-programming

[kaiwan]

Hi @redbilledpanda ,

Thanks for writing..
It's peculiar that you're running into issues with a stable 5.4 kernel.
Could you try simplifying: download a stable kernel source tree from kernel.org (f.e. the latest 5.10 LTS as of now
is 5.10.169; here:
https://mirrors.edge.kernel.org/pub/linux/kernel/v5.x/linux-5.10.169.tar.xz
).

Build it via the usual steps as shown in the book.
Am working on the 2nd edition; here's a simple wrapper script to build the kernel:
https://github.com/PacktPublishing/Linux-Kernel-Programming_2E/blob/main/ch2/kbuild.sh

HTH! Lemme know..

[redbilledpanda]

Hello @kaiwan

IMHO, the kernel build process is so tightly coupled with the underlying build infrastructure that certain things are taken for granted. Problems like the ones I mentioned above arise when we try do go against the established norm, like using a recent build system to compile a relatively older kernel. I am actually a bit aghast at the huge amount of complexity that has gone into the build process itself vis-a-vis getting the configuration just right enough to build and boot. For instance, building the 5.10.169 version of the kernel on my machine gives this error:

make[1]: *** No rule to make target 'debian/canonical-revoked-certs.pem', needed by 'certs/x509_revocation_list'.  Stop.
make[1]: *** Waiting for unfinished jobs....

As this post mentions, I could actually disable the whole key related shebang or else I would have to actually install a certificate that the kernel build process recognizes.

So I went with the second option of installing certfificates, taking help from the above link. After that, I was able to finally boot into this kernel, sigh! By the way, how is it that even with this new kernel, the existing window manager (LXDE in my case) starts without any issue as if nothing happened? Can we say that the window manager startup daemon (it has to be a daemon after all right?) is abstracted from the rootfs layout process? If not how to explain it?

Anyways, thanks for being so responsive and prompt and for writing such a detailed book. After ULK and LKD which are great but way too old now, this will be my go to book. Additionally, it's good to know that a second edition is also on the cards! I hope it'll be coming out soon and be available on Safari (the learning platform).

[kaiwan]

Hello @kaiwan

IMHO, the kernel build process is so tightly coupled with the underlying build infrastructure that certain things are taken for granted. Problems like the ones I mentioned above arise when we try do go against the established norm, like using a recent build system to compile a relatively older kernel. I am actually a bit aghast at the huge amount of complexity that has gone into the build process itself vis-a-vis getting the configuration just right enough to build and boot.

Hm, yeah... we have to take the good with the bad..

For instance, building the 5.10.169 version of the kernel on my machine gives this error:

make[1]: *** No rule to make target 'debian/canonical-revoked-certs.pem', needed by 'certs/x509_revocation_list'.  Stop.
make[1]: *** Waiting for unfinished jobs....

As this post mentions, I could actually disable the whole key related shebang or else I would have to actually install a certificate that the kernel build process recognizes.

Yes.. my helper script simply disables it...

So I went with the second option of installing certfificates, taking help from the above link. After that, I was able to finally boot into this kernel, sigh!

Nice!

By the way, how is it that even with this new kernel, the existing window manager (LXDE in my case) starts without any issue as if nothing happened? Can we say that the window manager startup daemon (it has to be a daemon after all right?) is abstracted from the rootfs layout process? If not how to explain it?

It's just this- the kernel has very loose coupling with the rootfs; thus, it simply works.. things break when you use a very old rootfs with newer kernels (>=3.x) or vice-versa..

Anyways, thanks for being so responsive and prompt and for writing such a detailed book. After ULK and LKD which are great but way too old now, this will be my go to book.

Thanks you!

Additionally, it's good to know that a second edition is also on the cards! I hope it'll be coming out soon and be available on Safari (the learning platform).
Well, it'll be available on the Packt platform at least...
Thanks for writing in, and the very best with kernel adventures :-)