[SOLVED] Kernel 3.5.4 - uniprocessor system fails to boot with multi-processor option off

VolumetricSteve · 10-10-2015, 06:30 PM

I've been carefully documenting what changes I make to my kernel since my last weird issue.

On a Pentium 3 933MHz, which I feel qualifies pretty solidly as a uniprocessor system, only boots if "Symmetric multi-processing support" is built into the kernel.

I straight up don't understand why.

My config is derived from the make i386defconfig command.

Keruskerfuerst · 10-11-2015, 11:13 AM

Today, all modern processors have more than one core.

I think the developers only write code for multiprocessor systems.

Just turn "Symmetric multi-processing support" on and the system runs.

Why are you using such a old CPU?

VolumetricSteve · 10-11-2015, 05:03 PM

Long story short, I'm building an arcade machine.

Most people do write code for multi-core systems now, you're right..but the option remains in the kernel for uniprocessor systems, which is why this confuses me. The option is there, my system is right for it, and yet it doesn't work. It could very well be there is some other part of the startup process that is looking for the multi-threaded code - even if it can only execute in one thread. I'd just like to know for sure.

Thanks for your reply.

VolumetricSteve · 10-26-2015, 07:40 AM

Figured it out.

I was using ext4 as my main filesystem, but I was also compiling without TLB huge pages...which I believe is a requirement of ext4, thus, none of my filesystems could mount properly.

My configs that worked with SMP on must also have had TLB huge pages on and I didn't realize it.

I've since fixed the issue, and moved to ext2 because I don't need journaling.