Quick'n'dirty user-level switching

August 25, 2020

My goal for the last couple of days has been getting my hobby OS project to the point where it can switch into user-mode. I was technically at that point yesterday, but with an unsatisfying caveat: the instant the processor entered user-mode (by means of the sysret instruction) it would page fault, and for a reasons I still don’t understand, my interrupt handler wasn’t invoked.

Today I managed to get user-level to (deliberately) infinitely loop. I can use a debugger to verify that the processor is in-fact in user-mode. To prevent the fault upon entering user-mode, I adjusted some access flags in the paging hierarchy to permit the processor to access the memory containing the user program and stack, while the processor is in user-mode.

I was wrong about the bootloader’s ELF-loading. I incorrectly assumed that it would set up paging to match the virtual addresses specified in an ELF-file, but it only loads code/data into a specific region of virtual memory, and sets up paging for that region. This means in order to have the user program be loaded at its expected virtual address, I need to identify unused physical memory and map some of it at this address.

In the interest of just getting something simple working, I’m currently setting access bits such that user-mode has access to some code from the kernel, and the kernel’s stack. That way I can just write a while(1); function in the kernel code, and run the sysret instruction such that it “returns” to user-mode at the beginning of this function.