OSDev.org

rizxt wrote:So, I know in order for me to return to ring 0 from ring 3, I need a TSS.

I don't plan on doing that quite yet, I just want to get into userspace.

rizxt wrote:I know I need entries in my GDT for user code/data. I don't know what the entries should look like, or differ from the kernel entries, or what order the GDT needs to be laid out in.

[GLOBAL Gdt64]
Gdt64:
	; Invalid segment
	DQ 0x0000000000000000 ; 0x0
	; Kernel code: RW, executable, code/data segment, present, 64-bit, ring 0
	DQ 0x00209A0000000000 ; 0x8
	; Kernel data: RW, data, code/data segment, present, ring 0
	DQ 0x0000920000000000 ; 0x10
	; User data: RW, data, code/data segment, present, ring 3
	DQ 0x0020F20000000000 ; 0x18
	; User code: RW, executable, code/data segment, present, 64-bit, ring 3
	DQ 0x0020FA0000000000 ; 0x20
[GLOBAL TSSEntry]
TSSEntry:
	DQ 0x0000000000000000 ; 0x28
	DQ 0x0000000000000000

rizxt wrote:And then finally, what do I do to officiate my entry into userspace?

AndrewAPrice wrote:

rizxt wrote:So, I know in order for me to return to ring 0 from ring 3, I need a TSS.

I don't plan on doing that quite yet, I just want to get into userspace.

You probably want a way to syscall back into the kernel unless you map 0xb8000 to usermode. Are you in protected mode or long mode?

I found the TSS straight foward to work with in long mode. The most difficult part is splitting up the TSS's address + size to write it into the GDT. The only thing you'd need in the actual TSS right now is the stack pointer to use for when you enter an interrupt. But, I don't know if you actually need a TSS if you keep interrupts disabled.

rizxt wrote:I know I need entries in my GDT for user code/data. I don't know what the entries should look like, or differ from the kernel entries, or what order the GDT needs to be laid out in.

This is for long mode. I have 5 segments in my GDT (kernel + userspace, code + data), and a reference to the TSS.

Code: Select all

[GLOBAL Gdt64]
Gdt64:
	; Invalid segment
	DQ 0x0000000000000000 ; 0x0
	; Kernel code: RW, executable, code/data segment, present, 64-bit, ring 0
	DQ 0x00209A0000000000 ; 0x8
	; Kernel data: RW, data, code/data segment, present, ring 0
	DQ 0x0000920000000000 ; 0x10
	; User data: RW, data, code/data segment, present, ring 3
	DQ 0x0020F20000000000 ; 0x18
	; User code: RW, executable, code/data segment, present, 64-bit, ring 3
	DQ 0x0020FA0000000000 ; 0x20
[GLOBAL TSSEntry]
TSSEntry:
	DQ 0x0000000000000000 ; 0x28
	DQ 0x0000000000000000

rizxt wrote:And then finally, what do I do to officiate my entry into userspace?

Set your segments, then jump into it. See Getting to Ring 3#Entering_Ring_3

rizxt wrote:Long mode. Also, reading that article, I noticed it uses IRET. The difference between x86 and x86_64 IRET is important, I believe. And it appears that article is written for x86.

JumpToUserMode:

    # rdi = user args
    # rsi = entry point in user space
    # rdx = user space stack

    movq $0x1b, %rax    # Selector 0x18 (User Data) + RPL 3
    movw %ax, %ds
    movw %ax, %es

    # Build a fake iret frame
    pushq %rax          # Selector 0x18 (User Data) + RPL 3
    pushq %rdx          # User space stack
    pushq $0x202        # rflags = interrupt enable + reserved bit
    pushq $0x23         # Selector 0x20 (User Code) + RPL 3
    pushq %rsi          # Entry point in user space

    iretq

foliagecanine wrote:Just curious, is this the kernel, a program, or test code that you are trying to execute in userspace?

rizxt wrote:

foliagecanine wrote:Just curious, is this the kernel, a program, or test code that you are trying to execute in userspace?

This is part of the kernel executable. I'm trying to get to userspace so I can implement syscalls + test them, and also implement a UI outside of ring 0.