metalsvm/arch/x86/kernel/gdt.c

/* 
 * Copyright 2010 Stefan Lankes, Chair for Operating Systems,
 *                               RWTH Aachen University
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 *
 * This file is part of MetalSVM.
 */

#include <metalsvm/stdio.h>
#include <metalsvm/string.h>
#include <metalsvm/stdlib.h>
#include <metalsvm/tasks.h>
#include <metalsvm/errno.h>
#include <metalsvm/processor.h>
#include <asm/gdt.h>
#include <asm/tss.h>
#include <asm/page.h>

gdt_ptr_t				gp;
static tss_t				task_state_segments[MAX_CORES] __attribute__ ((aligned (PAGE_SIZE)));
// currently, our kernel has full access to the ioports
static gdt_entry_t			gdt[GDT_ENTRIES] = {[0 ... GDT_ENTRIES-1] = {0, 0, 0, 0, 0, 0}};

/* 
 * This is defined in entry.asm. We use this to properly reload
 * the new segment registers
 */
extern void gdt_flush(void);

size_t* get_current_stack(void)
{
	task_t* curr_task = per_core(current_task);

	// determine and set esp0
#ifdef CONFIG_X86_32
	task_state_segments[CORE_ID].esp0 = (size_t) curr_task->stack + KERNEL_STACK_SIZE - 16; // => stack is 16byte aligned 
#else
	task_state_segments[CORE_ID].rsp0 = (size_t) curr_task->stack + KERNEL_STACK_SIZE - 16; // => stack is 16byte aligned 
#endif

	// use new page table
	write_cr3(virt_to_phys((size_t)curr_task->pgd));

	return curr_task->last_stack_pointer;
}

int arch_fork(task_t* task)
{
	struct state* state;
	task_t* curr_task = per_core(current_task);
	size_t esp, state_size;

	if (BUILTIN_EXPECT(!task, 0))
		return -EINVAL;

	if (BUILTIN_EXPECT(!task->stack, 0))
		return -EINVAL;

#ifdef CONFIG_X86_32
	state_size = sizeof(struct state) - 2*sizeof(size_t);
#else
	state_size = sizeof(struct state);
#endif

	// copy kernel stack of the current task
	mb();
	memcpy(task->stack, curr_task->stack, KERNEL_STACK_SIZE);

#ifdef CONFIG_X86_32
	asm volatile ("mov %%esp, %0" : "=m"(esp));
	esp -= (size_t) curr_task->stack;
	esp += (size_t) task->stack; 

	state = (struct state*) (esp - state_size);
	//memset(state, 0x00, state_size);

	asm volatile ("pusha; pop %0" : "=m"(state->edi));
	asm volatile ("pop %0" : "=m"(state->esi));
	asm volatile ("pop %0" : "=m"(state->ebp));
	asm volatile ("add $4, %%esp" ::: "%esp");
	asm volatile ("pop %0" : "=m"(state->ebx));
	asm volatile ("pop %0" : "=m"(state->edx));
	asm volatile ("pop %0" : "=m"(state->ecx));
	asm volatile ("pop %0" : "=m"(state->eax));

	state->esp = esp;
	task->last_stack_pointer = (size_t*) state;
	state->int_no = 0xB16B00B5;
	state->error =  0xC03DB4B3;
	state->cs = 0x08;
	state->ds = state->es = 0x10;
	// store the current EFLAGS
	asm volatile ("pushf; pop %0" : "=m"(state->eflags));
	// enable interrupts
	state->eflags |= (1 << 9);
	// This will be the entry point for the new task. read_ip cleanups the stack
	asm volatile ("push %0; call read_ip" :: "r"(&state->eip) : "%eax");
#else
#warning Currently, not supported!
	return -1;
#endif

	return 0;
}

int create_default_frame(task_t* task, entry_point_t ep, void* arg)
{
	size_t *stack;
	struct state *stptr;
	size_t state_size;

	if (BUILTIN_EXPECT(!task, 0))
		return -EINVAL;	

	if (BUILTIN_EXPECT(!task->stack, 0))
		return -EINVAL;

	memset(task->stack, 0xCD, KERNEL_STACK_SIZE);

	/* The difference between setting up a task for SW-task-switching
	 * and not for HW-task-switching is setting up a stack and not a TSS.
	 * This is the stack which will be activated and popped off for iret later.
	 */
	stack = (size_t*) (task->stack + KERNEL_STACK_SIZE - 16); // => stack is 16byte aligned 

	/* The next three things on the stack are a marker for debugging purposes, ... */
	*stack-- = 0xDEADBEEF;
#ifdef CONFIG_X86_32
	/* the first-function-to-be-called's arguments, ... */
	*stack-- = (size_t) arg;
#endif
	/* and the "caller" we shall return to.
	 * This procedure cleans the task after exit. */
	*stack = (size_t) leave_kernel_task;

	/* Next bunch on the stack is the initial register state. 
	 * The stack must look like the stack of a task which was
	 * scheduled away previously. */

	/* In 64bit mode, he stack pointer (SS:RSP) is pushed unconditionally on interrupts.
	 * In legacy modes, this push is conditional and based on a change in current privilege level (CPL).*/
#ifdef CONFIG_X86_32
	state_size = sizeof(struct state) - 2*sizeof(size_t);
#else
	state_size = sizeof(struct state);
#endif
	stack = (size_t*) ((size_t) stack - state_size);

	stptr = (struct state *) stack;
	memset(stptr, 0x00, state_size);
#ifdef CONFIG_X86_32
	stptr->esp = (size_t)stack + state_size;
#else
	stptr->rsp = (size_t)stack + state_size;
	/* the first-function-to-be-called's arguments, ... */
	stptr->rdi = (size_t) arg;
#endif
	stptr->int_no = 0xB16B00B5;
	stptr->error =  0xC03DB4B3;

	/* The instruction pointer shall be set on the first function to be called
	 * after IRETing */
#ifdef CONFIG_X86_32
	stptr->eip = (size_t)ep;
#else
	stptr->rip = (size_t)ep;
#endif
	stptr->cs = 0x08;
#ifdef CONFIG_X86_32
	stptr->eflags = 0x1202;
	stptr->ds = stptr->es = 0x10;
	// the creation of a kernel tasks didn't change the IOPL level
	// => useresp & ss is not required
#else
	stptr->rflags = 0x1202;
	stptr->ss = 0x10;
	stptr->userrsp = stptr->rsp;
#endif

	/* Set the task's stack pointer entry to the stack we have crafted right now. */
	task->last_stack_pointer = (size_t*)stack;

	return 0;
}

/* Setup a descriptor in the Global Descriptor Table */
static void gdt_set_gate(int num, unsigned long base, unsigned long limit,
			  unsigned char access, unsigned char gran)
{
	configure_gdt_entry(&gdt[num], base, limit, access, gran);
}

void configure_gdt_entry(gdt_entry_t *dest_entry, unsigned long base, unsigned long limit,
		unsigned char access, unsigned char gran)
{
	/* Setup the descriptor base address */
	dest_entry->base_low = (base & 0xFFFF);
	dest_entry->base_middle = (base >> 16) & 0xFF;
	dest_entry->base_high = (base >> 24) & 0xFF;

	/* Setup the descriptor limits */
	dest_entry->limit_low = (limit & 0xFFFF);
	dest_entry->granularity = ((limit >> 16) & 0x0F);

	/* Finally, set up the granularity and access flags */
	dest_entry->granularity |= (gran & 0xF0);
	dest_entry->access = access;
}

/* 
 * This will setup the special GDT
 * pointer, set up the entries in our GDT, and then
 * finally call gdt_flush() in our assembler file in order
 * to tell the processor where the new GDT is and update the
 * new segment registers 
 */
void gdt_install(void)
{
	unsigned int i;
	unsigned long mode, limit;

	memset(task_state_segments, 0x00, MAX_CORES*sizeof(tss_t));

#ifdef CONFIG_X86_32
	mode = GDT_FLAG_32_BIT;
	limit = 0xFFFFFFFF;
#elif defined(CONFIG_X86_64)
	mode = GDT_FLAG_64_BIT;
	limit = 0;
#else
#error invalid mode
#endif

	/* Setup the GDT pointer and limit */
	gp.limit = (sizeof(gdt_entry_t) * GDT_ENTRIES) - 1;
	gp.base = (size_t) &gdt;

	/* Our NULL descriptor */
	gdt_set_gate(0, 0, 0, 0, 0);

	/* 
	 * The second entry is our Code Segment. The base address
	 * is 0, the limit is 4 GByte, it uses 4KByte granularity,
	 * uses 32-bit opcodes, and is a Code Segment descriptor.
	 */
	gdt_set_gate(1, 0, limit,
		GDT_FLAG_RING0 | GDT_FLAG_SEGMENT | GDT_FLAG_CODESEG | GDT_FLAG_PRESENT,
		GDT_FLAG_4K_GRAN | mode);

	/* 
	 * The third entry is our Data Segment. It's EXACTLY the
	 * same as our code segment, but the descriptor type in
	 * this entry's access byte says it's a Data Segment 
	 */
	gdt_set_gate(2, 0, limit,
		GDT_FLAG_RING0 | GDT_FLAG_SEGMENT | GDT_FLAG_DATASEG | GDT_FLAG_PRESENT,
		GDT_FLAG_4K_GRAN | mode);
		
	/*
	 * Create code segement for userspace applications (ring 3)
	 */
	gdt_set_gate(3, 0, limit,
		GDT_FLAG_RING3 | GDT_FLAG_SEGMENT | GDT_FLAG_CODESEG | GDT_FLAG_PRESENT,
		GDT_FLAG_4K_GRAN | mode);

	/*
	 * Create data segement for userspace applications (ring 3)
	 */
	gdt_set_gate(4, 0, limit,
                GDT_FLAG_RING3 | GDT_FLAG_SEGMENT | GDT_FLAG_DATASEG | GDT_FLAG_PRESENT,
                GDT_FLAG_4K_GRAN | mode);

	/*
	 * Create TSS for each task at ring0 (we use these segments for task switching)
	 */
	for(i=0; i<MAX_CORES; i++) {
#ifdef CONFIG_X86_32
		/* set default values */
		task_state_segments[i].eflags = 0x1202;
		task_state_segments[i].ss0 = 0x10;		// data segment
		task_state_segments[i].esp0 = 0xDEADBEEF;	// invalid pseudo address
		gdt_set_gate(5+i, (unsigned long) (task_state_segments+i), sizeof(tss_t)-1,
			GDT_FLAG_PRESENT | GDT_FLAG_TSS | GDT_FLAG_RING0, mode);
#elif defined(CONFIG_X86_64)
		task_state_segments[i].rsp0 = 0xDEADBEEF;       // invalid pseudo address
		gdt_set_gate(5+i*2, (unsigned long) (task_state_segments+i), sizeof(tss_t)-1,
			GDT_FLAG_PRESENT | GDT_FLAG_TSS | GDT_FLAG_RING0, mode);
#endif
	}

	/* Flush out the old GDT and install the new changes! */
	gdt_flush();
}