metalsvm/mm/memory.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/mmu.h>
#include <metalsvm/spinlock.h>
#include <metalsvm/time.h>
#include <metalsvm/elf.h>
#include <metalsvm/processor.h>
#include <metalsvm/page.h>
#ifdef CONFIG_MULTIBOOT
#include <asm/multiboot.h>
#endif
#ifdef CONFIG_ROCKCREEK
#include <asm/scc.h>
#endif

/* 
 * 0 => free
 * 1 => occupied
 * 
 * Set whole address space as occupied
 */
static uint8_t	 	bitmap[BITMAP_SIZE] = {[0 ... BITMAP_SIZE-1] = 0xFF};
static spinlock_t	bitmap_lock = SPINLOCK_INIT;
static size_t		alloc_start;
atomic_int32_t		total_pages = ATOMIC_INIT(0);
atomic_int32_t		total_allocated_pages = ATOMIC_INIT(0);
atomic_int32_t		total_available_pages = ATOMIC_INIT(0);
atomic_int32_t		total_kernel_pages = ATOMIC_INIT(0);

/* 
 * Note that linker symbols are not variables, they have no memory allocated for
 * maintaining a value, rather their address is their value.
 */
extern const void kernel_start;
extern const void kernel_end;

inline static int page_marked(unsigned int i)
{
	unsigned int index = i / 8;
	unsigned int mod = i % 8;

	return  (bitmap[index] & (1 << mod));
}

inline static int page_unmarked(unsigned int i)
{
	return !page_marked(i);
}

inline static void page_set_mark(unsigned int i)
{
	unsigned int index = i / 8;
	unsigned int mod = i % 8;

	//if (page_marked(i))
	//	kprintf("page %u is alread marked\n", i);

        bitmap[index] = bitmap[index] | (1 << mod);
}

inline static void page_clear_mark(unsigned int i)
{
	unsigned int index = i / 8;
	unsigned int mod = i % 8;

	if (page_unmarked(i))
		kprintf("page %u is already unmarked\n", i);

	bitmap[index] = bitmap[index] & ~(1 << mod);
}

int mmu_init(void)
{
	size_t kernel_size;
	unsigned int i; 
	size_t addr, end_addr;

#ifdef CONFIG_MULTIBOOT
	if (mb_info && (mb_info->flags & (1 << 6))) {
		multiboot_memory_map_t* mmap = (multiboot_memory_map_t*) mb_info->mmap_addr;
		multiboot_memory_map_t* mmap_end = (void*) ((size_t) mb_info->mmap_addr + mb_info->mmap_length);

		while (mmap < mmap_end) {
			if (mmap->type == MULTIBOOT_MEMORY_AVAILABLE) {
				/* set the available memory as "unused" */
				addr = mmap->addr;
				end_addr = addr + mmap->len;
 
				while (addr < end_addr) {
					page_clear_mark(addr / PAGE_SIZE);
					addr += PAGE_SIZE;
					atomic_int32_inc(&total_pages);
					atomic_int32_inc(&total_available_pages);
				}
			}
			mmap++;
		}
	} else {
		kputs("Unable to initialize the memory management subsystem\n");
		while(1) { 
			NOP8;
		}
	}

	/* 
	 * Modules like the init ram disk are already loaded.
	 * Therefore, we set these pages as used.
	 */
	if (mb_info && (mb_info->flags & (1 << 3))) {
		multiboot_module_t* mmodule = (multiboot_module_t*) mb_info->mods_addr;

		for(i=0; i<mb_info->mods_count; i++, mmodule++) {
			for(addr=mmodule->mod_start; addr<mmodule->mod_end; addr+=PAGE_SIZE) {
				page_set_mark(addr / PAGE_SIZE);
				atomic_int32_inc(&total_allocated_pages);
				atomic_int32_dec(&total_available_pages);
			}
		}
	} 

#elif defined(CONFIG_ROCKCREEK)
	for(i=0; i<SCC_PMEM_REGIONS; i++) {
		addr = scc_info.private_mem[i].low;
		end_addr = scc_info.private_mem[i].high;

		while(addr < end_addr) {
			page_clear_mark(addr / PAGE_SIZE);
			if (addr < addr + PAGE_SIZE)
				break;
			addr += PAGE_SIZE;
			atomic_int32_inc(&total_pages);
			atomic_int32_inc(&total_available_pages);
		}
	}
#else
	#error Currently, MetalSVM supports only the Multiboot specification or the RockCreek processor!
#endif

	kernel_size = (size_t) &kernel_end - (size_t) &kernel_start;
	if (kernel_size % PAGE_SIZE)
		kernel_size += PAGE_SIZE - kernel_size % PAGE_SIZE;
	atomic_int32_add(&total_allocated_pages, kernel_size/PAGE_SIZE);
	atomic_int32_sub(&total_available_pages, kernel_size/PAGE_SIZE);

	/* set kernel space as used */
	for(i=(size_t) &kernel_start / PAGE_SIZE; i < (size_t) &kernel_end / PAGE_SIZE; i++)
		page_set_mark(i);
	if ((size_t) &kernel_end % PAGE_SIZE)
		page_set_mark(i);

	alloc_start = (size_t) &kernel_end / PAGE_SIZE;	
	if ((size_t) &kernel_end % PAGE_SIZE)
		alloc_start++;

	return paging_init();
}

/*
 * Use first fit algorithm to find a suitable physical memory region
 */
static size_t task_get_pages(task_t* task, uint32_t npages)
{
	uint32_t i, j, l;
	uint32_t k = 0;
	size_t ret = 0;

	if (BUILTIN_EXPECT(!npages, 0))
		return ret;

	spinlock_lock(&bitmap_lock);
	i = alloc_start;
next_try:
	while((k < BITMAP_SIZE) && page_marked(i)) {
		k++;
		i = (i+1) % BITMAP_SIZE;
	}

	if (k >= BITMAP_SIZE)
		goto oom;

	for(j=1; (j<npages) && (i+j < BITMAP_SIZE) && (k < BITMAP_SIZE); j++, k++) {
		if (page_marked(i+j)) {
			i = (i+j) % BITMAP_SIZE;
			goto next_try;
		}
	}

	if (i+j >= BITMAP_SIZE) {
		i = 0;
		goto next_try;
	}

	if (k >= BITMAP_SIZE)
		goto oom;

	ret = i*PAGE_SIZE;
	//kprintf("alloc: ret 0x%x, i = %d, j = %d, npages = %d\n", ret, i, j, npages);
	for(l=i; l<i+j; l++)
		page_set_mark(l);

	alloc_start = i+j;
	spinlock_unlock(&bitmap_lock);

	atomic_int32_add(&total_allocated_pages, npages);
	atomic_int32_sub(&total_available_pages, npages);
	if (task && task->ustack)
		atomic_int32_add(&(task->mem_usage), npages);
	else
		atomic_int32_add(&total_kernel_pages, npages);

	return ret;

oom:
	spinlock_unlock(&bitmap_lock);

	return ret;
}

size_t get_pages(uint32_t npages)
{
	return task_get_pages(per_core(current_task), npages);
}

void* mem_allocation(size_t sz, uint32_t flags)
{
	size_t phyaddr, viraddr;
	uint32_t npages = sz / PAGE_SIZE;
	task_t* task = per_core(current_task);

	if (sz % PAGE_SIZE)
		npages++;

	phyaddr = task_get_pages(task, npages);
	if (BUILTIN_EXPECT(!phyaddr, 0))
		return 0;

	spinlock_lock(task->pgd_lock);
	viraddr = map_region(task, 0, phyaddr, npages, flags);
	spinlock_unlock(task->pgd_lock);

	return (void*) viraddr;
}

void* kmalloc(size_t sz)
{
	uint32_t flags;
	task_t* task = per_core(current_task);

	if (task->ustack)
		flags = MAP_USER_SPACE|MAP_HEAP;
	else
		flags = MAP_KERNEL_SPACE|MAP_HEAP;

	return mem_allocation(sz, flags);
}

void* create_stack(task_t* task, size_t sz)
{
	size_t viraddr;
	uint32_t npages = sz / PAGE_SIZE;

	if (sz % PAGE_SIZE)
		npages++;

	spinlock_lock(task->pgd_lock);

	size_t addr = (size_t) task_get_pages(task, npages);
	if (BUILTIN_EXPECT(!addr, 0))
		return 0;

	spinlock_lock(task->pgd_lock);
	/*
	 * We need only user-level stacks. Kernel stacks already initialized as
	 * static array.
	 */
	viraddr = map_region(task, 0, addr, npages, MAP_USER_SPACE|MAP_STACK);
	spinlock_unlock(task->pgd_lock);

	memset((unsigned char*)viraddr, 0xCD, sz);

	return (void*) viraddr;
}

static void task_free(task_t* task, void* addr, size_t sz)
{
	uint32_t index, npages, i;
	size_t phyaddr;

	if (BUILTIN_EXPECT(!addr && !sz, 0))
		return;

        npages = sz / PAGE_SIZE;
	if (sz % PAGE_SIZE)
		npages++;


	spinlock_lock(task->pgd_lock); 
	vm_free(task, (size_t) addr, npages);
	spinlock_unlock(task->pgd_lock);

	spinlock_lock(&bitmap_lock);
	for(i=0; i<npages; i++) {
		phyaddr = virt_to_phys(task, (size_t) addr+i*PAGE_SIZE);
		index = phyaddr / PAGE_SIZE;
		page_unmarked(index);
	}
	spinlock_unlock(&bitmap_lock);

	atomic_int32_sub(&total_allocated_pages, npages);
	atomic_int32_add(&total_available_pages, npages);
	if (task && task->ustack)
		atomic_int32_sub(&(task->mem_usage), npages);
	else
		atomic_int32_sub(&total_kernel_pages, npages);
}

void kfree(void* addr, size_t sz)
{
	task_free(per_core(current_task), addr, sz);
}