Linux内核页表初始化

Linux在内核启动过程中start_kernel->setup_arch会调用如下两个函数对页表进行初始化和建立。

static inline void prepare_page_table(void)
{
    unsigned long addr;

    /*
     * Clear out all the mappings below the kernel image.
     */
    //初始化0~MODULES_VADDR（0xBF000000）地址空间的页表,0xBF000000~0xC0000000的16M空间用来存放kernel module library
    //关于PGDIR_SIZE=2M和pmd_clear下面单独叙述
    for (addr = 0; addr < MODULES_VADDR; addr += PGDIR_SIZE)
        pmd_clear(pmd_off_k(addr));

#ifdef CONFIG_XIP_KERNEL
    //XIP内核（即kernel image存放在Nor flash等可偏上执行的存储体中）存放在kernel module area，不能影响kernel image所在的地址空间的页表，etext为kernel代码段结束地址，加一页，应该是为kernel数据段保留足够的空间
    /* The XIP kernel is mapped in the module area -- skip over it */
    addr = ((unsigned long)_etext + PGDIR_SIZE - 1) & PGDIR_MASK;
#endif
    for ( ; addr < PAGE_OFFSET; addr += PGDIR_SIZE)//否则kernel module在使用时才动态加载，现在需要初始化该区域的页表
        pmd_clear(pmd_off_k(addr));

    /*
     * Clear out all the kernel space mappings, except for the first
     * memory bank, up to the end of the vmalloc region.
     */
    //由于第一个Bank存储着kernel image，这段空间的页表在head.S已经建立完成，以后无论什么动作都不能影响这段页表
    for (addr = __phys_to_virt(bank_phys_end(&meminfo.bank[0]));
         addr < VMALLOC_END; addr += PGDIR_SIZE)
        pmd_clear(pmd_off_k(addr));
}

由于ARM采用两级映射，pmd不占用字段，pmd=pgd，而这里一次清了两个pgd，和pgd的定义正好对应，所以在for循环时每次步长为PGDIR_SIZE（2M），根本的原因是ARM linux使用了ARM的两个段，在用户态为__USER_CS__USER_DS，在内核态为__KERNEL_CS和__KERNEL_DS

初始化页表里也会flush对应的TLB。

typedef unsigned long pgd_t[2];

#define pmd_clear(pmdp)			\
	do {				\
		pmdp[0] = __pmd(0);	\
		pmdp[1] = __pmd(0);	\
		clean_pmd_entry(pmdp);	\
	} while (0)

static inline pmd_t *pmd_off_k(unsigned long virt)
{
	return pmd_off(pgd_offset_k(virt), virt);
}

#define pgd_offset_k(addr)	pgd_offset(&init_mm, addr)

/* to find an entry in a page-table-directory */
#define pgd_index(addr)		((addr) >> PGDIR_SHIFT)

#define pgd_offset(mm, addr)	((mm)->pgd+pgd_index(addr))

/* to find an entry in a kernel page-table-directory */
#define pgd_offset_k(addr)	pgd_offset(&init_mm, addr)

/* Find an entry in the second-level page table.. */
#define pmd_offset(dir, addr)	((pmd_t *)(dir))

/* Find an entry in the third-level page table.. */
#define __pte_index(addr)	(((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))

map_lowmem函数会为地段物理内存的每个bank逐一调用map_memory_bank做页表映射，再调用create_mapping建立页表

/*
 * Create the page directory entries and any necessary
 * page tables for the mapping specified by `md'.  We
 * are able to cope here with varying sizes and address
 * offsets, and we take full advantage of sections and
 * supersections.
 */
static void __init create_mapping(struct map_desc *md)
{
	unsigned long phys, addr, length, end;
	const struct mem_type *type;
	pgd_t *pgd;
        //在内核为个bank建立页表时，它虚拟地址不可能落在user空间
	if (md->virtual != vectors_base() && md->virtual < TASK_SIZE) {
		printk(KERN_WARNING "BUG: not creating mapping for "
		       "0x%08llx at 0x%08lx in user region\n",
		       __pfn_to_phys((u64)md->pfn), md->virtual);
		return;
	}
//高端内存通过alloc_page+kmap的形式映射，在初始化阶段不可能有bank的虚拟地址落在VMALLOC线性地址空间
	if ((md->type == MT_DEVICE || md->type == MT_ROM) &&
	    md->virtual >= PAGE_OFFSET && md->virtual < VMALLOC_END) {
		printk(KERN_WARNING "BUG: mapping for 0x%08llx at 0x%08lx "
		       "overlaps vmalloc space\n",
		       __pfn_to_phys((u64)md->pfn), md->virtual);
	}

	type = &mem_types[md->type];
//如果有大于4G的物理内存，单独建立
	/*
	 * Catch 36-bit addresses
	 */
	if (md->pfn >= 0x100000) {
		create_36bit_mapping(md, type);
		return;
	}

	addr = md->virtual & PAGE_MASK;
	phys = (unsigned long)__pfn_to_phys(md->pfn);
	length = PAGE_ALIGN(md->length + (md->virtual & ~PAGE_MASK));//将bank整页对齐

	if (type->prot_l1 == 0 && ((addr | phys | length) & ~SECTION_MASK)) {//不能section(1M)对齐的
		printk(KERN_WARNING "BUG: map for 0x%08lx at 0x%08lx can not "
		       "be mapped using pages, ignoring.\n",
		       __pfn_to_phys(md->pfn), addr);
		return;
	}

	pgd = pgd_offset_k(addr);
	end = addr + length;
	do {
		unsigned long next = pgd_addr_end(addr, end);

		alloc_init_section(pgd, addr, next, phys, type);//真正建立页表的函数

		phys += next - addr;
		addr = next;
	} while (pgd++, addr != end);
}

static void __init alloc_init_section(pgd_t *pgd, unsigned long addr,
				      unsigned long end, unsigned long phys,
				      const struct mem_type *type)
{
	pmd_t *pmd = pmd_offset(pgd, addr);

	/*
	 * Try a section mapping - end, addr and phys must all be aligned
	 * to a section boundary.  Note that PMDs refer to the individual
	 * L1 entries, whereas PGDs refer to a group of L1 entries making
	 * up one logical pointer to an L2 table.
	 */
	if (((addr | end | phys) & ~SECTION_MASK) == 0) {//完整的一个section（1M）直接生成段页表
		pmd_t *p = pmd;

		if (addr & SECTION_SIZE)
			pmd++;

		do {
			*pmd = __pmd(phys | type->prot_sect);//填充段页表地址和section相关属性
			phys += SECTION_SIZE;
		} while (pmd++, addr += SECTION_SIZE, addr != end);

		flush_pmd_entry(p);
	} else {//不是段对齐的，要分配二级页表
		/*
		 * No need to loop; pte's aren't interested in the
		 * individual L1 entries.
		 */
		alloc_init_pte(pmd, addr, end, __phys_to_pfn(phys), type);
	}
}

static void __init alloc_init_pte(pmd_t *pmd, unsigned long addr,
                  unsigned long end, unsigned long pfn,
                  const struct mem_type *type)
{
    pte_t *pte;

    if (pmd_none(*pmd)) {//check pmd指向的L1页表中的页表项是否存在，不存在的话使用Bootmem分配所需的二级页表空间
        pte = alloc_bootmem_low_pages(2 * PTRS_PER_PTE * sizeof(pte_t));
        __pmd_populate(pmd, __pa(pte) | type->prot_l1);//填充二级页表，即pte的物理地址和prot_l1
    }

    pte = pte_offset_kernel(pmd, addr);//建立pte
    do {
        set_pte_ext(pte, pfn_pte(pfn, __pgprot(type->prot_pte)), 0);//set_pte_ext与平台相关，它完成硬件页表和内核页表两者的创建
        pfn++;
    } while (pte++, addr += PAGE_SIZE, addr != end);}