memory.c

LINUX1.0源代码,代码条理清晰

	return 0;
}

void do_no_page(unsigned long error_code, unsigned long address,
	struct task_struct *tsk, unsigned long user_esp)
{
	unsigned long tmp;
	unsigned long page;
	struct vm_area_struct * mpnt;

	page = get_empty_pgtable(tsk,address);
	if (!page)
		return;
	page &= PAGE_MASK;
	page += PAGE_PTR(address);
	tmp = *(unsigned long *) page;
	if (tmp & PAGE_PRESENT)
		return;
	++tsk->rss;
	if (tmp) {
		++tsk->maj_flt;
		swap_in((unsigned long *) page);
		return;
	}
	address &= 0xfffff000;
	tmp = 0;
	for (mpnt = tsk->mmap; mpnt != NULL; mpnt = mpnt->vm_next) {
		if (address < mpnt->vm_start)
			break;
		if (address >= mpnt->vm_end) {
			tmp = mpnt->vm_end;
			continue;
		}
		if (!mpnt->vm_ops || !mpnt->vm_ops->nopage) {
			++tsk->min_flt;
			get_empty_page(tsk,address);
			return;
		}
		mpnt->vm_ops->nopage(error_code, mpnt, address);
		return;
	}
	if (tsk != current)
		goto ok_no_page;
	if (address >= tsk->end_data && address < tsk->brk)
		goto ok_no_page;
	if (mpnt && mpnt == tsk->stk_vma &&
	    address - tmp > mpnt->vm_start - address &&
	    tsk->rlim[RLIMIT_STACK].rlim_cur > mpnt->vm_end - address) {
		mpnt->vm_start = address;
		goto ok_no_page;
	}
	tsk->tss.cr2 = address;
	current->tss.error_code = error_code;
	current->tss.trap_no = 14;
	send_sig(SIGSEGV,tsk,1);
	if (error_code & 4)	/* user level access? */
		return;
ok_no_page:
	++tsk->min_flt;
	get_empty_page(tsk,address);
}

/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
 */
asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long error_code)
{
	unsigned long address;
	unsigned long user_esp = 0;
	unsigned int bit;

	/* get the address */
	__asm__("movl %%cr2,%0":"=r" (address));
	if (address < TASK_SIZE) {
		if (error_code & 4) {	/* user mode access? */
			if (regs->eflags & VM_MASK) {
				bit = (address - 0xA0000) >> PAGE_SHIFT;
				if (bit < 32)
					current->screen_bitmap |= 1 << bit;
			} else 
				user_esp = regs->esp;
		}
		if (error_code & 1)
			do_wp_page(error_code, address, current, user_esp);
		else
			do_no_page(error_code, address, current, user_esp);
		return;
	}
	address -= TASK_SIZE;
	if (wp_works_ok < 0 && address == 0 && (error_code & PAGE_PRESENT)) {
		wp_works_ok = 1;
		pg0[0] = PAGE_SHARED;
		printk("This processor honours the WP bit even when in supervisor mode. Good.\n");
		return;
	}
	if (address < PAGE_SIZE) {
		printk("Unable to handle kernel NULL pointer dereference");
		pg0[0] = PAGE_SHARED;
	} else
		printk("Unable to handle kernel paging request");
	printk(" at address %08lx\n",address);
	die_if_kernel("Oops", regs, error_code);
	do_exit(SIGKILL);
}

/*
 * BAD_PAGE is the page that is used for page faults when linux
 * is out-of-memory. Older versions of linux just did a
 * do_exit(), but using this instead means there is less risk
 * for a process dying in kernel mode, possibly leaving a inode
 * unused etc..
 *
 * BAD_PAGETABLE is the accompanying page-table: it is initialized
 * to point to BAD_PAGE entries.
 *
 * ZERO_PAGE is a special page that is used for zero-initialized
 * data and COW.
 */
unsigned long __bad_pagetable(void)
{
	extern char empty_bad_page_table[PAGE_SIZE];

	__asm__ __volatile__("cld ; rep ; stosl":
		:"a" (BAD_PAGE + PAGE_TABLE),
		 "D" ((long) empty_bad_page_table),
		 "c" (PTRS_PER_PAGE)
		:"di","cx");
	return (unsigned long) empty_bad_page_table;
}

unsigned long __bad_page(void)
{
	extern char empty_bad_page[PAGE_SIZE];

	__asm__ __volatile__("cld ; rep ; stosl":
		:"a" (0),
		 "D" ((long) empty_bad_page),
		 "c" (PTRS_PER_PAGE)
		:"di","cx");
	return (unsigned long) empty_bad_page;
}

unsigned long __zero_page(void)
{
	extern char empty_zero_page[PAGE_SIZE];

	__asm__ __volatile__("cld ; rep ; stosl":
		:"a" (0),
		 "D" ((long) empty_zero_page),
		 "c" (PTRS_PER_PAGE)
		:"di","cx");
	return (unsigned long) empty_zero_page;
}

void show_mem(void)
{
	int i,free = 0,total = 0,reserved = 0;
	int shared = 0;

	printk("Mem-info:\n");
	printk("Free pages:      %6dkB\n",nr_free_pages<<(PAGE_SHIFT-10));
	printk("Secondary pages: %6dkB\n",nr_secondary_pages<<(PAGE_SHIFT-10));
	printk("Free swap:       %6dkB\n",nr_swap_pages<<(PAGE_SHIFT-10));
	i = high_memory >> PAGE_SHIFT;
	while (i-- > 0) {
		total++;
		if (mem_map[i] & MAP_PAGE_RESERVED)
			reserved++;
		else if (!mem_map[i])
			free++;
		else
			shared += mem_map[i]-1;
	}
	printk("%d pages of RAM\n",total);
	printk("%d free pages\n",free);
	printk("%d reserved pages\n",reserved);
	printk("%d pages shared\n",shared);
	show_buffers();
}

/*
 * paging_init() sets up the page tables - note that the first 4MB are
 * already mapped by head.S.
 *
 * This routines also unmaps the page at virtual kernel address 0, so
 * that we can trap those pesky NULL-reference errors in the kernel.
 */
unsigned long paging_init(unsigned long start_mem, unsigned long end_mem)
{
	unsigned long * pg_dir;
	unsigned long * pg_table;
	unsigned long tmp;
	unsigned long address;

/*
 * Physical page 0 is special; it's not touched by Linux since BIOS
 * and SMM (for laptops with [34]86/SL chips) may need it.  It is read
 * and write protected to detect null pointer references in the
 * kernel.
 */
#if 0
	memset((void *) 0, 0, PAGE_SIZE);
#endif
	start_mem = PAGE_ALIGN(start_mem);
	address = 0;
	pg_dir = swapper_pg_dir;
	while (address < end_mem) {
		tmp = *(pg_dir + 768);		/* at virtual addr 0xC0000000 */
		if (!tmp) {
			tmp = start_mem | PAGE_TABLE;
			*(pg_dir + 768) = tmp;
			start_mem += PAGE_SIZE;
		}
		*pg_dir = tmp;			/* also map it in at 0x0000000 for init */
		pg_dir++;
		pg_table = (unsigned long *) (tmp & PAGE_MASK);
		for (tmp = 0 ; tmp < PTRS_PER_PAGE ; tmp++,pg_table++) {
			if (address < end_mem)
				*pg_table = address | PAGE_SHARED;
			else
				*pg_table = 0;
			address += PAGE_SIZE;
		}
	}
	invalidate();
	return start_mem;
}

void mem_init(unsigned long start_low_mem,
	      unsigned long start_mem, unsigned long end_mem)
{
	int codepages = 0;
	int reservedpages = 0;
	int datapages = 0;
	unsigned long tmp;
	unsigned short * p;
	extern int etext;

	cli();
	end_mem &= PAGE_MASK;
	high_memory = end_mem;
	start_mem +=  0x0000000f;
	start_mem &= ~0x0000000f;
	tmp = MAP_NR(end_mem);
	mem_map = (unsigned short *) start_mem;
	p = mem_map + tmp;
	start_mem = (unsigned long) p;
	while (p > mem_map)
		*--p = MAP_PAGE_RESERVED;
	start_low_mem = PAGE_ALIGN(start_low_mem);
	start_mem = PAGE_ALIGN(start_mem);
	while (start_low_mem < 0xA0000) {
		mem_map[MAP_NR(start_low_mem)] = 0;
		start_low_mem += PAGE_SIZE;
	}
	while (start_mem < end_mem) {
		mem_map[MAP_NR(start_mem)] = 0;
		start_mem += PAGE_SIZE;
	}
#ifdef CONFIG_SOUND
	sound_mem_init();
#endif
	free_page_list = 0;
	nr_free_pages = 0;
	for (tmp = 0 ; tmp < end_mem ; tmp += PAGE_SIZE) {
		if (mem_map[MAP_NR(tmp)]) {
			if (tmp >= 0xA0000 && tmp < 0x100000)
				reservedpages++;
			else if (tmp < (unsigned long) &etext)
				codepages++;
			else
				datapages++;
			continue;
		}
		*(unsigned long *) tmp = free_page_list;
		free_page_list = tmp;
		nr_free_pages++;
	}
	tmp = nr_free_pages << PAGE_SHIFT;
	printk("Memory: %luk/%luk available (%dk kernel code, %dk reserved, %dk data)\n",
		tmp >> 10,
		end_mem >> 10,
		codepages << (PAGE_SHIFT-10),
		reservedpages << (PAGE_SHIFT-10),
		datapages << (PAGE_SHIFT-10));
/* test if the WP bit is honoured in supervisor mode */
	wp_works_ok = -1;
	pg0[0] = PAGE_READONLY;
	invalidate();
	__asm__ __volatile__("movb 0,%%al ; movb %%al,0": : :"ax", "memory");
	pg0[0] = 0;
	invalidate();
	if (wp_works_ok < 0)
		wp_works_ok = 0;
	return;
}

void si_meminfo(struct sysinfo *val)
{
	int i;

	i = high_memory >> PAGE_SHIFT;
	val->totalram = 0;
	val->freeram = 0;
	val->sharedram = 0;
	val->bufferram = buffermem;
	while (i-- > 0)  {
		if (mem_map[i] & MAP_PAGE_RESERVED)
			continue;
		val->totalram++;
		if (!mem_map[i]) {
			val->freeram++;
			continue;
		}
		val->sharedram += mem_map[i]-1;
	}
	val->totalram <<= PAGE_SHIFT;
	val->freeram <<= PAGE_SHIFT;
	val->sharedram <<= PAGE_SHIFT;
	return;
}


/* This handles a generic mmap of a disk file */
void file_mmap_nopage(int error_code, struct vm_area_struct * area, unsigned long address)
{
	struct inode * inode = area->vm_inode;
	unsigned int block;
	unsigned long page;
	int nr[8];
	int i, j;
	int prot = area->vm_page_prot;

	address &= PAGE_MASK;
	block = address - area->vm_start + area->vm_offset;
	block >>= inode->i_sb->s_blocksize_bits;

	page = get_free_page(GFP_KERNEL);
	if (share_page(area, area->vm_task, inode, address, error_code, page)) {
		++area->vm_task->min_flt;
		return;
	}

	++area->vm_task->maj_flt;
	if (!page) {
		oom(current);
		put_page(area->vm_task, BAD_PAGE, address, PAGE_PRIVATE);
		return;
	}
	for (i=0, j=0; i< PAGE_SIZE ; j++, block++, i += inode->i_sb->s_blocksize)
		nr[j] = bmap(inode,block);
	if (error_code & PAGE_RW)
		prot |= PAGE_RW | PAGE_DIRTY;
	page = bread_page(page, inode->i_dev, nr, inode->i_sb->s_blocksize, prot);

	if (!(prot & PAGE_RW)) {
		if (share_page(area, area->vm_task, inode, address, error_code, page))
			return;
	}
	if (put_page(area->vm_task,page,address,prot))
		return;
	free_page(page);
	oom(current);
}

void file_mmap_free(struct vm_area_struct * area)
{
	if (area->vm_inode)
		iput(area->vm_inode);
#if 0
	if (area->vm_inode)
		printk("Free inode %x:%d (%d)\n",area->vm_inode->i_dev, 
				 area->vm_inode->i_ino, area->vm_inode->i_count);
#endif
}

/*
 * Compare the contents of the mmap entries, and decide if we are allowed to
 * share the pages
 */
int file_mmap_share(struct vm_area_struct * area1, 
		    struct vm_area_struct * area2, 
		    unsigned long address)
{
	if (area1->vm_inode != area2->vm_inode)
		return 0;
	if (area1->vm_start != area2->vm_start)
		return 0;
	if (area1->vm_end != area2->vm_end)
		return 0;
	if (area1->vm_offset != area2->vm_offset)
		return 0;
	if (area1->vm_page_prot != area2->vm_page_prot)
		return 0;
	return 1;
}

struct vm_operations_struct file_mmap = {
	NULL,			/* open */
	file_mmap_free,		/* close */
	file_mmap_nopage,	/* nopage */
	NULL,			/* wppage */
	file_mmap_share,	/* share */
	NULL,			/* unmap */
};