ubd_kern.c

来自「linux 内核源代码」· C语言 代码 · 共 1,485 行 · 第 1/3 页

		ubd_end_request(req, 0, 0);
		return;
	}
	ubd_end_request(req, bytes, 1);
}

static LIST_HEAD(restart);

/* XXX - move this inside ubd_intr. */
/* Called without dev->lock held, and only in interrupt context. */
static void ubd_handler(void)
{
	struct io_thread_req *req;
	struct request *rq;
	struct ubd *ubd;
	struct list_head *list, *next_ele;
	unsigned long flags;
	int n;

	while(1){
		n = os_read_file(thread_fd, &req,
				 sizeof(struct io_thread_req *));
		if(n != sizeof(req)){
			if(n == -EAGAIN)
				break;
			printk(KERN_ERR "spurious interrupt in ubd_handler, "
			       "err = %d\n", -n);
			return;
		}

		rq = req->req;
		rq->nr_sectors -= req->length >> 9;
		if(rq->nr_sectors == 0)
			ubd_finish(rq, rq->hard_nr_sectors << 9);
		kfree(req);
	}
	reactivate_fd(thread_fd, UBD_IRQ);

	list_for_each_safe(list, next_ele, &restart){
		ubd = container_of(list, struct ubd, restart);
		list_del_init(&ubd->restart);
		spin_lock_irqsave(&ubd->lock, flags);
		do_ubd_request(ubd->queue);
		spin_unlock_irqrestore(&ubd->lock, flags);
	}
}

static irqreturn_t ubd_intr(int irq, void *dev)
{
	ubd_handler();
	return IRQ_HANDLED;
}

/* Only changed by ubd_init, which is an initcall. */
static int io_pid = -1;

void kill_io_thread(void)
{
	if(io_pid != -1)
		os_kill_process(io_pid, 1);
}

__uml_exitcall(kill_io_thread);

static inline int ubd_file_size(struct ubd *ubd_dev, __u64 *size_out)
{
	char *file;

	file = ubd_dev->cow.file ? ubd_dev->cow.file : ubd_dev->file;
	return os_file_size(file, size_out);
}

static void ubd_close_dev(struct ubd *ubd_dev)
{
	os_close_file(ubd_dev->fd);
	if(ubd_dev->cow.file == NULL)
		return;

	os_close_file(ubd_dev->cow.fd);
	vfree(ubd_dev->cow.bitmap);
	ubd_dev->cow.bitmap = NULL;
}

static int ubd_open_dev(struct ubd *ubd_dev)
{
	struct openflags flags;
	char **back_ptr;
	int err, create_cow, *create_ptr;
	int fd;

	ubd_dev->openflags = ubd_dev->boot_openflags;
	create_cow = 0;
	create_ptr = (ubd_dev->cow.file != NULL) ? &create_cow : NULL;
	back_ptr = ubd_dev->no_cow ? NULL : &ubd_dev->cow.file;

	fd = open_ubd_file(ubd_dev->file, &ubd_dev->openflags, ubd_dev->shared,
				back_ptr, &ubd_dev->cow.bitmap_offset,
				&ubd_dev->cow.bitmap_len, &ubd_dev->cow.data_offset,
				create_ptr);

	if((fd == -ENOENT) && create_cow){
		fd = create_cow_file(ubd_dev->file, ubd_dev->cow.file,
					  ubd_dev->openflags, 1 << 9, PAGE_SIZE,
					  &ubd_dev->cow.bitmap_offset,
					  &ubd_dev->cow.bitmap_len,
					  &ubd_dev->cow.data_offset);
		if(fd >= 0){
			printk(KERN_INFO "Creating \"%s\" as COW file for "
			       "\"%s\"\n", ubd_dev->file, ubd_dev->cow.file);
		}
	}

	if(fd < 0){
		printk("Failed to open '%s', errno = %d\n", ubd_dev->file,
		       -fd);
		return fd;
	}
	ubd_dev->fd = fd;

	if(ubd_dev->cow.file != NULL){
		blk_queue_max_sectors(ubd_dev->queue, 8 * sizeof(long));

		err = -ENOMEM;
		ubd_dev->cow.bitmap = vmalloc(ubd_dev->cow.bitmap_len);
		if(ubd_dev->cow.bitmap == NULL){
			printk(KERN_ERR "Failed to vmalloc COW bitmap\n");
			goto error;
		}
		flush_tlb_kernel_vm();

		err = read_cow_bitmap(ubd_dev->fd, ubd_dev->cow.bitmap,
				      ubd_dev->cow.bitmap_offset,
				      ubd_dev->cow.bitmap_len);
		if(err < 0)
			goto error;

		flags = ubd_dev->openflags;
		flags.w = 0;
		err = open_ubd_file(ubd_dev->cow.file, &flags, ubd_dev->shared, NULL,
				    NULL, NULL, NULL, NULL);
		if(err < 0) goto error;
		ubd_dev->cow.fd = err;
	}
	return 0;
 error:
	os_close_file(ubd_dev->fd);
	return err;
}

static void ubd_device_release(struct device *dev)
{
	struct ubd *ubd_dev = dev->driver_data;

	blk_cleanup_queue(ubd_dev->queue);
	*ubd_dev = ((struct ubd) DEFAULT_UBD);
}

static int ubd_disk_register(int major, u64 size, int unit,
			     struct gendisk **disk_out)
{
	struct gendisk *disk;

	disk = alloc_disk(1 << UBD_SHIFT);
	if(disk == NULL)
		return -ENOMEM;

	disk->major = major;
	disk->first_minor = unit << UBD_SHIFT;
	disk->fops = &ubd_blops;
	set_capacity(disk, size / 512);
	if(major == MAJOR_NR)
		sprintf(disk->disk_name, "ubd%c", 'a' + unit);
	else
		sprintf(disk->disk_name, "ubd_fake%d", unit);

	/* sysfs register (not for ide fake devices) */
	if (major == MAJOR_NR) {
		ubd_devs[unit].pdev.id   = unit;
		ubd_devs[unit].pdev.name = DRIVER_NAME;
		ubd_devs[unit].pdev.dev.release = ubd_device_release;
		ubd_devs[unit].pdev.dev.driver_data = &ubd_devs[unit];
		platform_device_register(&ubd_devs[unit].pdev);
		disk->driverfs_dev = &ubd_devs[unit].pdev.dev;
	}

	disk->private_data = &ubd_devs[unit];
	disk->queue = ubd_devs[unit].queue;
	add_disk(disk);

	*disk_out = disk;
	return 0;
}

#define ROUND_BLOCK(n) ((n + ((1 << 9) - 1)) & (-1 << 9))

static int ubd_add(int n, char **error_out)
{
	struct ubd *ubd_dev = &ubd_devs[n];
	int err = 0;

	if(ubd_dev->file == NULL)
		goto out;

	err = ubd_file_size(ubd_dev, &ubd_dev->size);
	if(err < 0){
		*error_out = "Couldn't determine size of device's file";
		goto out;
	}

	ubd_dev->size = ROUND_BLOCK(ubd_dev->size);

	INIT_LIST_HEAD(&ubd_dev->restart);
	sg_init_table(ubd_dev->sg, MAX_SG);

	err = -ENOMEM;
	ubd_dev->queue = blk_init_queue(do_ubd_request, &ubd_dev->lock);
	if (ubd_dev->queue == NULL) {
		*error_out = "Failed to initialize device queue";
		goto out;
	}
	ubd_dev->queue->queuedata = ubd_dev;

	blk_queue_max_hw_segments(ubd_dev->queue, MAX_SG);
	err = ubd_disk_register(MAJOR_NR, ubd_dev->size, n, &ubd_gendisk[n]);
	if(err){
		*error_out = "Failed to register device";
		goto out_cleanup;
	}

	if(fake_major != MAJOR_NR)
		ubd_disk_register(fake_major, ubd_dev->size, n,
				  &fake_gendisk[n]);

	/* perhaps this should also be under the "if (fake_major)" above */
	/* using the fake_disk->disk_name and also the fakehd_set name */
	if (fake_ide)
		make_ide_entries(ubd_gendisk[n]->disk_name);

	err = 0;
out:
	return err;

out_cleanup:
	blk_cleanup_queue(ubd_dev->queue);
	goto out;
}

static int ubd_config(char *str, char **error_out)
{
	int n, ret;

	/* This string is possibly broken up and stored, so it's only
	 * freed if ubd_setup_common fails, or if only general options
	 * were set.
	 */
	str = kstrdup(str, GFP_KERNEL);
	if (str == NULL) {
		*error_out = "Failed to allocate memory";
		return -ENOMEM;
	}

	ret = ubd_setup_common(str, &n, error_out);
	if (ret)
		goto err_free;

	if (n == -1) {
		ret = 0;
		goto err_free;
	}

	mutex_lock(&ubd_lock);
	ret = ubd_add(n, error_out);
	if (ret)
		ubd_devs[n].file = NULL;
	mutex_unlock(&ubd_lock);

out:
	return ret;

err_free:
	kfree(str);
	goto out;
}

static int ubd_get_config(char *name, char *str, int size, char **error_out)
{
	struct ubd *ubd_dev;
	int n, len = 0;

	n = parse_unit(&name);
	if((n >= MAX_DEV) || (n < 0)){
		*error_out = "ubd_get_config : device number out of range";
		return -1;
	}

	ubd_dev = &ubd_devs[n];
	mutex_lock(&ubd_lock);

	if(ubd_dev->file == NULL){
		CONFIG_CHUNK(str, size, len, "", 1);
		goto out;
	}

	CONFIG_CHUNK(str, size, len, ubd_dev->file, 0);

	if(ubd_dev->cow.file != NULL){
		CONFIG_CHUNK(str, size, len, ",", 0);
		CONFIG_CHUNK(str, size, len, ubd_dev->cow.file, 1);
	}
	else CONFIG_CHUNK(str, size, len, "", 1);

 out:
	mutex_unlock(&ubd_lock);
	return len;
}

static int ubd_id(char **str, int *start_out, int *end_out)
{
	int n;

	n = parse_unit(str);
	*start_out = 0;
	*end_out = MAX_DEV - 1;
	return n;
}

static int ubd_remove(int n, char **error_out)
{
	struct gendisk *disk = ubd_gendisk[n];
	struct ubd *ubd_dev;
	int err = -ENODEV;

	mutex_lock(&ubd_lock);

	ubd_dev = &ubd_devs[n];

	if(ubd_dev->file == NULL)
		goto out;

	/* you cannot remove a open disk */
	err = -EBUSY;
	if(ubd_dev->count > 0)
		goto out;

	ubd_gendisk[n] = NULL;
	if(disk != NULL){
		del_gendisk(disk);
		put_disk(disk);
	}

	if(fake_gendisk[n] != NULL){
		del_gendisk(fake_gendisk[n]);
		put_disk(fake_gendisk[n]);
		fake_gendisk[n] = NULL;
	}

	err = 0;
	platform_device_unregister(&ubd_dev->pdev);
out:
	mutex_unlock(&ubd_lock);
	return err;
}

/* All these are called by mconsole in process context and without
 * ubd-specific locks.  The structure itself is const except for .list.
 */
static struct mc_device ubd_mc = {
	.list		= LIST_HEAD_INIT(ubd_mc.list),
	.name		= "ubd",
	.config		= ubd_config,
	.get_config	= ubd_get_config,
	.id		= ubd_id,
	.remove		= ubd_remove,
};

static int __init ubd_mc_init(void)
{
	mconsole_register_dev(&ubd_mc);
	return 0;
}

__initcall(ubd_mc_init);

static int __init ubd0_init(void)
{
	struct ubd *ubd_dev = &ubd_devs[0];

	mutex_lock(&ubd_lock);
	if(ubd_dev->file == NULL)
		ubd_dev->file = "root_fs";
	mutex_unlock(&ubd_lock);

	return 0;
}

__initcall(ubd0_init);

/* Used in ubd_init, which is an initcall */
static struct platform_driver ubd_driver = {
	.driver = {
		.name  = DRIVER_NAME,
	},
};

static int __init ubd_init(void)
{
	char *error;
	int i, err;

	if (register_blkdev(MAJOR_NR, "ubd"))
		return -1;

	if (fake_major != MAJOR_NR) {
		char name[sizeof("ubd_nnn\0")];

		snprintf(name, sizeof(name), "ubd_%d", fake_major);
		if (register_blkdev(fake_major, "ubd"))
			return -1;
	}
	platform_driver_register(&ubd_driver);
	mutex_lock(&ubd_lock);
	for (i = 0; i < MAX_DEV; i++){
		err = ubd_add(i, &error);
		if(err)
			printk(KERN_ERR "Failed to initialize ubd device %d :"
			       "%s\n", i, error);
	}
	mutex_unlock(&ubd_lock);
	return 0;
}

late_initcall(ubd_init);

static int __init ubd_driver_init(void){
	unsigned long stack;
	int err;

	/* Set by CONFIG_BLK_DEV_UBD_SYNC or ubd=sync.*/
	if(global_openflags.s){
		printk(KERN_INFO "ubd: Synchronous mode\n");
		/* Letting ubd=sync be like using ubd#s= instead of ubd#= is
		 * enough. So use anyway the io thread. */
	}
	stack = alloc_stack(0, 0);
	io_pid = start_io_thread(stack + PAGE_SIZE - sizeof(void *),
				 &thread_fd);
	if(io_pid < 0){
		printk(KERN_ERR
		       "ubd : Failed to start I/O thread (errno = %d) - "
		       "falling back to synchronous I/O\n", -io_pid);
		io_pid = -1;
		return 0;
	}
	err = um_request_irq(UBD_IRQ, thread_fd, IRQ_READ, ubd_intr,
			     IRQF_DISABLED, "ubd", ubd_devs);
	if(err != 0)
		printk(KERN_ERR "um_request_irq failed - errno = %d\n", -err);
	return 0;
}

device_initcall(ubd_driver_init);

static int ubd_open(struct inode *inode, struct file *filp)
{
	struct gendisk *disk = inode->i_bdev->bd_disk;
	struct ubd *ubd_dev = disk->private_data;
	int err = 0;

	if(ubd_dev->count == 0){
		err = ubd_open_dev(ubd_dev);
		if(err){
			printk(KERN_ERR "%s: Can't open \"%s\": errno = %d\n",
			       disk->disk_name, ubd_dev->file, -err);
			goto out;
		}
	}
	ubd_dev->count++;
	set_disk_ro(disk, !ubd_dev->openflags.w);

	/* This should no more be needed. And it didn't work anyway to exclude
	 * read-write remounting of filesystems.*/
	/*if((filp->f_mode & FMODE_WRITE) && !ubd_dev->openflags.w){
	        if(--ubd_dev->count == 0) ubd_close_dev(ubd_dev);
	        err = -EROFS;
	}*/
 out:
	return err;
}

static int ubd_release(struct inode * inode, struct file * file)
{
	struct gendisk *disk = inode->i_bdev->bd_disk;
	struct ubd *ubd_dev = disk->private_data;

	if(--ubd_dev->count == 0)