blkfront: Fix blkfront backend switch race (bdev open) (13961743) · Commits · 戴 / test

We need not mind if users grab a late handle on a closing disk. We probably even should not. But we have to make sure it's not a dead one already Let the bdev deal with a gendisk deleted under its feet. Takes the info mutex to decide a race against backend closing. Signed-off-by:

Daniel Stodden <daniel.stodden@citrix.com> Signed-off-by:

Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com> Signed-off-by:

Jens Axboe <jaxboe@fusionio.com>

drivers/block/xen-blkfront.c

+24 −7

Original line number	Diff line number	Diff line
		@@ -1118,16 +1118,33 @@ static int blkfront_is_ready(struct xenbus_device *dev)

		static int blkif_open(struct block_device *bdev, fmode_t mode)
		{
		struct blkfront_info *info = bdev->bd_disk->private_data;

		if (!info->xbdev)
		return -ENODEV;
		struct gendisk *disk = bdev->bd_disk;
		struct blkfront_info *info;
		int err = 0;

		lock_kernel();
		info->users++;
		unlock_kernel();

		return 0;
		info = disk->private_data;
		if (!info) {
		/* xbdev gone */
		err = -ERESTARTSYS;
		goto out;
		}

		mutex_lock(&info->mutex);

		if (!info->gd)
		/* xbdev is closed */
		err = -ERESTARTSYS;

		mutex_unlock(&info->mutex);

		if (!err)
		++info->users;

		unlock_kernel();
		out:
		return err;
		}

		static int blkif_release(struct gendisk *disk, fmode_t mode)

Admin message