Re: [PATCH 2/3] block: fail SCSI passthrough ioctls on partition devices

[Paolo Bonzini <pbonzini@redhat.com>]

On 12/23/2011 11:46 PM, Linus Torvalds wrote:
> It sounds like people didn't even*think*  of the potential issues this
> patch can bring.  I'd absolutely be insane to apply them for -rc7.

Fair enough, I obviously cannot say they aren't intrusive.

Anyway, I set to change the patches to use ENOIOCTLCMD.  I did some
research and found the following commit:

commit d9ecdea7ed7467db32ec160f4eca46c279255606
Author: Christoph Hellwig <hch@lst.de>
Date:   Sat Jun 20 21:29:41 2009 +0200

    virtio_blk: ioctl return value fix
    
    Block driver ioctl methods must return ENOTTY and not -ENOIOCTLCMD if
    they expect the block layer to handle generic ioctls.
    
    This triggered a BLKROSET failure in xfsqa #200.
    
    Signed-off-by: Christoph Hellwig <hch@lst.de>
    Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

which indeed matches the current code in block/ioctl.c:

	case BLKROSET:
		ret = __blkdev_driver_ioctl(bdev, mode, cmd, arg);
		/* -EINVAL to handle old uncorrected drivers */
		if (ret != -EINVAL && ret != -ENOTTY)
			return ret;
		if (!capable(CAP_SYS_ADMIN))
			return -EACCES;
		if (get_user(n, (int __user *)(arg)))
			return -EFAULT;
		set_device_ro(bdev, n);
		return 0;

Hence, changing scsi_verify_blk_ioctl to return ENOIOCTLCMD is not
really possible.  I can make it return a boolean value, but I do not
like it: does true mean "pass this ioctl" or "forbid this ioctl"?

Would you apply the patches as they are or do you want me to squash in
something like this?

diff --git a/block/scsi_ioctl.c b/block/scsi_ioctl.c
index a6bedfe..bb94c88 100644
--- a/block/scsi_ioctl.c
+++ b/block/scsi_ioctl.c
@@ -710,6 +710,14 @@ int scsi_verify_blk_ioctl(struct block_device *bd, unsigned int cmd)
 	case SG_SET_RESERVED_SIZE:
 	case SG_EMULATED_HOST:
 		return 0;
+
+	case CDROMEJECT:
+		/* This is also unsafe for partition devices, but
+		 * "eject /mnt/usb-drive" invokes it.  Warn about it
+		 * and keep backwards compatibility.  */
+		printk_ratelimited(KERN_WARNING
+				   "sending CDROMEJECT ioctl to a partition\n");
+		return 0;
 	default:
 		break;
 	}

... perhaps allowing it only for CAP_SYS_RAWIO?

Paolo