On 21/09/2017 16:53, Christoph Hellwig wrote:
> This looks ok to me, but do we even need to keep the special
> cases above? Is there anything relying on the safe but not very
> useful ioctls?
No idea, I stuck to the usual "don't break userspace" rule.
Honestly I doubt anything is using most of those ioctls _in general_,
not just on a partition.
Paolo
> Condensing the thing down to:
>
> int scsi_verify_blk_ioctl(struct block_device *bd, unsigned int cmd)
> {
> if (bd && bd == bd->bd_contains)
> return 0;
> if (capable(CAP_SYS_RAWIO))
> return 0;
> return -ENOIOCTLCMD;
> }
>
> would certainly be nice.
