Re: [RFC PATCH] nvme: allow NVME_IOCTL_IO_CMD on controller char dev even when multiple ns

[Christoph Hellwig]

Well, there is at least one good reason for not allowing this
retroactively:

Old users of the ioctl could have complete garbage in the field, and
might send the command to a random namespace now instead of the first
one.

So unles we have a very good reason I think we should keep
NVME_IOCTL_IO_CMD on the controller char dev deprecated and maybe
eventually remove it.