io_opt fixups

[Christoph Hellwig]

Hi all,

I recently noticed that on my test VMs with the Qemu NVMe emulations I
see a zero max_setors_kb limit in sysfs.  It turns out Qemu advertises
a one-LBA optimal write size, which is a bit silly.

This series handles this odd case properly both in the block layer and
the nvme driver as a sort of defense in depth.

Diffstat:
 block/blk-settings.c     |    5 ++---
 drivers/nvme/host/core.c |    3 ++-
 2 files changed, 4 insertions(+), 4 deletions(-)