Looks good to me. I'm OK with removing scsi_low and all related code for FreeBSD 12.

I came across it during some other clean-up work.

Resolved in r322943

Rewrote the xpt_announce_periph and xpt_denounce_periph functions, the
previous work tried too hard and failed at code sharing. Also add in
seatbelts for transports that don't support sbufs yet, and add a tunable
knob to force the non-use of sbufs

Pending our discussion on IRC, would you mind if I took over this review and extended it?

I appreciate the effort on this, but I almost wonder if it's better to let the old hardware die off. I have no way to test this, but my initial scan looks ok. Which driver will have priority at boot if both the aac and aacraid variants are present? They both will return BUS_PROBE_DEFAULT, so it'll be indeterminate.

Need input on naming. sbuf_putbuf() was named since it wraps putbuf(). Is this the best name, though? Also, should it be kernel only, and the userland be something else that suggests printing to stdout vs stderr vs some other file descriptor?

Update patch to provide a kernel version and userland version
of sbuf_putbuf()

Expose _xpt_path_string() as xpt_path_sbuf()

Fix an uninitialized use of msi_msgs.