On Mon, 10 Dec 2018 10:31:57 +0000
Schrempf Frieder [off-list ref] wrote:

quoted

quoted

Yes, I need to validate op->addr.nbytes else LUT would going to be programmed for 0 addrlen.
I have checked this on the target.  

Also agree there. Some operations have 0 address bytes. We could also
test addr.buswidth, but I'm fine with the addr.nbytes test too.  

The "if (op->addr.nbytes)" is needed of course, but I think the default 
case in the switch statement (and for other reasons the whole switch 
statement) is not needed and rather a check for op->addr.nbytes > 4 
should be added to nxp_fspi_supports_op(). I wrongly assumed this check 
already exists in nxp_fspi_supports_op().

Ok, then this check on the max number of address bytes should indeed be
moved to the supports_op() implementation.