Hi Geert, On Thursday, November 15, 2018, Geert Uytterhoeven wrote: > > As for validating the values, the only thing I can really check is that: > > of_args.args[2] == RZA2_NPINS > > > > Of course, now that I say that, I realize that if/when it does come time > > to expand this driver beyond the 1 SOC that exists today, I will have > > to stop using that hard coded RZA2_NPINS value...but I'll deal with that > > when the time comes. > > Not that there is an inherent danger in taking of_args.args[2] without > validation to support future SoCs without driver changes. > port_names[] is a fixed size array, so an SoC with more pins will > cause memory access beyond the end of the array. Possibly there are > other locations that need to be changed. > > So IMHO it's better to have explicit support for different SoCs using > different compatible values, so you can store the number of pins in > of_device_id.data, and validate against that. I will fill in of_device_id.data with the number of ports for each SOC (only 1 SOC today) and then use that number throughout the driver instead of the hard coded number. I'll also check that of_args.args[2] matches of_device_id.data. Then, it should be real easy to add support for another SOC when it comes along. RZ/T1 which has the same controller has ports up to Port "U" for example. Chris
