On Sun, Nov 08, 2015 at 11:13:33AM +0100, Marcus Weseloh wrote: > 2015-11-08 10:35 GMT+01:00 Mark Brown <broonie@xxxxxxxxxx>: > > This is not needed, the core will always ensure that the transfer speed > > is set per on each transfer. > Ah, so we could always use tfr->speed_hz and completely ignore the > spi->max_speed_hz for the clock calculations, correct? Yes. > I did notice however, that the SPI system doesn't validate the > speed_hz value. So if I leave out the above quoted sanity checking, > then tfr->speed_hz could also contain values like -400, which trips up > the spi system in such a way that I could only repair with a reboot. I > was testing this with spidev... so maybe spidev should be patched to > check the speed value for validity? No, why would it make sense to have this validation in spidev - it's just one SPI device. It's better to add the validation to the SPI core.
Attachment:
signature.asc
Description: PGP signature