On Sat, Jul 10, 2021 at 10:32:46AM -0500, David Lechner wrote: > On 7/10/21 7:18 AM, Uwe Kleine-König wrote: > > If it's not broken for you without my patch, your spi bus driver doesn't > > honor .cs_change in the last transfer. Out of interest: Which bus are > > you using? I wonder if the driver should refuse the request if it cannot > > honer .cs_change?! (spi-imx does honor it only if gpios are used as chip > > select, the native chip selects cannot do that.) > I'm using spi-davinci. It uses the standard spi_transfer_one_message() > which handles cs_change. But I suspect when the SPI_CS_WORD flag is set, > and the message is big enough to use DMA, the hardware is probably > automatically toggling CS after the last transfer before the cs_change > logic asserts it again. > So unless there is a valid use case where we need both SPI_CS_WORD > and cs_change, I don't think we need to fix spi-davinci. In theory it does mean something, in practice the number of controllers that can implement it must be approaching zero.
Attachment:
signature.asc
Description: PGP signature
