On Fri, May 22, 2020 at 10:57 PM dillon min <dillon.minfei@xxxxxxxxx> wrote: > > hi Mark, > > Thanks for reviewing. > > On Fri, May 22, 2020 at 7:36 PM Mark Brown <broonie@xxxxxxxxxx> wrote: > > > > On Mon, May 18, 2020 at 07:09:20PM +0800, dillon.minfei@xxxxxxxxx wrote: > > > > > 2, use stm32 spi's "In full-duplex (BIDIMODE=0 and RXONLY=0)", as tx_buf is > > > null, we must add dummy data sent out before read data. > > > so, add stm32f4_spi_tx_dummy() to handle this situation. > > > > There are flags SPI_CONTROLLER_MUST_TX and SPI_CONTROLLER_MUST_RX flags > > that the driver can set if it needs to, no need to open code this in the > > driver. > > Yes, after check SPI_CONTROLLER_MUST_TX in drivers/spi/spi.c , it's > indeed to meet > this situation, i will try it and sumbmit a new patch. > > thanks. > > best regards > > Dillon Hi Mark, There might be a conflict with 'SPI_CONTROLLER_MUST_TX' and 'SPI_3WIRE' mode, i need to know the SPI_3WIRE direction, currently i get this information from 'struct spi_device' and 'struct spi_transfer' if ((spi_device->mode & SPI_3WIRE) && (spi_transfer->tx_buf == NULL) && (spi_transfer->rx_buf != NULL)) this is a SPI_3WIRE_RX transfer if ((spi_device->mode & SPI_3WIRE) && (spi_transfer->tx_buf != NULL) && (spi_transfer->rx_buf == NULL)) this is a SPI_3WIRE_TX transfer but, after spi-core create a dummy tx_buf or rx_buf, then i can't get the correct spi_3wire direction. actually, this dummy tx_buf is useless for SPI_3WIRE. it's has meaning for SPI_SIMPLE_RX mode, simulate SPI_FULL_DUMPLEX how do you think? thanks best regards Dillon