The Freescale MPC8xxx had a special quirk for handling a single hardwired chipselect, the case when we're using neither GPIO nor native chip select: when inspecting the device tree and finding zero "cs-gpios" on the device node the code would assume we have a single hardwired chipselect that leaves the device always selected. This quirk is not handled by the new core code, so we need to check the "cs-gpios" explicitly in the driver and set pdata->max_chipselect = 1 which will later fall through to the SPI master ->num_chipselect. Make sure not to assign the chip select handler in this case: there is no handling needed since the chip is always selected, and this is what the old code did as well. Cc: Christophe Leroy <christophe.leroy@xxxxxx> Reported-by: Christophe Leroy <christophe.leroy@xxxxxx> Fixes: 0f0581b24bd0 ("spi: fsl: Convert to use CS GPIO descriptors") Signed-off-by: Linus Walleij <linus.walleij@xxxxxxxxxx> --- ChangeLog v2->v3: - Resend with the other patches. ChangeLog v1->v2: - Reordered patches. - Expanded comment a bit. --- drivers/spi/spi-fsl-spi.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-) diff --git a/drivers/spi/spi-fsl-spi.c b/drivers/spi/spi-fsl-spi.c index c87e9c4506c2..4b70887cf443 100644 --- a/drivers/spi/spi-fsl-spi.c +++ b/drivers/spi/spi-fsl-spi.c @@ -728,8 +728,18 @@ static int of_fsl_spi_probe(struct platform_device *ofdev) } } #endif - - pdata->cs_control = fsl_spi_cs_control; + /* + * Handle the case where we have one hardwired (always selected) + * device on the first "chipselect". Else we let the core code + * handle any GPIOs or native chip selects and assign the + * appropriate callback for dealing with the CS lines. This isn't + * supported on the GRLIB variant. + */ + ret = gpiod_count(dev, "cs"); + if (ret <= 0) + pdata->max_chipselect = 1; + else + pdata->cs_control = fsl_spi_cs_control; } ret = of_address_to_resource(np, 0, &mem); -- 2.23.0