Re: [PATCH 06/32] spi: dw: Add support for the Kendryte K210 SoC

On 2020/11/10 6:22, Serge Semin wrote:

On Sat, Nov 07, 2020 at 05:13:54PM +0900, Damien Le Moal wrote:

quoted

The DW SPI master of the Kendryte K210 RISC-V SoC uses the 32-bits
ctrlr0 register format. This SoC is also quite slow and gets significant
SD card performance improvements from using no-delay polled transfers.
Add the dw_spi_k210_init() function tied to the
"canaan,kendryte-k210-spi" compatible string to set the
DW_SPI_CAP_DFS_32 and DW_SPI_CAP_POLL_NODELAY DW SPI capability fields
for this SoC.

Signed-off-by: Damien Le Moal <redacted>
---
 drivers/spi/spi-dw-mmio.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/drivers/spi/spi-dw-mmio.c b/drivers/spi/spi-dw-mmio.c
index 3f1bc384cb45..a00def6c5b39 100644
--- a/drivers/spi/spi-dw-mmio.c
+++ b/drivers/spi/spi-dw-mmio.c

@@ -223,6 +223,14 @@ static int dw_spi_keembay_init(struct platform_device *pdev,
 	return 0;
 }
 
+static int dw_spi_k210_init(struct platform_device *pdev,
+			    struct dw_spi_mmio *dwsmmio)
+{
+	dwsmmio->dws.caps = DW_SPI_CAP_DFS_32 | DW_SPI_CAP_POLL_NODELAY;
+
+	return 0;
+}
+
 static int dw_spi_mmio_probe(struct platform_device *pdev)
 {
 	int (*init_func)(struct platform_device *pdev,

@@ -340,6 +348,7 @@ static const struct of_device_id dw_spi_mmio_of_match[] = {
 	{ .compatible = "snps,dwc-ssi-1.01a", .data = dw_spi_dwc_ssi_init},
 	{ .compatible = "intel,keembay-ssi", .data = dw_spi_keembay_init},
 	{ .compatible = "microchip,sparx5-spi", dw_spi_mscc_sparx5_init},

quoted

+	{ .compatible = "canaan,kendryte-k210-spi", .data = dw_spi_k210_init},

Other than the comments from Sean and Mark regarding the DFS_32
feature runtime detectability, I couldn't find a patch with adding the
new new compatible string into the DW APB SSI DT schema. Have I missed
it? If I haven't could you add one to the next version of the series?

Yes, I will. I forgot to change the DW DT binding doc for this. I did add a
patch for the "polling" property but forgot the compatible string.

In any case, I think that this new compatible string change can be dropped by
switching to automatically detecting the DFS32 and using a different solution
than the polling property change I sent for the RX fifo overflow problem.

I am still going through all the emails trying to understand what to try next to
avoid the polling "hack".

Thanks !

-Sergey

quoted

 	{ /* end of table */}
 };
 MODULE_DEVICE_TABLE(of, dw_spi_mmio_of_match);
-- 
2.28.0

-- 
Damien Le Moal
Western Digital Research