On Montag, 4. September 2017 09:59:24 CEST Maxime Ripard wrote:

On Mon, Sep 04, 2017 at 12:40:53AM +0200, Stefan Brüns wrote:

quoted

For the H3, the burst lengths field offsets in the channel configuration
register differs from earlier SoC generations.

Using the A31 register macros actually configured the H3 controller
do to bursts of length 1 always, which although working leads to higher
bus utilisation.

Signed-off-by: Stefan Brüns <redacted>
---

 drivers/dma/sun6i-dma.c | 28 +++++++++++++++++++++-------
 1 file changed, 21 insertions(+), 7 deletions(-)

diff --git a/drivers/dma/sun6i-dma.c b/drivers/dma/sun6i-dma.c
index 1d9b3be30d22..f1a139f0102f 100644
--- a/drivers/dma/sun6i-dma.c
+++ b/drivers/dma/sun6i-dma.c

@@ -68,13 +68,15 @@

 #define DMA_CHAN_CFG_SRC_DRQ(x)		((x) & 0x1f)
 #define DMA_CHAN_CFG_SRC_IO_MODE	BIT(5)
 #define DMA_CHAN_CFG_SRC_LINEAR_MODE	(0 << 5)

-#define DMA_CHAN_CFG_SRC_BURST(x)	(((x) & 0x3) << 7)
+#define DMA_CHAN_CFG_SRC_BURST_A31(x)	(((x) & 0x3) << 7)
+#define DMA_CHAN_CFG_SRC_BURST_H3(x)	(((x) & 0x3) << 6)

 #define DMA_CHAN_CFG_SRC_WIDTH(x)	(((x) & 0x3) << 9)
 
 #define DMA_CHAN_CFG_DST_DRQ(x)		(DMA_CHAN_CFG_SRC_DRQ(x) << 16)
 #define DMA_CHAN_CFG_DST_IO_MODE	(DMA_CHAN_CFG_SRC_IO_MODE << 16)
 #define DMA_CHAN_CFG_DST_LINEAR_MODE	(DMA_CHAN_CFG_SRC_LINEAR_MODE << 16)

-#define DMA_CHAN_CFG_DST_BURST(x)	(DMA_CHAN_CFG_SRC_BURST(x) << 16)
+#define DMA_CHAN_CFG_DST_BURST_A31(x)	(DMA_CHAN_CFG_SRC_BURST_A31(x) <<

16) +#define DMA_CHAN_CFG_DST_BURST_H3(x)	(DMA_CHAN_CFG_SRC_BURST_H3(x)
<< 16)> 
 #define DMA_CHAN_CFG_DST_WIDTH(x)	(DMA_CHAN_CFG_SRC_WIDTH(x) << 16)
 
 #define DMA_CHAN_CUR_SRC	0x10

@@ -554,11 +556,17 @@ static int set_config(struct sun6i_dma_dev *sdev,

 	if (dst_width < 0)
 	
 		return dst_width;

-	*p_cfg = DMA_CHAN_CFG_SRC_BURST(src_burst) |
-		DMA_CHAN_CFG_SRC_WIDTH(src_width) |
-		DMA_CHAN_CFG_DST_BURST(dst_burst) |
+	*p_cfg = DMA_CHAN_CFG_SRC_WIDTH(src_width) |

 		DMA_CHAN_CFG_DST_WIDTH(dst_width);

+	if (sdev->cfg->dmac_variant == DMAC_VARIANT_H3) {
+		*p_cfg |= DMA_CHAN_CFG_SRC_BURST_H3(src_burst) |
+			  DMA_CHAN_CFG_DST_BURST_H3(dst_burst);
+	} else {
+		*p_cfg |= DMA_CHAN_CFG_SRC_BURST_A31(src_burst) |
+			  DMA_CHAN_CFG_DST_BURST_A31(dst_burst);
+	}
+

I guess we have two options to support that properly. We could either
have a different function that would generate that register value
based on the parameters we have, or duplicate the set_config function
entirely, with function pointer stored in the configuration.

I think I prefer the former, as it reduces the code duplication.

Duplicating "set_config" would also mean duplicating sun6i_dma_prep_dma_memcpy 
- there are two hunks which change setting of the burst length register value.

A function pointer in the config would work.

Kind regards,

Stefan

-- 
You received this message because you are subscribed to the Google Groups "linux-sunxi" group.
To unsubscribe from this group and stop receiving emails from it, send an email to linux-sunxi+unsubscribe-/JYPxA39Uh5TLH3MbocFF+G/Ez6ZCGd0@public.gmane.org
For more options, visit https://groups.google.com/d/optout.