Thread (26 messages) 26 messages, 4 authors, 2015-10-20

[PATCH v2 1/8] spi: imx: Fix DMA transfer

From: Anton Bondarenko <hidden>
Date: 2015-10-20 23:04:00
Also in: linux-spi, lkml

On 08.10.2015 11:19, Robin Gong wrote:
On Thu, Oct 01, 2015 at 12:02:41AM +0000, Bondarenko, Anton wrote:
quoted
quoted
quoted
@@ -201,9 +202,8 @@ static bool spi_imx_can_dma(struct spi_master *master, struct spi_device *spi,
  {
  	struct spi_imx_data *spi_imx = spi_master_get_devdata(master);

-	if (spi_imx->dma_is_inited
-	    && transfer->len > spi_imx->rx_wml * sizeof(u32)
-	    && transfer->len > spi_imx->tx_wml * sizeof(u32))
+	if (spi_imx->dma_is_inited &&
+	    (transfer->len > spi_imx->wml * sizeof(u32)))
Add Sascha in the loop. I don't think "* sizeof(u32)", since even 1 byte data
will consume one position of 32bit FIFO Thus if here
spi_imx->wml = spi_imx_get_fifosize(spi_imx) / 2 = 32, the threshold value
which judge DMA mode used or not should be 32 not 32 * 4.
Of course, it will not cause any function break since both DMA and PIO can work
,but I think we'd better correct it.
I agree, in case of 1 byte SPI word we do not need to multiply by 4.
But for 16 bit and 32 bit SPI words it's necessary. This part is
addressed in patch 3.
I could remove "* sizeof(u32)" for now.
I still think don't need *sizeof(u32) even for 16bit and 32bit, whatever bits
used as one spi word(<32bits), one spi word consume one position of SPI FIFO
(32bit).
Will be removed in V3 for this patch.
quoted
quoted
quoted
  		return true;
  	return false;
  }
Regards,
Anton
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help