Merge branch 'sg_nents' into rdma.git for-next

From Maor Gottlieb
====================

Fix the use of nents and orig_nents in the sg table append helpers. The
nents should be used by the DMA layer to store the number of DMA mapped
sges, the orig_nents is the number of CPU sges.

Since the sg append logic doesn't always create a SGL with exactly
orig_nents entries store a total_nents as well to allow the table to be
properly free'd and reorganize the freeing logic to share across all the
use cases.

====================

Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

* 'sg_nents':
  RDMA: Use the sg_table directly and remove the opencoded version from umem
  lib/scatterlist: Fix wrong update of orig_nents
  lib/scatterlist: Provide a dedicated function to support table append

1 files changed, 33 insertions, 19 deletions

diff --git a/drivers/spi/spi-imx.c b/drivers/spi/spi-imx.c
index 39dc02e366f4..fa68e9817929 100644
--- a/drivers/spi/spi-imx.c
+++ b/drivers/spi/spi-imx.c
@@ -505,8 +505,10 @@ static int mx51_ecspi_prepare_message(struct spi_imx_data *spi_imx,
 				      struct spi_message *msg)
 {
 	struct spi_device *spi = msg->spi;
+	struct spi_transfer *xfer;
 	u32 ctrl = MX51_ECSPI_CTRL_ENABLE;
-	u32 testreg;
+	u32 min_speed_hz = ~0U;
+	u32 testreg, delay;
 	u32 cfg = readl(spi_imx->base + MX51_ECSPI_CONFIG);
 
 	/* set Master or Slave mode */
@@ -567,6 +569,35 @@ static int mx51_ecspi_prepare_message(struct spi_imx_data *spi_imx,
 
 	writel(cfg, spi_imx->base + MX51_ECSPI_CONFIG);
 
+	/*
+	 * Wait until the changes in the configuration register CONFIGREG
+	 * propagate into the hardware. It takes exactly one tick of the
+	 * SCLK clock, but we will wait two SCLK clock just to be sure. The
+	 * effect of the delay it takes for the hardware to apply changes
+	 * is noticable if the SCLK clock run very slow. In such a case, if
+	 * the polarity of SCLK should be inverted, the GPIO ChipSelect might
+	 * be asserted before the SCLK polarity changes, which would disrupt
+	 * the SPI communication as the device on the other end would consider
+	 * the change of SCLK polarity as a clock tick already.
+	 *
+	 * Because spi_imx->spi_bus_clk is only set in bitbang prepare_message
+	 * callback, iterate over all the transfers in spi_message, find the
+	 * one with lowest bus frequency, and use that bus frequency for the
+	 * delay calculation. In case all transfers have speed_hz == 0, then
+	 * min_speed_hz is ~0 and the resulting delay is zero.
+	 */
+	list_for_each_entry(xfer, &msg->transfers, transfer_list) {
+		if (!xfer->speed_hz)
+			continue;
+		min_speed_hz = min(xfer->speed_hz, min_speed_hz);
+	}
+
+	delay = (2 * 1000000) / min_speed_hz;
+	if (likely(delay < 10))	/* SCLK is faster than 200 kHz */
+		udelay(delay);
+	else			/* SCLK is _very_ slow */
+		usleep_range(delay, delay + 10);
+
 	return 0;
 }
 
@@ -574,7 +605,7 @@ static int mx51_ecspi_prepare_transfer(struct spi_imx_data *spi_imx,
 				       struct spi_device *spi)
 {
 	u32 ctrl = readl(spi_imx->base + MX51_ECSPI_CTRL);
-	u32 clk, delay;
+	u32 clk;
 
 	/* Clear BL field and set the right value */
 	ctrl &= ~MX51_ECSPI_CTRL_BL_MASK;
@@ -596,23 +627,6 @@ static int mx51_ecspi_prepare_transfer(struct spi_imx_data *spi_imx,
 
 	writel(ctrl, spi_imx->base + MX51_ECSPI_CTRL);
 
-	/*
-	 * Wait until the changes in the configuration register CONFIGREG
-	 * propagate into the hardware. It takes exactly one tick of the
-	 * SCLK clock, but we will wait two SCLK clock just to be sure. The
-	 * effect of the delay it takes for the hardware to apply changes
-	 * is noticable if the SCLK clock run very slow. In such a case, if
-	 * the polarity of SCLK should be inverted, the GPIO ChipSelect might
-	 * be asserted before the SCLK polarity changes, which would disrupt
-	 * the SPI communication as the device on the other end would consider
-	 * the change of SCLK polarity as a clock tick already.
-	 */
-	delay = (2 * 1000000) / clk;
-	if (likely(delay < 10))	/* SCLK is faster than 100 kHz */
-		udelay(delay);
-	else			/* SCLK is _very_ slow */
-		usleep_range(delay, delay + 10);
-
 	return 0;
 }