[PATCH 0/4] mxs/spi: Fixes and improvements

[PATCH 0/4] mxs/spi: Fixes and improvements

[Marek Vasut]

I humbly present fixes for further issues found in the SPI driver
for i.MX23/i.MX28.

The first patch rectifies the handling of DMA chaining in the driver,
hopefully this time it's correct. Sadly, documentation on this matter
is missing.

The further patches are mostly code improvements, cleanups and fixes
for minor details.

Marek Vasut (4):
  mxs/spi: Fix issues when doing long continuous transfer
  mxs/spi: Increment the transfer length only if transfer succeeded
  mxs/spi: Decrement the DMA/PIO border
  mxs/spi: Rework the mxs_ssp_timeout to be more readable

 drivers/spi/spi-mxs.c |  165 +++++++++++++++++++++++++++++--------------------
 1 file changed, 99 insertions(+), 66 deletions(-)

Cc: Chris Ball <cjb-2X9k7bc8m7Mdnm+yROfE0A@public.gmane.org>
Cc: Fabio Estevam <fabio.estevam-KZfg59tc24xl57MIdRCFDg@public.gmane.org>
Cc: Grant Likely <grant.likely-s3s/WqlpOiPyB63q8FvJNQ@public.gmane.org>
Cc: Mark Brown <broonie-yzvPICuk2AATkU/dhu1WVueM+bqZidxxQQ4Iyu8u01E@public.gmane.org>
Cc: Shawn Guo <shawn.guo-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>

-- 
1.7.10.4


------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/

[PATCH 1/4] mxs/spi: Fix issues when doing long continuous transfer

[Marek Vasut]

When doing long continuous transfer, eg. from SPI flash via /dev/mtd,
the driver dies. This is caused by a bug in the DMA chaining. Rework
the DMA transfer code so that this issue does not happen any longer.

This involves proper allocation of correct amount of sg-list members.
Also, this means proper creation of DMA descriptors. There is actually an
important catch to this, the data transfer descriptors must be interleaved
with PIO register write descriptor, otherwise the transfer stalls. This
can be done in one descriptor, but due to the limitation of the DMA API,
it's not possible.

It turns out that in order for the SPI DMA to properly support
continuous transfers longer than 65280 bytes, there are some very
important parts that were left out from the documentation about about
the PIO transfer that is used.

Firstly, the XFER_SIZE register is not written with the whole length
of a transfer, but is written by each and every chained descriptor
with the length of the descriptors data buffer.

Next, unlike the demo code supplied by FSL, which only writes one PIO
word per descriptor, this does not apply if the descriptors are chained,
since the XFER_SIZE register must be written. Therefore, it is essential
to use four PIO words, CTRL0, CMD0, CMD1, XFER_SIZE. CMD0 and CMD1 are
written with zero, since they don't apply. The DMA programs the PIO words
in an incrementing order, so four PIO words.

Finally, unlike the demo code supplied by FSL, the SSP_CTRL0_IGNORE_CRC
must not be set during the whole transfer, but it must be set only on the
last descriptor in the chain.

Lastly, this code lends code from drivers/mtd/nand/omap2.c, which solves
trouble when the buffer supplied to the DMA transfer was vmalloc()'d. So
with this patch, it's safe to use /dev/mtdblockX interface again.

Signed-off-by: Marek Vasut <marex-ynQEQJNshbs@public.gmane.org>
Cc: Chris Ball <cjb-2X9k7bc8m7Mdnm+yROfE0A@public.gmane.org>
Cc: Fabio Estevam <fabio.estevam-KZfg59tc24xl57MIdRCFDg@public.gmane.org>
Cc: Grant Likely <grant.likely-s3s/WqlpOiPyB63q8FvJNQ@public.gmane.org>
Cc: Mark Brown <broonie-yzvPICuk2AATkU/dhu1WVueM+bqZidxxQQ4Iyu8u01E@public.gmane.org>
Cc: Shawn Guo <shawn.guo-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>
---
 drivers/spi/spi-mxs.c |  141 ++++++++++++++++++++++++++++++-------------------
 1 file changed, 88 insertions(+), 53 deletions(-)

diff --git a/drivers/spi/spi-mxs.c b/drivers/spi/spi-mxs.c
index d49634a..3add52e 100644
--- a/drivers/spi/spi-mxs.c
+++ b/drivers/spi/spi-mxs.c
@@ -53,9 +53,9 @@
 
 #define DRIVER_NAME		"mxs-spi"
 
-#define SSP_TIMEOUT		1000	/* 1000 ms */
+/* Use 10S timeout for very long transfers, it should suffice. */
+#define SSP_TIMEOUT		10000
 
-#define SG_NUM			4
 #define SG_MAXLEN		0xff00
 
 struct mxs_spi {
@@ -219,61 +219,94 @@ static int mxs_spi_txrx_dma(struct mxs_spi *spi, int cs,
 			    int *first, int *last, int write)
 {
 	struct mxs_ssp *ssp = &spi->ssp;
-	struct dma_async_tx_descriptor *desc;
-	struct scatterlist sg[SG_NUM];
+	struct dma_async_tx_descriptor *desc = NULL;
+	const bool vmalloced_buf = is_vmalloc_addr(buf);
+	const int desc_len = vmalloced_buf ? PAGE_SIZE : SG_MAXLEN;
+	const int sgs = DIV_ROUND_UP(len, desc_len);
 	int sg_count;
-	uint32_t pio = BM_SSP_CTRL0_DATA_XFER | mxs_spi_cs_to_reg(cs);
-	int ret;
-
-	if (len > SG_NUM * SG_MAXLEN) {
-		dev_err(ssp->dev, "Data chunk too big for DMA\n");
+	int min, ret;
+	uint32_t ctrl0;
+	struct page *vm_page;
+	void *sg_buf;
+	struct {
+		uint32_t		pio[4];
+		struct scatterlist	sg;
+	} *dma_xfer;
+
+	if (!len)
 		return -EINVAL;
-	}
+
+	dma_xfer = kzalloc(sizeof(*dma_xfer) * sgs, GFP_KERNEL);
+	if (!dma_xfer)
+		return -ENOMEM;
 
 	INIT_COMPLETION(spi->c);
 
+	ctrl0 = readl(ssp->base + HW_SSP_CTRL0);
+	ctrl0 |= BM_SSP_CTRL0_DATA_XFER | mxs_spi_cs_to_reg(cs);
+
 	if (*first)
-		pio |= BM_SSP_CTRL0_LOCK_CS;
-	if (*last)
-		pio |= BM_SSP_CTRL0_IGNORE_CRC;
+		ctrl0 |= BM_SSP_CTRL0_LOCK_CS;
 	if (!write)
-		pio |= BM_SSP_CTRL0_READ;
-
-	if (ssp->devid == IMX23_SSP)
-		pio |= len;
-	else
-		writel(len, ssp->base + HW_SSP_XFER_SIZE);
-
-	/* Queue the PIO register write transfer. */
-	desc = dmaengine_prep_slave_sg(ssp->dmach,
-			(struct scatterlist *)&pio,
-			1, DMA_TRANS_NONE, 0);
-	if (!desc) {
-		dev_err(ssp->dev,
-			"Failed to get PIO reg. write descriptor.\n");
-		return -EINVAL;
-	}
+		ctrl0 |= BM_SSP_CTRL0_READ;
 
 	/* Queue the DMA data transfer. */
-	sg_init_table(sg, (len / SG_MAXLEN) + 1);
-	sg_count = 0;
-	while (len) {
-		sg_set_buf(&sg[sg_count++], buf, min(len, SG_MAXLEN));
-		len -= min(len, SG_MAXLEN);
-		buf += min(len, SG_MAXLEN);
-	}
-	dma_map_sg(ssp->dev, sg, sg_count,
-		write ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
-
-	desc = dmaengine_prep_slave_sg(ssp->dmach, sg, sg_count,
-			write ? DMA_MEM_TO_DEV : DMA_DEV_TO_MEM,
-			DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
-
-	if (!desc) {
-		dev_err(ssp->dev,
-			"Failed to get DMA data write descriptor.\n");
-		ret = -EINVAL;
-		goto err;
+	for (sg_count = 0; sg_count < sgs; sg_count++) {
+		min = min(len, desc_len);
+
+		/* Prepare the transfer descriptor. */
+		if ((sg_count + 1 == sgs) && *last)
+			ctrl0 |= BM_SSP_CTRL0_IGNORE_CRC;
+
+		if (ssp->devid == IMX23_SSP)
+			ctrl0 |= min;
+
+		dma_xfer[sg_count].pio[0] = ctrl0;
+		dma_xfer[sg_count].pio[3] = min;
+
+		if (vmalloced_buf) {
+			vm_page = vmalloc_to_page(buf);
+			if (!vm_page) {
+				ret = -ENOMEM;
+				goto err_vmalloc;
+			}
+			sg_buf = page_address(vm_page) +
+				((size_t)buf & ~PAGE_MASK);
+		} else {
+			sg_buf = buf;
+		}
+
+		sg_init_one(&dma_xfer[sg_count].sg, sg_buf, min);
+		ret = dma_map_sg(ssp->dev, &dma_xfer[sg_count].sg, 1,
+			write ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
+
+		len -= min;
+		buf += min;
+
+		/* Queue the PIO register write transfer. */
+		desc = dmaengine_prep_slave_sg(ssp->dmach,
+				(struct scatterlist *)dma_xfer[sg_count].pio,
+				(ssp->devid == IMX23_SSP) ? 1 : 4,
+				DMA_TRANS_NONE,
+				sg_count ? DMA_PREP_INTERRUPT : 0);
+		if (!desc) {
+			dev_err(ssp->dev,
+				"Failed to get PIO reg. write descriptor.\n");
+			ret = -EINVAL;
+			goto err_mapped;
+		}
+
+		desc = dmaengine_prep_slave_sg(ssp->dmach,
+				&dma_xfer[sg_count].sg, 1,
+				write ? DMA_MEM_TO_DEV : DMA_DEV_TO_MEM,
+				DMA_PREP_INTERRUPT | DMA_CTRL_ACK);
+
+		if (!desc) {
+			dev_err(ssp->dev,
+				"Failed to get DMA data write descriptor.\n");
+			ret = -EINVAL;
+			goto err_mapped;
+		}
 	}
 
 	/*
@@ -289,21 +322,23 @@ static int mxs_spi_txrx_dma(struct mxs_spi *spi, int cs,
 
 	ret = wait_for_completion_timeout(&spi->c,
 				msecs_to_jiffies(SSP_TIMEOUT));
-
 	if (!ret) {
 		dev_err(ssp->dev, "DMA transfer timeout\n");
 		ret = -ETIMEDOUT;
-		goto err;
+		goto err_vmalloc;
 	}
 
 	ret = 0;
 
-err:
-	for (--sg_count; sg_count >= 0; sg_count--) {
-		dma_unmap_sg(ssp->dev, &sg[sg_count], 1,
+err_vmalloc:
+	while (--sg_count >= 0) {
+err_mapped:
+		dma_unmap_sg(ssp->dev, &dma_xfer[sg_count].sg, 1,
 			write ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
 	}
 
+	kfree(dma_xfer);
+
 	return ret;
 }
 
-- 
1.7.10.4


------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/

[PATCH 2/4] mxs/spi: Increment the transfer length only if transfer succeeded

[Marek Vasut]

The transfer function incremented (struct spi_message)->actual_length
unconditionally, even if the transfer failed. Rectify this by incrementing
this only if transfer succeeded.

Signed-off-by: Marek Vasut <marex-ynQEQJNshbs@public.gmane.org>
Cc: Chris Ball <cjb-2X9k7bc8m7Mdnm+yROfE0A@public.gmane.org>
Cc: Fabio Estevam <fabio.estevam-KZfg59tc24xl57MIdRCFDg@public.gmane.org>
Cc: Grant Likely <grant.likely-s3s/WqlpOiPyB63q8FvJNQ@public.gmane.org>
Cc: Mark Brown <broonie-yzvPICuk2AATkU/dhu1WVueM+bqZidxxQQ4Iyu8u01E@public.gmane.org>
Cc: Shawn Guo <shawn.guo-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>
---
 drivers/spi/spi-mxs.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/spi/spi-mxs.c b/drivers/spi/spi-mxs.c
index 3add52e..282bdd7 100644
--- a/drivers/spi/spi-mxs.c
+++ b/drivers/spi/spi-mxs.c
@@ -473,12 +473,12 @@ static int mxs_spi_transfer_one(struct spi_master *master,
 						&first, &last, 0);
 		}
 
-		m->actual_length += t->len;
 		if (status) {
 			stmp_reset_block(ssp->base);
 			break;
 		}
 
+		m->actual_length += t->len;
 		first = last = 0;
 	}
 
-- 
1.7.10.4


------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/

[PATCH 3/4] mxs/spi: Decrement the DMA/PIO border

[Marek Vasut]

This driver checks the length of transfer to be made and based
on this information, either chooses to transfer data via DMA or
PIO. Decrement this border further to gain better performace eg.
during SPI flash writes.

Empiric measurement shows that this gives extra 3kB/s write speed
with a M25P80 flash clocked at 40MHz.

Signed-off-by: Marek Vasut <marex-ynQEQJNshbs@public.gmane.org>
Cc: Chris Ball <cjb-2X9k7bc8m7Mdnm+yROfE0A@public.gmane.org>
Cc: Fabio Estevam <fabio.estevam-KZfg59tc24xl57MIdRCFDg@public.gmane.org>
Cc: Grant Likely <grant.likely-s3s/WqlpOiPyB63q8FvJNQ@public.gmane.org>
Cc: Mark Brown <broonie-yzvPICuk2AATkU/dhu1WVueM+bqZidxxQQ4Iyu8u01E@public.gmane.org>
Cc: Shawn Guo <shawn.guo-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>
---
 drivers/spi/spi-mxs.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/spi/spi-mxs.c b/drivers/spi/spi-mxs.c
index 282bdd7..6650ded 100644
--- a/drivers/spi/spi-mxs.c
+++ b/drivers/spi/spi-mxs.c
@@ -445,7 +445,7 @@ static int mxs_spi_transfer_one(struct spi_master *master,
 		 * DMA only: 2.164808 seconds, 473.0KB/s
 		 * Combined: 1.676276 seconds, 610.9KB/s
 		 */
-		if (t->len <= 256) {
+		if (t->len < 32) {
 			writel(BM_SSP_CTRL1_DMA_ENABLE,
 				ssp->base + HW_SSP_CTRL1(ssp) +
 				STMP_OFFSET_REG_CLR);
-- 
1.7.10.4


------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/

[PATCH 4/4] mxs/spi: Rework the mxs_ssp_timeout to be more readable

[Marek Vasut]

Rework the mxs_ssp_timeout() function to make it a bit more readable
and hopefully less error prone. Also, have only one successful exit
from the function and one failing exit instead of two.

Finally, discard the udelay() from this function altogether, as this
tightloop is quick enough it's pointless.

Signed-off-by: Marek Vasut <marex-ynQEQJNshbs@public.gmane.org>
Cc: Chris Ball <cjb-2X9k7bc8m7Mdnm+yROfE0A@public.gmane.org>
Cc: Fabio Estevam <fabio.estevam-KZfg59tc24xl57MIdRCFDg@public.gmane.org>
Cc: Grant Likely <grant.likely-s3s/WqlpOiPyB63q8FvJNQ@public.gmane.org>
Cc: Mark Brown <broonie-yzvPICuk2AATkU/dhu1WVueM+bqZidxxQQ4Iyu8u01E@public.gmane.org>
Cc: Shawn Guo <shawn.guo-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>
---
 drivers/spi/spi-mxs.c |   20 +++++++++-----------
 1 file changed, 9 insertions(+), 11 deletions(-)

diff --git a/drivers/spi/spi-mxs.c b/drivers/spi/spi-mxs.c
index 6650ded..2c3665a 100644
--- a/drivers/spi/spi-mxs.c
+++ b/drivers/spi/spi-mxs.c
@@ -177,25 +177,23 @@ static inline void mxs_spi_disable(struct mxs_spi *spi)
 
 static int mxs_ssp_wait(struct mxs_spi *spi, int offset, int mask, bool set)
 {
-	unsigned long timeout = jiffies + msecs_to_jiffies(SSP_TIMEOUT);
+	const unsigned long timeout = jiffies + msecs_to_jiffies(SSP_TIMEOUT);
 	struct mxs_ssp *ssp = &spi->ssp;
 	uint32_t reg;
 
-	while (1) {
+	do {
 		reg = readl_relaxed(ssp->base + offset);
 
-		if (set && ((reg & mask) == mask))
-			break;
+		if (!set)
+			reg = ~reg;
 
-		if (!set && ((~reg & mask) == mask))
-			break;
+		reg &= mask;
 
-		udelay(1);
+		if (reg == mask)
+			return 0;
+	} while (time_before(jiffies, timeout));
 
-		if (time_after(jiffies, timeout))
-			return -ETIMEDOUT;
-	}
-	return 0;
+	return -ETIMEDOUT;
 }
 
 static void mxs_ssp_dma_irq_callback(void *param)
-- 
1.7.10.4


------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/

Re: [PATCH 0/4] mxs/spi: Fixes and improvements

[Mark Brown]

On Tue, Sep 04, 2012 at 04:40:14AM +0200, Marek Vasut wrote:
> I humbly present fixes for further issues found in the SPI driver
> for i.MX23/i.MX28.

Applied all, thanks.

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/

Re: [PATCH 0/4] mxs/spi: Fixes and improvements

[Marek Vasut]

Dear Mark Brown,

> On Tue, Sep 04, 2012 at 04:40:14AM +0200, Marek Vasut wrote:
> > I humbly present fixes for further issues found in the SPI driver
> > for i.MX23/i.MX28.
> 
> Applied all, thanks.

Thanks Mark, sorry for pulling the bugs in.

Best regards,
Marek Vasut

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and 
threat landscape has changed and how IT managers can respond. Discussions 
will include endpoint security, mobile security and the latest in malware 
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/