Skip to contentSkip to conversation

spi: stm32: various driver fixes

PATCH In review Alain Volmat 3 participants 8 messages Aug 10, 2020 – Sep 14, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

v2 cover spi: stm32: various driver fixes

This serie is a reduced version of the serie [spi: stm32: various driver enhancements] previously sent.

Alain Volmat (1):
spi: stm32: always perform registers configuration prior to transfer

Amelie Delaunay (3):
spi: stm32: fix fifo threshold level in case of short transfer
spi: stm32: fix stm32_spi_prepare_mbr in case of odd clk_rate
spi: stm32: fixes suspend/resume management

Antonio Borneo (1):
spi: stm32h7: fix race condition at end of transfer

 drivers/spi/spi-stm32.c | 98 ++++++++++++++++++++++++++++++-------------------
 1 file changed, 61 insertions(+), 37 deletions(-)
Signature · 1 line
v2: fix conditional statement within [spi: stm32: fix fifo threshold level in case of short transfer]
#1 lore raw

PATCH v2 1/5 spi: stm32h7: fix race condition at end of transfer

From: Antonio Borneo <[email protected]>

The caller of stm32_spi_transfer_one(), spi_transfer_one_message(), is waiting for us to call spi_finalize_current_transfer() and will eventually schedule a new transfer, if available.
We should guarantee that the spi controller is really available before calling spi_finalize_current_transfer().

Move the call to spi_finalize_current_transfer() _after_ the call to stm32_spi_disable().

---
v2: identical to v1

 drivers/spi/spi-stm32.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/spi/spi-stm32.c +1 -1
index 4a21feae0103..814a3ec3b8ad 100644
@@ -971,8 +971,8 @@ static irqreturn_t stm32h7_spi_irq_thread(int irq, void *dev_id)
 	spin_unlock_irqrestore(&spi->lock, flags);
 
 	if (end) {
-		spi_finalize_current_transfer(master);
 		stm32h7_spi_disable(spi);
+		spi_finalize_current_transfer(master);
 	}
 
 	return IRQ_HANDLED;
Signature · 1 line
2.7.4
#2 lore raw

PATCH v2 2/5 spi: stm32: fix fifo threshold level in case of short transfer

In reply to Alain Volmat

From: Amelie Delaunay <[email protected]>

When transfer is shorter than half of the fifo, set the data packet size up to transfer size instead of up to half of the fifo.
Check also that threshold is set at least to 1 data frame.

---
v2: fix conditional statement

 drivers/spi/spi-stm32.c | 26 ++++++++++++++++++--------
 1 file changed, 18 insertions(+), 8 deletions(-)
drivers/spi/spi-stm32.c +18 -8
index 814a3ec3b8ad..e5450233f3f8 100644
@@ -467,20 +467,27 @@ static int stm32_spi_prepare_mbr(struct stm32_spi *spi, u32 speed_hz,
 /**
  * stm32h7_spi_prepare_fthlv - Determine FIFO threshold level
  * @spi: pointer to the spi controller data structure
+ * @xfer_len: length of the message to be transferred
  */
-static u32 stm32h7_spi_prepare_fthlv(struct stm32_spi *spi)
+static u32 stm32h7_spi_prepare_fthlv(struct stm32_spi *spi, u32 xfer_len)
 {
-	u32 fthlv, half_fifo;
+	u32 fthlv, half_fifo, packet;
 
 	/* data packet should not exceed 1/2 of fifo space */
 	half_fifo = (spi->fifo_size / 2);
 
+	/* data_packet should not exceed transfer length */
+	if (half_fifo > xfer_len)
+		packet = xfer_len;
+	else
+		packet = half_fifo;
+
 	if (spi->cur_bpw <= 8)
-		fthlv = half_fifo;
+		fthlv = packet;
 	else if (spi->cur_bpw <= 16)
-		fthlv = half_fifo / 2;
+		fthlv = packet / 2;
 	else
-		fthlv = half_fifo / 4;
+		fthlv = packet / 4;
 
 	/* align packet size with data registers access */
 	if (spi->cur_bpw > 8)
@@ -488,6 +495,9 @@ static u32 stm32h7_spi_prepare_fthlv(struct stm32_spi *spi)
 	else
 		fthlv -= (fthlv % 4); /* multiple of 4 */
 
+	if (!fthlv)
+		fthlv = 1;
+
 	return fthlv;
 }
 
@@ -1393,7 +1403,7 @@ static void stm32h7_spi_set_bpw(struct stm32_spi *spi)
 	cfg1_setb |= (bpw << STM32H7_SPI_CFG1_DSIZE_SHIFT) &
 		     STM32H7_SPI_CFG1_DSIZE;
 
-	spi->cur_fthlv = stm32h7_spi_prepare_fthlv(spi);
+	spi->cur_fthlv = stm32h7_spi_prepare_fthlv(spi, spi->cur_xferlen);
 	fthlv = spi->cur_fthlv - 1;
 
 	cfg1_clrb |= STM32H7_SPI_CFG1_FTHLV;
@@ -1588,6 +1598,8 @@ static int stm32_spi_transfer_one_setup(struct stm32_spi *spi,
 
 	spin_lock_irqsave(&spi->lock, flags);
 
+	spi->cur_xferlen = transfer->len;
+
 	if (spi->cur_bpw != transfer->bits_per_word) {
 		spi->cur_bpw = transfer->bits_per_word;
 		spi->cfg->set_bpw(spi);
@@ -1635,8 +1647,6 @@ static int stm32_spi_transfer_one_setup(struct stm32_spi *spi,
 			goto out;
 	}
 
-	spi->cur_xferlen = transfer->len;
-
 	dev_dbg(spi->dev, "transfer communication mode set to %d\n",
 		spi->cur_comm);
 	dev_dbg(spi->dev,
Signature · 1 line
2.7.4
#3 lore raw

PATCH v2 2/5 spi: stm32: fix fifo threshold level in case of short transfer

Quoted 3 lines

When transfer is shorter than half of the fifo, set the data packet size up to transfer size instead of up to half of the fifo.
Check also that threshold is set at least to 1 data frame.

Through a git-bisect we have identified this patch as problematic. We have an application that uses a SPI protocol with telegrams of length 2 to 16 bytes. Due to this patch we have errors in the data transfer of the MOSI direction. We use SPI in PIO mode.

Please explain what this patch should improve or what exactly is changed in the behaviour.

best regards,
Ludwig Zenz

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

#4 lore raw

PATCH v2 3/5 spi: stm32: fix stm32_spi_prepare_mbr in case of odd clk_rate

In reply to Alain Volmat

From: Amelie Delaunay <[email protected]>

Fix spi->clk_rate when it is odd to the nearest lowest even value because minimum SPI divider is 2.

---
v2: identical to v1

 drivers/spi/spi-stm32.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/spi/spi-stm32.c +2 -1
index e5450233f3f8..571dea72bf7e 100644
@@ -441,7 +441,8 @@ static int stm32_spi_prepare_mbr(struct stm32_spi *spi, u32 speed_hz,
 {
 	u32 div, mbrdiv;
 
-	div = DIV_ROUND_UP(spi->clk_rate, speed_hz);
+	/* Ensure spi->clk_rate is even */
+	div = DIV_ROUND_UP(spi->clk_rate & ~0x1, speed_hz);
 
 	/*
 	 * SPI framework set xfer->speed_hz to master->max_speed_hz if
Signature · 1 line
2.7.4
#5 lore raw

PATCH v2 4/5 spi: stm32: fixes suspend/resume management

In reply to Alain Volmat

From: Amelie Delaunay <[email protected]>

This patch adds pinctrl power management, and reconfigure spi controller in case of resume.

Fixes 038ac869c9d2 ("spi: stm32: add runtime PM support")

---
v2: identical to v1

 drivers/spi/spi-stm32.c | 27 ++++++++++++++++++++++++---
 1 file changed, 24 insertions(+), 3 deletions(-)
drivers/spi/spi-stm32.c +24 -3
index 571dea72bf7e..9b90a22543fd 100644
@@ -13,6 +13,7 @@
 #include <linux/iopoll.h>
 #include <linux/module.h>
 #include <linux/of_platform.h>
+#include <linux/pinctrl/consumer.h>
 #include <linux/pm_runtime.h>
 #include <linux/reset.h>
 #include <linux/spi/spi.h>
@@ -2007,6 +2008,8 @@ static int stm32_spi_remove(struct platform_device *pdev)
 
 	pm_runtime_disable(&pdev->dev);
 
+	pinctrl_pm_select_sleep_state(&pdev->dev);
+
 	return 0;
 }
 
@@ -2018,13 +2021,18 @@ static int stm32_spi_runtime_suspend(struct device *dev)
 
 	clk_disable_unprepare(spi->clk);
 
-	return 0;
+	return pinctrl_pm_select_sleep_state(dev);
 }
 
 static int stm32_spi_runtime_resume(struct device *dev)
 {
 	struct spi_master *master = dev_get_drvdata(dev);
 	struct stm32_spi *spi = spi_master_get_devdata(master);
+	int ret;
+
+	ret = pinctrl_pm_select_default_state(dev);
+	if (ret)
+		return ret;
 
 	return clk_prepare_enable(spi->clk);
 }
@@ -2054,10 +2062,23 @@ static int stm32_spi_resume(struct device *dev)
 		return ret;
 
 	ret = spi_master_resume(master);
-	if (ret)
+	if (ret) {
 		clk_disable_unprepare(spi->clk);
+		return ret;
+	}
 
-	return ret;
+	ret = pm_runtime_get_sync(dev);
+	if (ret) {
+		dev_err(dev, "Unable to power device:%d\n", ret);
+		return ret;
+	}
+
+	spi->cfg->config(spi);
+
+	pm_runtime_mark_last_busy(dev);
+	pm_runtime_put_autosuspend(dev);
+
+	return 0;
 }
 #endif
 
Signature · 1 line
2.7.4
#6 lore raw

PATCH v2 5/5 spi: stm32: always perform registers configuration prior to transfer

In reply to Alain Volmat

SPI registers content may have been lost upon suspend/resume sequence. So, always compute and apply the necessary configuration in stm32_spi_transfer_one_setup routine.

---
v2: identical to v1

 drivers/spi/spi-stm32.c | 42 +++++++++++++++++-------------------------
 1 file changed, 17 insertions(+), 25 deletions(-)
drivers/spi/spi-stm32.c +17 -25
index 9b90a22543fd..d4b33b358a31 100644
@@ -1597,41 +1597,33 @@ static int stm32_spi_transfer_one_setup(struct stm32_spi *spi,
 	unsigned long flags;
 	unsigned int comm_type;
 	int nb_words, ret = 0;
+	int mbr;
 
 	spin_lock_irqsave(&spi->lock, flags);
 
 	spi->cur_xferlen = transfer->len;
 
-	if (spi->cur_bpw != transfer->bits_per_word) {
-		spi->cur_bpw = transfer->bits_per_word;
-		spi->cfg->set_bpw(spi);
-	}
-
-	if (spi->cur_speed != transfer->speed_hz) {
-		int mbr;
-
-		/* Update spi->cur_speed with real clock speed */
-		mbr = stm32_spi_prepare_mbr(spi, transfer->speed_hz,
-					    spi->cfg->baud_rate_div_min,
-					    spi->cfg->baud_rate_div_max);
-		if (mbr < 0) {
-			ret = mbr;
-			goto out;
-		}
+	spi->cur_bpw = transfer->bits_per_word;
+	spi->cfg->set_bpw(spi);
 
-		transfer->speed_hz = spi->cur_speed;
-		stm32_spi_set_mbr(spi, mbr);
+	/* Update spi->cur_speed with real clock speed */
+	mbr = stm32_spi_prepare_mbr(spi, transfer->speed_hz,
+				    spi->cfg->baud_rate_div_min,
+				    spi->cfg->baud_rate_div_max);
+	if (mbr < 0) {
+		ret = mbr;
+		goto out;
 	}
 
-	comm_type = stm32_spi_communication_type(spi_dev, transfer);
-	if (spi->cur_comm != comm_type) {
-		ret = spi->cfg->set_mode(spi, comm_type);
+	transfer->speed_hz = spi->cur_speed;
+	stm32_spi_set_mbr(spi, mbr);
 
-		if (ret < 0)
-			goto out;
+	comm_type = stm32_spi_communication_type(spi_dev, transfer);
+	ret = spi->cfg->set_mode(spi, comm_type);
+	if (ret < 0)
+		goto out;
 
-		spi->cur_comm = comm_type;
-	}
+	spi->cur_comm = comm_type;
 
 	if (spi->cfg->set_data_idleness)
 		spi->cfg->set_data_idleness(spi, transfer->len);
Signature · 1 line
2.7.4
maintainer
#7 lore raw
In reply to Alain Volmat
Quoted 6 lines · Alain Volmat

This serie is a reduced version of the serie [spi: stm32: various driver enhancements] previously sent.

Alain Volmat (1):
spi: stm32: always perform registers configuration prior to transfer

Quoted 6 lines

Amelie Delaunay (3):
spi: stm32: fix fifo threshold level in case of short transfer
spi: stm32: fix stm32_spi_prepare_mbr in case of odd clk_rate
spi: stm32: fixes suspend/resume management

[...]

Applied to

https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git for-next

Thanks!

[1/5] spi: stm32h7: fix race condition at end of transfer

      commit: 135dd873d3c76d812ae64c668adef3f2c59ed27f
[2/5] spi: stm32: fix fifo threshold level in case of short transfer
      commit: 3373e9004acc0603242622b4378c64bc01d21b5f
[3/5] spi: stm32: fix stm32_spi_prepare_mbr in case of odd clk_rate
      commit: 9cc61973bf9385b19ff5dda4a2a7e265fcba85e4
[4/5] spi: stm32: fixes suspend/resume management
      commit: db96bf976a4fc65439be0b4524c0d41427d98814
[5/5] spi: stm32: always perform registers configuration prior to transfer
      commit: 60ccb3515fc61a0124c70aa37317f75b67560024

All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.

You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.

If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.

Please add any relevant lists and maintainers to the CCs when replying to this mail.

Thanks,
Mark