[PATCH v4 0/3] Fix STM32 I2C dma operations

[PATCH v4 0/3] Fix STM32 I2C dma operations

[Clément Le Goffic]

This patch series aims to fix some issues inside the driver's DMA
handling.
It also uses newer I2C DMA API.

Signed-off-by: Clément Le Goffic <clement.legoffic@foss.st.com>
---
Changes in v4:
- Patch[1]: Remove all `chan_dev` variable occurrencies
- Patch[2]:
    - Refine commit message
    - Use the dma_callback to factorize the code
- Patch[3]: Refine commit message
- Link to v3: https://lore.kernel.org/r/20250630-i2c-upstream-v3-0-7a23ab26683a@foss.st.com

Changes in v3:
- Add Alain Volmat's "Acked-by" on patch 1 and 2
- Link to v2: https://lore.kernel.org/r/20250627-i2c-upstream-v2-0-8c14523481dc@foss.st.com

Changes in v2:
- Fix the dev used in dma_unmap also in the error path of
  `stm32_i2c_prep_dma_xfer`
- Add a dma_unmap_single also in the ITs error handler
- Add Alain Volmat's "Acked-by" on patch 3
- Link to v1: https://lore.kernel.org/r/20250616-i2c-upstream-v1-0-42d3d5374e65@foss.st.com

---
Clément Le Goffic (3):
      i2c: stm32: fix the device used for the DMA map
      i2c: stm32f7: unmap DMA mapped buffer
      i2c: stm32f7: support i2c_*_dma_safe_msg_buf APIs

 drivers/i2c/busses/i2c-stm32.c   |  8 +++---
 drivers/i2c/busses/i2c-stm32f7.c | 56 +++++++++++++++++++++-------------------
 2 files changed, 33 insertions(+), 31 deletions(-)
---
base-commit: d0b3b7b22dfa1f4b515fd3a295b3fd958f9e81af
change-id: 20250527-i2c-upstream-e5b69501359a

Best regards,
--  
Clément Le Goffic <clement.legoffic@foss.st.com>

[PATCH v4 1/3] i2c: stm32: fix the device used for the DMA map

[Clément Le Goffic]

If the DMA mapping failed, it produced an error log with the wrong
device name:
"stm32-dma3 40400000.dma-controller: rejecting DMA map of vmalloc memory"
Fix this issue by replacing the dev with the I2C dev.

Fixes: bb8822cbbc53 ("i2c: i2c-stm32: Add generic DMA API")
Acked-by: Alain Volmat <alain.volmat@foss.st.com>
Signed-off-by: Clément Le Goffic <clement.legoffic@foss.st.com>
---
 drivers/i2c/busses/i2c-stm32.c   | 8 +++-----
 drivers/i2c/busses/i2c-stm32f7.c | 4 ++--
 2 files changed, 5 insertions(+), 7 deletions(-)

diff --git a/drivers/i2c/busses/i2c-stm32.c b/drivers/i2c/busses/i2c-stm32.c
index 157c64e27d0b..f84ec056e36d 100644
--- a/drivers/i2c/busses/i2c-stm32.c
+++ b/drivers/i2c/busses/i2c-stm32.c
@@ -102,7 +102,6 @@ int stm32_i2c_prep_dma_xfer(struct device *dev, struct stm32_i2c_dma *dma,
 			    void *dma_async_param)
 {
 	struct dma_async_tx_descriptor *txdesc;
-	struct device *chan_dev;
 	int ret;
 
 	if (rd_wr) {
@@ -116,11 +115,10 @@ int stm32_i2c_prep_dma_xfer(struct device *dev, struct stm32_i2c_dma *dma,
 	}
 
 	dma->dma_len = len;
-	chan_dev = dma->chan_using->device->dev;
 
-	dma->dma_buf = dma_map_single(chan_dev, buf, dma->dma_len,
+	dma->dma_buf = dma_map_single(dev, buf, dma->dma_len,
 				      dma->dma_data_dir);
-	if (dma_mapping_error(chan_dev, dma->dma_buf)) {
+	if (dma_mapping_error(dev, dma->dma_buf)) {
 		dev_err(dev, "DMA mapping failed\n");
 		return -EINVAL;
 	}
@@ -150,7 +148,7 @@ int stm32_i2c_prep_dma_xfer(struct device *dev, struct stm32_i2c_dma *dma,
 	return 0;
 
 err:
-	dma_unmap_single(chan_dev, dma->dma_buf, dma->dma_len,
+	dma_unmap_single(dev, dma->dma_buf, dma->dma_len,
 			 dma->dma_data_dir);
 	return ret;
 }
diff --git a/drivers/i2c/busses/i2c-stm32f7.c b/drivers/i2c/busses/i2c-stm32f7.c
index e4aaeb2262d0..817d081460c2 100644
--- a/drivers/i2c/busses/i2c-stm32f7.c
+++ b/drivers/i2c/busses/i2c-stm32f7.c
@@ -741,10 +741,10 @@ static void stm32f7_i2c_dma_callback(void *arg)
 {
 	struct stm32f7_i2c_dev *i2c_dev = (struct stm32f7_i2c_dev *)arg;
 	struct stm32_i2c_dma *dma = i2c_dev->dma;
-	struct device *dev = dma->chan_using->device->dev;
 
 	stm32f7_i2c_disable_dma_req(i2c_dev);
-	dma_unmap_single(dev, dma->dma_buf, dma->dma_len, dma->dma_data_dir);
+	dma_unmap_single(i2c_dev->dev, dma->dma_buf, dma->dma_len,
+			 dma->dma_data_dir);
 	complete(&dma->dma_complete);
 }
 

-- 
2.43.0

[PATCH v4 2/3] i2c: stm32f7: unmap DMA mapped buffer

[Clément Le Goffic]

Before each I2C transfer using DMA, the I2C buffer is DMA'pped to make
sure the memory buffer is DMA'able. This is handle in the function
`stm32_i2c_prep_dma_xfer()`.
If the transfer fails for any reason the I2C buffer must be unmap.
Use the dma_callback to factorize the code and fix this issue.

Note that the `stm32f7_i2c_dma_callback()` is now called in case of DMA
transfer success and error and that the `complete()` on the dma_complete
completion structure is done inconditionnally in case of transfer
success or error as well as the `dmaengine_terminate_async()`.
This is allowed as a `complete()` in case transfer error has no effect
as well as a `dmaengine_terminate_async()` on a transfer success.

Also fix the unneeded cast and remove not more needed variables.

Fixes: 7ecc8cfde553 ("i2c: i2c-stm32f7: Add DMA support")
Acked-by: Alain Volmat <alain.volmat@foss.st.com>
Signed-off-by: Clément Le Goffic <clement.legoffic@foss.st.com>
---
 drivers/i2c/busses/i2c-stm32f7.c | 20 +++++++-------------
 1 file changed, 7 insertions(+), 13 deletions(-)

diff --git a/drivers/i2c/busses/i2c-stm32f7.c b/drivers/i2c/busses/i2c-stm32f7.c
index 817d081460c2..73a7b8894c0d 100644
--- a/drivers/i2c/busses/i2c-stm32f7.c
+++ b/drivers/i2c/busses/i2c-stm32f7.c
@@ -739,10 +739,11 @@ static void stm32f7_i2c_disable_dma_req(struct stm32f7_i2c_dev *i2c_dev)
 
 static void stm32f7_i2c_dma_callback(void *arg)
 {
-	struct stm32f7_i2c_dev *i2c_dev = (struct stm32f7_i2c_dev *)arg;
+	struct stm32f7_i2c_dev *i2c_dev = arg;
 	struct stm32_i2c_dma *dma = i2c_dev->dma;
 
 	stm32f7_i2c_disable_dma_req(i2c_dev);
+	dmaengine_terminate_async(dma->chan_using);
 	dma_unmap_single(i2c_dev->dev, dma->dma_buf, dma->dma_len,
 			 dma->dma_data_dir);
 	complete(&dma->dma_complete);
@@ -1510,7 +1511,6 @@ static irqreturn_t stm32f7_i2c_handle_isr_errs(struct stm32f7_i2c_dev *i2c_dev,
 	u16 addr = f7_msg->addr;
 	void __iomem *base = i2c_dev->base;
 	struct device *dev = i2c_dev->dev;
-	struct stm32_i2c_dma *dma = i2c_dev->dma;
 
 	/* Bus error */
 	if (status & STM32F7_I2C_ISR_BERR) {
@@ -1551,10 +1551,8 @@ static irqreturn_t stm32f7_i2c_handle_isr_errs(struct stm32f7_i2c_dev *i2c_dev,
 	}
 
 	/* Disable dma */
-	if (i2c_dev->use_dma) {
-		stm32f7_i2c_disable_dma_req(i2c_dev);
-		dmaengine_terminate_async(dma->chan_using);
-	}
+	if (i2c_dev->use_dma)
+		stm32f7_i2c_dma_callback(i2c_dev);
 
 	i2c_dev->master_mode = false;
 	complete(&i2c_dev->complete);
@@ -1600,7 +1598,6 @@ static irqreturn_t stm32f7_i2c_isr_event_thread(int irq, void *data)
 {
 	struct stm32f7_i2c_dev *i2c_dev = data;
 	struct stm32f7_i2c_msg *f7_msg = &i2c_dev->f7_msg;
-	struct stm32_i2c_dma *dma = i2c_dev->dma;
 	void __iomem *base = i2c_dev->base;
 	u32 status, mask;
 	int ret;
@@ -1619,10 +1616,8 @@ static irqreturn_t stm32f7_i2c_isr_event_thread(int irq, void *data)
 		dev_dbg(i2c_dev->dev, "<%s>: Receive NACK (addr %x)\n",
 			__func__, f7_msg->addr);
 		writel_relaxed(STM32F7_I2C_ICR_NACKCF, base + STM32F7_I2C_ICR);
-		if (i2c_dev->use_dma) {
-			stm32f7_i2c_disable_dma_req(i2c_dev);
-			dmaengine_terminate_async(dma->chan_using);
-		}
+		if (i2c_dev->use_dma)
+			stm32f7_i2c_dma_callback(i2c_dev);
 		f7_msg->result = -ENXIO;
 	}
 
@@ -1640,8 +1635,7 @@ static irqreturn_t stm32f7_i2c_isr_event_thread(int irq, void *data)
 			ret = wait_for_completion_timeout(&i2c_dev->dma->dma_complete, HZ);
 			if (!ret) {
 				dev_dbg(i2c_dev->dev, "<%s>: Timed out\n", __func__);
-				stm32f7_i2c_disable_dma_req(i2c_dev);
-				dmaengine_terminate_async(dma->chan_using);
+				stm32f7_i2c_dma_callback(i2c_dev);
 				f7_msg->result = -ETIMEDOUT;
 			}
 		}

-- 
2.43.0

[PATCH v4 3/3] i2c: stm32f7: support i2c_*_dma_safe_msg_buf APIs

[Clément Le Goffic]

`i2c_*_dma_safe_msg_buf` APIs operate on a `struct i2c_msg`.
The get operation make sure the I2C buffer is DMA'able according to its
buffer length, or if the memory use is DMA coherent for example and
return a valid pointer for safe DMA access to be used.
The put operation release the pointer.
Prefer using generic API's than relying on private tests.

Acked-by: Alain Volmat <alain.volmat@foss.st.com>
Signed-off-by: Clément Le Goffic <clement.legoffic@foss.st.com>
---
 drivers/i2c/busses/i2c-stm32f7.c | 32 +++++++++++++++++++++-----------
 1 file changed, 21 insertions(+), 11 deletions(-)

diff --git a/drivers/i2c/busses/i2c-stm32f7.c b/drivers/i2c/busses/i2c-stm32f7.c
index 73a7b8894c0d..994232646789 100644
--- a/drivers/i2c/busses/i2c-stm32f7.c
+++ b/drivers/i2c/busses/i2c-stm32f7.c
@@ -741,11 +741,14 @@ static void stm32f7_i2c_dma_callback(void *arg)
 {
 	struct stm32f7_i2c_dev *i2c_dev = arg;
 	struct stm32_i2c_dma *dma = i2c_dev->dma;
+	struct stm32f7_i2c_msg *f7_msg = &i2c_dev->f7_msg;
 
 	stm32f7_i2c_disable_dma_req(i2c_dev);
 	dmaengine_terminate_async(dma->chan_using);
 	dma_unmap_single(i2c_dev->dev, dma->dma_buf, dma->dma_len,
 			 dma->dma_data_dir);
+	if (!f7_msg->smbus)
+		i2c_put_dma_safe_msg_buf(f7_msg->buf, i2c_dev->msg, true);
 	complete(&dma->dma_complete);
 }
 
@@ -881,6 +884,7 @@ static void stm32f7_i2c_xfer_msg(struct stm32f7_i2c_dev *i2c_dev,
 {
 	struct stm32f7_i2c_msg *f7_msg = &i2c_dev->f7_msg;
 	void __iomem *base = i2c_dev->base;
+	u8 *dma_buf;
 	u32 cr1, cr2;
 	int ret;
 
@@ -930,17 +934,23 @@ static void stm32f7_i2c_xfer_msg(struct stm32f7_i2c_dev *i2c_dev,
 
 	/* Configure DMA or enable RX/TX interrupt */
 	i2c_dev->use_dma = false;
-	if (i2c_dev->dma && f7_msg->count >= STM32F7_I2C_DMA_LEN_MIN
-	    && !i2c_dev->atomic) {
-		ret = stm32_i2c_prep_dma_xfer(i2c_dev->dev, i2c_dev->dma,
-					      msg->flags & I2C_M_RD,
-					      f7_msg->count, f7_msg->buf,
-					      stm32f7_i2c_dma_callback,
-					      i2c_dev);
-		if (!ret)
-			i2c_dev->use_dma = true;
-		else
-			dev_warn(i2c_dev->dev, "can't use DMA\n");
+	if (i2c_dev->dma && !i2c_dev->atomic) {
+		dma_buf = i2c_get_dma_safe_msg_buf(msg, STM32F7_I2C_DMA_LEN_MIN);
+		if (dma_buf) {
+			f7_msg->buf = dma_buf;
+			ret = stm32_i2c_prep_dma_xfer(i2c_dev->dev, i2c_dev->dma,
+						      msg->flags & I2C_M_RD,
+						      f7_msg->count, f7_msg->buf,
+						      stm32f7_i2c_dma_callback,
+						      i2c_dev);
+			if (ret) {
+				dev_warn(i2c_dev->dev, "can't use DMA\n");
+				i2c_put_dma_safe_msg_buf(f7_msg->buf, msg, false);
+				f7_msg->buf = msg->buf;
+			} else {
+				i2c_dev->use_dma = true;
+			}
+		}
 	}
 
 	if (!i2c_dev->use_dma) {

-- 
2.43.0

Re: [PATCH v4 0/3] Fix STM32 I2C dma operations

[Alain Volmat]

Hi Clément,

On Fri, Jul 04, 2025 at 10:39:13AM +0200, Clément Le Goffic wrote:
> This patch series aims to fix some issues inside the driver's DMA
> handling.
> It also uses newer I2C DMA API.
> 
> Signed-off-by: Clément Le Goffic <clement.legoffic@foss.st.com>
> ---
> Changes in v4:
> - Patch[1]: Remove all `chan_dev` variable occurrencies
> - Patch[2]:
>     - Refine commit message
>     - Use the dma_callback to factorize the code
> - Patch[3]: Refine commit message
> - Link to v3: https://lore.kernel.org/r/20250630-i2c-upstream-v3-0-7a23ab26683a@foss.st.com
> 
> Changes in v3:
> - Add Alain Volmat's "Acked-by" on patch 1 and 2
> - Link to v2: https://lore.kernel.org/r/20250627-i2c-upstream-v2-0-8c14523481dc@foss.st.com
> 
> Changes in v2:
> - Fix the dev used in dma_unmap also in the error path of
>   `stm32_i2c_prep_dma_xfer`
> - Add a dma_unmap_single also in the ITs error handler
> - Add Alain Volmat's "Acked-by" on patch 3
> - Link to v1: https://lore.kernel.org/r/20250616-i2c-upstream-v1-0-42d3d5374e65@foss.st.com
> 
> ---
> Clément Le Goffic (3):
>       i2c: stm32: fix the device used for the DMA map
>       i2c: stm32f7: unmap DMA mapped buffer
>       i2c: stm32f7: support i2c_*_dma_safe_msg_buf APIs
> 
>  drivers/i2c/busses/i2c-stm32.c   |  8 +++---
>  drivers/i2c/busses/i2c-stm32f7.c | 56 +++++++++++++++++++++-------------------
>  2 files changed, 33 insertions(+), 31 deletions(-)
> ---

Thanks for this new version of the serie.
This all looks good to me. My Acked-by are already set since v3 so
nothing more from me.

Regards,
Alain

> base-commit: d0b3b7b22dfa1f4b515fd3a295b3fd958f9e81af
> change-id: 20250527-i2c-upstream-e5b69501359a
> 
> Best regards,
> --  
> Clément Le Goffic <clement.legoffic@foss.st.com>
> 

Re: [PATCH v4 0/3] Fix STM32 I2C dma operations

[Andi Shyti]

Hi Clement,

Thanks for following up on the reviews.

> Clément Le Goffic (3):
>       i2c: stm32: fix the device used for the DMA map
>       i2c: stm32f7: unmap DMA mapped buffer

I applied the two above in i2c/i2c-host-fixes. I'm not sure we
need the "Fixes:" tag in the first patch, though, as we are not
fixing a real bug. But I'm keeping it there for the time being.

>       i2c: stm32f7: support i2c_*_dma_safe_msg_buf APIs

This one depends on the previous two to be appliex, so that for
now I added it in i2c/i2c-host-next. I will place it in the
proper branch after the weekly pull request.

Thanks,
Andi