[PATCH v4 1/3] i2c: imx: Fix race condition in dma read

[PATCH v4 1/3] i2c: imx: Fix race condition in dma read

[Esben Haabendal]

From: Esben Haabendal <eha@deif.com>

This fixes a race condition, where the DMAEN bit ends up being set after
I2C slave has transmitted a byte following the dummy read.  When that
happens, an interrupt is generated instead, and no DMA request is generated
to kickstart the DMA read, and a timeout happens after DMA_TIMEOUT (1 sec).

Fixed by setting the DMAEN bit before the dummy read.

Signed-off-by: Esben Haabendal <eha@deif.com>
Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
---
 drivers/i2c/busses/i2c-imx.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/i2c/busses/i2c-imx.c b/drivers/i2c/busses/i2c-imx.c
index 498c5e891649..ad6adefb64da 100644
--- a/drivers/i2c/busses/i2c-imx.c
+++ b/drivers/i2c/busses/i2c-imx.c
@@ -668,9 +668,6 @@ static int i2c_imx_dma_read(struct imx_i2c_struct *i2c_imx,
 	struct imx_i2c_dma *dma = i2c_imx->dma;
 	struct device *dev = &i2c_imx->adapter.dev;
 
-	temp = imx_i2c_read_reg(i2c_imx, IMX_I2C_I2CR);
-	temp |= I2CR_DMAEN;
-	imx_i2c_write_reg(temp, i2c_imx, IMX_I2C_I2CR);
 
 	dma->chan_using = dma->chan_rx;
 	dma->dma_transfer_dir = DMA_DEV_TO_MEM;
@@ -783,6 +780,7 @@ static int i2c_imx_read(struct imx_i2c_struct *i2c_imx, struct i2c_msg *msgs, bo
 	int i, result;
 	unsigned int temp;
 	int block_data = msgs->flags & I2C_M_RECV_LEN;
+	int use_dma = i2c_imx->dma && msgs->len >= DMA_THRESHOLD && !block_data;
 
 	dev_dbg(&i2c_imx->adapter.dev,
 		"<%s> write slave address: addr=0x%x\n",
@@ -809,12 +807,14 @@ static int i2c_imx_read(struct imx_i2c_struct *i2c_imx, struct i2c_msg *msgs, bo
 	 */
 	if ((msgs->len - 1) || block_data)
 		temp &= ~I2CR_TXAK;
+	if (use_dma)
+		temp |= I2CR_DMAEN;
 	imx_i2c_write_reg(temp, i2c_imx, IMX_I2C_I2CR);
 	imx_i2c_read_reg(i2c_imx, IMX_I2C_I2DR); /* dummy read */
 
 	dev_dbg(&i2c_imx->adapter.dev, "<%s> read data\n", __func__);
 
-	if (i2c_imx->dma && msgs->len >= DMA_THRESHOLD && !block_data)
+	if (use_dma)
 		return i2c_imx_dma_read(i2c_imx, msgs, is_lastmsg);
 
 	/* read data */
-- 
2.18.0

Re: [PATCH v4 1/3] i2c: imx: Fix race condition in dma read

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 646 bytes --]

On Thu, Aug 16, 2018 at 10:43:12AM +0200, Esben Haabendal wrote:
> From: Esben Haabendal <eha@deif.com>
> 
> This fixes a race condition, where the DMAEN bit ends up being set after
> I2C slave has transmitted a byte following the dummy read.  When that
> happens, an interrupt is generated instead, and no DMA request is generated
> to kickstart the DMA read, and a timeout happens after DMA_TIMEOUT (1 sec).
> 
> Fixed by setting the DMAEN bit before the dummy read.
> 
> Signed-off-by: Esben Haabendal <eha@deif.com>
> Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>

Applied to for-next and added stable, thanks!


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

[PATCH v4 2/3] i2c: imx: Simplify stopped state tracking

[Esben Haabendal]

From: Esben Haabendal <eha@deif.com>

Always update the stopped state when busy status have been checked.
This is identical to what was done before, with the exception of error
handling.
Without this change, some errors cause the stopped state to be left in
incorrect state in i2c_imx_stop(), i2c_imx_dma_read(), i2c_imx_read() and
i2c_imx_xfer().

Signed-off-by: Esben Haabendal <eha@deif.com>
Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
---
 drivers/i2c/busses/i2c-imx.c | 15 +++++++--------
 1 file changed, 7 insertions(+), 8 deletions(-)

diff --git a/drivers/i2c/busses/i2c-imx.c b/drivers/i2c/busses/i2c-imx.c
index ad6adefb64da..c406700789e1 100644
--- a/drivers/i2c/busses/i2c-imx.c
+++ b/drivers/i2c/busses/i2c-imx.c
@@ -421,10 +421,14 @@ static int i2c_imx_bus_busy(struct imx_i2c_struct *i2c_imx, int for_busy)
 			return -EAGAIN;
 		}
 
-		if (for_busy && (temp & I2SR_IBB))
+		if (for_busy && (temp & I2SR_IBB)) {
+			i2c_imx->stopped = 0;
 			break;
-		if (!for_busy && !(temp & I2SR_IBB))
+		}
+		if (!for_busy && !(temp & I2SR_IBB)) {
+			i2c_imx->stopped = 1;
 			break;
+		}
 		if (time_after(jiffies, orig_jiffies + msecs_to_jiffies(500))) {
 			dev_dbg(&i2c_imx->adapter.dev,
 				"<%s> I2C bus is busy\n", __func__);
@@ -538,7 +542,6 @@ static int i2c_imx_start(struct imx_i2c_struct *i2c_imx)
 	result = i2c_imx_bus_busy(i2c_imx, 1);
 	if (result)
 		return result;
-	i2c_imx->stopped = 0;
 
 	temp |= I2CR_IIEN | I2CR_MTX | I2CR_TXAK;
 	temp &= ~I2CR_DMAEN;
@@ -567,10 +570,8 @@ static void i2c_imx_stop(struct imx_i2c_struct *i2c_imx)
 		udelay(i2c_imx->disable_delay);
 	}
 
-	if (!i2c_imx->stopped) {
+	if (!i2c_imx->stopped)
 		i2c_imx_bus_busy(i2c_imx, 0);
-		i2c_imx->stopped = 1;
-	}
 
 	/* Disable I2C controller */
 	temp = i2c_imx->hwdata->i2cr_ien_opcode ^ I2CR_IEN,
@@ -724,7 +725,6 @@ static int i2c_imx_dma_read(struct imx_i2c_struct *i2c_imx,
 		temp &= ~(I2CR_MSTA | I2CR_MTX);
 		imx_i2c_write_reg(temp, i2c_imx, IMX_I2C_I2CR);
 		i2c_imx_bus_busy(i2c_imx, 0);
-		i2c_imx->stopped = 1;
 	} else {
 		/*
 		 * For i2c master receiver repeat restart operation like:
@@ -850,7 +850,6 @@ static int i2c_imx_read(struct imx_i2c_struct *i2c_imx, struct i2c_msg *msgs, bo
 				temp &= ~(I2CR_MSTA | I2CR_MTX);
 				imx_i2c_write_reg(temp, i2c_imx, IMX_I2C_I2CR);
 				i2c_imx_bus_busy(i2c_imx, 0);
-				i2c_imx->stopped = 1;
 			} else {
 				/*
 				 * For i2c master receiver repeat restart operation like:
-- 
2.18.0

Re: [PATCH v4 2/3] i2c: imx: Simplify stopped state tracking

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 686 bytes --]

On Thu, Aug 16, 2018 at 10:43:13AM +0200, Esben Haabendal wrote:
> From: Esben Haabendal <eha@deif.com>
> 
> Always update the stopped state when busy status have been checked.
> This is identical to what was done before, with the exception of error
> handling.
> Without this change, some errors cause the stopped state to be left in
> incorrect state in i2c_imx_stop(), i2c_imx_dma_read(), i2c_imx_read() and
> i2c_imx_xfer().
> 
> Signed-off-by: Esben Haabendal <eha@deif.com>
> Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>

Applied to for-next, thanks! I did not add stable because the cover
latter said minor issue? Not really sure here, though...


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH v4 2/3] i2c: imx: Simplify stopped state tracking

[Esben Haabendal]

Wolfram Sang <wsa@the-dreams.de> writes:

> On Thu, Aug 16, 2018 at 10:43:13AM +0200, Esben Haabendal wrote:
>> From: Esben Haabendal <eha@deif.com>
>> 
>> Always update the stopped state when busy status have been checked.
>> This is identical to what was done before, with the exception of error
>> handling.
>> Without this change, some errors cause the stopped state to be left in
>> incorrect state in i2c_imx_stop(), i2c_imx_dma_read(), i2c_imx_read() and
>> i2c_imx_xfer().
>> 
>> Signed-off-by: Esben Haabendal <eha@deif.com>
>> Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
>
> Applied to for-next, thanks! I did not add stable because the cover
> latter said minor issue? Not really sure here, though...

I agree, it does not need to go into stable.

/Esben

[PATCH v4 3/3] arm: dts: ls1021a: Enable I2C DMA support

[Esben Haabendal]

From: Esben Haabendal <eha@deif.com>

Gives substantial performance improvement for transfers larger than 16
bytes (DMA_THRESHOLD).  Smaller transfers are unaffected.

Signed-off-by: Esben Haabendal <eha@deif.com>
---
 arch/arm/boot/dts/ls1021a.dtsi | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/arch/arm/boot/dts/ls1021a.dtsi b/arch/arm/boot/dts/ls1021a.dtsi
index c55d479971cc..1e5640701c65 100644
--- a/arch/arm/boot/dts/ls1021a.dtsi
+++ b/arch/arm/boot/dts/ls1021a.dtsi
@@ -363,6 +363,8 @@
 			interrupts = <GIC_SPI 88 IRQ_TYPE_LEVEL_HIGH>;
 			clock-names = "i2c";
 			clocks = <&clockgen 4 1>;
+			dma-names = "tx", "rx";
+			dmas = <&edma0 1 39>, <&edma0 1 38>;
 			status = "disabled";
 		};
 
@@ -374,6 +376,8 @@
 			interrupts = <GIC_SPI 89 IRQ_TYPE_LEVEL_HIGH>;
 			clock-names = "i2c";
 			clocks = <&clockgen 4 1>;
+			dma-names = "tx", "rx";
+			dmas = <&edma0 1 37>, <&edma0 1 36>;
 			status = "disabled";
 		};
 
@@ -385,6 +389,8 @@
 			interrupts = <GIC_SPI 90 IRQ_TYPE_LEVEL_HIGH>;
 			clock-names = "i2c";
 			clocks = <&clockgen 4 1>;
+			dma-names = "tx", "rx";
+			dmas = <&edma0 1 35>, <&edma0 1 34>;
 			status = "disabled";
 		};
 
-- 
2.18.0

Re: [PATCH v4 3/3] arm: dts: ls1021a: Enable I2C DMA support

[Uwe Kleine-König]

On Thu, Aug 16, 2018 at 10:43:14AM +0200, Esben Haabendal wrote:
> From: Esben Haabendal <eha@deif.com>
> 
> Gives substantial performance improvement for transfers larger than 16
> bytes (DMA_THRESHOLD).  Smaller transfers are unaffected.
> 
> Signed-off-by: Esben Haabendal <eha@deif.com>

Looks reasonable to me.

Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |

Re: [PATCH v4 3/3] arm: dts: ls1021a: Enable I2C DMA support

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 411 bytes --]

On Thu, Aug 16, 2018 at 10:43:14AM +0200, Esben Haabendal wrote:
> From: Esben Haabendal <eha@deif.com>
> 
> Gives substantial performance improvement for transfers larger than 16
> bytes (DMA_THRESHOLD).  Smaller transfers are unaffected.
> 
> Signed-off-by: Esben Haabendal <eha@deif.com>

This needs to go via arm-soc. The two other patches will be in v4.19
with my pull request later this week.


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [PATCH v4 3/3] arm: dts: ls1021a: Enable I2C DMA support

[Esben Haabendal]

Wolfram Sang <wsa@the-dreams.de> writes:

> On Thu, Aug 16, 2018 at 10:43:14AM +0200, Esben Haabendal wrote:
>> From: Esben Haabendal <eha@deif.com>
>> 
>> Gives substantial performance improvement for transfers larger than 16
>> bytes (DMA_THRESHOLD).  Smaller transfers are unaffected.
>> 
>> Signed-off-by: Esben Haabendal <eha@deif.com>
>
> This needs to go via arm-soc. The two other patches will be in v4.19
> with my pull request later this week.

Ok, do I need to send to additional list(s)?
linux-arm-kernel@lists.infradead.org?

/Esben

Re: [PATCH v4 3/3] arm: dts: ls1021a: Enable I2C DMA support

[Wolfram Sang]

[-- Attachment #1: Type: text/plain, Size: 171 bytes --]


> Ok, do I need to send to additional list(s)?
> linux-arm-kernel@lists.infradead.org?

I'd say resend this patch after rc1 is out and the other 2 patches are
upstream.


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]