[PATCH v6 1/6] i2c: at91: fix a race condition when using the DMA controller

[Cyrille Pitchen]

For TX transactions, the TXCOMP bit in the Status Register is cleared
when the first data is written into the Transmit Holding Register.

In the lines from at91_do_twi_transfer():
at91_twi_write_data_dma(dev);
at91_twi_write(dev, AT91_TWI_IER, AT91_TWI_TXCOMP);

the TXCOMP interrupt may be enabled before the DMA controller has
actually started to write into the THR. In such a case, the TXCOMP bit
is still set into the Status Register so the interrupt is triggered
immediately. The driver understands that a transaction completion has
occurred but this transaction hasn't started yet. Hence the TXCOMP
interrupt is no longer enabled by at91_do_twi_transfer() but instead
by at91_twi_write_data_dma_callback().

Also, the TXCOMP bit in the Status Register in not a clear on read flag
but a snapshot of the transmission state at the time the Status
Register is read.
When a NACK error is dectected by the I2C controller, the TXCOMP, NACK
and TXRDY bits are set together to 1 in the SR. If enabled, the TXCOMP
interrupt is triggered at the same time. Also setting the TXRDY to 1
triggers the DMA controller to write the next data into the THR. Such
a write resets the TXCOMP bit to 0 in the SR. So depending on when the
interrupt handler reads the SR, it may fail to detect the NACK error
if it relies on the TXCOMP bit. The NACK bit and its interrupt should
be used instead.

For RX transactions, the TXCOMP bit in the Status Register is cleared
when the START bit is set into the Control Register. However to unify
the management of the TXCOMP bit when the DMA controller is used, the
TXCOMP interrupt is now enabled by the DMA callbacks for both TX and
RX transfers.

Signed-off-by: Cyrille Pitchen <cyrille.pitchen@xxxxxxxxx>
Cc: stable@xxxxxxxxxxxxxxx #3.10 and later
---
 drivers/i2c/busses/i2c-at91.c | 70 ++++++++++++++++++++++++++++++++-----------
 1 file changed, 53 insertions(+), 17 deletions(-)

diff --git a/drivers/i2c/busses/i2c-at91.c b/drivers/i2c/busses/i2c-at91.c
index ff23d1b..9bd10a9 100644
--- a/drivers/i2c/busses/i2c-at91.c
+++ b/drivers/i2c/busses/i2c-at91.c
@@ -65,6 +65,9 @@
 #define	AT91_TWI_UNRE		0x0080	/* Underrun Error */
 #define	AT91_TWI_NACK		0x0100	/* Not Acknowledged */
 
+#define	AT91_TWI_INT_MASK \
+	(AT91_TWI_TXCOMP | AT91_TWI_RXRDY | AT91_TWI_TXRDY | AT91_TWI_NACK)
+
 #define	AT91_TWI_IER		0x0024	/* Interrupt Enable Register */
 #define	AT91_TWI_IDR		0x0028	/* Interrupt Disable Register */
 #define	AT91_TWI_IMR		0x002c	/* Interrupt Mask Register */
@@ -119,13 +122,12 @@ static void at91_twi_write(struct at91_twi_dev *dev, unsigned reg, unsigned val)
 
 static void at91_disable_twi_interrupts(struct at91_twi_dev *dev)
 {
-	at91_twi_write(dev, AT91_TWI_IDR,
-		       AT91_TWI_TXCOMP | AT91_TWI_RXRDY | AT91_TWI_TXRDY);
+	at91_twi_write(dev, AT91_TWI_IDR, AT91_TWI_INT_MASK);
 }
 
 static void at91_twi_irq_save(struct at91_twi_dev *dev)
 {
-	dev->imr = at91_twi_read(dev, AT91_TWI_IMR) & 0x7;
+	dev->imr = at91_twi_read(dev, AT91_TWI_IMR) & AT91_TWI_INT_MASK;
 	at91_disable_twi_interrupts(dev);
 }
 
@@ -215,6 +217,14 @@ static void at91_twi_write_data_dma_callback(void *data)
 	dma_unmap_single(dev->dev, sg_dma_address(&dev->dma.sg),
 			 dev->buf_len, DMA_TO_DEVICE);
 
+	/*
+	 * When this callback is called, THR/TX FIFO is likely not to be empty
+	 * yet. So we have to wait for TXCOMP or NACK bits to be set into the
+	 * Status Register to be sure that the STOP bit has been sent and the
+	 * transfer is completed. The NACK interrupt has already been enabled,
+	 * we just have to enable TXCOMP one.
+	 */
+	at91_twi_write(dev, AT91_TWI_IER, AT91_TWI_TXCOMP);
 	at91_twi_write(dev, AT91_TWI_CR, AT91_TWI_STOP);
 }
 
@@ -309,7 +319,7 @@ static void at91_twi_read_data_dma_callback(void *data)
 	/* The last two bytes have to be read without using dma */
 	dev->buf += dev->buf_len - 2;
 	dev->buf_len = 2;
-	at91_twi_write(dev, AT91_TWI_IER, AT91_TWI_RXRDY);
+	at91_twi_write(dev, AT91_TWI_IER, AT91_TWI_RXRDY | AT91_TWI_TXCOMP);
 }
 
 static void at91_twi_read_data_dma(struct at91_twi_dev *dev)
@@ -370,7 +380,7 @@ static irqreturn_t atmel_twi_interrupt(int irq, void *dev_id)
 	/* catch error flags */
 	dev->transfer_status |= status;
 
-	if (irqstatus & AT91_TWI_TXCOMP) {
+	if (irqstatus & (AT91_TWI_TXCOMP | AT91_TWI_NACK)) {
 		at91_disable_twi_interrupts(dev);
 		complete(&dev->cmd_complete);
 	}
@@ -384,6 +394,34 @@ static int at91_do_twi_transfer(struct at91_twi_dev *dev)
 	unsigned long time_left;
 	bool has_unre_flag = dev->pdata->has_unre_flag;
 
+	/*
+	 * WARNING: the TXCOMP bit in the Status Register is NOT a clear on
+	 * read flag but shows the state of the transmission at the time the
+	 * Status Register is read. According to the programmer datasheet,
+	 * TXCOMP is set when both holding register and internal shifter are
+	 * empty and STOP condition has been sent.
+	 * Consequently, we should enable NACK interrupt rather than TXCOMP to
+	 * detect transmission failure.
+	 *
+	 * Besides, the TXCOMP bit is already set before the i2c transaction
+	 * has been started. For read transactions, this bit is cleared when
+	 * writing the START bit into the Control Register. So the
+	 * corresponding interrupt can safely be enabled just after.
+	 * However for write transactions managed by the CPU, we first write
+	 * into THR, so TXCOMP is cleared. Then we can safely enable TXCOMP
+	 * interrupt. If TXCOMP interrupt were enabled before writing into THR,
+	 * the interrupt handler would be called immediately and the i2c command
+	 * would be reported as completed.
+	 * Also when a write transaction is managed by the DMA controller,
+	 * enabling the TXCOMP interrupt in this function may lead to a race
+	 * condition since we don't know whether the TXCOMP interrupt is enabled
+	 * before or after the DMA has started to write into THR. So the TXCOMP
+	 * interrupt is enabled later by at91_twi_write_data_dma_callback().
+	 * Immediately after in that DMA callback, we still need to send the
+	 * STOP condition manually writing the corresponding bit into the
+	 * Control Register.
+	 */
+
 	dev_dbg(dev->dev, "transfer: %s %d bytes.\n",
 		(dev->msg->flags & I2C_M_RD) ? "read" : "write", dev->buf_len);
 
@@ -414,26 +452,24 @@ static int at91_do_twi_transfer(struct at91_twi_dev *dev)
 		 * seems to be the best solution.
 		 */
 		if (dev->use_dma && (dev->buf_len > AT91_I2C_DMA_THRESHOLD)) {
+			at91_twi_write(dev, AT91_TWI_IER, AT91_TWI_NACK);
 			at91_twi_read_data_dma(dev);
-			/*
-			 * It is important to enable TXCOMP irq here because
-			 * doing it only when transferring the last two bytes
-			 * will mask NACK errors since TXCOMP is set when a
-			 * NACK occurs.
-			 */
-			at91_twi_write(dev, AT91_TWI_IER,
-			       AT91_TWI_TXCOMP);
-		} else
+		} else {
 			at91_twi_write(dev, AT91_TWI_IER,
-			       AT91_TWI_TXCOMP | AT91_TWI_RXRDY);
+				       AT91_TWI_TXCOMP |
+				       AT91_TWI_NACK |
+				       AT91_TWI_RXRDY);
+		}
 	} else {
 		if (dev->use_dma && (dev->buf_len > AT91_I2C_DMA_THRESHOLD)) {
+			at91_twi_write(dev, AT91_TWI_IER, AT91_TWI_NACK);
 			at91_twi_write_data_dma(dev);
-			at91_twi_write(dev, AT91_TWI_IER, AT91_TWI_TXCOMP);
 		} else {
 			at91_twi_write_next_byte(dev);
 			at91_twi_write(dev, AT91_TWI_IER,
-				AT91_TWI_TXCOMP | AT91_TWI_TXRDY);
+				       AT91_TWI_TXCOMP |
+				       AT91_TWI_NACK |
+				       AT91_TWI_TXRDY);
 		}
 	}
 
-- 
1.8.2.2

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/