Re: [PATCH v3 2/2] i2c: aspeed: Acknowledge Tx done with and without ACK irq late

[Quan Nguyen]

On 08/12/2023 10:31, Quan Nguyen wrote:
> Commit 2be6b47211e1 ("i2c: aspeed: Acknowledge most interrupts early in
> interrupt handler") acknowledges most interrupts early before the slave
> irq handler is executed, except for the "Receive Done Interrupt status"
> which is acknowledged late in the interrupt.
> However, it has been observed that the early acknowledgment of "Transmit
> Done Interrupt Status" (with ACK or NACK) often causes the interrupt to
> be raised in READ REQUEST state, that shows the
> "Unexpected ACK on read request." complaint messages.
>
> Assuming that the "Transmit Done" interrupt should only be acknowledged
> once it is truly processed, this commit fixes that issue by acknowledging
> interrupts for both ACK and NACK cases late in the interrupt handler.
>
> Fixes: 2be6b47211e1 ("i2c: aspeed: Acknowledge most interrupts early in interrupt handler")
> Signed-off-by: Quan Nguyen <quan@xxxxxxxxxxxxxxxxxxxxxx>
> ---
> v3:
>    + Fix the unconditinal write when ack the irqs               [Andrew]
>    + Refactor the code to enhance code readability                [Quan]
>    + Fix grammar in commit message                                [Quan]
>
> v2:
>    + Split to separate series                                     [Joel]
>    + Added the Fixes line                                         [Joel]
>    + Fixed multiline comment                                      [Joel]
>    + Refactor irq clearing code                          [Joel, Guenter]
>    + Revised commit message                                       [Joel]
>    + Revised commit message                                       [Quan]
>    + About a note to remind why the readl() should immediately follow the
> writel() to fix the race condition when clearing irq status from commit
> c926c87b8e36 ("i2c: aspeed: Avoid i2c interrupt status clear race
> condition"), I think it looks straight forward in this patch and decided
> not to add that note.                                            [Joel]
>
> v1:
>    + First introduced in
> https://lore.kernel.org/all/20210519074934.20712-1-quan@xxxxxxxxxxxxxxxxxxxxxx/
> ---
>   drivers/i2c/busses/i2c-aspeed.c | 21 +++++++++++++--------
>   1 file changed, 13 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/i2c/busses/i2c-aspeed.c b/drivers/i2c/busses/i2c-aspeed.c
> index 1c2a4f4c4e1b..967a26dd4ffa 100644
> --- a/drivers/i2c/busses/i2c-aspeed.c
> +++ b/drivers/i2c/busses/i2c-aspeed.c
> @@ -617,13 +617,19 @@ static u32 aspeed_i2c_master_irq(struct aspeed_i2c_bus *bus, u32 irq_status)
>   static irqreturn_t aspeed_i2c_bus_irq(int irq, void *dev_id)
>   {
>   	struct aspeed_i2c_bus *bus = dev_id;
> -	u32 irq_received, irq_remaining, irq_handled;
> +	u32 irq_received, irq_remaining, irq_handled, irq_ack_last;
>   
>   	spin_lock(&bus->lock);
>   	irq_received = readl(bus->base + ASPEED_I2C_INTR_STS_REG);
> -	/* Ack all interrupts except for Rx done */
> -	writel(irq_received & ~ASPEED_I2CD_INTR_RX_DONE,
> -	       bus->base + ASPEED_I2C_INTR_STS_REG);
> +
> +	/*
> +	 * Early acking of INTR_RX_DONE and INTR_TX_[ACK|NAK] would indicate HW to
> +	 * start receiving or sending new data, and this may cause a race condition
> +	 * as the irq handler has not yet handled these irqs but is being acked.
> +	 * Let's ack them late at the end of the irq handler when those are truly processed.
> +	 */
> +	irq_ack_last = ASPEED_I2CD_INTR_RX_DONE | ASPEED_I2CD_INTR_TX_ACK | ASPEED_I2CD_INTR_TX_NAK;
> +	writel(irq_received & ~irq_ack_last, bus->base + ASPEED_I2C_INTR_STS_REG);

As there will be v4, I'd like to improve this line of code a bit.

During debug, there are significant number of times when the 
irq_received has the value of 0x00000004 (INTR_RX_DONE) or 0x00000001 
(INTR_TX_ACK). This makes the "irq_received & ~irq_ack_last" turn out to 
be 0.

AFAIK, it does not make sense to keep repeatedly writing the 0 to the HW 
then try to re-read it with readl() every time. I'd like to change this 
HW access conditionally to avoid those unnecessary access.

The change would looks like below:

diff --git a/drivers/i2c/busses/i2c-aspeed.c 
b/drivers/i2c/busses/i2c-aspeed.c
index 5511fd46a65e..0f67218cf04a 100644
--- a/drivers/i2c/busses/i2c-aspeed.c
+++ b/drivers/i2c/busses/i2c-aspeed.c
@@ -93,6 +93,10 @@
 		 ASPEED_I2CD_INTR_RX_DONE | \
 		 ASPEED_I2CD_INTR_TX_NAK |  \
 		 ASPEED_I2CD_INTR_TX_ACK)
+#define ASPEED_I2CD_INTR_ACK_RX_TX         \
+		(ASPEED_I2CD_INTR_RX_DONE | \
+		 ASPEED_I2CD_INTR_TX_ACK |  \
+		 ASPEED_I2CD_INTR_TX_NAK)

 /* 0x14 : I2CD Command/Status Register   */
 #define ASPEED_I2CD_SCL_LINE_STS			BIT(18)
@@ -622,10 +626,19 @@ static irqreturn_t aspeed_i2c_bus_irq(int irq, 
void *dev_id)

 	spin_lock(&bus->lock);
 	irq_received = readl(bus->base + ASPEED_I2C_INTR_STS_REG);
-	/* Ack all interrupts except for Rx done */
-	writel(irq_received & ~ASPEED_I2CD_INTR_RX_DONE,
-	       bus->base + ASPEED_I2C_INTR_STS_REG);
-	readl(bus->base + ASPEED_I2C_INTR_STS_REG);
+
+	/*
+	 * Early acking of INTR_RX_DONE and INTR_TX_[ACK|NAK] would
+        ...
+	 */
+	if (irq_received & ~ASPEED_I2CD_INTR_ACK_RX_TX) {
+		writel(irq_received & ~ASPEED_I2CD_INTR_ACK_RX_TX,
+		       bus->base + ASPEED_I2C_INTR_STS_REG);
+		readl(bus->base + ASPEED_I2C_INTR_STS_REG);
+	}
+
 	irq_received &= ASPEED_I2CD_INTR_RECV_MASK;
 	irq_remaining = irq_received;

Thanks,
- Quan