[PATCH 1/1] i3c: master: svc: skip address resend on repeat START

[PATCH 1/1] i3c: master: svc: skip address resend on repeat START

[Frank Li]

According to the I3C specification, address arbitration only happens during
the START. Repeated START do not initiate arbitration, and In-Band
Interrupts (IBIs) cannot occur at this stage.

Resending the address upon a NACK in a repeat START is therefore redundant
and unnecessary. Avoid redundant retries, improving efficiency and ensuring
protocol compliance.

Signed-off-by: Frank Li <Frank.Li@nxp.com>
---
 drivers/i3c/master/svc-i3c-master.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/i3c/master/svc-i3c-master.c b/drivers/i3c/master/svc-i3c-master.c
index 85e16de208d3b..5212c2cc6855f 100644
--- a/drivers/i3c/master/svc-i3c-master.c
+++ b/drivers/i3c/master/svc-i3c-master.c
@@ -1281,9 +1281,9 @@ static int svc_i3c_master_write(struct svc_i3c_master *master,
 static int svc_i3c_master_xfer(struct svc_i3c_master *master,
 			       bool rnw, unsigned int xfer_type, u8 addr,
 			       u8 *in, const u8 *out, unsigned int xfer_len,
-			       unsigned int *actual_len, bool continued)
+			       unsigned int *actual_len, bool continued, bool repeat_start)
 {
-	int retry = 2;
+	int retry = repeat_start ? 1 : 2;
 	u32 reg;
 	int ret;
 
@@ -1468,7 +1468,7 @@ static void svc_i3c_master_start_xfer_locked(struct svc_i3c_master *master)
 		ret = svc_i3c_master_xfer(master, cmd->rnw, xfer->type,
 					  cmd->addr, cmd->in, cmd->out,
 					  cmd->len, &cmd->actual_len,
-					  cmd->continued);
+					  cmd->continued, !!i);
 		/* cmd->xfer is NULL if I2C or CCC transfer */
 		if (cmd->xfer)
 			cmd->xfer->actual_len = cmd->actual_len;
-- 
2.34.1


-- 
linux-i3c mailing list
linux-i3c@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-i3c

Re: [PATCH 1/1] i3c: master: svc: skip address resend on repeat START

[Miquel Raynal]

On 16/04/2025 at 12:47:50 -04, Frank Li <Frank.Li@nxp.com> wrote:

> According to the I3C specification, address arbitration only happens during
> the START. Repeated START do not initiate arbitration, and In-Band
> Interrupts (IBIs) cannot occur at this stage.
>
> Resending the address upon a NACK in a repeat START is therefore redundant
> and unnecessary. Avoid redundant retries, improving efficiency and ensuring
> protocol compliance.
>
> Signed-off-by: Frank Li <Frank.Li@nxp.com>
> ---
>  drivers/i3c/master/svc-i3c-master.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/i3c/master/svc-i3c-master.c b/drivers/i3c/master/svc-i3c-master.c
> index 85e16de208d3b..5212c2cc6855f 100644
> --- a/drivers/i3c/master/svc-i3c-master.c
> +++ b/drivers/i3c/master/svc-i3c-master.c
> @@ -1281,9 +1281,9 @@ static int svc_i3c_master_write(struct svc_i3c_master *master,
>  static int svc_i3c_master_xfer(struct svc_i3c_master *master,
>  			       bool rnw, unsigned int xfer_type, u8 addr,
>  			       u8 *in, const u8 *out, unsigned int xfer_len,
> -			       unsigned int *actual_len, bool continued)
> +			       unsigned int *actual_len, bool continued, bool repeat_start)
>  {
> -	int retry = 2;
> +	int retry = repeat_start ? 1 : 2;
>  	u32 reg;
>  	int ret;
>  
> @@ -1468,7 +1468,7 @@ static void svc_i3c_master_start_xfer_locked(struct svc_i3c_master *master)
>  		ret = svc_i3c_master_xfer(master, cmd->rnw, xfer->type,
>  					  cmd->addr, cmd->in, cmd->out,
>  					  cmd->len, &cmd->actual_len,
> -					  cmd->continued);
> +					  cmd->continued, !!i);

Could we just pass 'i' here? Or maybe 'i > 0' which might be even clearer.

>  		/* cmd->xfer is NULL if I2C or CCC transfer */
>  		if (cmd->xfer)
>  			cmd->xfer->actual_len = cmd->actual_len;

Thanks,
Miquèl

-- 
linux-i3c mailing list
linux-i3c@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-i3c