[v1] octeontx2-af: CN10KB: Fix FIFO length calculation for RPM2

[net PATCH] octeontx2-af: CN10KB: Fix FIFO length calculation for RPM2

Posted by Naveen Mamindlapalli 1 year, 11 months ago

From: Nithin Dabilpuram <ndabilpuram@marvell.com>

RPM0 and RPM1 on the CN10KB SoC have 8 LMACs each, whereas RPM2
has only 4 LMACs. Similarly, the RPM0 and RPM1 have 256KB FIFO,
whereas RPM2 has 128KB FIFO. This patch fixes an issue with
improper TX credit programming for the RPM2 link.

Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Signed-off-by: Naveen Mamindlapalli <naveenm@marvell.com>
---
 drivers/net/ethernet/marvell/octeontx2/af/rpm.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rpm.c b/drivers/net/ethernet/marvell/octeontx2/af/rpm.c
index 4728ba34b0e3..76218f1cb459 100644
--- a/drivers/net/ethernet/marvell/octeontx2/af/rpm.c
+++ b/drivers/net/ethernet/marvell/octeontx2/af/rpm.c
@@ -506,6 +506,7 @@ u32 rpm2_get_lmac_fifo_len(void *rpmd, int lmac_id)
 	rpm_t *rpm = rpmd;
 	u8 num_lmacs;
 	u32 fifo_len;
+	u16 max_lmac;
 
 	lmac_info = rpm_read(rpm, 0, RPM2_CMRX_RX_LMACS);
 	/* LMACs are divided into two groups and each group
@@ -513,7 +514,11 @@ u32 rpm2_get_lmac_fifo_len(void *rpmd, int lmac_id)
 	 * Group0 lmac_id range {0..3}
 	 * Group1 lmac_id range {4..7}
 	 */
-	fifo_len = rpm->mac_ops->fifo_len / 2;
+	max_lmac = (rpm_read(rpm, 0, CGX_CONST) >> 24) & 0xFF;
+	if (max_lmac > 4)
+		fifo_len = rpm->mac_ops->fifo_len / 2;
+	else
+		fifo_len = rpm->mac_ops->fifo_len;
 
 	if (lmac_id < 4) {
 		num_lmacs = hweight8(lmac_info & 0xF);
-- 
2.39.0.198.ga38d39a4c5

Re: [net PATCH] octeontx2-af: CN10KB: Fix FIFO length calculation for RPM2

Posted by Simon Horman 1 year, 11 months ago

On Fri, Jan 05, 2024 at 12:24:23PM +0530, Naveen Mamindlapalli wrote:
> From: Nithin Dabilpuram <ndabilpuram@marvell.com>
> 
> RPM0 and RPM1 on the CN10KB SoC have 8 LMACs each, whereas RPM2
> has only 4 LMACs. Similarly, the RPM0 and RPM1 have 256KB FIFO,
> whereas RPM2 has 128KB FIFO. This patch fixes an issue with
> improper TX credit programming for the RPM2 link.
> 
> Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
> Signed-off-by: Naveen Mamindlapalli <naveenm@marvell.com>

If this is a fix for a user-visible bug then it should
have a Fixes tag. Else it should be targeted at net-next.

Also, as a potential follow-up, it looks like this
file (driver?) could benefit from use of GETMASK/FIELD_GET/FIELD_PREP.
But, IMHO, there is no need to do that for this change
which is in a style consistent with the rest of the file.

> ---
>  drivers/net/ethernet/marvell/octeontx2/af/rpm.c | 7 ++++++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rpm.c b/drivers/net/ethernet/marvell/octeontx2/af/rpm.c
> index 4728ba34b0e3..76218f1cb459 100644
> --- a/drivers/net/ethernet/marvell/octeontx2/af/rpm.c
> +++ b/drivers/net/ethernet/marvell/octeontx2/af/rpm.c
> @@ -506,6 +506,7 @@ u32 rpm2_get_lmac_fifo_len(void *rpmd, int lmac_id)
>  	rpm_t *rpm = rpmd;
>  	u8 num_lmacs;
>  	u32 fifo_len;
> +	u16 max_lmac;
>  
>  	lmac_info = rpm_read(rpm, 0, RPM2_CMRX_RX_LMACS);
>  	/* LMACs are divided into two groups and each group
> @@ -513,7 +514,11 @@ u32 rpm2_get_lmac_fifo_len(void *rpmd, int lmac_id)
>  	 * Group0 lmac_id range {0..3}
>  	 * Group1 lmac_id range {4..7}
>  	 */
> -	fifo_len = rpm->mac_ops->fifo_len / 2;
> +	max_lmac = (rpm_read(rpm, 0, CGX_CONST) >> 24) & 0xFF;
> +	if (max_lmac > 4)
> +		fifo_len = rpm->mac_ops->fifo_len / 2;
> +	else
> +		fifo_len = rpm->mac_ops->fifo_len;
>  
>  	if (lmac_id < 4) {
>  		num_lmacs = hweight8(lmac_info & 0xF);
> -- 
> 2.39.0.198.ga38d39a4c5
>