Re: [PATCH net-next v6 2/2] net: dsa: mv88e6xxx: add support for credit based shaper

O Tue, Jun 09, 2026 at 12:05:41PM +0200, Paolo Abeni wrote:

On 6/4/26 3:44 PM, Cedric Jehasse via B4 Relay wrote:

quoted

@@ -5679,6 +5686,27 @@ static const struct mv88e6xxx_ops mv88e6393x_ops = {
 	.tcam_ops = &mv88e6393_tcam_ops,
 };
 
+static const struct mv88e6xxx_qav_info mv88e6352_qav_info = {
+	.rate_unit = 32,
+	.rate_mask = GENMASK(14, 0),
+	.hilimit_mask = GENMASK(14, 0),
+	.queue_mask = GENMASK(3, 0),
+};
+
+static const struct mv88e6xxx_qav_info mv88e6341_qav_info = {
+	.rate_unit = 64,
+	.rate_mask = GENMASK(15, 0),
+	.hilimit_mask = GENMASK(14, 0),
+	.queue_mask = GENMASK(3, 0),
+};
+
+static const struct mv88e6xxx_qav_info mv88e6390_qav_info = {
+	.rate_unit = 64,
+	.rate_mask = GENMASK(15, 0),
+	.hilimit_mask = GENMASK(13, 0),
+	.queue_mask = GENMASK(7, 0),
+};

Sashiko mentions that the above masks are not consistent with the switch
datasheet.

I'm quite surprised the LLM could check such fine grain detail, could
you please double check if the comment makes sense:

https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20260604-net-next-mv88e6xxx-cbs-v6-0-b68c01142d1e%40luminex.be

The datasheet i have doesn't mention a QPriEn bit in the QPri Rate register,
that bit is marked as reserved in the datasheet. That comment doesn't seem
valid.
However, reviewing this i've spotted a mistake in the hilimit_mask for the 6341.
I will fix it.

Cedric