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

[Luke Howard]

> On 28 May 2026, at 7:19 pm, Cedric Jehasse <cedric.jehasse@xxxxxxxxx> wrote:
> 
> On Thu, May 28, 2026 at 1:52 AM Luke Howard <lukeh@xxxxxxxx> wrote:
>> Would also be good to add support for the 6341, which as 4 queues but uses the same Qav multipliers as the 6390.
> I'll have a look at the 6341

As an aside, the 6341 driver is buggy. It assumes the chip does not have a dedicated ATU FID register because it has 256 databases. This is incorrect and breaks VLAN-aware bridging. I will post a patch when the merge window reopens.

On the hardware side, the ATU hashing algorithm appears to easily collide. This causes problems with things such as installing the broadcast address into multiple FIDs, not to mention 802.1Q Dynamic Reservation Entries. The only workaround is to use the ‘direct’ algorithm which may have a performance impact (I am presuming it is a linear search; the data sheet warns against using it).

Luke