Re: [PATCH net-next v7 2/2] net: dsa: mv88e6xxx: add support for credit based shaper

[Cedric Jehasse]

O Fri, Jun 12, 2026 at 03:36:56PM -0700, Jakub Kicinski wrote:
> On Tue, 09 Jun 2026 14:10:51 +0200 Cedric Jehasse via B4 Relay wrote:
> > From: Cedric Jehasse <cedric.jehasse@xxxxxxxxxx>
> > 
> > Some of the chips supported by this driver have credit based shaper
> > support. Support is added for the 6341, 6352, 6390 and 6393 families.
> > This is configured using the Qav registers in the AVB register block.
> > There are small differences in the Qav registers between the chip
> > families (eg. the unit used for the rate and number of bits in the
> > registers). mv88e6xxx_qav_info is introduced to configure this per chip.
> > 
> > Eg. setting up 20mbps credit based shaper on a 1GBit link:
> > tc qdisc add dev p8 parent root handle 100: mqprio \
> >                 num_tc 8 \
> >                 map 0 0 6 7 0 5 0 0 0 0 0 0 0 0 0 0 \
> >                 queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 \
> >                 hw 0
> > 
> > tc qdisc replace dev p8 parent 100:8 cbs locredit -1470 hicredit 30 \
> >     sendslope -980000 idleslope 20000 offload 1
> 
> Are DSA ports multi-queue? I would have expected a DSA driver to
> offload PRIO not MQPRIO.
> 
> I seem to recall other discussion on the ML on the topic.
> It'd be great to get some review tags from folks familiar with 
> the device.

As mentioned by Luke, this patch adds offloading for CBS, not MQPIO. Other dsa
driver already do this (eg. microchip ksz driver).

> 
> > Note: only idleslope and hicredit can be programmed in the switch
> > registers, other parameters won't affect settings.
> 
> > +static int mv88e6xxx_setup_tc_cbs(struct dsa_switch *ds, int port,
> > +				  struct tc_cbs_qopt_offload *cbs)
> 
> please stick an extack into struct tc_cbs_qopt_offload and use it to
> report the reason for rejection back to the user
> 

Do you mean modifying cbs_enable_offload to report a driver specified reason isof
reportig the generic "Specified device failed to setup cbs hardware offload"?
What should be done for other drivers who handle TC_SETUP_QDISC_CBS, but don't
fill in extack? Should cbs_enable_offload fill in extack with "Specified device
failed to setup cbs hardware offload"?

Cedric