RE: [PATCH v4 net-next 7/8] net: mscc: ocelot: use index to set vcap policer

[Xiaoliang Yang]

On Wed, Sep 22, 2021 at 13:18:37 +0000, Vladimir Oltean wrote:
> > Policer was previously automatically assigned from the highest index
> > to the lowest index from policer pool. But police action of tc flower
> > now uses index to set an police entry. This patch uses the police
> > index to set vcap policers, so that one policer can be shared by multiple
> rules.
> >
> > Signed-off-by: Xiaoliang Yang <xiaoliang.yang_1@xxxxxxx>
> > ---
> > +#define VSC9959_VCAP_POLICER_BASE	63
> > +#define VSC9959_VCAP_POLICER_MAX	383
> >
> 
> > +#define VSC7514_VCAP_POLICER_BASE			128
> > +#define VSC7514_VCAP_POLICER_MAX			191
> 
> I think this deserves an explanation.
> 
> The VSC7514 driver uses the max number of policers as 383 (0x17f) ever since
> commit b596229448dd ("net: mscc: ocelot: Add support for tcam"), aka the
> very beginning.
> 
> Yet, the documentation at "3.10.1 Policer Allocation"
> https://ww1.microchip.com/downloads/en/DeviceDoc/VMDS-10491.pdf
> says very clearly that there are only 192 policers indeed.
> 
> What's going on?

In commit commit b596229448dd ("net: mscc: ocelot: Add support for tcam"), Horatiu Vultur define the max number of policers as 383:
+#define OCELOT_POLICER_DISCARD 0x17f
VCAP IS2 use this policer to set drop action. I did not change this and set the VCAP policers with 128-191 according to the VSC7514 document.

I don't know why 383 was used as the maximum value of policer in the original code. Can Microchip people check the code or the documentation for errors?

> 
> Also, FWIW, Seville has this policer allocation:
> 
>       0 ----+----------------------+
>             |  Port Policers (11)  |
>      11 ----+----------------------+
>             |  VCAP Policers (21)  |
>      32 ----+----------------------+
>             |   QoS Policers (88)  |
>     120 ----+----------------------+
>             |  VCAP Policers (43)  |
>     162 ----+----------------------+

I didn't find Seville's document, if this allocation is right, I will add it in Seville driver.

Thanks,
Xiaoliang