Re: [RFC net-next v4 6/9] netdev-genl: Support setting per-NAPI config values

[Joe Damato <jdamato@fastly.com>]

On Tue, Oct 08, 2024 at 04:19:19PM -0700, Jakub Kicinski wrote:
> On Tue, 8 Oct 2024 16:00:41 -0700 Joe Damato wrote:
> > > Make sure you edit the spec, not the output. Looks like there may be 
> > > a problem here (napi-id vs id in the attributes).  
> > 
> > I'm not sure I follow this part, sorry if I'm just missing something
> > here.
> > 
> > I was referring to NETDEV_A_NAPI_DEFER_HARD_IRQS which in RFCv4 is
> > listed as NLA_S32 (in this patch):
> > 
> > static const struct nla_policy netdev_napi_set_nl_policy[NETDEV_A_NAPI_GRO_FLUSH_TIMEOUT + 1] = {
> >      [NETDEV_A_NAPI_ID] = { .type = NLA_U32, },
> >      [NETDEV_A_NAPI_DEFER_HARD_IRQS] = { .type = NLA_S32 },
> > 
> > However, in the yaml spec (patch 2/9):
> > 
> > +      -
> > +        name: defer-hard-irqs
> > +        doc: The number of consecutive empty polls before IRQ deferral ends
> > +             and hardware IRQs are re-enabled.
> > +        type: u32
> > +        checks:
> > +          max: s32-max
> > 
> > So the type is u32 but with a "checks" to match what happens now in
> > sysfs.
> > 
> > That's why I mentioned changing NLA_S32 to NLA_U32.
> > 
> > Am I missing something?
> 
> YNL will generate the correct code for your - the right type
> and the right range validation. Run the command below to see.
> 
> > Not sure what you meant by "napi-id vs id" ?
> 
> I can't apply the series now, but when it was posted the YNL code
> generation failed here complaining about napi-id not existing in
> the attribute set in which it is used. In the napi attribute set
> the NAPI ID is called just "id", not "napi-id".

Ah, I see what you mean now. It should have been obvious, but the
-gen* files are, uh, auto-generated ;)

And yes, I see now that the attribute set names it "id", so I've
fixed it and the command runs clean and I'll include the generated
output this time in the v5.