Re: [PATCH net-next v2 6/9] net: microchip: sparx5: Adding basic rule management in VCAP API

[Steen Hegelund]

Hi Casper,

On Thu, 2022-10-20 at 09:41 +0200, Casper Andersson wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
> 
> On 2022-10-19 13:42, Steen Hegelund wrote:
> > +/* Write VCAP cache content to the VCAP HW instance */
> > +static int vcap_write_rule(struct vcap_rule_internal *ri)
> > +{
> > +     struct vcap_admin *admin = ri->admin;
> > +     int sw_idx, ent_idx = 0, act_idx = 0;
> > +     u32 addr = ri->addr;
> > +
> > +     if (!ri->size || !ri->keyset_sw_regs || !ri->actionset_sw_regs) {
> > +             pr_err("%s:%d: rule is empty\n", __func__, __LINE__);
> > +             return -EINVAL;
> > +     }
> > +     /* Use the values in the streams to write the VCAP cache */
> > +     for (sw_idx = 0; sw_idx < ri->size; sw_idx++, addr++) {
> > +             ri->vctrl->ops->cache_write(ri->ndev, admin,
> > +                                     VCAP_SEL_ENTRY, ent_idx,
> > +                                     ri->keyset_sw_regs);
> > +             ri->vctrl->ops->cache_write(ri->ndev, admin,
> > +                                     VCAP_SEL_ACTION, act_idx,
> > +                                     ri->actionset_sw_regs);
> > +             ri->vctrl->ops->update(ri->ndev, admin, VCAP_CMD_WRITE,
> > +                                VCAP_SEL_ALL, addr);
> 
> Arguments not aligned with opening parenthesis.

I will fix that.

> 
> >  /* Validate a rule with respect to available port keys */
> >  int vcap_val_rule(struct vcap_rule *rule, u16 l3_proto)
> >  {
> >       struct vcap_rule_internal *ri = to_intrule(rule);
> > +     enum vcap_keyfield_set keysets[10];
> > +     struct vcap_keyset_list kslist;
> > +     int ret;
> > 
> >       /* This validation will be much expanded later */
> > +     ret = vcap_api_check(ri->vctrl);
> > +     if (ret)
> > +             return ret;
> >       if (!ri->admin) {
> >               ri->data.exterr = VCAP_ERR_NO_ADMIN;
> >               return -EINVAL;
> > @@ -113,14 +304,41 @@ int vcap_val_rule(struct vcap_rule *rule, u16 l3_proto)
> >               ri->data.exterr = VCAP_ERR_NO_KEYSET_MATCH;
> >               return -EINVAL;
> >       }
> > +     /* prepare for keyset validation */
> > +     keysets[0] = ri->data.keyset;
> > +     kslist.keysets = keysets;
> > +     kslist.cnt = 1;
> > +     /* Pick a keyset that is supported in the port lookups */
> > +     ret = ri->vctrl->ops->validate_keyset(ri->ndev, ri->admin, rule, &kslist,
> > +                                           l3_proto);
> > +     if (ret < 0) {
> > +             pr_err("%s:%d: keyset validation failed: %d\n",
> > +                    __func__, __LINE__, ret);
> > +             ri->data.exterr = VCAP_ERR_NO_PORT_KEYSET_MATCH;
> > +             return ret;
> > +     }
> >       if (ri->data.actionset == VCAP_AFS_NO_VALUE) {
> >               ri->data.exterr = VCAP_ERR_NO_ACTIONSET_MATCH;
> >               return -EINVAL;
> >       }
> > -     return 0;
> > +     vcap_add_type_keyfield(rule);
> > +     /* Add default fields to this rule */
> > +     ri->vctrl->ops->add_default_fields(ri->ndev, ri->admin, rule);
> > +
> > +     /* Rule size is the maximum of the entry and action subword count */
> > +     ri->size = max(ri->keyset_sw, ri->actionset_sw);
> > +
> > +     /* Finally check if there is room for the rule in the VCAP */
> > +     return vcap_rule_space(ri->admin, ri->size);
> >  }
> >  EXPORT_SYMBOL_GPL(vcap_val_rule);
> 
> Validating a rule also modifies it. I think validation and modification
> should generally be kept apart. But it looks like it might be hard with
> the current design since you need to add the fields to then check the
> space it takes, and the rule sizes can depend on the hardware.

Hmm.  You got a point.  I just wanted to keep the number of API calls down a bit, so the validation
also ensures that the rule has a keyset type field and any other default fields that is needed in
the particular VCAP and setting the keyset defines the size of the rule, so it all needs to fit
together in the end.
For now I would like to keep this as just one validation call.

> 
> Tested on Microchip PCB135 switch.
> 
> Tested-by: Casper Andersson <casper.casan@xxxxxxxxx>
> Reviewed-by: Casper Andersson <casper.casan@xxxxxxxxx>
> 

BR
Steen