RE: [PATCH v4 1/1] ethdev: parsing multiple representor devargs string

[Harman Kalra <hkalra@marvell.com>]

Hi Chaoyong,

Please see responses inline


<snip>

> >                         return ret;
> >         }
> >
> ...
> > a/drivers/net/nfp/flower/nfp_flower_representor.c
> > b/drivers/net/nfp/flower/nfp_flower_representor.c
> > index 5f7c1fa737..63fe37c8d7 100644
> > --- a/drivers/net/nfp/flower/nfp_flower_representor.c
> > +++ b/drivers/net/nfp/flower/nfp_flower_representor.c
> > @@ -792,8 +792,8 @@ nfp_flower_repr_create(struct nfp_app_fw_flower
> > *app_fw_flower)
> >
> >         /* Now parse PCI device args passed for representor info */
> >         if (pci_dev->device.devargs != NULL) {
> > -               ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
> > &eth_da);
> > -               if (ret != 0) {
> > +               ret =
> > + rte_eth_devargs_parse(pci_dev->device.devargs->args,
> > &eth_da, 1);
> > +               if (ret < 0) {
> >                         PMD_INIT_LOG(ERR, "devarg parse failed");
> >                         return -EINVAL;
> >                 }
> 
> Sorry, I don't quite understand why change 'ret != 0' to 'ret < 0'?
> Compare return value with 0 or NULL is the specification for our PMD, we
> prefer not to change it if don't have a good reason.
> Thanks.

To support multiple representors under one PCI BDF, "eth_devargs args" parameter
passed to rte_eth_devargs_parse() is now an array which gets populated with multiple
"struct rte_eth_devargs" elements viz different pfvf representor devargs. 

We are proposing a change to the return value of this API, I.e. negative means error condition
and positive value refers to no of "struct rte_eth_devargs" elements populated. So it can't be
zero, else we wont know how many elements were populated in the array.

Thanks
Harman

> 
> > --
> > 2.18.0