Re: [dpdk-dev] [PATCH 2/5] ethdev: add new ops of 'check_command_supported' and 'rx_classification_filter_ctl'

["Zhang, Helin" <helin.zhang@intel.com>]

> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Sent: Thursday, July 24, 2014 3:57 PM
> To: Zhang, Helin
> Cc: dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH 2/5] ethdev: add new ops of
> 'check_command_supported' and 'rx_classification_filter_ctl'
> 
> 2014-07-24 14:42, Helin Zhang:
> > Two ops of 'check_command_supported' and 'rx_classification_filter_ctl'
> > are added.
> > * 'check_command_supported' is for capability discovery. In anothoer
> >   word, it is to check if specific feature/command is supported by
> >   the specific port.
> 
> This generic service is really needed to add NIC-specific functions.
> I'd suggest to name it "is_supported".
> 
I tried to name it as the same style of rte_cpu_check_supported().
As you indicated, 'is_commmand_supported' might be a better name.

> > * 'rx_classification_filter_ctl' is for receive classifcation filter
> >   configuring. e.g. hash function configuration, flow director
> >   configuration. It is a common API where a lot of commands can
> >   be implemented for different sub features.
> 
> Not sure about this one. You are hiding specific API in an opaque structure.
> By the way, it should be in another patch.
> 
We want to implement several common API for NIC specific features, to avoid creating quite a lot of ops in 'struct eth_dev_ops'.
The idea came from ioctl.

> > +/**
> > + * Check if the command is supported by an Ethernet device.
> > + *
> > + * @param port_id
> > + *   The port identifier of the Ethernet device.
> > + * @param cmd
> > + *   The command.
> > + *
> > + * @return
> > + *   - (> 0) The command is supported.
> > + *   - (0) The command is not supported.
> > + *   - (-ENOTSUP) if hardware doesn't support.
> > + *   - (-ENODEV) if <port_id> is invalid.
> > + */
> > +int rte_eth_dev_check_command_supported(uint8_t port_id, uint32_t
> cmd);
> 
> What are the possible commands?
> You should define enum/constants here.
> 
The idea came from ioctl(). Enum could be a choice. The commands was defined in rte_eth_features.h. Comments should be added here for telling that.

> --
> Thomas