[dpdk-dev] [PATCH v2 3/5] ethdev: convert remaining apps to new offload API

[Thomas Monjalon]

04/07/2018 13:16, Andrew Rybchenko:
> On 07/03/2018 12:27 AM, Thomas Monjalon wrote:
> > --- a/doc/guides/sample_app_ug/link_status_intr.rst
> > +++ b/doc/guides/sample_app_ug/link_status_intr.rst
> > @@ -137,10 +137,7 @@ The global configuration is stored in a static structure:
> >       static const struct rte_eth_conf port_conf = {
> >           .rxmode = {
> >               .split_hdr_size = 0,
> > -            .header_split = 0,   /**< Header Split disabled */
> > -            .hw_ip_checksum = 0, /**< IP checksum offload disabled */
> > -            .hw_vlan_filter = 0, /**< VLAN filtering disabled */
> > -            .hw_strip_crc= 0,    /**< CRC stripped by hardware */
> > +            .offloads = DEV_RX_OFFLOAD_CRC_STRIP,
> 
> Is it intended that CRC strip was disabled before and now it is becoming 
> enabled?

Yes, I consider the comment to be the real intent.

> > --- a/examples/bbdev_app/main.c
> > +++ b/examples/bbdev_app/main.c
> > @@ -64,11 +64,7 @@ static const struct rte_eth_conf port_conf = {
> >   		.mq_mode = ETH_MQ_RX_NONE,
> >   		.max_rx_pkt_len = ETHER_MAX_LEN,
> >   		.split_hdr_size = 0,
> > -		.header_split = 0, /**< Header Split disabled */
> > -		.hw_ip_checksum = 0, /**< IP checksum offload disabled */
> > -		.hw_vlan_filter = 0, /**< VLAN filtering disabled */
> > -		.jumbo_frame = 0, /**< Jumbo Frame Support disabled */
> > -		.hw_strip_crc = 0, /**< CRC stripped by hardware */
> > +		.offloads = DEV_RX_OFFLOAD_CRC_STRIP,
> 
> Is it intended that CRC strip was disabled before and now it is becoming 
> enabled?

Yes, I consider the comment to be the real intent.

> > --- a/test/test/test_pmd_perf.c
> > +++ b/test/test/test_pmd_perf.c
> > @@ -97,11 +90,6 @@ static struct rte_eth_txconf tx_conf = {
> >   	},
> >   	.tx_free_thresh = 32, /* Use PMD default values */
> >   	.tx_rs_thresh = 32, /* Use PMD default values */
> > -	.txq_flags = (ETH_TXQ_FLAGS_NOMULTSEGS |
> > -		      ETH_TXQ_FLAGS_NOVLANOFFL |
> > -		      ETH_TXQ_FLAGS_NOXSUMSCTP |
> > -		      ETH_TXQ_FLAGS_NOXSUMUDP |
> > -		      ETH_TXQ_FLAGS_NOXSUMTCP)
> >   };
> >   
> >   enum {
> > @@ -808,38 +796,29 @@ test_set_rxtx_conf(cmdline_fixed_string_t mode)
> >   
> >   	if (!strcmp(mode, "vector")) {
> >   		/* vector rx, tx */
> > -		tx_conf.txq_flags = 0xf01;
> 
> I'd say that 100% correct equivalent would be:
> tx_conf.offloads &= ~(DEV_TX_OFFLOAD_VLAN_INSERT |
>            DEV_TX_OFFLOAD_IPV4_CKSUM | DEV_TX_OFFLOAD_UDP_CKSUM |
>            DEV_TX_OFFLOAD_TCP_CKSUM | DEV_TX_OFFLOAD_SCTP_CKSUM |
>            DEV_TX_OFFLOAD_MULTI_SEGS);

I'd say it is a really crappy code, and probably tuned for Intel devices only.

> 
> I guess the function may be called few times with different mode set.
> If so, similar fixes should be applied below as well.
> 
> >   		tx_conf.tx_rs_thresh = 32;
> >   		tx_conf.tx_free_thresh = 32;
> > -		port_conf.rxmode.hw_ip_checksum = 0;
> > -		port_conf.rxmode.enable_scatter = 0;
> 
> port_conf.rxmode.offloads &= ~(DEV_RX_OFFLOAD_CHECKSUM | 
> DEV_RX_OFFLOAD_SCATTER);
> 
> >   		return 0;
> >   	} else if (!strcmp(mode, "scalar")) {
> >   		/* bulk alloc rx, full-featured tx */
> > -		tx_conf.txq_flags = 0;
> 
> I think here we should enable offloads listed above to have 
> full-featured Tx:
> tx_conf.offloads |=  ...
> 
> >   		tx_conf.tx_rs_thresh = 32;
> >   		tx_conf.tx_free_thresh = 32;
> > -		port_conf.rxmode.hw_ip_checksum = 1;
> > -		port_conf.rxmode.enable_scatter = 0;
> > +		port_conf.rxmode.offloads |= DEV_RX_OFFLOAD_CHECKSUM;
> 
> port_conf.rxmode.offloads &= ~DEV_RX_OFFLOAD_SCATTER;
> 
> >   		return 0;
> >   	} else if (!strcmp(mode, "hybrid")) {
> >   		/* bulk alloc rx, vector tx
> >   		 * when vec macro not define,
> >   		 * using the same rx/tx as scalar
> >   		 */
> > -		tx_conf.txq_flags = 0xf01;
> 
> As in similar case above.
> 
> >   		tx_conf.tx_rs_thresh = 32;
> >   		tx_conf.tx_free_thresh = 32;
> > -		port_conf.rxmode.hw_ip_checksum = 1;
> > -		port_conf.rxmode.enable_scatter = 0;
> > +		port_conf.rxmode.offloads |= DEV_RX_OFFLOAD_CHECKSUM;
> 
> As in similar case above
> 
> >   		return 0;
> >   	} else if (!strcmp(mode, "full")) {
> >   		/* full feature rx,tx pair */
> > -		tx_conf.txq_flags = 0x0;   /* must condition */
> 
> As in similar case above.
> 
> >   		tx_conf.tx_rs_thresh = 32;
> >   		tx_conf.tx_free_thresh = 32;
> > -		port_conf.rxmode.hw_ip_checksum = 0;
> > -		port_conf.rxmode.enable_scatter = 1; /* must condition */
> > +		port_conf.rxmode.offloads |= DEV_RX_OFFLOAD_SCATTER;
> 
> port_conf.rxmode.offloads &= ~DEV_RX_OFFLOAD_CHECKSUM;
> 
> >   		return 0;
> >   	}
> >   
> 
> In general I think that it would be really good to avoid changes in
> behaviour when technical changes are done.

I agree, but in this case, it is impossible to know what was the real intent.
And I am perfectly fine breaking bad code.
The other option is to just remove the file. Maybe the best option?