On 4/23/2018 1:33 PM, Sunil Kumar Kori wrote:
> Fixes: 16e2c27f4fc7 ("net/dpaa: support new ethdev offload APIs")
> Cc: shreyansh.j...@nxp.com
>
> Signed-off-by: Sunil Kumar Kori <sunil.k...@nxp.com>
> ---
> drivers/net/dpaa/dpaa_ethdev.c | 34 +++++++++++++++++++++++++++-------
> 1 file changed, 27 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
> index b2740b4..81001cb 100644
> --- a/drivers/net/dpaa/dpaa_ethdev.c
> +++ b/drivers/net/dpaa/dpaa_ethdev.c
> @@ -45,6 +45,25 @@
> #include <fsl_bman.h>
> #include <fsl_fman.h>
>
> +/* Non-Supported Rx offloads */
> +static uint64_t dev_rx_offloads_not_supported =
> + DEV_RX_OFFLOAD_TCP_LRO |
> + DEV_RX_OFFLOAD_MACSEC_STRIP |
> + DEV_RX_OFFLOAD_HEADER_SPLIT |
> + DEV_RX_OFFLOAD_VLAN_EXTEND |
> + DEV_RX_OFFLOAD_SECURITY;
> +
> +/* Non-Supported Tx offloads */
> +static uint64_t dev_tx_offloads_not_supported =
> + DEV_TX_OFFLOAD_TCP_TSO |
> + DEV_TX_OFFLOAD_UDP_TSO |
> + DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
> + DEV_TX_OFFLOAD_GRE_TNL_TSO |
> + DEV_TX_OFFLOAD_IPIP_TNL_TSO |
> + DEV_TX_OFFLOAD_GENEVE_TNL_TSO |
> + DEV_TX_OFFLOAD_MACSEC_INSERT |
> + DEV_TX_OFFLOAD_SECURITY;
> +
> /* Keep track of whether QMAN and BMAN have been globally initialized */
> static int is_global_init;
> /* At present we only allow up to 4 push mode queues - as each of this queue
> @@ -150,16 +169,17 @@ dpaa_eth_dev_configure(struct rte_eth_dev *dev)
> PMD_INIT_FUNC_TRACE();
>
> dpaa_eth_dev_info(dev, &dev_info);
> - if (((~(dev_info.rx_offload_capa) & rx_offloads) != 0)) {
> - DPAA_PMD_ERR("Some Rx offloads are not supported "
> - "requested 0x%" PRIx64 " supported 0x%" PRIx64,
> + /* Rx offloads validation */
> + if (dev_rx_offloads_not_supported & rx_offloads) {
> + DPAA_PMD_ERR(
> + "Rx offloads not supported - Requested 0x%" PRIx64 " supported
> 0x%" PRIx64,
> rx_offloads, dev_info.rx_offload_capa);
> return -ENOTSUP;
> }
There are three group of offloads from PMD point of view:
- Supported, reported via rx_offload_capa, setting these will work fine
- Not supported, listed in dev_rx_offloads_not_supported, application setting
this will get an error with and error log.
- Not configurable, supported by hardware but not configurable by application,
these goes silently, can you print a warning for this case?
And it can be easier to keep list of these offloads, otherwise you will need to
update your unsupported list when a new type of offloading introduced. And
developers may more interested to see the list of not configurable, and leaving
rest as unsupported.
