RE: [PATCH] eventdev/eth_rx: fix timestamp field register in mbuf

["Naga Harish K, S V" <s.v.naga.harish.k@intel.com>]

> -----Original Message-----
> From: Rahul Bhansali <rbhansali@marvell.com>
> Sent: Monday, September 18, 2023 1:56 PM
> To: dev@dpdk.org; Naga Harish K, S V <s.v.naga.harish.k@intel.com>; Jerin
> Jacob <jerinj@marvell.com>; Kundapura, Ganapati
> <ganapati.kundapura@intel.com>
> Cc: Rahul Bhansali <rbhansali@marvell.com>; stable@dpdk.org
> Subject: [PATCH] eventdev/eth_rx: fix timestamp field register in mbuf
> 
> For eventdev internal port, timestamp dynamic field registration in mbuf is not
> required as that will be done from net device.
> For SW eventdev, Rx timestamp field registration will be done during Rx queue
> add operation as per device capabilities and offload configuration.
> 
> Fixes: 83ab470d1259 ("eventdev/eth_rx: use timestamp as dynamic mbuf
> field")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Rahul Bhansali <rbhansali@marvell.com>
> ---
>  lib/eventdev/rte_event_eth_rx_adapter.c | 19 ++++++++++++-------
>  1 file changed, 12 insertions(+), 7 deletions(-)
> 
> diff --git a/lib/eventdev/rte_event_eth_rx_adapter.c
> b/lib/eventdev/rte_event_eth_rx_adapter.c
> index 3ebfa5366d..5a5fade466 100644
> --- a/lib/eventdev/rte_event_eth_rx_adapter.c
> +++ b/lib/eventdev/rte_event_eth_rx_adapter.c
> @@ -2472,13 +2472,6 @@ rxa_create(uint8_t id, uint8_t dev_id,
>  	if (conf_cb == rxa_default_conf_cb)
>  		rx_adapter->default_cb_arg = 1;
> 
> -	if (rte_mbuf_dyn_rx_timestamp_register(
> -			&event_eth_rx_timestamp_dynfield_offset,
> -			&event_eth_rx_timestamp_dynflag) != 0) {
> -		RTE_EDEV_LOG_ERR("Error registering timestamp field in
> mbuf\n");
> -		return -rte_errno;
> -	}
> -
>  	rte_eventdev_trace_eth_rx_adapter_create(id, dev_id, conf_cb,
>  		conf_arg);
>  	return 0;
> @@ -2738,6 +2731,7 @@ rte_event_eth_rx_adapter_queue_add(uint8_t id,
>  					1);
>  		}
>  	} else {
> +		uint64_t dev_offloads;
>  		rte_spinlock_lock(&rx_adapter->rx_lock);
>  		dev_info->internal_event_port = 0;
>  		ret = rxa_init_service(rx_adapter, id); @@ -2749,6 +2743,17
> @@ rte_event_eth_rx_adapter_queue_add(uint8_t id,
>  				rxa_sw_adapter_queue_count(rx_adapter));
>  		}
>  		rte_spinlock_unlock(&rx_adapter->rx_lock);
> +
> +		dev_offloads = dev_info->dev->data-
> >dev_conf.rxmode.offloads;

This is a one-time operation and need not happen for every queue_add.
Move this registration to "rxa_init_service()" function which executes only once for creating rte_service.

Also, no need to check for offload capabilities and directly do the registration inside
Rxa_init_service as done before in rxa_create.
Mbuf field is global to the entire application and need not be done based on ethdev offload capabilities.


> +		if (dev_offloads & RTE_ETH_RX_OFFLOAD_TIMESTAMP) {
> +			if (rte_mbuf_dyn_rx_timestamp_register(
> +
> 	&event_eth_rx_timestamp_dynfield_offset,
> +
> 	&event_eth_rx_timestamp_dynflag) != 0) {
> +				RTE_EDEV_LOG_ERR("Error registering
> timestamp field in mbuf\n");
> +				return -rte_errno;
> +			}
> +		}
> +
>  	}
> 
>  	rte_eventdev_trace_eth_rx_adapter_queue_add(id, eth_dev_id,
> --
> 2.25.1