[PATCH] net/i40e: avx512 fast-free path bug fix

[Morten Brørup]

> From: Kamalakshitha Aligeri [mailto:kamalakshitha.aligeri at arm.com]
> Sent: Tuesday, 7 March 2023 20.32
> 
> In i40e_tx_free_bufs_avx512 fast-free path, when cache is NULL,
> non fast-free path is being executed. Fixed the bug by calling
> rte_mempool_generic_put API that handles the cache==NULL case.
> 
> Fixes: 5171b4ee6b6b ("net/i40e: optimize Tx by using AVX512")
> Cc: leyi.rong at intel.com
> Cc: stable at dpdk.org
> 
> Signed-off-by: Kamalakshitha Aligeri <kamalakshitha.aligeri at arm.com>
> Reviewed-by: Ruifeng Wang <ruifeng.wang at arm.com>
> Reviewed-by: Feifei Wang <feifei.wang2 at arm.com>
> ---
>  .mailmap                                |  1 +
>  drivers/net/i40e/i40e_rxtx_vec_avx512.c | 12 ++++--------
>  2 files changed, 5 insertions(+), 8 deletions(-)
> 
> diff --git a/.mailmap b/.mailmap
> index a9f4f28fba..2581d0efe7 100644
> --- a/.mailmap
> +++ b/.mailmap
> @@ -677,6 +677,7 @@ Kai Ji <kai.ji at intel.com>
>  Kaiwen Deng <kaiwenx.deng at intel.com>
>  Kalesh AP <kalesh-anakkur.purayil at broadcom.com>
>  Kamalakannan R <kamalakannan.r at intel.com>
> +Kamalakshitha Aligeri <kamalakshitha.aligeri at arm.com>
>  Kamil Bednarczyk <kamil.bednarczyk at intel.com>
>  Kamil Chalupnik <kamilx.chalupnik at intel.com>
>  Kamil Rytarowski <kamil.rytarowski at caviumnetworks.com>
> diff --git a/drivers/net/i40e/i40e_rxtx_vec_avx512.c
> b/drivers/net/i40e/i40e_rxtx_vec_avx512.c
> index d3c7bfd121..ad0893324d 100644
> --- a/drivers/net/i40e/i40e_rxtx_vec_avx512.c
> +++ b/drivers/net/i40e/i40e_rxtx_vec_avx512.c
> @@ -783,16 +783,13 @@ i40e_tx_free_bufs_avx512(struct i40e_tx_queue
> *txq)
>  		struct rte_mempool_cache *cache =
> rte_mempool_default_cache(mp,
>  				rte_lcore_id());
> 
> -		if (!cache || cache->len == 0)
> -			goto normal;
> -
> -		cache_objs = &cache->objs[cache->len];
> -
> -		if (n > RTE_MEMPOOL_CACHE_MAX_SIZE) {
> -			rte_mempool_ops_enqueue_bulk(mp, (void *)txep, n);
> +		if (!cache || n > RTE_MEMPOOL_CACHE_MAX_SIZE) {
> +			rte_mempool_generic_put(mp, (void *)txep, n, cache);
>  			goto done;
>  		}
> 
> +		cache_objs = &cache->objs[cache->len];
> +
>  		/* The cache follows the following algorithm
>  		 *   1. Add the objects to the cache
>  		 *   2. Anything greater than the cache min value (if it
> @@ -824,7 +821,6 @@ i40e_tx_free_bufs_avx512(struct i40e_tx_queue *txq)
>  		goto done;
>  	}
> 
> -normal:
>  	m = rte_pktmbuf_prefree_seg(txep[0].mbuf);
>  	if (likely(m)) {
>  		free[0] = m;
> --
> 2.25.1
> 

An improvement of the copy-paste code we are aiming to replace by proper use of the mempool API.

But still an improvement.

Acked-by: Morten Brørup <mb at smartsharesystems.com>