On Wed, 9 Jul 2025 21:20:10 -0400
Jie Liu <liuj...@linkdatatechnology.com> wrote:
> s32 __rte_cold __sxe_rx_queue_mbufs_alloc(struct sxe_rx_queue *rxq)
> +{
> + struct sxe_rx_buffer *buf_ring = rxq->buffer_ring;
> + s32 ret = 0;
> + u64 dma_addr;
> + u16 i;
> +
> + for (i = 0; i < rxq->ring_depth; i++) {
> + volatile union sxe_rx_data_desc *desc;
> + struct rte_mbuf *mbuf = rte_mbuf_raw_alloc(rxq->mb_pool);
> +
> + if (mbuf == NULL) {
> + PMD_LOG_ERR(DRV, "rx mbuf alloc failed queue_id=%u",
> + (u16)rxq->queue_id);
> + ret = -ENOMEM;
> + goto l_end;
> + }
> +
Logging at error level could cause a lot of messages if pool gets exhausted
under
load. Better to just have a rx_nombuf statistic.
