The iavf driver has IPSec offload support on Tx, so add hooks to the common Tx function to support that. Do so in a way that has zero performance impact for drivers which do not have IPSec support, by passing in compile-time NULL constants for the function pointers, which can be optimized away by the compiler.
Signed-off-by: Bruce Richardson <[email protected]> --- drivers/net/intel/common/tx_scalar_fns.h | 60 ++++++++++++++++++++++-- drivers/net/intel/i40e/i40e_rxtx.c | 4 +- drivers/net/intel/ice/ice_rxtx.c | 4 +- 3 files changed, 60 insertions(+), 8 deletions(-) diff --git a/drivers/net/intel/common/tx_scalar_fns.h b/drivers/net/intel/common/tx_scalar_fns.h index 70b22f1da0..8c0de26537 100644 --- a/drivers/net/intel/common/tx_scalar_fns.h +++ b/drivers/net/intel/common/tx_scalar_fns.h @@ -152,6 +152,24 @@ typedef uint16_t (*ci_get_ctx_desc_fn)(uint64_t ol_flags, const struct rte_mbuf const union ci_tx_offload *tx_offload, const struct ci_tx_queue *txq, uint32_t *td_offset, uint64_t *qw0, uint64_t *qw1); +/* gets IPsec descriptor information and returns number of descriptors needed (0 or 1) */ +typedef uint16_t (*get_ipsec_desc_t)(const struct rte_mbuf *mbuf, + const struct ci_tx_queue *txq, + void **ipsec_metadata, + uint64_t *qw0, + uint64_t *qw1); +/* calculates segment length for IPsec + TSO combinations */ +typedef uint16_t (*calc_ipsec_segment_len_t)(const struct rte_mbuf *mb_seg, + uint64_t ol_flags, + const void *ipsec_metadata, + uint16_t tlen); + +/** IPsec descriptor operations for drivers that support inline IPsec crypto. */ +struct ci_ipsec_ops { + get_ipsec_desc_t get_ipsec_desc; + calc_ipsec_segment_len_t calc_segment_len; +}; + /* gets current timestamp tail index */ typedef uint16_t (*get_ts_tail_t)(struct ci_tx_queue *txq); /* writes a timestamp descriptor and returns new tail index */ @@ -171,6 +189,7 @@ ci_xmit_pkts(struct ci_tx_queue *txq, struct rte_mbuf **tx_pkts, uint16_t nb_pkts, ci_get_ctx_desc_fn get_ctx_desc, + const struct ci_ipsec_ops *ipsec_ops, const struct ci_timesstamp_queue_fns *ts_fns) { volatile struct ci_tx_desc *ci_tx_ring; @@ -206,6 +225,9 @@ ci_xmit_pkts(struct ci_tx_queue *txq, (void)ci_tx_xmit_cleanup(txq); for (nb_tx = 0; nb_tx < nb_pkts; nb_tx++) { + void *ipsec_md = NULL; + uint16_t nb_ipsec = 0; + uint64_t ipsec_qw0 = 0, ipsec_qw1 = 0; uint64_t cd_qw0, cd_qw1; tx_pkt = *tx_pkts++; @@ -225,17 +247,22 @@ ci_xmit_pkts(struct ci_tx_queue *txq, nb_ctx = get_ctx_desc(ol_flags, tx_pkt, &tx_offload, txq, &td_offset, &cd_qw0, &cd_qw1); + /* Get IPsec descriptor information if IPsec ops provided */ + if (ipsec_ops != NULL) + nb_ipsec = ipsec_ops->get_ipsec_desc(tx_pkt, txq, &ipsec_md, + &ipsec_qw0, &ipsec_qw1); + /* The number of descriptors that must be allocated for * a packet equals to the number of the segments of that - * packet plus the number of context descriptor if needed. + * packet plus the number of context and IPsec descriptors if needed. * Recalculate the needed tx descs when TSO enabled in case * the mbuf data size exceeds max data size that hw allows * per tx desc. */ if (ol_flags & (RTE_MBUF_F_TX_TCP_SEG | RTE_MBUF_F_TX_UDP_SEG)) - nb_used = (uint16_t)(ci_calc_pkt_desc(tx_pkt) + nb_ctx); + nb_used = (uint16_t)(ci_calc_pkt_desc(tx_pkt) + nb_ctx + nb_ipsec); else - nb_used = (uint16_t)(tx_pkt->nb_segs + nb_ctx); + nb_used = (uint16_t)(tx_pkt->nb_segs + nb_ctx + nb_ipsec); tx_last = (uint16_t)(tx_id + nb_used - 1); /* Circular ring */ @@ -288,6 +315,26 @@ ci_xmit_pkts(struct ci_tx_queue *txq, tx_id = txe->next_id; txe = txn; } + + if (ipsec_ops != NULL && nb_ipsec > 0) { + /* Setup TX IPsec descriptor if required */ + uint64_t *ipsec_txd = RTE_CAST_PTR(uint64_t *, &ci_tx_ring[tx_id]); + + txn = &sw_ring[txe->next_id]; + RTE_MBUF_PREFETCH_TO_FREE(txn->mbuf); + if (txe->mbuf) { + rte_pktmbuf_free_seg(txe->mbuf); + txe->mbuf = NULL; + } + + ipsec_txd[0] = ipsec_qw0; + ipsec_txd[1] = ipsec_qw1; + + txe->last_id = tx_last; + tx_id = txe->next_id; + txe = txn; + } + m_seg = tx_pkt; do { @@ -299,7 +346,12 @@ ci_xmit_pkts(struct ci_tx_queue *txq, txe->mbuf = m_seg; /* Setup TX Descriptor */ - slen = m_seg->data_len; + /* Calculate segment length, using IPsec callback if provided */ + if (ipsec_ops != NULL) + slen = ipsec_ops->calc_segment_len(m_seg, ol_flags, ipsec_md, 0); + else + slen = m_seg->data_len; + buf_dma_addr = rte_mbuf_data_iova(m_seg); while ((ol_flags & (RTE_MBUF_F_TX_TCP_SEG | RTE_MBUF_F_TX_UDP_SEG)) && diff --git a/drivers/net/intel/i40e/i40e_rxtx.c b/drivers/net/intel/i40e/i40e_rxtx.c index ecec70e0ac..e22fcfff60 100644 --- a/drivers/net/intel/i40e/i40e_rxtx.c +++ b/drivers/net/intel/i40e/i40e_rxtx.c @@ -1015,8 +1015,8 @@ get_context_desc(uint64_t ol_flags, const struct rte_mbuf *tx_pkt, uint16_t i40e_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) { - /* i40e does not support timestamp queues, so pass NULL for ts_fns */ - return ci_xmit_pkts(tx_queue, tx_pkts, nb_pkts, get_context_desc, NULL); + /* i40e does not support IPsec or timestamp queues, so pass NULL for both */ + return ci_xmit_pkts(tx_queue, tx_pkts, nb_pkts, get_context_desc, NULL, NULL); } static __rte_always_inline int diff --git a/drivers/net/intel/ice/ice_rxtx.c b/drivers/net/intel/ice/ice_rxtx.c index 384676cfc2..49ed6b8399 100644 --- a/drivers/net/intel/ice/ice_rxtx.c +++ b/drivers/net/intel/ice/ice_rxtx.c @@ -3100,9 +3100,9 @@ ice_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) struct ci_tx_queue *txq = (struct ci_tx_queue *)tx_queue; if (txq->tsq != NULL && txq->tsq->ts_flag > 0) - return ci_xmit_pkts(txq, tx_pkts, nb_pkts, get_context_desc, &ts_fns); + return ci_xmit_pkts(txq, tx_pkts, nb_pkts, get_context_desc, NULL, &ts_fns); - return ci_xmit_pkts(txq, tx_pkts, nb_pkts, get_context_desc, NULL); + return ci_xmit_pkts(txq, tx_pkts, nb_pkts, get_context_desc, NULL, NULL); } static __rte_always_inline int -- 2.51.0
