[dpdk-dev] [PATCH v2] net/ixgbe: fix IPsec Rx with SSE on

[Radu Nicolau <radu.nicolau@intel.com>]

Fixes: 9a0752f498d2 ("net/ixgbe: enable inline IPsec")
And small performance improvement

Signed-off-by: Radu Nicolau <radu.nicolau@intel.com>
Acked-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
---

v2: fixed build on gcc 4.8.5

 drivers/net/ixgbe/ixgbe_rxtx_vec_sse.c | 80 +++++++++++++---------------------
 1 file changed, 31 insertions(+), 49 deletions(-)

diff --git a/drivers/net/ixgbe/ixgbe_rxtx_vec_sse.c b/drivers/net/ixgbe/ixgbe_rxtx_vec_sse.c
index 158ccd9..7258ff9 100644
--- a/drivers/net/ixgbe/ixgbe_rxtx_vec_sse.c
+++ b/drivers/net/ixgbe/ixgbe_rxtx_vec_sse.c
@@ -126,54 +126,36 @@ ixgbe_rxq_rearm(struct ixgbe_rx_queue *rxq)
 static inline void
 desc_to_olflags_v_ipsec(__m128i descs[4], struct rte_mbuf **rx_pkts)
 {
-	__m128i sterr0, sterr1, sterr2, sterr3;
-	__m128i tmp1, tmp2, tmp3, tmp4;
-	__m128i rearm0, rearm1, rearm2, rearm3;
-
-	const __m128i ipsec_sterr_msk = _mm_set_epi32(
-		0, IXGBE_RXDADV_IPSEC_STATUS_SECP |
-			IXGBE_RXDADV_IPSEC_ERROR_AUTH_FAILED,
-		0, 0);
-	const __m128i ipsec_proc_msk  = _mm_set_epi32(
-		0, IXGBE_RXDADV_IPSEC_STATUS_SECP, 0, 0);
-	const __m128i ipsec_err_flag  = _mm_set_epi32(
-		0, PKT_RX_SEC_OFFLOAD_FAILED | PKT_RX_SEC_OFFLOAD,
-		0, 0);
-	const __m128i ipsec_proc_flag = _mm_set_epi32(
-		0, PKT_RX_SEC_OFFLOAD, 0, 0);
-
-	rearm0 = _mm_load_si128((__m128i *)&rx_pkts[0]->rearm_data);
-	rearm1 = _mm_load_si128((__m128i *)&rx_pkts[1]->rearm_data);
-	rearm2 = _mm_load_si128((__m128i *)&rx_pkts[2]->rearm_data);
-	rearm3 = _mm_load_si128((__m128i *)&rx_pkts[3]->rearm_data);
-	sterr0 = _mm_and_si128(descs[0], ipsec_sterr_msk);
-	sterr1 = _mm_and_si128(descs[1], ipsec_sterr_msk);
-	sterr2 = _mm_and_si128(descs[2], ipsec_sterr_msk);
-	sterr3 = _mm_and_si128(descs[3], ipsec_sterr_msk);
-	tmp1 = _mm_cmpeq_epi32(sterr0, ipsec_sterr_msk);
-	tmp2 = _mm_cmpeq_epi32(sterr0, ipsec_proc_msk);
-	tmp3 = _mm_cmpeq_epi32(sterr1, ipsec_sterr_msk);
-	tmp4 = _mm_cmpeq_epi32(sterr1, ipsec_proc_msk);
-	sterr0 = _mm_or_si128(_mm_and_si128(tmp1, ipsec_err_flag),
-				_mm_and_si128(tmp2, ipsec_proc_flag));
-	sterr1 = _mm_or_si128(_mm_and_si128(tmp3, ipsec_err_flag),
-				_mm_and_si128(tmp4, ipsec_proc_flag));
-	tmp1 = _mm_cmpeq_epi32(sterr2, ipsec_sterr_msk);
-	tmp2 = _mm_cmpeq_epi32(sterr2, ipsec_proc_msk);
-	tmp3 = _mm_cmpeq_epi32(sterr3, ipsec_sterr_msk);
-	tmp4 = _mm_cmpeq_epi32(sterr3, ipsec_proc_msk);
-	sterr2 = _mm_or_si128(_mm_and_si128(tmp1, ipsec_err_flag),
-				_mm_and_si128(tmp2, ipsec_proc_flag));
-	sterr3 = _mm_or_si128(_mm_and_si128(tmp3, ipsec_err_flag),
-				_mm_and_si128(tmp4, ipsec_proc_flag));
-	rearm0 = _mm_or_si128(rearm0, sterr0);
-	rearm1 = _mm_or_si128(rearm1, sterr1);
-	rearm2 = _mm_or_si128(rearm2, sterr2);
-	rearm3 = _mm_or_si128(rearm3, sterr3);
-	_mm_store_si128((__m128i *)&rx_pkts[0]->rearm_data, rearm0);
-	_mm_store_si128((__m128i *)&rx_pkts[1]->rearm_data, rearm1);
-	_mm_store_si128((__m128i *)&rx_pkts[2]->rearm_data, rearm2);
-	_mm_store_si128((__m128i *)&rx_pkts[3]->rearm_data, rearm3);
+	__m128i sterr, rearm, tmp_e, tmp_p;
+	uint32_t *rearm0 = (uint32_t *)rx_pkts[0]->rearm_data + 2;
+	uint32_t *rearm1 = (uint32_t *)rx_pkts[1]->rearm_data + 2;
+	uint32_t *rearm2 = (uint32_t *)rx_pkts[2]->rearm_data + 2;
+	uint32_t *rearm3 = (uint32_t *)rx_pkts[3]->rearm_data + 2;
+	const __m128i ipsec_sterr_msk =
+			_mm_set1_epi32(IXGBE_RXDADV_IPSEC_STATUS_SECP |
+				       IXGBE_RXDADV_IPSEC_ERROR_AUTH_FAILED);
+	const __m128i ipsec_proc_msk  =
+			_mm_set1_epi32(IXGBE_RXDADV_IPSEC_STATUS_SECP);
+	const __m128i ipsec_err_flag  =
+			_mm_set1_epi32(PKT_RX_SEC_OFFLOAD_FAILED |
+				       PKT_RX_SEC_OFFLOAD);
+	const __m128i ipsec_proc_flag = _mm_set1_epi32(PKT_RX_SEC_OFFLOAD);
+
+	rearm = _mm_set_epi32(*rearm3, *rearm2, *rearm1, *rearm0);
+	sterr = _mm_set_epi32(_mm_extract_epi32(descs[3], 2),
+			      _mm_extract_epi32(descs[2], 2),
+			      _mm_extract_epi32(descs[1], 2),
+			      _mm_extract_epi32(descs[0], 2));
+	sterr = _mm_and_si128(sterr, ipsec_sterr_msk);
+	tmp_e = _mm_cmpeq_epi32(sterr, ipsec_sterr_msk);
+	tmp_p = _mm_cmpeq_epi32(sterr, ipsec_proc_msk);
+	sterr = _mm_or_si128(_mm_and_si128(tmp_e, ipsec_err_flag),
+				_mm_and_si128(tmp_p, ipsec_proc_flag));
+	rearm = _mm_or_si128(rearm, sterr);
+	*rearm0 = _mm_extract_epi32(rearm, 0);
+	*rearm1 = _mm_extract_epi32(rearm, 1);
+	*rearm2 = _mm_extract_epi32(rearm, 2);
+	*rearm3 = _mm_extract_epi32(rearm, 3);
 }
 #endif
 
@@ -533,7 +515,7 @@ _recv_raw_pkts_vec(struct ixgbe_rx_queue *rxq, struct rte_mbuf **rx_pkts,
 
 #ifdef RTE_LIBRTE_SECURITY
 		if (unlikely(use_ipsec))
-			desc_to_olflags_v_ipsec(descs, rx_pkts);
+			desc_to_olflags_v_ipsec(descs, &rx_pkts[pos]);
 #endif
 
 		/* D.2 pkt 3,4 set in_port/nb_seg and remove crc */
-- 
2.7.5