From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga05.intel.com (mga05.intel.com [192.55.52.43]) by dpdk.org (Postfix) with ESMTP id 279DA1B223 for ; Tue, 31 Oct 2017 16:47:34 +0100 (CET) Received: from orsmga003.jf.intel.com ([10.7.209.27]) by fmsmga105.fm.intel.com with ESMTP; 31 Oct 2017 08:47:33 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.44,324,1505804400"; d="scan'208";a="1031684757" Received: from irsmsx151.ger.corp.intel.com ([163.33.192.59]) by orsmga003.jf.intel.com with ESMTP; 31 Oct 2017 08:47:32 -0700 Received: from irsmsx112.ger.corp.intel.com (10.108.20.5) by IRSMSX151.ger.corp.intel.com (163.33.192.59) with Microsoft SMTP Server (TLS) id 14.3.319.2; Tue, 31 Oct 2017 15:47:32 +0000 Received: from irsmsx105.ger.corp.intel.com ([169.254.7.67]) by irsmsx112.ger.corp.intel.com ([169.254.1.12]) with mapi id 14.03.0319.002; Tue, 31 Oct 2017 15:47:31 +0000 From: "Ananyev, Konstantin" To: "Nicolau, Radu" , "dev@dpdk.org" Thread-Topic: [PATCH] net/ixgbe: fix IPsec Rx with SSE on Thread-Index: AQHTTyenv+avbuCTdU+CHrDwIRc23KL+IIOg Date: Tue, 31 Oct 2017 15:47:30 +0000 Message-ID: <2601191342CEEE43887BDE71AB9772585FAB5C14@irsmsx105.ger.corp.intel.com> References: <1509110679-23926-1-git-send-email-radu.nicolau@intel.com> In-Reply-To: <1509110679-23926-1-git-send-email-radu.nicolau@intel.com> Accept-Language: en-IE, en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: x-titus-metadata-40: eyJDYXRlZ29yeUxhYmVscyI6IiIsIk1ldGFkYXRhIjp7Im5zIjoiaHR0cDpcL1wvd3d3LnRpdHVzLmNvbVwvbnNcL0ludGVsMyIsImlkIjoiMTQzNTU1YTctZmIxZS00YjJmLTkzMGEtYWNhNDk0NWFlNWMyIiwicHJvcHMiOlt7Im4iOiJDVFBDbGFzc2lmaWNhdGlvbiIsInZhbHMiOlt7InZhbHVlIjoiQ1RQX0lDIn1dfV19LCJTdWJqZWN0TGFiZWxzIjpbXSwiVE1DVmVyc2lvbiI6IjE2LjUuOS4zIiwiVHJ1c3RlZExhYmVsSGFzaCI6IjNMdEJSQkx3dGNNNW9VaWtVSHNKRG9JY2NzbjN3NFF5MFMrQTdYeHcxaUU9In0= x-ctpclassification: CTP_IC dlp-product: dlpe-windows dlp-version: 11.0.0.116 dlp-reaction: no-action x-originating-ip: [163.33.239.180] Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Subject: Re: [dpdk-dev] [PATCH] net/ixgbe: fix IPsec Rx with SSE on X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 31 Oct 2017 15:47:35 -0000 > -----Original Message----- > From: Nicolau, Radu > Sent: Friday, October 27, 2017 2:25 PM > To: dev@dpdk.org > Cc: Ananyev, Konstantin ; Nicolau, Radu > Subject: [PATCH] net/ixgbe: fix IPsec Rx with SSE on >=20 > Fixes: 9a0752f498d2 ("net/ixgbe: enable inline IPsec") > And small performance improvement >=20 > Signed-off-by: Radu Nicolau > --- > drivers/net/ixgbe/ixgbe_rxtx_vec_sse.c | 79 +++++++++++++---------------= ------ > 1 file changed, 30 insertions(+), 49 deletions(-) >=20 > diff --git a/drivers/net/ixgbe/ixgbe_rxtx_vec_sse.c b/drivers/net/ixgbe/i= xgbe_rxtx_vec_sse.c > index 158ccd9..6d7ed5d 100644 > --- a/drivers/net/ixgbe/ixgbe_rxtx_vec_sse.c > +++ b/drivers/net/ixgbe/ixgbe_rxtx_vec_sse.c > @@ -126,54 +126,35 @@ 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 =3D _mm_set_epi32( > - 0, IXGBE_RXDADV_IPSEC_STATUS_SECP | > - IXGBE_RXDADV_IPSEC_ERROR_AUTH_FAILED, > - 0, 0); > - const __m128i ipsec_proc_msk =3D _mm_set_epi32( > - 0, IXGBE_RXDADV_IPSEC_STATUS_SECP, 0, 0); > - const __m128i ipsec_err_flag =3D _mm_set_epi32( > - 0, PKT_RX_SEC_OFFLOAD_FAILED | PKT_RX_SEC_OFFLOAD, > - 0, 0); > - const __m128i ipsec_proc_flag =3D _mm_set_epi32( > - 0, PKT_RX_SEC_OFFLOAD, 0, 0); > - > - rearm0 =3D _mm_load_si128((__m128i *)&rx_pkts[0]->rearm_data); > - rearm1 =3D _mm_load_si128((__m128i *)&rx_pkts[1]->rearm_data); > - rearm2 =3D _mm_load_si128((__m128i *)&rx_pkts[2]->rearm_data); > - rearm3 =3D _mm_load_si128((__m128i *)&rx_pkts[3]->rearm_data); > - sterr0 =3D _mm_and_si128(descs[0], ipsec_sterr_msk); > - sterr1 =3D _mm_and_si128(descs[1], ipsec_sterr_msk); > - sterr2 =3D _mm_and_si128(descs[2], ipsec_sterr_msk); > - sterr3 =3D _mm_and_si128(descs[3], ipsec_sterr_msk); > - tmp1 =3D _mm_cmpeq_epi32(sterr0, ipsec_sterr_msk); > - tmp2 =3D _mm_cmpeq_epi32(sterr0, ipsec_proc_msk); > - tmp3 =3D _mm_cmpeq_epi32(sterr1, ipsec_sterr_msk); > - tmp4 =3D _mm_cmpeq_epi32(sterr1, ipsec_proc_msk); > - sterr0 =3D _mm_or_si128(_mm_and_si128(tmp1, ipsec_err_flag), > - _mm_and_si128(tmp2, ipsec_proc_flag)); > - sterr1 =3D _mm_or_si128(_mm_and_si128(tmp3, ipsec_err_flag), > - _mm_and_si128(tmp4, ipsec_proc_flag)); > - tmp1 =3D _mm_cmpeq_epi32(sterr2, ipsec_sterr_msk); > - tmp2 =3D _mm_cmpeq_epi32(sterr2, ipsec_proc_msk); > - tmp3 =3D _mm_cmpeq_epi32(sterr3, ipsec_sterr_msk); > - tmp4 =3D _mm_cmpeq_epi32(sterr3, ipsec_proc_msk); > - sterr2 =3D _mm_or_si128(_mm_and_si128(tmp1, ipsec_err_flag), > - _mm_and_si128(tmp2, ipsec_proc_flag)); > - sterr3 =3D _mm_or_si128(_mm_and_si128(tmp3, ipsec_err_flag), > - _mm_and_si128(tmp4, ipsec_proc_flag)); > - rearm0 =3D _mm_or_si128(rearm0, sterr0); > - rearm1 =3D _mm_or_si128(rearm1, sterr1); > - rearm2 =3D _mm_or_si128(rearm2, sterr2); > - rearm3 =3D _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; > + const __m128i ipsec_sterr_msk =3D > + _mm_set1_epi32(IXGBE_RXDADV_IPSEC_STATUS_SECP | > + IXGBE_RXDADV_IPSEC_ERROR_AUTH_FAILED); > + const __m128i ipsec_proc_msk =3D > + _mm_set1_epi32(IXGBE_RXDADV_IPSEC_STATUS_SECP); > + const __m128i ipsec_err_flag =3D > + _mm_set1_epi32(PKT_RX_SEC_OFFLOAD_FAILED | > + PKT_RX_SEC_OFFLOAD); > + const __m128i ipsec_proc_flag =3D _mm_set1_epi32(PKT_RX_SEC_OFFLOAD); > + > + rearm =3D _mm_set_epi32(((uint32_t *)rx_pkts[0]->rearm_data)[2], > + ((uint32_t *)rx_pkts[1]->rearm_data)[2], > + ((uint32_t *)rx_pkts[2]->rearm_data)[2], > + ((uint32_t *)rx_pkts[3]->rearm_data)[2]); > + sterr =3D _mm_set_epi32(_mm_extract_epi32(descs[0], 2), > + _mm_extract_epi32(descs[1], 2), > + _mm_extract_epi32(descs[2], 2), > + _mm_extract_epi32(descs[3], 2)); > + sterr =3D _mm_and_si128(sterr, ipsec_sterr_msk); > + tmp_e =3D _mm_cmpeq_epi32(sterr, ipsec_sterr_msk); > + tmp_p =3D _mm_cmpeq_epi32(sterr, ipsec_proc_msk); > + sterr =3D _mm_or_si128(_mm_and_si128(tmp_e, ipsec_err_flag), > + _mm_and_si128(tmp_p, ipsec_proc_flag)); > + rearm =3D _mm_or_si128(rearm, sterr); > + ((uint32_t *)rx_pkts[0]->rearm_data)[2] =3D _mm_extract_epi32(rearm, 3)= ; > + ((uint32_t *)rx_pkts[1]->rearm_data)[2] =3D _mm_extract_epi32(rearm, 2)= ; > + ((uint32_t *)rx_pkts[2]->rearm_data)[2] =3D _mm_extract_epi32(rearm, 1)= ; > + ((uint32_t *)rx_pkts[3]->rearm_data)[2] =3D _mm_extract_epi32(rearm, 0)= ; > } > #endif >=20 > @@ -533,7 +514,7 @@ _recv_raw_pkts_vec(struct ixgbe_rx_queue *rxq, struct= rte_mbuf **rx_pkts, >=20 > #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 >=20 > /* D.2 pkt 3,4 set in_port/nb_seg and remove crc */ > -- Acked-by: Konstantin Ananyev > 2.7.5