From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga11.intel.com (mga11.intel.com [192.55.52.93]) by dpdk.org (Postfix) with ESMTP id DD5D3FFA for ; Fri, 3 Jul 2015 10:32:44 +0200 (CEST) Received: from fmsmga003.fm.intel.com ([10.253.24.29]) by fmsmga102.fm.intel.com with ESMTP; 03 Jul 2015 01:32:44 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.15,398,1432623600"; d="scan'208";a="518279589" Received: from shvmail01.sh.intel.com ([10.239.29.42]) by FMSMGA003.fm.intel.com with ESMTP; 03 Jul 2015 01:32:43 -0700 Received: from shecgisg004.sh.intel.com (shecgisg004.sh.intel.com [10.239.29.89]) by shvmail01.sh.intel.com with ESMTP id t638WeJo028855; Fri, 3 Jul 2015 16:32:40 +0800 Received: from shecgisg004.sh.intel.com (localhost [127.0.0.1]) by shecgisg004.sh.intel.com (8.13.6/8.13.6/SuSE Linux 0.8) with ESMTP id t638Wadt019540; Fri, 3 Jul 2015 16:32:38 +0800 Received: (from hzhan75@localhost) by shecgisg004.sh.intel.com (8.13.6/8.13.6/Submit) id t638WaDg019536; Fri, 3 Jul 2015 16:32:36 +0800 From: Helin Zhang To: dev@dpdk.org Date: Fri, 3 Jul 2015 16:32:09 +0800 Message-Id: <1435912347-19499-2-git-send-email-helin.zhang@intel.com> X-Mailer: git-send-email 1.7.4.1 In-Reply-To: <1435912347-19499-1-git-send-email-helin.zhang@intel.com> References: <1435024235-19483-1-git-send-email-helin.zhang@intel.com> <1435912347-19499-1-git-send-email-helin.zhang@intel.com> Subject: [dpdk-dev] [PATCH v9 01/19] mbuf: redefine packet_type in rte_mbuf X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 03 Jul 2015 08:32:45 -0000 In order to unify the packet type, the field of 'packet_type' in 'struct rte_mbuf' needs to be extended from 16 to 32 bits. Accordingly, some fields in 'struct rte_mbuf' are re-organized to support this change for Vector PMD. As 'struct rte_kni_mbuf' for KNI should be right mapped to 'struct rte_mbuf', it should be modified accordingly. In ixgbe PMD driver, corresponding changes are added for the mbuf changes, especially the bit masks of packet type for 'ol_flags' are replaced by unified packet type. In addition, more packet types (UDP, TCP and SCTP) are supported in vectorized ixgbe PMD. To avoid breaking ABI compatibility, all the changes would be enabled by RTE_NEXT_ABI, which is disabled by default. Note that around 2% performance drop (64B) was observed of doing 4 ports (1 port per 82599 card) IO forwarding on the same SNB core. Signed-off-by: Helin Zhang Signed-off-by: Cunming Liang --- drivers/net/ixgbe/ixgbe_rxtx_vec.c | 75 +++++++++++++++++++++- .../linuxapp/eal/include/exec-env/rte_kni_common.h | 6 ++ lib/librte_mbuf/rte_mbuf.h | 26 ++++++++ 3 files changed, 105 insertions(+), 2 deletions(-) v2 changes: * Enlarged the packet_type field from 16 bits to 32 bits. * Redefined the packet type sub-fields. * Updated the 'struct rte_kni_mbuf' for KNI according to the mbuf changes. v3 changes: * Put the mbuf layout changes into a single patch. * Disabled vector ixgbe PMD by default, as mbuf layout changed. v5 changes: * Re-worded the commit logs. v6 changes: * Disabled the code changes for unified packet type by default, to avoid breaking ABI compatibility. v7 changes: * Renamed RTE_UNIFIED_PKT_TYPE to RTE_NEXT_ABI. * Integrated with changes of QinQ stripping/insertion. v8 changes: * Moved the field of 'vlan_tci_outer' in 'struct rte_mbuf' to the end of the 1st cache line, to avoid breaking any vectorized PMD storing. v9 changes: * Put the mbuf changes and vector PMD changes together, as they are tightly relevant. diff --git a/drivers/net/ixgbe/ixgbe_rxtx_vec.c b/drivers/net/ixgbe/ixgbe_rxtx_vec.c index abd10f6..ccea7cd 100644 --- a/drivers/net/ixgbe/ixgbe_rxtx_vec.c +++ b/drivers/net/ixgbe/ixgbe_rxtx_vec.c @@ -134,6 +134,12 @@ ixgbe_rxq_rearm(struct ixgbe_rx_queue *rxq) */ #ifdef RTE_IXGBE_RX_OLFLAGS_ENABLE +#ifdef RTE_NEXT_ABI +#define OLFLAGS_MASK_V (((uint64_t)PKT_RX_VLAN_PKT << 48) | \ + ((uint64_t)PKT_RX_VLAN_PKT << 32) | \ + ((uint64_t)PKT_RX_VLAN_PKT << 16) | \ + ((uint64_t)PKT_RX_VLAN_PKT)) +#else #define OLFLAGS_MASK ((uint16_t)(PKT_RX_VLAN_PKT | PKT_RX_IPV4_HDR |\ PKT_RX_IPV4_HDR_EXT | PKT_RX_IPV6_HDR |\ PKT_RX_IPV6_HDR_EXT)) @@ -142,11 +148,26 @@ ixgbe_rxq_rearm(struct ixgbe_rx_queue *rxq) ((uint64_t)OLFLAGS_MASK << 16) | \ ((uint64_t)OLFLAGS_MASK)) #define PTYPE_SHIFT (1) +#endif /* RTE_NEXT_ABI */ + #define VTAG_SHIFT (3) static inline void desc_to_olflags_v(__m128i descs[4], struct rte_mbuf **rx_pkts) { +#ifdef RTE_NEXT_ABI + __m128i vtag0, vtag1; + union { + uint16_t e[4]; + uint64_t dword; + } vol; + + vtag0 = _mm_unpackhi_epi16(descs[0], descs[1]); + vtag1 = _mm_unpackhi_epi16(descs[2], descs[3]); + vtag1 = _mm_unpacklo_epi32(vtag0, vtag1); + vtag1 = _mm_srli_epi16(vtag1, VTAG_SHIFT); + vol.dword = _mm_cvtsi128_si64(vtag1) & OLFLAGS_MASK_V; +#else __m128i ptype0, ptype1, vtag0, vtag1; union { uint16_t e[4]; @@ -166,6 +187,7 @@ desc_to_olflags_v(__m128i descs[4], struct rte_mbuf **rx_pkts) ptype1 = _mm_or_si128(ptype1, vtag1); vol.dword = _mm_cvtsi128_si64(ptype1) & OLFLAGS_MASK_V; +#endif /* RTE_NEXT_ABI */ rx_pkts[0]->ol_flags = vol.e[0]; rx_pkts[1]->ol_flags = vol.e[1]; @@ -196,6 +218,18 @@ _recv_raw_pkts_vec(struct ixgbe_rx_queue *rxq, struct rte_mbuf **rx_pkts, int pos; uint64_t var; __m128i shuf_msk; +#ifdef RTE_NEXT_ABI + __m128i crc_adjust = _mm_set_epi16( + 0, 0, 0, /* ignore non-length fields */ + -rxq->crc_len, /* sub crc on data_len */ + 0, /* ignore high-16bits of pkt_len */ + -rxq->crc_len, /* sub crc on pkt_len */ + 0, 0 /* ignore pkt_type field */ + ); + __m128i dd_check, eop_check; + __m128i desc_mask = _mm_set_epi32(0xFFFFFFFF, 0xFFFFFFFF, + 0xFFFFFFFF, 0xFFFF07F0); +#else __m128i crc_adjust = _mm_set_epi16( 0, 0, 0, 0, /* ignore non-length fields */ 0, /* ignore high-16bits of pkt_len */ @@ -204,6 +238,7 @@ _recv_raw_pkts_vec(struct ixgbe_rx_queue *rxq, struct rte_mbuf **rx_pkts, 0 /* ignore pkt_type field */ ); __m128i dd_check, eop_check; +#endif /* RTE_NEXT_ABI */ if (unlikely(nb_pkts < RTE_IXGBE_VPMD_RX_BURST)) return 0; @@ -232,6 +267,18 @@ _recv_raw_pkts_vec(struct ixgbe_rx_queue *rxq, struct rte_mbuf **rx_pkts, eop_check = _mm_set_epi64x(0x0000000200000002LL, 0x0000000200000002LL); /* mask to shuffle from desc. to mbuf */ +#ifdef RTE_NEXT_ABI + shuf_msk = _mm_set_epi8( + 7, 6, 5, 4, /* octet 4~7, 32bits rss */ + 15, 14, /* octet 14~15, low 16 bits vlan_macip */ + 13, 12, /* octet 12~13, 16 bits data_len */ + 0xFF, 0xFF, /* skip high 16 bits pkt_len, zero out */ + 13, 12, /* octet 12~13, low 16 bits pkt_len */ + 0xFF, 0xFF, /* skip high 16 bits pkt_type */ + 1, /* octet 1, 8 bits pkt_type field */ + 0 /* octet 0, 4 bits offset 4 pkt_type field */ + ); +#else shuf_msk = _mm_set_epi8( 7, 6, 5, 4, /* octet 4~7, 32bits rss */ 0xFF, 0xFF, /* skip high 16 bits vlan_macip, zero out */ @@ -241,18 +288,28 @@ _recv_raw_pkts_vec(struct ixgbe_rx_queue *rxq, struct rte_mbuf **rx_pkts, 13, 12, /* octet 12~13, 16 bits data_len */ 0xFF, 0xFF /* skip pkt_type field */ ); +#endif /* RTE_NEXT_ABI */ /* Cache is empty -> need to scan the buffer rings, but first move * the next 'n' mbufs into the cache */ sw_ring = &rxq->sw_ring[rxq->rx_tail]; - /* - * A. load 4 packet in one loop +#ifdef RTE_NEXT_ABI + /* A. load 4 packet in one loop + * [A*. mask out 4 unused dirty field in desc] * B. copy 4 mbuf point from swring to rx_pkts * C. calc the number of DD bits among the 4 packets * [C*. extract the end-of-packet bit, if requested] * D. fill info. from desc to mbuf */ +#else + /* A. load 4 packet in one loop + * B. copy 4 mbuf point from swring to rx_pkts + * C. calc the number of DD bits among the 4 packets + * [C*. extract the end-of-packet bit, if requested] + * D. fill info. from desc to mbuf + */ +#endif /* RTE_NEXT_ABI */ for (pos = 0, nb_pkts_recd = 0; pos < RTE_IXGBE_VPMD_RX_BURST; pos += RTE_IXGBE_DESCS_PER_LOOP, rxdp += RTE_IXGBE_DESCS_PER_LOOP) { @@ -289,6 +346,16 @@ _recv_raw_pkts_vec(struct ixgbe_rx_queue *rxq, struct rte_mbuf **rx_pkts, /* B.2 copy 2 mbuf point into rx_pkts */ _mm_storeu_si128((__m128i *)&rx_pkts[pos+2], mbp2); +#ifdef RTE_NEXT_ABI + /* A* mask out 0~3 bits RSS type */ + descs[3] = _mm_and_si128(descs[3], desc_mask); + descs[2] = _mm_and_si128(descs[2], desc_mask); + + /* A* mask out 0~3 bits RSS type */ + descs[1] = _mm_and_si128(descs[1], desc_mask); + descs[0] = _mm_and_si128(descs[0], desc_mask); +#endif /* RTE_NEXT_ABI */ + /* avoid compiler reorder optimization */ rte_compiler_barrier(); @@ -301,7 +368,11 @@ _recv_raw_pkts_vec(struct ixgbe_rx_queue *rxq, struct rte_mbuf **rx_pkts, /* C.1 4=>2 filter staterr info only */ sterr_tmp1 = _mm_unpackhi_epi32(descs[1], descs[0]); +#ifdef RTE_NEXT_ABI + /* set ol_flags with vlan packet type */ +#else /* set ol_flags with packet type and vlan tag */ +#endif /* RTE_NEXT_ABI */ desc_to_olflags_v(descs, &rx_pkts[pos]); /* D.2 pkt 3,4 set in_port/nb_seg and remove crc */ diff --git a/lib/librte_eal/linuxapp/eal/include/exec-env/rte_kni_common.h b/lib/librte_eal/linuxapp/eal/include/exec-env/rte_kni_common.h index 1e55c2d..e9f38bd 100644 --- a/lib/librte_eal/linuxapp/eal/include/exec-env/rte_kni_common.h +++ b/lib/librte_eal/linuxapp/eal/include/exec-env/rte_kni_common.h @@ -117,9 +117,15 @@ struct rte_kni_mbuf { uint16_t data_off; /**< Start address of data in segment buffer. */ char pad1[4]; uint64_t ol_flags; /**< Offload features. */ +#ifdef RTE_NEXT_ABI + char pad2[4]; + uint32_t pkt_len; /**< Total pkt len: sum of all segment data_len. */ + uint16_t data_len; /**< Amount of data in segment buffer. */ +#else char pad2[2]; uint16_t data_len; /**< Amount of data in segment buffer. */ uint32_t pkt_len; /**< Total pkt len: sum of all segment data_len. */ +#endif /* fields on second cache line */ char pad3[8] __attribute__((__aligned__(RTE_CACHE_LINE_SIZE))); diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h index 80419df..ac29da3 100644 --- a/lib/librte_mbuf/rte_mbuf.h +++ b/lib/librte_mbuf/rte_mbuf.h @@ -276,6 +276,28 @@ struct rte_mbuf { /* remaining bytes are set on RX when pulling packet from descriptor */ MARKER rx_descriptor_fields1; +#ifdef RTE_NEXT_ABI + /* + * The packet type, which is the combination of outer/inner L2, L3, L4 + * and tunnel types. + */ + union { + uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */ + struct { + uint32_t l2_type:4; /**< (Outer) L2 type. */ + uint32_t l3_type:4; /**< (Outer) L3 type. */ + uint32_t l4_type:4; /**< (Outer) L4 type. */ + uint32_t tun_type:4; /**< Tunnel type. */ + uint32_t inner_l2_type:4; /**< Inner L2 type. */ + uint32_t inner_l3_type:4; /**< Inner L3 type. */ + uint32_t inner_l4_type:4; /**< Inner L4 type. */ + }; + }; + + uint32_t pkt_len; /**< Total pkt len: sum of all segments. */ + uint16_t data_len; /**< Amount of data in segment buffer. */ + uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */ +#else /* RTE_NEXT_ABI */ /** * The packet type, which is used to indicate ordinary packet and also * tunneled packet format, i.e. each number is represented a type of @@ -287,6 +309,7 @@ struct rte_mbuf { uint32_t pkt_len; /**< Total pkt len: sum of all segments. */ uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */ uint16_t vlan_tci_outer; /**< Outer VLAN Tag Control Identifier (CPU order) */ +#endif /* RTE_NEXT_ABI */ union { uint32_t rss; /**< RSS hash result if RSS enabled */ struct { @@ -307,6 +330,9 @@ struct rte_mbuf { } hash; /**< hash information */ uint32_t seqn; /**< Sequence number. See also rte_reorder_insert() */ +#ifdef RTE_NEXT_ABI + uint16_t vlan_tci_outer; /**< Outer VLAN Tag Control Identifier (CPU order) */ +#endif /* RTE_NEXT_ABI */ /* second cache line - fields only used in slow path or on TX */ MARKER cacheline1 __rte_cache_aligned; -- 1.9.3