DPDK patches and discussions
 help / color / mirror / Atom feed
From: Maciej Gajdzica <maciejx.t.gajdzica@intel.com>
To: dev@dpdk.org
Subject: [dpdk-dev] [PATCH v2 08/11] ip_pipeline: added new implementation of passthrough pipeline
Date: Thu, 25 Jun 2015 13:15:11 +0200	[thread overview]
Message-ID: <1435230914-8174-9-git-send-email-maciejx.t.gajdzica@intel.com> (raw)
In-Reply-To: <1435230914-8174-1-git-send-email-maciejx.t.gajdzica@intel.com>

From: Jasvinder Singh <jasvinder.singh@intel.com>

Passthrough pipeline implementation is split to two files.
pipeline_passthrough.c file handles front-end functions (cli commands
parsing) pipeline_passthrough_ops.c contains implementation of functions
done by pipeline (back-end).

Signed-off-by: Jasvinder Singh <jasvinder.singh@intel.com>
---
 examples/ip_pipeline/Makefile                      |    2 +
 examples/ip_pipeline/config/pt1.cfg                |    9 +
 examples/ip_pipeline/config/pt2.cfg                |   15 +
 examples/ip_pipeline/config/pt3.cfg                |   21 +
 examples/ip_pipeline/init.c                        |    2 +
 examples/ip_pipeline/pipeline/hash_func.h          |  351 ++++++++++
 .../ip_pipeline/pipeline/pipeline_actions_common.h |  119 ++++
 .../ip_pipeline/pipeline/pipeline_passthrough.c    |  192 +----
 .../ip_pipeline/pipeline/pipeline_passthrough.h    |   41 ++
 .../ip_pipeline/pipeline/pipeline_passthrough_be.c |  741 ++++++++++++++++++++
 .../ip_pipeline/pipeline/pipeline_passthrough_be.h |   41 ++
 11 files changed, 1355 insertions(+), 179 deletions(-)
 create mode 100644 examples/ip_pipeline/config/pt1.cfg
 create mode 100644 examples/ip_pipeline/config/pt2.cfg
 create mode 100644 examples/ip_pipeline/config/pt3.cfg
 create mode 100644 examples/ip_pipeline/pipeline/hash_func.h
 create mode 100644 examples/ip_pipeline/pipeline/pipeline_actions_common.h
 create mode 100644 examples/ip_pipeline/pipeline/pipeline_passthrough.h
 create mode 100644 examples/ip_pipeline/pipeline/pipeline_passthrough_be.c
 create mode 100644 examples/ip_pipeline/pipeline/pipeline_passthrough_be.h

diff --git a/examples/ip_pipeline/Makefile b/examples/ip_pipeline/Makefile
index f255338..930dc61 100644
--- a/examples/ip_pipeline/Makefile
+++ b/examples/ip_pipeline/Makefile
@@ -60,6 +60,8 @@ SRCS-$(CONFIG_RTE_LIBRTE_PIPELINE) += pipeline_common_be.c
 SRCS-$(CONFIG_RTE_LIBRTE_PIPELINE) += pipeline_common_fe.c
 SRCS-$(CONFIG_RTE_LIBRTE_PIPELINE) += pipeline_master_be.c
 SRCS-$(CONFIG_RTE_LIBRTE_PIPELINE) += pipeline_master.c
+SRCS-$(CONFIG_RTE_LIBRTE_PIPELINE) += pipeline_passthrough_be.c
+SRCS-$(CONFIG_RTE_LIBRTE_PIPELINE) += pipeline_passthrough.c
 
 CFLAGS += -I$(SRCDIR) -I$(SRCDIR)/pipeline
 CFLAGS += -O3
diff --git a/examples/ip_pipeline/config/pt1.cfg b/examples/ip_pipeline/config/pt1.cfg
new file mode 100644
index 0000000..c9cdc78
--- /dev/null
+++ b/examples/ip_pipeline/config/pt1.cfg
@@ -0,0 +1,9 @@
+[PIPELINE0]

+type = MASTER

+core = 0

+

+[PIPELINE1]

+type = PASS-THROUGH

+core = s0c1

+pktq_in = RXQ0.0 RXQ1.0 RXQ2.0 RXQ3.0

+pktq_out = TXQ0.0 TXQ1.0 TXQ2.0 TXQ3.0

diff --git a/examples/ip_pipeline/config/pt2.cfg b/examples/ip_pipeline/config/pt2.cfg
new file mode 100644
index 0000000..860cab6
--- /dev/null
+++ b/examples/ip_pipeline/config/pt2.cfg
@@ -0,0 +1,15 @@
+[PIPELINE0]

+type = MASTER

+core = 0

+

+[PIPELINE1]

+type = PASS-THROUGH

+core = s0c1

+pktq_in = RXQ0.0 RXQ1.0 RXQ2.0 RXQ3.0

+pktq_out = SWQ0 SWQ1 SWQ2 SWQ3

+

+[PIPELINE2]

+type = PASS-THROUGH

+core = s0c2

+pktq_in = SWQ0 SWQ1 SWQ2 SWQ3

+pktq_out = TXQ0.0 TXQ1.0 TXQ2.0 TXQ3.0

diff --git a/examples/ip_pipeline/config/pt3.cfg b/examples/ip_pipeline/config/pt3.cfg
new file mode 100644
index 0000000..e6159d0
--- /dev/null
+++ b/examples/ip_pipeline/config/pt3.cfg
@@ -0,0 +1,21 @@
+[PIPELINE0]

+type = MASTER

+core = 0

+

+[PIPELINE1]

+type = PASS-THROUGH

+core = s0c1

+pktq_in = RXQ0.0 RXQ1.0 RXQ2.0 RXQ3.0

+pktq_out = SWQ0 SWQ1 SWQ2 SWQ3

+

+[PIPELINE2]

+type = PASS-THROUGH

+core = s0c2

+pktq_in = SWQ0 SWQ1 SWQ2 SWQ3

+pktq_out = SWQ4 SWQ5 SWQ6 SWQ7

+

+[PIPELINE3]

+type = PASS-THROUGH

+core = s0c3

+pktq_in = SWQ4 SWQ5 SWQ6 SWQ7

+pktq_out = TXQ0.0 TXQ1.0 TXQ2.0 TXQ3.0

diff --git a/examples/ip_pipeline/init.c b/examples/ip_pipeline/init.c
index b362af0..711c243 100644
--- a/examples/ip_pipeline/init.c
+++ b/examples/ip_pipeline/init.c
@@ -45,6 +45,7 @@
 #include "pipeline.h"
 #include "pipeline_common_fe.h"
 #include "pipeline_master.h"
+#include "pipeline_passthrough.h"
 
 #define APP_NAME_SIZE	32
 
@@ -1189,6 +1190,7 @@ int app_init(struct app_params *app)
 
 	app_pipeline_common_cmd_push(app);
 	app_pipeline_type_register(app, &pipeline_master);
+	app_pipeline_type_register(app, &pipeline_passthrough);
 
 	app_init_pipelines(app);
 	app_init_threads(app);
diff --git a/examples/ip_pipeline/pipeline/hash_func.h b/examples/ip_pipeline/pipeline/hash_func.h
new file mode 100644
index 0000000..0d9c019
--- /dev/null
+++ b/examples/ip_pipeline/pipeline/hash_func.h
@@ -0,0 +1,351 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#ifndef __INCLUDE_HASH_FUNC_H__
+#define __INCLUDE_HASH_FUNC_H__
+
+static inline uint64_t
+hash_xor_key8(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t xor0;
+
+	xor0 = seed ^ k[0];
+
+	return (xor0 >> 32) ^ xor0;
+}
+
+static inline uint64_t
+hash_xor_key16(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t xor0;
+
+	xor0 = (k[0] ^ seed) ^ k[1];
+
+	return (xor0 >> 32) ^ xor0;
+}
+
+static inline uint64_t
+hash_xor_key24(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t xor0;
+
+	xor0 = (k[0] ^ seed) ^ k[1];
+
+	xor0 ^= k[2]; 
+
+	return (xor0 >> 32) ^ xor0;
+}
+
+static inline uint64_t
+hash_xor_key32(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t xor0, xor1;
+
+	xor0 = (k[0] ^ seed) ^ k[1];
+	xor1 = k[2] ^ k[3];
+
+	xor0 ^= xor1; 
+
+	return (xor0 >> 32) ^ xor0;
+}
+
+static inline uint64_t
+hash_xor_key40(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t xor0, xor1;
+
+	xor0 = (k[0] ^ seed) ^ k[1];
+	xor1 = k[2] ^ k[3];
+
+	xor0 ^= xor1;
+
+	xor0 ^= k[4];
+
+	return (xor0 >> 32) ^ xor0;
+}
+
+static inline uint64_t
+hash_xor_key48(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t xor0, xor1, xor2;
+
+	xor0 = (k[0] ^ seed) ^ k[1];
+	xor1 = k[2] ^ k[3];
+	xor2 = k[4] ^ k[5];
+
+	xor0 ^= xor1;
+
+	xor0 ^= xor2;
+
+	return (xor0 >> 32) ^ xor0;
+}
+
+static inline uint64_t
+hash_xor_key56(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t xor0, xor1, xor2;
+
+	xor0 = (k[0] ^ seed) ^ k[1];
+	xor1 = k[2] ^ k[3];
+	xor2 = k[4] ^ k[5];
+
+	xor0 ^= xor1;
+	xor2 ^= k[6];
+
+	xor0 ^= xor2;
+
+	return (xor0 >> 32) ^ xor0;
+}
+
+static inline uint64_t
+hash_xor_key64(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t xor0, xor1, xor2, xor3;
+
+	xor0 = (k[0] ^ seed) ^ k[1];
+	xor1 = k[2] ^ k[3];
+	xor2 = k[4] ^ k[5];
+	xor3 = k[6] ^ k[7];
+
+	xor0 ^= xor1;
+	xor2 ^= xor3;
+
+	xor0 ^= xor2;
+
+	return (xor0 >> 32) ^ xor0;
+}
+
+#if defined(__x86_64__)
+
+#include <x86intrin.h>
+
+static inline uint64_t
+hash_crc_key8(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t crc0;
+
+	crc0 = _mm_crc32_u64(seed, k[0]);
+
+	return crc0;
+}
+
+static inline uint64_t
+hash_crc_key16(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t k0, crc0, crc1;
+
+	k0 = k[0];
+
+	crc0 = _mm_crc32_u64(k0, seed);
+	crc1 = _mm_crc32_u64(k0 >> 32, k[1]);
+
+	crc0 ^= crc1;
+
+	return crc0;
+}
+
+static inline uint64_t
+hash_crc_key24(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t k0, k2, crc0, crc1;
+	
+	k0 = k[0];
+	k2 = k[2];
+
+	crc0 = _mm_crc32_u64(k0, seed);
+	crc1 = _mm_crc32_u64(k0 >> 32, k[1]);
+		
+	crc0 = _mm_crc32_u64(crc0, k2);
+
+	crc0 ^= crc1;
+
+	return crc0;
+}
+
+static inline uint64_t
+hash_crc_key32(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t k0, k2, crc0, crc1, crc2, crc3;
+	
+	k0 = k[0];
+	k2 = k[2];
+
+	crc0 = _mm_crc32_u64(k0, seed);
+	crc1 = _mm_crc32_u64(k0 >> 32, k[1]);
+	
+	crc2 = _mm_crc32_u64(k2, k[3]);
+	crc3 = k2 >> 32;
+	
+	crc0 = _mm_crc32_u64(crc0, crc1);
+	crc1 = _mm_crc32_u64(crc2, crc3);
+
+	crc0 ^= crc1;
+
+	return crc0;
+}
+
+static inline uint64_t
+hash_crc_key40(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t k0, k2, crc0, crc1, crc2, crc3;
+	
+	k0 = k[0];
+	k2 = k[2];
+
+	crc0 = _mm_crc32_u64(k0, seed);
+	crc1 = _mm_crc32_u64(k0 >> 32, k[1]);
+	
+	crc2 = _mm_crc32_u64(k2, k[3]);
+	crc3 = _mm_crc32_u64(k2 >> 32, k[4]);
+	
+	crc0 = _mm_crc32_u64(crc0, crc1);
+	crc1 = _mm_crc32_u64(crc2, crc3);
+
+	crc0 ^= crc1;
+
+	return crc0;
+}
+
+static inline uint64_t
+hash_crc_key48(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t k0, k2, k5, crc0, crc1, crc2, crc3;
+	
+	k0 = k[0];
+	k2 = k[2];
+	k5 = k[5];
+
+	crc0 = _mm_crc32_u64(k0, seed);
+	crc1 = _mm_crc32_u64(k0 >> 32, k[1]);
+	
+	crc2 = _mm_crc32_u64(k2, k[3]);
+	crc3 = _mm_crc32_u64(k2 >> 32, k[4]);
+
+	crc0 = _mm_crc32_u64(crc0, (crc1 << 32) ^ crc2);
+	crc1 = _mm_crc32_u64(crc3, k5);
+
+	crc0 ^= crc1;
+
+	return crc0;
+}
+
+static inline uint64_t
+hash_crc_key56(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t k0, k2, k5, crc0, crc1, crc2, crc3, crc4, crc5;
+	
+	k0 = k[0];
+	k2 = k[2];
+	k5 = k[5];
+
+	crc0 = _mm_crc32_u64(k0, seed);
+	crc1 = _mm_crc32_u64(k0 >> 32, k[1]);
+	
+	crc2 = _mm_crc32_u64(k2, k[3]);
+	crc3 = _mm_crc32_u64(k2 >> 32, k[4]);
+
+	crc4 = _mm_crc32_u64(k5, k[6]);
+	crc5 = k5 >> 32;
+	
+	crc0 = _mm_crc32_u64(crc0, (crc1 << 32) ^ crc2);
+	crc1 = _mm_crc32_u64(crc3, (crc4 << 32) ^ crc5);
+
+	crc0 ^= crc1;
+
+	return crc0;
+}
+
+static inline uint64_t
+hash_crc_key64(void *key, __rte_unused uint32_t key_size, uint64_t seed)
+{
+	uint64_t *k = key;
+	uint64_t k0, k2, k5, crc0, crc1, crc2, crc3, crc4, crc5;
+	
+	k0 = k[0];
+	k2 = k[2];
+	k5 = k[5];
+
+	crc0 = _mm_crc32_u64(k0, seed);
+	crc1 = _mm_crc32_u64(k0 >> 32, k[1]);
+	
+	crc2 = _mm_crc32_u64(k2, k[3]);
+	crc3 = _mm_crc32_u64(k2 >> 32, k[4]);
+
+	crc4 = _mm_crc32_u64(k5, k[6]);
+	crc5 = _mm_crc32_u64(k5 >> 32, k[7]);
+	
+	crc0 = _mm_crc32_u64(crc0, (crc1 << 32) ^ crc2);
+	crc1 = _mm_crc32_u64(crc3, (crc4 << 32) ^ crc5);
+
+	crc0 ^= crc1;
+
+	return crc0;
+}
+
+#define hash_default_key8                        hash_crc_key8
+#define hash_default_key16                       hash_crc_key16
+#define hash_default_key24                       hash_crc_key24
+#define hash_default_key32                       hash_crc_key32
+#define hash_default_key40                       hash_crc_key40
+#define hash_default_key48                       hash_crc_key48
+#define hash_default_key56                       hash_crc_key56
+#define hash_default_key64                       hash_crc_key64
+
+#else
+
+#define hash_default_key8                        hash_xor_key8
+#define hash_default_key16                       hash_xor_key16
+#define hash_default_key24                       hash_xor_key24
+#define hash_default_key32                       hash_xor_key32
+#define hash_default_key40                       hash_xor_key40
+#define hash_default_key48                       hash_xor_key48
+#define hash_default_key56                       hash_xor_key56
+#define hash_default_key64                       hash_xor_key64
+
+#endif
+
+#endif
diff --git a/examples/ip_pipeline/pipeline/pipeline_actions_common.h b/examples/ip_pipeline/pipeline/pipeline_actions_common.h
new file mode 100644
index 0000000..3389ced
--- /dev/null
+++ b/examples/ip_pipeline/pipeline/pipeline_actions_common.h
@@ -0,0 +1,119 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#ifndef __INCLUDE_PIPELINE_ACTIONS_COMMON_H__
+#define __INCLUDE_PIPELINE_ACTIONS_COMMON_H__
+
+#define PIPELINE_PORT_IN_AH(f_ah, f_pkt_work, f_pkt4_work)		    \
+static int															\
+f_ah(																\
+	struct rte_mbuf **pkts,											\
+	uint32_t n_pkts,												\
+	uint64_t *pkts_mask,											\
+	void *arg)														\
+{																	\
+	uint32_t i;														\
+																	\
+	for (i = 0; i < (n_pkts & (~0x3LLU)); i += 4)					\
+		f_pkt4_work(&pkts[i], arg);									\
+																	\
+	for ( ; i < n_pkts; i++)										\
+		f_pkt_work(pkts[i], arg);									\
+																	\
+	*pkts_mask = (~0LLU) >> (64 - n_pkts);							\
+																	\
+	return 0;														\
+}
+
+#define PIPELINE_TABLE_AH_HIT(f_ah, f_pkt_work, f_pkt4_work)		\
+static int															\
+f_ah(																\
+	struct rte_mbuf **pkts,											\
+	uint64_t *pkts_mask,											\
+	struct rte_pipeline_table_entry **entries,						\
+	void *arg)														\
+{																	\
+	uint64_t pkts_in_mask = *pkts_mask;								\
+																	\
+	if ((pkts_in_mask & (pkts_in_mask + 1)) == 0) {					\
+		uint64_t n_pkts = __builtin_popcountll(pkts_in_mask);		\
+		uint32_t i;													\
+																	\
+		for (i = 0; i < (n_pkts & (~0x3LLU)); i += 4)				\
+			f_pkt4_work(&pkts[i], &entries[i], arg);				\
+																	\
+		for ( ; i < n_pkts; i++)									\
+			f_pkt_work(pkts[i], entries[i], arg);					\
+	} else															\
+		for ( ; pkts_in_mask; ) {									\
+			uint32_t pos = __builtin_ctzll(pkts_in_mask);			\
+			uint64_t pkt_mask = 1LLU << pos;						\
+																	\
+			pkts_in_mask &= ~pkt_mask;								\
+			f_pkt_work(pkts[pos], entries[pos], arg);				\
+		}															\
+																	\
+	return 0;														\
+}
+
+#define PIPELINE_TABLE_AH_MISS(f_ah, f_pkt_work, f_pkt4_work)		\
+static int															\
+f_ah(																\
+	struct rte_mbuf **pkts,											\
+	uint64_t *pkts_mask,											\
+	struct rte_pipeline_table_entry *entry,							\
+	void *arg)														\
+{																	\
+	uint64_t pkts_in_mask = *pkts_mask;								\
+																	\
+	if ((pkts_in_mask & (pkts_in_mask + 1)) == 0) {					\
+		uint64_t n_pkts = __builtin_popcountll(pkts_in_mask);		\
+		uint32_t i;													\
+																	\
+		for (i = 0; i < (n_pkts & (~0x3LLU)); i += 4)				\
+			f_pkt4_work(&pkts[i], entry, arg);						\
+																	\
+		for ( ; i < n_pkts; i++)									\
+			f_pkt_work(pkts[i], entry, arg);						\
+	} else															\
+		for ( ; pkts_in_mask; ) {									\
+			uint32_t pos = __builtin_ctzll(pkts_in_mask);			\
+			uint64_t pkt_mask = 1LLU << pos;						\
+																	\
+			pkts_in_mask &= ~pkt_mask;								\
+			f_pkt_work(pkts[pos], entry, arg);						\
+		}															\
+																	\
+	return 0;														\
+}
+
+#endif
diff --git a/examples/ip_pipeline/pipeline/pipeline_passthrough.c b/examples/ip_pipeline/pipeline/pipeline_passthrough.c
index 948b2c1..fc2cae5 100644
--- a/examples/ip_pipeline/pipeline/pipeline_passthrough.c
+++ b/examples/ip_pipeline/pipeline/pipeline_passthrough.c
@@ -1,7 +1,7 @@
 /*-
  *   BSD LICENSE
  *
- *   Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
+ *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
  *   All rights reserved.
  *
  *   Redistribution and use in source and binary forms, with or without
@@ -31,183 +31,17 @@
  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
 
-#include <stdio.h>
-#include <stdlib.h>
-#include <stdint.h>
+#include "pipeline_passthrough.h"
+#include "pipeline_passthrough_be.h"
 
-#include <rte_malloc.h>
-#include <rte_log.h>
+static struct pipeline_fe_ops pipeline_passthrough_fe_ops = {
+	.f_init = NULL,
+	.f_free = NULL,
+	.cmds = NULL,
+};
 
-#include <rte_port_ring.h>
-#include <rte_table_stub.h>
-#include <rte_pipeline.h>
-
-#include "main.h"
-
-void
-app_main_loop_pipeline_passthrough(void) {
-	struct rte_pipeline_params pipeline_params = {
-		.name = "pipeline",
-		.socket_id = rte_socket_id(),
-	};
-
-	struct rte_pipeline *p;
-	uint32_t port_in_id[APP_MAX_PORTS];
-	uint32_t port_out_id[APP_MAX_PORTS];
-	uint32_t table_id[APP_MAX_PORTS];
-	uint32_t i;
-
-	uint32_t core_id = rte_lcore_id();
-	struct app_core_params *core_params = app_get_core_params(core_id);
-
-	if ((core_params == NULL) || (core_params->core_type != APP_CORE_PT))
-		rte_panic("Core %u misconfiguration\n", core_id);
-
-	RTE_LOG(INFO, USER1, "Core %u is doing pass-through\n", core_id);
-
-	/* Pipeline configuration */
-	p = rte_pipeline_create(&pipeline_params);
-	if (p == NULL)
-		rte_panic("%s: Unable to configure the pipeline\n", __func__);
-
-	/* Input port configuration */
-	for (i = 0; i < app.n_ports; i++) {
-		struct rte_port_ring_reader_params port_ring_params = {
-			.ring = app.rings[core_params->swq_in[i]],
-		};
-
-		struct rte_pipeline_port_in_params port_params = {
-			.ops = &rte_port_ring_reader_ops,
-			.arg_create = (void *) &port_ring_params,
-			.f_action = NULL,
-			.arg_ah = NULL,
-			.burst_size = app.bsz_swq_rd,
-		};
-
-		if (rte_pipeline_port_in_create(p, &port_params,
-			&port_in_id[i])) {
-			rte_panic("%s: Unable to configure input port for "
-				"ring %d\n", __func__, i);
-		}
-	}
-
-	/* Output port configuration */
-	for (i = 0; i < app.n_ports; i++) {
-		struct rte_port_ring_writer_params port_ring_params = {
-			.ring = app.rings[core_params->swq_out[i]],
-			.tx_burst_sz = app.bsz_swq_wr,
-		};
-
-		struct rte_pipeline_port_out_params port_params = {
-			.ops = &rte_port_ring_writer_ops,
-			.arg_create = (void *) &port_ring_params,
-			.f_action = NULL,
-			.f_action_bulk = NULL,
-			.arg_ah = NULL,
-		};
-
-		if (rte_pipeline_port_out_create(p, &port_params,
-			&port_out_id[i])) {
-			rte_panic("%s: Unable to configure output port for "
-				"ring %d\n", __func__, i);
-		}
-	}
-
-	/* Table configuration */
-	for (i = 0; i < app.n_ports; i++) {
-		struct rte_pipeline_table_params table_params = {
-			.ops = &rte_table_stub_ops,
-			.arg_create = NULL,
-			.f_action_hit = NULL,
-			.f_action_miss = NULL,
-			.arg_ah = NULL,
-			.action_data_size = 0,
-		};
-
-		if (rte_pipeline_table_create(p, &table_params, &table_id[i]))
-			rte_panic("%s: Unable to configure table %u\n",
-				__func__, i);
-	}
-
-	/* Interconnecting ports and tables */
-	for (i = 0; i < app.n_ports; i++) {
-		if (rte_pipeline_port_in_connect_to_table(p, port_in_id[i],
-			table_id[i])) {
-			rte_panic("%s: Unable to connect input port %u to "
-				"table %u\n", __func__, port_in_id[i],
-				table_id[i]);
-		}
-	}
-
-	/* Add entries to tables */
-	for (i = 0; i < app.n_ports; i++) {
-		struct rte_pipeline_table_entry default_entry = {
-			.action = RTE_PIPELINE_ACTION_PORT,
-			{.port_id = port_out_id[i]},
-		};
-
-		struct rte_pipeline_table_entry *default_entry_ptr;
-
-		if (rte_pipeline_table_default_entry_add(p, table_id[i],
-			&default_entry, &default_entry_ptr))
-			rte_panic("%s: Unable to add default entry to "
-				"table %u\n", __func__, table_id[i]);
-	}
-
-	/* Enable input ports */
-	for (i = 0; i < app.n_ports; i++)
-		if (rte_pipeline_port_in_enable(p, port_in_id[i]))
-			rte_panic("Unable to enable input port %u\n",
-				port_in_id[i]);
-
-	/* Check pipeline consistency */
-	if (rte_pipeline_check(p) < 0)
-		rte_panic("%s: Pipeline consistency check failed\n", __func__);
-
-	/* Run-time */
-	for (i = 0; ; i++) {
-		rte_pipeline_run(p);
-
-		if ((i & APP_FLUSH) == 0)
-			rte_pipeline_flush(p);
-	}
-}
-
-void
-app_main_loop_passthrough(void) {
-	struct app_mbuf_array *m;
-	uint32_t i;
-
-	uint32_t core_id = rte_lcore_id();
-	struct app_core_params *core_params = app_get_core_params(core_id);
-
-	if ((core_params == NULL) || (core_params->core_type != APP_CORE_PT))
-		rte_panic("Core %u misconfiguration\n", core_id);
-
-	RTE_LOG(INFO, USER1, "Core %u is doing pass-through (no pipeline)\n",
-		core_id);
-
-	m = rte_malloc_socket(NULL, sizeof(struct app_mbuf_array),
-		RTE_CACHE_LINE_SIZE, rte_socket_id());
-	if (m == NULL)
-		rte_panic("%s: cannot allocate buffer space\n", __func__);
-
-	for (i = 0; ; i = ((i + 1) & (app.n_ports - 1))) {
-		int ret;
-
-		ret = rte_ring_sc_dequeue_bulk(
-			app.rings[core_params->swq_in[i]],
-			(void **) m->array,
-			app.bsz_swq_rd);
-
-		if (ret == -ENOENT)
-			continue;
-
-		do {
-			ret = rte_ring_sp_enqueue_bulk(
-				app.rings[core_params->swq_out[i]],
-				(void **) m->array,
-				app.bsz_swq_wr);
-		} while (ret < 0);
-	}
-}
+struct pipeline_type pipeline_passthrough = {
+	.name = "PASS-THROUGH",
+	.be_ops = &pipeline_passthrough_be_ops,
+	.fe_ops = &pipeline_passthrough_fe_ops,
+};
diff --git a/examples/ip_pipeline/pipeline/pipeline_passthrough.h b/examples/ip_pipeline/pipeline/pipeline_passthrough.h
new file mode 100644
index 0000000..420a876
--- /dev/null
+++ b/examples/ip_pipeline/pipeline/pipeline_passthrough.h
@@ -0,0 +1,41 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __INCLUDE_PIPELINE_PASSTHROUGH_H__
+#define __INCLUDE_PIPELINE_PASSTHROUGH_H__
+
+#include "pipeline.h"
+
+extern struct pipeline_type pipeline_passthrough;
+
+#endif
diff --git a/examples/ip_pipeline/pipeline/pipeline_passthrough_be.c b/examples/ip_pipeline/pipeline/pipeline_passthrough_be.c
new file mode 100644
index 0000000..3edf7ac
--- /dev/null
+++ b/examples/ip_pipeline/pipeline/pipeline_passthrough_be.c
@@ -0,0 +1,741 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <string.h>
+
+#include <rte_common.h>
+#include <rte_malloc.h>
+#include <rte_byteorder.h>
+#include <rte_table_stub.h>
+#include <rte_table_hash.h>
+#include <rte_pipeline.h>
+
+#include "pipeline_passthrough_be.h"
+#include "pipeline_actions_common.h"
+#include "hash_func.h"
+
+enum flow_key_type {
+	FLOW_KEY_QINQ,
+	FLOW_KEY_IPV4_5TUPLE,
+	FLOW_KEY_IPV6_5TUPLE,
+};
+
+struct pipeline_passthrough {
+	struct pipeline p;
+
+	uint32_t key_type_valid;
+	enum flow_key_type key_type;
+	uint32_t key_offset_rd;
+	uint32_t key_offset_wr;
+	uint32_t hash_offset;
+
+	rte_table_hash_op_hash f_hash;
+	rte_pipeline_port_in_action_handler f_port_in_ah;
+} __rte_cache_aligned;
+
+static pipeline_msg_req_handler handlers[] = {
+	[PIPELINE_MSG_REQ_PING] = pipeline_msg_req_ping_handler,
+	[PIPELINE_MSG_REQ_STATS_PORT_IN] = pipeline_msg_req_stats_port_in_handler,
+	[PIPELINE_MSG_REQ_STATS_PORT_OUT] = pipeline_msg_req_stats_port_out_handler,
+	[PIPELINE_MSG_REQ_STATS_TABLE] = pipeline_msg_req_stats_table_handler,
+	[PIPELINE_MSG_REQ_PORT_IN_ENABLE] = pipeline_msg_req_port_in_enable_handler,
+	[PIPELINE_MSG_REQ_PORT_IN_DISABLE] = pipeline_msg_req_port_in_disable_handler,
+	[PIPELINE_MSG_REQ_CUSTOM] = pipeline_msg_req_invalid_handler,
+};
+
+static inline void
+pkt_work_key_qinq(
+	struct rte_mbuf *pkt,
+	void *arg)
+{
+	struct pipeline_passthrough *p_pt = arg;
+	uint64_t *key_rd = RTE_MBUF_METADATA_UINT64_PTR(pkt, p_pt->key_offset_rd);
+	uint64_t *key_wr = RTE_MBUF_METADATA_UINT64_PTR(pkt, p_pt->key_offset_wr);
+	uint32_t *hash = RTE_MBUF_METADATA_UINT32_PTR(pkt, p_pt->hash_offset);
+
+	/* Read */
+	uint64_t key_qinq = *key_rd & rte_bswap64(0x00000FFF00000FFFLLU);
+
+	/* Compute */
+	uint32_t hash_qinq = p_pt->f_hash(&key_qinq, 8, 0);
+
+	/* Write */
+	*key_wr = key_qinq;
+	*hash = hash_qinq;
+}
+
+static inline void
+pkt4_work_key_qinq(
+	struct rte_mbuf **pkt,
+	void *arg)
+{
+	struct pipeline_passthrough *p_pt = arg;
+
+	uint64_t *key_rd0 = RTE_MBUF_METADATA_UINT64_PTR(pkt[0], p_pt->key_offset_rd);
+	uint64_t *key_wr0 = RTE_MBUF_METADATA_UINT64_PTR(pkt[0], p_pt->key_offset_wr);
+	uint32_t *hash0 = RTE_MBUF_METADATA_UINT32_PTR(pkt[0], p_pt->hash_offset);
+
+	uint64_t *key_rd1 = RTE_MBUF_METADATA_UINT64_PTR(pkt[1], p_pt->key_offset_rd);
+	uint64_t *key_wr1 = RTE_MBUF_METADATA_UINT64_PTR(pkt[1], p_pt->key_offset_wr);
+	uint32_t *hash1 = RTE_MBUF_METADATA_UINT32_PTR(pkt[1], p_pt->hash_offset);
+
+	uint64_t *key_rd2 = RTE_MBUF_METADATA_UINT64_PTR(pkt[2], p_pt->key_offset_rd);
+	uint64_t *key_wr2 = RTE_MBUF_METADATA_UINT64_PTR(pkt[2], p_pt->key_offset_wr);
+	uint32_t *hash2 = RTE_MBUF_METADATA_UINT32_PTR(pkt[2], p_pt->hash_offset);
+
+	uint64_t *key_rd3 = RTE_MBUF_METADATA_UINT64_PTR(pkt[3], p_pt->key_offset_rd);
+	uint64_t *key_wr3 = RTE_MBUF_METADATA_UINT64_PTR(pkt[3], p_pt->key_offset_wr);
+	uint32_t *hash3 = RTE_MBUF_METADATA_UINT32_PTR(pkt[3], p_pt->hash_offset);
+
+	/* Read */
+	uint64_t key_qinq0 = *key_rd0 & rte_bswap64(0x00000FFF00000FFFLLU);
+	uint64_t key_qinq1 = *key_rd1 & rte_bswap64(0x00000FFF00000FFFLLU);
+	uint64_t key_qinq2 = *key_rd2 & rte_bswap64(0x00000FFF00000FFFLLU);
+	uint64_t key_qinq3 = *key_rd3 & rte_bswap64(0x00000FFF00000FFFLLU);
+
+	/* Compute */
+	uint32_t hash_qinq0 = p_pt->f_hash(&key_qinq0, 8, 0);
+	uint32_t hash_qinq1 = p_pt->f_hash(&key_qinq1, 8, 0);
+	uint32_t hash_qinq2 = p_pt->f_hash(&key_qinq2, 8, 0);
+	uint32_t hash_qinq3 = p_pt->f_hash(&key_qinq3, 8, 0);
+
+	/* Write */
+	*key_wr0 = key_qinq0;
+	*key_wr1 = key_qinq1;
+	*key_wr2 = key_qinq2;
+	*key_wr3 = key_qinq3;
+
+	*hash0 = hash_qinq0;
+	*hash1 = hash_qinq1;
+	*hash2 = hash_qinq2;
+	*hash3 = hash_qinq3;
+}
+
+PIPELINE_PORT_IN_AH(port_in_ah_key_qinq, pkt_work_key_qinq, pkt4_work_key_qinq);
+
+static inline void
+pkt_work_key_ipv4(
+	struct rte_mbuf *pkt,
+	void *arg)
+{
+	struct pipeline_passthrough *p_pt = arg;
+	uint64_t *key_rd = RTE_MBUF_METADATA_UINT64_PTR(pkt, p_pt->key_offset_rd);
+	uint64_t *key_wr = RTE_MBUF_METADATA_UINT64_PTR(pkt, p_pt->key_offset_wr);
+	uint32_t *hash = RTE_MBUF_METADATA_UINT32_PTR(pkt, p_pt->hash_offset);
+	uint64_t key_ipv4[2];
+	uint32_t hash_ipv4;
+
+	/* Read */
+	key_ipv4[0] = key_rd[0] & rte_bswap64(0x00FF0000FFFFFFFFLLU);
+	key_ipv4[1] = key_rd[1];
+
+	/* Compute */
+	hash_ipv4 = p_pt->f_hash(key_ipv4, 16, 0);
+
+	/* Write */
+	key_wr[0] = key_ipv4[0];
+	key_wr[1] = key_ipv4[1];
+	*hash = hash_ipv4;
+}
+
+static inline void
+pkt4_work_key_ipv4(
+	struct rte_mbuf **pkt,
+	void *arg)
+{
+	struct pipeline_passthrough *p_pt = arg;
+
+	uint64_t *key_rd0 = RTE_MBUF_METADATA_UINT64_PTR(pkt[0], p_pt->key_offset_rd);
+	uint64_t *key_wr0 = RTE_MBUF_METADATA_UINT64_PTR(pkt[0], p_pt->key_offset_wr);
+	uint32_t *hash0 = RTE_MBUF_METADATA_UINT32_PTR(pkt[0], p_pt->hash_offset);
+
+	uint64_t *key_rd1 = RTE_MBUF_METADATA_UINT64_PTR(pkt[1], p_pt->key_offset_rd);
+	uint64_t *key_wr1 = RTE_MBUF_METADATA_UINT64_PTR(pkt[1], p_pt->key_offset_wr);
+	uint32_t *hash1 = RTE_MBUF_METADATA_UINT32_PTR(pkt[1], p_pt->hash_offset);
+
+	uint64_t *key_rd2 = RTE_MBUF_METADATA_UINT64_PTR(pkt[2], p_pt->key_offset_rd);
+	uint64_t *key_wr2 = RTE_MBUF_METADATA_UINT64_PTR(pkt[2], p_pt->key_offset_wr);
+	uint32_t *hash2 = RTE_MBUF_METADATA_UINT32_PTR(pkt[2], p_pt->hash_offset);
+
+	uint64_t *key_rd3 = RTE_MBUF_METADATA_UINT64_PTR(pkt[3], p_pt->key_offset_rd);
+	uint64_t *key_wr3 = RTE_MBUF_METADATA_UINT64_PTR(pkt[3], p_pt->key_offset_wr);
+	uint32_t *hash3 = RTE_MBUF_METADATA_UINT32_PTR(pkt[3], p_pt->hash_offset);
+
+	uint64_t key_ipv4_0[2];
+	uint64_t key_ipv4_1[2];
+	uint64_t key_ipv4_2[2];
+	uint64_t key_ipv4_3[2];
+
+	uint32_t hash_ipv4_0;
+	uint32_t hash_ipv4_1;
+	uint32_t hash_ipv4_2;
+	uint32_t hash_ipv4_3;
+
+	/* Read */
+	key_ipv4_0[0] = key_rd0[0] & rte_bswap64(0x00FF0000FFFFFFFFLLU);
+	key_ipv4_1[0] = key_rd1[0] & rte_bswap64(0x00FF0000FFFFFFFFLLU);
+	key_ipv4_2[0] = key_rd2[0] & rte_bswap64(0x00FF0000FFFFFFFFLLU);
+	key_ipv4_3[0] = key_rd3[0] & rte_bswap64(0x00FF0000FFFFFFFFLLU);
+
+	key_ipv4_0[1] = key_rd0[1];
+	key_ipv4_1[1] = key_rd1[1];
+	key_ipv4_2[1] = key_rd2[1];
+	key_ipv4_3[1] = key_rd3[1];
+
+	/* Compute */
+	hash_ipv4_0 = p_pt->f_hash(key_ipv4_0, 16, 0);
+	hash_ipv4_1 = p_pt->f_hash(key_ipv4_1, 16, 0);
+	hash_ipv4_2 = p_pt->f_hash(key_ipv4_2, 16, 0);
+	hash_ipv4_3 = p_pt->f_hash(key_ipv4_3, 16, 0);
+
+	/* Write */
+	key_wr0[0] = key_ipv4_0[0];
+	key_wr1[0] = key_ipv4_1[0];
+	key_wr2[0] = key_ipv4_2[0];
+	key_wr3[0] = key_ipv4_3[0];
+
+	key_wr0[1] = key_ipv4_0[1];
+	key_wr1[1] = key_ipv4_1[1];
+	key_wr2[1] = key_ipv4_2[1];
+	key_wr3[1] = key_ipv4_3[1];
+
+	*hash0 = hash_ipv4_0;
+	*hash1 = hash_ipv4_1;
+	*hash2 = hash_ipv4_2;
+	*hash3 = hash_ipv4_3;
+}
+
+PIPELINE_PORT_IN_AH(port_in_ah_key_ipv4, pkt_work_key_ipv4, pkt4_work_key_ipv4);
+
+static inline void
+pkt_work_key_ipv6(
+	struct rte_mbuf *pkt,
+	void *arg)
+{
+	struct pipeline_passthrough *p_pt = arg;
+	uint64_t *key_rd = RTE_MBUF_METADATA_UINT64_PTR(pkt, p_pt->key_offset_rd);
+	uint64_t *key_wr = RTE_MBUF_METADATA_UINT64_PTR(pkt, p_pt->key_offset_wr);
+	uint32_t *hash = RTE_MBUF_METADATA_UINT32_PTR(pkt, p_pt->hash_offset);
+	uint64_t key_ipv6[8];
+	uint32_t hash_ipv6;
+
+	/* Read */
+	key_ipv6[0] = key_rd[0] & rte_bswap64(0x0000FF00FFFFFFFFLLU);
+	key_ipv6[1] = key_rd[1];
+	key_ipv6[2] = key_rd[2];
+	key_ipv6[3] = key_rd[3];
+	key_ipv6[4] = key_rd[4];
+	key_ipv6[5] = 0;
+	key_ipv6[6] = 0;
+	key_ipv6[7] = 0;	
+
+	/* Compute */
+	hash_ipv6 = p_pt->f_hash(key_ipv6, 64, 0);
+
+	/* Write */
+	key_wr[0] = key_ipv6[0];
+	key_wr[1] = key_ipv6[1];
+	key_wr[2] = key_ipv6[2];
+	key_wr[3] = key_ipv6[3];
+	key_wr[4] = key_ipv6[4];
+	key_wr[5] = 0;
+	key_wr[6] = 0;
+	key_wr[7] = 0;
+	*hash = hash_ipv6;
+}
+
+static inline void
+pkt4_work_key_ipv6(
+	struct rte_mbuf **pkt,
+	void *arg)
+{
+	struct pipeline_passthrough *p_pt = arg;
+
+	uint64_t *key_rd0 = RTE_MBUF_METADATA_UINT64_PTR(pkt[0], p_pt->key_offset_rd);
+	uint64_t *key_wr0 = RTE_MBUF_METADATA_UINT64_PTR(pkt[0], p_pt->key_offset_wr);
+	uint32_t *hash0 = RTE_MBUF_METADATA_UINT32_PTR(pkt[0], p_pt->hash_offset);
+
+	uint64_t *key_rd1 = RTE_MBUF_METADATA_UINT64_PTR(pkt[1], p_pt->key_offset_rd);
+	uint64_t *key_wr1 = RTE_MBUF_METADATA_UINT64_PTR(pkt[1], p_pt->key_offset_wr);
+	uint32_t *hash1 = RTE_MBUF_METADATA_UINT32_PTR(pkt[1], p_pt->hash_offset);
+
+	uint64_t *key_rd2 = RTE_MBUF_METADATA_UINT64_PTR(pkt[2], p_pt->key_offset_rd);
+	uint64_t *key_wr2 = RTE_MBUF_METADATA_UINT64_PTR(pkt[2], p_pt->key_offset_wr);
+	uint32_t *hash2 = RTE_MBUF_METADATA_UINT32_PTR(pkt[2], p_pt->hash_offset);
+
+	uint64_t *key_rd3 = RTE_MBUF_METADATA_UINT64_PTR(pkt[3], p_pt->key_offset_rd);
+	uint64_t *key_wr3 = RTE_MBUF_METADATA_UINT64_PTR(pkt[3], p_pt->key_offset_wr);
+	uint32_t *hash3 = RTE_MBUF_METADATA_UINT32_PTR(pkt[3], p_pt->hash_offset);
+
+	uint64_t key_ipv6_0[8];
+	uint64_t key_ipv6_1[8];
+	uint64_t key_ipv6_2[8];
+	uint64_t key_ipv6_3[8];
+
+	uint32_t hash_ipv6_0;
+	uint32_t hash_ipv6_1;
+	uint32_t hash_ipv6_2;
+	uint32_t hash_ipv6_3;
+
+	/* Read */
+	key_ipv6_0[0] = key_rd0[0] & rte_bswap64(0x0000FF00FFFFFFFFLLU);
+	key_ipv6_1[0] = key_rd1[0] & rte_bswap64(0x0000FF00FFFFFFFFLLU);
+	key_ipv6_2[0] = key_rd2[0] & rte_bswap64(0x0000FF00FFFFFFFFLLU);
+	key_ipv6_3[0] = key_rd3[0] & rte_bswap64(0x0000FF00FFFFFFFFLLU);
+
+	key_ipv6_0[1] = key_rd0[1];
+	key_ipv6_1[1] = key_rd1[1];
+	key_ipv6_2[1] = key_rd2[1];
+	key_ipv6_3[1] = key_rd3[1];
+
+	key_ipv6_0[2] = key_rd0[2];
+	key_ipv6_1[2] = key_rd1[2];
+	key_ipv6_2[2] = key_rd2[2];
+	key_ipv6_3[2] = key_rd3[2];
+
+	key_ipv6_0[3] = key_rd0[3];
+	key_ipv6_1[3] = key_rd1[3];
+	key_ipv6_2[3] = key_rd2[3];
+	key_ipv6_3[3] = key_rd3[3];
+
+	key_ipv6_0[4] = key_rd0[4];
+	key_ipv6_1[4] = key_rd1[4];
+	key_ipv6_2[4] = key_rd2[4];
+	key_ipv6_3[4] = key_rd3[4];
+
+	key_ipv6_0[5] = 0;
+	key_ipv6_1[5] = 0;
+	key_ipv6_2[5] = 0;
+	key_ipv6_3[5] = 0;
+
+	key_ipv6_0[6] = 0;
+	key_ipv6_1[6] = 0;
+	key_ipv6_2[6] = 0;
+	key_ipv6_3[6] = 0;
+
+	key_ipv6_0[7] = 0;
+	key_ipv6_1[7] = 0;
+	key_ipv6_2[7] = 0;
+	key_ipv6_3[7] = 0;
+
+	/* Compute */
+	hash_ipv6_0 = p_pt->f_hash(key_ipv6_0, 64, 0);
+	hash_ipv6_1 = p_pt->f_hash(key_ipv6_1, 64, 0);
+	hash_ipv6_2 = p_pt->f_hash(key_ipv6_2, 64, 0);
+	hash_ipv6_3 = p_pt->f_hash(key_ipv6_3, 64, 0);
+
+	/* Write */
+	key_wr0[0] = key_ipv6_0[0];
+	key_wr1[0] = key_ipv6_1[0];
+	key_wr2[0] = key_ipv6_2[0];
+	key_wr3[0] = key_ipv6_3[0];
+
+	key_wr0[1] = key_ipv6_0[1];
+	key_wr1[1] = key_ipv6_1[1];
+	key_wr2[1] = key_ipv6_2[1];
+	key_wr3[1] = key_ipv6_3[1];
+
+	key_wr0[2] = key_ipv6_0[2];
+	key_wr1[2] = key_ipv6_1[2];
+	key_wr2[2] = key_ipv6_2[2];
+	key_wr3[2] = key_ipv6_3[2];
+
+	key_wr0[3] = key_ipv6_0[3];
+	key_wr1[3] = key_ipv6_1[3];
+	key_wr2[3] = key_ipv6_2[3];
+	key_wr3[3] = key_ipv6_3[3];
+
+	key_wr0[4] = key_ipv6_0[4];
+	key_wr1[4] = key_ipv6_1[4];
+	key_wr2[4] = key_ipv6_2[4];
+	key_wr3[4] = key_ipv6_3[4];
+
+	key_wr0[5] = 0;
+	key_wr0[5] = 0;
+	key_wr0[5] = 0;
+	key_wr0[5] = 0;
+
+	key_wr0[6] = 0;
+	key_wr0[6] = 0;
+	key_wr0[6] = 0;
+	key_wr0[6] = 0;
+
+	key_wr0[7] = 0;
+	key_wr0[7] = 0;
+	key_wr0[7] = 0;
+	key_wr0[7] = 0;
+
+	*hash0 = hash_ipv6_0;
+	*hash1 = hash_ipv6_1;
+	*hash2 = hash_ipv6_2;
+	*hash3 = hash_ipv6_3;
+}
+
+PIPELINE_PORT_IN_AH(port_in_ah_key_ipv6, pkt_work_key_ipv6, pkt4_work_key_ipv6);
+
+static int
+pipeline_passthrough_parse_args(struct pipeline_passthrough *p,
+	struct pipeline_params *params)
+{
+	uint32_t key_type_present = 0;
+	uint32_t key_offset_rd_present = 0;
+	uint32_t key_offset_wr_present = 0;
+	uint32_t hash_offset_present = 0;
+	uint32_t i;
+
+	for (i = 0; i < params->n_args; i++) {
+		char *arg_name = params->args_name[i];
+		char *arg_value = params->args_value[i];
+
+		/* key_type */
+		if (strcmp(arg_name, "key_type") == 0) {
+			if (key_type_present)
+				return -1;
+			key_type_present = 1;
+
+			if ((strcmp(arg_value, "q-in-q") == 0) ||
+				(strcmp(arg_value, "qinq") == 0))
+				p->key_type = FLOW_KEY_QINQ;
+			else if (strcmp(arg_value, "ipv4_5tuple") == 0)
+				p->key_type = FLOW_KEY_IPV4_5TUPLE;
+			else if (strcmp(arg_value, "ipv6_5tuple") == 0)
+				p->key_type = FLOW_KEY_IPV6_5TUPLE;
+			else
+				return -1;
+
+			p->key_type_valid = 1;
+
+			continue;
+		}
+
+		/* key_offset_rd */
+		if (strcmp(arg_name, "key_offset_rd") == 0) {
+			if (key_offset_rd_present)
+				return -1;
+			key_offset_rd_present = 1;
+
+			p->key_offset_rd = atoi(arg_value);
+
+			continue;
+		}
+
+		/* key_offset_wr */
+		if (strcmp(arg_name, "key_offset_wr") == 0) {
+			if (key_offset_wr_present)
+				return -1;
+			key_offset_wr_present = 1;
+
+			p->key_offset_wr = atoi(arg_value);
+
+			continue;
+		}
+
+		/* hash_offset */
+		if (strcmp(arg_name, "hash_offset") == 0) {
+			if (hash_offset_present)
+				return -1;
+			hash_offset_present = 1;
+
+			p->hash_offset = atoi(arg_value);
+
+			continue;
+		}
+
+		/* any other */
+		return -1;
+	}
+
+	/* Check that mandatory arguments are present */
+	if ((key_offset_rd_present != key_type_present) ||
+		(key_offset_wr_present != key_type_present) ||
+		(hash_offset_present != key_type_present))
+		return -1;
+
+	return 0;
+}
+
+static void*
+pipeline_passthrough_init(struct pipeline_params *params,
+	__rte_unused void *arg)
+{
+	struct pipeline *p;
+	struct pipeline_passthrough *p_pt;
+	uint32_t size, i;
+
+	/* Check input arguments */
+	if ((params == NULL) ||
+		(params->n_ports_in == 0) ||
+		(params->n_ports_out == 0) ||
+		(params->n_ports_in < params->n_ports_out) ||
+		(params->n_ports_in % params->n_ports_out))
+		return NULL;
+	
+	/* Memory allocation */
+	size = RTE_CACHE_LINE_ROUNDUP(sizeof(struct pipeline_passthrough));
+	p = rte_zmalloc(NULL, size, RTE_CACHE_LINE_SIZE);
+	p_pt = (struct pipeline_passthrough *) p;
+	if (p == NULL)
+		return NULL;
+
+	strcpy(p->name, params->name);
+	p->log_level = params->log_level;
+
+	PLOG(p, HIGH, "Pass-through");
+
+	/* Parse arguments */
+	if (pipeline_passthrough_parse_args(p_pt, params))
+		return NULL;
+
+	if (p_pt->key_type_valid == 0) {
+		p_pt->f_hash = NULL;
+		p_pt->f_port_in_ah = NULL;
+	} else
+		switch (p_pt->key_type) {
+			case FLOW_KEY_QINQ:
+				p_pt->f_hash = hash_default_key8;
+				p_pt->f_port_in_ah = port_in_ah_key_qinq;
+				break;
+
+			case FLOW_KEY_IPV4_5TUPLE:
+				p_pt->f_hash = hash_default_key16;
+				p_pt->f_port_in_ah = port_in_ah_key_ipv4;
+				break;
+
+			case FLOW_KEY_IPV6_5TUPLE:
+				p_pt->f_hash = hash_default_key64;
+				p_pt->f_port_in_ah = port_in_ah_key_ipv6;
+				break;
+
+			default:
+				p_pt->f_hash = NULL;
+				p_pt->f_port_in_ah = NULL;
+		}
+
+	/* Pipeline */
+	{
+		struct rte_pipeline_params pipeline_params = {
+			.name = "PASS-THROUGH",
+			.socket_id = params->socket_id,
+			.offset_port_id = 0,
+		};
+
+		p->p = rte_pipeline_create(&pipeline_params);
+		if (p->p == NULL) {
+			rte_free(p);
+			return NULL;
+		}
+	}
+
+	/* Input ports */
+	p->n_ports_in = params->n_ports_in;
+	for (i = 0; i < p->n_ports_in; i++) {
+		struct rte_pipeline_port_in_params port_params = {
+			.ops = pipeline_port_in_params_get_ops(&params->port_in[i]),
+			.arg_create = pipeline_port_in_params_convert(&params->port_in[i]),
+			.f_action = p_pt->f_port_in_ah,
+			.arg_ah = p_pt,
+			.burst_size = params->port_in[i].burst_size,
+		};
+
+		int status = rte_pipeline_port_in_create(p->p,
+			&port_params,
+			&p->port_in_id[i]);
+
+		if (status) {
+			rte_pipeline_free(p->p);
+			rte_free(p);
+			return NULL;
+		}
+	}
+
+	/* Output ports */
+	p->n_ports_out = params->n_ports_out;
+	for (i = 0; i < p->n_ports_out; i++) {
+		struct rte_pipeline_port_out_params port_params = {
+			.ops = pipeline_port_out_params_get_ops(&params->port_out[i]),
+			.arg_create = pipeline_port_out_params_convert(&params->port_out[i]),
+			.f_action = NULL,
+			.f_action_bulk = NULL,
+			.arg_ah = NULL,
+		};
+
+		int status = rte_pipeline_port_out_create(p->p,
+			&port_params,
+			&p->port_out_id[i]);
+		
+		if (status) {
+			rte_pipeline_free(p->p);
+			rte_free(p);
+			return NULL;
+		}
+	}
+
+	/* Tables */
+	p->n_tables = p->n_ports_in;
+	for (i = 0; i < p->n_ports_in; i++) {
+		struct rte_pipeline_table_params table_params = {
+			.ops = &rte_table_stub_ops,
+			.arg_create = NULL,
+			.f_action_hit = NULL,
+			.f_action_miss = NULL,
+			.arg_ah = NULL,
+			.action_data_size = 0,
+		};
+
+		int status = rte_pipeline_table_create(p->p,
+			&table_params,
+			&p->table_id[i]);
+		
+		
+		if (status) {
+			rte_pipeline_free(p->p);
+			rte_free(p);
+			return NULL;
+		}
+	}
+
+	/* Connecting input ports to tables */
+	for (i = 0; i < p->n_ports_in; i++) {
+		int status = rte_pipeline_port_in_connect_to_table(p->p,
+			p->port_in_id[i],
+			p->table_id[i]);
+
+		if (status) {
+			rte_pipeline_free(p->p);
+			rte_free(p);
+			return NULL;
+		}
+	}
+
+	/* Add entries to tables */
+	for (i = 0; i < p->n_ports_in; i++) {
+		struct rte_pipeline_table_entry default_entry = {
+			.action = RTE_PIPELINE_ACTION_PORT,
+			{.port_id = p->port_out_id[i / (p->n_ports_in / p->n_ports_out)]},
+		};
+
+		
+		struct rte_pipeline_table_entry *default_entry_ptr;
+
+		int status = rte_pipeline_table_default_entry_add(p->p,
+			p->table_id[i],
+			&default_entry,
+			&default_entry_ptr);
+		
+		if (status) {
+			rte_pipeline_free(p->p);
+			rte_free(p);
+			return NULL;
+		}
+	}
+
+	/* Enable input ports */
+	for (i = 0; i < p->n_ports_in; i++) {
+		int status = rte_pipeline_port_in_enable(p->p,
+			p->port_in_id[i]);
+		
+		if (status) {
+			rte_pipeline_free(p->p);
+			rte_free(p);
+			return NULL;
+		}
+	}
+
+	/* Check pipeline consistency */
+	if (rte_pipeline_check(p->p) < 0) {
+		rte_pipeline_free(p->p);
+		rte_free(p);
+		return NULL;
+	}
+	
+	/* Message queues */
+	p->n_msgq = params->n_msgq;
+	for (i = 0; i < p->n_msgq; i++)
+		p->msgq_in[i] = params->msgq_in[i];
+	for (i = 0; i < p->n_msgq; i++)
+		p->msgq_out[i] = params->msgq_out[i];
+	
+	/* Message handlers */
+	memcpy(p->handlers, handlers, sizeof(p->handlers));
+
+	return p;
+}
+
+static int
+pipeline_passthrough_free(void *pipeline)
+{
+	struct pipeline *p = (struct pipeline *) pipeline;
+
+	/* Check input arguments */
+	if (p == NULL)
+		return -1;
+	
+	/* Free resources */
+	rte_pipeline_free(p->p);
+	rte_free(p);
+	return 0;
+}
+
+static int
+pipeline_passthrough_timer(void *pipeline)
+{
+	struct pipeline *p = (struct pipeline *) pipeline;
+
+	pipeline_msg_req_handle(p);
+	rte_pipeline_flush(p->p);
+
+	return 0;
+}
+
+static int
+pipeline_passthrough_track(void *pipeline, uint32_t port_in, uint32_t *port_out)
+{
+	struct pipeline *p = (struct pipeline *) pipeline;
+
+	/* Check input arguments */
+	if ((p == NULL) ||
+		(port_in >= p->n_ports_in) ||
+		(port_out == NULL))
+		return -1;
+	
+	*port_out = port_in / p->n_ports_in;
+	return 0;
+}
+
+struct pipeline_be_ops pipeline_passthrough_be_ops = {
+		.f_init = pipeline_passthrough_init,
+		.f_free = pipeline_passthrough_free,
+		.f_run = NULL,
+		.f_timer = pipeline_passthrough_timer,
+		.f_track = pipeline_passthrough_track,
+};
diff --git a/examples/ip_pipeline/pipeline/pipeline_passthrough_be.h b/examples/ip_pipeline/pipeline/pipeline_passthrough_be.h
new file mode 100644
index 0000000..9d5e3db
--- /dev/null
+++ b/examples/ip_pipeline/pipeline/pipeline_passthrough_be.h
@@ -0,0 +1,41 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __INCLUDE_PIPELINE_PASSTHROUGH_BE_H__
+#define __INCLUDE_PIPELINE_PASSTHROUGH_BE_H__
+
+#include "pipeline_common_be.h"
+
+extern struct pipeline_be_ops pipeline_passthrough_be_ops;
+
+#endif
-- 
1.7.9.5

  parent reply	other threads:[~2015-06-25 11:30 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-06-25 11:15 [dpdk-dev] [PATCH v2 00/11] ip_pipeline: ip_pipeline application enhancements Maciej Gajdzica
2015-06-25 11:15 ` [dpdk-dev] [PATCH v2 01/11] ip_pipeline: add parsing for config files with new syntax Maciej Gajdzica
2015-06-25 11:15 ` [dpdk-dev] [PATCH v2 02/11] ip_pipeline: added config checks Maciej Gajdzica
2015-06-25 11:15 ` [dpdk-dev] [PATCH v2 03/11] ip_pipeline: modified init to match new params struct Maciej Gajdzica
2015-06-25 11:15 ` [dpdk-dev] [PATCH v2 04/11] ip_pipeline: moved pipelines to separate folder Maciej Gajdzica
2015-06-25 11:15 ` [dpdk-dev] [PATCH v2 05/11] ip_pipeline: added master pipeline Maciej Gajdzica
2015-06-25 11:15 ` [dpdk-dev] [PATCH v2 06/11] ip_pipeline: added application thread Maciej Gajdzica
2015-06-25 11:15 ` [dpdk-dev] [PATCH v2 07/11] ip_pipeline: moved config files to separate folder Maciej Gajdzica
2015-06-25 11:15 ` Maciej Gajdzica [this message]
2015-06-25 11:15 ` [dpdk-dev] [PATCH v2 09/11] ip_pipeline: added new implementation of firewall pipeline Maciej Gajdzica
2015-06-25 11:15 ` [dpdk-dev] [PATCH v2 10/11] ip_pipeline: added new implementation of routing pipeline Maciej Gajdzica
2015-06-25 11:15 ` [dpdk-dev] [PATCH v2 11/11] ip_pipeline: added new implementation of flow classification pipeline Maciej Gajdzica
2015-06-25 12:19 ` [dpdk-dev] [PATCH v2 00/11] ip_pipeline: ip_pipeline application enhancements Dumitrescu, Cristian

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1435230914-8174-9-git-send-email-maciejx.t.gajdzica@intel.com \
    --to=maciejx.t.gajdzica@intel.com \
    --cc=dev@dpdk.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).