diff mbox

[3/3] linux-dpdk: Rework buffer management

Message ID 1412591001-6850-4-git-send-email-ciprian.barbu@linaro.org
State Superseded
Headers show

Commit Message

Ciprian Barbu Oct. 6, 2014, 10:23 a.m. UTC
Signed-off-by: Ciprian Barbu <ciprian.barbu@linaro.org>
---
 platform/linux-dpdk/Makefile.am                    |   2 +-
 platform/linux-dpdk/include/odp_buffer_internal.h  |   8 +-
 platform/linux-dpdk/include/odp_packet_internal.h  |   8 +-
 .../linux-dpdk/include/odp_packet_io_internal.h    |   5 -
 platform/linux-dpdk/odp_buffer.c                   |  26 +-
 platform/linux-dpdk/odp_buffer_pool.c              | 151 +++++++-
 platform/linux-dpdk/odp_packet.c                   | 140 ++++---
 platform/linux-dpdk/odp_packet_dpdk.c              |  13 +-
 platform/linux-dpdk/odp_queue.c                    |  20 +-
 platform/linux-dpdk/odp_schedule.c                 | 421 +++++++++++++++++++++
 10 files changed, 690 insertions(+), 104 deletions(-)
 create mode 100644 platform/linux-dpdk/odp_schedule.c
diff mbox

Patch

diff --git a/platform/linux-dpdk/Makefile.am b/platform/linux-dpdk/Makefile.am
index 5c9d503..3d68729 100644
--- a/platform/linux-dpdk/Makefile.am
+++ b/platform/linux-dpdk/Makefile.am
@@ -80,7 +80,7 @@  __LIB__libodp_la_SOURCES = \
 			   odp_queue.c \
 			   ../linux-generic/odp_ring.c \
 			   ../linux-generic/odp_rwlock.c \
-			   ../linux-generic/odp_schedule.c \
+			   odp_schedule.c \
 			   ../linux-generic/odp_shared_memory.c \
 			   ../linux-generic/odp_spinlock.c \
 			   ../linux-generic/odp_system_info.c \
diff --git a/platform/linux-dpdk/include/odp_buffer_internal.h b/platform/linux-dpdk/include/odp_buffer_internal.h
index f87ec80..5406606 100644
--- a/platform/linux-dpdk/include/odp_buffer_internal.h
+++ b/platform/linux-dpdk/include/odp_buffer_internal.h
@@ -59,8 +59,12 @@  typedef union odp_buffer_bits_t {
 struct odp_buffer_hdr_t;
 
 
-typedef struct rte_mbuf odp_buffer_hdr_t;
-
+typedef struct odp_buffer_hdr_t {
+	struct rte_mbuf mb;            /* Underlying DPDK rte_mbuf */
+	struct odp_buffer_hdr_t *next; /* Next buf in a list */
+	int type;                      /* ODP buffer type; not DPDK buf type */
+	uint32_t index;                /* Index in the rte_mempool */
+} odp_buffer_hdr_t;
 
 int odp_buffer_snprint(char *str, size_t n, odp_buffer_t buf);
 
diff --git a/platform/linux-dpdk/include/odp_packet_internal.h b/platform/linux-dpdk/include/odp_packet_internal.h
index ccaba5a..1cd44b1 100644
--- a/platform/linux-dpdk/include/odp_packet_internal.h
+++ b/platform/linux-dpdk/include/odp_packet_internal.h
@@ -113,15 +113,9 @@  typedef struct {
 	uint32_t l3_offset; /**< offset to L3 hdr, e.g. IPv4, IPv6 */
 	uint32_t l4_offset; /**< offset to L4 hdr (TCP, UDP, SCTP, also ICMP) */
 
-	uint32_t frame_len;
-
-	uint64_t user_ctx;        /* user context */
+	uint64_t user_ctx;  /**< user context */
 
 	odp_pktio_t input;
-
-	uint32_t pad;
-	uint8_t payload[];
-
 } odp_packet_hdr_t;
 
 /**
diff --git a/platform/linux-dpdk/include/odp_packet_io_internal.h b/platform/linux-dpdk/include/odp_packet_io_internal.h
index 08abea7..9263349 100644
--- a/platform/linux-dpdk/include/odp_packet_io_internal.h
+++ b/platform/linux-dpdk/include/odp_packet_io_internal.h
@@ -31,11 +31,6 @@  struct pktio_entry {
 	odp_queue_t inq_default;	/**< default input queue, if set */
 	odp_queue_t outq_default;	/**< default out queue */
 	odp_pktio_params_t params;	/**< pktio parameters */
-	pkt_sock_t pkt_sock;		/**< using socket API for IO */
-	pkt_sock_mmap_t pkt_sock_mmap;	/**< using socket mmap API for IO */
-#ifdef ODP_HAVE_NETMAP
-	pkt_netmap_t pkt_nm;		/**< using netmap API for IO */
-#endif
 	pkt_dpdk_t pkt_dpdk;		/**< using DPDK API for IO */
 };
 
diff --git a/platform/linux-dpdk/odp_buffer.c b/platform/linux-dpdk/odp_buffer.c
index e2f8942..ccd7a1a 100644
--- a/platform/linux-dpdk/odp_buffer.c
+++ b/platform/linux-dpdk/odp_buffer.c
@@ -16,7 +16,7 @@  void *odp_buffer_addr(odp_buffer_t buf)
 {
 	odp_buffer_hdr_t *hdr = odp_buf_to_hdr(buf);
 
-	return hdr->buf_addr;
+	return hdr->mb.buf_addr;
 }
 
 
@@ -24,7 +24,7 @@  size_t odp_buffer_size(odp_buffer_t buf)
 {
 	odp_buffer_hdr_t *hdr = odp_buf_to_hdr(buf);
 
-	return hdr->buf_len;
+	return hdr->mb.buf_len;
 }
 
 
@@ -38,11 +38,9 @@  int odp_buffer_type(odp_buffer_t buf)
 
 int odp_buffer_is_valid(odp_buffer_t buf)
 {
-	odp_buffer_bits_t handle;
-
-	handle.u32 = buf;
-
-	return (handle.index != ODP_BUFFER_INVALID_INDEX);
+	/* We could call rte_mbuf_sanity_check, but that panics
+	 * and aborts the program */
+	return (void *) buf != NULL;
 }
 
 
@@ -61,17 +59,19 @@  int odp_buffer_snprint(char *str, size_t n, odp_buffer_t buf)
 	len += snprintf(&str[len], n-len,
 			"Buffer\n");
 	len += snprintf(&str[len], n-len,
-			"  pool         %"PRIu64"\n", (int64_t) hdr->pool);
+			"  pool         %"PRIu64"\n", (int64_t) hdr->mb.pool);
+	len += snprintf(&str[len], n-len,
+			"  phy_addr     %"PRIu64"\n", hdr->mb.buf_physaddr);
 	len += snprintf(&str[len], n-len,
-			"  phy_addr     %"PRIu64"\n", hdr->buf_physaddr);
+			"  addr         %p\n",        hdr->mb.buf_addr);
 	len += snprintf(&str[len], n-len,
-			"  addr         %p\n",        hdr->buf_addr);
+			"  size         %u\n",        hdr->mb.buf_len);
 	len += snprintf(&str[len], n-len,
-			"  size         %u\n",        hdr->buf_len);
+			"  ref_count    %i\n",        hdr->mb.refcnt);
 	len += snprintf(&str[len], n-len,
-			"  ref_count    %i\n",        hdr->refcnt);
+			"  dpdk type    %i\n",        hdr->mb.type);
 	len += snprintf(&str[len], n-len,
-			"  type         %i\n",        hdr->type);
+			"  odp type     %i\n",        hdr->type);
 
 	return len;
 }
diff --git a/platform/linux-dpdk/odp_buffer_pool.c b/platform/linux-dpdk/odp_buffer_pool.c
index 7f28f9d..40abee1 100644
--- a/platform/linux-dpdk/odp_buffer_pool.c
+++ b/platform/linux-dpdk/odp_buffer_pool.c
@@ -9,6 +9,7 @@ 
 #include <odp_buffer_pool_internal.h>
 #include <odp_buffer_internal.h>
 #include <odp_packet_internal.h>
+#include <odp_timer_internal.h>
 #include <odp_shared_memory.h>
 #include <odp_align.h>
 #include <odp_internal.h>
@@ -44,6 +45,13 @@ 
 
 #define NULL_INDEX ((uint32_t)-1)
 
+union buffer_type_any_u {
+	odp_buffer_hdr_t  buf;
+	odp_packet_hdr_t  pkt;
+	odp_timeout_hdr_t tmo;
+};
+
+typedef union buffer_type_any_u odp_any_buffer_hdr_t;
 
 typedef union pool_entry_u {
 	struct pool_entry_s s;
@@ -59,7 +67,7 @@  typedef struct pool_table_t {
 } pool_table_t;
 
 
-/* The pool table */
+/* The pool table ptr - resides in shared memory */
 static pool_table_t *pool_tbl;
 
 /* Pool entry pointers (for inlining) */
@@ -101,31 +109,150 @@  int odp_buffer_pool_init_global(void)
 	return 0;
 }
 
+struct mbuf_ctor_arg {
+	uint16_t seg_buf_offset; /* To skip the ODP buf/pkt/tmo header */
+	uint16_t seg_buf_size;   /* total sz: offset + user sz + HDROOM */
+	int buf_type;
+};
+
+struct mbuf_pool_ctor_arg {
+	uint16_t seg_buf_size; /* size of mbuf: user specified sz + HDROOM */
+};
+
+static void
+odp_dpdk_mbuf_pool_ctor(struct rte_mempool *mp,
+			void *opaque_arg)
+{
+	struct mbuf_pool_ctor_arg      *mbp_ctor_arg;
+	struct rte_pktmbuf_pool_private *mbp_priv;
+
+	if (mp->private_data_size < sizeof(struct rte_pktmbuf_pool_private)) {
+		ODP_ERR("%s(%s) private_data_size %d < %d",
+			__func__, mp->name, (int) mp->private_data_size,
+			(int) sizeof(struct rte_pktmbuf_pool_private));
+		return;
+	}
+	mbp_ctor_arg = (struct mbuf_pool_ctor_arg *) opaque_arg;
+	mbp_priv = rte_mempool_get_priv(mp);
+	mbp_priv->mbuf_data_room_size = mbp_ctor_arg->seg_buf_size;
+}
+
+/* ODP DPDK mbuf constructor.
+ * This is a combination of rte_pktmbuf_init in rte_mbuf.c
+ * and testpmd_mbuf_ctor in testpmd.c
+ */
+static void
+odp_dpdk_mbuf_ctor(struct rte_mempool *mp,
+		   void *opaque_arg,
+		   void *raw_mbuf,
+		   unsigned i)
+{
+	struct mbuf_ctor_arg *mb_ctor_arg;
+	struct rte_mbuf *mb = raw_mbuf;
+	struct odp_buffer_hdr_t *buf_hdr;
+
+	/* The rte_mbuf is at the begninning in all cases */
+	mb_ctor_arg = (struct mbuf_ctor_arg *) opaque_arg;
+	mb = (struct rte_mbuf *) raw_mbuf;
+
+	RTE_MBUF_ASSERT(mp->elt_size >= sizeof(struct rte_mbuf));
+
+	memset(mb, 0, mp->elt_size);
+
+	/* Start of buffer is just after the ODP type specific header
+	 * which contains in the very beginning the rte_mbuf struct */
+	mb->buf_addr     = (char *) mb + mb_ctor_arg->seg_buf_offset;
+	mb->buf_physaddr = rte_mempool_virt2phy(mp, mb) +
+			mb_ctor_arg->seg_buf_offset;
+	mb->buf_len      = mb_ctor_arg->seg_buf_size;
+
+	/* keep some headroom between start of buffer and data */
+	if (mb_ctor_arg->buf_type == ODP_BUFFER_TYPE_PACKET ||
+	    mb_ctor_arg->buf_type == ODP_BUFFER_TYPE_ANY)
+		mb->pkt.data = (char *) mb->buf_addr + RTE_PKTMBUF_HEADROOM;
+	else
+		mb->pkt.data = mb->buf_addr;
+
+	/* init some constant fields */
+	mb->type         = RTE_MBUF_PKT;
+	mb->pool         = mp;
+	mb->pkt.nb_segs  = 1;
+	mb->pkt.in_port  = 0xff;
+	mb->ol_flags     = 0;
+	mb->pkt.vlan_macip.data = 0;
+	mb->pkt.hash.rss = 0;
+
+	/* Save index, might be useful for debugging purposes */
+	buf_hdr = (struct odp_buffer_hdr_t *) raw_mbuf;
+	buf_hdr->index = i;
+}
 
 odp_buffer_pool_t odp_buffer_pool_create(const char *name,
 					 void *base_addr, uint64_t size,
 					 size_t buf_size, size_t buf_align,
 					 int buf_type)
 {
-	struct rte_mempool *pktmbuf_pool = NULL;
+	struct rte_mempool *pool = NULL;
+	struct mbuf_pool_ctor_arg mbp_ctor_arg;
+	struct mbuf_ctor_arg mb_ctor_arg;
+	unsigned mb_size;
+	size_t hdr_size;
+
+	/* Not used for rte_mempool; the new ODP buffer management introduces
+	 * rte_mempool_create_from_region where base_addr makes sense */
+	(void)base_addr;
+
+	/* buf_align will be removed soon, no need to wory about it */
+	(void)buf_align;
+
 	ODP_DBG("odp_buffer_pool_create: %s, %lx, %u, %u, %u, %d\n", name,
 		(uint64_t) base_addr, (unsigned) size,
 		(unsigned) buf_size, (unsigned) buf_align,
 		buf_type);
 
-	pktmbuf_pool =
-		rte_mempool_create(name, NB_MBUF,
-				   MBUF_SIZE, MAX_PKT_BURST,
-				   sizeof(struct rte_pktmbuf_pool_private),
-				   rte_pktmbuf_pool_init, NULL,
-				   rte_pktmbuf_init, NULL,
-				   rte_socket_id(), 0);
-	if (pktmbuf_pool == NULL) {
+	switch (buf_type) {
+	case ODP_BUFFER_TYPE_RAW:
+		hdr_size = sizeof(odp_buffer_hdr_t);
+		mbp_ctor_arg.seg_buf_size = (uint16_t) buf_size;
+		break;
+	case ODP_BUFFER_TYPE_PACKET:
+		hdr_size = sizeof(odp_packet_hdr_t);
+		mbp_ctor_arg.seg_buf_size =
+			(uint16_t) (RTE_PKTMBUF_HEADROOM + buf_size);
+		break;
+	case ODP_BUFFER_TYPE_TIMEOUT:
+		hdr_size = sizeof(odp_timeout_hdr_t);
+		mbp_ctor_arg.seg_buf_size = (uint16_t) buf_size;
+		break;
+	case ODP_BUFFER_TYPE_ANY:
+		hdr_size = sizeof(odp_any_buffer_hdr_t);
+		mbp_ctor_arg.seg_buf_size =
+			(uint16_t) (RTE_PKTMBUF_HEADROOM + buf_size);
+		break;
+	default:
+		ODP_ERR("odp_buffer_pool_create: Bad type %i\n", buf_type);
+		return ODP_BUFFER_POOL_INVALID;
+		break;
+	}
+
+	mb_ctor_arg.seg_buf_offset =
+		(uint16_t) ODP_CACHE_LINE_SIZE_ROUNDUP(hdr_size);
+	mb_ctor_arg.seg_buf_size = mbp_ctor_arg.seg_buf_size;
+	mb_ctor_arg.buf_type = buf_type;
+	mb_size = mb_ctor_arg.seg_buf_offset + mb_ctor_arg.seg_buf_size;
+
+	pool = rte_mempool_create(name, NB_MBUF,
+				  mb_size, MAX_PKT_BURST,
+				  sizeof(struct rte_pktmbuf_pool_private),
+				  odp_dpdk_mbuf_pool_ctor, &mbp_ctor_arg,
+				  odp_dpdk_mbuf_ctor, &mb_ctor_arg,
+				  rte_socket_id(), 0);
+	if (pool == NULL) {
 		ODP_ERR("Cannot init DPDK mbuf pool\n");
-		return -1;
+		return ODP_BUFFER_POOL_INVALID;
 	}
 
-	return (odp_buffer_pool_t) pktmbuf_pool;
+	return (odp_buffer_pool_t) pool;
 }
 
 
diff --git a/platform/linux-dpdk/odp_packet.c b/platform/linux-dpdk/odp_packet.c
index 458b0db..1cd190d 100644
--- a/platform/linux-dpdk/odp_packet.c
+++ b/platform/linux-dpdk/odp_packet.c
@@ -23,13 +23,13 @@  static inline uint8_t parse_ipv6(odp_packet_hdr_t *pkt_hdr,
 void odp_packet_init(odp_packet_t pkt)
 {
 	odp_packet_hdr_t *const pkt_hdr = odp_packet_hdr(pkt);
-	const size_t start_offset = ODP_FIELD_SIZEOF(odp_packet_hdr_t, buf_hdr);
-	uint8_t *start;
-	size_t len;
+	struct rte_mbuf *mb;
+	void *start;
 
-	start = (uint8_t *)pkt_hdr + start_offset;
-	len = ODP_OFFSETOF(odp_packet_hdr_t, payload) - start_offset;
-	memset(start, 0, len);
+	mb = &pkt_hdr->buf_hdr.mb;
+
+	start = mb->buf_addr;
+	memset(start, 0, mb->buf_len);
 
 	pkt_hdr->l2_offset = (uint32_t) ODP_PACKET_OFFSET_INVALID;
 	pkt_hdr->l3_offset = (uint32_t) ODP_PACKET_OFFSET_INVALID;
@@ -46,18 +46,47 @@  odp_buffer_t odp_buffer_from_packet(odp_packet_t pkt)
 	return (odp_buffer_t)pkt;
 }
 
-void odp_packet_set_len(odp_packet_t pkt, size_t len)
+/* Advance the pkt data pointer and set len in one call */
+static int odp_packet_set_offset_len(odp_packet_t pkt, size_t frame_offset,
+				     size_t len)
 {
-	/* for rte_pktmbuf */
-	odp_buffer_hdr_t *buf_hdr = odp_buf_to_hdr(odp_buffer_from_packet(pkt));
-	buf_hdr->pkt.data_len = len;
+	struct rte_mbuf *mb = &(odp_packet_hdr(pkt)->buf_hdr.mb);
+	uint16_t offset;
+	uint16_t data_len;
+
+	/* The pkt buf may have been pulled back into the headroom
+	 * so we cannot rely on finding the data right after the
+	 * ODP header and HEADROOM */
+	offset = (uint16_t)((unsigned long)mb->pkt.data -
+			    (unsigned long)mb->buf_addr);
+	ODP_ASSERT(mb->buf_len >= offset, "Corrupted mbuf");
+	data_len = mb->buf_len - offset;
+
+	if (data_len < frame_offset) {
+		ODP_ERR("Frame offset too big");
+		return -1;
+	}
+	mb->pkt.data = (void *)((char *)mb->pkt.data + frame_offset);
+	data_len -= frame_offset;
+
+	if (data_len < len) {
+		ODP_ERR("Packet len too big");
+		return -1;
+	}
+	mb->pkt.pkt_len = len;
 
-	odp_packet_hdr(pkt)->frame_len = len;
+	return 0;
+}
+
+void odp_packet_set_len(odp_packet_t pkt, size_t len)
+{
+	(void)odp_packet_set_offset_len(pkt, 0, len);
 }
 
 size_t odp_packet_get_len(odp_packet_t pkt)
 {
-	return odp_packet_hdr(pkt)->frame_len;
+	struct rte_mbuf *mb = &(odp_packet_hdr(pkt)->buf_hdr.mb);
+	return mb->pkt.pkt_len;
 }
 
 uint8_t *odp_packet_buf_addr(odp_packet_t pkt)
@@ -67,7 +96,8 @@  uint8_t *odp_packet_buf_addr(odp_packet_t pkt)
 
 uint8_t *odp_packet_start(odp_packet_t pkt)
 {
-	return odp_packet_buf_addr(pkt) + odp_packet_hdr(pkt)->frame_offset;
+	struct rte_mbuf *mb = &(odp_packet_hdr(pkt)->buf_hdr.mb);
+	return mb->pkt.data;
 }
 
 
@@ -78,7 +108,7 @@  uint8_t *odp_packet_l2(odp_packet_t pkt)
 	if (odp_unlikely(offset == ODP_PACKET_OFFSET_INVALID))
 		return NULL;
 
-	return odp_packet_buf_addr(pkt) + offset;
+	return odp_packet_start(pkt) + offset;
 }
 
 size_t odp_packet_l2_offset(odp_packet_t pkt)
@@ -98,7 +128,7 @@  uint8_t *odp_packet_l3(odp_packet_t pkt)
 	if (odp_unlikely(offset == ODP_PACKET_OFFSET_INVALID))
 		return NULL;
 
-	return odp_packet_buf_addr(pkt) + offset;
+	return odp_packet_start(pkt) + offset;
 }
 
 size_t odp_packet_l3_offset(odp_packet_t pkt)
@@ -118,7 +148,7 @@  uint8_t *odp_packet_l4(odp_packet_t pkt)
 	if (odp_unlikely(offset == ODP_PACKET_OFFSET_INVALID))
 		return NULL;
 
-	return odp_packet_buf_addr(pkt) + offset;
+	return odp_packet_start(pkt) + offset;
 }
 
 size_t odp_packet_l4_offset(odp_packet_t pkt)
@@ -152,9 +182,13 @@  void odp_packet_parse(odp_packet_t pkt, size_t len, size_t frame_offset)
 	size_t offset = 0;
 	uint8_t ip_proto = 0;
 
+	/* The frame_offset is not relevant for frames from DPDK */
 	pkt_hdr->input_flags.eth = 1;
-	pkt_hdr->frame_offset = frame_offset;
-	pkt_hdr->frame_len = len;
+	(void) frame_offset;
+	pkt_hdr->frame_offset = 0;
+	if (odp_packet_set_offset_len(pkt, 0, len)) {
+		return;
+	}
 
 	if (odp_unlikely(len < ODPH_ETH_LEN_MIN)) {
 		pkt_hdr->error_flags.frame_len = 1;
@@ -165,7 +199,7 @@  void odp_packet_parse(odp_packet_t pkt, size_t len, size_t frame_offset)
 
 	/* Assume valid L2 header, no CRC/FCS check in SW */
 	pkt_hdr->input_flags.l2 = 1;
-	pkt_hdr->l2_offset = frame_offset;
+	pkt_hdr->l2_offset = 0;
 
 	eth = (odph_ethhdr_t *)odp_packet_start(pkt);
 	ethtype = odp_be_to_cpu_16(eth->type);
@@ -189,7 +223,7 @@  void odp_packet_parse(odp_packet_t pkt, size_t len, size_t frame_offset)
 	case ODPH_ETHTYPE_IPV4:
 		pkt_hdr->input_flags.ipv4 = 1;
 		pkt_hdr->input_flags.l3 = 1;
-		pkt_hdr->l3_offset = frame_offset + ODPH_ETHHDR_LEN + offset;
+		pkt_hdr->l3_offset = ODPH_ETHHDR_LEN + offset;
 		ipv4 = (odph_ipv4hdr_t *)odp_packet_l3(pkt);
 		ip_proto = parse_ipv4(pkt_hdr, ipv4, &offset);
 		break;
@@ -304,6 +338,7 @@  void odp_packet_print(odp_packet_t pkt)
 {
 	int max_len = 512;
 	char str[max_len];
+	uint8_t *p;
 	int len = 0;
 	int n = max_len-1;
 	odp_packet_hdr_t *hdr = odp_packet_hdr(pkt);
@@ -325,51 +360,60 @@  void odp_packet_print(odp_packet_t pkt)
 	len += snprintf(&str[len], n-len,
 			"  l4_offset    %u\n", hdr->l4_offset);
 	len += snprintf(&str[len], n-len,
-			"  frame_len    %u\n", hdr->frame_len);
+			"  frame_len    %u\n", hdr->buf_hdr.mb.pkt.pkt_len);
 	len += snprintf(&str[len], n-len,
 			"  input        %u\n", hdr->input);
 	str[len] = '\0';
 
 	printf("\n%s\n", str);
+	rte_pktmbuf_dump(&hdr->buf_hdr.mb, 32);
+
+	p = odp_packet_start(pkt);
+	printf("00000000: %02X %02X %02X %02X %02X %02X %02X %02X\n",
+	       p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7]);
+	printf("00000008: %02X %02X %02X %02X %02X %02X %02X %02X\n",
+	       p[8], p[9], p[10], p[11], p[12], p[13], p[14], p[15]);
+
 }
 
+/* For now we can only copy between packets of the same segment size
+ * We should probably refine this API, maybe introduce a clone API */
 int odp_packet_copy(odp_packet_t pkt_dst, odp_packet_t pkt_src)
 {
-	odp_packet_hdr_t *const pkt_hdr_dst = odp_packet_hdr(pkt_dst);
-	odp_packet_hdr_t *const pkt_hdr_src = odp_packet_hdr(pkt_src);
-	const size_t start_offset = ODP_FIELD_SIZEOF(odp_packet_hdr_t, buf_hdr);
-	uint8_t *start_src;
-	uint8_t *start_dst;
-	size_t len;
+	struct rte_mbuf *mb_dst, *mb_src;
+	uint8_t nb_segs, i;
 
-	if (pkt_dst == ODP_PACKET_INVALID || pkt_src == ODP_PACKET_INVALID)
-		return -1;
+	ODP_ASSERT(odp_buffer_type(pkt_dst) == ODP_BUFFER_TYPE_PACKET &&
+		   odp_buffer_type(pkt_src) == ODP_BUFFER_TYPE_PACKET,
+		   "dst_pkt or src_pkt not of type ODP_BUFFER_TYPE_PACKET");
 
-	/* if (pkt_hdr_dst->buf_hdr.size < */
-	/*	pkt_hdr_src->frame_len + pkt_hdr_src->frame_offset) */
-	if (pkt_hdr_dst->buf_hdr.buf_len <
-		pkt_hdr_src->frame_len + pkt_hdr_src->frame_offset)
+	if (pkt_dst == ODP_PACKET_INVALID || pkt_src == ODP_PACKET_INVALID)
 		return -1;
 
-	/* Copy packet header */
-	start_dst = (uint8_t *)pkt_hdr_dst + start_offset;
-	start_src = (uint8_t *)pkt_hdr_src + start_offset;
-	len = ODP_OFFSETOF(odp_packet_hdr_t, payload) - start_offset;
-	memcpy(start_dst, start_src, len);
+	mb_dst = &(odp_packet_hdr(pkt_dst)->buf_hdr.mb);
+	mb_src = &(odp_packet_hdr(pkt_src)->buf_hdr.mb);
 
-	/* Copy frame payload */
-	start_dst = (uint8_t *)odp_packet_start(pkt_dst);
-	start_src = (uint8_t *)odp_packet_start(pkt_src);
-	len = pkt_hdr_src->frame_len;
-	memcpy(start_dst, start_src, len);
+	if (mb_dst->pkt.nb_segs != mb_src->pkt.nb_segs) {
+		ODP_ERR("Different nb_segs in pkt_dst and pkt_src");
+		return -1;
+	}
 
-	/* Copy useful things from the buffer header */
-	/* pkt_hdr_dst->buf_hdr.cur_offset = pkt_hdr_src->buf_hdr.cur_offset; */
+	nb_segs = mb_src->pkt.nb_segs;
 
-	/* Create a copy of the scatter list */
-	/* odp_buffer_copy_scatter(odp_buffer_from_packet(pkt_dst), */
-	/*			odp_buffer_from_packet(pkt_src)); */
+	if (mb_dst->buf_len < mb_src->buf_len) {
+		ODP_ERR("dst_pkt smaller than src_pkt");
+		return -1;
+	}
 
+	for (i = 0; i < nb_segs; i++) {
+		if (mb_src == NULL || mb_dst == NULL) {
+			ODP_ERR("Corrupted packets");
+			return -1;
+		}
+		memcpy(mb_dst->buf_addr, mb_src->buf_addr, mb_src->buf_len);
+		mb_dst = mb_dst->pkt.next;
+		mb_src = mb_src->pkt.next;
+	}
 	return 0;
 }
 
diff --git a/platform/linux-dpdk/odp_packet_dpdk.c b/platform/linux-dpdk/odp_packet_dpdk.c
index d5c8e80..9fe4bbd 100644
--- a/platform/linux-dpdk/odp_packet_dpdk.c
+++ b/platform/linux-dpdk/odp_packet_dpdk.c
@@ -157,17 +157,18 @@  int close_pkt_dpdk(pkt_dpdk_t * const pkt_dpdk)
 }
 
 int recv_pkt_dpdk(pkt_dpdk_t * const pkt_dpdk, odp_packet_t pkt_table[],
-		unsigned len)
+		  unsigned len)
 {
-	struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
 	uint16_t nb_rx, i = 0;
 
-	memset(pkts_burst, 0 , sizeof(pkts_burst));
 	nb_rx = rte_eth_rx_burst((uint8_t)pkt_dpdk->portid,
 				 (uint16_t)pkt_dpdk->queueid,
-				 (struct rte_mbuf **)pkts_burst, (uint16_t)len);
-	for (i = 0; i < nb_rx; i++)
-		pkt_table[i] = (odp_packet_t)pkts_burst[i];
+				 (struct rte_mbuf **)pkt_table, (uint16_t)len);
+	for (i = 0; i < nb_rx; i++) {
+		odp_packet_hdr_t *pkt_hdr = odp_packet_hdr(pkt_table[i]);
+		struct rte_mbuf *mb = &pkt_hdr->buf_hdr.mb;
+		odp_packet_parse(pkt_table[i], mb->pkt.pkt_len, 0);
+	}
 	return nb_rx;
 }
 
diff --git a/platform/linux-dpdk/odp_queue.c b/platform/linux-dpdk/odp_queue.c
index 7cc8116..03f7c2c 100644
--- a/platform/linux-dpdk/odp_queue.c
+++ b/platform/linux-dpdk/odp_queue.c
@@ -241,11 +241,11 @@  int queue_enq(queue_entry_t *queue, odp_buffer_hdr_t *buf_hdr)
 		/* Empty queue */
 		queue->s.head = buf_hdr;
 		queue->s.tail = buf_hdr;
-		buf_hdr->pkt.next = NULL;
+		buf_hdr->next = NULL;
 	} else {
-		queue->s.tail->pkt.next = buf_hdr;
+		queue->s.tail->next = buf_hdr;
 		queue->s.tail = buf_hdr;
-		buf_hdr->pkt.next = NULL;
+		buf_hdr->next = NULL;
 	}
 
 	if (queue->s.status == QUEUE_STATUS_NOTSCHED) {
@@ -269,17 +269,17 @@  int queue_enq_multi(queue_entry_t *queue, odp_buffer_hdr_t *buf_hdr[], int num)
 	odp_buffer_hdr_t *tail;
 
 	for (i = 0; i < num - 1; i++)
-		buf_hdr[i]->pkt.next = buf_hdr[i+1];
+		buf_hdr[i]->next = buf_hdr[i+1];
 
 	tail = buf_hdr[num-1];
-	buf_hdr[num-1]->pkt.next = NULL;
+	buf_hdr[num-1]->next = NULL;
 
 	LOCK(&queue->s.lock);
 	/* Empty queue */
 	if (queue->s.head == NULL)
 		queue->s.head = buf_hdr[0];
 	else
-		queue->s.tail->pkt.next = buf_hdr[0];
+		queue->s.tail->next = buf_hdr[0];
 
 	queue->s.tail = tail;
 
@@ -340,8 +340,8 @@  odp_buffer_hdr_t *queue_deq(queue_entry_t *queue)
 			queue->s.status = QUEUE_STATUS_NOTSCHED;
 	} else {
 		buf_hdr       = queue->s.head;
-		queue->s.head = buf_hdr->pkt.next;
-		buf_hdr->pkt.next = NULL;
+		queue->s.head = buf_hdr->next;
+		buf_hdr->next = NULL;
 
 		if (queue->s.head == NULL) {
 			/* Queue is now empty */
@@ -372,8 +372,8 @@  int queue_deq_multi(queue_entry_t *queue, odp_buffer_hdr_t *buf_hdr[], int num)
 		for (; i < num && hdr; i++) {
 			buf_hdr[i]       = hdr;
 			/* odp_prefetch(hdr->addr); */
-			hdr              = hdr->pkt.next;
-			buf_hdr[i]->pkt.next = NULL;
+			hdr              = hdr->next;
+			buf_hdr[i]->next = NULL;
 		}
 
 		queue->s.head = hdr;
diff --git a/platform/linux-dpdk/odp_schedule.c b/platform/linux-dpdk/odp_schedule.c
new file mode 100644
index 0000000..cb559e4
--- /dev/null
+++ b/platform/linux-dpdk/odp_schedule.c
@@ -0,0 +1,421 @@ 
+/* Copyright (c) 2013, Linaro Limited
+ * All rights reserved.
+ *
+ * SPDX-License-Identifier:     BSD-3-Clause
+ */
+
+#include <odp_schedule.h>
+#include <odp_schedule_internal.h>
+#include <odp_align.h>
+#include <odp_queue.h>
+#include <odp_shared_memory.h>
+#include <odp_buffer.h>
+#include <odp_buffer_pool.h>
+#include <odp_internal.h>
+#include <odp_config.h>
+#include <odp_debug.h>
+#include <odp_thread.h>
+#include <odp_time.h>
+#include <odp_spinlock.h>
+#include <odp_hints.h>
+
+#include <odp_queue_internal.h>
+
+
+/* Limits to number of scheduled queues */
+#define SCHED_POOL_SIZE (256*1024)
+
+/* Scheduler sub queues */
+#define QUEUES_PER_PRIO  4
+
+/* TODO: random or queue based selection */
+#define SEL_PRI_QUEUE(x) ((QUEUES_PER_PRIO-1) & (queue_to_id(x)))
+
+/* Maximum number of dequeues */
+#define MAX_DEQ 4
+
+
+/* Mask of queues per priority */
+typedef uint8_t pri_mask_t;
+
+ODP_STATIC_ASSERT((8 * sizeof(pri_mask_t)) >=
+		  QUEUES_PER_PRIO, "pri_mask_t_is_too_small");
+
+
+typedef struct {
+	odp_queue_t       pri_queue[ODP_CONFIG_SCHED_PRIOS][QUEUES_PER_PRIO];
+	pri_mask_t        pri_mask[ODP_CONFIG_SCHED_PRIOS];
+	odp_spinlock_t    mask_lock;
+	odp_buffer_pool_t pool;
+} sched_t;
+
+typedef struct {
+	odp_queue_t queue;
+
+} queue_desc_t;
+
+typedef struct {
+	odp_queue_t  pri_queue;
+	odp_buffer_t desc_buf;
+
+	odp_buffer_t buf[MAX_DEQ];
+	int num;
+	int index;
+	odp_queue_t queue;
+	int pause;
+
+} sched_local_t;
+
+/* Global scheduler context */
+static sched_t *sched;
+
+/* Thread local scheduler context */
+static __thread sched_local_t sched_local;
+
+
+static inline odp_queue_t select_pri_queue(odp_queue_t queue, int prio)
+{
+	int id = SEL_PRI_QUEUE(queue);
+	return sched->pri_queue[prio][id];
+}
+
+
+int odp_schedule_init_global(void)
+{
+	odp_shm_t shm;
+	odp_buffer_pool_t pool;
+	int i, j;
+
+	ODP_DBG("Schedule init ... ");
+
+	shm = odp_shm_reserve("odp_scheduler",
+				sizeof(sched_t),
+				ODP_CACHE_LINE_SIZE, 0);
+
+	sched = odp_shm_addr(shm);
+
+	if (sched == NULL) {
+		ODP_ERR("Schedule init: Shm reserve failed.\n");
+		return -1;
+	}
+
+	pool = odp_buffer_pool_create("odp_sched_pool", NULL,
+				      SCHED_POOL_SIZE, sizeof(queue_desc_t),
+				      ODP_CACHE_LINE_SIZE,
+				      ODP_BUFFER_TYPE_RAW);
+
+	if (pool == ODP_BUFFER_POOL_INVALID) {
+		ODP_ERR("Schedule init: Pool create failed.\n");
+		return -1;
+	}
+
+	sched->pool = pool;
+	odp_spinlock_init(&sched->mask_lock);
+
+	for (i = 0; i < ODP_CONFIG_SCHED_PRIOS; i++) {
+		odp_queue_t queue;
+		char name[] = "odp_priXX_YY";
+
+		name[7] = '0' + i / 10;
+		name[8] = '0' + i - 10*(i / 10);
+
+		for (j = 0; j < QUEUES_PER_PRIO; j++) {
+			name[10] = '0' + j / 10;
+			name[11] = '0' + j - 10*(j / 10);
+
+			queue = odp_queue_create(name,
+						 ODP_QUEUE_TYPE_POLL, NULL);
+
+			if (queue == ODP_QUEUE_INVALID) {
+				ODP_ERR("Sched init: Queue create failed.\n");
+				return -1;
+			}
+
+			sched->pri_queue[i][j] = queue;
+			sched->pri_mask[i]     = 0;
+		}
+	}
+
+	ODP_DBG("done\n");
+
+	return 0;
+}
+
+
+int odp_schedule_init_local(void)
+{
+	int i;
+
+	sched_local.pri_queue = ODP_QUEUE_INVALID;
+	sched_local.desc_buf  = ODP_BUFFER_INVALID;
+
+	for (i = 0; i < MAX_DEQ; i++)
+		sched_local.buf[i] = ODP_BUFFER_INVALID;
+
+	sched_local.num   = 0;
+	sched_local.index = 0;
+	sched_local.queue = ODP_QUEUE_INVALID;
+	sched_local.pause = 0;
+
+	return 0;
+}
+
+
+void odp_schedule_mask_set(odp_queue_t queue, int prio)
+{
+	int id = SEL_PRI_QUEUE(queue);
+
+	odp_spinlock_lock(&sched->mask_lock);
+	sched->pri_mask[prio] |= 1 << id;
+	odp_spinlock_unlock(&sched->mask_lock);
+}
+
+
+odp_buffer_t odp_schedule_buffer_alloc(odp_queue_t queue)
+{
+	odp_buffer_t buf;
+
+	buf = odp_buffer_alloc(sched->pool);
+
+	if (buf != ODP_BUFFER_INVALID) {
+		queue_desc_t *desc;
+		desc        = odp_buffer_addr(buf);
+		desc->queue = queue;
+	}
+
+	return buf;
+}
+
+
+void odp_schedule_queue(odp_queue_t queue, int prio)
+{
+	odp_buffer_t desc_buf;
+	odp_queue_t  pri_queue;
+
+	pri_queue = select_pri_queue(queue, prio);
+	desc_buf  = queue_sched_buf(queue);
+
+	odp_queue_enq(pri_queue, desc_buf);
+}
+
+
+void odp_schedule_release_atomic(void)
+{
+	if (sched_local.pri_queue != ODP_QUEUE_INVALID &&
+	    sched_local.num       == 0) {
+		/* Release current atomic queue */
+		odp_queue_enq(sched_local.pri_queue, sched_local.desc_buf);
+		sched_local.pri_queue = ODP_QUEUE_INVALID;
+	}
+}
+
+
+static inline int copy_bufs(odp_buffer_t out_buf[], unsigned int max)
+{
+	int i = 0;
+
+	while (sched_local.num && max) {
+		out_buf[i] = sched_local.buf[sched_local.index];
+		sched_local.index++;
+		sched_local.num--;
+		max--;
+		i++;
+	}
+
+	return i;
+}
+
+
+/*
+ * Schedule queues
+ *
+ * TODO: SYNC_ORDERED not implemented yet
+ */
+static int schedule(odp_queue_t *out_queue, odp_buffer_t out_buf[],
+		    unsigned int max_num, unsigned int max_deq)
+{
+	int i, j;
+	int thr;
+	int ret;
+
+	if (sched_local.num) {
+		ret = copy_bufs(out_buf, max_num);
+
+		if (out_queue)
+			*out_queue = sched_local.queue;
+
+		return ret;
+	}
+
+	odp_schedule_release_atomic();
+
+	if (odp_unlikely(sched_local.pause))
+		return 0;
+
+	thr = odp_thread_id();
+
+	for (i = 0; i < ODP_CONFIG_SCHED_PRIOS; i++) {
+		int id;
+
+		if (sched->pri_mask[i] == 0)
+			continue;
+
+		id = thr & (QUEUES_PER_PRIO-1);
+
+		for (j = 0; j < QUEUES_PER_PRIO; j++, id++) {
+			odp_queue_t  pri_q;
+			odp_buffer_t desc_buf;
+
+			if (id >= QUEUES_PER_PRIO)
+				id = 0;
+
+			if (odp_unlikely((sched->pri_mask[i] & (1 << id)) == 0))
+				continue;
+
+			pri_q    = sched->pri_queue[i][id];
+			desc_buf = odp_queue_deq(pri_q);
+
+			if (desc_buf != ODP_BUFFER_INVALID) {
+				queue_desc_t *desc;
+				odp_queue_t queue;
+				int num;
+
+				desc  = odp_buffer_addr(desc_buf);
+				queue = desc->queue;
+
+				num = odp_queue_deq_multi(queue,
+							  sched_local.buf,
+							  max_deq);
+
+				if (num == 0) {
+					/* Remove empty queue from scheduling,
+					 * except packet input queues
+					 */
+					if (odp_queue_type(queue) ==
+					    ODP_QUEUE_TYPE_PKTIN)
+						odp_queue_enq(pri_q, desc_buf);
+
+					continue;
+				}
+
+				sched_local.num   = num;
+				sched_local.index = 0;
+				ret = copy_bufs(out_buf, max_num);
+
+				sched_local.queue = queue;
+
+				if (queue_sched_atomic(queue)) {
+					/* Hold queue during atomic access */
+					sched_local.pri_queue = pri_q;
+					sched_local.desc_buf  = desc_buf;
+				} else {
+					/* Continue scheduling the queue */
+					odp_queue_enq(pri_q, desc_buf);
+				}
+
+				/* Output the source queue handle */
+				if (out_queue)
+					*out_queue = queue;
+
+				return ret;
+			}
+		}
+	}
+
+	return 0;
+}
+
+
+static int schedule_loop(odp_queue_t *out_queue, uint64_t wait,
+			  odp_buffer_t out_buf[],
+			  unsigned int max_num, unsigned int max_deq)
+{
+	uint64_t start_cycle, cycle, diff;
+	int ret;
+
+	start_cycle = 0;
+
+	while (1) {
+		ret = schedule(out_queue, out_buf, max_num, max_deq);
+
+		if (ret)
+			break;
+
+		if (wait == ODP_SCHED_WAIT)
+			continue;
+
+		if (wait == ODP_SCHED_NO_WAIT)
+			break;
+
+		if (start_cycle == 0) {
+			start_cycle = odp_time_get_cycles();
+			continue;
+		}
+
+		cycle = odp_time_get_cycles();
+		diff  = odp_time_diff_cycles(start_cycle, cycle);
+
+		if (wait < diff)
+			break;
+	}
+
+	return ret;
+}
+
+
+odp_buffer_t odp_schedule(odp_queue_t *out_queue, uint64_t wait)
+{
+	odp_buffer_t buf;
+
+	buf = ODP_BUFFER_INVALID;
+
+	schedule_loop(out_queue, wait, &buf, 1, MAX_DEQ);
+
+	return buf;
+}
+
+
+odp_buffer_t odp_schedule_one(odp_queue_t *out_queue, uint64_t wait)
+{
+	odp_buffer_t buf;
+
+	buf = ODP_BUFFER_INVALID;
+
+	schedule_loop(out_queue, wait, &buf, 1, 1);
+
+	return buf;
+}
+
+
+int odp_schedule_multi(odp_queue_t *out_queue, uint64_t wait,
+		       odp_buffer_t out_buf[], unsigned int num)
+{
+	return schedule_loop(out_queue, wait, out_buf, num, MAX_DEQ);
+}
+
+
+void odp_schedule_pause(void)
+{
+	sched_local.pause = 1;
+}
+
+
+void odp_schedule_resume(void)
+{
+	sched_local.pause = 0;
+}
+
+
+uint64_t odp_schedule_wait_time(uint64_t ns)
+{
+	if (ns <= ODP_SCHED_NO_WAIT)
+		ns = ODP_SCHED_NO_WAIT + 1;
+
+	return odp_time_ns_to_cycles(ns);
+}
+
+
+int odp_schedule_num_prio(void)
+{
+	return ODP_CONFIG_SCHED_PRIOS;
+}