ip_frag: add IPv6 fragmentation support
authorAnatoly Burakov <anatoly.burakov@intel.com>
Wed, 28 May 2014 17:32:43 +0000 (18:32 +0100)
committerThomas Monjalon <thomas.monjalon@6wind.com>
Mon, 16 Jun 2014 16:55:05 +0000 (18:55 +0200)
Mostly a copy-paste of IPv4.

Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
Acked-by: Thomas Monjalon <thomas.monjalon@6wind.com>
lib/librte_ip_frag/Makefile
lib/librte_ip_frag/rte_ip_frag.h
lib/librte_ip_frag/rte_ipv6_fragmentation.c [new file with mode: 0644]

index a723770..02817f0 100644 (file)
@@ -40,6 +40,7 @@ CFLAGS += $(WERROR_FLAGS) -I$(SRCDIR)
 #source files
 SRCS-$(CONFIG_RTE_LIBRTE_IP_FRAG) += rte_ipv4_fragmentation.c
 SRCS-$(CONFIG_RTE_LIBRTE_IP_FRAG) += rte_ipv4_reassembly.c
+SRCS-$(CONFIG_RTE_LIBRTE_IP_FRAG) += rte_ipv6_fragmentation.c
 SRCS-$(CONFIG_RTE_LIBRTE_IP_FRAG) += rte_ip_frag_common.c
 SRCS-$(CONFIG_RTE_LIBRTE_IP_FRAG) += ip_frag_internal.c
 
index 877fbc0..5c5e502 100644 (file)
@@ -173,6 +173,35 @@ rte_ip_frag_table_destroy( struct rte_ip_frag_tbl *tbl)
        rte_free(tbl);
 }
 
+/**
+ * This function implements the fragmentation of IPv6 packets.
+ *
+ * @param pkt_in
+ *   The input packet.
+ * @param pkts_out
+ *   Array storing the output fragments.
+ * @param nb_pkts_out
+ *   Number of fragments.
+ * @param mtu_size
+ *   Size in bytes of the Maximum Transfer Unit (MTU) for the outgoing IPv6
+ *   datagrams. This value includes the size of the IPv6 header.
+ * @param pool_direct
+ *   MBUF pool used for allocating direct buffers for the output fragments.
+ * @param pool_indirect
+ *   MBUF pool used for allocating indirect buffers for the output fragments.
+ * @return
+ *   Upon successful completion - number of output fragments placed
+ *   in the pkts_out array.
+ *   Otherwise - (-1) * errno.
+ */
+int32_t
+rte_ipv6_fragment_packet(struct rte_mbuf *pkt_in,
+               struct rte_mbuf **pkts_out,
+               uint16_t nb_pkts_out,
+               uint16_t mtu_size,
+               struct rte_mempool *pool_direct,
+               struct rte_mempool *pool_indirect);
+
 /**
  * IPv4 fragmentation.
  *
diff --git a/lib/librte_ip_frag/rte_ipv6_fragmentation.c b/lib/librte_ip_frag/rte_ipv6_fragmentation.c
new file mode 100644 (file)
index 0000000..e8f137c
--- /dev/null
@@ -0,0 +1,219 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2014 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 <stddef.h>
+#include <stdint.h>
+#include <errno.h>
+
+#include <rte_byteorder.h>
+#include <rte_memcpy.h>
+#include <rte_ip.h>
+
+#include "rte_ip_frag.h"
+#include "ip_frag_common.h"
+
+/**
+ * @file
+ * RTE IPv6 Fragmentation
+ *
+ * Implementation of IPv6 fragmentation.
+ *
+ */
+
+/* Fragment Extension Header */
+#define        IPV6_HDR_MF_SHIFT                       0
+#define        IPV6_HDR_FO_SHIFT                       3
+#define        IPV6_HDR_MF_MASK                        (1 << IPV6_HDR_MF_SHIFT)
+#define        IPV6_HDR_FO_MASK                        ((1 << IPV6_HDR_FO_SHIFT) - 1)
+
+static inline void
+__fill_ipv6hdr_frag(struct ipv6_hdr *dst,
+               const struct ipv6_hdr *src, uint16_t len, uint16_t fofs,
+               uint32_t mf)
+{
+       struct ipv6_extension_fragment *fh;
+
+       rte_memcpy(dst, src, sizeof(*dst));
+       dst->payload_len = rte_cpu_to_be_16(len);
+       dst->proto = IPPROTO_FRAGMENT;
+
+       fh = (struct ipv6_extension_fragment *) ++dst;
+       fh->next_header = src->proto;
+       fh->reserved1   = 0;
+       fh->frag_offset = rte_cpu_to_be_16(fofs);
+       fh->reserved2   = 0;
+       fh->more_frags  = rte_cpu_to_be_16(mf);
+       fh->id = 0;
+}
+
+static inline void
+__free_fragments(struct rte_mbuf *mb[], uint32_t num)
+{
+       uint32_t i;
+       for (i = 0; i < num; i++)
+               rte_pktmbuf_free(mb[i]);
+}
+
+/**
+ * IPv6 fragmentation.
+ *
+ * This function implements the fragmentation of IPv6 packets.
+ *
+ * @param pkt_in
+ *   The input packet.
+ * @param pkts_out
+ *   Array storing the output fragments.
+ * @param mtu_size
+ *   Size in bytes of the Maximum Transfer Unit (MTU) for the outgoing IPv6
+ *   datagrams. This value includes the size of the IPv6 header.
+ * @param pool_direct
+ *   MBUF pool used for allocating direct buffers for the output fragments.
+ * @param pool_indirect
+ *   MBUF pool used for allocating indirect buffers for the output fragments.
+ * @return
+ *   Upon successful completion - number of output fragments placed
+ *   in the pkts_out array.
+ *   Otherwise - (-1) * <errno>.
+ */
+int32_t
+rte_ipv6_fragment_packet(struct rte_mbuf *pkt_in,
+       struct rte_mbuf **pkts_out,
+       uint16_t nb_pkts_out,
+       uint16_t mtu_size,
+       struct rte_mempool *pool_direct,
+       struct rte_mempool *pool_indirect)
+{
+       struct rte_mbuf *in_seg = NULL;
+       struct ipv6_hdr *in_hdr;
+       uint32_t out_pkt_pos, in_seg_data_pos;
+       uint32_t more_in_segs;
+       uint16_t fragment_offset, frag_size;
+
+       frag_size = (uint16_t)(mtu_size - sizeof(struct ipv6_hdr));
+
+       /* Fragment size should be a multiple of 8. */
+       RTE_IP_FRAG_ASSERT((frag_size & IPV6_HDR_FO_MASK) == 0);
+
+       /* Check that pkts_out is big enough to hold all fragments */
+       if (unlikely (frag_size * nb_pkts_out <
+           (uint16_t)(pkt_in->pkt.pkt_len - sizeof (struct ipv6_hdr))))
+               return (-EINVAL);
+
+       in_hdr = (struct ipv6_hdr *) pkt_in->pkt.data;
+
+       in_seg = pkt_in;
+       in_seg_data_pos = sizeof(struct ipv6_hdr);
+       out_pkt_pos = 0;
+       fragment_offset = 0;
+
+       more_in_segs = 1;
+       while (likely(more_in_segs)) {
+               struct rte_mbuf *out_pkt = NULL, *out_seg_prev = NULL;
+               uint32_t more_out_segs;
+               struct ipv6_hdr *out_hdr;
+
+               /* Allocate direct buffer */
+               out_pkt = rte_pktmbuf_alloc(pool_direct);
+               if (unlikely(out_pkt == NULL)) {
+                       __free_fragments(pkts_out, out_pkt_pos);
+                       return (-ENOMEM);
+               }
+
+               /* Reserve space for the IP header that will be built later */
+               out_pkt->pkt.data_len = sizeof(struct ipv6_hdr) + sizeof(struct ipv6_extension_fragment);
+               out_pkt->pkt.pkt_len  = sizeof(struct ipv6_hdr) + sizeof(struct ipv6_extension_fragment);
+
+               out_seg_prev = out_pkt;
+               more_out_segs = 1;
+               while (likely(more_out_segs && more_in_segs)) {
+                       struct rte_mbuf *out_seg = NULL;
+                       uint32_t len;
+
+                       /* Allocate indirect buffer */
+                       out_seg = rte_pktmbuf_alloc(pool_indirect);
+                       if (unlikely(out_seg == NULL)) {
+                               rte_pktmbuf_free(out_pkt);
+                               __free_fragments(pkts_out, out_pkt_pos);
+                               return (-ENOMEM);
+                       }
+                       out_seg_prev->pkt.next = out_seg;
+                       out_seg_prev = out_seg;
+
+                       /* Prepare indirect buffer */
+                       rte_pktmbuf_attach(out_seg, in_seg);
+                       len = mtu_size - out_pkt->pkt.pkt_len;
+                       if (len > (in_seg->pkt.data_len - in_seg_data_pos)) {
+                               len = in_seg->pkt.data_len - in_seg_data_pos;
+                       }
+                       out_seg->pkt.data = (char *) in_seg->pkt.data + (uint16_t) in_seg_data_pos;
+                       out_seg->pkt.data_len = (uint16_t)len;
+                       out_pkt->pkt.pkt_len = (uint16_t)(len +
+                           out_pkt->pkt.pkt_len);
+                       out_pkt->pkt.nb_segs += 1;
+                       in_seg_data_pos += len;
+
+                       /* Current output packet (i.e. fragment) done ? */
+                       if (unlikely(out_pkt->pkt.pkt_len >= mtu_size)) {
+                               more_out_segs = 0;
+                       }
+
+                       /* Current input segment done ? */
+                       if (unlikely(in_seg_data_pos == in_seg->pkt.data_len)) {
+                               in_seg = in_seg->pkt.next;
+                               in_seg_data_pos = 0;
+
+                               if (unlikely(in_seg == NULL)) {
+                                       more_in_segs = 0;
+                               }
+                       }
+               }
+
+               /* Build the IP header */
+
+               out_hdr = (struct ipv6_hdr *) out_pkt->pkt.data;
+
+               __fill_ipv6hdr_frag(out_hdr, in_hdr,
+                   (uint16_t) out_pkt->pkt.pkt_len - sizeof(struct ipv6_hdr),
+                   fragment_offset, more_in_segs);
+
+               fragment_offset = (uint16_t)(fragment_offset +
+                   out_pkt->pkt.pkt_len - sizeof(struct ipv6_hdr)
+                       - sizeof(struct ipv6_extension_fragment));
+
+               /* Write the fragment to the output list */
+               pkts_out[out_pkt_pos] = out_pkt;
+               out_pkt_pos ++;
+       }
+
+       return (out_pkt_pos);
+}