1 /* SPDX-License-Identifier: BSD-3-Clause
2 * Copyright(c) 1982, 1986, 1990, 1993
3 * The Regents of the University of California.
4 * Copyright(c) 2010-2014 Intel Corporation.
5 * Copyright(c) 2014 6WIND S.A.
20 #ifdef RTE_EXEC_ENV_WINDOWS
23 #include <sys/socket.h>
24 #include <sys/types.h>
25 #include <netinet/in.h>
26 #include <arpa/inet.h>
27 #include <netinet/ip.h>
28 #include <netinet/ip6.h>
31 #include <rte_byteorder.h>
44 uint8_t version_ihl; /**< version and header length */
46 #if RTE_BYTE_ORDER == RTE_LITTLE_ENDIAN
47 uint8_t ihl:4; /**< header length */
48 uint8_t version:4; /**< version */
49 #elif RTE_BYTE_ORDER == RTE_BIG_ENDIAN
50 uint8_t version:4; /**< version */
51 uint8_t ihl:4; /**< header length */
55 uint8_t type_of_service; /**< type of service */
56 rte_be16_t total_length; /**< length of packet */
57 rte_be16_t packet_id; /**< packet ID */
58 rte_be16_t fragment_offset; /**< fragmentation offset */
59 uint8_t time_to_live; /**< time to live */
60 uint8_t next_proto_id; /**< protocol ID */
61 rte_be16_t hdr_checksum; /**< header checksum */
62 rte_be32_t src_addr; /**< source address */
63 rte_be32_t dst_addr; /**< destination address */
66 /** Create IPv4 address */
67 #define RTE_IPV4(a, b, c, d) ((uint32_t)(((a) & 0xff) << 24) | \
68 (((b) & 0xff) << 16) | \
69 (((c) & 0xff) << 8) | \
72 /** Maximal IPv4 packet length (including a header) */
73 #define RTE_IPV4_MAX_PKT_LEN 65535
75 /** Internet header length mask for version_ihl field */
76 #define RTE_IPV4_HDR_IHL_MASK (0x0f)
78 * Internet header length field multiplier (IHL field specifies overall header
79 * length in number of 4-byte words)
81 #define RTE_IPV4_IHL_MULTIPLIER (4)
83 /* Type of Service fields */
84 #define RTE_IPV4_HDR_DSCP_MASK (0xfc)
85 #define RTE_IPV4_HDR_ECN_MASK (0x03)
86 #define RTE_IPV4_HDR_ECN_CE RTE_IPV4_HDR_ECN_MASK
88 /* Fragment Offset * Flags. */
89 #define RTE_IPV4_HDR_DF_SHIFT 14
90 #define RTE_IPV4_HDR_MF_SHIFT 13
91 #define RTE_IPV4_HDR_FO_SHIFT 3
93 #define RTE_IPV4_HDR_DF_FLAG (1 << RTE_IPV4_HDR_DF_SHIFT)
94 #define RTE_IPV4_HDR_MF_FLAG (1 << RTE_IPV4_HDR_MF_SHIFT)
96 #define RTE_IPV4_HDR_OFFSET_MASK ((1 << RTE_IPV4_HDR_MF_SHIFT) - 1)
98 #define RTE_IPV4_HDR_OFFSET_UNITS 8
103 #define RTE_IPV4_ANY ((uint32_t)0x00000000) /**< 0.0.0.0 */
104 #define RTE_IPV4_LOOPBACK ((uint32_t)0x7f000001) /**< 127.0.0.1 */
105 #define RTE_IPV4_BROADCAST ((uint32_t)0xe0000000) /**< 224.0.0.0 */
106 #define RTE_IPV4_ALLHOSTS_GROUP ((uint32_t)0xe0000001) /**< 224.0.0.1 */
107 #define RTE_IPV4_ALLRTRS_GROUP ((uint32_t)0xe0000002) /**< 224.0.0.2 */
108 #define RTE_IPV4_MAX_LOCAL_GROUP ((uint32_t)0xe00000ff) /**< 224.0.0.255 */
111 * IPv4 Multicast-related macros
113 #define RTE_IPV4_MIN_MCAST \
114 RTE_IPV4(224, 0, 0, 0) /**< Minimal IPv4-multicast address */
115 #define RTE_IPV4_MAX_MCAST \
116 RTE_IPV4(239, 255, 255, 255) /**< Maximum IPv4 multicast address */
118 #define RTE_IS_IPV4_MCAST(x) \
119 ((x) >= RTE_IPV4_MIN_MCAST && (x) <= RTE_IPV4_MAX_MCAST)
120 /**< check if IPv4 address is multicast */
122 /* IPv4 default fields values */
123 #define RTE_IPV4_MIN_IHL (0x5)
124 #define RTE_IPV4_VHL_DEF ((IPVERSION << 4) | RTE_IPV4_MIN_IHL)
127 * Get the length of an IPv4 header.
130 * Pointer to the IPv4 header.
132 * The length of the IPv4 header (with options if present) in bytes.
134 static inline uint8_t
135 rte_ipv4_hdr_len(const struct rte_ipv4_hdr *ipv4_hdr)
137 return (uint8_t)((ipv4_hdr->version_ihl & RTE_IPV4_HDR_IHL_MASK) *
138 RTE_IPV4_IHL_MULTIPLIER);
142 * @internal Calculate a sum of all words in the buffer.
143 * Helper routine for the rte_raw_cksum().
146 * Pointer to the buffer.
148 * Length of the buffer.
150 * Initial value of the sum.
152 * sum += Sum of all words in the buffer.
154 static inline uint32_t
155 __rte_raw_cksum(const void *buf, size_t len, uint32_t sum)
157 /* workaround gcc strict-aliasing warning */
158 uintptr_t ptr = (uintptr_t)buf;
159 typedef uint16_t __attribute__((__may_alias__)) u16_p;
160 const u16_p *u16_buf = (const u16_p *)ptr;
162 while (len >= (sizeof(*u16_buf) * 4)) {
167 len -= sizeof(*u16_buf) * 4;
170 while (len >= sizeof(*u16_buf)) {
172 len -= sizeof(*u16_buf);
176 /* if length is in odd bytes */
179 *(uint8_t *)&left = *(const uint8_t *)u16_buf;
187 * @internal Reduce a sum to the non-complemented checksum.
188 * Helper routine for the rte_raw_cksum().
193 * The non-complemented checksum.
195 static inline uint16_t
196 __rte_raw_cksum_reduce(uint32_t sum)
198 sum = ((sum & 0xffff0000) >> 16) + (sum & 0xffff);
199 sum = ((sum & 0xffff0000) >> 16) + (sum & 0xffff);
200 return (uint16_t)sum;
204 * Process the non-complemented checksum of a buffer.
207 * Pointer to the buffer.
209 * Length of the buffer.
211 * The non-complemented checksum.
213 static inline uint16_t
214 rte_raw_cksum(const void *buf, size_t len)
218 sum = __rte_raw_cksum(buf, len, 0);
219 return __rte_raw_cksum_reduce(sum);
223 * Compute the raw (non complemented) checksum of a packet.
226 * The pointer to the mbuf.
228 * The offset in bytes to start the checksum.
230 * The length in bytes of the data to checksum.
232 * A pointer to the checksum, filled on success.
234 * 0 on success, -1 on error (bad length or offset).
237 rte_raw_cksum_mbuf(const struct rte_mbuf *m, uint32_t off, uint32_t len,
240 const struct rte_mbuf *seg;
243 uint32_t seglen, done;
245 /* easy case: all data in the first segment */
246 if (off + len <= rte_pktmbuf_data_len(m)) {
247 *cksum = rte_raw_cksum(rte_pktmbuf_mtod_offset(m,
248 const char *, off), len);
252 if (unlikely(off + len > rte_pktmbuf_pkt_len(m)))
253 return -1; /* invalid params, return a dummy value */
255 /* else browse the segment to find offset */
257 for (seg = m; seg != NULL; seg = seg->next) {
258 seglen = rte_pktmbuf_data_len(seg);
263 RTE_ASSERT(seg != NULL);
267 buf = rte_pktmbuf_mtod_offset(seg, const char *, off);
269 /* all in one segment */
270 *cksum = rte_raw_cksum(buf, len);
274 /* hard case: process checksum of several segments */
278 tmp = __rte_raw_cksum(buf, seglen, 0);
280 tmp = rte_bswap16((uint16_t)tmp);
286 buf = rte_pktmbuf_mtod(seg, const char *);
287 seglen = rte_pktmbuf_data_len(seg);
288 if (seglen > len - done)
292 *cksum = __rte_raw_cksum_reduce(sum);
297 * Process the IPv4 checksum of an IPv4 header.
299 * The checksum field must be set to 0 by the caller.
302 * The pointer to the contiguous IPv4 header.
304 * The complemented checksum to set in the IP packet.
306 static inline uint16_t
307 rte_ipv4_cksum(const struct rte_ipv4_hdr *ipv4_hdr)
310 cksum = rte_raw_cksum(ipv4_hdr, rte_ipv4_hdr_len(ipv4_hdr));
311 return (uint16_t)~cksum;
315 * Process the pseudo-header checksum of an IPv4 header.
317 * The checksum field must be set to 0 by the caller.
319 * Depending on the ol_flags, the pseudo-header checksum expected by the
320 * drivers is not the same. For instance, when TSO is enabled, the IP
321 * payload length must not be included in the packet.
323 * When ol_flags is 0, it computes the standard pseudo-header checksum.
326 * The pointer to the contiguous IPv4 header.
328 * The ol_flags of the associated mbuf.
330 * The non-complemented checksum to set in the L4 header.
332 static inline uint16_t
333 rte_ipv4_phdr_cksum(const struct rte_ipv4_hdr *ipv4_hdr, uint64_t ol_flags)
335 struct ipv4_psd_header {
336 uint32_t src_addr; /* IP address of source host. */
337 uint32_t dst_addr; /* IP address of destination host. */
338 uint8_t zero; /* zero. */
339 uint8_t proto; /* L4 protocol type. */
340 uint16_t len; /* L4 length. */
345 psd_hdr.src_addr = ipv4_hdr->src_addr;
346 psd_hdr.dst_addr = ipv4_hdr->dst_addr;
348 psd_hdr.proto = ipv4_hdr->next_proto_id;
349 if (ol_flags & PKT_TX_TCP_SEG) {
352 l3_len = rte_be_to_cpu_16(ipv4_hdr->total_length);
353 psd_hdr.len = rte_cpu_to_be_16((uint16_t)(l3_len -
354 rte_ipv4_hdr_len(ipv4_hdr)));
356 return rte_raw_cksum(&psd_hdr, sizeof(psd_hdr));
360 * @internal Calculate the non-complemented IPv4 L4 checksum
362 static inline uint16_t
363 __rte_ipv4_udptcp_cksum(const struct rte_ipv4_hdr *ipv4_hdr, const void *l4_hdr)
366 uint32_t l3_len, l4_len;
369 ip_hdr_len = rte_ipv4_hdr_len(ipv4_hdr);
370 l3_len = rte_be_to_cpu_16(ipv4_hdr->total_length);
371 if (l3_len < ip_hdr_len)
374 l4_len = l3_len - ip_hdr_len;
376 cksum = rte_raw_cksum(l4_hdr, l4_len);
377 cksum += rte_ipv4_phdr_cksum(ipv4_hdr, 0);
379 cksum = ((cksum & 0xffff0000) >> 16) + (cksum & 0xffff);
381 return (uint16_t)cksum;
385 * Process the IPv4 UDP or TCP checksum.
387 * The layer 4 checksum must be set to 0 in the L4 header by the caller.
390 * The pointer to the contiguous IPv4 header.
392 * The pointer to the beginning of the L4 header.
394 * The complemented checksum to set in the L4 header.
396 static inline uint16_t
397 rte_ipv4_udptcp_cksum(const struct rte_ipv4_hdr *ipv4_hdr, const void *l4_hdr)
399 uint16_t cksum = __rte_ipv4_udptcp_cksum(ipv4_hdr, l4_hdr);
404 * Per RFC 768: If the computed checksum is zero for UDP,
405 * it is transmitted as all ones
406 * (the equivalent in one's complement arithmetic).
408 if (cksum == 0 && ipv4_hdr->next_proto_id == IPPROTO_UDP)
415 * Validate the IPv4 UDP or TCP checksum.
417 * In case of UDP, the caller must first check if udp_hdr->dgram_cksum is 0
418 * (i.e. no checksum).
421 * The pointer to the contiguous IPv4 header.
423 * The pointer to the beginning of the L4 header.
425 * Return 0 if the checksum is correct, else -1.
429 rte_ipv4_udptcp_cksum_verify(const struct rte_ipv4_hdr *ipv4_hdr,
432 uint16_t cksum = __rte_ipv4_udptcp_cksum(ipv4_hdr, l4_hdr);
443 struct rte_ipv6_hdr {
444 rte_be32_t vtc_flow; /**< IP version, traffic class & flow label. */
445 rte_be16_t payload_len; /**< IP payload size, including ext. headers */
446 uint8_t proto; /**< Protocol, next header. */
447 uint8_t hop_limits; /**< Hop limits. */
448 uint8_t src_addr[16]; /**< IP address of source host. */
449 uint8_t dst_addr[16]; /**< IP address of destination host(s). */
452 /* IPv6 vtc_flow: IPv / TC / flow_label */
453 #define RTE_IPV6_HDR_FL_SHIFT 0
454 #define RTE_IPV6_HDR_TC_SHIFT 20
455 #define RTE_IPV6_HDR_FL_MASK ((1u << RTE_IPV6_HDR_TC_SHIFT) - 1)
456 #define RTE_IPV6_HDR_TC_MASK (0xff << RTE_IPV6_HDR_TC_SHIFT)
457 #define RTE_IPV6_HDR_DSCP_MASK (0xfc << RTE_IPV6_HDR_TC_SHIFT)
458 #define RTE_IPV6_HDR_ECN_MASK (0x03 << RTE_IPV6_HDR_TC_SHIFT)
459 #define RTE_IPV6_HDR_ECN_CE RTE_IPV6_HDR_ECN_MASK
461 #define RTE_IPV6_MIN_MTU 1280 /**< Minimum MTU for IPv6, see RFC 8200. */
464 * Process the pseudo-header checksum of an IPv6 header.
466 * Depending on the ol_flags, the pseudo-header checksum expected by the
467 * drivers is not the same. For instance, when TSO is enabled, the IPv6
468 * payload length must not be included in the packet.
470 * When ol_flags is 0, it computes the standard pseudo-header checksum.
473 * The pointer to the contiguous IPv6 header.
475 * The ol_flags of the associated mbuf.
477 * The non-complemented checksum to set in the L4 header.
479 static inline uint16_t
480 rte_ipv6_phdr_cksum(const struct rte_ipv6_hdr *ipv6_hdr, uint64_t ol_flags)
484 rte_be32_t len; /* L4 length. */
485 rte_be32_t proto; /* L4 protocol - top 3 bytes must be zero */
488 psd_hdr.proto = (uint32_t)(ipv6_hdr->proto << 24);
489 if (ol_flags & PKT_TX_TCP_SEG) {
492 psd_hdr.len = ipv6_hdr->payload_len;
495 sum = __rte_raw_cksum(ipv6_hdr->src_addr,
496 sizeof(ipv6_hdr->src_addr) + sizeof(ipv6_hdr->dst_addr),
498 sum = __rte_raw_cksum(&psd_hdr, sizeof(psd_hdr), sum);
499 return __rte_raw_cksum_reduce(sum);
503 * @internal Calculate the non-complemented IPv6 L4 checksum
505 static inline uint16_t
506 __rte_ipv6_udptcp_cksum(const struct rte_ipv6_hdr *ipv6_hdr, const void *l4_hdr)
511 l4_len = rte_be_to_cpu_16(ipv6_hdr->payload_len);
513 cksum = rte_raw_cksum(l4_hdr, l4_len);
514 cksum += rte_ipv6_phdr_cksum(ipv6_hdr, 0);
516 cksum = ((cksum & 0xffff0000) >> 16) + (cksum & 0xffff);
518 return (uint16_t)cksum;
522 * Process the IPv6 UDP or TCP checksum.
524 * The IPv6 header must not be followed by extension headers. The layer 4
525 * checksum must be set to 0 in the L4 header by the caller.
528 * The pointer to the contiguous IPv6 header.
530 * The pointer to the beginning of the L4 header.
532 * The complemented checksum to set in the L4 header.
534 static inline uint16_t
535 rte_ipv6_udptcp_cksum(const struct rte_ipv6_hdr *ipv6_hdr, const void *l4_hdr)
537 uint16_t cksum = __rte_ipv6_udptcp_cksum(ipv6_hdr, l4_hdr);
542 * Per RFC 768: If the computed checksum is zero for UDP,
543 * it is transmitted as all ones
544 * (the equivalent in one's complement arithmetic).
546 if (cksum == 0 && ipv6_hdr->proto == IPPROTO_UDP)
553 * Validate the IPv6 UDP or TCP checksum.
555 * In case of UDP, the caller must first check if udp_hdr->dgram_cksum is 0:
556 * this is either invalid or means no checksum in some situations. See 8.1
557 * (Upper-Layer Checksums) in RFC 8200.
560 * The pointer to the contiguous IPv6 header.
562 * The pointer to the beginning of the L4 header.
564 * Return 0 if the checksum is correct, else -1.
568 rte_ipv6_udptcp_cksum_verify(const struct rte_ipv6_hdr *ipv6_hdr,
571 uint16_t cksum = __rte_ipv6_udptcp_cksum(ipv6_hdr, l4_hdr);
579 /** IPv6 fragment extension header. */
580 #define RTE_IPV6_EHDR_MF_SHIFT 0
581 #define RTE_IPV6_EHDR_MF_MASK 1
582 #define RTE_IPV6_EHDR_FO_SHIFT 3
583 #define RTE_IPV6_EHDR_FO_MASK (~((1 << RTE_IPV6_EHDR_FO_SHIFT) - 1))
584 #define RTE_IPV6_EHDR_FO_ALIGN (1 << RTE_IPV6_EHDR_FO_SHIFT)
586 #define RTE_IPV6_FRAG_USED_MASK (RTE_IPV6_EHDR_MF_MASK | RTE_IPV6_EHDR_FO_MASK)
588 #define RTE_IPV6_GET_MF(x) ((x) & RTE_IPV6_EHDR_MF_MASK)
589 #define RTE_IPV6_GET_FO(x) ((x) >> RTE_IPV6_EHDR_FO_SHIFT)
591 #define RTE_IPV6_SET_FRAG_DATA(fo, mf) \
592 (((fo) & RTE_IPV6_EHDR_FO_MASK) | ((mf) & RTE_IPV6_EHDR_MF_MASK))
594 struct rte_ipv6_fragment_ext {
595 uint8_t next_header; /**< Next header type */
596 uint8_t reserved; /**< Reserved */
597 rte_be16_t frag_data; /**< All fragmentation data */
598 rte_be32_t id; /**< Packet ID */
601 /* IPv6 fragment extension header size */
602 #define RTE_IPV6_FRAG_HDR_SIZE sizeof(struct rte_ipv6_fragment_ext)
605 * Parse next IPv6 header extension
607 * This function checks if proto number is an IPv6 extensions and parses its
608 * data if so, providing information on next header and extension length.
611 * Pointer to an extension raw data.
613 * Protocol number extracted from the "next header" field from
614 * the IPv6 header or the previous extension.
616 * Extension data length.
618 * next protocol number if proto is an IPv6 extension, -EINVAL otherwise
622 rte_ipv6_get_next_ext(const uint8_t *p, int proto, size_t *ext_len)
629 *ext_len = (*p + 2) * sizeof(uint32_t);
632 case IPPROTO_HOPOPTS:
633 case IPPROTO_ROUTING:
634 case IPPROTO_DSTOPTS:
636 *ext_len = (*p + 1) * sizeof(uint64_t);
639 case IPPROTO_FRAGMENT:
641 *ext_len = RTE_IPV6_FRAG_HDR_SIZE;
655 #endif /* _RTE_IP_H_ */