4 * Copyright(c) 2015 Vladimir Medvedkin <medvedkinv@gmail.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * * Neither the name of Intel Corporation nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
40 * toeplitz hash functions.
48 * Software implementation of the Toeplitz hash function used by RSS.
49 * Can be used either for packet distribution on single queue NIC
50 * or for simulating of RSS computation on specific NIC (for example
51 * after GRE header decapsulating)
55 #include <rte_byteorder.h>
56 #include <rte_config.h>
58 #include <rte_common.h>
60 #if defined(RTE_ARCH_X86) || defined(RTE_MACHINE_CPUFLAG_NEON)
65 /* Byte swap mask used for converting IPv6 address
66 * 4-byte chunks to CPU byte order
68 static const __m128i rte_thash_ipv6_bswap_mask = {
69 0x0405060700010203ULL, 0x0C0D0E0F08090A0BULL};
73 * length in dwords of input tuple to
74 * calculate hash of ipv4 header only
76 #define RTE_THASH_V4_L3_LEN ((sizeof(struct rte_ipv4_tuple) - \
77 sizeof(((struct rte_ipv4_tuple *)0)->sctp_tag)) / 4)
80 * length in dwords of input tuple to
81 * calculate hash of ipv4 header +
84 #define RTE_THASH_V4_L4_LEN ((sizeof(struct rte_ipv4_tuple)) / 4)
87 * length in dwords of input tuple to
88 * calculate hash of ipv6 header only
90 #define RTE_THASH_V6_L3_LEN ((sizeof(struct rte_ipv6_tuple) - \
91 sizeof(((struct rte_ipv6_tuple *)0)->sctp_tag)) / 4)
94 * length in dwords of input tuple to
95 * calculate hash of ipv6 header +
98 #define RTE_THASH_V6_L4_LEN ((sizeof(struct rte_ipv6_tuple)) / 4)
102 * addresses and ports/sctp_tag have to be CPU byte order
104 struct rte_ipv4_tuple {
119 * Addresses have to be filled by rte_thash_load_v6_addr()
120 * ports/sctp_tag have to be CPU byte order
122 struct rte_ipv6_tuple {
123 uint8_t src_addr[16];
124 uint8_t dst_addr[16];
135 union rte_thash_tuple {
136 struct rte_ipv4_tuple v4;
137 struct rte_ipv6_tuple v6;
139 } __attribute__((aligned(XMM_SIZE)));
145 * Prepare special converted key to use with rte_softrss_be()
147 * pointer to original RSS key
149 * pointer to target RSS key
154 rte_convert_rss_key(const uint32_t *orig, uint32_t *targ, int len)
158 for (i = 0; i < (len >> 2); i++)
159 targ[i] = rte_be_to_cpu_32(orig[i]);
163 * Prepare and load IPv6 addresses (src and dst)
166 * Pointer to ipv6 header of the original packet
168 * Pointer to rte_ipv6_tuple structure
171 rte_thash_load_v6_addrs(const struct ipv6_hdr *orig, union rte_thash_tuple *targ)
174 __m128i ipv6 = _mm_loadu_si128((const __m128i *)orig->src_addr);
175 *(__m128i *)targ->v6.src_addr =
176 _mm_shuffle_epi8(ipv6, rte_thash_ipv6_bswap_mask);
177 ipv6 = _mm_loadu_si128((const __m128i *)orig->dst_addr);
178 *(__m128i *)targ->v6.dst_addr =
179 _mm_shuffle_epi8(ipv6, rte_thash_ipv6_bswap_mask);
180 #elif defined(RTE_MACHINE_CPUFLAG_NEON)
181 uint8x16_t ipv6 = vld1q_u8((uint8_t const *)orig->src_addr);
182 vst1q_u8((uint8_t *)targ->v6.src_addr, vrev32q_u8(ipv6));
183 ipv6 = vld1q_u8((uint8_t const *)orig->dst_addr);
184 vst1q_u8((uint8_t *)targ->v6.dst_addr, vrev32q_u8(ipv6));
187 for (i = 0; i < 4; i++) {
188 *((uint32_t *)targ->v6.src_addr + i) =
189 rte_be_to_cpu_32(*((const uint32_t *)orig->src_addr + i));
190 *((uint32_t *)targ->v6.dst_addr + i) =
191 rte_be_to_cpu_32(*((const uint32_t *)orig->dst_addr + i));
197 * Generic implementation. Can be used with original rss_key
199 * Pointer to input tuple
201 * Length of input_tuple in 4-bytes chunks
203 * Pointer to RSS hash key.
205 * Calculated hash value.
207 static inline uint32_t
208 rte_softrss(uint32_t *input_tuple, uint32_t input_len,
209 const uint8_t *rss_key)
211 uint32_t i, j, map, ret = 0;
213 for (j = 0; j < input_len; j++) {
214 for (map = input_tuple[j]; map; map &= (map - 1)) {
216 ret ^= rte_cpu_to_be_32(((const uint32_t *)rss_key)[j]) << (31 - i) |
217 (uint32_t)((uint64_t)(rte_cpu_to_be_32(((const uint32_t *)rss_key)[j + 1])) >>
225 * Optimized implementation.
226 * If you want the calculated hash value matches NIC RSS value
227 * you have to use special converted key with rte_convert_rss_key() fn.
229 * Pointer to input tuple
231 * Length of input_tuple in 4-bytes chunks
233 * Pointer to RSS hash key.
235 * Calculated hash value.
237 static inline uint32_t
238 rte_softrss_be(uint32_t *input_tuple, uint32_t input_len,
239 const uint8_t *rss_key)
241 uint32_t i, j, map, ret = 0;
243 for (j = 0; j < input_len; j++) {
244 for (map = input_tuple[j]; map; map &= (map - 1)) {
246 ret ^= ((const uint32_t *)rss_key)[j] << (31 - i) |
247 (uint32_t)((uint64_t)(((const uint32_t *)rss_key)[j + 1]) >> (i + 1));
257 #endif /* _RTE_THASH_H */