1 /*
2  * Copyright (c) 2008-2016 Solarflare Communications Inc.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions are met:
7  *
8  * 1. Redistributions of source code must retain the above copyright notice,
9  *    this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright notice,
11  *    this list of conditions and the following disclaimer in the documentation
12  *    and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
15  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
16  * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
17  * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
18  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
19  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
20  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
21  * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
22  * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
23  * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
24  * EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
25  *
26  * The views and conclusions contained in the software and documentation are
27  * those of the authors and should not be interpreted as representing official
28  * policies, either expressed or implied, of the FreeBSD Project.
29  */
30 
31 #include <sys/param.h>
32 #include <sys/int_limits.h>
33 #include <sys/byteorder.h>
34 #include <sys/random.h>
35 #include <sys/types.h>
36 #include <sys/kmem.h>
37 #include <netinet/in.h>
38 #include "sfxge.h"
39 #include "efx.h"
40 
41 /*
42  * The largest amount of the data which the hash may be calculated over
43  * is a 4-tuple of source/destination IPv6 addresses (2 x 16 bytes)
44  * and source/destination TCP port numbers (2 x 2 bytes), adding up to 40 bytes
45  */
46 #define	SFXGE_TOEPLITZ_IN_MAX \
47 	(2 * (sizeof (struct in6_addr) + sizeof (in_port_t)))
48 #define	SFXGE_TOEPLITZ_CACHE_SIZE (SFXGE_TOEPLITZ_IN_MAX * (UINT8_MAX + 1))
49 
50 static uint32_t
toeplitz_hash(const uint32_t * cache,const uint8_t * input,unsigned pos,unsigned datalen)51 toeplitz_hash(const uint32_t *cache, const uint8_t *input,
52     unsigned pos, unsigned datalen)
53 {
54 	uint32_t hash = 0;
55 	for (; datalen != 0; datalen--, pos++, input++) {
56 		hash ^= cache[pos * (UINT8_MAX + 1) + *input];
57 	}
58 
59 	return (hash);
60 }
61 
62 uint32_t
sfxge_toeplitz_hash(sfxge_t * sp,unsigned int addr_size,uint8_t * src_addr,uint16_t src_port,uint8_t * dst_addr,uint16_t dst_port)63 sfxge_toeplitz_hash(sfxge_t *sp, unsigned int addr_size,
64     uint8_t *src_addr, uint16_t src_port, uint8_t *dst_addr, uint16_t dst_port)
65 {
66 	uint32_t hash = 0;
67 	unsigned pos = 0;
68 
69 	hash ^= toeplitz_hash(sp->s_toeplitz_cache, src_addr, pos, addr_size);
70 	pos += addr_size;
71 	hash ^= toeplitz_hash(sp->s_toeplitz_cache, dst_addr, pos, addr_size);
72 	pos += addr_size;
73 	if (src_port != 0 || dst_port != 0) {
74 		hash ^= toeplitz_hash(sp->s_toeplitz_cache,
75 		    (const uint8_t *)&src_port, pos, sizeof (src_port));
76 		pos += sizeof (src_port);
77 		hash ^= toeplitz_hash(sp->s_toeplitz_cache,
78 		    (const uint8_t *)&dst_port, pos, sizeof (dst_port));
79 	}
80 	return (hash);
81 }
82 
83 /*
84  * The algorithm to calculate RSS Toeplitz hash is essentially as follows:
85  * - Regard a Toeplitz key and an input as bit strings, with the
86  * most significant bit of the first byte being the first bit
87  * - Let's have a 32-bit window sliding over the Toeplitz key bit by bit
88  * - Let the initial value of the hash be zero
89  * - Then for every bit in the input that is set to 1, XOR the value of the
90  *   window at a given bit position into the resulting hash
91  *
92  * First we note that since XOR is commutative and associative, the
93  * resulting hash is just a XOR of subhashes for every input bit:
94  *        H = H_0 XOR H_1 XOR ... XOR H_n               (1)
95  * Then we note that every H_i is only dependent on the value of i and
96  * the value of i'th bit of input, but not on any preceding or following
97  * input bits.
98  * Then we note that (1) holds also for any bit sequences,
99  * e.g. for bytes of input:
100  *       H = H_0_7 XOR H_8_15 XOR ... XOR H_(n-7)_n     (2)
101  * and every
102  *       H_i_j = H_i XOR H_(i+1) ... XOR H_j.           (3)
103  *
104  * It naturally follows than H_i_(i+7) only depends on the value of the byte
105  * and the position of the byte in the input.
106  * Therefore we may pre-calculate the value of each byte sub-hash H_i_(i+7)
107  * for each possible byte value and each possible byte input position, and
108  * then just assemble the hash of the packet byte-by-byte instead of
109  * bit-by-bit.
110  *
111  * The amount of memory required for such a cache is not prohibitive:
112  * - we have at most 36 bytes of input, each holding 256 possible values
113  * - and the hash is 32-bit wide
114  * - hence, we need only 36 * 256 * 4 = 36kBytes of cache.
115  *
116  * The performance gain, at least on synthetic benchmarks, is significant:
117  * cache lookup is about 15 times faster than direct hash calculation
118  */
119 const uint32_t *
toeplitz_cache_init(const uint8_t * key)120 toeplitz_cache_init(const uint8_t *key)
121 {
122 	uint32_t *cache = kmem_alloc(SFXGE_TOEPLITZ_CACHE_SIZE *
123 	    sizeof (uint32_t), KM_SLEEP);
124 	unsigned i;
125 
126 	for (i = 0; i < SFXGE_TOEPLITZ_IN_MAX; i++, key++) {
127 		uint32_t key_bits[NBBY] = { 0 };
128 		unsigned j;
129 		unsigned mask;
130 		unsigned byte;
131 
132 #if defined(BE_IN32)
133 		key_bits[0] = BE_IN32(key);
134 #else
135 		key_bits[0] = BE_32(*(uint32_t *)key);
136 #endif
137 		for (j = 1, mask = 1 << (NBBY - 1); j < NBBY; j++, mask >>= 1) {
138 			key_bits[j] = key_bits[j - 1] << 1;
139 			if ((key[sizeof (uint32_t)] & mask) != 0)
140 				key_bits[j] |= 1;
141 		}
142 
143 		for (byte = 0; byte <= UINT8_MAX; byte++) {
144 			uint32_t res = 0;
145 			for (j = 0, mask = 1 << (NBBY - 1);
146 			    j < NBBY;
147 			    j++, mask >>= 1) {
148 				if (byte & mask)
149 					res ^= key_bits[j];
150 			}
151 			cache[i * (UINT8_MAX + 1) + byte] = res;
152 		}
153 	}
154 	return (cache);
155 }
156 
157 
158 int
sfxge_toeplitz_hash_init(sfxge_t * sp)159 sfxge_toeplitz_hash_init(sfxge_t *sp)
160 {
161 	int rc;
162 	uint8_t toeplitz_key[SFXGE_TOEPLITZ_KEY_LEN];
163 
164 	(void) random_get_pseudo_bytes(toeplitz_key, sizeof (toeplitz_key));
165 
166 	if ((rc = efx_rx_scale_mode_set(sp->s_enp, EFX_RX_HASHALG_TOEPLITZ,
167 	    (1 << EFX_RX_HASH_IPV4) | (1 << EFX_RX_HASH_TCPIPV4) |
168 	    (1 << EFX_RX_HASH_IPV6) | (1 << EFX_RX_HASH_TCPIPV6), B_TRUE)) != 0)
169 		return (rc);
170 
171 	if ((rc = efx_rx_scale_key_set(sp->s_enp, toeplitz_key,
172 	    sizeof (toeplitz_key))) != 0)
173 		return (rc);
174 
175 	sp->s_toeplitz_cache = toeplitz_cache_init(toeplitz_key);
176 
177 	return (0);
178 }
179 
180 void
sfxge_toeplitz_hash_fini(sfxge_t * sp)181 sfxge_toeplitz_hash_fini(sfxge_t *sp)
182 {
183 	if (sp->s_toeplitz_cache != NULL) {
184 		kmem_free((void *)sp->s_toeplitz_cache,
185 		    SFXGE_TOEPLITZ_CACHE_SIZE);
186 		sp->s_toeplitz_cache = NULL;
187 	}
188 }
189