2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2006, Myricom Inc.
5 * Copyright (c) 2008, Intel Corporation.
6 * Copyright (c) 2016-2021 Mellanox Technologies.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 #ifndef TCP_LRO_ENTRIES
39 /* Define default number of LRO entries per RX queue */
40 #define TCP_LRO_ENTRIES 8
44 * Flags for ACK entry for compression
45 * the bottom 8 bits has the th_flags.
46 * LRO itself adds only the TSTMP flags
47 * to indicate if either of the types
48 * of timestamps are filled and the
49 * HAS_TSTMP option to indicate if the
50 * TCP timestamp option is valid.
52 * The other 5 flag bits are for processing
56 #define TSTMP_LRO 0x0100
57 #define TSTMP_HDWR 0x0200
58 #define HAS_TSTMP 0x0400
60 * Default number of interrupts on the same cpu in a row
61 * that will cause us to declare a "affinity cpu".
63 #define TCP_LRO_CPU_DECLARATION_THRESH 50
70 uint8_t lro_type; /* internal */
71 #define LRO_TYPE_NONE 0
72 #define LRO_TYPE_IPV4_TCP 1
73 #define LRO_TYPE_IPV6_TCP 2
74 #define LRO_TYPE_IPV4_UDP 3
75 #define LRO_TYPE_IPV6_UDP 4
77 #define LRO_FLAG_DECRYPTED 1
78 uint16_t vlan_id; /* VLAN identifier */
79 uint16_t s_port; /* source TCP/UDP port */
80 uint16_t d_port; /* destination TCP/UDP port */
81 uint32_t vxlan_vni; /* VXLAN virtual network identifier */
89 } s_addr; /* source IPv4/IPv6 address */
97 } d_addr; /* destination IPv4/IPv6 address */
99 } __aligned(sizeof(u_long));
101 #define LRO_RAW_ADDRESS_MAX \
102 (sizeof(union lro_address) / sizeof(u_long))
104 /* Optimize address comparison by comparing one unsigned long at a time: */
107 lro_address_compare(const union lro_address *pa, const union lro_address *pb)
109 if (pa->lro_type == LRO_TYPE_NONE && pb->lro_type == LRO_TYPE_NONE) {
111 } else for (unsigned i = 0; i < LRO_RAW_ADDRESS_MAX; i++) {
112 if (pa->raw[i] != pb->raw[i])
119 union lro_address data;
130 uint16_t total_hdr_len;
133 /* This structure is zeroed frequently, try to keep it small. */
135 LIST_ENTRY(lro_entry) next;
136 LIST_ENTRY(lro_entry) hash_next;
139 struct mbuf *m_last_mbuf;
140 struct lro_parser outer;
141 struct lro_parser inner;
142 uint32_t next_seq; /* tcp_seq */
143 uint32_t ack_seq; /* tcp_seq */
147 uint16_t uncompressed;
150 uint8_t timestamp : 1;
151 uint8_t needs_merge : 1;
152 struct bintime alloc_time; /* time when entry was allocated */
155 LIST_HEAD(lro_head, lro_entry);
157 struct lro_mbuf_sort {
162 /* NB: This is part of driver structs. */
165 struct lro_mbuf_sort *lro_mbuf_data;
166 struct bintime lro_last_queue_time; /* last time data was queued */
168 uint64_t lro_flushed;
169 uint64_t lro_bad_csum;
171 unsigned lro_mbuf_count;
172 unsigned lro_mbuf_max;
173 unsigned short lro_ackcnt_lim; /* max # of aggregated ACKs */
174 unsigned short lro_cpu; /* Guess at the cpu we have affinity too */
175 unsigned lro_length_lim; /* max len of aggregated data */
177 uint32_t lro_last_cpu;
178 uint32_t lro_cnt_of_same_cpu;
179 struct lro_head *lro_hash;
180 struct lro_head lro_active;
181 struct lro_head lro_free;
182 uint8_t lro_cpu_is_set; /* Flag to say its ok to set the CPU on the inp */
186 uint64_t timestamp; /* hardware or sofware timestamp, valid if TSTMP_LRO or TSTMP_HDRW set */
187 uint32_t seq; /* th_seq value */
188 uint32_t ack; /* th_ack value */
189 uint32_t ts_value; /* If ts option value, valid if HAS_TSTMP is set */
190 uint32_t ts_echo; /* If ts option echo, valid if HAS_TSTMP is set */
191 uint16_t win; /* TCP window */
192 uint16_t flags; /* Flags to say if TS is present and type of timestamp and th_flags */
193 uint8_t codepoint; /* IP level codepoint including ECN bits */
194 uint8_t ack_val_set; /* Classification of ack used by the stack */
195 uint8_t pad[2]; /* To 32 byte boundary */
198 /* We use two M_PROTO on the mbuf */
199 #define M_ACKCMP M_PROTO4 /* Indicates LRO is sending in a Ack-compression mbuf */
200 #define M_LRO_EHDRSTRP M_PROTO6 /* Indicates that LRO has stripped the etherenet header */
202 #define TCP_LRO_LENGTH_MAX (65535 - 255) /* safe value with room for outer headers */
203 #define TCP_LRO_ACKCNT_MAX 65535 /* unlimited */
205 int tcp_lro_init(struct lro_ctrl *);
206 int tcp_lro_init_args(struct lro_ctrl *, struct ifnet *, unsigned, unsigned);
207 void tcp_lro_free(struct lro_ctrl *);
208 void tcp_lro_flush_inactive(struct lro_ctrl *, const struct timeval *);
209 void tcp_lro_flush(struct lro_ctrl *, struct lro_entry *);
210 void tcp_lro_flush_all(struct lro_ctrl *);
211 int tcp_lro_rx(struct lro_ctrl *, struct mbuf *, uint32_t);
212 void tcp_lro_queue_mbuf(struct lro_ctrl *, struct mbuf *);
213 void tcp_lro_reg_mbufq(void);
214 void tcp_lro_dereg_mbufq(void);
216 #define TCP_LRO_NO_ENTRIES -2
217 #define TCP_LRO_CANNOT -1
218 #define TCP_LRO_NOT_SUPPORTED 1
220 #endif /* _TCP_LRO_H_ */