ceph/src/seastar/dpdk/drivers/net/bonding/rte_eth_bond_alb.c

   1 /*-
   2  *   BSD LICENSE
   3  *
   4  *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
   5  *   All rights reserved.
   6  *
   7  *   Redistribution and use in source and binary forms, with or without
   8  *   modification, are permitted provided that the following conditions
   9  *   are met:
  10  *
  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
  16  *       distribution.
  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.
  20  *
  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.
  32  */
  33
  34 #include "rte_eth_bond_private.h"
  35 #include "rte_eth_bond_alb.h"
  36
  37 static inline uint8_t
  38 simple_hash(uint8_t *hash_start, int hash_size)
  39 {
  40         int i;
  41         uint8_t hash;
  42
  43         hash = 0;
  44         for (i = 0; i < hash_size; ++i)
  45                 hash ^= hash_start[i];
  46
  47         return hash;
  48 }
  49
  50 static uint8_t
  51 calculate_slave(struct bond_dev_private *internals)
  52 {
  53         uint8_t idx;
  54
  55         idx = (internals->mode6.last_slave + 1) % internals->active_slave_count;
  56         internals->mode6.last_slave = idx;
  57         return internals->active_slaves[idx];
  58 }
  59
  60 int
  61 bond_mode_alb_enable(struct rte_eth_dev *bond_dev)
  62 {
  63         struct bond_dev_private *internals = bond_dev->data->dev_private;
  64         struct client_data *hash_table = internals->mode6.client_table;
  65
  66         uint16_t data_size;
  67         char mem_name[RTE_ETH_NAME_MAX_LEN];
  68         int socket_id = bond_dev->data->numa_node;
  69
  70         /* Fill hash table with initial values */
  71         memset(hash_table, 0, sizeof(struct client_data) * ALB_HASH_TABLE_SIZE);
  72         rte_spinlock_init(&internals->mode6.lock);
  73         internals->mode6.last_slave = ALB_NULL_INDEX;
  74         internals->mode6.ntt = 0;
  75
  76         /* Initialize memory pool for ARP packets to send */
  77         if (internals->mode6.mempool == NULL) {
  78                 /*
  79                  * 256 is size of ETH header, ARP header and nested VLAN headers.
  80                  * The value is chosen to be cache aligned.
  81                  */
  82                 data_size = 256 + RTE_PKTMBUF_HEADROOM;
  83                 snprintf(mem_name, sizeof(mem_name), "%s_MODE6", bond_dev->data->name);
  84                 internals->mode6.mempool = rte_pktmbuf_pool_create(mem_name,
  85                         512 * RTE_MAX_ETHPORTS,
  86                         RTE_MEMPOOL_CACHE_MAX_SIZE >= 32 ?
  87                                 32 : RTE_MEMPOOL_CACHE_MAX_SIZE,
  88                         0, data_size, socket_id);
  89
  90                 if (internals->mode6.mempool == NULL) {
  91                         RTE_LOG(ERR, PMD, "%s: Failed to initialize ALB mempool.\n",
  92                                         bond_dev->data->name);
  93                         goto mempool_alloc_error;
  94                 }
  95         }
  96
  97         return 0;
  98
  99 mempool_alloc_error:
 100         return -ENOMEM;
 101 }
 102
 103 void bond_mode_alb_arp_recv(struct ether_hdr *eth_h, uint16_t offset,
 104                 struct bond_dev_private *internals) {
 105         struct arp_hdr *arp;
 106
 107         struct client_data *hash_table = internals->mode6.client_table;
 108         struct client_data *client_info;
 109
 110         uint8_t hash_index;
 111
 112         arp = (struct arp_hdr *) ((char *) (eth_h + 1) + offset);
 113
 114         /* ARP Requests are forwarded to the application with no changes */
 115         if (arp->arp_op != rte_cpu_to_be_16(ARP_OP_REPLY))
 116                 return;
 117
 118         /* From now on, we analyze only ARP Reply packets */
 119         hash_index = simple_hash((uint8_t *) &arp->arp_data.arp_sip,
 120                         sizeof(arp->arp_data.arp_sip));
 121         client_info = &hash_table[hash_index];
 122
 123         /*
 124          * We got reply for ARP Request send by the application. We need to
 125          * update client table when received data differ from what is stored
 126          * in ALB table and issue sending update packet to that slave.
 127          */
 128         rte_spinlock_lock(&internals->mode6.lock);
 129         if (client_info->in_use == 0 ||
 130                         client_info->app_ip != arp->arp_data.arp_tip ||
 131                         client_info->cli_ip != arp->arp_data.arp_sip ||
 132                         !is_same_ether_addr(&client_info->cli_mac, &arp->arp_data.arp_sha) ||
 133                         client_info->vlan_count != offset / sizeof(struct vlan_hdr) ||
 134                         memcmp(client_info->vlan, eth_h + 1, offset) != 0
 135         ) {
 136                 client_info->in_use = 1;
 137                 client_info->app_ip = arp->arp_data.arp_tip;
 138                 client_info->cli_ip = arp->arp_data.arp_sip;
 139                 ether_addr_copy(&arp->arp_data.arp_sha, &client_info->cli_mac);
 140                 client_info->slave_idx = calculate_slave(internals);
 141                 rte_eth_macaddr_get(client_info->slave_idx, &client_info->app_mac);
 142                 ether_addr_copy(&client_info->app_mac, &arp->arp_data.arp_tha);
 143                 memcpy(client_info->vlan, eth_h + 1, offset);
 144                 client_info->vlan_count = offset / sizeof(struct vlan_hdr);
 145         }
 146         internals->mode6.ntt = 1;
 147         rte_spinlock_unlock(&internals->mode6.lock);
 148 }
 149
 150 uint8_t
 151 bond_mode_alb_arp_xmit(struct ether_hdr *eth_h, uint16_t offset,
 152                 struct bond_dev_private *internals)
 153 {
 154         struct arp_hdr *arp;
 155
 156         struct client_data *hash_table = internals->mode6.client_table;
 157         struct client_data *client_info;
 158
 159         uint8_t hash_index;
 160
 161         struct ether_addr bonding_mac;
 162
 163         arp = (struct arp_hdr *)((char *)(eth_h + 1) + offset);
 164
 165         /*
 166          * Traffic with src MAC other than bonding should be sent on
 167          * current primary port.
 168          */
 169         rte_eth_macaddr_get(internals->port_id, &bonding_mac);
 170         if (!is_same_ether_addr(&bonding_mac, &arp->arp_data.arp_sha)) {
 171                 rte_eth_macaddr_get(internals->current_primary_port,
 172                                 &arp->arp_data.arp_sha);
 173                 return internals->current_primary_port;
 174         }
 175
 176         hash_index = simple_hash((uint8_t *)&arp->arp_data.arp_tip,
 177                         sizeof(uint32_t));
 178         client_info = &hash_table[hash_index];
 179
 180         rte_spinlock_lock(&internals->mode6.lock);
 181         if (arp->arp_op == rte_cpu_to_be_16(ARP_OP_REPLY)) {
 182                 if (client_info->in_use) {
 183                         if (client_info->app_ip == arp->arp_data.arp_sip &&
 184                                 client_info->cli_ip == arp->arp_data.arp_tip) {
 185                                 /* Entry is already assigned to this client */
 186                                 if (!is_broadcast_ether_addr(&arp->arp_data.arp_tha)) {
 187                                         ether_addr_copy(&arp->arp_data.arp_tha,
 188                                                         &client_info->cli_mac);
 189                                 }
 190                                 rte_eth_macaddr_get(client_info->slave_idx,
 191                                                 &client_info->app_mac);
 192                                 ether_addr_copy(&client_info->app_mac, &arp->arp_data.arp_sha);
 193                                 memcpy(client_info->vlan, eth_h + 1, offset);
 194                                 client_info->vlan_count = offset / sizeof(struct vlan_hdr);
 195                                 rte_spinlock_unlock(&internals->mode6.lock);
 196                                 return client_info->slave_idx;
 197                         }
 198                 }
 199
 200                 /* Assign new slave to this client and update src mac in ARP */
 201                 client_info->in_use = 1;
 202                 client_info->ntt = 0;
 203                 client_info->app_ip = arp->arp_data.arp_sip;
 204                 ether_addr_copy(&arp->arp_data.arp_tha, &client_info->cli_mac);
 205                 client_info->cli_ip = arp->arp_data.arp_tip;
 206                 client_info->slave_idx = calculate_slave(internals);
 207                 rte_eth_macaddr_get(client_info->slave_idx, &client_info->app_mac);
 208                 ether_addr_copy(&client_info->app_mac, &arp->arp_data.arp_sha);
 209                 memcpy(client_info->vlan, eth_h + 1, offset);
 210                 client_info->vlan_count = offset / sizeof(struct vlan_hdr);
 211                 rte_spinlock_unlock(&internals->mode6.lock);
 212                 return client_info->slave_idx;
 213         }
 214
 215         /* If packet is not ARP Reply, send it on current primary port. */
 216         rte_spinlock_unlock(&internals->mode6.lock);
 217         rte_eth_macaddr_get(internals->current_primary_port,
 218                         &arp->arp_data.arp_sha);
 219         return internals->current_primary_port;
 220 }
 221
 222 uint8_t
 223 bond_mode_alb_arp_upd(struct client_data *client_info,
 224                 struct rte_mbuf *pkt, struct bond_dev_private *internals)
 225 {
 226         struct ether_hdr *eth_h;
 227         struct arp_hdr *arp_h;
 228         uint8_t slave_idx;
 229
 230         rte_spinlock_lock(&internals->mode6.lock);
 231         eth_h = rte_pktmbuf_mtod(pkt, struct ether_hdr *);
 232
 233         ether_addr_copy(&client_info->app_mac, &eth_h->s_addr);
 234         ether_addr_copy(&client_info->cli_mac, &eth_h->d_addr);
 235         if (client_info->vlan_count > 0)
 236                 eth_h->ether_type = rte_cpu_to_be_16(ETHER_TYPE_VLAN);
 237         else
 238                 eth_h->ether_type = rte_cpu_to_be_16(ETHER_TYPE_ARP);
 239
 240         arp_h = (struct arp_hdr *)((char *)eth_h + sizeof(struct ether_hdr)
 241                         + client_info->vlan_count * sizeof(struct vlan_hdr));
 242
 243         memcpy(eth_h + 1, client_info->vlan,
 244                         client_info->vlan_count * sizeof(struct vlan_hdr));
 245
 246         ether_addr_copy(&client_info->app_mac, &arp_h->arp_data.arp_sha);
 247         arp_h->arp_data.arp_sip = client_info->app_ip;
 248         ether_addr_copy(&client_info->cli_mac, &arp_h->arp_data.arp_tha);
 249         arp_h->arp_data.arp_tip = client_info->cli_ip;
 250
 251         arp_h->arp_hrd = rte_cpu_to_be_16(ARP_HRD_ETHER);
 252         arp_h->arp_pro = rte_cpu_to_be_16(ETHER_TYPE_IPv4);
 253         arp_h->arp_hln = ETHER_ADDR_LEN;
 254         arp_h->arp_pln = sizeof(uint32_t);
 255         arp_h->arp_op = rte_cpu_to_be_16(ARP_OP_REPLY);
 256
 257         slave_idx = client_info->slave_idx;
 258         rte_spinlock_unlock(&internals->mode6.lock);
 259
 260         return slave_idx;
 261 }
 262
 263 void
 264 bond_mode_alb_client_list_upd(struct rte_eth_dev *bond_dev)
 265 {
 266         struct bond_dev_private *internals = bond_dev->data->dev_private;
 267         struct client_data *client_info;
 268
 269         int i;
 270
 271         /* If active slave count is 0, it's pointless to refresh alb table */
 272         if (internals->active_slave_count <= 0)
 273                 return;
 274
 275         rte_spinlock_lock(&internals->mode6.lock);
 276         internals->mode6.last_slave = ALB_NULL_INDEX;
 277
 278         for (i = 0; i < ALB_HASH_TABLE_SIZE; i++) {
 279                 client_info = &internals->mode6.client_table[i];
 280                 if (client_info->in_use) {
 281                         client_info->slave_idx = calculate_slave(internals);
 282                         rte_eth_macaddr_get(client_info->slave_idx, &client_info->app_mac);
 283                         internals->mode6.ntt = 1;
 284                 }
 285         }
 286         rte_spinlock_unlock(&internals->mode6.lock);
 287 }