1 /*- 2 * BSD LICENSE 3 * 4 * Copyright(c) 2010-2014 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 #ifndef _VIRTQUEUE_H_ 35 #define _VIRTQUEUE_H_ 36 37 #include <stdint.h> 38 39 #include <rte_atomic.h> 40 #include <rte_memory.h> 41 #include <rte_memzone.h> 42 #include <rte_mempool.h> 43 44 #include "virtio_pci.h" 45 #include "virtio_ring.h" 46 #include "virtio_logs.h" 47 48 struct rte_mbuf; 49 50 /* 51 * Per virtio_config.h in Linux. 52 * For virtio_pci on SMP, we don't need to order with respect to MMIO 53 * accesses through relaxed memory I/O windows, so smp_mb() et al are 54 * sufficient. 55 * 56 */ 57 #define virtio_mb() rte_smp_mb() 58 #define virtio_rmb() rte_smp_rmb() 59 #define virtio_wmb() rte_smp_wmb() 60 61 #ifdef RTE_PMD_PACKET_PREFETCH 62 #define rte_packet_prefetch(p) rte_prefetch1(p) 63 #else 64 #define rte_packet_prefetch(p) do {} while(0) 65 #endif 66 67 #define VIRTQUEUE_MAX_NAME_SZ 32 68 69 #ifdef RTE_VIRTIO_USER 70 #define MBUF_DATA_DMA_ADDR(mb, offset) \ 71 ((uint64_t)((uintptr_t)(*(void **)((uintptr_t)mb + offset)) \ 72 + (mb)->data_off)) 73 #else /* RTE_VIRTIO_USER */ 74 #define MBUF_DATA_DMA_ADDR(mb, offset) rte_mbuf_data_dma_addr(mb) 75 #endif /* RTE_VIRTIO_USER */ 76 77 #define VTNET_SQ_RQ_QUEUE_IDX 0 78 #define VTNET_SQ_TQ_QUEUE_IDX 1 79 #define VTNET_SQ_CQ_QUEUE_IDX 2 80 81 enum { VTNET_RQ = 0, VTNET_TQ = 1, VTNET_CQ = 2 }; 82 /** 83 * The maximum virtqueue size is 2^15. Use that value as the end of 84 * descriptor chain terminator since it will never be a valid index 85 * in the descriptor table. This is used to verify we are correctly 86 * handling vq_free_cnt. 87 */ 88 #define VQ_RING_DESC_CHAIN_END 32768 89 90 /** 91 * Control the RX mode, ie. promiscuous, allmulti, etc... 92 * All commands require an "out" sg entry containing a 1 byte 93 * state value, zero = disable, non-zero = enable. Commands 94 * 0 and 1 are supported with the VIRTIO_NET_F_CTRL_RX feature. 95 * Commands 2-5 are added with VIRTIO_NET_F_CTRL_RX_EXTRA. 96 */ 97 #define VIRTIO_NET_CTRL_RX 0 98 #define VIRTIO_NET_CTRL_RX_PROMISC 0 99 #define VIRTIO_NET_CTRL_RX_ALLMULTI 1 100 #define VIRTIO_NET_CTRL_RX_ALLUNI 2 101 #define VIRTIO_NET_CTRL_RX_NOMULTI 3 102 #define VIRTIO_NET_CTRL_RX_NOUNI 4 103 #define VIRTIO_NET_CTRL_RX_NOBCAST 5 104 105 /** 106 * Control the MAC 107 * 108 * The MAC filter table is managed by the hypervisor, the guest should 109 * assume the size is infinite. Filtering should be considered 110 * non-perfect, ie. based on hypervisor resources, the guest may 111 * received packets from sources not specified in the filter list. 112 * 113 * In addition to the class/cmd header, the TABLE_SET command requires 114 * two out scatterlists. Each contains a 4 byte count of entries followed 115 * by a concatenated byte stream of the ETH_ALEN MAC addresses. The 116 * first sg list contains unicast addresses, the second is for multicast. 117 * This functionality is present if the VIRTIO_NET_F_CTRL_RX feature 118 * is available. 119 * 120 * The ADDR_SET command requests one out scatterlist, it contains a 121 * 6 bytes MAC address. This functionality is present if the 122 * VIRTIO_NET_F_CTRL_MAC_ADDR feature is available. 123 */ 124 struct virtio_net_ctrl_mac { 125 uint32_t entries; 126 uint8_t macs[][ETHER_ADDR_LEN]; 127 } __attribute__((__packed__)); 128 129 #define VIRTIO_NET_CTRL_MAC 1 130 #define VIRTIO_NET_CTRL_MAC_TABLE_SET 0 131 #define VIRTIO_NET_CTRL_MAC_ADDR_SET 1 132 133 /** 134 * Control VLAN filtering 135 * 136 * The VLAN filter table is controlled via a simple ADD/DEL interface. 137 * VLAN IDs not added may be filtered by the hypervisor. Del is the 138 * opposite of add. Both commands expect an out entry containing a 2 139 * byte VLAN ID. VLAN filtering is available with the 140 * VIRTIO_NET_F_CTRL_VLAN feature bit. 141 */ 142 #define VIRTIO_NET_CTRL_VLAN 2 143 #define VIRTIO_NET_CTRL_VLAN_ADD 0 144 #define VIRTIO_NET_CTRL_VLAN_DEL 1 145 146 struct virtio_net_ctrl_hdr { 147 uint8_t class; 148 uint8_t cmd; 149 } __attribute__((packed)); 150 151 typedef uint8_t virtio_net_ctrl_ack; 152 153 #define VIRTIO_NET_OK 0 154 #define VIRTIO_NET_ERR 1 155 156 #define VIRTIO_MAX_CTRL_DATA 2048 157 158 struct virtio_pmd_ctrl { 159 struct virtio_net_ctrl_hdr hdr; 160 virtio_net_ctrl_ack status; 161 uint8_t data[VIRTIO_MAX_CTRL_DATA]; 162 }; 163 164 struct vq_desc_extra { 165 void *cookie; 166 uint16_t ndescs; 167 }; 168 169 struct virtqueue { 170 struct virtio_hw *hw; /**< virtio_hw structure pointer. */ 171 struct vring vq_ring; /**< vring keeping desc, used and avail */ 172 /** 173 * Last consumed descriptor in the used table, 174 * trails vq_ring.used->idx. 175 */ 176 uint16_t vq_used_cons_idx; 177 uint16_t vq_nentries; /**< vring desc numbers */ 178 uint16_t vq_free_cnt; /**< num of desc available */ 179 uint16_t vq_avail_idx; /**< sync until needed */ 180 uint16_t vq_free_thresh; /**< free threshold */ 181 182 void *vq_ring_virt_mem; /**< linear address of vring*/ 183 unsigned int vq_ring_size; 184 185 phys_addr_t vq_ring_mem; /**< physical address of vring */ 186 /**< use virtual address for virtio-user. */ 187 188 /** 189 * Head of the free chain in the descriptor table. If 190 * there are no free descriptors, this will be set to 191 * VQ_RING_DESC_CHAIN_END. 192 */ 193 uint16_t vq_desc_head_idx; 194 uint16_t vq_desc_tail_idx; 195 uint16_t vq_queue_index; /**< PCI queue index */ 196 uint16_t offset; /**< relative offset to obtain addr in mbuf */ 197 uint16_t *notify_addr; 198 int configured; 199 struct rte_mbuf **sw_ring; /**< RX software ring. */ 200 struct vq_desc_extra vq_descx[0]; 201 }; 202 203 /* If multiqueue is provided by host, then we suppport it. */ 204 #define VIRTIO_NET_CTRL_MQ 4 205 #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET 0 206 #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MIN 1 207 #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MAX 0x8000 208 209 #define VIRTIO_NET_CTRL_MAC_ADDR_SET 1 210 211 /** 212 * This is the first element of the scatter-gather list. If you don't 213 * specify GSO or CSUM features, you can simply ignore the header. 214 */ 215 struct virtio_net_hdr { 216 #define VIRTIO_NET_HDR_F_NEEDS_CSUM 1 /**< Use csum_start,csum_offset*/ 217 uint8_t flags; 218 #define VIRTIO_NET_HDR_GSO_NONE 0 /**< Not a GSO frame */ 219 #define VIRTIO_NET_HDR_GSO_TCPV4 1 /**< GSO frame, IPv4 TCP (TSO) */ 220 #define VIRTIO_NET_HDR_GSO_UDP 3 /**< GSO frame, IPv4 UDP (UFO) */ 221 #define VIRTIO_NET_HDR_GSO_TCPV6 4 /**< GSO frame, IPv6 TCP */ 222 #define VIRTIO_NET_HDR_GSO_ECN 0x80 /**< TCP has ECN set */ 223 uint8_t gso_type; 224 uint16_t hdr_len; /**< Ethernet + IP + tcp/udp hdrs */ 225 uint16_t gso_size; /**< Bytes to append to hdr_len per frame */ 226 uint16_t csum_start; /**< Position to start checksumming from */ 227 uint16_t csum_offset; /**< Offset after that to place checksum */ 228 }; 229 230 /** 231 * This is the version of the header to use when the MRG_RXBUF 232 * feature has been negotiated. 233 */ 234 struct virtio_net_hdr_mrg_rxbuf { 235 struct virtio_net_hdr hdr; 236 uint16_t num_buffers; /**< Number of merged rx buffers */ 237 }; 238 239 /* Region reserved to allow for transmit header and indirect ring */ 240 #define VIRTIO_MAX_TX_INDIRECT 8 241 struct virtio_tx_region { 242 struct virtio_net_hdr_mrg_rxbuf tx_hdr; 243 struct vring_desc tx_indir[VIRTIO_MAX_TX_INDIRECT] 244 __attribute__((__aligned__(16))); 245 }; 246 247 /* Chain all the descriptors in the ring with an END */ 248 static inline void 249 vring_desc_init(struct vring_desc *dp, uint16_t n) 250 { 251 uint16_t i; 252 253 for (i = 0; i < n - 1; i++) 254 dp[i].next = (uint16_t)(i + 1); 255 dp[i].next = VQ_RING_DESC_CHAIN_END; 256 } 257 258 /** 259 * Tell the backend not to interrupt us. 260 */ 261 void virtqueue_disable_intr(struct virtqueue *vq); 262 /** 263 * Dump virtqueue internal structures, for debug purpose only. 264 */ 265 void virtqueue_dump(struct virtqueue *vq); 266 /** 267 * Get all mbufs to be freed. 268 */ 269 struct rte_mbuf *virtqueue_detatch_unused(struct virtqueue *vq); 270 271 static inline int 272 virtqueue_full(const struct virtqueue *vq) 273 { 274 return vq->vq_free_cnt == 0; 275 } 276 277 #define VIRTQUEUE_NUSED(vq) ((uint16_t)((vq)->vq_ring.used->idx - (vq)->vq_used_cons_idx)) 278 279 static inline void 280 vq_update_avail_idx(struct virtqueue *vq) 281 { 282 virtio_wmb(); 283 vq->vq_ring.avail->idx = vq->vq_avail_idx; 284 } 285 286 static inline void 287 vq_update_avail_ring(struct virtqueue *vq, uint16_t desc_idx) 288 { 289 uint16_t avail_idx; 290 /* 291 * Place the head of the descriptor chain into the next slot and make 292 * it usable to the host. The chain is made available now rather than 293 * deferring to virtqueue_notify() in the hopes that if the host is 294 * currently running on another CPU, we can keep it processing the new 295 * descriptor. 296 */ 297 avail_idx = (uint16_t)(vq->vq_avail_idx & (vq->vq_nentries - 1)); 298 if (unlikely(vq->vq_ring.avail->ring[avail_idx] != desc_idx)) 299 vq->vq_ring.avail->ring[avail_idx] = desc_idx; 300 vq->vq_avail_idx++; 301 } 302 303 static inline int 304 virtqueue_kick_prepare(struct virtqueue *vq) 305 { 306 return !(vq->vq_ring.used->flags & VRING_USED_F_NO_NOTIFY); 307 } 308 309 static inline void 310 virtqueue_notify(struct virtqueue *vq) 311 { 312 /* 313 * Ensure updated avail->idx is visible to host. 314 * For virtio on IA, the notificaiton is through io port operation 315 * which is a serialization instruction itself. 316 */ 317 vq->hw->vtpci_ops->notify_queue(vq->hw, vq); 318 } 319 320 #ifdef RTE_LIBRTE_VIRTIO_DEBUG_DUMP 321 #define VIRTQUEUE_DUMP(vq) do { \ 322 uint16_t used_idx, nused; \ 323 used_idx = (vq)->vq_ring.used->idx; \ 324 nused = (uint16_t)(used_idx - (vq)->vq_used_cons_idx); \ 325 PMD_INIT_LOG(DEBUG, \ 326 "VQ: - size=%d; free=%d; used=%d; desc_head_idx=%d;" \ 327 " avail.idx=%d; used_cons_idx=%d; used.idx=%d;" \ 328 " avail.flags=0x%x; used.flags=0x%x", \ 329 (vq)->vq_nentries, (vq)->vq_free_cnt, nused, \ 330 (vq)->vq_desc_head_idx, (vq)->vq_ring.avail->idx, \ 331 (vq)->vq_used_cons_idx, (vq)->vq_ring.used->idx, \ 332 (vq)->vq_ring.avail->flags, (vq)->vq_ring.used->flags); \ 333 } while (0) 334 #else 335 #define VIRTQUEUE_DUMP(vq) do { } while (0) 336 #endif 337 338 #endif /* _VIRTQUEUE_H_ */ 339