4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
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.
39 #include <rte_atomic.h>
40 #include <rte_memory.h>
41 #include <rte_memzone.h>
42 #include <rte_mempool.h>
44 #include "virtio_pci.h"
45 #include "virtio_ring.h"
46 #include "virtio_logs.h"
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
57 #define virtio_mb() rte_smp_mb()
58 #define virtio_rmb() rte_smp_rmb()
59 #define virtio_wmb() rte_smp_wmb()
61 #ifdef RTE_PMD_PACKET_PREFETCH
62 #define rte_packet_prefetch(p) rte_prefetch1(p)
64 #define rte_packet_prefetch(p) do {} while(0)
67 #define VIRTQUEUE_MAX_NAME_SZ 32
69 #define RTE_MBUF_DATA_DMA_ADDR(mb) \
70 (uint64_t) ((mb)->buf_physaddr + (mb)->data_off)
72 #define VTNET_SQ_RQ_QUEUE_IDX 0
73 #define VTNET_SQ_TQ_QUEUE_IDX 1
74 #define VTNET_SQ_CQ_QUEUE_IDX 2
76 enum { VTNET_RQ = 0, VTNET_TQ = 1, VTNET_CQ = 2 };
78 * The maximum virtqueue size is 2^15. Use that value as the end of
79 * descriptor chain terminator since it will never be a valid index
80 * in the descriptor table. This is used to verify we are correctly
81 * handling vq_free_cnt.
83 #define VQ_RING_DESC_CHAIN_END 32768
86 * Control the RX mode, ie. promiscuous, allmulti, etc...
87 * All commands require an "out" sg entry containing a 1 byte
88 * state value, zero = disable, non-zero = enable. Commands
89 * 0 and 1 are supported with the VIRTIO_NET_F_CTRL_RX feature.
90 * Commands 2-5 are added with VIRTIO_NET_F_CTRL_RX_EXTRA.
92 #define VIRTIO_NET_CTRL_RX 0
93 #define VIRTIO_NET_CTRL_RX_PROMISC 0
94 #define VIRTIO_NET_CTRL_RX_ALLMULTI 1
95 #define VIRTIO_NET_CTRL_RX_ALLUNI 2
96 #define VIRTIO_NET_CTRL_RX_NOMULTI 3
97 #define VIRTIO_NET_CTRL_RX_NOUNI 4
98 #define VIRTIO_NET_CTRL_RX_NOBCAST 5
103 * The MAC filter table is managed by the hypervisor, the guest should
104 * assume the size is infinite. Filtering should be considered
105 * non-perfect, ie. based on hypervisor resources, the guest may
106 * received packets from sources not specified in the filter list.
108 * In addition to the class/cmd header, the TABLE_SET command requires
109 * two out scatterlists. Each contains a 4 byte count of entries followed
110 * by a concatenated byte stream of the ETH_ALEN MAC addresses. The
111 * first sg list contains unicast addresses, the second is for multicast.
112 * This functionality is present if the VIRTIO_NET_F_CTRL_RX feature
115 * The ADDR_SET command requests one out scatterlist, it contains a
116 * 6 bytes MAC address. This functionality is present if the
117 * VIRTIO_NET_F_CTRL_MAC_ADDR feature is available.
119 struct virtio_net_ctrl_mac {
121 uint8_t macs[][ETHER_ADDR_LEN];
122 } __attribute__((__packed__));
124 #define VIRTIO_NET_CTRL_MAC 1
125 #define VIRTIO_NET_CTRL_MAC_TABLE_SET 0
126 #define VIRTIO_NET_CTRL_MAC_ADDR_SET 1
129 * Control VLAN filtering
131 * The VLAN filter table is controlled via a simple ADD/DEL interface.
132 * VLAN IDs not added may be filtered by the hypervisor. Del is the
133 * opposite of add. Both commands expect an out entry containing a 2
134 * byte VLAN ID. VLAN filtering is available with the
135 * VIRTIO_NET_F_CTRL_VLAN feature bit.
137 #define VIRTIO_NET_CTRL_VLAN 2
138 #define VIRTIO_NET_CTRL_VLAN_ADD 0
139 #define VIRTIO_NET_CTRL_VLAN_DEL 1
141 struct virtio_net_ctrl_hdr {
144 } __attribute__((packed));
146 typedef uint8_t virtio_net_ctrl_ack;
148 #define VIRTIO_NET_OK 0
149 #define VIRTIO_NET_ERR 1
151 #define VIRTIO_MAX_CTRL_DATA 2048
153 struct virtio_pmd_ctrl {
154 struct virtio_net_ctrl_hdr hdr;
155 virtio_net_ctrl_ack status;
156 uint8_t data[VIRTIO_MAX_CTRL_DATA];
160 struct virtio_hw *hw; /**< virtio_hw structure pointer. */
161 const struct rte_memzone *mz; /**< mem zone to populate RX ring. */
162 const struct rte_memzone *virtio_net_hdr_mz; /**< memzone to populate hdr. */
163 struct rte_mempool *mpool; /**< mempool for mbuf allocation */
164 uint16_t queue_id; /**< DPDK queue index. */
165 uint8_t port_id; /**< Device port identifier. */
166 uint16_t vq_queue_index; /**< PCI queue index */
168 void *vq_ring_virt_mem; /**< linear address of vring*/
169 unsigned int vq_ring_size;
170 phys_addr_t vq_ring_mem; /**< physical address of vring */
172 struct vring vq_ring; /**< vring keeping desc, used and avail */
173 uint16_t vq_free_cnt; /**< num of desc available */
174 uint16_t vq_nentries; /**< vring desc numbers */
175 uint16_t vq_free_thresh; /**< free threshold */
177 * Head of the free chain in the descriptor table. If
178 * there are no free descriptors, this will be set to
179 * VQ_RING_DESC_CHAIN_END.
181 uint16_t vq_desc_head_idx;
182 uint16_t vq_desc_tail_idx;
184 * Last consumed descriptor in the used table,
185 * trails vq_ring.used->idx.
187 uint16_t vq_used_cons_idx;
188 uint16_t vq_avail_idx;
189 uint64_t mbuf_initializer; /**< value to init mbufs. */
190 phys_addr_t virtio_net_hdr_mem; /**< hdr for each xmit packet */
192 struct rte_mbuf **sw_ring; /**< RX software ring. */
193 /* dummy mbuf, for wraparound when processing RX ring. */
194 struct rte_mbuf fake_mbuf;
202 /* Size bins in array as RFC 2819, undersized [0], 64 [1], etc */
203 uint64_t size_bins[8];
205 struct vq_desc_extra {
211 /* If multiqueue is provided by host, then we suppport it. */
212 #define VIRTIO_NET_CTRL_MQ 4
213 #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET 0
214 #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MIN 1
215 #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MAX 0x8000
217 #define VIRTIO_NET_CTRL_MAC_ADDR_SET 1
220 * This is the first element of the scatter-gather list. If you don't
221 * specify GSO or CSUM features, you can simply ignore the header.
223 struct virtio_net_hdr {
224 #define VIRTIO_NET_HDR_F_NEEDS_CSUM 1 /**< Use csum_start,csum_offset*/
226 #define VIRTIO_NET_HDR_GSO_NONE 0 /**< Not a GSO frame */
227 #define VIRTIO_NET_HDR_GSO_TCPV4 1 /**< GSO frame, IPv4 TCP (TSO) */
228 #define VIRTIO_NET_HDR_GSO_UDP 3 /**< GSO frame, IPv4 UDP (UFO) */
229 #define VIRTIO_NET_HDR_GSO_TCPV6 4 /**< GSO frame, IPv6 TCP */
230 #define VIRTIO_NET_HDR_GSO_ECN 0x80 /**< TCP has ECN set */
232 uint16_t hdr_len; /**< Ethernet + IP + tcp/udp hdrs */
233 uint16_t gso_size; /**< Bytes to append to hdr_len per frame */
234 uint16_t csum_start; /**< Position to start checksumming from */
235 uint16_t csum_offset; /**< Offset after that to place checksum */
239 * This is the version of the header to use when the MRG_RXBUF
240 * feature has been negotiated.
242 struct virtio_net_hdr_mrg_rxbuf {
243 struct virtio_net_hdr hdr;
244 uint16_t num_buffers; /**< Number of merged rx buffers */
248 * Tell the backend not to interrupt us.
250 void virtqueue_disable_intr(struct virtqueue *vq);
252 * Dump virtqueue internal structures, for debug purpose only.
254 void virtqueue_dump(struct virtqueue *vq);
256 * Get all mbufs to be freed.
258 struct rte_mbuf *virtqueue_detatch_unused(struct virtqueue *vq);
261 virtqueue_full(const struct virtqueue *vq)
263 return vq->vq_free_cnt == 0;
266 #define VIRTQUEUE_NUSED(vq) ((uint16_t)((vq)->vq_ring.used->idx - (vq)->vq_used_cons_idx))
269 vq_update_avail_idx(struct virtqueue *vq)
272 vq->vq_ring.avail->idx = vq->vq_avail_idx;
276 vq_update_avail_ring(struct virtqueue *vq, uint16_t desc_idx)
280 * Place the head of the descriptor chain into the next slot and make
281 * it usable to the host. The chain is made available now rather than
282 * deferring to virtqueue_notify() in the hopes that if the host is
283 * currently running on another CPU, we can keep it processing the new
286 avail_idx = (uint16_t)(vq->vq_avail_idx & (vq->vq_nentries - 1));
287 vq->vq_ring.avail->ring[avail_idx] = desc_idx;
292 virtqueue_kick_prepare(struct virtqueue *vq)
294 return !(vq->vq_ring.used->flags & VRING_USED_F_NO_NOTIFY);
298 virtqueue_notify(struct virtqueue *vq)
301 * Ensure updated avail->idx is visible to host.
302 * For virtio on IA, the notificaiton is through io port operation
303 * which is a serialization instruction itself.
305 VIRTIO_WRITE_REG_2(vq->hw, VIRTIO_PCI_QUEUE_NOTIFY, vq->vq_queue_index);
308 #ifdef RTE_LIBRTE_VIRTIO_DEBUG_DUMP
309 #define VIRTQUEUE_DUMP(vq) do { \
310 uint16_t used_idx, nused; \
311 used_idx = (vq)->vq_ring.used->idx; \
312 nused = (uint16_t)(used_idx - (vq)->vq_used_cons_idx); \
313 PMD_INIT_LOG(DEBUG, \
314 "VQ: - size=%d; free=%d; used=%d; desc_head_idx=%d;" \
315 " avail.idx=%d; used_cons_idx=%d; used.idx=%d;" \
316 " avail.flags=0x%x; used.flags=0x%x", \
317 (vq)->vq_nentries, (vq)->vq_free_cnt, nused, \
318 (vq)->vq_desc_head_idx, (vq)->vq_ring.avail->idx, \
319 (vq)->vq_used_cons_idx, (vq)->vq_ring.used->idx, \
320 (vq)->vq_ring.avail->flags, (vq)->vq_ring.used->flags); \
323 #define VIRTQUEUE_DUMP(vq) do { } while (0)
326 #endif /* _VIRTQUEUE_H_ */