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
56 * This driver is for virtio_pci on SMP and therefore can assume
57 * weaker (compiler barriers)
59 #define virtio_mb() rte_mb()
60 #define virtio_rmb() rte_compiler_barrier()
61 #define virtio_wmb() rte_compiler_barrier()
63 #ifdef RTE_PMD_PACKET_PREFETCH
64 #define rte_packet_prefetch(p) rte_prefetch1(p)
66 #define rte_packet_prefetch(p) do {} while(0)
69 #define VIRTQUEUE_MAX_NAME_SZ 32
71 #define RTE_MBUF_DATA_DMA_ADDR(mb) \
72 (uint64_t) ((mb)->buf_physaddr + (mb)->data_off)
74 #define VTNET_SQ_RQ_QUEUE_IDX 0
75 #define VTNET_SQ_TQ_QUEUE_IDX 1
76 #define VTNET_SQ_CQ_QUEUE_IDX 2
78 enum { VTNET_RQ = 0, VTNET_TQ = 1, VTNET_CQ = 2 };
80 * The maximum virtqueue size is 2^15. Use that value as the end of
81 * descriptor chain terminator since it will never be a valid index
82 * in the descriptor table. This is used to verify we are correctly
83 * handling vq_free_cnt.
85 #define VQ_RING_DESC_CHAIN_END 32768
88 * Control the RX mode, ie. promiscuous, allmulti, etc...
89 * All commands require an "out" sg entry containing a 1 byte
90 * state value, zero = disable, non-zero = enable. Commands
91 * 0 and 1 are supported with the VIRTIO_NET_F_CTRL_RX feature.
92 * Commands 2-5 are added with VIRTIO_NET_F_CTRL_RX_EXTRA.
94 #define VIRTIO_NET_CTRL_RX 0
95 #define VIRTIO_NET_CTRL_RX_PROMISC 0
96 #define VIRTIO_NET_CTRL_RX_ALLMULTI 1
97 #define VIRTIO_NET_CTRL_RX_ALLUNI 2
98 #define VIRTIO_NET_CTRL_RX_NOMULTI 3
99 #define VIRTIO_NET_CTRL_RX_NOUNI 4
100 #define VIRTIO_NET_CTRL_RX_NOBCAST 5
105 * The MAC filter table is managed by the hypervisor, the guest should
106 * assume the size is infinite. Filtering should be considered
107 * non-perfect, ie. based on hypervisor resources, the guest may
108 * received packets from sources not specified in the filter list.
110 * In addition to the class/cmd header, the TABLE_SET command requires
111 * two out scatterlists. Each contains a 4 byte count of entries followed
112 * by a concatenated byte stream of the ETH_ALEN MAC addresses. The
113 * first sg list contains unicast addresses, the second is for multicast.
114 * This functionality is present if the VIRTIO_NET_F_CTRL_RX feature
117 * The ADDR_SET command requests one out scatterlist, it contains a
118 * 6 bytes MAC address. This functionality is present if the
119 * VIRTIO_NET_F_CTRL_MAC_ADDR feature is available.
121 struct virtio_net_ctrl_mac {
123 uint8_t macs[][ETHER_ADDR_LEN];
124 } __attribute__((__packed__));
126 #define VIRTIO_NET_CTRL_MAC 1
127 #define VIRTIO_NET_CTRL_MAC_TABLE_SET 0
128 #define VIRTIO_NET_CTRL_MAC_ADDR_SET 1
131 * Control VLAN filtering
133 * The VLAN filter table is controlled via a simple ADD/DEL interface.
134 * VLAN IDs not added may be filtered by the hypervisor. Del is the
135 * opposite of add. Both commands expect an out entry containing a 2
136 * byte VLAN ID. VLAN filtering is available with the
137 * VIRTIO_NET_F_CTRL_VLAN feature bit.
139 #define VIRTIO_NET_CTRL_VLAN 2
140 #define VIRTIO_NET_CTRL_VLAN_ADD 0
141 #define VIRTIO_NET_CTRL_VLAN_DEL 1
143 struct virtio_net_ctrl_hdr {
146 } __attribute__((packed));
148 typedef uint8_t virtio_net_ctrl_ack;
150 #define VIRTIO_NET_OK 0
151 #define VIRTIO_NET_ERR 1
153 #define VIRTIO_MAX_CTRL_DATA 2048
155 struct virtio_pmd_ctrl {
156 struct virtio_net_ctrl_hdr hdr;
157 virtio_net_ctrl_ack status;
158 uint8_t data[VIRTIO_MAX_CTRL_DATA];
162 struct virtio_hw *hw; /**< virtio_hw structure pointer. */
163 const struct rte_memzone *mz; /**< mem zone to populate RX ring. */
164 const struct rte_memzone *virtio_net_hdr_mz; /**< memzone to populate hdr. */
165 struct rte_mempool *mpool; /**< mempool for mbuf allocation */
166 uint16_t queue_id; /**< DPDK queue index. */
167 uint8_t port_id; /**< Device port identifier. */
168 uint16_t vq_queue_index; /**< PCI queue index */
170 void *vq_ring_virt_mem; /**< linear address of vring*/
171 unsigned int vq_ring_size;
172 phys_addr_t vq_ring_mem; /**< physical address of vring */
174 struct vring vq_ring; /**< vring keeping desc, used and avail */
175 uint16_t vq_free_cnt; /**< num of desc available */
176 uint16_t vq_nentries; /**< vring desc numbers */
177 uint16_t vq_free_thresh; /**< free threshold */
179 * Head of the free chain in the descriptor table. If
180 * there are no free descriptors, this will be set to
181 * VQ_RING_DESC_CHAIN_END.
183 uint16_t vq_desc_head_idx;
184 uint16_t vq_desc_tail_idx;
186 * Last consumed descriptor in the used table,
187 * trails vq_ring.used->idx.
189 uint16_t vq_used_cons_idx;
190 uint16_t vq_avail_idx;
191 uint64_t mbuf_initializer; /**< value to init mbufs. */
192 phys_addr_t virtio_net_hdr_mem; /**< hdr for each xmit packet */
194 struct rte_mbuf **sw_ring; /**< RX software ring. */
195 /* dummy mbuf, for wraparound when processing RX ring. */
196 struct rte_mbuf fake_mbuf;
204 /* Size bins in array as RFC 2819, undersized [0], 64 [1], etc */
205 uint64_t size_bins[8];
207 struct vq_desc_extra {
213 /* If multiqueue is provided by host, then we suppport it. */
214 #define VIRTIO_NET_CTRL_MQ 4
215 #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET 0
216 #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MIN 1
217 #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MAX 0x8000
219 #define VIRTIO_NET_CTRL_MAC_ADDR_SET 1
222 * This is the first element of the scatter-gather list. If you don't
223 * specify GSO or CSUM features, you can simply ignore the header.
225 struct virtio_net_hdr {
226 #define VIRTIO_NET_HDR_F_NEEDS_CSUM 1 /**< Use csum_start,csum_offset*/
228 #define VIRTIO_NET_HDR_GSO_NONE 0 /**< Not a GSO frame */
229 #define VIRTIO_NET_HDR_GSO_TCPV4 1 /**< GSO frame, IPv4 TCP (TSO) */
230 #define VIRTIO_NET_HDR_GSO_UDP 3 /**< GSO frame, IPv4 UDP (UFO) */
231 #define VIRTIO_NET_HDR_GSO_TCPV6 4 /**< GSO frame, IPv6 TCP */
232 #define VIRTIO_NET_HDR_GSO_ECN 0x80 /**< TCP has ECN set */
234 uint16_t hdr_len; /**< Ethernet + IP + tcp/udp hdrs */
235 uint16_t gso_size; /**< Bytes to append to hdr_len per frame */
236 uint16_t csum_start; /**< Position to start checksumming from */
237 uint16_t csum_offset; /**< Offset after that to place checksum */
241 * This is the version of the header to use when the MRG_RXBUF
242 * feature has been negotiated.
244 struct virtio_net_hdr_mrg_rxbuf {
245 struct virtio_net_hdr hdr;
246 uint16_t num_buffers; /**< Number of merged rx buffers */
250 * Tell the backend not to interrupt us.
252 void virtqueue_disable_intr(struct virtqueue *vq);
254 * Dump virtqueue internal structures, for debug purpose only.
256 void virtqueue_dump(struct virtqueue *vq);
258 * Get all mbufs to be freed.
260 struct rte_mbuf *virtqueue_detatch_unused(struct virtqueue *vq);
263 virtqueue_full(const struct virtqueue *vq)
265 return vq->vq_free_cnt == 0;
268 #define VIRTQUEUE_NUSED(vq) ((uint16_t)((vq)->vq_ring.used->idx - (vq)->vq_used_cons_idx))
271 vq_update_avail_idx(struct virtqueue *vq)
274 vq->vq_ring.avail->idx = vq->vq_avail_idx;
278 vq_update_avail_ring(struct virtqueue *vq, uint16_t desc_idx)
282 * Place the head of the descriptor chain into the next slot and make
283 * it usable to the host. The chain is made available now rather than
284 * deferring to virtqueue_notify() in the hopes that if the host is
285 * currently running on another CPU, we can keep it processing the new
288 avail_idx = (uint16_t)(vq->vq_avail_idx & (vq->vq_nentries - 1));
289 vq->vq_ring.avail->ring[avail_idx] = desc_idx;
294 virtqueue_kick_prepare(struct virtqueue *vq)
296 return !(vq->vq_ring.used->flags & VRING_USED_F_NO_NOTIFY);
300 virtqueue_notify(struct virtqueue *vq)
303 * Ensure updated avail->idx is visible to host.
304 * For virtio on IA, the notificaiton is through io port operation
305 * which is a serialization instruction itself.
307 VIRTIO_WRITE_REG_2(vq->hw, VIRTIO_PCI_QUEUE_NOTIFY, vq->vq_queue_index);
310 #ifdef RTE_LIBRTE_VIRTIO_DEBUG_DUMP
311 #define VIRTQUEUE_DUMP(vq) do { \
312 uint16_t used_idx, nused; \
313 used_idx = (vq)->vq_ring.used->idx; \
314 nused = (uint16_t)(used_idx - (vq)->vq_used_cons_idx); \
315 PMD_INIT_LOG(DEBUG, \
316 "VQ: - size=%d; free=%d; used=%d; desc_head_idx=%d;" \
317 " avail.idx=%d; used_cons_idx=%d; used.idx=%d;" \
318 " avail.flags=0x%x; used.flags=0x%x", \
319 (vq)->vq_nentries, (vq)->vq_free_cnt, nused, \
320 (vq)->vq_desc_head_idx, (vq)->vq_ring.avail->idx, \
321 (vq)->vq_used_cons_idx, (vq)->vq_ring.used->idx, \
322 (vq)->vq_ring.avail->flags, (vq)->vq_ring.used->flags); \
325 #define VIRTQUEUE_DUMP(vq) do { } while (0)
328 #endif /* _VIRTQUEUE_H_ */