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_mempool.h>
43 #include "virtio_pci.h"
44 #include "virtio_ring.h"
45 #include "virtio_logs.h"
46 #include "virtio_rxtx.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 #ifdef RTE_VIRTIO_USER
71 * Return the physical address (or virtual address in case of
72 * virtio-user) of mbuf data buffer.
74 * The address is firstly casted to the word size (sizeof(uintptr_t))
75 * before casting it to uint64_t. This is to make it work with different
76 * combination of word size (64 bit and 32 bit) and virtio device
77 * (virtio-pci and virtio-user).
79 #define VIRTIO_MBUF_ADDR(mb, vq) \
80 ((uint64_t)(*(uintptr_t *)((uintptr_t)(mb) + (vq)->offset)))
82 #define VIRTIO_MBUF_ADDR(mb, vq) ((mb)->buf_iova)
86 * Return the physical address (or virtual address in case of
87 * virtio-user) of mbuf data buffer, taking care of mbuf data offset
89 #define VIRTIO_MBUF_DATA_DMA_ADDR(mb, vq) \
90 (VIRTIO_MBUF_ADDR(mb, vq) + (mb)->data_off)
92 #define VTNET_SQ_RQ_QUEUE_IDX 0
93 #define VTNET_SQ_TQ_QUEUE_IDX 1
94 #define VTNET_SQ_CQ_QUEUE_IDX 2
96 enum { VTNET_RQ = 0, VTNET_TQ = 1, VTNET_CQ = 2 };
98 * The maximum virtqueue size is 2^15. Use that value as the end of
99 * descriptor chain terminator since it will never be a valid index
100 * in the descriptor table. This is used to verify we are correctly
101 * handling vq_free_cnt.
103 #define VQ_RING_DESC_CHAIN_END 32768
106 * Control the RX mode, ie. promiscuous, allmulti, etc...
107 * All commands require an "out" sg entry containing a 1 byte
108 * state value, zero = disable, non-zero = enable. Commands
109 * 0 and 1 are supported with the VIRTIO_NET_F_CTRL_RX feature.
110 * Commands 2-5 are added with VIRTIO_NET_F_CTRL_RX_EXTRA.
112 #define VIRTIO_NET_CTRL_RX 0
113 #define VIRTIO_NET_CTRL_RX_PROMISC 0
114 #define VIRTIO_NET_CTRL_RX_ALLMULTI 1
115 #define VIRTIO_NET_CTRL_RX_ALLUNI 2
116 #define VIRTIO_NET_CTRL_RX_NOMULTI 3
117 #define VIRTIO_NET_CTRL_RX_NOUNI 4
118 #define VIRTIO_NET_CTRL_RX_NOBCAST 5
123 * The MAC filter table is managed by the hypervisor, the guest should
124 * assume the size is infinite. Filtering should be considered
125 * non-perfect, ie. based on hypervisor resources, the guest may
126 * received packets from sources not specified in the filter list.
128 * In addition to the class/cmd header, the TABLE_SET command requires
129 * two out scatterlists. Each contains a 4 byte count of entries followed
130 * by a concatenated byte stream of the ETH_ALEN MAC addresses. The
131 * first sg list contains unicast addresses, the second is for multicast.
132 * This functionality is present if the VIRTIO_NET_F_CTRL_RX feature
135 * The ADDR_SET command requests one out scatterlist, it contains a
136 * 6 bytes MAC address. This functionality is present if the
137 * VIRTIO_NET_F_CTRL_MAC_ADDR feature is available.
139 struct virtio_net_ctrl_mac {
141 uint8_t macs[][ETHER_ADDR_LEN];
142 } __attribute__((__packed__));
144 #define VIRTIO_NET_CTRL_MAC 1
145 #define VIRTIO_NET_CTRL_MAC_TABLE_SET 0
146 #define VIRTIO_NET_CTRL_MAC_ADDR_SET 1
149 * Control VLAN filtering
151 * The VLAN filter table is controlled via a simple ADD/DEL interface.
152 * VLAN IDs not added may be filtered by the hypervisor. Del is the
153 * opposite of add. Both commands expect an out entry containing a 2
154 * byte VLAN ID. VLAN filtering is available with the
155 * VIRTIO_NET_F_CTRL_VLAN feature bit.
157 #define VIRTIO_NET_CTRL_VLAN 2
158 #define VIRTIO_NET_CTRL_VLAN_ADD 0
159 #define VIRTIO_NET_CTRL_VLAN_DEL 1
161 struct virtio_net_ctrl_hdr {
164 } __attribute__((packed));
166 typedef uint8_t virtio_net_ctrl_ack;
168 #define VIRTIO_NET_OK 0
169 #define VIRTIO_NET_ERR 1
171 #define VIRTIO_MAX_CTRL_DATA 2048
173 struct virtio_pmd_ctrl {
174 struct virtio_net_ctrl_hdr hdr;
175 virtio_net_ctrl_ack status;
176 uint8_t data[VIRTIO_MAX_CTRL_DATA];
179 struct vq_desc_extra {
185 struct virtio_hw *hw; /**< virtio_hw structure pointer. */
186 struct vring vq_ring; /**< vring keeping desc, used and avail */
188 * Last consumed descriptor in the used table,
189 * trails vq_ring.used->idx.
191 uint16_t vq_used_cons_idx;
192 uint16_t vq_nentries; /**< vring desc numbers */
193 uint16_t vq_free_cnt; /**< num of desc available */
194 uint16_t vq_avail_idx; /**< sync until needed */
195 uint16_t vq_free_thresh; /**< free threshold */
197 void *vq_ring_virt_mem; /**< linear address of vring*/
198 unsigned int vq_ring_size;
201 struct virtnet_rx rxq;
202 struct virtnet_tx txq;
203 struct virtnet_ctl cq;
206 rte_iova_t vq_ring_mem; /**< physical address of vring,
207 * or virtual address for virtio_user. */
210 * Head of the free chain in the descriptor table. If
211 * there are no free descriptors, this will be set to
212 * VQ_RING_DESC_CHAIN_END.
214 uint16_t vq_desc_head_idx;
215 uint16_t vq_desc_tail_idx;
216 uint16_t vq_queue_index; /**< PCI queue index */
217 uint16_t offset; /**< relative offset to obtain addr in mbuf */
218 uint16_t *notify_addr;
219 struct rte_mbuf **sw_ring; /**< RX software ring. */
220 struct vq_desc_extra vq_descx[0];
223 /* If multiqueue is provided by host, then we suppport it. */
224 #define VIRTIO_NET_CTRL_MQ 4
225 #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET 0
226 #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MIN 1
227 #define VIRTIO_NET_CTRL_MQ_VQ_PAIRS_MAX 0x8000
229 #define VIRTIO_NET_CTRL_MAC_ADDR_SET 1
232 * This is the first element of the scatter-gather list. If you don't
233 * specify GSO or CSUM features, you can simply ignore the header.
235 struct virtio_net_hdr {
236 #define VIRTIO_NET_HDR_F_NEEDS_CSUM 1 /**< Use csum_start,csum_offset*/
237 #define VIRTIO_NET_HDR_F_DATA_VALID 2 /**< Checksum is valid */
239 #define VIRTIO_NET_HDR_GSO_NONE 0 /**< Not a GSO frame */
240 #define VIRTIO_NET_HDR_GSO_TCPV4 1 /**< GSO frame, IPv4 TCP (TSO) */
241 #define VIRTIO_NET_HDR_GSO_UDP 3 /**< GSO frame, IPv4 UDP (UFO) */
242 #define VIRTIO_NET_HDR_GSO_TCPV6 4 /**< GSO frame, IPv6 TCP */
243 #define VIRTIO_NET_HDR_GSO_ECN 0x80 /**< TCP has ECN set */
245 uint16_t hdr_len; /**< Ethernet + IP + tcp/udp hdrs */
246 uint16_t gso_size; /**< Bytes to append to hdr_len per frame */
247 uint16_t csum_start; /**< Position to start checksumming from */
248 uint16_t csum_offset; /**< Offset after that to place checksum */
252 * This is the version of the header to use when the MRG_RXBUF
253 * feature has been negotiated.
255 struct virtio_net_hdr_mrg_rxbuf {
256 struct virtio_net_hdr hdr;
257 uint16_t num_buffers; /**< Number of merged rx buffers */
260 /* Region reserved to allow for transmit header and indirect ring */
261 #define VIRTIO_MAX_TX_INDIRECT 8
262 struct virtio_tx_region {
263 struct virtio_net_hdr_mrg_rxbuf tx_hdr;
264 struct vring_desc tx_indir[VIRTIO_MAX_TX_INDIRECT]
265 __attribute__((__aligned__(16)));
268 /* Chain all the descriptors in the ring with an END */
270 vring_desc_init(struct vring_desc *dp, uint16_t n)
274 for (i = 0; i < n - 1; i++)
275 dp[i].next = (uint16_t)(i + 1);
276 dp[i].next = VQ_RING_DESC_CHAIN_END;
280 * Tell the backend not to interrupt us.
283 virtqueue_disable_intr(struct virtqueue *vq)
285 vq->vq_ring.avail->flags |= VRING_AVAIL_F_NO_INTERRUPT;
289 * Tell the backend to interrupt us.
292 virtqueue_enable_intr(struct virtqueue *vq)
294 vq->vq_ring.avail->flags &= (~VRING_AVAIL_F_NO_INTERRUPT);
298 * Dump virtqueue internal structures, for debug purpose only.
300 void virtqueue_dump(struct virtqueue *vq);
302 * Get all mbufs to be freed.
304 struct rte_mbuf *virtqueue_detatch_unused(struct virtqueue *vq);
306 /* Flush the elements in the used ring. */
307 void virtqueue_flush(struct virtqueue *vq);
310 virtqueue_full(const struct virtqueue *vq)
312 return vq->vq_free_cnt == 0;
315 #define VIRTQUEUE_NUSED(vq) ((uint16_t)((vq)->vq_ring.used->idx - (vq)->vq_used_cons_idx))
317 void vq_ring_free_chain(struct virtqueue *vq, uint16_t desc_idx);
320 vq_update_avail_idx(struct virtqueue *vq)
323 vq->vq_ring.avail->idx = vq->vq_avail_idx;
327 vq_update_avail_ring(struct virtqueue *vq, uint16_t desc_idx)
331 * Place the head of the descriptor chain into the next slot and make
332 * it usable to the host. The chain is made available now rather than
333 * deferring to virtqueue_notify() in the hopes that if the host is
334 * currently running on another CPU, we can keep it processing the new
337 avail_idx = (uint16_t)(vq->vq_avail_idx & (vq->vq_nentries - 1));
338 if (unlikely(vq->vq_ring.avail->ring[avail_idx] != desc_idx))
339 vq->vq_ring.avail->ring[avail_idx] = desc_idx;
344 virtqueue_kick_prepare(struct virtqueue *vq)
346 return !(vq->vq_ring.used->flags & VRING_USED_F_NO_NOTIFY);
350 virtqueue_notify(struct virtqueue *vq)
353 * Ensure updated avail->idx is visible to host.
354 * For virtio on IA, the notificaiton is through io port operation
355 * which is a serialization instruction itself.
357 VTPCI_OPS(vq->hw)->notify_queue(vq->hw, vq);
360 #ifdef RTE_LIBRTE_VIRTIO_DEBUG_DUMP
361 #define VIRTQUEUE_DUMP(vq) do { \
362 uint16_t used_idx, nused; \
363 used_idx = (vq)->vq_ring.used->idx; \
364 nused = (uint16_t)(used_idx - (vq)->vq_used_cons_idx); \
365 PMD_INIT_LOG(DEBUG, \
366 "VQ: - size=%d; free=%d; used=%d; desc_head_idx=%d;" \
367 " avail.idx=%d; used_cons_idx=%d; used.idx=%d;" \
368 " avail.flags=0x%x; used.flags=0x%x", \
369 (vq)->vq_nentries, (vq)->vq_free_cnt, nused, \
370 (vq)->vq_desc_head_idx, (vq)->vq_ring.avail->idx, \
371 (vq)->vq_used_cons_idx, (vq)->vq_ring.used->idx, \
372 (vq)->vq_ring.avail->flags, (vq)->vq_ring.used->flags); \
375 #define VIRTQUEUE_DUMP(vq) do { } while (0)
378 #endif /* _VIRTQUEUE_H_ */