4 * Copyright(c) 2010-2016 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.
40 #include <rte_ethdev.h>
41 #include <rte_memcpy.h>
42 #include <rte_string_fns.h>
43 #include <rte_memzone.h>
44 #include <rte_malloc.h>
45 #include <rte_atomic.h>
46 #include <rte_branch_prediction.h>
48 #include <rte_ether.h>
49 #include <rte_common.h>
50 #include <rte_errno.h>
52 #include <rte_memory.h>
56 #include "virtio_ethdev.h"
57 #include "virtio_pci.h"
58 #include "virtio_logs.h"
59 #include "virtqueue.h"
60 #include "virtio_rxtx.h"
62 static int eth_virtio_dev_uninit(struct rte_eth_dev *eth_dev);
63 static int virtio_dev_configure(struct rte_eth_dev *dev);
64 static int virtio_dev_start(struct rte_eth_dev *dev);
65 static void virtio_dev_stop(struct rte_eth_dev *dev);
66 static void virtio_dev_promiscuous_enable(struct rte_eth_dev *dev);
67 static void virtio_dev_promiscuous_disable(struct rte_eth_dev *dev);
68 static void virtio_dev_allmulticast_enable(struct rte_eth_dev *dev);
69 static void virtio_dev_allmulticast_disable(struct rte_eth_dev *dev);
70 static void virtio_dev_info_get(struct rte_eth_dev *dev,
71 struct rte_eth_dev_info *dev_info);
72 static int virtio_dev_link_update(struct rte_eth_dev *dev,
73 __rte_unused int wait_to_complete);
75 static void virtio_set_hwaddr(struct virtio_hw *hw);
76 static void virtio_get_hwaddr(struct virtio_hw *hw);
78 static void virtio_dev_stats_get(struct rte_eth_dev *dev,
79 struct rte_eth_stats *stats);
80 static int virtio_dev_xstats_get(struct rte_eth_dev *dev,
81 struct rte_eth_xstat *xstats, unsigned n);
82 static int virtio_dev_xstats_get_names(struct rte_eth_dev *dev,
83 struct rte_eth_xstat_name *xstats_names,
85 static void virtio_dev_stats_reset(struct rte_eth_dev *dev);
86 static void virtio_dev_free_mbufs(struct rte_eth_dev *dev);
87 static int virtio_vlan_filter_set(struct rte_eth_dev *dev,
88 uint16_t vlan_id, int on);
89 static void virtio_mac_addr_add(struct rte_eth_dev *dev,
90 struct ether_addr *mac_addr,
91 uint32_t index, uint32_t vmdq __rte_unused);
92 static void virtio_mac_addr_remove(struct rte_eth_dev *dev, uint32_t index);
93 static void virtio_mac_addr_set(struct rte_eth_dev *dev,
94 struct ether_addr *mac_addr);
96 static int virtio_dev_queue_stats_mapping_set(
97 __rte_unused struct rte_eth_dev *eth_dev,
98 __rte_unused uint16_t queue_id,
99 __rte_unused uint8_t stat_idx,
100 __rte_unused uint8_t is_rx);
103 * The set of PCI devices this driver supports
105 static const struct rte_pci_id pci_id_virtio_map[] = {
106 { RTE_PCI_DEVICE(VIRTIO_PCI_VENDORID, VIRTIO_PCI_LEGACY_DEVICEID_NET) },
107 { RTE_PCI_DEVICE(VIRTIO_PCI_VENDORID, VIRTIO_PCI_MODERN_DEVICEID_NET) },
108 { .vendor_id = 0, /* sentinel */ },
111 struct rte_virtio_xstats_name_off {
112 char name[RTE_ETH_XSTATS_NAME_SIZE];
116 /* [rt]x_qX_ is prepended to the name string here */
117 static const struct rte_virtio_xstats_name_off rte_virtio_rxq_stat_strings[] = {
118 {"good_packets", offsetof(struct virtnet_rx, stats.packets)},
119 {"good_bytes", offsetof(struct virtnet_rx, stats.bytes)},
120 {"errors", offsetof(struct virtnet_rx, stats.errors)},
121 {"multicast_packets", offsetof(struct virtnet_rx, stats.multicast)},
122 {"broadcast_packets", offsetof(struct virtnet_rx, stats.broadcast)},
123 {"undersize_packets", offsetof(struct virtnet_rx, stats.size_bins[0])},
124 {"size_64_packets", offsetof(struct virtnet_rx, stats.size_bins[1])},
125 {"size_65_127_packets", offsetof(struct virtnet_rx, stats.size_bins[2])},
126 {"size_128_255_packets", offsetof(struct virtnet_rx, stats.size_bins[3])},
127 {"size_256_511_packets", offsetof(struct virtnet_rx, stats.size_bins[4])},
128 {"size_512_1023_packets", offsetof(struct virtnet_rx, stats.size_bins[5])},
129 {"size_1024_1518_packets", offsetof(struct virtnet_rx, stats.size_bins[6])},
130 {"size_1519_max_packets", offsetof(struct virtnet_rx, stats.size_bins[7])},
133 /* [rt]x_qX_ is prepended to the name string here */
134 static const struct rte_virtio_xstats_name_off rte_virtio_txq_stat_strings[] = {
135 {"good_packets", offsetof(struct virtnet_tx, stats.packets)},
136 {"good_bytes", offsetof(struct virtnet_tx, stats.bytes)},
137 {"errors", offsetof(struct virtnet_tx, stats.errors)},
138 {"multicast_packets", offsetof(struct virtnet_tx, stats.multicast)},
139 {"broadcast_packets", offsetof(struct virtnet_tx, stats.broadcast)},
140 {"undersize_packets", offsetof(struct virtnet_tx, stats.size_bins[0])},
141 {"size_64_packets", offsetof(struct virtnet_tx, stats.size_bins[1])},
142 {"size_65_127_packets", offsetof(struct virtnet_tx, stats.size_bins[2])},
143 {"size_128_255_packets", offsetof(struct virtnet_tx, stats.size_bins[3])},
144 {"size_256_511_packets", offsetof(struct virtnet_tx, stats.size_bins[4])},
145 {"size_512_1023_packets", offsetof(struct virtnet_tx, stats.size_bins[5])},
146 {"size_1024_1518_packets", offsetof(struct virtnet_tx, stats.size_bins[6])},
147 {"size_1519_max_packets", offsetof(struct virtnet_tx, stats.size_bins[7])},
150 #define VIRTIO_NB_RXQ_XSTATS (sizeof(rte_virtio_rxq_stat_strings) / \
151 sizeof(rte_virtio_rxq_stat_strings[0]))
152 #define VIRTIO_NB_TXQ_XSTATS (sizeof(rte_virtio_txq_stat_strings) / \
153 sizeof(rte_virtio_txq_stat_strings[0]))
156 virtio_send_command(struct virtnet_ctl *cvq, struct virtio_pmd_ctrl *ctrl,
157 int *dlen, int pkt_num)
161 virtio_net_ctrl_ack status = ~0;
162 struct virtio_pmd_ctrl result;
163 struct virtqueue *vq;
165 ctrl->status = status;
167 if (!cvq || !cvq->vq) {
168 PMD_INIT_LOG(ERR, "Control queue is not supported.");
172 head = vq->vq_desc_head_idx;
174 PMD_INIT_LOG(DEBUG, "vq->vq_desc_head_idx = %d, status = %d, "
175 "vq->hw->cvq = %p vq = %p",
176 vq->vq_desc_head_idx, status, vq->hw->cvq, vq);
178 if ((vq->vq_free_cnt < ((uint32_t)pkt_num + 2)) || (pkt_num < 1))
181 memcpy(cvq->virtio_net_hdr_mz->addr, ctrl,
182 sizeof(struct virtio_pmd_ctrl));
185 * Format is enforced in qemu code:
186 * One TX packet for header;
187 * At least one TX packet per argument;
188 * One RX packet for ACK.
190 vq->vq_ring.desc[head].flags = VRING_DESC_F_NEXT;
191 vq->vq_ring.desc[head].addr = cvq->virtio_net_hdr_mem;
192 vq->vq_ring.desc[head].len = sizeof(struct virtio_net_ctrl_hdr);
194 i = vq->vq_ring.desc[head].next;
196 for (k = 0; k < pkt_num; k++) {
197 vq->vq_ring.desc[i].flags = VRING_DESC_F_NEXT;
198 vq->vq_ring.desc[i].addr = cvq->virtio_net_hdr_mem
199 + sizeof(struct virtio_net_ctrl_hdr)
200 + sizeof(ctrl->status) + sizeof(uint8_t)*sum;
201 vq->vq_ring.desc[i].len = dlen[k];
204 i = vq->vq_ring.desc[i].next;
207 vq->vq_ring.desc[i].flags = VRING_DESC_F_WRITE;
208 vq->vq_ring.desc[i].addr = cvq->virtio_net_hdr_mem
209 + sizeof(struct virtio_net_ctrl_hdr);
210 vq->vq_ring.desc[i].len = sizeof(ctrl->status);
213 vq->vq_desc_head_idx = vq->vq_ring.desc[i].next;
215 vq_update_avail_ring(vq, head);
216 vq_update_avail_idx(vq);
218 PMD_INIT_LOG(DEBUG, "vq->vq_queue_index = %d", vq->vq_queue_index);
220 virtqueue_notify(vq);
223 while (VIRTQUEUE_NUSED(vq) == 0) {
228 while (VIRTQUEUE_NUSED(vq)) {
229 uint32_t idx, desc_idx, used_idx;
230 struct vring_used_elem *uep;
232 used_idx = (uint32_t)(vq->vq_used_cons_idx
233 & (vq->vq_nentries - 1));
234 uep = &vq->vq_ring.used->ring[used_idx];
235 idx = (uint32_t) uep->id;
238 while (vq->vq_ring.desc[desc_idx].flags & VRING_DESC_F_NEXT) {
239 desc_idx = vq->vq_ring.desc[desc_idx].next;
243 vq->vq_ring.desc[desc_idx].next = vq->vq_desc_head_idx;
244 vq->vq_desc_head_idx = idx;
246 vq->vq_used_cons_idx++;
250 PMD_INIT_LOG(DEBUG, "vq->vq_free_cnt=%d\nvq->vq_desc_head_idx=%d",
251 vq->vq_free_cnt, vq->vq_desc_head_idx);
253 memcpy(&result, cvq->virtio_net_hdr_mz->addr,
254 sizeof(struct virtio_pmd_ctrl));
256 return result.status;
260 virtio_set_multiple_queues(struct rte_eth_dev *dev, uint16_t nb_queues)
262 struct virtio_hw *hw = dev->data->dev_private;
263 struct virtio_pmd_ctrl ctrl;
267 ctrl.hdr.class = VIRTIO_NET_CTRL_MQ;
268 ctrl.hdr.cmd = VIRTIO_NET_CTRL_MQ_VQ_PAIRS_SET;
269 memcpy(ctrl.data, &nb_queues, sizeof(uint16_t));
271 dlen[0] = sizeof(uint16_t);
273 ret = virtio_send_command(hw->cvq, &ctrl, dlen, 1);
275 PMD_INIT_LOG(ERR, "Multiqueue configured but send command "
276 "failed, this is too late now...");
284 virtio_dev_queue_release(struct virtqueue *vq)
286 struct virtio_hw *hw;
291 hw->vtpci_ops->del_queue(hw, vq);
293 rte_free(vq->sw_ring);
298 int virtio_dev_queue_setup(struct rte_eth_dev *dev,
301 uint16_t vtpci_queue_idx,
303 unsigned int socket_id,
306 char vq_name[VIRTQUEUE_MAX_NAME_SZ];
307 char vq_hdr_name[VIRTQUEUE_MAX_NAME_SZ];
308 const struct rte_memzone *mz = NULL, *hdr_mz = NULL;
309 unsigned int vq_size, size;
310 struct virtio_hw *hw = dev->data->dev_private;
311 struct virtnet_rx *rxvq = NULL;
312 struct virtnet_tx *txvq = NULL;
313 struct virtnet_ctl *cvq = NULL;
314 struct virtqueue *vq;
315 const char *queue_names[] = {"rvq", "txq", "cvq"};
316 size_t sz_vq, sz_q = 0, sz_hdr_mz = 0;
317 void *sw_ring = NULL;
320 PMD_INIT_LOG(DEBUG, "setting up queue: %u", vtpci_queue_idx);
323 * Read the virtqueue size from the Queue Size field
324 * Always power of 2 and if 0 virtqueue does not exist
326 vq_size = hw->vtpci_ops->get_queue_num(hw, vtpci_queue_idx);
327 PMD_INIT_LOG(DEBUG, "vq_size: %u nb_desc:%u", vq_size, nb_desc);
329 PMD_INIT_LOG(ERR, "virtqueue does not exist");
333 if (!rte_is_power_of_2(vq_size)) {
334 PMD_INIT_LOG(ERR, "virtqueue size is not powerof 2");
338 snprintf(vq_name, sizeof(vq_name), "port%d_%s%d",
339 dev->data->port_id, queue_names[queue_type], queue_idx);
341 sz_vq = RTE_ALIGN_CEIL(sizeof(*vq) +
342 vq_size * sizeof(struct vq_desc_extra),
343 RTE_CACHE_LINE_SIZE);
344 if (queue_type == VTNET_RQ) {
345 sz_q = sz_vq + sizeof(*rxvq);
346 } else if (queue_type == VTNET_TQ) {
347 sz_q = sz_vq + sizeof(*txvq);
349 * For each xmit packet, allocate a virtio_net_hdr
350 * and indirect ring elements
352 sz_hdr_mz = vq_size * sizeof(struct virtio_tx_region);
353 } else if (queue_type == VTNET_CQ) {
354 sz_q = sz_vq + sizeof(*cvq);
355 /* Allocate a page for control vq command, data and status */
356 sz_hdr_mz = PAGE_SIZE;
359 vq = rte_zmalloc_socket(vq_name, sz_q, RTE_CACHE_LINE_SIZE, socket_id);
361 PMD_INIT_LOG(ERR, "can not allocate vq");
365 vq->vq_queue_index = vtpci_queue_idx;
366 vq->vq_nentries = vq_size;
368 if (nb_desc == 0 || nb_desc > vq_size)
370 vq->vq_free_cnt = nb_desc;
373 * Reserve a memzone for vring elements
375 size = vring_size(vq_size, VIRTIO_PCI_VRING_ALIGN);
376 vq->vq_ring_size = RTE_ALIGN_CEIL(size, VIRTIO_PCI_VRING_ALIGN);
377 PMD_INIT_LOG(DEBUG, "vring_size: %d, rounded_vring_size: %d",
378 size, vq->vq_ring_size);
380 mz = rte_memzone_reserve_aligned(vq_name, vq->vq_ring_size, socket_id,
381 0, VIRTIO_PCI_VRING_ALIGN);
383 if (rte_errno == EEXIST)
384 mz = rte_memzone_lookup(vq_name);
391 memset(mz->addr, 0, sizeof(mz->len));
393 vq->vq_ring_mem = mz->phys_addr;
394 vq->vq_ring_virt_mem = mz->addr;
395 PMD_INIT_LOG(DEBUG, "vq->vq_ring_mem: 0x%" PRIx64,
396 (uint64_t)mz->phys_addr);
397 PMD_INIT_LOG(DEBUG, "vq->vq_ring_virt_mem: 0x%" PRIx64,
398 (uint64_t)(uintptr_t)mz->addr);
401 snprintf(vq_hdr_name, sizeof(vq_hdr_name), "port%d_%s%d_hdr",
402 dev->data->port_id, queue_names[queue_type],
404 hdr_mz = rte_memzone_reserve_aligned(vq_hdr_name, sz_hdr_mz,
406 RTE_CACHE_LINE_SIZE);
407 if (hdr_mz == NULL) {
408 if (rte_errno == EEXIST)
409 hdr_mz = rte_memzone_lookup(vq_hdr_name);
410 if (hdr_mz == NULL) {
417 if (queue_type == VTNET_RQ) {
418 size_t sz_sw = (RTE_PMD_VIRTIO_RX_MAX_BURST + vq_size) *
419 sizeof(vq->sw_ring[0]);
421 sw_ring = rte_zmalloc_socket("sw_ring", sz_sw,
422 RTE_CACHE_LINE_SIZE, socket_id);
424 PMD_INIT_LOG(ERR, "can not allocate RX soft ring");
429 vq->sw_ring = sw_ring;
430 rxvq = (struct virtnet_rx *)RTE_PTR_ADD(vq, sz_vq);
432 rxvq->port_id = dev->data->port_id;
433 rxvq->queue_id = queue_idx;
436 } else if (queue_type == VTNET_TQ) {
437 txvq = (struct virtnet_tx *)RTE_PTR_ADD(vq, sz_vq);
439 txvq->port_id = dev->data->port_id;
440 txvq->queue_id = queue_idx;
442 txvq->virtio_net_hdr_mz = hdr_mz;
443 txvq->virtio_net_hdr_mem = hdr_mz->phys_addr;
446 } else if (queue_type == VTNET_CQ) {
447 cvq = (struct virtnet_ctl *)RTE_PTR_ADD(vq, sz_vq);
450 cvq->virtio_net_hdr_mz = hdr_mz;
451 cvq->virtio_net_hdr_mem = hdr_mz->phys_addr;
452 memset(cvq->virtio_net_hdr_mz->addr, 0, PAGE_SIZE);
456 /* For virtio_user case (that is when dev->pci_dev is NULL), we use
457 * virtual address. And we need properly set _offset_, please see
458 * VIRTIO_MBUF_DATA_DMA_ADDR in virtqueue.h for more information.
461 vq->offset = offsetof(struct rte_mbuf, buf_physaddr);
463 vq->vq_ring_mem = (uintptr_t)mz->addr;
464 vq->offset = offsetof(struct rte_mbuf, buf_addr);
465 if (queue_type == VTNET_TQ)
466 txvq->virtio_net_hdr_mem = (uintptr_t)hdr_mz->addr;
467 else if (queue_type == VTNET_CQ)
468 cvq->virtio_net_hdr_mem = (uintptr_t)hdr_mz->addr;
471 if (queue_type == VTNET_TQ) {
472 struct virtio_tx_region *txr;
476 memset(txr, 0, vq_size * sizeof(*txr));
477 for (i = 0; i < vq_size; i++) {
478 struct vring_desc *start_dp = txr[i].tx_indir;
480 vring_desc_init(start_dp, RTE_DIM(txr[i].tx_indir));
482 /* first indirect descriptor is always the tx header */
483 start_dp->addr = txvq->virtio_net_hdr_mem
485 + offsetof(struct virtio_tx_region, tx_hdr);
487 start_dp->len = hw->vtnet_hdr_size;
488 start_dp->flags = VRING_DESC_F_NEXT;
492 if (hw->vtpci_ops->setup_queue(hw, vq) < 0) {
493 PMD_INIT_LOG(ERR, "setup_queue failed");
494 virtio_dev_queue_release(vq);
503 rte_memzone_free(hdr_mz);
504 rte_memzone_free(mz);
511 virtio_dev_cq_queue_setup(struct rte_eth_dev *dev, uint16_t vtpci_queue_idx,
514 struct virtnet_ctl *cvq;
516 struct virtio_hw *hw = dev->data->dev_private;
518 PMD_INIT_FUNC_TRACE();
519 ret = virtio_dev_queue_setup(dev, VTNET_CQ, VTNET_SQ_CQ_QUEUE_IDX,
520 vtpci_queue_idx, 0, socket_id, (void **)&cvq);
522 PMD_INIT_LOG(ERR, "control vq initialization failed");
531 virtio_free_queues(struct rte_eth_dev *dev)
535 for (i = 0; i < dev->data->nb_rx_queues; i++)
536 virtio_dev_rx_queue_release(dev->data->rx_queues[i]);
538 dev->data->nb_rx_queues = 0;
540 for (i = 0; i < dev->data->nb_tx_queues; i++)
541 virtio_dev_tx_queue_release(dev->data->tx_queues[i]);
543 dev->data->nb_tx_queues = 0;
547 virtio_dev_close(struct rte_eth_dev *dev)
549 struct virtio_hw *hw = dev->data->dev_private;
551 PMD_INIT_LOG(DEBUG, "virtio_dev_close");
553 if (hw->started == 1)
554 virtio_dev_stop(dev);
557 if (dev->data->dev_flags & RTE_ETH_DEV_INTR_LSC)
558 vtpci_irq_config(hw, VIRTIO_MSI_NO_VECTOR);
560 virtio_dev_free_mbufs(dev);
561 virtio_free_queues(dev);
565 virtio_dev_promiscuous_enable(struct rte_eth_dev *dev)
567 struct virtio_hw *hw = dev->data->dev_private;
568 struct virtio_pmd_ctrl ctrl;
572 if (!vtpci_with_feature(hw, VIRTIO_NET_F_CTRL_RX)) {
573 PMD_INIT_LOG(INFO, "host does not support rx control\n");
577 ctrl.hdr.class = VIRTIO_NET_CTRL_RX;
578 ctrl.hdr.cmd = VIRTIO_NET_CTRL_RX_PROMISC;
582 ret = virtio_send_command(hw->cvq, &ctrl, dlen, 1);
584 PMD_INIT_LOG(ERR, "Failed to enable promisc");
588 virtio_dev_promiscuous_disable(struct rte_eth_dev *dev)
590 struct virtio_hw *hw = dev->data->dev_private;
591 struct virtio_pmd_ctrl ctrl;
595 if (!vtpci_with_feature(hw, VIRTIO_NET_F_CTRL_RX)) {
596 PMD_INIT_LOG(INFO, "host does not support rx control\n");
600 ctrl.hdr.class = VIRTIO_NET_CTRL_RX;
601 ctrl.hdr.cmd = VIRTIO_NET_CTRL_RX_PROMISC;
605 ret = virtio_send_command(hw->cvq, &ctrl, dlen, 1);
607 PMD_INIT_LOG(ERR, "Failed to disable promisc");
611 virtio_dev_allmulticast_enable(struct rte_eth_dev *dev)
613 struct virtio_hw *hw = dev->data->dev_private;
614 struct virtio_pmd_ctrl ctrl;
618 if (!vtpci_with_feature(hw, VIRTIO_NET_F_CTRL_RX)) {
619 PMD_INIT_LOG(INFO, "host does not support rx control\n");
623 ctrl.hdr.class = VIRTIO_NET_CTRL_RX;
624 ctrl.hdr.cmd = VIRTIO_NET_CTRL_RX_ALLMULTI;
628 ret = virtio_send_command(hw->cvq, &ctrl, dlen, 1);
630 PMD_INIT_LOG(ERR, "Failed to enable allmulticast");
634 virtio_dev_allmulticast_disable(struct rte_eth_dev *dev)
636 struct virtio_hw *hw = dev->data->dev_private;
637 struct virtio_pmd_ctrl ctrl;
641 if (!vtpci_with_feature(hw, VIRTIO_NET_F_CTRL_RX)) {
642 PMD_INIT_LOG(INFO, "host does not support rx control\n");
646 ctrl.hdr.class = VIRTIO_NET_CTRL_RX;
647 ctrl.hdr.cmd = VIRTIO_NET_CTRL_RX_ALLMULTI;
651 ret = virtio_send_command(hw->cvq, &ctrl, dlen, 1);
653 PMD_INIT_LOG(ERR, "Failed to disable allmulticast");
656 #define VLAN_TAG_LEN 4 /* 802.3ac tag (not DMA'd) */
658 virtio_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
660 struct virtio_hw *hw = dev->data->dev_private;
661 uint32_t ether_hdr_len = ETHER_HDR_LEN + VLAN_TAG_LEN +
663 uint32_t frame_size = mtu + ether_hdr_len;
665 if (mtu < ETHER_MIN_MTU || frame_size > VIRTIO_MAX_RX_PKTLEN) {
666 PMD_INIT_LOG(ERR, "MTU should be between %d and %d\n",
667 ETHER_MIN_MTU, VIRTIO_MAX_RX_PKTLEN - ether_hdr_len);
674 * dev_ops for virtio, bare necessities for basic operation
676 static const struct eth_dev_ops virtio_eth_dev_ops = {
677 .dev_configure = virtio_dev_configure,
678 .dev_start = virtio_dev_start,
679 .dev_stop = virtio_dev_stop,
680 .dev_close = virtio_dev_close,
681 .promiscuous_enable = virtio_dev_promiscuous_enable,
682 .promiscuous_disable = virtio_dev_promiscuous_disable,
683 .allmulticast_enable = virtio_dev_allmulticast_enable,
684 .allmulticast_disable = virtio_dev_allmulticast_disable,
685 .mtu_set = virtio_mtu_set,
686 .dev_infos_get = virtio_dev_info_get,
687 .stats_get = virtio_dev_stats_get,
688 .xstats_get = virtio_dev_xstats_get,
689 .xstats_get_names = virtio_dev_xstats_get_names,
690 .stats_reset = virtio_dev_stats_reset,
691 .xstats_reset = virtio_dev_stats_reset,
692 .link_update = virtio_dev_link_update,
693 .rx_queue_setup = virtio_dev_rx_queue_setup,
694 .rx_queue_release = virtio_dev_rx_queue_release,
695 .tx_queue_setup = virtio_dev_tx_queue_setup,
696 .tx_queue_release = virtio_dev_tx_queue_release,
697 /* collect stats per queue */
698 .queue_stats_mapping_set = virtio_dev_queue_stats_mapping_set,
699 .vlan_filter_set = virtio_vlan_filter_set,
700 .mac_addr_add = virtio_mac_addr_add,
701 .mac_addr_remove = virtio_mac_addr_remove,
702 .mac_addr_set = virtio_mac_addr_set,
706 virtio_dev_atomic_read_link_status(struct rte_eth_dev *dev,
707 struct rte_eth_link *link)
709 struct rte_eth_link *dst = link;
710 struct rte_eth_link *src = &(dev->data->dev_link);
712 if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
713 *(uint64_t *)src) == 0)
720 * Atomically writes the link status information into global
721 * structure rte_eth_dev.
724 * - Pointer to the structure rte_eth_dev to read from.
725 * - Pointer to the buffer to be saved with the link status.
728 * - On success, zero.
729 * - On failure, negative value.
732 virtio_dev_atomic_write_link_status(struct rte_eth_dev *dev,
733 struct rte_eth_link *link)
735 struct rte_eth_link *dst = &(dev->data->dev_link);
736 struct rte_eth_link *src = link;
738 if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
739 *(uint64_t *)src) == 0)
746 virtio_update_stats(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
750 for (i = 0; i < dev->data->nb_tx_queues; i++) {
751 const struct virtnet_tx *txvq = dev->data->tx_queues[i];
755 stats->opackets += txvq->stats.packets;
756 stats->obytes += txvq->stats.bytes;
757 stats->oerrors += txvq->stats.errors;
759 if (i < RTE_ETHDEV_QUEUE_STAT_CNTRS) {
760 stats->q_opackets[i] = txvq->stats.packets;
761 stats->q_obytes[i] = txvq->stats.bytes;
765 for (i = 0; i < dev->data->nb_rx_queues; i++) {
766 const struct virtnet_rx *rxvq = dev->data->rx_queues[i];
770 stats->ipackets += rxvq->stats.packets;
771 stats->ibytes += rxvq->stats.bytes;
772 stats->ierrors += rxvq->stats.errors;
774 if (i < RTE_ETHDEV_QUEUE_STAT_CNTRS) {
775 stats->q_ipackets[i] = rxvq->stats.packets;
776 stats->q_ibytes[i] = rxvq->stats.bytes;
780 stats->rx_nombuf = dev->data->rx_mbuf_alloc_failed;
783 static int virtio_dev_xstats_get_names(struct rte_eth_dev *dev,
784 struct rte_eth_xstat_name *xstats_names,
785 __rte_unused unsigned limit)
791 unsigned nstats = dev->data->nb_tx_queues * VIRTIO_NB_TXQ_XSTATS +
792 dev->data->nb_rx_queues * VIRTIO_NB_RXQ_XSTATS;
794 if (xstats_names != NULL) {
795 /* Note: limit checked in rte_eth_xstats_names() */
797 for (i = 0; i < dev->data->nb_rx_queues; i++) {
798 struct virtqueue *rxvq = dev->data->rx_queues[i];
801 for (t = 0; t < VIRTIO_NB_RXQ_XSTATS; t++) {
802 snprintf(xstats_names[count].name,
803 sizeof(xstats_names[count].name),
805 rte_virtio_rxq_stat_strings[t].name);
810 for (i = 0; i < dev->data->nb_tx_queues; i++) {
811 struct virtqueue *txvq = dev->data->tx_queues[i];
814 for (t = 0; t < VIRTIO_NB_TXQ_XSTATS; t++) {
815 snprintf(xstats_names[count].name,
816 sizeof(xstats_names[count].name),
818 rte_virtio_txq_stat_strings[t].name);
828 virtio_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
834 unsigned nstats = dev->data->nb_tx_queues * VIRTIO_NB_TXQ_XSTATS +
835 dev->data->nb_rx_queues * VIRTIO_NB_RXQ_XSTATS;
840 for (i = 0; i < dev->data->nb_rx_queues; i++) {
841 struct virtnet_rx *rxvq = dev->data->rx_queues[i];
848 for (t = 0; t < VIRTIO_NB_RXQ_XSTATS; t++) {
849 xstats[count].value = *(uint64_t *)(((char *)rxvq) +
850 rte_virtio_rxq_stat_strings[t].offset);
855 for (i = 0; i < dev->data->nb_tx_queues; i++) {
856 struct virtnet_tx *txvq = dev->data->tx_queues[i];
863 for (t = 0; t < VIRTIO_NB_TXQ_XSTATS; t++) {
864 xstats[count].value = *(uint64_t *)(((char *)txvq) +
865 rte_virtio_txq_stat_strings[t].offset);
874 virtio_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
876 virtio_update_stats(dev, stats);
880 virtio_dev_stats_reset(struct rte_eth_dev *dev)
884 for (i = 0; i < dev->data->nb_tx_queues; i++) {
885 struct virtnet_tx *txvq = dev->data->tx_queues[i];
889 txvq->stats.packets = 0;
890 txvq->stats.bytes = 0;
891 txvq->stats.errors = 0;
892 txvq->stats.multicast = 0;
893 txvq->stats.broadcast = 0;
894 memset(txvq->stats.size_bins, 0,
895 sizeof(txvq->stats.size_bins[0]) * 8);
898 for (i = 0; i < dev->data->nb_rx_queues; i++) {
899 struct virtnet_rx *rxvq = dev->data->rx_queues[i];
903 rxvq->stats.packets = 0;
904 rxvq->stats.bytes = 0;
905 rxvq->stats.errors = 0;
906 rxvq->stats.multicast = 0;
907 rxvq->stats.broadcast = 0;
908 memset(rxvq->stats.size_bins, 0,
909 sizeof(rxvq->stats.size_bins[0]) * 8);
914 virtio_set_hwaddr(struct virtio_hw *hw)
916 vtpci_write_dev_config(hw,
917 offsetof(struct virtio_net_config, mac),
918 &hw->mac_addr, ETHER_ADDR_LEN);
922 virtio_get_hwaddr(struct virtio_hw *hw)
924 if (vtpci_with_feature(hw, VIRTIO_NET_F_MAC)) {
925 vtpci_read_dev_config(hw,
926 offsetof(struct virtio_net_config, mac),
927 &hw->mac_addr, ETHER_ADDR_LEN);
929 eth_random_addr(&hw->mac_addr[0]);
930 virtio_set_hwaddr(hw);
935 virtio_mac_table_set(struct virtio_hw *hw,
936 const struct virtio_net_ctrl_mac *uc,
937 const struct virtio_net_ctrl_mac *mc)
939 struct virtio_pmd_ctrl ctrl;
942 if (!vtpci_with_feature(hw, VIRTIO_NET_F_CTRL_MAC_ADDR)) {
943 PMD_DRV_LOG(INFO, "host does not support mac table");
947 ctrl.hdr.class = VIRTIO_NET_CTRL_MAC;
948 ctrl.hdr.cmd = VIRTIO_NET_CTRL_MAC_TABLE_SET;
950 len[0] = uc->entries * ETHER_ADDR_LEN + sizeof(uc->entries);
951 memcpy(ctrl.data, uc, len[0]);
953 len[1] = mc->entries * ETHER_ADDR_LEN + sizeof(mc->entries);
954 memcpy(ctrl.data + len[0], mc, len[1]);
956 err = virtio_send_command(hw->cvq, &ctrl, len, 2);
958 PMD_DRV_LOG(NOTICE, "mac table set failed: %d", err);
962 virtio_mac_addr_add(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
963 uint32_t index, uint32_t vmdq __rte_unused)
965 struct virtio_hw *hw = dev->data->dev_private;
966 const struct ether_addr *addrs = dev->data->mac_addrs;
968 struct virtio_net_ctrl_mac *uc, *mc;
970 if (index >= VIRTIO_MAX_MAC_ADDRS) {
971 PMD_DRV_LOG(ERR, "mac address index %u out of range", index);
975 uc = alloca(VIRTIO_MAX_MAC_ADDRS * ETHER_ADDR_LEN + sizeof(uc->entries));
977 mc = alloca(VIRTIO_MAX_MAC_ADDRS * ETHER_ADDR_LEN + sizeof(mc->entries));
980 for (i = 0; i < VIRTIO_MAX_MAC_ADDRS; i++) {
981 const struct ether_addr *addr
982 = (i == index) ? mac_addr : addrs + i;
983 struct virtio_net_ctrl_mac *tbl
984 = is_multicast_ether_addr(addr) ? mc : uc;
986 memcpy(&tbl->macs[tbl->entries++], addr, ETHER_ADDR_LEN);
989 virtio_mac_table_set(hw, uc, mc);
993 virtio_mac_addr_remove(struct rte_eth_dev *dev, uint32_t index)
995 struct virtio_hw *hw = dev->data->dev_private;
996 struct ether_addr *addrs = dev->data->mac_addrs;
997 struct virtio_net_ctrl_mac *uc, *mc;
1000 if (index >= VIRTIO_MAX_MAC_ADDRS) {
1001 PMD_DRV_LOG(ERR, "mac address index %u out of range", index);
1005 uc = alloca(VIRTIO_MAX_MAC_ADDRS * ETHER_ADDR_LEN + sizeof(uc->entries));
1007 mc = alloca(VIRTIO_MAX_MAC_ADDRS * ETHER_ADDR_LEN + sizeof(mc->entries));
1010 for (i = 0; i < VIRTIO_MAX_MAC_ADDRS; i++) {
1011 struct virtio_net_ctrl_mac *tbl;
1013 if (i == index || is_zero_ether_addr(addrs + i))
1016 tbl = is_multicast_ether_addr(addrs + i) ? mc : uc;
1017 memcpy(&tbl->macs[tbl->entries++], addrs + i, ETHER_ADDR_LEN);
1020 virtio_mac_table_set(hw, uc, mc);
1024 virtio_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr)
1026 struct virtio_hw *hw = dev->data->dev_private;
1028 memcpy(hw->mac_addr, mac_addr, ETHER_ADDR_LEN);
1030 /* Use atomic update if available */
1031 if (vtpci_with_feature(hw, VIRTIO_NET_F_CTRL_MAC_ADDR)) {
1032 struct virtio_pmd_ctrl ctrl;
1033 int len = ETHER_ADDR_LEN;
1035 ctrl.hdr.class = VIRTIO_NET_CTRL_MAC;
1036 ctrl.hdr.cmd = VIRTIO_NET_CTRL_MAC_ADDR_SET;
1038 memcpy(ctrl.data, mac_addr, ETHER_ADDR_LEN);
1039 virtio_send_command(hw->cvq, &ctrl, &len, 1);
1040 } else if (vtpci_with_feature(hw, VIRTIO_NET_F_MAC))
1041 virtio_set_hwaddr(hw);
1045 virtio_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1047 struct virtio_hw *hw = dev->data->dev_private;
1048 struct virtio_pmd_ctrl ctrl;
1051 if (!vtpci_with_feature(hw, VIRTIO_NET_F_CTRL_VLAN))
1054 ctrl.hdr.class = VIRTIO_NET_CTRL_VLAN;
1055 ctrl.hdr.cmd = on ? VIRTIO_NET_CTRL_VLAN_ADD : VIRTIO_NET_CTRL_VLAN_DEL;
1056 memcpy(ctrl.data, &vlan_id, sizeof(vlan_id));
1057 len = sizeof(vlan_id);
1059 return virtio_send_command(hw->cvq, &ctrl, &len, 1);
1063 virtio_negotiate_features(struct virtio_hw *hw)
1065 uint64_t host_features;
1067 /* Prepare guest_features: feature that driver wants to support */
1068 hw->guest_features = VIRTIO_PMD_GUEST_FEATURES;
1069 PMD_INIT_LOG(DEBUG, "guest_features before negotiate = %" PRIx64,
1070 hw->guest_features);
1072 /* Read device(host) feature bits */
1073 host_features = hw->vtpci_ops->get_features(hw);
1074 PMD_INIT_LOG(DEBUG, "host_features before negotiate = %" PRIx64,
1078 * Negotiate features: Subset of device feature bits are written back
1079 * guest feature bits.
1081 hw->guest_features = vtpci_negotiate_features(hw, host_features);
1082 PMD_INIT_LOG(DEBUG, "features after negotiate = %" PRIx64,
1083 hw->guest_features);
1086 if (!vtpci_with_feature(hw, VIRTIO_F_VERSION_1)) {
1088 "VIRTIO_F_VERSION_1 features is not enabled.");
1091 vtpci_set_status(hw, VIRTIO_CONFIG_STATUS_FEATURES_OK);
1092 if (!(vtpci_get_status(hw) & VIRTIO_CONFIG_STATUS_FEATURES_OK)) {
1094 "failed to set FEATURES_OK status!");
1103 * Process Virtio Config changed interrupt and call the callback
1104 * if link state changed.
1107 virtio_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
1110 struct rte_eth_dev *dev = param;
1111 struct virtio_hw *hw = dev->data->dev_private;
1114 /* Read interrupt status which clears interrupt */
1115 isr = vtpci_isr(hw);
1116 PMD_DRV_LOG(INFO, "interrupt status = %#x", isr);
1118 if (rte_intr_enable(&dev->pci_dev->intr_handle) < 0)
1119 PMD_DRV_LOG(ERR, "interrupt enable failed");
1121 if (isr & VIRTIO_PCI_ISR_CONFIG) {
1122 if (virtio_dev_link_update(dev, 0) == 0)
1123 _rte_eth_dev_callback_process(dev,
1124 RTE_ETH_EVENT_INTR_LSC);
1130 rx_func_get(struct rte_eth_dev *eth_dev)
1132 struct virtio_hw *hw = eth_dev->data->dev_private;
1133 if (vtpci_with_feature(hw, VIRTIO_NET_F_MRG_RXBUF))
1134 eth_dev->rx_pkt_burst = &virtio_recv_mergeable_pkts;
1136 eth_dev->rx_pkt_burst = &virtio_recv_pkts;
1140 virtio_init_device(struct rte_eth_dev *eth_dev)
1142 struct virtio_hw *hw = eth_dev->data->dev_private;
1143 struct virtio_net_config *config;
1144 struct virtio_net_config local_config;
1145 struct rte_pci_device *pci_dev = eth_dev->pci_dev;
1147 /* Reset the device although not necessary at startup */
1150 /* Tell the host we've noticed this device. */
1151 vtpci_set_status(hw, VIRTIO_CONFIG_STATUS_ACK);
1153 /* Tell the host we've known how to drive the device. */
1154 vtpci_set_status(hw, VIRTIO_CONFIG_STATUS_DRIVER);
1155 if (virtio_negotiate_features(hw) < 0)
1158 /* If host does not support status then disable LSC */
1159 if (!vtpci_with_feature(hw, VIRTIO_NET_F_STATUS))
1160 eth_dev->data->dev_flags &= ~RTE_ETH_DEV_INTR_LSC;
1162 eth_dev->data->dev_flags |= RTE_ETH_DEV_INTR_LSC;
1164 rte_eth_copy_pci_info(eth_dev, pci_dev);
1166 rx_func_get(eth_dev);
1168 /* Setting up rx_header size for the device */
1169 if (vtpci_with_feature(hw, VIRTIO_NET_F_MRG_RXBUF) ||
1170 vtpci_with_feature(hw, VIRTIO_F_VERSION_1))
1171 hw->vtnet_hdr_size = sizeof(struct virtio_net_hdr_mrg_rxbuf);
1173 hw->vtnet_hdr_size = sizeof(struct virtio_net_hdr);
1175 /* Copy the permanent MAC address to: virtio_hw */
1176 virtio_get_hwaddr(hw);
1177 ether_addr_copy((struct ether_addr *) hw->mac_addr,
1178 ð_dev->data->mac_addrs[0]);
1180 "PORT MAC: %02X:%02X:%02X:%02X:%02X:%02X",
1181 hw->mac_addr[0], hw->mac_addr[1], hw->mac_addr[2],
1182 hw->mac_addr[3], hw->mac_addr[4], hw->mac_addr[5]);
1184 if (vtpci_with_feature(hw, VIRTIO_NET_F_CTRL_VQ)) {
1185 config = &local_config;
1187 vtpci_read_dev_config(hw,
1188 offsetof(struct virtio_net_config, mac),
1189 &config->mac, sizeof(config->mac));
1191 if (vtpci_with_feature(hw, VIRTIO_NET_F_STATUS)) {
1192 vtpci_read_dev_config(hw,
1193 offsetof(struct virtio_net_config, status),
1194 &config->status, sizeof(config->status));
1197 "VIRTIO_NET_F_STATUS is not supported");
1201 if (vtpci_with_feature(hw, VIRTIO_NET_F_MQ)) {
1202 vtpci_read_dev_config(hw,
1203 offsetof(struct virtio_net_config, max_virtqueue_pairs),
1204 &config->max_virtqueue_pairs,
1205 sizeof(config->max_virtqueue_pairs));
1208 "VIRTIO_NET_F_MQ is not supported");
1209 config->max_virtqueue_pairs = 1;
1213 (VIRTIO_MAX_RX_QUEUES < config->max_virtqueue_pairs) ?
1214 VIRTIO_MAX_RX_QUEUES : config->max_virtqueue_pairs;
1216 (VIRTIO_MAX_TX_QUEUES < config->max_virtqueue_pairs) ?
1217 VIRTIO_MAX_TX_QUEUES : config->max_virtqueue_pairs;
1219 virtio_dev_cq_queue_setup(eth_dev,
1220 config->max_virtqueue_pairs * 2,
1223 PMD_INIT_LOG(DEBUG, "config->max_virtqueue_pairs=%d",
1224 config->max_virtqueue_pairs);
1225 PMD_INIT_LOG(DEBUG, "config->status=%d", config->status);
1227 "PORT MAC: %02X:%02X:%02X:%02X:%02X:%02X",
1228 config->mac[0], config->mac[1],
1229 config->mac[2], config->mac[3],
1230 config->mac[4], config->mac[5]);
1232 hw->max_rx_queues = 1;
1233 hw->max_tx_queues = 1;
1236 PMD_INIT_LOG(DEBUG, "hw->max_rx_queues=%d hw->max_tx_queues=%d",
1237 hw->max_rx_queues, hw->max_tx_queues);
1239 PMD_INIT_LOG(DEBUG, "port %d vendorID=0x%x deviceID=0x%x",
1240 eth_dev->data->port_id, pci_dev->id.vendor_id,
1241 pci_dev->id.device_id);
1243 virtio_dev_cq_start(eth_dev);
1249 * This function is based on probe() function in virtio_pci.c
1250 * It returns 0 on success.
1253 eth_virtio_dev_init(struct rte_eth_dev *eth_dev)
1255 struct virtio_hw *hw = eth_dev->data->dev_private;
1256 struct rte_pci_device *pci_dev;
1257 uint32_t dev_flags = RTE_ETH_DEV_DETACHABLE;
1260 RTE_BUILD_BUG_ON(RTE_PKTMBUF_HEADROOM < sizeof(struct virtio_net_hdr_mrg_rxbuf));
1262 eth_dev->dev_ops = &virtio_eth_dev_ops;
1263 eth_dev->tx_pkt_burst = &virtio_xmit_pkts;
1265 if (rte_eal_process_type() == RTE_PROC_SECONDARY) {
1266 rx_func_get(eth_dev);
1270 /* Allocate memory for storing MAC addresses */
1271 eth_dev->data->mac_addrs = rte_zmalloc("virtio", VIRTIO_MAX_MAC_ADDRS * ETHER_ADDR_LEN, 0);
1272 if (eth_dev->data->mac_addrs == NULL) {
1274 "Failed to allocate %d bytes needed to store MAC addresses",
1275 VIRTIO_MAX_MAC_ADDRS * ETHER_ADDR_LEN);
1279 pci_dev = eth_dev->pci_dev;
1282 ret = vtpci_init(pci_dev, hw, &dev_flags);
1287 eth_dev->data->dev_flags = dev_flags;
1289 /* reset device and negotiate features */
1290 ret = virtio_init_device(eth_dev);
1294 /* Setup interrupt callback */
1295 if (eth_dev->data->dev_flags & RTE_ETH_DEV_INTR_LSC)
1296 rte_intr_callback_register(&pci_dev->intr_handle,
1297 virtio_interrupt_handler, eth_dev);
1303 eth_virtio_dev_uninit(struct rte_eth_dev *eth_dev)
1305 struct rte_pci_device *pci_dev;
1306 struct virtio_hw *hw = eth_dev->data->dev_private;
1308 PMD_INIT_FUNC_TRACE();
1310 if (rte_eal_process_type() == RTE_PROC_SECONDARY)
1313 /* Close it anyway since there's no way to know if closed */
1314 virtio_dev_close(eth_dev);
1316 pci_dev = eth_dev->pci_dev;
1318 eth_dev->dev_ops = NULL;
1319 eth_dev->tx_pkt_burst = NULL;
1320 eth_dev->rx_pkt_burst = NULL;
1323 virtio_dev_queue_release(hw->cvq->vq);
1325 rte_free(eth_dev->data->mac_addrs);
1326 eth_dev->data->mac_addrs = NULL;
1328 /* reset interrupt callback */
1329 if (eth_dev->data->dev_flags & RTE_ETH_DEV_INTR_LSC)
1330 rte_intr_callback_unregister(&pci_dev->intr_handle,
1331 virtio_interrupt_handler,
1333 rte_eal_pci_unmap_device(pci_dev);
1335 PMD_INIT_LOG(DEBUG, "dev_uninit completed");
1340 static struct eth_driver rte_virtio_pmd = {
1343 .name = "net_virtio",
1345 .id_table = pci_id_virtio_map,
1346 .drv_flags = RTE_PCI_DRV_DETACHABLE,
1347 .probe = rte_eth_dev_pci_probe,
1348 .remove = rte_eth_dev_pci_remove,
1350 .eth_dev_init = eth_virtio_dev_init,
1351 .eth_dev_uninit = eth_virtio_dev_uninit,
1352 .dev_private_size = sizeof(struct virtio_hw),
1355 RTE_INIT(rte_virtio_pmd_init);
1357 rte_virtio_pmd_init(void)
1359 if (rte_eal_iopl_init() != 0) {
1360 PMD_INIT_LOG(ERR, "IOPL call failed - cannot use virtio PMD");
1364 rte_eal_pci_register(&rte_virtio_pmd.pci_drv);
1368 * Configure virtio device
1369 * It returns 0 on success.
1372 virtio_dev_configure(struct rte_eth_dev *dev)
1374 const struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
1375 struct virtio_hw *hw = dev->data->dev_private;
1377 PMD_INIT_LOG(DEBUG, "configure");
1379 if (rxmode->hw_ip_checksum) {
1380 PMD_DRV_LOG(ERR, "HW IP checksum not supported");
1384 hw->vlan_strip = rxmode->hw_vlan_strip;
1386 if (rxmode->hw_vlan_filter
1387 && !vtpci_with_feature(hw, VIRTIO_NET_F_CTRL_VLAN)) {
1389 "vlan filtering not available on this host");
1393 if (dev->data->dev_flags & RTE_ETH_DEV_INTR_LSC)
1394 if (vtpci_irq_config(hw, 0) == VIRTIO_MSI_NO_VECTOR) {
1395 PMD_DRV_LOG(ERR, "failed to set config vector");
1404 virtio_dev_start(struct rte_eth_dev *dev)
1406 uint16_t nb_queues, i;
1407 struct virtio_hw *hw = dev->data->dev_private;
1408 struct virtnet_rx *rxvq;
1409 struct virtnet_tx *txvq __rte_unused;
1411 /* check if lsc interrupt feature is enabled */
1412 if (dev->data->dev_conf.intr_conf.lsc) {
1413 if (!(dev->data->dev_flags & RTE_ETH_DEV_INTR_LSC)) {
1414 PMD_DRV_LOG(ERR, "link status not supported by host");
1418 if (rte_intr_enable(&dev->pci_dev->intr_handle) < 0) {
1419 PMD_DRV_LOG(ERR, "interrupt enable failed");
1424 /* Initialize Link state */
1425 virtio_dev_link_update(dev, 0);
1427 /* On restart after stop do not touch queues */
1431 /* Do final configuration before rx/tx engine starts */
1432 virtio_dev_rxtx_start(dev);
1433 vtpci_reinit_complete(hw);
1437 /*Notify the backend
1438 *Otherwise the tap backend might already stop its queue due to fullness.
1439 *vhost backend will have no chance to be waked up
1441 nb_queues = dev->data->nb_rx_queues;
1442 if (nb_queues > 1) {
1443 if (virtio_set_multiple_queues(dev, nb_queues) != 0)
1447 PMD_INIT_LOG(DEBUG, "nb_queues=%d", nb_queues);
1449 for (i = 0; i < nb_queues; i++) {
1450 rxvq = dev->data->rx_queues[i];
1451 virtqueue_notify(rxvq->vq);
1454 PMD_INIT_LOG(DEBUG, "Notified backend at initialization");
1456 for (i = 0; i < dev->data->nb_rx_queues; i++) {
1457 rxvq = dev->data->rx_queues[i];
1458 VIRTQUEUE_DUMP(rxvq->vq);
1461 for (i = 0; i < dev->data->nb_tx_queues; i++) {
1462 txvq = dev->data->tx_queues[i];
1463 VIRTQUEUE_DUMP(txvq->vq);
1469 static void virtio_dev_free_mbufs(struct rte_eth_dev *dev)
1471 struct rte_mbuf *buf;
1472 int i, mbuf_num = 0;
1474 for (i = 0; i < dev->data->nb_rx_queues; i++) {
1475 struct virtnet_rx *rxvq = dev->data->rx_queues[i];
1478 "Before freeing rxq[%d] used and unused buf", i);
1479 VIRTQUEUE_DUMP(rxvq->vq);
1481 PMD_INIT_LOG(DEBUG, "rx_queues[%d]=%p", i, rxvq);
1482 while ((buf = virtqueue_detatch_unused(rxvq->vq)) != NULL) {
1483 rte_pktmbuf_free(buf);
1487 PMD_INIT_LOG(DEBUG, "free %d mbufs", mbuf_num);
1489 "After freeing rxq[%d] used and unused buf", i);
1490 VIRTQUEUE_DUMP(rxvq->vq);
1493 for (i = 0; i < dev->data->nb_tx_queues; i++) {
1494 struct virtnet_tx *txvq = dev->data->tx_queues[i];
1497 "Before freeing txq[%d] used and unused bufs",
1499 VIRTQUEUE_DUMP(txvq->vq);
1502 while ((buf = virtqueue_detatch_unused(txvq->vq)) != NULL) {
1503 rte_pktmbuf_free(buf);
1507 PMD_INIT_LOG(DEBUG, "free %d mbufs", mbuf_num);
1509 "After freeing txq[%d] used and unused buf", i);
1510 VIRTQUEUE_DUMP(txvq->vq);
1515 * Stop device: disable interrupt and mark link down
1518 virtio_dev_stop(struct rte_eth_dev *dev)
1520 struct rte_eth_link link;
1521 struct virtio_hw *hw = dev->data->dev_private;
1523 PMD_INIT_LOG(DEBUG, "stop");
1527 if (dev->data->dev_conf.intr_conf.lsc)
1528 rte_intr_disable(&dev->pci_dev->intr_handle);
1530 memset(&link, 0, sizeof(link));
1531 virtio_dev_atomic_write_link_status(dev, &link);
1535 virtio_dev_link_update(struct rte_eth_dev *dev, __rte_unused int wait_to_complete)
1537 struct rte_eth_link link, old;
1539 struct virtio_hw *hw = dev->data->dev_private;
1540 memset(&link, 0, sizeof(link));
1541 virtio_dev_atomic_read_link_status(dev, &link);
1543 link.link_duplex = ETH_LINK_FULL_DUPLEX;
1544 link.link_speed = SPEED_10G;
1546 if (vtpci_with_feature(hw, VIRTIO_NET_F_STATUS)) {
1547 PMD_INIT_LOG(DEBUG, "Get link status from hw");
1548 vtpci_read_dev_config(hw,
1549 offsetof(struct virtio_net_config, status),
1550 &status, sizeof(status));
1551 if ((status & VIRTIO_NET_S_LINK_UP) == 0) {
1552 link.link_status = ETH_LINK_DOWN;
1553 PMD_INIT_LOG(DEBUG, "Port %d is down",
1554 dev->data->port_id);
1556 link.link_status = ETH_LINK_UP;
1557 PMD_INIT_LOG(DEBUG, "Port %d is up",
1558 dev->data->port_id);
1561 link.link_status = ETH_LINK_UP;
1563 virtio_dev_atomic_write_link_status(dev, &link);
1565 return (old.link_status == link.link_status) ? -1 : 0;
1569 virtio_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1571 struct virtio_hw *hw = dev->data->dev_private;
1574 dev_info->driver_name = dev->driver->pci_drv.driver.name;
1576 dev_info->driver_name = "virtio_user PMD";
1577 dev_info->max_rx_queues = (uint16_t)hw->max_rx_queues;
1578 dev_info->max_tx_queues = (uint16_t)hw->max_tx_queues;
1579 dev_info->min_rx_bufsize = VIRTIO_MIN_RX_BUFSIZE;
1580 dev_info->max_rx_pktlen = VIRTIO_MAX_RX_PKTLEN;
1581 dev_info->max_mac_addrs = VIRTIO_MAX_MAC_ADDRS;
1582 dev_info->default_txconf = (struct rte_eth_txconf) {
1583 .txq_flags = ETH_TXQ_FLAGS_NOOFFLOADS
1588 * It enables testpmd to collect per queue stats.
1591 virtio_dev_queue_stats_mapping_set(__rte_unused struct rte_eth_dev *eth_dev,
1592 __rte_unused uint16_t queue_id, __rte_unused uint8_t stat_idx,
1593 __rte_unused uint8_t is_rx)
1598 DRIVER_EXPORT_NAME(net_virtio, __COUNTER__);
1599 DRIVER_REGISTER_PCI_TABLE(net_virtio, pci_id_virtio_map);