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.
35 #include <sys/types.h>
38 #include <sys/types.h>
39 #include <sys/socket.h>
41 #include <rte_malloc.h>
42 #include <rte_kvargs.h>
43 #include <rte_ethdev_vdev.h>
45 #include <rte_alarm.h>
47 #include "virtio_ethdev.h"
48 #include "virtio_logs.h"
49 #include "virtio_pci.h"
50 #include "virtqueue.h"
51 #include "virtio_rxtx.h"
52 #include "virtio_user/virtio_user_dev.h"
54 #define virtio_user_get_dev(hw) \
55 ((struct virtio_user_dev *)(hw)->virtio_user_dev)
58 virtio_user_delayed_handler(void *param)
60 struct virtio_hw *hw = (struct virtio_hw *)param;
61 struct rte_eth_dev *dev = &rte_eth_devices[hw->port_id];
63 rte_intr_callback_unregister(dev->intr_handle,
64 virtio_interrupt_handler,
69 virtio_user_read_dev_config(struct virtio_hw *hw, size_t offset,
70 void *dst, int length)
73 struct virtio_user_dev *dev = virtio_user_get_dev(hw);
75 if (offset == offsetof(struct virtio_net_config, mac) &&
76 length == ETHER_ADDR_LEN) {
77 for (i = 0; i < ETHER_ADDR_LEN; ++i)
78 ((uint8_t *)dst)[i] = dev->mac_addr[i];
82 if (offset == offsetof(struct virtio_net_config, status)) {
85 if (dev->vhostfd >= 0) {
89 flags = fcntl(dev->vhostfd, F_GETFL);
90 fcntl(dev->vhostfd, F_SETFL, flags | O_NONBLOCK);
91 r = recv(dev->vhostfd, buf, 128, MSG_PEEK);
92 if (r == 0 || (r < 0 && errno != EAGAIN)) {
93 dev->status &= (~VIRTIO_NET_S_LINK_UP);
94 PMD_DRV_LOG(ERR, "virtio-user port %u is down",
96 /* Only client mode is available now. Once the
97 * connection is broken, it can never be up
98 * again. Besides, this function could be called
99 * in the process of interrupt handling,
100 * callback cannot be unregistered here, set an
104 virtio_user_delayed_handler,
107 dev->status |= VIRTIO_NET_S_LINK_UP;
109 fcntl(dev->vhostfd, F_SETFL, flags & (~O_NONBLOCK));
111 *(uint16_t *)dst = dev->status;
114 if (offset == offsetof(struct virtio_net_config, max_virtqueue_pairs))
115 *(uint16_t *)dst = dev->max_queue_pairs;
119 virtio_user_write_dev_config(struct virtio_hw *hw, size_t offset,
120 const void *src, int length)
123 struct virtio_user_dev *dev = virtio_user_get_dev(hw);
125 if ((offset == offsetof(struct virtio_net_config, mac)) &&
126 (length == ETHER_ADDR_LEN))
127 for (i = 0; i < ETHER_ADDR_LEN; ++i)
128 dev->mac_addr[i] = ((const uint8_t *)src)[i];
130 PMD_DRV_LOG(ERR, "not supported offset=%zu, len=%d",
135 virtio_user_reset(struct virtio_hw *hw)
137 struct virtio_user_dev *dev = virtio_user_get_dev(hw);
139 if (dev->status & VIRTIO_CONFIG_STATUS_DRIVER_OK)
140 virtio_user_stop_device(dev);
144 virtio_user_set_status(struct virtio_hw *hw, uint8_t status)
146 struct virtio_user_dev *dev = virtio_user_get_dev(hw);
148 if (status & VIRTIO_CONFIG_STATUS_DRIVER_OK)
149 virtio_user_start_device(dev);
150 else if (status == VIRTIO_CONFIG_STATUS_RESET)
151 virtio_user_reset(hw);
152 dev->status = status;
156 virtio_user_get_status(struct virtio_hw *hw)
158 struct virtio_user_dev *dev = virtio_user_get_dev(hw);
164 virtio_user_get_features(struct virtio_hw *hw)
166 struct virtio_user_dev *dev = virtio_user_get_dev(hw);
168 /* unmask feature bits defined in vhost user protocol */
169 return dev->device_features & VIRTIO_PMD_SUPPORTED_GUEST_FEATURES;
173 virtio_user_set_features(struct virtio_hw *hw, uint64_t features)
175 struct virtio_user_dev *dev = virtio_user_get_dev(hw);
177 dev->features = features & dev->device_features;
181 virtio_user_get_isr(struct virtio_hw *hw __rte_unused)
183 /* rxq interrupts and config interrupt are separated in virtio-user,
184 * here we only report config change.
186 return VIRTIO_PCI_ISR_CONFIG;
190 virtio_user_set_config_irq(struct virtio_hw *hw __rte_unused,
191 uint16_t vec __rte_unused)
197 virtio_user_set_queue_irq(struct virtio_hw *hw __rte_unused,
198 struct virtqueue *vq __rte_unused,
201 /* pretend we have done that */
205 /* This function is to get the queue size, aka, number of descs, of a specified
206 * queue. Different with the VHOST_USER_GET_QUEUE_NUM, which is used to get the
207 * max supported queues.
210 virtio_user_get_queue_num(struct virtio_hw *hw, uint16_t queue_id __rte_unused)
212 struct virtio_user_dev *dev = virtio_user_get_dev(hw);
214 /* Currently, each queue has same queue size */
215 return dev->queue_size;
219 virtio_user_setup_queue(struct virtio_hw *hw, struct virtqueue *vq)
221 struct virtio_user_dev *dev = virtio_user_get_dev(hw);
222 uint16_t queue_idx = vq->vq_queue_index;
223 uint64_t desc_addr, avail_addr, used_addr;
225 desc_addr = (uintptr_t)vq->vq_ring_virt_mem;
226 avail_addr = desc_addr + vq->vq_nentries * sizeof(struct vring_desc);
227 used_addr = RTE_ALIGN_CEIL(avail_addr + offsetof(struct vring_avail,
228 ring[vq->vq_nentries]),
229 VIRTIO_PCI_VRING_ALIGN);
231 dev->vrings[queue_idx].num = vq->vq_nentries;
232 dev->vrings[queue_idx].desc = (void *)(uintptr_t)desc_addr;
233 dev->vrings[queue_idx].avail = (void *)(uintptr_t)avail_addr;
234 dev->vrings[queue_idx].used = (void *)(uintptr_t)used_addr;
240 virtio_user_del_queue(struct virtio_hw *hw, struct virtqueue *vq)
242 /* For legacy devices, write 0 to VIRTIO_PCI_QUEUE_PFN port, QEMU
243 * correspondingly stops the ioeventfds, and reset the status of
245 * For modern devices, set queue desc, avail, used in PCI bar to 0,
246 * not see any more behavior in QEMU.
248 * Here we just care about what information to deliver to vhost-user
249 * or vhost-kernel. So we just close ioeventfd for now.
251 struct virtio_user_dev *dev = virtio_user_get_dev(hw);
253 close(dev->callfds[vq->vq_queue_index]);
254 close(dev->kickfds[vq->vq_queue_index]);
258 virtio_user_notify_queue(struct virtio_hw *hw, struct virtqueue *vq)
261 struct virtio_user_dev *dev = virtio_user_get_dev(hw);
263 if (hw->cvq && (hw->cvq->vq == vq)) {
264 virtio_user_handle_cq(dev, vq->vq_queue_index);
268 if (write(dev->kickfds[vq->vq_queue_index], &buf, sizeof(buf)) < 0)
269 PMD_DRV_LOG(ERR, "failed to kick backend: %s",
273 const struct virtio_pci_ops virtio_user_ops = {
274 .read_dev_cfg = virtio_user_read_dev_config,
275 .write_dev_cfg = virtio_user_write_dev_config,
276 .reset = virtio_user_reset,
277 .get_status = virtio_user_get_status,
278 .set_status = virtio_user_set_status,
279 .get_features = virtio_user_get_features,
280 .set_features = virtio_user_set_features,
281 .get_isr = virtio_user_get_isr,
282 .set_config_irq = virtio_user_set_config_irq,
283 .set_queue_irq = virtio_user_set_queue_irq,
284 .get_queue_num = virtio_user_get_queue_num,
285 .setup_queue = virtio_user_setup_queue,
286 .del_queue = virtio_user_del_queue,
287 .notify_queue = virtio_user_notify_queue,
290 static const char *valid_args[] = {
291 #define VIRTIO_USER_ARG_QUEUES_NUM "queues"
292 VIRTIO_USER_ARG_QUEUES_NUM,
293 #define VIRTIO_USER_ARG_CQ_NUM "cq"
294 VIRTIO_USER_ARG_CQ_NUM,
295 #define VIRTIO_USER_ARG_MAC "mac"
297 #define VIRTIO_USER_ARG_PATH "path"
298 VIRTIO_USER_ARG_PATH,
299 #define VIRTIO_USER_ARG_QUEUE_SIZE "queue_size"
300 VIRTIO_USER_ARG_QUEUE_SIZE,
301 #define VIRTIO_USER_ARG_INTERFACE_NAME "iface"
302 VIRTIO_USER_ARG_INTERFACE_NAME,
306 #define VIRTIO_USER_DEF_CQ_EN 0
307 #define VIRTIO_USER_DEF_Q_NUM 1
308 #define VIRTIO_USER_DEF_Q_SZ 256
311 get_string_arg(const char *key __rte_unused,
312 const char *value, void *extra_args)
314 if (!value || !extra_args)
317 *(char **)extra_args = strdup(value);
319 if (!*(char **)extra_args)
326 get_integer_arg(const char *key __rte_unused,
327 const char *value, void *extra_args)
329 if (!value || !extra_args)
332 *(uint64_t *)extra_args = strtoull(value, NULL, 0);
337 static struct rte_vdev_driver virtio_user_driver;
339 static struct rte_eth_dev *
340 virtio_user_eth_dev_alloc(struct rte_vdev_device *vdev)
342 struct rte_eth_dev *eth_dev;
343 struct rte_eth_dev_data *data;
344 struct virtio_hw *hw;
345 struct virtio_user_dev *dev;
347 eth_dev = rte_eth_vdev_allocate(vdev, sizeof(*hw));
349 PMD_INIT_LOG(ERR, "cannot alloc rte_eth_dev");
353 data = eth_dev->data;
354 hw = eth_dev->data->dev_private;
356 dev = rte_zmalloc(NULL, sizeof(*dev), 0);
358 PMD_INIT_LOG(ERR, "malloc virtio_user_dev failed");
359 rte_eth_dev_release_port(eth_dev);
364 hw->port_id = data->port_id;
365 dev->port_id = data->port_id;
366 virtio_hw_internal[hw->port_id].vtpci_ops = &virtio_user_ops;
368 * MSIX is required to enable LSC (see virtio_init_device).
369 * Here just pretend that we support msix.
373 hw->use_simple_rxtx = 0;
374 hw->virtio_user_dev = dev;
375 data->dev_flags = RTE_ETH_DEV_DETACHABLE;
380 virtio_user_eth_dev_free(struct rte_eth_dev *eth_dev)
382 struct rte_eth_dev_data *data = eth_dev->data;
383 struct virtio_hw *hw = data->dev_private;
385 rte_free(hw->virtio_user_dev);
387 rte_eth_dev_release_port(eth_dev);
390 /* Dev initialization routine. Invoked once for each virtio vdev at
391 * EAL init time, see rte_eal_dev_init().
392 * Returns 0 on success.
395 virtio_user_pmd_probe(struct rte_vdev_device *dev)
397 struct rte_kvargs *kvlist = NULL;
398 struct rte_eth_dev *eth_dev;
399 struct virtio_hw *hw;
400 uint64_t queues = VIRTIO_USER_DEF_Q_NUM;
401 uint64_t cq = VIRTIO_USER_DEF_CQ_EN;
402 uint64_t queue_size = VIRTIO_USER_DEF_Q_SZ;
405 char *mac_addr = NULL;
408 kvlist = rte_kvargs_parse(rte_vdev_device_args(dev), valid_args);
410 PMD_INIT_LOG(ERR, "error when parsing param");
414 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_PATH) == 1) {
415 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_PATH,
416 &get_string_arg, &path) < 0) {
417 PMD_INIT_LOG(ERR, "error to parse %s",
418 VIRTIO_USER_ARG_PATH);
422 PMD_INIT_LOG(ERR, "arg %s is mandatory for virtio_user",
423 VIRTIO_USER_ARG_QUEUE_SIZE);
427 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_INTERFACE_NAME) == 1) {
428 if (is_vhost_user_by_type(path)) {
430 "arg %s applies only to vhost-kernel backend",
431 VIRTIO_USER_ARG_INTERFACE_NAME);
435 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_INTERFACE_NAME,
436 &get_string_arg, &ifname) < 0) {
437 PMD_INIT_LOG(ERR, "error to parse %s",
438 VIRTIO_USER_ARG_INTERFACE_NAME);
443 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_MAC) == 1) {
444 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_MAC,
445 &get_string_arg, &mac_addr) < 0) {
446 PMD_INIT_LOG(ERR, "error to parse %s",
447 VIRTIO_USER_ARG_MAC);
452 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_QUEUE_SIZE) == 1) {
453 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_QUEUE_SIZE,
454 &get_integer_arg, &queue_size) < 0) {
455 PMD_INIT_LOG(ERR, "error to parse %s",
456 VIRTIO_USER_ARG_QUEUE_SIZE);
461 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_QUEUES_NUM) == 1) {
462 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_QUEUES_NUM,
463 &get_integer_arg, &queues) < 0) {
464 PMD_INIT_LOG(ERR, "error to parse %s",
465 VIRTIO_USER_ARG_QUEUES_NUM);
470 if (rte_kvargs_count(kvlist, VIRTIO_USER_ARG_CQ_NUM) == 1) {
471 if (rte_kvargs_process(kvlist, VIRTIO_USER_ARG_CQ_NUM,
472 &get_integer_arg, &cq) < 0) {
473 PMD_INIT_LOG(ERR, "error to parse %s",
474 VIRTIO_USER_ARG_CQ_NUM);
477 } else if (queues > 1) {
481 if (queues > 1 && cq == 0) {
482 PMD_INIT_LOG(ERR, "multi-q requires ctrl-q");
486 if (queues > VIRTIO_MAX_VIRTQUEUE_PAIRS) {
487 PMD_INIT_LOG(ERR, "arg %s %" PRIu64 " exceeds the limit %u",
488 VIRTIO_USER_ARG_QUEUES_NUM, queues,
489 VIRTIO_MAX_VIRTQUEUE_PAIRS);
493 if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
494 eth_dev = virtio_user_eth_dev_alloc(dev);
496 PMD_INIT_LOG(ERR, "virtio_user fails to alloc device");
500 hw = eth_dev->data->dev_private;
501 if (virtio_user_dev_init(hw->virtio_user_dev, path, queues, cq,
502 queue_size, mac_addr, &ifname) < 0) {
503 PMD_INIT_LOG(ERR, "virtio_user_dev_init fails");
504 virtio_user_eth_dev_free(eth_dev);
508 eth_dev = rte_eth_dev_attach_secondary(rte_vdev_device_name(dev));
513 /* previously called by rte_pci_probe() for physical dev */
514 if (eth_virtio_dev_init(eth_dev) < 0) {
515 PMD_INIT_LOG(ERR, "eth_virtio_dev_init fails");
516 virtio_user_eth_dev_free(eth_dev);
523 rte_kvargs_free(kvlist);
533 /** Called by rte_eth_dev_detach() */
535 virtio_user_pmd_remove(struct rte_vdev_device *vdev)
538 struct rte_eth_dev *eth_dev;
539 struct virtio_hw *hw;
540 struct virtio_user_dev *dev;
545 name = rte_vdev_device_name(vdev);
546 PMD_DRV_LOG(INFO, "Un-Initializing %s", name);
547 eth_dev = rte_eth_dev_allocated(name);
551 /* make sure the device is stopped, queues freed */
552 rte_eth_dev_close(eth_dev->data->port_id);
554 hw = eth_dev->data->dev_private;
555 dev = hw->virtio_user_dev;
556 virtio_user_dev_uninit(dev);
558 rte_free(eth_dev->data->dev_private);
559 rte_free(eth_dev->data);
560 rte_eth_dev_release_port(eth_dev);
565 static struct rte_vdev_driver virtio_user_driver = {
566 .probe = virtio_user_pmd_probe,
567 .remove = virtio_user_pmd_remove,
570 RTE_PMD_REGISTER_VDEV(net_virtio_user, virtio_user_driver);
571 RTE_PMD_REGISTER_ALIAS(net_virtio_user, virtio_user);
572 RTE_PMD_REGISTER_PARAM_STRING(net_virtio_user,