2 * Copyright (c) 2009, Microsoft Corporation.
4 * This program is free software; you can redistribute it and/or modify it
5 * under the terms and conditions of the GNU General Public License,
6 * version 2, as published by the Free Software Foundation.
8 * This program is distributed in the hope it will be useful, but WITHOUT
9 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
13 * You should have received a copy of the GNU General Public License along with
14 * this program; if not, see <http://www.gnu.org/licenses/>.
17 * Haiyang Zhang <haiyangz@microsoft.com>
18 * Hank Janssen <hjanssen@microsoft.com>
20 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22 #include <linux/init.h>
23 #include <linux/atomic.h>
24 #include <linux/module.h>
25 #include <linux/highmem.h>
26 #include <linux/device.h>
28 #include <linux/delay.h>
29 #include <linux/netdevice.h>
30 #include <linux/inetdevice.h>
31 #include <linux/etherdevice.h>
32 #include <linux/skbuff.h>
33 #include <linux/if_vlan.h>
35 #include <linux/slab.h>
37 #include <net/route.h>
39 #include <net/pkt_sched.h>
41 #include "hyperv_net.h"
43 struct net_device_context {
44 /* point back to our device context */
45 struct hv_device *device_ctx;
46 struct delayed_work dwork;
47 struct work_struct work;
50 #define RING_SIZE_MIN 64
51 static int ring_size = 128;
52 module_param(ring_size, int, S_IRUGO);
53 MODULE_PARM_DESC(ring_size, "Ring buffer size (# of pages)");
55 static void do_set_multicast(struct work_struct *w)
57 struct net_device_context *ndevctx =
58 container_of(w, struct net_device_context, work);
59 struct netvsc_device *nvdev;
60 struct rndis_device *rdev;
62 nvdev = hv_get_drvdata(ndevctx->device_ctx);
63 if (nvdev == NULL || nvdev->ndev == NULL)
66 rdev = nvdev->extension;
70 if (nvdev->ndev->flags & IFF_PROMISC)
71 rndis_filter_set_packet_filter(rdev,
72 NDIS_PACKET_TYPE_PROMISCUOUS);
74 rndis_filter_set_packet_filter(rdev,
75 NDIS_PACKET_TYPE_BROADCAST |
76 NDIS_PACKET_TYPE_ALL_MULTICAST |
77 NDIS_PACKET_TYPE_DIRECTED);
80 static void netvsc_set_multicast_list(struct net_device *net)
82 struct net_device_context *net_device_ctx = netdev_priv(net);
84 schedule_work(&net_device_ctx->work);
87 static int netvsc_open(struct net_device *net)
89 struct net_device_context *net_device_ctx = netdev_priv(net);
90 struct hv_device *device_obj = net_device_ctx->device_ctx;
91 struct netvsc_device *nvdev;
92 struct rndis_device *rdev;
95 netif_carrier_off(net);
97 /* Open up the device */
98 ret = rndis_filter_open(device_obj);
100 netdev_err(net, "unable to open device (ret %d).\n", ret);
104 netif_start_queue(net);
106 nvdev = hv_get_drvdata(device_obj);
107 rdev = nvdev->extension;
108 if (!rdev->link_state)
109 netif_carrier_on(net);
114 static int netvsc_close(struct net_device *net)
116 struct net_device_context *net_device_ctx = netdev_priv(net);
117 struct hv_device *device_obj = net_device_ctx->device_ctx;
120 netif_tx_disable(net);
122 /* Make sure netvsc_set_multicast_list doesn't re-enable filter! */
123 cancel_work_sync(&net_device_ctx->work);
124 ret = rndis_filter_close(device_obj);
126 netdev_err(net, "unable to close device (ret %d).\n", ret);
131 static void netvsc_xmit_completion(void *context)
133 struct hv_netvsc_packet *packet = (struct hv_netvsc_packet *)context;
134 struct sk_buff *skb = (struct sk_buff *)
135 (unsigned long)packet->completion.send.send_completion_tid;
140 dev_kfree_skb_any(skb);
143 static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
145 struct net_device_context *net_device_ctx = netdev_priv(net);
146 struct hv_netvsc_packet *packet;
148 unsigned int i, num_pages, npg_data;
150 /* Add multipages for skb->data and additional 2 for RNDIS */
151 npg_data = (((unsigned long)skb->data + skb_headlen(skb) - 1)
152 >> PAGE_SHIFT) - ((unsigned long)skb->data >> PAGE_SHIFT) + 1;
153 num_pages = skb_shinfo(skb)->nr_frags + npg_data + 2;
155 /* Allocate a netvsc packet based on # of frags. */
156 packet = kzalloc(sizeof(struct hv_netvsc_packet) +
157 (num_pages * sizeof(struct hv_page_buffer)) +
158 sizeof(struct rndis_filter_packet) +
159 NDIS_VLAN_PPI_SIZE, GFP_ATOMIC);
161 /* out of memory, drop packet */
162 netdev_err(net, "unable to allocate hv_netvsc_packet\n");
165 net->stats.tx_dropped++;
169 packet->vlan_tci = skb->vlan_tci;
171 packet->extension = (void *)(unsigned long)packet +
172 sizeof(struct hv_netvsc_packet) +
173 (num_pages * sizeof(struct hv_page_buffer));
175 /* If the rndis msg goes beyond 1 page, we will add 1 later */
176 packet->page_buf_cnt = num_pages - 1;
178 /* Initialize it from the skb */
179 packet->total_data_buflen = skb->len;
181 /* Start filling in the page buffers starting after RNDIS buffer. */
182 packet->page_buf[1].pfn = virt_to_phys(skb->data) >> PAGE_SHIFT;
183 packet->page_buf[1].offset
184 = (unsigned long)skb->data & (PAGE_SIZE - 1);
186 packet->page_buf[1].len = skb_headlen(skb);
188 packet->page_buf[1].len = PAGE_SIZE
189 - packet->page_buf[1].offset;
191 for (i = 2; i <= npg_data; i++) {
192 packet->page_buf[i].pfn = virt_to_phys(skb->data
193 + PAGE_SIZE * (i-1)) >> PAGE_SHIFT;
194 packet->page_buf[i].offset = 0;
195 packet->page_buf[i].len = PAGE_SIZE;
198 packet->page_buf[npg_data].len = (((unsigned long)skb->data
199 + skb_headlen(skb) - 1) & (PAGE_SIZE - 1)) + 1;
201 /* Additional fragments are after SKB data */
202 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
203 const skb_frag_t *f = &skb_shinfo(skb)->frags[i];
205 packet->page_buf[i+npg_data+1].pfn =
206 page_to_pfn(skb_frag_page(f));
207 packet->page_buf[i+npg_data+1].offset = f->page_offset;
208 packet->page_buf[i+npg_data+1].len = skb_frag_size(f);
211 /* Set the completion routine */
212 packet->completion.send.send_completion = netvsc_xmit_completion;
213 packet->completion.send.send_completion_ctx = packet;
214 packet->completion.send.send_completion_tid = (unsigned long)skb;
216 ret = rndis_filter_send(net_device_ctx->device_ctx,
219 net->stats.tx_bytes += skb->len;
220 net->stats.tx_packets++;
223 if (ret != -EAGAIN) {
224 dev_kfree_skb_any(skb);
225 net->stats.tx_dropped++;
229 return (ret == -EAGAIN) ? NETDEV_TX_BUSY : NETDEV_TX_OK;
233 * netvsc_linkstatus_callback - Link up/down notification
235 void netvsc_linkstatus_callback(struct hv_device *device_obj,
238 struct net_device *net;
239 struct net_device_context *ndev_ctx;
240 struct netvsc_device *net_device;
241 struct rndis_device *rdev;
243 net_device = hv_get_drvdata(device_obj);
244 rdev = net_device->extension;
246 rdev->link_state = status != 1;
248 net = net_device->ndev;
250 if (!net || net->reg_state != NETREG_REGISTERED)
253 ndev_ctx = netdev_priv(net);
255 schedule_delayed_work(&ndev_ctx->dwork, 0);
256 schedule_delayed_work(&ndev_ctx->dwork, msecs_to_jiffies(20));
258 schedule_delayed_work(&ndev_ctx->dwork, 0);
263 * netvsc_recv_callback - Callback when we receive a packet from the
264 * "wire" on the specified device.
266 int netvsc_recv_callback(struct hv_device *device_obj,
267 struct hv_netvsc_packet *packet)
269 struct net_device *net;
272 net = ((struct netvsc_device *)hv_get_drvdata(device_obj))->ndev;
273 if (!net || net->reg_state != NETREG_REGISTERED) {
274 packet->status = NVSP_STAT_FAIL;
278 /* Allocate a skb - TODO direct I/O to pages? */
279 skb = netdev_alloc_skb_ip_align(net, packet->total_data_buflen);
280 if (unlikely(!skb)) {
281 ++net->stats.rx_dropped;
282 packet->status = NVSP_STAT_FAIL;
287 * Copy to skb. This copy is needed here since the memory pointed by
288 * hv_netvsc_packet cannot be deallocated
290 memcpy(skb_put(skb, packet->total_data_buflen), packet->data,
291 packet->total_data_buflen);
293 skb->protocol = eth_type_trans(skb, net);
294 skb->ip_summed = CHECKSUM_NONE;
295 if (packet->vlan_tci & VLAN_TAG_PRESENT)
296 __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
299 net->stats.rx_packets++;
300 net->stats.rx_bytes += packet->total_data_buflen;
303 * Pass the skb back up. Network stack will deallocate the skb when it
312 static void netvsc_get_drvinfo(struct net_device *net,
313 struct ethtool_drvinfo *info)
315 strlcpy(info->driver, KBUILD_MODNAME, sizeof(info->driver));
316 strlcpy(info->fw_version, "N/A", sizeof(info->fw_version));
319 static int netvsc_change_mtu(struct net_device *ndev, int mtu)
321 struct net_device_context *ndevctx = netdev_priv(ndev);
322 struct hv_device *hdev = ndevctx->device_ctx;
323 struct netvsc_device *nvdev = hv_get_drvdata(hdev);
324 struct netvsc_device_info device_info;
325 int limit = ETH_DATA_LEN;
327 if (nvdev == NULL || nvdev->destroy)
330 if (nvdev->nvsp_version == NVSP_PROTOCOL_VERSION_2)
333 if (mtu < 68 || mtu > limit)
336 nvdev->start_remove = true;
337 cancel_work_sync(&ndevctx->work);
338 netif_tx_disable(ndev);
339 rndis_filter_device_remove(hdev);
343 ndevctx->device_ctx = hdev;
344 hv_set_drvdata(hdev, ndev);
345 device_info.ring_size = ring_size;
346 rndis_filter_device_add(hdev, &device_info);
347 netif_wake_queue(ndev);
353 static int netvsc_set_mac_addr(struct net_device *ndev, void *p)
355 struct net_device_context *ndevctx = netdev_priv(ndev);
356 struct hv_device *hdev = ndevctx->device_ctx;
357 struct sockaddr *addr = p;
358 char save_adr[ETH_ALEN];
359 unsigned char save_aatype;
362 memcpy(save_adr, ndev->dev_addr, ETH_ALEN);
363 save_aatype = ndev->addr_assign_type;
365 err = eth_mac_addr(ndev, p);
369 err = rndis_filter_set_device_mac(hdev, addr->sa_data);
371 /* roll back to saved MAC */
372 memcpy(ndev->dev_addr, save_adr, ETH_ALEN);
373 ndev->addr_assign_type = save_aatype;
380 static const struct ethtool_ops ethtool_ops = {
381 .get_drvinfo = netvsc_get_drvinfo,
382 .get_link = ethtool_op_get_link,
385 static const struct net_device_ops device_ops = {
386 .ndo_open = netvsc_open,
387 .ndo_stop = netvsc_close,
388 .ndo_start_xmit = netvsc_start_xmit,
389 .ndo_set_rx_mode = netvsc_set_multicast_list,
390 .ndo_change_mtu = netvsc_change_mtu,
391 .ndo_validate_addr = eth_validate_addr,
392 .ndo_set_mac_address = netvsc_set_mac_addr,
396 * Send GARP packet to network peers after migrations.
397 * After Quick Migration, the network is not immediately operational in the
398 * current context when receiving RNDIS_STATUS_MEDIA_CONNECT event. So, add
399 * another netif_notify_peers() into a delayed work, otherwise GARP packet
400 * will not be sent after quick migration, and cause network disconnection.
401 * Also, we update the carrier status here.
403 static void netvsc_link_change(struct work_struct *w)
405 struct net_device_context *ndev_ctx;
406 struct net_device *net;
407 struct netvsc_device *net_device;
408 struct rndis_device *rdev;
413 ndev_ctx = container_of(w, struct net_device_context, dwork.work);
414 net_device = hv_get_drvdata(ndev_ctx->device_ctx);
415 rdev = net_device->extension;
416 net = net_device->ndev;
418 if (rdev->link_state) {
419 netif_carrier_off(net);
422 netif_carrier_on(net);
429 netdev_notify_peers(net);
433 static int netvsc_probe(struct hv_device *dev,
434 const struct hv_vmbus_device_id *dev_id)
436 struct net_device *net = NULL;
437 struct net_device_context *net_device_ctx;
438 struct netvsc_device_info device_info;
441 net = alloc_etherdev(sizeof(struct net_device_context));
445 netif_carrier_off(net);
447 net_device_ctx = netdev_priv(net);
448 net_device_ctx->device_ctx = dev;
449 hv_set_drvdata(dev, net);
450 INIT_DELAYED_WORK(&net_device_ctx->dwork, netvsc_link_change);
451 INIT_WORK(&net_device_ctx->work, do_set_multicast);
453 net->netdev_ops = &device_ops;
455 /* TODO: Add GSO and Checksum offload */
456 net->hw_features = 0;
457 net->features = NETIF_F_HW_VLAN_CTAG_TX;
459 SET_ETHTOOL_OPS(net, ðtool_ops);
460 SET_NETDEV_DEV(net, &dev->device);
462 /* Notify the netvsc driver of the new device */
463 device_info.ring_size = ring_size;
464 ret = rndis_filter_device_add(dev, &device_info);
466 netdev_err(net, "unable to add netvsc device (ret %d)\n", ret);
468 hv_set_drvdata(dev, NULL);
471 memcpy(net->dev_addr, device_info.mac_adr, ETH_ALEN);
473 ret = register_netdev(net);
475 pr_err("Unable to register netdev.\n");
476 rndis_filter_device_remove(dev);
479 schedule_delayed_work(&net_device_ctx->dwork, 0);
485 static int netvsc_remove(struct hv_device *dev)
487 struct net_device *net;
488 struct net_device_context *ndev_ctx;
489 struct netvsc_device *net_device;
491 net_device = hv_get_drvdata(dev);
492 net = net_device->ndev;
495 dev_err(&dev->device, "No net device to remove\n");
499 net_device->start_remove = true;
501 ndev_ctx = netdev_priv(net);
502 cancel_delayed_work_sync(&ndev_ctx->dwork);
503 cancel_work_sync(&ndev_ctx->work);
505 /* Stop outbound asap */
506 netif_tx_disable(net);
508 unregister_netdev(net);
511 * Call to the vsc driver to let it know that the device is being
514 rndis_filter_device_remove(dev);
520 static const struct hv_vmbus_device_id id_table[] = {
526 MODULE_DEVICE_TABLE(vmbus, id_table);
528 /* The one and only one */
529 static struct hv_driver netvsc_drv = {
530 .name = KBUILD_MODNAME,
531 .id_table = id_table,
532 .probe = netvsc_probe,
533 .remove = netvsc_remove,
536 static void __exit netvsc_drv_exit(void)
538 vmbus_driver_unregister(&netvsc_drv);
541 static int __init netvsc_drv_init(void)
543 if (ring_size < RING_SIZE_MIN) {
544 ring_size = RING_SIZE_MIN;
545 pr_info("Increased ring_size to %d (min allowed)\n",
548 return vmbus_driver_register(&netvsc_drv);
551 MODULE_LICENSE("GPL");
552 MODULE_DESCRIPTION("Microsoft Hyper-V network driver");
554 module_init(netvsc_drv_init);
555 module_exit(netvsc_drv_exit);