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, write to the Free Software Foundation, Inc., 59 Temple
15 * Place - Suite 330, Boston, MA 02111-1307 USA.
18 * Haiyang Zhang <haiyangz@microsoft.com>
19 * Hank Janssen <hjanssen@microsoft.com>
21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
23 #include <linux/init.h>
24 #include <linux/atomic.h>
25 #include <linux/module.h>
26 #include <linux/highmem.h>
27 #include <linux/device.h>
29 #include <linux/delay.h>
30 #include <linux/netdevice.h>
31 #include <linux/inetdevice.h>
32 #include <linux/etherdevice.h>
33 #include <linux/skbuff.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;
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;
93 /* Open up the device */
94 ret = rndis_filter_open(device_obj);
96 netdev_err(net, "unable to open device (ret %d).\n", ret);
100 netif_start_queue(net);
105 static int netvsc_close(struct net_device *net)
107 struct net_device_context *net_device_ctx = netdev_priv(net);
108 struct hv_device *device_obj = net_device_ctx->device_ctx;
111 netif_tx_disable(net);
113 /* Make sure netvsc_set_multicast_list doesn't re-enable filter! */
114 cancel_work_sync(&net_device_ctx->work);
115 ret = rndis_filter_close(device_obj);
117 netdev_err(net, "unable to close device (ret %d).\n", ret);
122 static void netvsc_xmit_completion(void *context)
124 struct hv_netvsc_packet *packet = (struct hv_netvsc_packet *)context;
125 struct sk_buff *skb = (struct sk_buff *)
126 (unsigned long)packet->completion.send.send_completion_tid;
131 dev_kfree_skb_any(skb);
134 static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
136 struct net_device_context *net_device_ctx = netdev_priv(net);
137 struct hv_netvsc_packet *packet;
139 unsigned int i, num_pages, npg_data;
141 /* Add multipages for skb->data and additional 2 for RNDIS */
142 npg_data = (((unsigned long)skb->data + skb_headlen(skb) - 1)
143 >> PAGE_SHIFT) - ((unsigned long)skb->data >> PAGE_SHIFT) + 1;
144 num_pages = skb_shinfo(skb)->nr_frags + npg_data + 2;
146 /* Allocate a netvsc packet based on # of frags. */
147 packet = kzalloc(sizeof(struct hv_netvsc_packet) +
148 (num_pages * sizeof(struct hv_page_buffer)) +
149 sizeof(struct rndis_filter_packet) +
150 NDIS_VLAN_PPI_SIZE, GFP_ATOMIC);
152 /* out of memory, drop packet */
153 netdev_err(net, "unable to allocate hv_netvsc_packet\n");
156 net->stats.tx_dropped++;
160 packet->vlan_tci = skb->vlan_tci;
162 packet->extension = (void *)(unsigned long)packet +
163 sizeof(struct hv_netvsc_packet) +
164 (num_pages * sizeof(struct hv_page_buffer));
166 /* If the rndis msg goes beyond 1 page, we will add 1 later */
167 packet->page_buf_cnt = num_pages - 1;
169 /* Initialize it from the skb */
170 packet->total_data_buflen = skb->len;
172 /* Start filling in the page buffers starting after RNDIS buffer. */
173 packet->page_buf[1].pfn = virt_to_phys(skb->data) >> PAGE_SHIFT;
174 packet->page_buf[1].offset
175 = (unsigned long)skb->data & (PAGE_SIZE - 1);
177 packet->page_buf[1].len = skb_headlen(skb);
179 packet->page_buf[1].len = PAGE_SIZE
180 - packet->page_buf[1].offset;
182 for (i = 2; i <= npg_data; i++) {
183 packet->page_buf[i].pfn = virt_to_phys(skb->data
184 + PAGE_SIZE * (i-1)) >> PAGE_SHIFT;
185 packet->page_buf[i].offset = 0;
186 packet->page_buf[i].len = PAGE_SIZE;
189 packet->page_buf[npg_data].len = (((unsigned long)skb->data
190 + skb_headlen(skb) - 1) & (PAGE_SIZE - 1)) + 1;
192 /* Additional fragments are after SKB data */
193 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
194 const skb_frag_t *f = &skb_shinfo(skb)->frags[i];
196 packet->page_buf[i+npg_data+1].pfn =
197 page_to_pfn(skb_frag_page(f));
198 packet->page_buf[i+npg_data+1].offset = f->page_offset;
199 packet->page_buf[i+npg_data+1].len = skb_frag_size(f);
202 /* Set the completion routine */
203 packet->completion.send.send_completion = netvsc_xmit_completion;
204 packet->completion.send.send_completion_ctx = packet;
205 packet->completion.send.send_completion_tid = (unsigned long)skb;
207 ret = rndis_filter_send(net_device_ctx->device_ctx,
210 net->stats.tx_bytes += skb->len;
211 net->stats.tx_packets++;
214 if (ret != -EAGAIN) {
215 dev_kfree_skb_any(skb);
216 net->stats.tx_dropped++;
220 return (ret == -EAGAIN) ? NETDEV_TX_BUSY : NETDEV_TX_OK;
224 * netvsc_linkstatus_callback - Link up/down notification
226 void netvsc_linkstatus_callback(struct hv_device *device_obj,
229 struct net_device *net;
230 struct net_device_context *ndev_ctx;
231 struct netvsc_device *net_device;
233 net_device = hv_get_drvdata(device_obj);
234 net = net_device->ndev;
237 netdev_err(net, "got link status but net device "
238 "not initialized yet\n");
243 netif_carrier_on(net);
244 netif_wake_queue(net);
245 ndev_ctx = netdev_priv(net);
246 schedule_delayed_work(&ndev_ctx->dwork, 0);
247 schedule_delayed_work(&ndev_ctx->dwork, msecs_to_jiffies(20));
249 netif_carrier_off(net);
250 netif_tx_disable(net);
255 * netvsc_recv_callback - Callback when we receive a packet from the
256 * "wire" on the specified device.
258 int netvsc_recv_callback(struct hv_device *device_obj,
259 struct hv_netvsc_packet *packet)
261 struct net_device *net;
264 net = ((struct netvsc_device *)hv_get_drvdata(device_obj))->ndev;
266 netdev_err(net, "got receive callback but net device"
267 " not initialized yet\n");
271 /* Allocate a skb - TODO direct I/O to pages? */
272 skb = netdev_alloc_skb_ip_align(net, packet->total_data_buflen);
273 if (unlikely(!skb)) {
274 ++net->stats.rx_dropped;
279 * Copy to skb. This copy is needed here since the memory pointed by
280 * hv_netvsc_packet cannot be deallocated
282 memcpy(skb_put(skb, packet->total_data_buflen), packet->data,
283 packet->total_data_buflen);
285 skb->protocol = eth_type_trans(skb, net);
286 skb->ip_summed = CHECKSUM_NONE;
287 skb->vlan_tci = packet->vlan_tci;
289 net->stats.rx_packets++;
290 net->stats.rx_bytes += packet->total_data_buflen;
293 * Pass the skb back up. Network stack will deallocate the skb when it
302 static void netvsc_get_drvinfo(struct net_device *net,
303 struct ethtool_drvinfo *info)
305 strcpy(info->driver, KBUILD_MODNAME);
306 strcpy(info->version, HV_DRV_VERSION);
307 strcpy(info->fw_version, "N/A");
310 static int netvsc_change_mtu(struct net_device *ndev, int mtu)
312 struct net_device_context *ndevctx = netdev_priv(ndev);
313 struct hv_device *hdev = ndevctx->device_ctx;
314 struct netvsc_device *nvdev = hv_get_drvdata(hdev);
315 struct netvsc_device_info device_info;
316 int limit = ETH_DATA_LEN;
318 if (nvdev == NULL || nvdev->destroy)
321 if (nvdev->nvsp_version == NVSP_PROTOCOL_VERSION_2)
324 if (mtu < 68 || mtu > limit)
327 nvdev->start_remove = true;
328 cancel_delayed_work_sync(&ndevctx->dwork);
329 cancel_work_sync(&ndevctx->work);
330 netif_tx_disable(ndev);
331 rndis_filter_device_remove(hdev);
335 ndevctx->device_ctx = hdev;
336 hv_set_drvdata(hdev, ndev);
337 device_info.ring_size = ring_size;
338 rndis_filter_device_add(hdev, &device_info);
339 netif_wake_queue(ndev);
344 static const struct ethtool_ops ethtool_ops = {
345 .get_drvinfo = netvsc_get_drvinfo,
346 .get_link = ethtool_op_get_link,
349 static const struct net_device_ops device_ops = {
350 .ndo_open = netvsc_open,
351 .ndo_stop = netvsc_close,
352 .ndo_start_xmit = netvsc_start_xmit,
353 .ndo_set_rx_mode = netvsc_set_multicast_list,
354 .ndo_change_mtu = netvsc_change_mtu,
355 .ndo_validate_addr = eth_validate_addr,
356 .ndo_set_mac_address = eth_mac_addr,
360 * Send GARP packet to network peers after migrations.
361 * After Quick Migration, the network is not immediately operational in the
362 * current context when receiving RNDIS_STATUS_MEDIA_CONNECT event. So, add
363 * another netif_notify_peers() into a delayed work, otherwise GARP packet
364 * will not be sent after quick migration, and cause network disconnection.
366 static void netvsc_send_garp(struct work_struct *w)
368 struct net_device_context *ndev_ctx;
369 struct net_device *net;
370 struct netvsc_device *net_device;
372 ndev_ctx = container_of(w, struct net_device_context, dwork.work);
373 net_device = hv_get_drvdata(ndev_ctx->device_ctx);
374 net = net_device->ndev;
375 netif_notify_peers(net);
379 static int netvsc_probe(struct hv_device *dev,
380 const struct hv_vmbus_device_id *dev_id)
382 struct net_device *net = NULL;
383 struct net_device_context *net_device_ctx;
384 struct netvsc_device_info device_info;
387 net = alloc_etherdev(sizeof(struct net_device_context));
391 /* Set initial state */
392 netif_carrier_off(net);
394 net_device_ctx = netdev_priv(net);
395 net_device_ctx->device_ctx = dev;
396 hv_set_drvdata(dev, net);
397 INIT_DELAYED_WORK(&net_device_ctx->dwork, netvsc_send_garp);
398 INIT_WORK(&net_device_ctx->work, do_set_multicast);
400 net->netdev_ops = &device_ops;
402 /* TODO: Add GSO and Checksum offload */
403 net->hw_features = NETIF_F_SG;
404 net->features = NETIF_F_SG | NETIF_F_HW_VLAN_TX;
406 SET_ETHTOOL_OPS(net, ðtool_ops);
407 SET_NETDEV_DEV(net, &dev->device);
409 ret = register_netdev(net);
411 pr_err("Unable to register netdev.\n");
416 /* Notify the netvsc driver of the new device */
417 device_info.ring_size = ring_size;
418 ret = rndis_filter_device_add(dev, &device_info);
420 netdev_err(net, "unable to add netvsc device (ret %d)\n", ret);
421 unregister_netdev(net);
423 hv_set_drvdata(dev, NULL);
426 memcpy(net->dev_addr, device_info.mac_adr, ETH_ALEN);
428 netif_carrier_on(net);
434 static int netvsc_remove(struct hv_device *dev)
436 struct net_device *net;
437 struct net_device_context *ndev_ctx;
438 struct netvsc_device *net_device;
440 net_device = hv_get_drvdata(dev);
441 net = net_device->ndev;
444 dev_err(&dev->device, "No net device to remove\n");
448 net_device->start_remove = true;
450 ndev_ctx = netdev_priv(net);
451 cancel_delayed_work_sync(&ndev_ctx->dwork);
452 cancel_work_sync(&ndev_ctx->work);
454 /* Stop outbound asap */
455 netif_tx_disable(net);
457 unregister_netdev(net);
460 * Call to the vsc driver to let it know that the device is being
463 rndis_filter_device_remove(dev);
469 static const struct hv_vmbus_device_id id_table[] = {
471 { VMBUS_DEVICE(0x63, 0x51, 0x61, 0xF8, 0x3E, 0xDF, 0xc5, 0x46,
472 0x91, 0x3F, 0xF2, 0xD2, 0xF9, 0x65, 0xED, 0x0E) },
476 MODULE_DEVICE_TABLE(vmbus, id_table);
478 /* The one and only one */
479 static struct hv_driver netvsc_drv = {
480 .name = KBUILD_MODNAME,
481 .id_table = id_table,
482 .probe = netvsc_probe,
483 .remove = netvsc_remove,
486 static void __exit netvsc_drv_exit(void)
488 vmbus_driver_unregister(&netvsc_drv);
491 static int __init netvsc_drv_init(void)
493 return vmbus_driver_register(&netvsc_drv);
496 MODULE_LICENSE("GPL");
497 MODULE_VERSION(HV_DRV_VERSION);
498 MODULE_DESCRIPTION("Microsoft Hyper-V network driver");
500 module_init(netvsc_drv_init);
501 module_exit(netvsc_drv_exit);