drivers/net/hyperv/netvsc_drv.c

   1 /*
   2  * Copyright (c) 2009, Microsoft Corporation.
   3  *
   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.
   7  *
   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
  11  * more details.
  12  *
  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.
  16  *
  17  * Authors:
  18  *   Haiyang Zhang <haiyangz@microsoft.com>
  19  *   Hank Janssen  <hjanssen@microsoft.com>
  20  */
  21 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  22
  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>
  28 #include <linux/io.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>
  34 #include <linux/in.h>
  35 #include <linux/slab.h>
  36 #include <net/arp.h>
  37 #include <net/route.h>
  38 #include <net/sock.h>
  39 #include <net/pkt_sched.h>
  40
  41 #include "hyperv_net.h"
  42
  43 struct net_device_context {
  44         /* point back to our device context */
  45         struct hv_device *device_ctx;
  46         struct delayed_work dwork;
  47 };
  48
  49
  50 static int ring_size = 128;
  51 module_param(ring_size, int, S_IRUGO);
  52 MODULE_PARM_DESC(ring_size, "Ring buffer size (# of pages)");
  53
  54 struct set_multicast_work {
  55         struct work_struct work;
  56         struct net_device *net;
  57 };
  58
  59 static void do_set_multicast(struct work_struct *w)
  60 {
  61         struct set_multicast_work *swk =
  62                 container_of(w, struct set_multicast_work, work);
  63         struct net_device *net = swk->net;
  64
  65         struct net_device_context *ndevctx = netdev_priv(net);
  66         struct netvsc_device *nvdev;
  67         struct rndis_device *rdev;
  68
  69         nvdev = hv_get_drvdata(ndevctx->device_ctx);
  70         if (nvdev == NULL)
  71                 return;
  72
  73         rdev = nvdev->extension;
  74         if (rdev == NULL)
  75                 return;
  76
  77         if (net->flags & IFF_PROMISC)
  78                 rndis_filter_set_packet_filter(rdev,
  79                         NDIS_PACKET_TYPE_PROMISCUOUS);
  80         else
  81                 rndis_filter_set_packet_filter(rdev,
  82                         NDIS_PACKET_TYPE_BROADCAST |
  83                         NDIS_PACKET_TYPE_ALL_MULTICAST |
  84                         NDIS_PACKET_TYPE_DIRECTED);
  85
  86         kfree(w);
  87 }
  88
  89 static void netvsc_set_multicast_list(struct net_device *net)
  90 {
  91         struct set_multicast_work *swk =
  92                 kmalloc(sizeof(struct set_multicast_work), GFP_ATOMIC);
  93         if (swk == NULL)
  94                 return;
  95
  96         swk->net = net;
  97         INIT_WORK(&swk->work, do_set_multicast);
  98         schedule_work(&swk->work);
  99 }
 100
 101 static int netvsc_open(struct net_device *net)
 102 {
 103         struct net_device_context *net_device_ctx = netdev_priv(net);
 104         struct hv_device *device_obj = net_device_ctx->device_ctx;
 105         int ret = 0;
 106
 107         /* Open up the device */
 108         ret = rndis_filter_open(device_obj);
 109         if (ret != 0) {
 110                 netdev_err(net, "unable to open device (ret %d).\n", ret);
 111                 return ret;
 112         }
 113
 114         netif_start_queue(net);
 115
 116         return ret;
 117 }
 118
 119 static int netvsc_close(struct net_device *net)
 120 {
 121         struct net_device_context *net_device_ctx = netdev_priv(net);
 122         struct hv_device *device_obj = net_device_ctx->device_ctx;
 123         int ret;
 124
 125         netif_stop_queue(net);
 126
 127         ret = rndis_filter_close(device_obj);
 128         if (ret != 0)
 129                 netdev_err(net, "unable to close device (ret %d).\n", ret);
 130
 131         return ret;
 132 }
 133
 134 static void netvsc_xmit_completion(void *context)
 135 {
 136         struct hv_netvsc_packet *packet = (struct hv_netvsc_packet *)context;
 137         struct sk_buff *skb = (struct sk_buff *)
 138                 (unsigned long)packet->completion.send.send_completion_tid;
 139
 140         kfree(packet);
 141
 142         if (skb)
 143                 dev_kfree_skb_any(skb);
 144 }
 145
 146 static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
 147 {
 148         struct net_device_context *net_device_ctx = netdev_priv(net);
 149         struct hv_netvsc_packet *packet;
 150         int ret;
 151         unsigned int i, num_pages;
 152
 153         /* Add 1 for skb->data and additional one for RNDIS */
 154         num_pages = skb_shinfo(skb)->nr_frags + 1 + 1;
 155
 156         /* Allocate a netvsc packet based on # of frags. */
 157         packet = kzalloc(sizeof(struct hv_netvsc_packet) +
 158                          (num_pages * sizeof(struct hv_page_buffer)) +
 159                          sizeof(struct rndis_filter_packet), GFP_ATOMIC);
 160         if (!packet) {
 161                 /* out of memory, drop packet */
 162                 netdev_err(net, "unable to allocate hv_netvsc_packet\n");
 163
 164                 dev_kfree_skb(skb);
 165                 net->stats.tx_dropped++;
 166                 return NETDEV_TX_BUSY;
 167         }
 168
 169         packet->extension = (void *)(unsigned long)packet +
 170                                 sizeof(struct hv_netvsc_packet) +
 171                                     (num_pages * sizeof(struct hv_page_buffer));
 172
 173         /* Setup the rndis header */
 174         packet->page_buf_cnt = num_pages;
 175
 176         /* Initialize it from the skb */
 177         packet->total_data_buflen       = skb->len;
 178
 179         /* Start filling in the page buffers starting after RNDIS buffer. */
 180         packet->page_buf[1].pfn = virt_to_phys(skb->data) >> PAGE_SHIFT;
 181         packet->page_buf[1].offset
 182                 = (unsigned long)skb->data & (PAGE_SIZE - 1);
 183         packet->page_buf[1].len = skb_headlen(skb);
 184
 185         /* Additional fragments are after SKB data */
 186         for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
 187                 const skb_frag_t *f = &skb_shinfo(skb)->frags[i];
 188
 189                 packet->page_buf[i+2].pfn = page_to_pfn(skb_frag_page(f));
 190                 packet->page_buf[i+2].offset = f->page_offset;
 191                 packet->page_buf[i+2].len = skb_frag_size(f);
 192         }
 193
 194         /* Set the completion routine */
 195         packet->completion.send.send_completion = netvsc_xmit_completion;
 196         packet->completion.send.send_completion_ctx = packet;
 197         packet->completion.send.send_completion_tid = (unsigned long)skb;
 198
 199         ret = rndis_filter_send(net_device_ctx->device_ctx,
 200                                   packet);
 201         if (ret == 0) {
 202                 net->stats.tx_bytes += skb->len;
 203                 net->stats.tx_packets++;
 204         } else {
 205                 /* we are shutting down or bus overloaded, just drop packet */
 206                 net->stats.tx_dropped++;
 207                 kfree(packet);
 208                 dev_kfree_skb_any(skb);
 209         }
 210
 211         return ret ? NETDEV_TX_BUSY : NETDEV_TX_OK;
 212 }
 213
 214 /*
 215  * netvsc_linkstatus_callback - Link up/down notification
 216  */
 217 void netvsc_linkstatus_callback(struct hv_device *device_obj,
 218                                        unsigned int status)
 219 {
 220         struct net_device *net;
 221         struct net_device_context *ndev_ctx;
 222         struct netvsc_device *net_device;
 223
 224         net_device = hv_get_drvdata(device_obj);
 225         net = net_device->ndev;
 226
 227         if (!net) {
 228                 netdev_err(net, "got link status but net device "
 229                                 "not initialized yet\n");
 230                 return;
 231         }
 232
 233         if (status == 1) {
 234                 netif_carrier_on(net);
 235                 netif_wake_queue(net);
 236                 ndev_ctx = netdev_priv(net);
 237                 schedule_delayed_work(&ndev_ctx->dwork, 0);
 238                 schedule_delayed_work(&ndev_ctx->dwork, msecs_to_jiffies(20));
 239         } else {
 240                 netif_carrier_off(net);
 241                 netif_stop_queue(net);
 242         }
 243 }
 244
 245 /*
 246  * netvsc_recv_callback -  Callback when we receive a packet from the
 247  * "wire" on the specified device.
 248  */
 249 int netvsc_recv_callback(struct hv_device *device_obj,
 250                                 struct hv_netvsc_packet *packet)
 251 {
 252         struct net_device *net = dev_get_drvdata(&device_obj->device);
 253         struct sk_buff *skb;
 254         struct netvsc_device *net_device;
 255
 256         net_device = hv_get_drvdata(device_obj);
 257         net = net_device->ndev;
 258
 259         if (!net) {
 260                 netdev_err(net, "got receive callback but net device"
 261                         " not initialized yet\n");
 262                 return 0;
 263         }
 264
 265         /* Allocate a skb - TODO direct I/O to pages? */
 266         skb = netdev_alloc_skb_ip_align(net, packet->total_data_buflen);
 267         if (unlikely(!skb)) {
 268                 ++net->stats.rx_dropped;
 269                 return 0;
 270         }
 271
 272         /*
 273          * Copy to skb. This copy is needed here since the memory pointed by
 274          * hv_netvsc_packet cannot be deallocated
 275          */
 276         memcpy(skb_put(skb, packet->total_data_buflen), packet->data,
 277                 packet->total_data_buflen);
 278
 279         skb->protocol = eth_type_trans(skb, net);
 280         skb->ip_summed = CHECKSUM_NONE;
 281
 282         net->stats.rx_packets++;
 283         net->stats.rx_bytes += skb->len;
 284
 285         /*
 286          * Pass the skb back up. Network stack will deallocate the skb when it
 287          * is done.
 288          * TODO - use NAPI?
 289          */
 290         netif_rx(skb);
 291
 292         return 0;
 293 }
 294
 295 static void netvsc_get_drvinfo(struct net_device *net,
 296                                struct ethtool_drvinfo *info)
 297 {
 298         strcpy(info->driver, "hv_netvsc");
 299         strcpy(info->version, HV_DRV_VERSION);
 300         strcpy(info->fw_version, "N/A");
 301 }
 302
 303 static const struct ethtool_ops ethtool_ops = {
 304         .get_drvinfo    = netvsc_get_drvinfo,
 305         .get_link       = ethtool_op_get_link,
 306 };
 307
 308 static const struct net_device_ops device_ops = {
 309         .ndo_open =                     netvsc_open,
 310         .ndo_stop =                     netvsc_close,
 311         .ndo_start_xmit =               netvsc_start_xmit,
 312         .ndo_set_rx_mode =              netvsc_set_multicast_list,
 313         .ndo_change_mtu =               eth_change_mtu,
 314         .ndo_validate_addr =            eth_validate_addr,
 315         .ndo_set_mac_address =          eth_mac_addr,
 316 };
 317
 318 /*
 319  * Send GARP packet to network peers after migrations.
 320  * After Quick Migration, the network is not immediately operational in the
 321  * current context when receiving RNDIS_STATUS_MEDIA_CONNECT event. So, add
 322  * another netif_notify_peers() into a delayed work, otherwise GARP packet
 323  * will not be sent after quick migration, and cause network disconnection.
 324  */
 325 static void netvsc_send_garp(struct work_struct *w)
 326 {
 327         struct net_device_context *ndev_ctx;
 328         struct net_device *net;
 329         struct netvsc_device *net_device;
 330
 331         ndev_ctx = container_of(w, struct net_device_context, dwork.work);
 332         net_device = hv_get_drvdata(ndev_ctx->device_ctx);
 333         net = net_device->ndev;
 334         netif_notify_peers(net);
 335 }
 336
 337
 338 static int netvsc_probe(struct hv_device *dev,
 339                         const struct hv_vmbus_device_id *dev_id)
 340 {
 341         struct net_device *net = NULL;
 342         struct net_device_context *net_device_ctx;
 343         struct netvsc_device_info device_info;
 344         int ret;
 345
 346         net = alloc_etherdev(sizeof(struct net_device_context));
 347         if (!net)
 348                 return -ENOMEM;
 349
 350         /* Set initial state */
 351         netif_carrier_off(net);
 352
 353         net_device_ctx = netdev_priv(net);
 354         net_device_ctx->device_ctx = dev;
 355         hv_set_drvdata(dev, net);
 356         INIT_DELAYED_WORK(&net_device_ctx->dwork, netvsc_send_garp);
 357
 358         net->netdev_ops = &device_ops;
 359
 360         /* TODO: Add GSO and Checksum offload */
 361         net->hw_features = NETIF_F_SG;
 362         net->features = NETIF_F_SG;
 363
 364         SET_ETHTOOL_OPS(net, &ethtool_ops);
 365         SET_NETDEV_DEV(net, &dev->device);
 366
 367         ret = register_netdev(net);
 368         if (ret != 0) {
 369                 pr_err("Unable to register netdev.\n");
 370                 free_netdev(net);
 371                 goto out;
 372         }
 373
 374         /* Notify the netvsc driver of the new device */
 375         device_info.ring_size = ring_size;
 376         ret = rndis_filter_device_add(dev, &device_info);
 377         if (ret != 0) {
 378                 netdev_err(net, "unable to add netvsc device (ret %d)\n", ret);
 379                 unregister_netdev(net);
 380                 free_netdev(net);
 381                 hv_set_drvdata(dev, NULL);
 382                 return ret;
 383         }
 384         memcpy(net->dev_addr, device_info.mac_adr, ETH_ALEN);
 385
 386         netif_carrier_on(net);
 387
 388 out:
 389         return ret;
 390 }
 391
 392 static int netvsc_remove(struct hv_device *dev)
 393 {
 394         struct net_device *net;
 395         struct net_device_context *ndev_ctx;
 396         struct netvsc_device *net_device;
 397
 398         net_device = hv_get_drvdata(dev);
 399         net = net_device->ndev;
 400
 401         if (net == NULL) {
 402                 dev_err(&dev->device, "No net device to remove\n");
 403                 return 0;
 404         }
 405
 406         ndev_ctx = netdev_priv(net);
 407         cancel_delayed_work_sync(&ndev_ctx->dwork);
 408
 409         /* Stop outbound asap */
 410         netif_stop_queue(net);
 411
 412         unregister_netdev(net);
 413
 414         /*
 415          * Call to the vsc driver to let it know that the device is being
 416          * removed
 417          */
 418         rndis_filter_device_remove(dev);
 419
 420         free_netdev(net);
 421         return 0;
 422 }
 423
 424 static const struct hv_vmbus_device_id id_table[] = {
 425         /* Network guid */
 426         { VMBUS_DEVICE(0x63, 0x51, 0x61, 0xF8, 0x3E, 0xDF, 0xc5, 0x46,
 427                        0x91, 0x3F, 0xF2, 0xD2, 0xF9, 0x65, 0xED, 0x0E) },
 428         { },
 429 };
 430
 431 MODULE_DEVICE_TABLE(vmbus, id_table);
 432
 433 /* The one and only one */
 434 static struct  hv_driver netvsc_drv = {
 435         .name = "netvsc",
 436         .id_table = id_table,
 437         .probe = netvsc_probe,
 438         .remove = netvsc_remove,
 439 };
 440
 441 static void __exit netvsc_drv_exit(void)
 442 {
 443         vmbus_driver_unregister(&netvsc_drv);
 444 }
 445
 446 static int __init netvsc_drv_init(void)
 447 {
 448         return vmbus_driver_register(&netvsc_drv);
 449 }
 450
 451 MODULE_LICENSE("GPL");
 452 MODULE_VERSION(HV_DRV_VERSION);
 453 MODULE_DESCRIPTION("Microsoft Hyper-V network driver");
 454
 455 module_init(netvsc_drv_init);
 456 module_exit(netvsc_drv_exit);