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         struct work_struct work;
  48 };
  49
  50
  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)");
  54
  55 static void do_set_multicast(struct work_struct *w)
  56 {
  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;
  61
  62         nvdev = hv_get_drvdata(ndevctx->device_ctx);
  63         if (nvdev == NULL || nvdev->ndev == NULL)
  64                 return;
  65
  66         rdev = nvdev->extension;
  67         if (rdev == NULL)
  68                 return;
  69
  70         if (nvdev->ndev->flags & IFF_PROMISC)
  71                 rndis_filter_set_packet_filter(rdev,
  72                         NDIS_PACKET_TYPE_PROMISCUOUS);
  73         else
  74                 rndis_filter_set_packet_filter(rdev,
  75                         NDIS_PACKET_TYPE_BROADCAST |
  76                         NDIS_PACKET_TYPE_ALL_MULTICAST |
  77                         NDIS_PACKET_TYPE_DIRECTED);
  78 }
  79
  80 static void netvsc_set_multicast_list(struct net_device *net)
  81 {
  82         struct net_device_context *net_device_ctx = netdev_priv(net);
  83
  84         schedule_work(&net_device_ctx->work);
  85 }
  86
  87 static int netvsc_open(struct net_device *net)
  88 {
  89         struct net_device_context *net_device_ctx = netdev_priv(net);
  90         struct hv_device *device_obj = net_device_ctx->device_ctx;
  91         int ret = 0;
  92
  93         /* Open up the device */
  94         ret = rndis_filter_open(device_obj);
  95         if (ret != 0) {
  96                 netdev_err(net, "unable to open device (ret %d).\n", ret);
  97                 return ret;
  98         }
  99
 100         netif_start_queue(net);
 101
 102         return ret;
 103 }
 104
 105 static int netvsc_close(struct net_device *net)
 106 {
 107         struct net_device_context *net_device_ctx = netdev_priv(net);
 108         struct hv_device *device_obj = net_device_ctx->device_ctx;
 109         int ret;
 110
 111         netif_tx_disable(net);
 112
 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);
 116         if (ret != 0)
 117                 netdev_err(net, "unable to close device (ret %d).\n", ret);
 118
 119         return ret;
 120 }
 121
 122 static void netvsc_xmit_completion(void *context)
 123 {
 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;
 127
 128         kfree(packet);
 129
 130         if (skb)
 131                 dev_kfree_skb_any(skb);
 132 }
 133
 134 static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
 135 {
 136         struct net_device_context *net_device_ctx = netdev_priv(net);
 137         struct hv_netvsc_packet *packet;
 138         int ret;
 139         unsigned int i, num_pages, npg_data;
 140
 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;
 145
 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);
 151         if (!packet) {
 152                 /* out of memory, drop packet */
 153                 netdev_err(net, "unable to allocate hv_netvsc_packet\n");
 154
 155                 dev_kfree_skb(skb);
 156                 net->stats.tx_dropped++;
 157                 return NETDEV_TX_OK;
 158         }
 159
 160         packet->vlan_tci = skb->vlan_tci;
 161
 162         packet->extension = (void *)(unsigned long)packet +
 163                                 sizeof(struct hv_netvsc_packet) +
 164                                     (num_pages * sizeof(struct hv_page_buffer));
 165
 166         /* If the rndis msg goes beyond 1 page, we will add 1 later */
 167         packet->page_buf_cnt = num_pages - 1;
 168
 169         /* Initialize it from the skb */
 170         packet->total_data_buflen = skb->len;
 171
 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);
 176         if (npg_data == 1)
 177                 packet->page_buf[1].len = skb_headlen(skb);
 178         else
 179                 packet->page_buf[1].len = PAGE_SIZE
 180                         - packet->page_buf[1].offset;
 181
 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;
 187         }
 188         if (npg_data > 1)
 189                 packet->page_buf[npg_data].len = (((unsigned long)skb->data
 190                         + skb_headlen(skb) - 1) & (PAGE_SIZE - 1)) + 1;
 191
 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];
 195
 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);
 200         }
 201
 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;
 206
 207         ret = rndis_filter_send(net_device_ctx->device_ctx,
 208                                   packet);
 209         if (ret == 0) {
 210                 net->stats.tx_bytes += skb->len;
 211                 net->stats.tx_packets++;
 212         } else {
 213                 kfree(packet);
 214         }
 215
 216         return ret ? NETDEV_TX_BUSY : NETDEV_TX_OK;
 217 }
 218
 219 /*
 220  * netvsc_linkstatus_callback - Link up/down notification
 221  */
 222 void netvsc_linkstatus_callback(struct hv_device *device_obj,
 223                                        unsigned int status)
 224 {
 225         struct net_device *net;
 226         struct net_device_context *ndev_ctx;
 227         struct netvsc_device *net_device;
 228
 229         net_device = hv_get_drvdata(device_obj);
 230         net = net_device->ndev;
 231
 232         if (!net) {
 233                 netdev_err(net, "got link status but net device "
 234                                 "not initialized yet\n");
 235                 return;
 236         }
 237
 238         if (status == 1) {
 239                 netif_carrier_on(net);
 240                 netif_wake_queue(net);
 241                 ndev_ctx = netdev_priv(net);
 242                 schedule_delayed_work(&ndev_ctx->dwork, 0);
 243                 schedule_delayed_work(&ndev_ctx->dwork, msecs_to_jiffies(20));
 244         } else {
 245                 netif_carrier_off(net);
 246                 netif_tx_disable(net);
 247         }
 248 }
 249
 250 /*
 251  * netvsc_recv_callback -  Callback when we receive a packet from the
 252  * "wire" on the specified device.
 253  */
 254 int netvsc_recv_callback(struct hv_device *device_obj,
 255                                 struct hv_netvsc_packet *packet)
 256 {
 257         struct net_device *net;
 258         struct sk_buff *skb;
 259
 260         net = ((struct netvsc_device *)hv_get_drvdata(device_obj))->ndev;
 261         if (!net) {
 262                 netdev_err(net, "got receive callback but net device"
 263                         " not initialized yet\n");
 264                 return 0;
 265         }
 266
 267         /* Allocate a skb - TODO direct I/O to pages? */
 268         skb = netdev_alloc_skb_ip_align(net, packet->total_data_buflen);
 269         if (unlikely(!skb)) {
 270                 ++net->stats.rx_dropped;
 271                 return 0;
 272         }
 273
 274         /*
 275          * Copy to skb. This copy is needed here since the memory pointed by
 276          * hv_netvsc_packet cannot be deallocated
 277          */
 278         memcpy(skb_put(skb, packet->total_data_buflen), packet->data,
 279                 packet->total_data_buflen);
 280
 281         skb->protocol = eth_type_trans(skb, net);
 282         skb->ip_summed = CHECKSUM_NONE;
 283         skb->vlan_tci = packet->vlan_tci;
 284
 285         net->stats.rx_packets++;
 286         net->stats.rx_bytes += packet->total_data_buflen;
 287
 288         /*
 289          * Pass the skb back up. Network stack will deallocate the skb when it
 290          * is done.
 291          * TODO - use NAPI?
 292          */
 293         netif_rx(skb);
 294
 295         return 0;
 296 }
 297
 298 static void netvsc_get_drvinfo(struct net_device *net,
 299                                struct ethtool_drvinfo *info)
 300 {
 301         strcpy(info->driver, KBUILD_MODNAME);
 302         strcpy(info->version, HV_DRV_VERSION);
 303         strcpy(info->fw_version, "N/A");
 304 }
 305
 306 static int netvsc_change_mtu(struct net_device *ndev, int mtu)
 307 {
 308         struct net_device_context *ndevctx = netdev_priv(ndev);
 309         struct hv_device *hdev =  ndevctx->device_ctx;
 310         struct netvsc_device *nvdev = hv_get_drvdata(hdev);
 311         struct netvsc_device_info device_info;
 312         int limit = ETH_DATA_LEN;
 313
 314         if (nvdev == NULL || nvdev->destroy)
 315                 return -ENODEV;
 316
 317         if (nvdev->nvsp_version == NVSP_PROTOCOL_VERSION_2)
 318                 limit = NETVSC_MTU;
 319
 320         if (mtu < 68 || mtu > limit)
 321                 return -EINVAL;
 322
 323         nvdev->start_remove = true;
 324         cancel_delayed_work_sync(&ndevctx->dwork);
 325         cancel_work_sync(&ndevctx->work);
 326         netif_tx_disable(ndev);
 327         rndis_filter_device_remove(hdev);
 328
 329         ndev->mtu = mtu;
 330
 331         ndevctx->device_ctx = hdev;
 332         hv_set_drvdata(hdev, ndev);
 333         device_info.ring_size = ring_size;
 334         rndis_filter_device_add(hdev, &device_info);
 335         netif_wake_queue(ndev);
 336
 337         return 0;
 338 }
 339
 340 static const struct ethtool_ops ethtool_ops = {
 341         .get_drvinfo    = netvsc_get_drvinfo,
 342         .get_link       = ethtool_op_get_link,
 343 };
 344
 345 static const struct net_device_ops device_ops = {
 346         .ndo_open =                     netvsc_open,
 347         .ndo_stop =                     netvsc_close,
 348         .ndo_start_xmit =               netvsc_start_xmit,
 349         .ndo_set_rx_mode =              netvsc_set_multicast_list,
 350         .ndo_change_mtu =               netvsc_change_mtu,
 351         .ndo_validate_addr =            eth_validate_addr,
 352         .ndo_set_mac_address =          eth_mac_addr,
 353 };
 354
 355 /*
 356  * Send GARP packet to network peers after migrations.
 357  * After Quick Migration, the network is not immediately operational in the
 358  * current context when receiving RNDIS_STATUS_MEDIA_CONNECT event. So, add
 359  * another netif_notify_peers() into a delayed work, otherwise GARP packet
 360  * will not be sent after quick migration, and cause network disconnection.
 361  */
 362 static void netvsc_send_garp(struct work_struct *w)
 363 {
 364         struct net_device_context *ndev_ctx;
 365         struct net_device *net;
 366         struct netvsc_device *net_device;
 367
 368         ndev_ctx = container_of(w, struct net_device_context, dwork.work);
 369         net_device = hv_get_drvdata(ndev_ctx->device_ctx);
 370         net = net_device->ndev;
 371         netif_notify_peers(net);
 372 }
 373
 374
 375 static int netvsc_probe(struct hv_device *dev,
 376                         const struct hv_vmbus_device_id *dev_id)
 377 {
 378         struct net_device *net = NULL;
 379         struct net_device_context *net_device_ctx;
 380         struct netvsc_device_info device_info;
 381         int ret;
 382
 383         net = alloc_etherdev(sizeof(struct net_device_context));
 384         if (!net)
 385                 return -ENOMEM;
 386
 387         /* Set initial state */
 388         netif_carrier_off(net);
 389
 390         net_device_ctx = netdev_priv(net);
 391         net_device_ctx->device_ctx = dev;
 392         hv_set_drvdata(dev, net);
 393         INIT_DELAYED_WORK(&net_device_ctx->dwork, netvsc_send_garp);
 394         INIT_WORK(&net_device_ctx->work, do_set_multicast);
 395
 396         net->netdev_ops = &device_ops;
 397
 398         /* TODO: Add GSO and Checksum offload */
 399         net->hw_features = NETIF_F_SG;
 400         net->features = NETIF_F_SG | NETIF_F_HW_VLAN_TX;
 401
 402         SET_ETHTOOL_OPS(net, &ethtool_ops);
 403         SET_NETDEV_DEV(net, &dev->device);
 404
 405         ret = register_netdev(net);
 406         if (ret != 0) {
 407                 pr_err("Unable to register netdev.\n");
 408                 free_netdev(net);
 409                 goto out;
 410         }
 411
 412         /* Notify the netvsc driver of the new device */
 413         device_info.ring_size = ring_size;
 414         ret = rndis_filter_device_add(dev, &device_info);
 415         if (ret != 0) {
 416                 netdev_err(net, "unable to add netvsc device (ret %d)\n", ret);
 417                 unregister_netdev(net);
 418                 free_netdev(net);
 419                 hv_set_drvdata(dev, NULL);
 420                 return ret;
 421         }
 422         memcpy(net->dev_addr, device_info.mac_adr, ETH_ALEN);
 423
 424         netif_carrier_on(net);
 425
 426 out:
 427         return ret;
 428 }
 429
 430 static int netvsc_remove(struct hv_device *dev)
 431 {
 432         struct net_device *net;
 433         struct net_device_context *ndev_ctx;
 434         struct netvsc_device *net_device;
 435
 436         net_device = hv_get_drvdata(dev);
 437         net = net_device->ndev;
 438
 439         if (net == NULL) {
 440                 dev_err(&dev->device, "No net device to remove\n");
 441                 return 0;
 442         }
 443
 444         net_device->start_remove = true;
 445
 446         ndev_ctx = netdev_priv(net);
 447         cancel_delayed_work_sync(&ndev_ctx->dwork);
 448         cancel_work_sync(&ndev_ctx->work);
 449
 450         /* Stop outbound asap */
 451         netif_tx_disable(net);
 452
 453         unregister_netdev(net);
 454
 455         /*
 456          * Call to the vsc driver to let it know that the device is being
 457          * removed
 458          */
 459         rndis_filter_device_remove(dev);
 460
 461         free_netdev(net);
 462         return 0;
 463 }
 464
 465 static const struct hv_vmbus_device_id id_table[] = {
 466         /* Network guid */
 467         { VMBUS_DEVICE(0x63, 0x51, 0x61, 0xF8, 0x3E, 0xDF, 0xc5, 0x46,
 468                        0x91, 0x3F, 0xF2, 0xD2, 0xF9, 0x65, 0xED, 0x0E) },
 469         { },
 470 };
 471
 472 MODULE_DEVICE_TABLE(vmbus, id_table);
 473
 474 /* The one and only one */
 475 static struct  hv_driver netvsc_drv = {
 476         .name = KBUILD_MODNAME,
 477         .id_table = id_table,
 478         .probe = netvsc_probe,
 479         .remove = netvsc_remove,
 480 };
 481
 482 static void __exit netvsc_drv_exit(void)
 483 {
 484         vmbus_driver_unregister(&netvsc_drv);
 485 }
 486
 487 static int __init netvsc_drv_init(void)
 488 {
 489         return vmbus_driver_register(&netvsc_drv);
 490 }
 491
 492 MODULE_LICENSE("GPL");
 493 MODULE_VERSION(HV_DRV_VERSION);
 494 MODULE_DESCRIPTION("Microsoft Hyper-V network driver");
 495
 496 module_init(netvsc_drv_init);
 497 module_exit(netvsc_drv_exit);