drivers/staging/hv/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/module.h>
  25 #include <linux/highmem.h>
  26 #include <linux/device.h>
  27 #include <linux/io.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/in.h>
  34 #include <linux/slab.h>
  35 #include <linux/dmi.h>
  36 #include <linux/pci.h>
  37 #include <net/arp.h>
  38 #include <net/route.h>
  39 #include <net/sock.h>
  40 #include <net/pkt_sched.h>
  41
  42 #include "hyperv.h"
  43 #include "hyperv_net.h"
  44
  45 struct net_device_context {
  46         /* point back to our device context */
  47         struct hv_device *device_ctx;
  48         unsigned long avail;
  49         struct work_struct work;
  50 };
  51
  52
  53 #define PACKET_PAGES_LOWATER  8
  54 /* Need this many pages to handle worst case fragmented packet */
  55 #define PACKET_PAGES_HIWATER  (MAX_SKB_FRAGS + 2)
  56
  57 static int ring_size = 128;
  58 module_param(ring_size, int, S_IRUGO);
  59 MODULE_PARM_DESC(ring_size, "Ring buffer size (# of pages)");
  60
  61 /* no-op so the netdev core doesn't return -EINVAL when modifying the the
  62  * multicast address list in SIOCADDMULTI. hv is setup to get all multicast
  63  * when it calls RndisFilterOnOpen() */
  64 static void netvsc_set_multicast_list(struct net_device *net)
  65 {
  66 }
  67
  68 static int netvsc_open(struct net_device *net)
  69 {
  70         struct net_device_context *net_device_ctx = netdev_priv(net);
  71         struct hv_device *device_obj = net_device_ctx->device_ctx;
  72         int ret = 0;
  73
  74         if (netif_carrier_ok(net)) {
  75                 /* Open up the device */
  76                 ret = rndis_filter_open(device_obj);
  77                 if (ret != 0) {
  78                         netdev_err(net, "unable to open device (ret %d).\n",
  79                                    ret);
  80                         return ret;
  81                 }
  82
  83                 netif_start_queue(net);
  84         } else {
  85                 netdev_err(net, "unable to open device...link is down.\n");
  86         }
  87
  88         return ret;
  89 }
  90
  91 static int netvsc_close(struct net_device *net)
  92 {
  93         struct net_device_context *net_device_ctx = netdev_priv(net);
  94         struct hv_device *device_obj = net_device_ctx->device_ctx;
  95         int ret;
  96
  97         netif_stop_queue(net);
  98
  99         ret = rndis_filter_close(device_obj);
 100         if (ret != 0)
 101                 netdev_err(net, "unable to close device (ret %d).\n", ret);
 102
 103         return ret;
 104 }
 105
 106 static void netvsc_xmit_completion(void *context)
 107 {
 108         struct hv_netvsc_packet *packet = (struct hv_netvsc_packet *)context;
 109         struct sk_buff *skb = (struct sk_buff *)
 110                 (unsigned long)packet->completion.send.send_completion_tid;
 111
 112         kfree(packet);
 113
 114         if (skb) {
 115                 struct net_device *net = skb->dev;
 116                 struct net_device_context *net_device_ctx = netdev_priv(net);
 117                 unsigned int num_pages = skb_shinfo(skb)->nr_frags + 2;
 118
 119                 dev_kfree_skb_any(skb);
 120
 121                 net_device_ctx->avail += num_pages;
 122                 if (net_device_ctx->avail >= PACKET_PAGES_HIWATER)
 123                         netif_wake_queue(net);
 124         }
 125 }
 126
 127 static int netvsc_start_xmit(struct sk_buff *skb, struct net_device *net)
 128 {
 129         struct net_device_context *net_device_ctx = netdev_priv(net);
 130         struct hv_netvsc_packet *packet;
 131         int ret;
 132         unsigned int i, num_pages;
 133
 134         /* Add 1 for skb->data and additional one for RNDIS */
 135         num_pages = skb_shinfo(skb)->nr_frags + 1 + 1;
 136         if (num_pages > net_device_ctx->avail)
 137                 return NETDEV_TX_BUSY;
 138
 139         /* Allocate a netvsc packet based on # of frags. */
 140         packet = kzalloc(sizeof(struct hv_netvsc_packet) +
 141                          (num_pages * sizeof(struct hv_page_buffer)) +
 142                          sizeof(struct rndis_filter_packet), GFP_ATOMIC);
 143         if (!packet) {
 144                 /* out of memory, silently drop packet */
 145                 netdev_err(net, "unable to allocate hv_netvsc_packet\n");
 146
 147                 dev_kfree_skb(skb);
 148                 net->stats.tx_dropped++;
 149                 return NETDEV_TX_OK;
 150         }
 151
 152         packet->extension = (void *)(unsigned long)packet +
 153                                 sizeof(struct hv_netvsc_packet) +
 154                                     (num_pages * sizeof(struct hv_page_buffer));
 155
 156         /* Setup the rndis header */
 157         packet->page_buf_cnt = num_pages;
 158
 159         /* TODO: Flush all write buffers/ memory fence ??? */
 160         /* wmb(); */
 161
 162         /* Initialize it from the skb */
 163         packet->total_data_buflen       = skb->len;
 164
 165         /* Start filling in the page buffers starting after RNDIS buffer. */
 166         packet->page_buf[1].pfn = virt_to_phys(skb->data) >> PAGE_SHIFT;
 167         packet->page_buf[1].offset
 168                 = (unsigned long)skb->data & (PAGE_SIZE - 1);
 169         packet->page_buf[1].len = skb_headlen(skb);
 170
 171         /* Additional fragments are after SKB data */
 172         for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
 173                 skb_frag_t *f = &skb_shinfo(skb)->frags[i];
 174
 175                 packet->page_buf[i+2].pfn = page_to_pfn(f->page);
 176                 packet->page_buf[i+2].offset = f->page_offset;
 177                 packet->page_buf[i+2].len = f->size;
 178         }
 179
 180         /* Set the completion routine */
 181         packet->completion.send.send_completion = netvsc_xmit_completion;
 182         packet->completion.send.send_completion_ctx = packet;
 183         packet->completion.send.send_completion_tid = (unsigned long)skb;
 184
 185         ret = rndis_filter_send(net_device_ctx->device_ctx,
 186                                   packet);
 187         if (ret == 0) {
 188                 net->stats.tx_bytes += skb->len;
 189                 net->stats.tx_packets++;
 190
 191                 net_device_ctx->avail -= num_pages;
 192                 if (net_device_ctx->avail < PACKET_PAGES_LOWATER)
 193                         netif_stop_queue(net);
 194         } else {
 195                 /* we are shutting down or bus overloaded, just drop packet */
 196                 net->stats.tx_dropped++;
 197                 netvsc_xmit_completion(packet);
 198         }
 199
 200         return NETDEV_TX_OK;
 201 }
 202
 203 /*
 204  * netvsc_linkstatus_callback - Link up/down notification
 205  */
 206 void netvsc_linkstatus_callback(struct hv_device *device_obj,
 207                                        unsigned int status)
 208 {
 209         struct net_device *net = dev_get_drvdata(&device_obj->device);
 210         struct net_device_context *ndev_ctx;
 211
 212         if (!net) {
 213                 netdev_err(net, "got link status but net device "
 214                                 "not initialized yet\n");
 215                 return;
 216         }
 217
 218         if (status == 1) {
 219                 netif_carrier_on(net);
 220                 netif_wake_queue(net);
 221                 netif_notify_peers(net);
 222                 ndev_ctx = netdev_priv(net);
 223                 schedule_work(&ndev_ctx->work);
 224         } else {
 225                 netif_carrier_off(net);
 226                 netif_stop_queue(net);
 227         }
 228 }
 229
 230 /*
 231  * netvsc_recv_callback -  Callback when we receive a packet from the
 232  * "wire" on the specified device.
 233  */
 234 int netvsc_recv_callback(struct hv_device *device_obj,
 235                                 struct hv_netvsc_packet *packet)
 236 {
 237         struct net_device *net = dev_get_drvdata(&device_obj->device);
 238         struct sk_buff *skb;
 239         void *data;
 240         int i;
 241         unsigned long flags;
 242
 243         if (!net) {
 244                 netdev_err(net, "got receive callback but net device"
 245                         " not initialized yet\n");
 246                 return 0;
 247         }
 248
 249         /* Allocate a skb - TODO direct I/O to pages? */
 250         skb = netdev_alloc_skb_ip_align(net, packet->total_data_buflen);
 251         if (unlikely(!skb)) {
 252                 ++net->stats.rx_dropped;
 253                 return 0;
 254         }
 255
 256         /* for kmap_atomic */
 257         local_irq_save(flags);
 258
 259         /*
 260          * Copy to skb. This copy is needed here since the memory pointed by
 261          * hv_netvsc_packet cannot be deallocated
 262          */
 263         for (i = 0; i < packet->page_buf_cnt; i++) {
 264                 data = kmap_atomic(pfn_to_page(packet->page_buf[i].pfn),
 265                                                KM_IRQ1);
 266                 data = (void *)(unsigned long)data +
 267                                 packet->page_buf[i].offset;
 268
 269                 memcpy(skb_put(skb, packet->page_buf[i].len), data,
 270                        packet->page_buf[i].len);
 271
 272                 kunmap_atomic((void *)((unsigned long)data -
 273                                        packet->page_buf[i].offset), KM_IRQ1);
 274         }
 275
 276         local_irq_restore(flags);
 277
 278         skb->protocol = eth_type_trans(skb, net);
 279         skb->ip_summed = CHECKSUM_NONE;
 280
 281         net->stats.rx_packets++;
 282         net->stats.rx_bytes += skb->len;
 283
 284         /*
 285          * Pass the skb back up. Network stack will deallocate the skb when it
 286          * is done.
 287          * TODO - use NAPI?
 288          */
 289         netif_rx(skb);
 290
 291         return 0;
 292 }
 293
 294 static void netvsc_get_drvinfo(struct net_device *net,
 295                                struct ethtool_drvinfo *info)
 296 {
 297         strcpy(info->driver, "hv_netvsc");
 298         strcpy(info->version, HV_DRV_VERSION);
 299         strcpy(info->fw_version, "N/A");
 300 }
 301
 302 static const struct ethtool_ops ethtool_ops = {
 303         .get_drvinfo    = netvsc_get_drvinfo,
 304         .get_link       = ethtool_op_get_link,
 305 };
 306
 307 static const struct net_device_ops device_ops = {
 308         .ndo_open =                     netvsc_open,
 309         .ndo_stop =                     netvsc_close,
 310         .ndo_start_xmit =               netvsc_start_xmit,
 311         .ndo_set_multicast_list =       netvsc_set_multicast_list,
 312         .ndo_change_mtu =               eth_change_mtu,
 313         .ndo_validate_addr =            eth_validate_addr,
 314         .ndo_set_mac_address =          eth_mac_addr,
 315 };
 316
 317 /*
 318  * Send GARP packet to network peers after migrations.
 319  * After Quick Migration, the network is not immediately operational in the
 320  * current context when receiving RNDIS_STATUS_MEDIA_CONNECT event. So, add
 321  * another netif_notify_peers() into a scheduled work, otherwise GARP packet
 322  * will not be sent after quick migration, and cause network disconnection.
 323  */
 324 static void netvsc_send_garp(struct work_struct *w)
 325 {
 326         struct net_device_context *ndev_ctx;
 327         struct net_device *net;
 328
 329         msleep(20);
 330         ndev_ctx = container_of(w, struct net_device_context, work);
 331         net = dev_get_drvdata(&ndev_ctx->device_ctx->device);
 332         netif_notify_peers(net);
 333 }
 334
 335
 336 static int netvsc_probe(struct hv_device *dev)
 337 {
 338         struct net_device *net = NULL;
 339         struct net_device_context *net_device_ctx;
 340         struct netvsc_device_info device_info;
 341         int ret;
 342
 343         net = alloc_etherdev(sizeof(struct net_device_context));
 344         if (!net)
 345                 return -1;
 346
 347         /* Set initial state */
 348         netif_carrier_off(net);
 349
 350         net_device_ctx = netdev_priv(net);
 351         net_device_ctx->device_ctx = dev;
 352         net_device_ctx->avail = ring_size;
 353         dev_set_drvdata(&dev->device, net);
 354         INIT_WORK(&net_device_ctx->work, netvsc_send_garp);
 355
 356         /* Notify the netvsc driver of the new device */
 357         device_info.ring_size = ring_size;
 358         ret = rndis_filte_device_add(dev, &device_info);
 359         if (ret != 0) {
 360                 free_netdev(net);
 361                 dev_set_drvdata(&dev->device, NULL);
 362
 363                 netdev_err(net, "unable to add netvsc device (ret %d)\n", ret);
 364                 return ret;
 365         }
 366
 367         /*
 368          * If carrier is still off ie we did not get a link status callback,
 369          * update it if necessary
 370          */
 371         /*
 372          * FIXME: We should use a atomic or test/set instead to avoid getting
 373          * out of sync with the device's link status
 374          */
 375         if (!netif_carrier_ok(net))
 376                 if (!device_info.link_state)
 377                         netif_carrier_on(net);
 378
 379         memcpy(net->dev_addr, device_info.mac_adr, ETH_ALEN);
 380
 381         net->netdev_ops = &device_ops;
 382
 383         /* TODO: Add GSO and Checksum offload */
 384         net->hw_features = NETIF_F_SG;
 385         net->features = NETIF_F_SG;
 386
 387         SET_ETHTOOL_OPS(net, &ethtool_ops);
 388         SET_NETDEV_DEV(net, &dev->device);
 389
 390         ret = register_netdev(net);
 391         if (ret != 0) {
 392                 /* Remove the device and release the resource */
 393                 rndis_filter_device_remove(dev);
 394                 free_netdev(net);
 395         }
 396
 397         return ret;
 398 }
 399
 400 static int netvsc_remove(struct hv_device *dev)
 401 {
 402         struct net_device *net = dev_get_drvdata(&dev->device);
 403         int ret;
 404
 405         if (net == NULL) {
 406                 dev_err(&dev->device, "No net device to remove\n");
 407                 return 0;
 408         }
 409
 410         /* Stop outbound asap */
 411         netif_stop_queue(net);
 412         /* netif_carrier_off(net); */
 413
 414         unregister_netdev(net);
 415
 416         /*
 417          * Call to the vsc driver to let it know that the device is being
 418          * removed
 419          */
 420         ret = rndis_filter_device_remove(dev);
 421         if (ret != 0) {
 422                 /* TODO: */
 423                 netdev_err(net, "unable to remove vsc device (ret %d)\n", ret);
 424         }
 425
 426         free_netdev(net);
 427         return ret;
 428 }
 429
 430 /* The one and only one */
 431 static struct  hv_driver netvsc_drv = {
 432         .probe = netvsc_probe,
 433         .remove = netvsc_remove,
 434 };
 435
 436 static void __exit netvsc_drv_exit(void)
 437 {
 438         vmbus_child_driver_unregister(&netvsc_drv.driver);
 439 }
 440
 441
 442 static const struct dmi_system_id __initconst
 443 hv_netvsc_dmi_table[] __maybe_unused  = {
 444         {
 445                 .ident = "Hyper-V",
 446                 .matches = {
 447                         DMI_MATCH(DMI_SYS_VENDOR, "Microsoft Corporation"),
 448                         DMI_MATCH(DMI_PRODUCT_NAME, "Virtual Machine"),
 449                         DMI_MATCH(DMI_BOARD_NAME, "Virtual Machine"),
 450                 },
 451         },
 452         { },
 453 };
 454 MODULE_DEVICE_TABLE(dmi, hv_netvsc_dmi_table);
 455
 456 static int __init netvsc_drv_init(void)
 457 {
 458         struct hv_driver *drv = &netvsc_drv;
 459         int ret;
 460
 461         pr_info("initializing....");
 462
 463         if (!dmi_check_system(hv_netvsc_dmi_table))
 464                 return -ENODEV;
 465
 466
 467         /* Callback to client driver to complete the initialization */
 468         netvsc_initialize(drv);
 469
 470         drv->driver.name = drv->name;
 471
 472         /* The driver belongs to vmbus */
 473         ret = vmbus_child_driver_register(&drv->driver);
 474
 475         return ret;
 476 }
 477
 478 static const struct pci_device_id __initconst
 479 hv_netvsc_pci_table[] __maybe_unused = {
 480         { PCI_DEVICE(0x1414, 0x5353) }, /* VGA compatible controller */
 481         { 0 }
 482 };
 483 MODULE_DEVICE_TABLE(pci, hv_netvsc_pci_table);
 484
 485 MODULE_LICENSE("GPL");
 486 MODULE_VERSION(HV_DRV_VERSION);
 487 MODULE_DESCRIPTION("Microsoft Hyper-V network driver");
 488
 489 module_init(netvsc_drv_init);
 490 module_exit(netvsc_drv_exit);