sys/dev/hyperv/netvsc/hv_netvsc_drv_freebsd.c

   1 /*-
   2  * Copyright (c) 2010-2012 Citrix Inc.
   3  * Copyright (c) 2009-2012 Microsoft Corp.
   4  * Copyright (c) 2012 NetApp Inc.
   5  * All rights reserved.
   6  *
   7  * Redistribution and use in source and binary forms, with or without
   8  * modification, are permitted provided that the following conditions
   9  * are met:
  10  * 1. Redistributions of source code must retain the above copyright
  11  *    notice unmodified, this list of conditions, and the following
  12  *    disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  18  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
  19  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
  20  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
  21  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  22  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  23  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  24  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  25  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
  26  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  27  */
  28
  29 /*-
  30  * Copyright (c) 2004-2006 Kip Macy
  31  * All rights reserved.
  32  *
  33  * Redistribution and use in source and binary forms, with or without
  34  * modification, are permitted provided that the following conditions
  35  * are met:
  36  * 1. Redistributions of source code must retain the above copyright
  37  *    notice, this list of conditions and the following disclaimer.
  38  * 2. Redistributions in binary form must reproduce the above copyright
  39  *    notice, this list of conditions and the following disclaimer in the
  40  *    documentation and/or other materials provided with the distribution.
  41  *
  42  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
  43  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  44  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  45  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
  46  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  47  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  48  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  49  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  50  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  51  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  52  * SUCH DAMAGE.
  53  */
  54
  55 #include <sys/cdefs.h>
  56 __FBSDID("$FreeBSD$");
  57
  58 #include "opt_inet6.h"
  59 #include "opt_inet.h"
  60
  61 #include <sys/param.h>
  62 #include <sys/systm.h>
  63 #include <sys/sockio.h>
  64 #include <sys/mbuf.h>
  65 #include <sys/malloc.h>
  66 #include <sys/module.h>
  67 #include <sys/kernel.h>
  68 #include <sys/socket.h>
  69 #include <sys/queue.h>
  70 #include <sys/lock.h>
  71 #include <sys/sx.h>
  72
  73 #include <net/if.h>
  74 #include <net/if_arp.h>
  75 #include <net/ethernet.h>
  76 #include <net/if_dl.h>
  77 #include <net/if_media.h>
  78
  79 #include <net/bpf.h>
  80
  81 #include <net/if_types.h>
  82 #include <net/if_vlan_var.h>
  83 #include <net/if.h>
  84
  85 #include <netinet/in_systm.h>
  86 #include <netinet/in.h>
  87 #include <netinet/ip.h>
  88 #include <netinet/if_ether.h>
  89 #include <netinet/tcp.h>
  90 #include <netinet/udp.h>
  91 #include <netinet/ip6.h>
  92
  93 #include <vm/vm.h>
  94 #include <vm/vm_param.h>
  95 #include <vm/vm_kern.h>
  96 #include <vm/pmap.h>
  97
  98 #include <machine/bus.h>
  99 #include <machine/resource.h>
 100 #include <machine/frame.h>
 101 #include <machine/vmparam.h>
 102
 103 #include <sys/bus.h>
 104 #include <sys/rman.h>
 105 #include <sys/mutex.h>
 106 #include <sys/errno.h>
 107 #include <sys/types.h>
 108 #include <machine/atomic.h>
 109
 110 #include <machine/intr_machdep.h>
 111
 112 #include <machine/in_cksum.h>
 113
 114 #include <dev/hyperv/include/hyperv.h>
 115 #include "hv_net_vsc.h"
 116 #include "hv_rndis.h"
 117 #include "hv_rndis_filter.h"
 118
 119
 120 /* Short for Hyper-V network interface */
 121 #define NETVSC_DEVNAME    "hn"
 122
 123 /*
 124  * It looks like offset 0 of buf is reserved to hold the softc pointer.
 125  * The sc pointer evidently not needed, and is not presently populated.
 126  * The packet offset is where the netvsc_packet starts in the buffer.
 127  */
 128 #define HV_NV_SC_PTR_OFFSET_IN_BUF         0
 129 #define HV_NV_PACKET_OFFSET_IN_BUF         16
 130
 131 /*
 132  * A unified flag for all outbound check sum flags is useful,
 133  * and it helps avoiding unnecessary check sum calculation in
 134  * network forwarding scenario.
 135  */
 136 #define HV_CSUM_FOR_OUTBOUND                                            \
 137     (CSUM_IP|CSUM_IP_UDP|CSUM_IP_TCP|CSUM_IP_SCTP|CSUM_IP_TSO|          \
 138     CSUM_IP_ISCSI|CSUM_IP6_UDP|CSUM_IP6_TCP|CSUM_IP6_SCTP|              \
 139     CSUM_IP6_TSO|CSUM_IP6_ISCSI)
 140
 141 /*
 142  * Data types
 143  */
 144
 145 struct hv_netvsc_driver_context {
 146         uint32_t                drv_inited;
 147 };
 148
 149 /*
 150  * Be aware that this sleepable mutex will exhibit WITNESS errors when
 151  * certain TCP and ARP code paths are taken.  This appears to be a
 152  * well-known condition, as all other drivers checked use a sleeping
 153  * mutex to protect their transmit paths.
 154  * Also Be aware that mutexes do not play well with semaphores, and there
 155  * is a conflicting semaphore in a certain channel code path.
 156  */
 157 #define NV_LOCK_INIT(_sc, _name) \
 158             mtx_init(&(_sc)->hn_lock, _name, MTX_NETWORK_LOCK, MTX_DEF)
 159 #define NV_LOCK(_sc)            mtx_lock(&(_sc)->hn_lock)
 160 #define NV_LOCK_ASSERT(_sc)     mtx_assert(&(_sc)->hn_lock, MA_OWNED)
 161 #define NV_UNLOCK(_sc)          mtx_unlock(&(_sc)->hn_lock)
 162 #define NV_LOCK_DESTROY(_sc)    mtx_destroy(&(_sc)->hn_lock)
 163
 164
 165 /*
 166  * Globals
 167  */
 168
 169 int hv_promisc_mode = 0;    /* normal mode by default */
 170
 171 /* The one and only one */
 172 static struct hv_netvsc_driver_context g_netvsc_drv;
 173
 174
 175 /*
 176  * Forward declarations
 177  */
 178 static void hn_stop(hn_softc_t *sc);
 179 static void hn_ifinit_locked(hn_softc_t *sc);
 180 static void hn_ifinit(void *xsc);
 181 static int  hn_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data);
 182 static int  hn_start_locked(struct ifnet *ifp);
 183 static void hn_start(struct ifnet *ifp);
 184
 185 /*
 186  * NetVsc get message transport protocol type
 187  */
 188 static uint32_t get_transport_proto_type(struct mbuf *m_head)
 189 {
 190         uint32_t ret_val = TRANSPORT_TYPE_NOT_IP;
 191         uint16_t ether_type = 0;
 192         int ether_len = 0;
 193         struct ether_vlan_header *eh;
 194 #ifdef INET
 195         struct ip *iph;
 196 #endif
 197 #ifdef INET6
 198         struct ip6_hdr *ip6;
 199 #endif
 200
 201         eh = mtod(m_head, struct ether_vlan_header*);
 202         if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) {
 203                 ether_len = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN;
 204                 ether_type = eh->evl_proto;
 205         } else {
 206                 ether_len = ETHER_HDR_LEN;
 207                 ether_type = eh->evl_encap_proto;
 208         }
 209
 210         switch (ntohs(ether_type)) {
 211 #ifdef INET6
 212         case ETHERTYPE_IPV6:
 213                 ip6 = (struct ip6_hdr *)(m_head->m_data + ether_len);
 214
 215                 if (IPPROTO_TCP == ip6->ip6_nxt) {
 216                         ret_val = TRANSPORT_TYPE_IPV6_TCP;
 217                 } else if (IPPROTO_UDP == ip6->ip6_nxt) {
 218                         ret_val = TRANSPORT_TYPE_IPV6_UDP;
 219                 }
 220                 break;
 221 #endif
 222 #ifdef INET
 223         case ETHERTYPE_IP:
 224                 iph = (struct ip *)(m_head->m_data + ether_len);
 225
 226                 if (IPPROTO_TCP == iph->ip_p) {
 227                         ret_val = TRANSPORT_TYPE_IPV4_TCP;
 228                 } else if (IPPROTO_UDP == iph->ip_p) {
 229                         ret_val = TRANSPORT_TYPE_IPV4_UDP;
 230                 }
 231                 break;
 232 #endif
 233         default:
 234                 ret_val = TRANSPORT_TYPE_NOT_IP;
 235                 break;
 236         }
 237
 238         return (ret_val);
 239 }
 240
 241 /*
 242  * NetVsc driver initialization
 243  * Note:  Filter init is no longer required
 244  */
 245 static int
 246 netvsc_drv_init(void)
 247 {
 248         return (0);
 249 }
 250
 251 /*
 252  * NetVsc global initialization entry point
 253  */
 254 static void
 255 netvsc_init(void)
 256 {
 257         if (bootverbose)
 258                 printf("Netvsc initializing... ");
 259
 260         /*
 261          * XXXKYS: cleanup initialization
 262          */
 263         if (!cold && !g_netvsc_drv.drv_inited) {
 264                 g_netvsc_drv.drv_inited = 1;
 265                 netvsc_drv_init();
 266                 if (bootverbose)
 267                         printf("done!\n");
 268         } else if (bootverbose)
 269                 printf("Already initialized!\n");
 270 }
 271
 272 /* {F8615163-DF3E-46c5-913F-F2D2F965ED0E} */
 273 static const hv_guid g_net_vsc_device_type = {
 274         .data = {0x63, 0x51, 0x61, 0xF8, 0x3E, 0xDF, 0xc5, 0x46,
 275                 0x91, 0x3F, 0xF2, 0xD2, 0xF9, 0x65, 0xED, 0x0E}
 276 };
 277
 278 /*
 279  * Standard probe entry point.
 280  *
 281  */
 282 static int
 283 netvsc_probe(device_t dev)
 284 {
 285         const char *p;
 286
 287         p = vmbus_get_type(dev);
 288         if (!memcmp(p, &g_net_vsc_device_type.data, sizeof(hv_guid))) {
 289                 device_set_desc(dev, "Synthetic Network Interface");
 290                 if (bootverbose)
 291                         printf("Netvsc probe... DONE \n");
 292
 293                 return (BUS_PROBE_DEFAULT);
 294         }
 295
 296         return (ENXIO);
 297 }
 298
 299 /*
 300  * Standard attach entry point.
 301  *
 302  * Called when the driver is loaded.  It allocates needed resources,
 303  * and initializes the "hardware" and software.
 304  */
 305 static int
 306 netvsc_attach(device_t dev)
 307 {
 308         struct hv_device *device_ctx = vmbus_get_devctx(dev);
 309         netvsc_device_info device_info;
 310         hn_softc_t *sc;
 311         int unit = device_get_unit(dev);
 312         struct ifnet *ifp;
 313         int ret;
 314
 315         netvsc_init();
 316
 317         sc = device_get_softc(dev);
 318         if (sc == NULL) {
 319                 return (ENOMEM);
 320         }
 321
 322         bzero(sc, sizeof(hn_softc_t));
 323         sc->hn_unit = unit;
 324         sc->hn_dev = dev;
 325
 326         NV_LOCK_INIT(sc, "NetVSCLock");
 327
 328         sc->hn_dev_obj = device_ctx;
 329
 330         ifp = sc->hn_ifp = sc->arpcom.ac_ifp = if_alloc(IFT_ETHER);
 331         ifp->if_softc = sc;
 332
 333         if_initname(ifp, device_get_name(dev), device_get_unit(dev));
 334         ifp->if_dunit = unit;
 335         ifp->if_dname = NETVSC_DEVNAME;
 336
 337         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 338         ifp->if_ioctl = hn_ioctl;
 339         ifp->if_start = hn_start;
 340         ifp->if_init = hn_ifinit;
 341         /* needed by hv_rf_on_device_add() code */
 342         ifp->if_mtu = ETHERMTU;
 343         IFQ_SET_MAXLEN(&ifp->if_snd, 512);
 344         ifp->if_snd.ifq_drv_maxlen = 511;
 345         IFQ_SET_READY(&ifp->if_snd);
 346
 347         /*
 348          * Tell upper layers that we support full VLAN capability.
 349          */
 350         ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header);
 351         ifp->if_capabilities |=
 352             IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU | IFCAP_HWCSUM | IFCAP_TSO;
 353         ifp->if_capenable |=
 354             IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU | IFCAP_HWCSUM | IFCAP_TSO;
 355         /*
 356          * Only enable UDP checksum offloading when it is on 2012R2 or
 357          * later. UDP checksum offloading doesn't work on earlier
 358          * Windows releases.
 359          */
 360         if (hv_vmbus_protocal_version >= HV_VMBUS_VERSION_WIN8_1)
 361                 ifp->if_hwassist = CSUM_TCP | CSUM_UDP | CSUM_TSO;
 362         else
 363                 ifp->if_hwassist = CSUM_TCP | CSUM_TSO;
 364
 365         ret = hv_rf_on_device_add(device_ctx, &device_info);
 366         if (ret != 0) {
 367                 if_free(ifp);
 368
 369                 return (ret);
 370         }
 371         if (device_info.link_state == 0) {
 372                 sc->hn_carrier = 1;
 373         }
 374
 375         ether_ifattach(ifp, device_info.mac_addr);
 376
 377         return (0);
 378 }
 379
 380 /*
 381  * Standard detach entry point
 382  */
 383 static int
 384 netvsc_detach(device_t dev)
 385 {
 386         struct hv_device *hv_device = vmbus_get_devctx(dev);
 387
 388         if (bootverbose)
 389                 printf("netvsc_detach\n");
 390
 391         /*
 392          * XXXKYS:  Need to clean up all our
 393          * driver state; this is the driver
 394          * unloading.
 395          */
 396
 397         /*
 398          * XXXKYS:  Need to stop outgoing traffic and unregister
 399          * the netdevice.
 400          */
 401
 402         hv_rf_on_device_remove(hv_device, HV_RF_NV_DESTROY_CHANNEL);
 403
 404         return (0);
 405 }
 406
 407 /*
 408  * Standard shutdown entry point
 409  */
 410 static int
 411 netvsc_shutdown(device_t dev)
 412 {
 413         return (0);
 414 }
 415
 416 /*
 417  * Send completion processing
 418  *
 419  * Note:  It looks like offset 0 of buf is reserved to hold the softc
 420  * pointer.  The sc pointer is not currently needed in this function, and
 421  * it is not presently populated by the TX function.
 422  */
 423 void
 424 netvsc_xmit_completion(void *context)
 425 {
 426         netvsc_packet *packet = (netvsc_packet *)context;
 427         struct mbuf *mb;
 428         uint8_t *buf;
 429
 430         mb = (struct mbuf *)(uintptr_t)packet->compl.send.send_completion_tid;
 431         buf = ((uint8_t *)packet) - HV_NV_PACKET_OFFSET_IN_BUF;
 432
 433         free(buf, M_NETVSC);
 434
 435         if (mb != NULL) {
 436                 m_freem(mb);
 437         }
 438 }
 439
 440 /*
 441  * Start a transmit of one or more packets
 442  */
 443 static int
 444 hn_start_locked(struct ifnet *ifp)
 445 {
 446         hn_softc_t *sc = ifp->if_softc;
 447         struct hv_device *device_ctx = vmbus_get_devctx(sc->hn_dev);
 448         netvsc_dev *net_dev = sc->net_dev;
 449         device_t dev = device_ctx->device;
 450         uint8_t *buf;
 451         netvsc_packet *packet;
 452         struct mbuf *m_head, *m;
 453         struct mbuf *mc_head = NULL;
 454         struct ether_vlan_header *eh;
 455         rndis_msg *rndis_mesg;
 456         rndis_packet *rndis_pkt;
 457         rndis_per_packet_info *rppi;
 458         ndis_8021q_info *rppi_vlan_info;
 459         rndis_tcp_ip_csum_info *csum_info;
 460         rndis_tcp_tso_info *tso_info;
 461         int ether_len;
 462         int i;
 463         int num_frags;
 464         int len;
 465         int retries = 0;
 466         int ret = 0;
 467         uint32_t rndis_msg_size = 0;
 468         uint32_t trans_proto_type;
 469         uint32_t send_buf_section_idx =
 470             NVSP_1_CHIMNEY_SEND_INVALID_SECTION_INDEX;
 471
 472         while (!IFQ_DRV_IS_EMPTY(&sc->hn_ifp->if_snd)) {
 473                 IFQ_DRV_DEQUEUE(&sc->hn_ifp->if_snd, m_head);
 474                 if (m_head == NULL) {
 475                         break;
 476                 }
 477
 478                 len = 0;
 479                 num_frags = 0;
 480
 481                 /* Walk the mbuf list computing total length and num frags */
 482                 for (m = m_head; m != NULL; m = m->m_next) {
 483                         if (m->m_len != 0) {
 484                                 num_frags++;
 485                                 len += m->m_len;
 486                         }
 487                 }
 488
 489                 /*
 490                  * Reserve the number of pages requested.  Currently,
 491                  * one page is reserved for the message in the RNDIS
 492                  * filter packet
 493                  */
 494                 num_frags += HV_RF_NUM_TX_RESERVED_PAGE_BUFS;
 495
 496                 /* If exceeds # page_buffers in netvsc_packet */
 497                 if (num_frags > NETVSC_PACKET_MAXPAGE) {
 498                         device_printf(dev, "exceed max page buffers,%d,%d\n",
 499                             num_frags, NETVSC_PACKET_MAXPAGE);
 500                         m_freem(m_head);
 501                         if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
 502                         return (EINVAL);
 503                 }
 504
 505                 /*
 506                  * Allocate a buffer with space for a netvsc packet plus a
 507                  * number of reserved areas.  First comes a (currently 16
 508                  * bytes, currently unused) reserved data area.  Second is
 509                  * the netvsc_packet. Third is an area reserved for an
 510                  * rndis_filter_packet struct. Fourth (optional) is a
 511                  * rndis_per_packet_info struct.
 512                  * Changed malloc to M_NOWAIT to avoid sleep under spin lock.
 513                  * No longer reserving extra space for page buffers, as they
 514                  * are already part of the netvsc_packet.
 515                  */
 516                 buf = malloc(HV_NV_PACKET_OFFSET_IN_BUF +
 517                         sizeof(netvsc_packet) +
 518                         sizeof(rndis_msg) +
 519                         RNDIS_VLAN_PPI_SIZE +
 520                         RNDIS_TSO_PPI_SIZE +
 521                         RNDIS_CSUM_PPI_SIZE,
 522                         M_NETVSC, M_ZERO | M_NOWAIT);
 523                 if (buf == NULL) {
 524                         device_printf(dev, "hn:malloc packet failed\n");
 525                         m_freem(m_head);
 526                         if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
 527                         return (ENOMEM);
 528                 }
 529
 530                 packet = (netvsc_packet *)(buf + HV_NV_PACKET_OFFSET_IN_BUF);
 531                 *(vm_offset_t *)buf = HV_NV_SC_PTR_OFFSET_IN_BUF;
 532
 533                 packet->is_data_pkt = TRUE;
 534
 535                 /* Set up the rndis header */
 536                 packet->page_buf_count = num_frags;
 537
 538                 /* Initialize it from the mbuf */
 539                 packet->tot_data_buf_len = len;
 540
 541                 /*
 542                  * extension points to the area reserved for the
 543                  * rndis_filter_packet, which is placed just after
 544                  * the netvsc_packet (and rppi struct, if present;
 545                  * length is updated later).
 546                  */
 547                 packet->rndis_mesg = packet + 1;
 548                 rndis_mesg = (rndis_msg *)packet->rndis_mesg;
 549                 rndis_mesg->ndis_msg_type = REMOTE_NDIS_PACKET_MSG;
 550
 551                 rndis_pkt = &rndis_mesg->msg.packet;
 552                 rndis_pkt->data_offset = sizeof(rndis_packet);
 553                 rndis_pkt->data_length = packet->tot_data_buf_len;
 554                 rndis_pkt->per_pkt_info_offset = sizeof(rndis_packet);
 555
 556                 rndis_msg_size = RNDIS_MESSAGE_SIZE(rndis_packet);
 557
 558                 /*
 559                  * If the Hyper-V infrastructure needs to embed a VLAN tag,
 560                  * initialize netvsc_packet and rppi struct values as needed.
 561                  */
 562                 if (m_head->m_flags & M_VLANTAG) {
 563                         /*
 564                          * set up some additional fields so the Hyper-V infrastructure will stuff the VLAN tag
 565                          * into the frame.
 566                          */
 567                         packet->vlan_tci = m_head->m_pkthdr.ether_vtag;
 568
 569                         rndis_msg_size += RNDIS_VLAN_PPI_SIZE;
 570
 571                         rppi = hv_set_rppi_data(rndis_mesg, RNDIS_VLAN_PPI_SIZE,
 572                             ieee_8021q_info);
 573
 574                         /* VLAN info immediately follows rppi struct */
 575                         rppi_vlan_info = (ndis_8021q_info *)((char*)rppi +
 576                             rppi->per_packet_info_offset);
 577                         /* FreeBSD does not support CFI or priority */
 578                         rppi_vlan_info->u1.s1.vlan_id =
 579                             packet->vlan_tci & 0xfff;
 580                 }
 581
 582                 /* Only check the flags for outbound and ignore the ones for inbound */
 583                 if (0 == (m_head->m_pkthdr.csum_flags & HV_CSUM_FOR_OUTBOUND)) {
 584                         goto pre_send;
 585                 }
 586
 587                 eh = mtod(m_head, struct ether_vlan_header*);
 588                 if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) {
 589                         ether_len = ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN;
 590                 } else {
 591                         ether_len = ETHER_HDR_LEN;
 592                 }
 593
 594                 trans_proto_type = get_transport_proto_type(m_head);
 595                 if (TRANSPORT_TYPE_NOT_IP == trans_proto_type) {
 596                         goto pre_send;
 597                 }
 598
 599                 /*
 600                  * TSO packet needless to setup the send side checksum
 601                  * offload.
 602                  */
 603                 if (m_head->m_pkthdr.csum_flags & CSUM_TSO) {
 604                         goto do_tso;
 605                 }
 606
 607                 /* setup checksum offload */
 608                 rndis_msg_size += RNDIS_CSUM_PPI_SIZE;
 609                 rppi = hv_set_rppi_data(rndis_mesg, RNDIS_CSUM_PPI_SIZE,
 610                     tcpip_chksum_info);
 611                 csum_info = (rndis_tcp_ip_csum_info *)((char*)rppi +
 612                     rppi->per_packet_info_offset);
 613
 614                 if (trans_proto_type & (TYPE_IPV4 << 16)) {
 615                         csum_info->xmit.is_ipv4 = 1;
 616                 } else {
 617                         csum_info->xmit.is_ipv6 = 1;
 618                 }
 619
 620                 if (trans_proto_type & TYPE_TCP) {
 621                         csum_info->xmit.tcp_csum = 1;
 622                         csum_info->xmit.tcp_header_offset = 0;
 623                 } else if (trans_proto_type & TYPE_UDP) {
 624                         csum_info->xmit.udp_csum = 1;
 625                 }
 626
 627                 goto pre_send;
 628
 629 do_tso:
 630                 /* setup TCP segmentation offload */
 631                 rndis_msg_size += RNDIS_TSO_PPI_SIZE;
 632                 rppi = hv_set_rppi_data(rndis_mesg, RNDIS_TSO_PPI_SIZE,
 633                     tcp_large_send_info);
 634
 635                 tso_info = (rndis_tcp_tso_info *)((char *)rppi +
 636                     rppi->per_packet_info_offset);
 637                 tso_info->lso_v2_xmit.type =
 638                     RNDIS_TCP_LARGE_SEND_OFFLOAD_V2_TYPE;
 639
 640 #ifdef INET
 641                 if (trans_proto_type & (TYPE_IPV4 << 16)) {
 642                         struct ip *ip =
 643                             (struct ip *)(m_head->m_data + ether_len);
 644                         unsigned long iph_len = ip->ip_hl << 2;
 645                         struct tcphdr *th =
 646                             (struct tcphdr *)((caddr_t)ip + iph_len);
 647
 648                         tso_info->lso_v2_xmit.ip_version =
 649                             RNDIS_TCP_LARGE_SEND_OFFLOAD_IPV4;
 650                         ip->ip_len = 0;
 651                         ip->ip_sum = 0;
 652
 653                         th->th_sum = in_pseudo(ip->ip_src.s_addr,
 654                             ip->ip_dst.s_addr,
 655                             htons(IPPROTO_TCP));
 656                 }
 657 #endif
 658 #if defined(INET6) && defined(INET)
 659                 else
 660 #endif
 661 #ifdef INET6
 662                 {
 663                         struct ip6_hdr *ip6 =
 664                             (struct ip6_hdr *)(m_head->m_data + ether_len);
 665                         struct tcphdr *th = (struct tcphdr *)(ip6 + 1);
 666
 667                         tso_info->lso_v2_xmit.ip_version =
 668                             RNDIS_TCP_LARGE_SEND_OFFLOAD_IPV6;
 669                         ip6->ip6_plen = 0;
 670                         th->th_sum = in6_cksum_pseudo(ip6, 0, IPPROTO_TCP, 0);
 671                 }
 672 #endif
 673                 tso_info->lso_v2_xmit.tcp_header_offset = 0;
 674                 tso_info->lso_v2_xmit.mss = m_head->m_pkthdr.tso_segsz;
 675
 676 pre_send:
 677                 rndis_mesg->msg_len = packet->tot_data_buf_len + rndis_msg_size;
 678                 packet->tot_data_buf_len = rndis_mesg->msg_len;
 679
 680                 /* send packet with send buffer */
 681                 if (packet->tot_data_buf_len < net_dev->send_section_size) {
 682                         send_buf_section_idx =
 683                             hv_nv_get_next_send_section(net_dev);
 684                         if (send_buf_section_idx !=
 685                             NVSP_1_CHIMNEY_SEND_INVALID_SECTION_INDEX) {
 686                                 char *dest = ((char *)net_dev->send_buf +
 687                                     send_buf_section_idx *
 688                                     net_dev->send_section_size);
 689
 690                                 memcpy(dest, rndis_mesg, rndis_msg_size);
 691                                 dest += rndis_msg_size;
 692                                 for (m = m_head; m != NULL; m = m->m_next) {
 693                                         if (m->m_len) {
 694                                                 memcpy(dest,
 695                                                     (void *)mtod(m, vm_offset_t),
 696                                                     m->m_len);
 697                                                 dest += m->m_len;
 698                                         }
 699                                 }
 700
 701                                 packet->send_buf_section_idx =
 702                                     send_buf_section_idx;
 703                                 packet->send_buf_section_size =
 704                                     packet->tot_data_buf_len;
 705                                 packet->page_buf_count = 0;
 706                                 goto do_send;
 707                         }
 708                 }
 709
 710                 /* send packet with page buffer */
 711                 packet->page_buffers[0].pfn =
 712                     atop(hv_get_phys_addr(rndis_mesg));
 713                 packet->page_buffers[0].offset =
 714                     (unsigned long)rndis_mesg & PAGE_MASK;
 715                 packet->page_buffers[0].length = rndis_msg_size;
 716
 717                 /*
 718                  * Fill the page buffers with mbuf info starting at index
 719                  * HV_RF_NUM_TX_RESERVED_PAGE_BUFS.
 720                  */
 721                 i = HV_RF_NUM_TX_RESERVED_PAGE_BUFS;
 722                 for (m = m_head; m != NULL; m = m->m_next) {
 723                         if (m->m_len) {
 724                                 vm_offset_t paddr =
 725                                     vtophys(mtod(m, vm_offset_t));
 726                                 packet->page_buffers[i].pfn =
 727                                     paddr >> PAGE_SHIFT;
 728                                 packet->page_buffers[i].offset =
 729                                     paddr & (PAGE_SIZE - 1);
 730                                 packet->page_buffers[i].length = m->m_len;
 731                                 i++;
 732                         }
 733                 }
 734
 735                 packet->send_buf_section_idx =
 736                     NVSP_1_CHIMNEY_SEND_INVALID_SECTION_INDEX;
 737                 packet->send_buf_section_size = 0;
 738
 739 do_send:
 740
 741                 /*
 742                  * If bpf, copy the mbuf chain.  This is less expensive than
 743                  * it appears; the mbuf clusters are not copied, only their
 744                  * reference counts are incremented.
 745                  * Needed to avoid a race condition where the completion
 746                  * callback is invoked, freeing the mbuf chain, before the
 747                  * bpf_mtap code has a chance to run.
 748                  */
 749                 if (ifp->if_bpf) {
 750                         mc_head = m_copypacket(m_head, M_DONTWAIT);
 751                 }
 752 retry_send:
 753                 /* Set the completion routine */
 754                 packet->compl.send.on_send_completion = netvsc_xmit_completion;
 755                 packet->compl.send.send_completion_context = packet;
 756                 packet->compl.send.send_completion_tid = (uint64_t)(uintptr_t)m_head;
 757
 758                 /* Removed critical_enter(), does not appear necessary */
 759                 ret = hv_nv_on_send(device_ctx, packet);
 760                 if (ret == 0) {
 761                         ifp->if_opackets++;
 762                         /* if bpf && mc_head, call bpf_mtap code */
 763                         if (mc_head) {
 764                                 ETHER_BPF_MTAP(ifp, mc_head);
 765                         }
 766                 } else {
 767                         retries++;
 768                         if (retries < 4) {
 769                                 goto retry_send;
 770                         }
 771
 772                         IF_PREPEND(&ifp->if_snd, m_head);
 773                         ifp->if_drv_flags |= IFF_DRV_OACTIVE;
 774
 775                         /*
 776                          * Null the mbuf pointer so the completion function
 777                          * does not free the mbuf chain.  We just pushed the
 778                          * mbuf chain back on the if_snd queue.
 779                          */
 780                         packet->compl.send.send_completion_tid = 0;
 781
 782                         /*
 783                          * Release the resources since we will not get any
 784                          * send completion
 785                          */
 786                         netvsc_xmit_completion(packet);
 787                         if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
 788                 }
 789
 790                 /* if bpf && mc_head, free the mbuf chain copy */
 791                 if (mc_head) {
 792                         m_freem(mc_head);
 793                 }
 794         }
 795
 796         return (ret);
 797 }
 798
 799 /*
 800  * Link up/down notification
 801  */
 802 void
 803 netvsc_linkstatus_callback(struct hv_device *device_obj, uint32_t status)
 804 {
 805         hn_softc_t *sc = device_get_softc(device_obj->device);
 806
 807         if (sc == NULL) {
 808                 return;
 809         }
 810
 811         if (status == 1) {
 812                 sc->hn_carrier = 1;
 813         } else {
 814                 sc->hn_carrier = 0;
 815         }
 816 }
 817
 818 /*
 819  * Append the specified data to the indicated mbuf chain,
 820  * Extend the mbuf chain if the new data does not fit in
 821  * existing space.
 822  *
 823  * This is a minor rewrite of m_append() from sys/kern/uipc_mbuf.c.
 824  * There should be an equivalent in the kernel mbuf code,
 825  * but there does not appear to be one yet.
 826  *
 827  * Differs from m_append() in that additional mbufs are
 828  * allocated with cluster size MJUMPAGESIZE, and filled
 829  * accordingly.
 830  *
 831  * Return 1 if able to complete the job; otherwise 0.
 832  */
 833 static int
 834 hv_m_append(struct mbuf *m0, int len, c_caddr_t cp)
 835 {
 836         struct mbuf *m, *n;
 837         int remainder, space;
 838
 839         for (m = m0; m->m_next != NULL; m = m->m_next)
 840                 ;
 841         remainder = len;
 842         space = M_TRAILINGSPACE(m);
 843         if (space > 0) {
 844                 /*
 845                  * Copy into available space.
 846                  */
 847                 if (space > remainder)
 848                         space = remainder;
 849                 bcopy(cp, mtod(m, caddr_t) + m->m_len, space);
 850                 m->m_len += space;
 851                 cp += space;
 852                 remainder -= space;
 853         }
 854         while (remainder > 0) {
 855                 /*
 856                  * Allocate a new mbuf; could check space
 857                  * and allocate a cluster instead.
 858                  */
 859                 n = m_getjcl(M_DONTWAIT, m->m_type, 0, MJUMPAGESIZE);
 860                 if (n == NULL)
 861                         break;
 862                 n->m_len = min(MJUMPAGESIZE, remainder);
 863                 bcopy(cp, mtod(n, caddr_t), n->m_len);
 864                 cp += n->m_len;
 865                 remainder -= n->m_len;
 866                 m->m_next = n;
 867                 m = n;
 868         }
 869         if (m0->m_flags & M_PKTHDR)
 870                 m0->m_pkthdr.len += len - remainder;
 871
 872         return (remainder == 0);
 873 }
 874
 875
 876 /*
 877  * Called when we receive a data packet from the "wire" on the
 878  * specified device
 879  *
 880  * Note:  This is no longer used as a callback
 881  */
 882 int
 883 netvsc_recv(struct hv_device *device_ctx, netvsc_packet *packet,
 884     rndis_tcp_ip_csum_info *csum_info)
 885 {
 886         hn_softc_t *sc = (hn_softc_t *)device_get_softc(device_ctx->device);
 887         struct mbuf *m_new;
 888         struct ifnet *ifp;
 889         device_t dev = device_ctx->device;
 890         int size;
 891
 892         if (sc == NULL) {
 893                 return (0); /* TODO: KYS how can this be! */
 894         }
 895
 896         ifp = sc->hn_ifp;
 897
 898         ifp = sc->arpcom.ac_ifp;
 899
 900         if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 901                 return (0);
 902         }
 903
 904         /*
 905          * Bail out if packet contains more data than configured MTU.
 906          */
 907         if (packet->tot_data_buf_len > (ifp->if_mtu + ETHER_HDR_LEN)) {
 908                 return (0);
 909         }
 910
 911         /*
 912          * Get an mbuf with a cluster.  For packets 2K or less,
 913          * get a standard 2K cluster.  For anything larger, get a
 914          * 4K cluster.  Any buffers larger than 4K can cause problems
 915          * if looped around to the Hyper-V TX channel, so avoid them.
 916          */
 917         size = MCLBYTES;
 918
 919         if (packet->tot_data_buf_len > MCLBYTES) {
 920                 /* 4096 */
 921                 size = MJUMPAGESIZE;
 922         }
 923
 924         m_new = m_getjcl(M_DONTWAIT, MT_DATA, M_PKTHDR, size);
 925
 926         if (m_new == NULL) {
 927                 device_printf(dev, "alloc mbuf failed.\n");
 928                 return (0);
 929         }
 930
 931         hv_m_append(m_new, packet->tot_data_buf_len,
 932                         packet->data);
 933
 934         m_new->m_pkthdr.rcvif = ifp;
 935
 936         /* receive side checksum offload */
 937         m_new->m_pkthdr.csum_flags = 0;
 938         if (NULL != csum_info) {
 939                 /* IP csum offload */
 940                 if (csum_info->receive.ip_csum_succeeded) {
 941                         m_new->m_pkthdr.csum_flags |=
 942                             (CSUM_IP_CHECKED | CSUM_IP_VALID);
 943                 }
 944
 945                 /* TCP csum offload */
 946                 if (csum_info->receive.tcp_csum_succeeded) {
 947                         m_new->m_pkthdr.csum_flags |=
 948                             (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
 949                         m_new->m_pkthdr.csum_data = 0xffff;
 950                 }
 951         }
 952
 953         if ((packet->vlan_tci != 0) &&
 954             (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) != 0) {
 955                 m_new->m_pkthdr.ether_vtag = packet->vlan_tci;
 956                 m_new->m_flags |= M_VLANTAG;
 957         }
 958
 959         /*
 960          * Note:  Moved RX completion back to hv_nv_on_receive() so all
 961          * messages (not just data messages) will trigger a response.
 962          */
 963
 964         ifp->if_ipackets++;
 965
 966         /* We're not holding the lock here, so don't release it */
 967         (*ifp->if_input)(ifp, m_new);
 968
 969         return (0);
 970 }
 971
 972 /*
 973  * Rules for using sc->temp_unusable:
 974  * 1.  sc->temp_unusable can only be read or written while holding NV_LOCK()
 975  * 2.  code reading sc->temp_unusable under NV_LOCK(), and finding
 976  *     sc->temp_unusable set, must release NV_LOCK() and exit
 977  * 3.  to retain exclusive control of the interface,
 978  *     sc->temp_unusable must be set by code before releasing NV_LOCK()
 979  * 4.  only code setting sc->temp_unusable can clear sc->temp_unusable
 980  * 5.  code setting sc->temp_unusable must eventually clear sc->temp_unusable
 981  */
 982
 983 /*
 984  * Standard ioctl entry point.  Called when the user wants to configure
 985  * the interface.
 986  */
 987 static int
 988 hn_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 989 {
 990         hn_softc_t *sc = ifp->if_softc;
 991         struct ifreq *ifr = (struct ifreq *)data;
 992 #ifdef INET
 993         struct ifaddr *ifa = (struct ifaddr *)data;
 994 #endif
 995         netvsc_device_info device_info;
 996         struct hv_device *hn_dev;
 997         int mask, error = 0;
 998         int retry_cnt = 500;
 999
1000         switch(cmd) {
1001
1002         case SIOCSIFADDR:
1003 #ifdef INET
1004                 if (ifa->ifa_addr->sa_family == AF_INET) {
1005                         ifp->if_flags |= IFF_UP;
1006                         if (!(ifp->if_drv_flags & IFF_DRV_RUNNING))
1007                                 hn_ifinit(sc);
1008                         arp_ifinit(ifp, ifa);
1009                 } else
1010 #endif
1011                 error = ether_ioctl(ifp, cmd, data);
1012                 break;
1013         case SIOCSIFMTU:
1014                 hn_dev = vmbus_get_devctx(sc->hn_dev);
1015
1016                 /* Check MTU value change */
1017                 if (ifp->if_mtu == ifr->ifr_mtu)
1018                         break;
1019
1020                 if (ifr->ifr_mtu > NETVSC_MAX_CONFIGURABLE_MTU) {
1021                         error = EINVAL;
1022                         break;
1023                 }
1024
1025                 /* Obtain and record requested MTU */
1026                 ifp->if_mtu = ifr->ifr_mtu;
1027
1028                 do {
1029                         NV_LOCK(sc);
1030                         if (!sc->temp_unusable) {
1031                                 sc->temp_unusable = TRUE;
1032                                 retry_cnt = -1;
1033                         }
1034                         NV_UNLOCK(sc);
1035                         if (retry_cnt > 0) {
1036                                 retry_cnt--;
1037                                 DELAY(5 * 1000);
1038                         }
1039                 } while (retry_cnt > 0);
1040
1041                 if (retry_cnt == 0) {
1042                         error = EINVAL;
1043                         break;
1044                 }
1045
1046                 /* We must remove and add back the device to cause the new
1047                  * MTU to take effect.  This includes tearing down, but not
1048                  * deleting the channel, then bringing it back up.
1049                  */
1050                 error = hv_rf_on_device_remove(hn_dev, HV_RF_NV_RETAIN_CHANNEL);
1051                 if (error) {
1052                         NV_LOCK(sc);
1053                         sc->temp_unusable = FALSE;
1054                         NV_UNLOCK(sc);
1055                         break;
1056                 }
1057                 error = hv_rf_on_device_add(hn_dev, &device_info);
1058                 if (error) {
1059                         NV_LOCK(sc);
1060                         sc->temp_unusable = FALSE;
1061                         NV_UNLOCK(sc);
1062                         break;
1063                 }
1064
1065                 hn_ifinit_locked(sc);
1066
1067                 NV_LOCK(sc);
1068                 sc->temp_unusable = FALSE;
1069                 NV_UNLOCK(sc);
1070                 break;
1071         case SIOCSIFFLAGS:
1072                 do {
1073                        NV_LOCK(sc);
1074                        if (!sc->temp_unusable) {
1075                                sc->temp_unusable = TRUE;
1076                                retry_cnt = -1;
1077                        }
1078                        NV_UNLOCK(sc);
1079                        if (retry_cnt > 0) {
1080                                 retry_cnt--;
1081                                 DELAY(5 * 1000);
1082                        }
1083                 } while (retry_cnt > 0);
1084
1085                 if (retry_cnt == 0) {
1086                        error = EINVAL;
1087                        break;
1088                 }
1089
1090                 if (ifp->if_flags & IFF_UP) {
1091                         /*
1092                          * If only the state of the PROMISC flag changed,
1093                          * then just use the 'set promisc mode' command
1094                          * instead of reinitializing the entire NIC. Doing
1095                          * a full re-init means reloading the firmware and
1096                          * waiting for it to start up, which may take a
1097                          * second or two.
1098                          */
1099 #ifdef notyet
1100                         /* Fixme:  Promiscuous mode? */
1101                         if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
1102                             ifp->if_flags & IFF_PROMISC &&
1103                             !(sc->hn_if_flags & IFF_PROMISC)) {
1104                                 /* do something here for Hyper-V */
1105                         } else if (ifp->if_drv_flags & IFF_DRV_RUNNING &&
1106                             !(ifp->if_flags & IFF_PROMISC) &&
1107                             sc->hn_if_flags & IFF_PROMISC) {
1108                                 /* do something here for Hyper-V */
1109                         } else
1110 #endif
1111                                 hn_ifinit_locked(sc);
1112                 } else {
1113                         if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
1114                                 hn_stop(sc);
1115                         }
1116                 }
1117                 NV_LOCK(sc);
1118                 sc->temp_unusable = FALSE;
1119                 NV_UNLOCK(sc);
1120                 sc->hn_if_flags = ifp->if_flags;
1121                 error = 0;
1122                 break;
1123         case SIOCSIFCAP:
1124                 mask = ifr->ifr_reqcap ^ ifp->if_capenable;
1125                 if (mask & IFCAP_TXCSUM) {
1126                         if (IFCAP_TXCSUM & ifp->if_capenable) {
1127                                 ifp->if_capenable &= ~IFCAP_TXCSUM;
1128                                 ifp->if_hwassist &= ~(CSUM_TCP | CSUM_UDP);
1129                         } else {
1130                                 ifp->if_capenable |= IFCAP_TXCSUM;
1131                                 /*
1132                                  * Only enable UDP checksum offloading on
1133                                  * Windows Server 2012R2 or later releases.
1134                                  */
1135                                 if (hv_vmbus_protocal_version >=
1136                                     HV_VMBUS_VERSION_WIN8_1) {
1137                                         ifp->if_hwassist |=
1138                                             (CSUM_TCP | CSUM_UDP);
1139                                 } else {
1140                                         ifp->if_hwassist |= CSUM_TCP;
1141                                 }
1142                         }
1143                 }
1144
1145                 if (mask & IFCAP_RXCSUM) {
1146                         if (IFCAP_RXCSUM & ifp->if_capenable) {
1147                                 ifp->if_capenable &= ~IFCAP_RXCSUM;
1148                         } else {
1149                                 ifp->if_capenable |= IFCAP_RXCSUM;
1150                         }
1151                 }
1152
1153                 if (mask & IFCAP_TSO4) {
1154                         ifp->if_capenable ^= IFCAP_TSO4;
1155                         ifp->if_hwassist ^= CSUM_IP_TSO;
1156                 }
1157
1158                 if (mask & IFCAP_TSO6) {
1159                         ifp->if_capenable ^= IFCAP_TSO6;
1160                         ifp->if_hwassist ^= CSUM_IP6_TSO;
1161                 }
1162
1163                 error = 0;
1164                 break;
1165         case SIOCADDMULTI:
1166         case SIOCDELMULTI:
1167 #ifdef notyet
1168                 /* Fixme:  Multicast mode? */
1169                 if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
1170                         NV_LOCK(sc);
1171                         netvsc_setmulti(sc);
1172                         NV_UNLOCK(sc);
1173                         error = 0;
1174                 }
1175 #endif
1176                 /* FALLTHROUGH */
1177         case SIOCSIFMEDIA:
1178         case SIOCGIFMEDIA:
1179                 error = EINVAL;
1180                 break;
1181         default:
1182                 error = ether_ioctl(ifp, cmd, data);
1183                 break;
1184         }
1185
1186         return (error);
1187 }
1188
1189 /*
1190  *
1191  */
1192 static void
1193 hn_stop(hn_softc_t *sc)
1194 {
1195         struct ifnet *ifp;
1196         int ret;
1197         struct hv_device *device_ctx = vmbus_get_devctx(sc->hn_dev);
1198
1199         ifp = sc->hn_ifp;
1200
1201         if (bootverbose)
1202                 printf(" Closing Device ...\n");
1203
1204         ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE);
1205         if_link_state_change(ifp, LINK_STATE_DOWN);
1206         sc->hn_initdone = 0;
1207
1208         ret = hv_rf_on_close(device_ctx);
1209 }
1210
1211 /*
1212  * FreeBSD transmit entry point
1213  */
1214 static void
1215 hn_start(struct ifnet *ifp)
1216 {
1217         hn_softc_t *sc;
1218
1219         sc = ifp->if_softc;
1220         NV_LOCK(sc);
1221         if (sc->temp_unusable) {
1222                 NV_UNLOCK(sc);
1223                 return;
1224         }
1225         hn_start_locked(ifp);
1226         NV_UNLOCK(sc);
1227 }
1228
1229 /*
1230  *
1231  */
1232 static void
1233 hn_ifinit_locked(hn_softc_t *sc)
1234 {
1235         struct ifnet *ifp;
1236         struct hv_device *device_ctx = vmbus_get_devctx(sc->hn_dev);
1237         int ret;
1238
1239         ifp = sc->hn_ifp;
1240
1241         if (ifp->if_drv_flags & IFF_DRV_RUNNING) {
1242                 return;
1243         }
1244
1245         hv_promisc_mode = 1;
1246
1247         ret = hv_rf_on_open(device_ctx);
1248         if (ret != 0) {
1249                 return;
1250         } else {
1251                 sc->hn_initdone = 1;
1252         }
1253         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1254         ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
1255         if_link_state_change(ifp, LINK_STATE_UP);
1256 }
1257
1258 /*
1259  *
1260  */
1261 static void
1262 hn_ifinit(void *xsc)
1263 {
1264         hn_softc_t *sc = xsc;
1265
1266         NV_LOCK(sc);
1267         if (sc->temp_unusable) {
1268                 NV_UNLOCK(sc);
1269                 return;
1270         }
1271         sc->temp_unusable = TRUE;
1272         NV_UNLOCK(sc);
1273
1274         hn_ifinit_locked(sc);
1275
1276         NV_LOCK(sc);
1277         sc->temp_unusable = FALSE;
1278         NV_UNLOCK(sc);
1279 }
1280
1281 #ifdef LATER
1282 /*
1283  *
1284  */
1285 static void
1286 hn_watchdog(struct ifnet *ifp)
1287 {
1288         hn_softc_t *sc;
1289         sc = ifp->if_softc;
1290
1291         printf("hn%d: watchdog timeout -- resetting\n", sc->hn_unit);
1292         hn_ifinit(sc);    /*???*/
1293         ifp->if_oerrors++;
1294 }
1295 #endif
1296
1297 static device_method_t netvsc_methods[] = {
1298         /* Device interface */
1299         DEVMETHOD(device_probe,         netvsc_probe),
1300         DEVMETHOD(device_attach,        netvsc_attach),
1301         DEVMETHOD(device_detach,        netvsc_detach),
1302         DEVMETHOD(device_shutdown,      netvsc_shutdown),
1303
1304         { 0, 0 }
1305 };
1306
1307 static driver_t netvsc_driver = {
1308         NETVSC_DEVNAME,
1309         netvsc_methods,
1310         sizeof(hn_softc_t)
1311 };
1312
1313 static devclass_t netvsc_devclass;
1314
1315 DRIVER_MODULE(hn, vmbus, netvsc_driver, netvsc_devclass, 0, 0);
1316 MODULE_VERSION(hn, 1);
1317 MODULE_DEPEND(hn, vmbus, 1, 1, 1);
1318 SYSINIT(netvsc_initx, SI_SUB_KTHREAD_IDLE, SI_ORDER_MIDDLE + 1, netvsc_init,
1319      NULL);
1320