drivers/hv/channel_mgmt.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/kernel.h>
  24 #include <linux/sched.h>
  25 #include <linux/wait.h>
  26 #include <linux/mm.h>
  27 #include <linux/slab.h>
  28 #include <linux/list.h>
  29 #include <linux/module.h>
  30 #include <linux/completion.h>
  31 #include <linux/hyperv.h>
  32
  33 #include "hyperv_vmbus.h"
  34
  35 static void init_vp_index(struct vmbus_channel *channel,
  36                           const uuid_le *type_guid);
  37
  38 /**
  39  * vmbus_prep_negotiate_resp() - Create default response for Hyper-V Negotiate message
  40  * @icmsghdrp: Pointer to msg header structure
  41  * @icmsg_negotiate: Pointer to negotiate message structure
  42  * @buf: Raw buffer channel data
  43  *
  44  * @icmsghdrp is of type &struct icmsg_hdr.
  45  * @negop is of type &struct icmsg_negotiate.
  46  * Set up and fill in default negotiate response message.
  47  *
  48  * The fw_version specifies the  framework version that
  49  * we can support and srv_version specifies the service
  50  * version we can support.
  51  *
  52  * Mainly used by Hyper-V drivers.
  53  */
  54 bool vmbus_prep_negotiate_resp(struct icmsg_hdr *icmsghdrp,
  55                                 struct icmsg_negotiate *negop, u8 *buf,
  56                                 int fw_version, int srv_version)
  57 {
  58         int icframe_major, icframe_minor;
  59         int icmsg_major, icmsg_minor;
  60         int fw_major, fw_minor;
  61         int srv_major, srv_minor;
  62         int i;
  63         bool found_match = false;
  64
  65         icmsghdrp->icmsgsize = 0x10;
  66         fw_major = (fw_version >> 16);
  67         fw_minor = (fw_version & 0xFFFF);
  68
  69         srv_major = (srv_version >> 16);
  70         srv_minor = (srv_version & 0xFFFF);
  71
  72         negop = (struct icmsg_negotiate *)&buf[
  73                 sizeof(struct vmbuspipe_hdr) +
  74                 sizeof(struct icmsg_hdr)];
  75
  76         icframe_major = negop->icframe_vercnt;
  77         icframe_minor = 0;
  78
  79         icmsg_major = negop->icmsg_vercnt;
  80         icmsg_minor = 0;
  81
  82         /*
  83          * Select the framework version number we will
  84          * support.
  85          */
  86
  87         for (i = 0; i < negop->icframe_vercnt; i++) {
  88                 if ((negop->icversion_data[i].major == fw_major) &&
  89                    (negop->icversion_data[i].minor == fw_minor)) {
  90                         icframe_major = negop->icversion_data[i].major;
  91                         icframe_minor = negop->icversion_data[i].minor;
  92                         found_match = true;
  93                 }
  94         }
  95
  96         if (!found_match)
  97                 goto fw_error;
  98
  99         found_match = false;
 100
 101         for (i = negop->icframe_vercnt;
 102                  (i < negop->icframe_vercnt + negop->icmsg_vercnt); i++) {
 103                 if ((negop->icversion_data[i].major == srv_major) &&
 104                    (negop->icversion_data[i].minor == srv_minor)) {
 105                         icmsg_major = negop->icversion_data[i].major;
 106                         icmsg_minor = negop->icversion_data[i].minor;
 107                         found_match = true;
 108                 }
 109         }
 110
 111         /*
 112          * Respond with the framework and service
 113          * version numbers we can support.
 114          */
 115
 116 fw_error:
 117         if (!found_match) {
 118                 negop->icframe_vercnt = 0;
 119                 negop->icmsg_vercnt = 0;
 120         } else {
 121                 negop->icframe_vercnt = 1;
 122                 negop->icmsg_vercnt = 1;
 123         }
 124
 125         negop->icversion_data[0].major = icframe_major;
 126         negop->icversion_data[0].minor = icframe_minor;
 127         negop->icversion_data[1].major = icmsg_major;
 128         negop->icversion_data[1].minor = icmsg_minor;
 129         return found_match;
 130 }
 131
 132 EXPORT_SYMBOL_GPL(vmbus_prep_negotiate_resp);
 133
 134 /*
 135  * alloc_channel - Allocate and initialize a vmbus channel object
 136  */
 137 static struct vmbus_channel *alloc_channel(void)
 138 {
 139         static atomic_t chan_num = ATOMIC_INIT(0);
 140         struct vmbus_channel *channel;
 141
 142         channel = kzalloc(sizeof(*channel), GFP_ATOMIC);
 143         if (!channel)
 144                 return NULL;
 145
 146         channel->id = atomic_inc_return(&chan_num);
 147         spin_lock_init(&channel->inbound_lock);
 148         spin_lock_init(&channel->lock);
 149
 150         INIT_LIST_HEAD(&channel->sc_list);
 151         INIT_LIST_HEAD(&channel->percpu_list);
 152
 153         return channel;
 154 }
 155
 156 /*
 157  * free_channel - Release the resources used by the vmbus channel object
 158  */
 159 static void free_channel(struct vmbus_channel *channel)
 160 {
 161         kfree(channel);
 162 }
 163
 164 static void percpu_channel_enq(void *arg)
 165 {
 166         struct vmbus_channel *channel = arg;
 167         int cpu = smp_processor_id();
 168
 169         list_add_tail(&channel->percpu_list, &hv_context.percpu_list[cpu]);
 170 }
 171
 172 static void percpu_channel_deq(void *arg)
 173 {
 174         struct vmbus_channel *channel = arg;
 175
 176         list_del(&channel->percpu_list);
 177 }
 178
 179
 180 void hv_process_channel_removal(struct vmbus_channel *channel, u32 relid)
 181 {
 182         struct vmbus_channel_relid_released msg;
 183         unsigned long flags;
 184         struct vmbus_channel *primary_channel;
 185
 186         memset(&msg, 0, sizeof(struct vmbus_channel_relid_released));
 187         msg.child_relid = relid;
 188         msg.header.msgtype = CHANNELMSG_RELID_RELEASED;
 189         vmbus_post_msg(&msg, sizeof(struct vmbus_channel_relid_released));
 190
 191         if (channel == NULL)
 192                 return;
 193
 194         if (channel->target_cpu != get_cpu()) {
 195                 put_cpu();
 196                 smp_call_function_single(channel->target_cpu,
 197                                          percpu_channel_deq, channel, true);
 198         } else {
 199                 percpu_channel_deq(channel);
 200                 put_cpu();
 201         }
 202
 203         if (channel->primary_channel == NULL) {
 204                 spin_lock_irqsave(&vmbus_connection.channel_lock, flags);
 205                 list_del(&channel->listentry);
 206                 spin_unlock_irqrestore(&vmbus_connection.channel_lock, flags);
 207         } else {
 208                 primary_channel = channel->primary_channel;
 209                 spin_lock_irqsave(&primary_channel->lock, flags);
 210                 list_del(&channel->sc_list);
 211                 primary_channel->num_sc--;
 212                 spin_unlock_irqrestore(&primary_channel->lock, flags);
 213         }
 214         free_channel(channel);
 215 }
 216
 217 void vmbus_free_channels(void)
 218 {
 219         struct vmbus_channel *channel, *tmp;
 220
 221         list_for_each_entry_safe(channel, tmp, &vmbus_connection.chn_list,
 222                 listentry) {
 223                 /* if we don't set rescind to true, vmbus_close_internal()
 224                  * won't invoke hv_process_channel_removal().
 225                  */
 226                 channel->rescind = true;
 227
 228                 vmbus_device_unregister(channel->device_obj);
 229         }
 230 }
 231
 232 /*
 233  * vmbus_process_offer - Process the offer by creating a channel/device
 234  * associated with this offer
 235  */
 236 static void vmbus_process_offer(struct vmbus_channel *newchannel)
 237 {
 238         struct vmbus_channel *channel;
 239         bool fnew = true;
 240         unsigned long flags;
 241
 242         /* Make sure this is a new offer */
 243         spin_lock_irqsave(&vmbus_connection.channel_lock, flags);
 244
 245         list_for_each_entry(channel, &vmbus_connection.chn_list, listentry) {
 246                 if (!uuid_le_cmp(channel->offermsg.offer.if_type,
 247                         newchannel->offermsg.offer.if_type) &&
 248                         !uuid_le_cmp(channel->offermsg.offer.if_instance,
 249                                 newchannel->offermsg.offer.if_instance)) {
 250                         fnew = false;
 251                         break;
 252                 }
 253         }
 254
 255         if (fnew)
 256                 list_add_tail(&newchannel->listentry,
 257                               &vmbus_connection.chn_list);
 258
 259         spin_unlock_irqrestore(&vmbus_connection.channel_lock, flags);
 260
 261         if (!fnew) {
 262                 /*
 263                  * Check to see if this is a sub-channel.
 264                  */
 265                 if (newchannel->offermsg.offer.sub_channel_index != 0) {
 266                         /*
 267                          * Process the sub-channel.
 268                          */
 269                         newchannel->primary_channel = channel;
 270                         spin_lock_irqsave(&channel->lock, flags);
 271                         list_add_tail(&newchannel->sc_list, &channel->sc_list);
 272                         channel->num_sc++;
 273                         spin_unlock_irqrestore(&channel->lock, flags);
 274                 } else
 275                         goto err_free_chan;
 276         }
 277
 278         init_vp_index(newchannel, &newchannel->offermsg.offer.if_type);
 279
 280         if (newchannel->target_cpu != get_cpu()) {
 281                 put_cpu();
 282                 smp_call_function_single(newchannel->target_cpu,
 283                                          percpu_channel_enq,
 284                                          newchannel, true);
 285         } else {
 286                 percpu_channel_enq(newchannel);
 287                 put_cpu();
 288         }
 289
 290         /*
 291          * This state is used to indicate a successful open
 292          * so that when we do close the channel normally, we
 293          * can cleanup properly
 294          */
 295         newchannel->state = CHANNEL_OPEN_STATE;
 296
 297         if (!fnew) {
 298                 if (channel->sc_creation_callback != NULL)
 299                         channel->sc_creation_callback(newchannel);
 300                 return;
 301         }
 302
 303         /*
 304          * Start the process of binding this offer to the driver
 305          * We need to set the DeviceObject field before calling
 306          * vmbus_child_dev_add()
 307          */
 308         newchannel->device_obj = vmbus_device_create(
 309                 &newchannel->offermsg.offer.if_type,
 310                 &newchannel->offermsg.offer.if_instance,
 311                 newchannel);
 312         if (!newchannel->device_obj)
 313                 goto err_deq_chan;
 314
 315         /*
 316          * Add the new device to the bus. This will kick off device-driver
 317          * binding which eventually invokes the device driver's AddDevice()
 318          * method.
 319          */
 320         if (vmbus_device_register(newchannel->device_obj) != 0) {
 321                 pr_err("unable to add child device object (relid %d)\n",
 322                         newchannel->offermsg.child_relid);
 323                 kfree(newchannel->device_obj);
 324                 goto err_deq_chan;
 325         }
 326         return;
 327
 328 err_deq_chan:
 329         spin_lock_irqsave(&vmbus_connection.channel_lock, flags);
 330         list_del(&newchannel->listentry);
 331         spin_unlock_irqrestore(&vmbus_connection.channel_lock, flags);
 332
 333         if (newchannel->target_cpu != get_cpu()) {
 334                 put_cpu();
 335                 smp_call_function_single(newchannel->target_cpu,
 336                                          percpu_channel_deq, newchannel, true);
 337         } else {
 338                 percpu_channel_deq(newchannel);
 339                 put_cpu();
 340         }
 341
 342 err_free_chan:
 343         free_channel(newchannel);
 344 }
 345
 346 enum {
 347         IDE = 0,
 348         SCSI,
 349         NIC,
 350         MAX_PERF_CHN,
 351 };
 352
 353 /*
 354  * This is an array of device_ids (device types) that are performance critical.
 355  * We attempt to distribute the interrupt load for these devices across
 356  * all available CPUs.
 357  */
 358 static const struct hv_vmbus_device_id hp_devs[] = {
 359         /* IDE */
 360         { HV_IDE_GUID, },
 361         /* Storage - SCSI */
 362         { HV_SCSI_GUID, },
 363         /* Network */
 364         { HV_NIC_GUID, },
 365         /* NetworkDirect Guest RDMA */
 366         { HV_ND_GUID, },
 367 };
 368
 369
 370 /*
 371  * We use this state to statically distribute the channel interrupt load.
 372  */
 373 static int next_numa_node_id;
 374
 375 /*
 376  * Starting with Win8, we can statically distribute the incoming
 377  * channel interrupt load by binding a channel to VCPU.
 378  * We do this in a hierarchical fashion:
 379  * First distribute the primary channels across available NUMA nodes
 380  * and then distribute the subchannels amongst the CPUs in the NUMA
 381  * node assigned to the primary channel.
 382  *
 383  * For pre-win8 hosts or non-performance critical channels we assign the
 384  * first CPU in the first NUMA node.
 385  */
 386 static void init_vp_index(struct vmbus_channel *channel, const uuid_le *type_guid)
 387 {
 388         u32 cur_cpu;
 389         int i;
 390         bool perf_chn = false;
 391         struct vmbus_channel *primary = channel->primary_channel;
 392         int next_node;
 393         struct cpumask available_mask;
 394
 395         for (i = IDE; i < MAX_PERF_CHN; i++) {
 396                 if (!memcmp(type_guid->b, hp_devs[i].guid,
 397                                  sizeof(uuid_le))) {
 398                         perf_chn = true;
 399                         break;
 400                 }
 401         }
 402         if ((vmbus_proto_version == VERSION_WS2008) ||
 403             (vmbus_proto_version == VERSION_WIN7) || (!perf_chn)) {
 404                 /*
 405                  * Prior to win8, all channel interrupts are
 406                  * delivered on cpu 0.
 407                  * Also if the channel is not a performance critical
 408                  * channel, bind it to cpu 0.
 409                  */
 410                 channel->numa_node = 0;
 411                 cpumask_set_cpu(0, &channel->alloced_cpus_in_node);
 412                 channel->target_cpu = 0;
 413                 channel->target_vp = hv_context.vp_index[0];
 414                 return;
 415         }
 416
 417         /*
 418          * We distribute primary channels evenly across all the available
 419          * NUMA nodes and within the assigned NUMA node we will assign the
 420          * first available CPU to the primary channel.
 421          * The sub-channels will be assigned to the CPUs available in the
 422          * NUMA node evenly.
 423          */
 424         if (!primary) {
 425                 while (true) {
 426                         next_node = next_numa_node_id++;
 427                         if (next_node == nr_node_ids)
 428                                 next_node = next_numa_node_id = 0;
 429                         if (cpumask_empty(cpumask_of_node(next_node)))
 430                                 continue;
 431                         break;
 432                 }
 433                 channel->numa_node = next_node;
 434                 primary = channel;
 435         }
 436
 437         if (cpumask_weight(&primary->alloced_cpus_in_node) ==
 438             cpumask_weight(cpumask_of_node(primary->numa_node))) {
 439                 /*
 440                  * We have cycled through all the CPUs in the node;
 441                  * reset the alloced map.
 442                  */
 443                 cpumask_clear(&primary->alloced_cpus_in_node);
 444         }
 445
 446         cpumask_xor(&available_mask, &primary->alloced_cpus_in_node,
 447                     cpumask_of_node(primary->numa_node));
 448
 449         cur_cpu = cpumask_next(-1, &available_mask);
 450         cpumask_set_cpu(cur_cpu, &primary->alloced_cpus_in_node);
 451
 452         channel->target_cpu = cur_cpu;
 453         channel->target_vp = hv_context.vp_index[cur_cpu];
 454 }
 455
 456 /*
 457  * vmbus_unload_response - Handler for the unload response.
 458  */
 459 static void vmbus_unload_response(struct vmbus_channel_message_header *hdr)
 460 {
 461         /*
 462          * This is a global event; just wakeup the waiting thread.
 463          * Once we successfully unload, we can cleanup the monitor state.
 464          */
 465         complete(&vmbus_connection.unload_event);
 466 }
 467
 468 void vmbus_initiate_unload(void)
 469 {
 470         struct vmbus_channel_message_header hdr;
 471
 472         init_completion(&vmbus_connection.unload_event);
 473         memset(&hdr, 0, sizeof(struct vmbus_channel_message_header));
 474         hdr.msgtype = CHANNELMSG_UNLOAD;
 475         vmbus_post_msg(&hdr, sizeof(struct vmbus_channel_message_header));
 476
 477         wait_for_completion(&vmbus_connection.unload_event);
 478 }
 479
 480 /*
 481  * vmbus_onoffer - Handler for channel offers from vmbus in parent partition.
 482  *
 483  */
 484 static void vmbus_onoffer(struct vmbus_channel_message_header *hdr)
 485 {
 486         struct vmbus_channel_offer_channel *offer;
 487         struct vmbus_channel *newchannel;
 488
 489         offer = (struct vmbus_channel_offer_channel *)hdr;
 490
 491         /* Allocate the channel object and save this offer. */
 492         newchannel = alloc_channel();
 493         if (!newchannel) {
 494                 pr_err("Unable to allocate channel object\n");
 495                 return;
 496         }
 497
 498         /*
 499          * By default we setup state to enable batched
 500          * reading. A specific service can choose to
 501          * disable this prior to opening the channel.
 502          */
 503         newchannel->batched_reading = true;
 504
 505         /*
 506          * Setup state for signalling the host.
 507          */
 508         newchannel->sig_event = (struct hv_input_signal_event *)
 509                                 (ALIGN((unsigned long)
 510                                 &newchannel->sig_buf,
 511                                 HV_HYPERCALL_PARAM_ALIGN));
 512
 513         newchannel->sig_event->connectionid.asu32 = 0;
 514         newchannel->sig_event->connectionid.u.id = VMBUS_EVENT_CONNECTION_ID;
 515         newchannel->sig_event->flag_number = 0;
 516         newchannel->sig_event->rsvdz = 0;
 517
 518         if (vmbus_proto_version != VERSION_WS2008) {
 519                 newchannel->is_dedicated_interrupt =
 520                                 (offer->is_dedicated_interrupt != 0);
 521                 newchannel->sig_event->connectionid.u.id =
 522                                 offer->connection_id;
 523         }
 524
 525         memcpy(&newchannel->offermsg, offer,
 526                sizeof(struct vmbus_channel_offer_channel));
 527         newchannel->monitor_grp = (u8)offer->monitorid / 32;
 528         newchannel->monitor_bit = (u8)offer->monitorid % 32;
 529
 530         vmbus_process_offer(newchannel);
 531 }
 532
 533 /*
 534  * vmbus_onoffer_rescind - Rescind offer handler.
 535  *
 536  * We queue a work item to process this offer synchronously
 537  */
 538 static void vmbus_onoffer_rescind(struct vmbus_channel_message_header *hdr)
 539 {
 540         struct vmbus_channel_rescind_offer *rescind;
 541         struct vmbus_channel *channel;
 542         unsigned long flags;
 543         struct device *dev;
 544
 545         rescind = (struct vmbus_channel_rescind_offer *)hdr;
 546         channel = relid2channel(rescind->child_relid);
 547
 548         if (channel == NULL) {
 549                 hv_process_channel_removal(NULL, rescind->child_relid);
 550                 return;
 551         }
 552
 553         spin_lock_irqsave(&channel->lock, flags);
 554         channel->rescind = true;
 555         spin_unlock_irqrestore(&channel->lock, flags);
 556
 557         if (channel->device_obj) {
 558                 /*
 559                  * We will have to unregister this device from the
 560                  * driver core.
 561                  */
 562                 dev = get_device(&channel->device_obj->device);
 563                 if (dev) {
 564                         vmbus_device_unregister(channel->device_obj);
 565                         put_device(dev);
 566                 }
 567         } else {
 568                 hv_process_channel_removal(channel,
 569                         channel->offermsg.child_relid);
 570         }
 571 }
 572
 573 /*
 574  * vmbus_onoffers_delivered -
 575  * This is invoked when all offers have been delivered.
 576  *
 577  * Nothing to do here.
 578  */
 579 static void vmbus_onoffers_delivered(
 580                         struct vmbus_channel_message_header *hdr)
 581 {
 582 }
 583
 584 /*
 585  * vmbus_onopen_result - Open result handler.
 586  *
 587  * This is invoked when we received a response to our channel open request.
 588  * Find the matching request, copy the response and signal the requesting
 589  * thread.
 590  */
 591 static void vmbus_onopen_result(struct vmbus_channel_message_header *hdr)
 592 {
 593         struct vmbus_channel_open_result *result;
 594         struct vmbus_channel_msginfo *msginfo;
 595         struct vmbus_channel_message_header *requestheader;
 596         struct vmbus_channel_open_channel *openmsg;
 597         unsigned long flags;
 598
 599         result = (struct vmbus_channel_open_result *)hdr;
 600
 601         /*
 602          * Find the open msg, copy the result and signal/unblock the wait event
 603          */
 604         spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
 605
 606         list_for_each_entry(msginfo, &vmbus_connection.chn_msg_list,
 607                                 msglistentry) {
 608                 requestheader =
 609                         (struct vmbus_channel_message_header *)msginfo->msg;
 610
 611                 if (requestheader->msgtype == CHANNELMSG_OPENCHANNEL) {
 612                         openmsg =
 613                         (struct vmbus_channel_open_channel *)msginfo->msg;
 614                         if (openmsg->child_relid == result->child_relid &&
 615                             openmsg->openid == result->openid) {
 616                                 memcpy(&msginfo->response.open_result,
 617                                        result,
 618                                        sizeof(
 619                                         struct vmbus_channel_open_result));
 620                                 complete(&msginfo->waitevent);
 621                                 break;
 622                         }
 623                 }
 624         }
 625         spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 626 }
 627
 628 /*
 629  * vmbus_ongpadl_created - GPADL created handler.
 630  *
 631  * This is invoked when we received a response to our gpadl create request.
 632  * Find the matching request, copy the response and signal the requesting
 633  * thread.
 634  */
 635 static void vmbus_ongpadl_created(struct vmbus_channel_message_header *hdr)
 636 {
 637         struct vmbus_channel_gpadl_created *gpadlcreated;
 638         struct vmbus_channel_msginfo *msginfo;
 639         struct vmbus_channel_message_header *requestheader;
 640         struct vmbus_channel_gpadl_header *gpadlheader;
 641         unsigned long flags;
 642
 643         gpadlcreated = (struct vmbus_channel_gpadl_created *)hdr;
 644
 645         /*
 646          * Find the establish msg, copy the result and signal/unblock the wait
 647          * event
 648          */
 649         spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
 650
 651         list_for_each_entry(msginfo, &vmbus_connection.chn_msg_list,
 652                                 msglistentry) {
 653                 requestheader =
 654                         (struct vmbus_channel_message_header *)msginfo->msg;
 655
 656                 if (requestheader->msgtype == CHANNELMSG_GPADL_HEADER) {
 657                         gpadlheader =
 658                         (struct vmbus_channel_gpadl_header *)requestheader;
 659
 660                         if ((gpadlcreated->child_relid ==
 661                              gpadlheader->child_relid) &&
 662                             (gpadlcreated->gpadl == gpadlheader->gpadl)) {
 663                                 memcpy(&msginfo->response.gpadl_created,
 664                                        gpadlcreated,
 665                                        sizeof(
 666                                         struct vmbus_channel_gpadl_created));
 667                                 complete(&msginfo->waitevent);
 668                                 break;
 669                         }
 670                 }
 671         }
 672         spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 673 }
 674
 675 /*
 676  * vmbus_ongpadl_torndown - GPADL torndown handler.
 677  *
 678  * This is invoked when we received a response to our gpadl teardown request.
 679  * Find the matching request, copy the response and signal the requesting
 680  * thread.
 681  */
 682 static void vmbus_ongpadl_torndown(
 683                         struct vmbus_channel_message_header *hdr)
 684 {
 685         struct vmbus_channel_gpadl_torndown *gpadl_torndown;
 686         struct vmbus_channel_msginfo *msginfo;
 687         struct vmbus_channel_message_header *requestheader;
 688         struct vmbus_channel_gpadl_teardown *gpadl_teardown;
 689         unsigned long flags;
 690
 691         gpadl_torndown = (struct vmbus_channel_gpadl_torndown *)hdr;
 692
 693         /*
 694          * Find the open msg, copy the result and signal/unblock the wait event
 695          */
 696         spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
 697
 698         list_for_each_entry(msginfo, &vmbus_connection.chn_msg_list,
 699                                 msglistentry) {
 700                 requestheader =
 701                         (struct vmbus_channel_message_header *)msginfo->msg;
 702
 703                 if (requestheader->msgtype == CHANNELMSG_GPADL_TEARDOWN) {
 704                         gpadl_teardown =
 705                         (struct vmbus_channel_gpadl_teardown *)requestheader;
 706
 707                         if (gpadl_torndown->gpadl == gpadl_teardown->gpadl) {
 708                                 memcpy(&msginfo->response.gpadl_torndown,
 709                                        gpadl_torndown,
 710                                        sizeof(
 711                                         struct vmbus_channel_gpadl_torndown));
 712                                 complete(&msginfo->waitevent);
 713                                 break;
 714                         }
 715                 }
 716         }
 717         spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 718 }
 719
 720 /*
 721  * vmbus_onversion_response - Version response handler
 722  *
 723  * This is invoked when we received a response to our initiate contact request.
 724  * Find the matching request, copy the response and signal the requesting
 725  * thread.
 726  */
 727 static void vmbus_onversion_response(
 728                 struct vmbus_channel_message_header *hdr)
 729 {
 730         struct vmbus_channel_msginfo *msginfo;
 731         struct vmbus_channel_message_header *requestheader;
 732         struct vmbus_channel_version_response *version_response;
 733         unsigned long flags;
 734
 735         version_response = (struct vmbus_channel_version_response *)hdr;
 736         spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
 737
 738         list_for_each_entry(msginfo, &vmbus_connection.chn_msg_list,
 739                                 msglistentry) {
 740                 requestheader =
 741                         (struct vmbus_channel_message_header *)msginfo->msg;
 742
 743                 if (requestheader->msgtype ==
 744                     CHANNELMSG_INITIATE_CONTACT) {
 745                         memcpy(&msginfo->response.version_response,
 746                               version_response,
 747                               sizeof(struct vmbus_channel_version_response));
 748                         complete(&msginfo->waitevent);
 749                 }
 750         }
 751         spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 752 }
 753
 754 /* Channel message dispatch table */
 755 struct vmbus_channel_message_table_entry
 756         channel_message_table[CHANNELMSG_COUNT] = {
 757         {CHANNELMSG_INVALID,                    0, NULL},
 758         {CHANNELMSG_OFFERCHANNEL,               0, vmbus_onoffer},
 759         {CHANNELMSG_RESCIND_CHANNELOFFER,       0, vmbus_onoffer_rescind},
 760         {CHANNELMSG_REQUESTOFFERS,              0, NULL},
 761         {CHANNELMSG_ALLOFFERS_DELIVERED,        1, vmbus_onoffers_delivered},
 762         {CHANNELMSG_OPENCHANNEL,                0, NULL},
 763         {CHANNELMSG_OPENCHANNEL_RESULT,         1, vmbus_onopen_result},
 764         {CHANNELMSG_CLOSECHANNEL,               0, NULL},
 765         {CHANNELMSG_GPADL_HEADER,               0, NULL},
 766         {CHANNELMSG_GPADL_BODY,                 0, NULL},
 767         {CHANNELMSG_GPADL_CREATED,              1, vmbus_ongpadl_created},
 768         {CHANNELMSG_GPADL_TEARDOWN,             0, NULL},
 769         {CHANNELMSG_GPADL_TORNDOWN,             1, vmbus_ongpadl_torndown},
 770         {CHANNELMSG_RELID_RELEASED,             0, NULL},
 771         {CHANNELMSG_INITIATE_CONTACT,           0, NULL},
 772         {CHANNELMSG_VERSION_RESPONSE,           1, vmbus_onversion_response},
 773         {CHANNELMSG_UNLOAD,                     0, NULL},
 774         {CHANNELMSG_UNLOAD_RESPONSE,            1, vmbus_unload_response},
 775 };
 776
 777 /*
 778  * vmbus_onmessage - Handler for channel protocol messages.
 779  *
 780  * This is invoked in the vmbus worker thread context.
 781  */
 782 void vmbus_onmessage(void *context)
 783 {
 784         struct hv_message *msg = context;
 785         struct vmbus_channel_message_header *hdr;
 786         int size;
 787
 788         hdr = (struct vmbus_channel_message_header *)msg->u.payload;
 789         size = msg->header.payload_size;
 790
 791         if (hdr->msgtype >= CHANNELMSG_COUNT) {
 792                 pr_err("Received invalid channel message type %d size %d\n",
 793                            hdr->msgtype, size);
 794                 print_hex_dump_bytes("", DUMP_PREFIX_NONE,
 795                                      (unsigned char *)msg->u.payload, size);
 796                 return;
 797         }
 798
 799         if (channel_message_table[hdr->msgtype].message_handler)
 800                 channel_message_table[hdr->msgtype].message_handler(hdr);
 801         else
 802                 pr_err("Unhandled channel message type %d\n", hdr->msgtype);
 803 }
 804
 805 /*
 806  * vmbus_request_offers - Send a request to get all our pending offers.
 807  */
 808 int vmbus_request_offers(void)
 809 {
 810         struct vmbus_channel_message_header *msg;
 811         struct vmbus_channel_msginfo *msginfo;
 812         int ret;
 813
 814         msginfo = kmalloc(sizeof(*msginfo) +
 815                           sizeof(struct vmbus_channel_message_header),
 816                           GFP_KERNEL);
 817         if (!msginfo)
 818                 return -ENOMEM;
 819
 820         msg = (struct vmbus_channel_message_header *)msginfo->msg;
 821
 822         msg->msgtype = CHANNELMSG_REQUESTOFFERS;
 823
 824
 825         ret = vmbus_post_msg(msg,
 826                                sizeof(struct vmbus_channel_message_header));
 827         if (ret != 0) {
 828                 pr_err("Unable to request offers - %d\n", ret);
 829
 830                 goto cleanup;
 831         }
 832
 833 cleanup:
 834         kfree(msginfo);
 835
 836         return ret;
 837 }
 838
 839 /*
 840  * Retrieve the (sub) channel on which to send an outgoing request.
 841  * When a primary channel has multiple sub-channels, we try to
 842  * distribute the load equally amongst all available channels.
 843  */
 844 struct vmbus_channel *vmbus_get_outgoing_channel(struct vmbus_channel *primary)
 845 {
 846         struct list_head *cur, *tmp;
 847         int cur_cpu;
 848         struct vmbus_channel *cur_channel;
 849         struct vmbus_channel *outgoing_channel = primary;
 850         int next_channel;
 851         int i = 1;
 852
 853         if (list_empty(&primary->sc_list))
 854                 return outgoing_channel;
 855
 856         next_channel = primary->next_oc++;
 857
 858         if (next_channel > (primary->num_sc)) {
 859                 primary->next_oc = 0;
 860                 return outgoing_channel;
 861         }
 862
 863         cur_cpu = hv_context.vp_index[get_cpu()];
 864         put_cpu();
 865         list_for_each_safe(cur, tmp, &primary->sc_list) {
 866                 cur_channel = list_entry(cur, struct vmbus_channel, sc_list);
 867                 if (cur_channel->state != CHANNEL_OPENED_STATE)
 868                         continue;
 869
 870                 if (cur_channel->target_vp == cur_cpu)
 871                         return cur_channel;
 872
 873                 if (i == next_channel)
 874                         return cur_channel;
 875
 876                 i++;
 877         }
 878
 879         return outgoing_channel;
 880 }
 881 EXPORT_SYMBOL_GPL(vmbus_get_outgoing_channel);
 882
 883 static void invoke_sc_cb(struct vmbus_channel *primary_channel)
 884 {
 885         struct list_head *cur, *tmp;
 886         struct vmbus_channel *cur_channel;
 887
 888         if (primary_channel->sc_creation_callback == NULL)
 889                 return;
 890
 891         list_for_each_safe(cur, tmp, &primary_channel->sc_list) {
 892                 cur_channel = list_entry(cur, struct vmbus_channel, sc_list);
 893
 894                 primary_channel->sc_creation_callback(cur_channel);
 895         }
 896 }
 897
 898 void vmbus_set_sc_create_callback(struct vmbus_channel *primary_channel,
 899                                 void (*sc_cr_cb)(struct vmbus_channel *new_sc))
 900 {
 901         primary_channel->sc_creation_callback = sc_cr_cb;
 902 }
 903 EXPORT_SYMBOL_GPL(vmbus_set_sc_create_callback);
 904
 905 bool vmbus_are_subchannels_present(struct vmbus_channel *primary)
 906 {
 907         bool ret;
 908
 909         ret = !list_empty(&primary->sc_list);
 910
 911         if (ret) {
 912                 /*
 913                  * Invoke the callback on sub-channel creation.
 914                  * This will present a uniform interface to the
 915                  * clients.
 916                  */
 917                 invoke_sc_cb(primary);
 918         }
 919
 920         return ret;
 921 }
 922 EXPORT_SYMBOL_GPL(vmbus_are_subchannels_present);