drivers/hv/channel.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/module.h>
  29 #include <linux/hyperv.h>
  30 #include <linux/uio.h>
  31 #include <linux/interrupt.h>
  32
  33 #include "hyperv_vmbus.h"
  34
  35 #define NUM_PAGES_SPANNED(addr, len) \
  36 ((PAGE_ALIGN(addr + len) >> PAGE_SHIFT) - (addr >> PAGE_SHIFT))
  37
  38 /*
  39  * vmbus_setevent- Trigger an event notification on the specified
  40  * channel.
  41  */
  42 static void vmbus_setevent(struct vmbus_channel *channel)
  43 {
  44         struct hv_monitor_page *monitorpage;
  45
  46         /*
  47          * For channels marked as in "low latency" mode
  48          * bypass the monitor page mechanism.
  49          */
  50         if ((channel->offermsg.monitor_allocated) &&
  51             (!channel->low_latency)) {
  52                 /* Each u32 represents 32 channels */
  53                 sync_set_bit(channel->offermsg.child_relid & 31,
  54                         (unsigned long *) vmbus_connection.send_int_page +
  55                         (channel->offermsg.child_relid >> 5));
  56
  57                 /* Get the child to parent monitor page */
  58                 monitorpage = vmbus_connection.monitor_pages[1];
  59
  60                 sync_set_bit(channel->monitor_bit,
  61                         (unsigned long *)&monitorpage->trigger_group
  62                                         [channel->monitor_grp].pending);
  63
  64         } else {
  65                 vmbus_set_event(channel);
  66         }
  67 }
  68
  69 /*
  70  * vmbus_open - Open the specified channel.
  71  */
  72 int vmbus_open(struct vmbus_channel *newchannel, u32 send_ringbuffer_size,
  73                      u32 recv_ringbuffer_size, void *userdata, u32 userdatalen,
  74                      void (*onchannelcallback)(void *context), void *context)
  75 {
  76         struct vmbus_channel_open_channel *open_msg;
  77         struct vmbus_channel_msginfo *open_info = NULL;
  78         void *in, *out;
  79         unsigned long flags;
  80         int ret, err = 0;
  81         struct page *page;
  82
  83         spin_lock_irqsave(&newchannel->lock, flags);
  84         if (newchannel->state == CHANNEL_OPEN_STATE) {
  85                 newchannel->state = CHANNEL_OPENING_STATE;
  86         } else {
  87                 spin_unlock_irqrestore(&newchannel->lock, flags);
  88                 return -EINVAL;
  89         }
  90         spin_unlock_irqrestore(&newchannel->lock, flags);
  91
  92         newchannel->onchannel_callback = onchannelcallback;
  93         newchannel->channel_callback_context = context;
  94
  95         /* Allocate the ring buffer */
  96         page = alloc_pages_node(cpu_to_node(newchannel->target_cpu),
  97                                 GFP_KERNEL|__GFP_ZERO,
  98                                 get_order(send_ringbuffer_size +
  99                                 recv_ringbuffer_size));
 100
 101         if (!page)
 102                 out = (void *)__get_free_pages(GFP_KERNEL|__GFP_ZERO,
 103                                                get_order(send_ringbuffer_size +
 104                                                recv_ringbuffer_size));
 105         else
 106                 out = (void *)page_address(page);
 107
 108         if (!out) {
 109                 err = -ENOMEM;
 110                 goto error0;
 111         }
 112
 113         in = (void *)((unsigned long)out + send_ringbuffer_size);
 114
 115         newchannel->ringbuffer_pages = out;
 116         newchannel->ringbuffer_pagecount = (send_ringbuffer_size +
 117                                            recv_ringbuffer_size) >> PAGE_SHIFT;
 118
 119         ret = hv_ringbuffer_init(
 120                 &newchannel->outbound, out, send_ringbuffer_size);
 121
 122         if (ret != 0) {
 123                 err = ret;
 124                 goto error0;
 125         }
 126
 127         ret = hv_ringbuffer_init(
 128                 &newchannel->inbound, in, recv_ringbuffer_size);
 129         if (ret != 0) {
 130                 err = ret;
 131                 goto error0;
 132         }
 133
 134
 135         /* Establish the gpadl for the ring buffer */
 136         newchannel->ringbuffer_gpadlhandle = 0;
 137
 138         ret = vmbus_establish_gpadl(newchannel,
 139                                          newchannel->outbound.ring_buffer,
 140                                          send_ringbuffer_size +
 141                                          recv_ringbuffer_size,
 142                                          &newchannel->ringbuffer_gpadlhandle);
 143
 144         if (ret != 0) {
 145                 err = ret;
 146                 goto error0;
 147         }
 148
 149         /* Create and init the channel open message */
 150         open_info = kmalloc(sizeof(*open_info) +
 151                            sizeof(struct vmbus_channel_open_channel),
 152                            GFP_KERNEL);
 153         if (!open_info) {
 154                 err = -ENOMEM;
 155                 goto error_gpadl;
 156         }
 157
 158         init_completion(&open_info->waitevent);
 159
 160         open_msg = (struct vmbus_channel_open_channel *)open_info->msg;
 161         open_msg->header.msgtype = CHANNELMSG_OPENCHANNEL;
 162         open_msg->openid = newchannel->offermsg.child_relid;
 163         open_msg->child_relid = newchannel->offermsg.child_relid;
 164         open_msg->ringbuffer_gpadlhandle = newchannel->ringbuffer_gpadlhandle;
 165         open_msg->downstream_ringbuffer_pageoffset = send_ringbuffer_size >>
 166                                                   PAGE_SHIFT;
 167         open_msg->target_vp = newchannel->target_vp;
 168
 169         if (userdatalen > MAX_USER_DEFINED_BYTES) {
 170                 err = -EINVAL;
 171                 goto error_gpadl;
 172         }
 173
 174         if (userdatalen)
 175                 memcpy(open_msg->userdata, userdata, userdatalen);
 176
 177         spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
 178         list_add_tail(&open_info->msglistentry,
 179                       &vmbus_connection.chn_msg_list);
 180         spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 181
 182         ret = vmbus_post_msg(open_msg,
 183                                sizeof(struct vmbus_channel_open_channel));
 184
 185         if (ret != 0) {
 186                 err = ret;
 187                 goto error1;
 188         }
 189
 190         wait_for_completion(&open_info->waitevent);
 191
 192         spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
 193         list_del(&open_info->msglistentry);
 194         spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 195
 196         if (open_info->response.open_result.status) {
 197                 err = -EAGAIN;
 198                 goto error_gpadl;
 199         }
 200
 201         newchannel->state = CHANNEL_OPENED_STATE;
 202         kfree(open_info);
 203         return 0;
 204
 205 error1:
 206         spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
 207         list_del(&open_info->msglistentry);
 208         spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 209
 210 error_gpadl:
 211         vmbus_teardown_gpadl(newchannel, newchannel->ringbuffer_gpadlhandle);
 212
 213 error0:
 214         free_pages((unsigned long)out,
 215                 get_order(send_ringbuffer_size + recv_ringbuffer_size));
 216         kfree(open_info);
 217         newchannel->state = CHANNEL_OPEN_STATE;
 218         return err;
 219 }
 220 EXPORT_SYMBOL_GPL(vmbus_open);
 221
 222 /* Used for Hyper-V Socket: a guest client's connect() to the host */
 223 int vmbus_send_tl_connect_request(const uuid_le *shv_guest_servie_id,
 224                                   const uuid_le *shv_host_servie_id)
 225 {
 226         struct vmbus_channel_tl_connect_request conn_msg;
 227
 228         memset(&conn_msg, 0, sizeof(conn_msg));
 229         conn_msg.header.msgtype = CHANNELMSG_TL_CONNECT_REQUEST;
 230         conn_msg.guest_endpoint_id = *shv_guest_servie_id;
 231         conn_msg.host_service_id = *shv_host_servie_id;
 232
 233         return vmbus_post_msg(&conn_msg, sizeof(conn_msg));
 234 }
 235 EXPORT_SYMBOL_GPL(vmbus_send_tl_connect_request);
 236
 237 /*
 238  * create_gpadl_header - Creates a gpadl for the specified buffer
 239  */
 240 static int create_gpadl_header(void *kbuffer, u32 size,
 241                                struct vmbus_channel_msginfo **msginfo)
 242 {
 243         int i;
 244         int pagecount;
 245         struct vmbus_channel_gpadl_header *gpadl_header;
 246         struct vmbus_channel_gpadl_body *gpadl_body;
 247         struct vmbus_channel_msginfo *msgheader;
 248         struct vmbus_channel_msginfo *msgbody = NULL;
 249         u32 msgsize;
 250
 251         int pfnsum, pfncount, pfnleft, pfncurr, pfnsize;
 252
 253         pagecount = size >> PAGE_SHIFT;
 254
 255         /* do we need a gpadl body msg */
 256         pfnsize = MAX_SIZE_CHANNEL_MESSAGE -
 257                   sizeof(struct vmbus_channel_gpadl_header) -
 258                   sizeof(struct gpa_range);
 259         pfncount = pfnsize / sizeof(u64);
 260
 261         if (pagecount > pfncount) {
 262                 /* we need a gpadl body */
 263                 /* fill in the header */
 264                 msgsize = sizeof(struct vmbus_channel_msginfo) +
 265                           sizeof(struct vmbus_channel_gpadl_header) +
 266                           sizeof(struct gpa_range) + pfncount * sizeof(u64);
 267                 msgheader =  kzalloc(msgsize, GFP_KERNEL);
 268                 if (!msgheader)
 269                         goto nomem;
 270
 271                 INIT_LIST_HEAD(&msgheader->submsglist);
 272                 msgheader->msgsize = msgsize;
 273
 274                 gpadl_header = (struct vmbus_channel_gpadl_header *)
 275                         msgheader->msg;
 276                 gpadl_header->rangecount = 1;
 277                 gpadl_header->range_buflen = sizeof(struct gpa_range) +
 278                                          pagecount * sizeof(u64);
 279                 gpadl_header->range[0].byte_offset = 0;
 280                 gpadl_header->range[0].byte_count = size;
 281                 for (i = 0; i < pfncount; i++)
 282                         gpadl_header->range[0].pfn_array[i] = slow_virt_to_phys(
 283                                 kbuffer + PAGE_SIZE * i) >> PAGE_SHIFT;
 284                 *msginfo = msgheader;
 285
 286                 pfnsum = pfncount;
 287                 pfnleft = pagecount - pfncount;
 288
 289                 /* how many pfns can we fit */
 290                 pfnsize = MAX_SIZE_CHANNEL_MESSAGE -
 291                           sizeof(struct vmbus_channel_gpadl_body);
 292                 pfncount = pfnsize / sizeof(u64);
 293
 294                 /* fill in the body */
 295                 while (pfnleft) {
 296                         if (pfnleft > pfncount)
 297                                 pfncurr = pfncount;
 298                         else
 299                                 pfncurr = pfnleft;
 300
 301                         msgsize = sizeof(struct vmbus_channel_msginfo) +
 302                                   sizeof(struct vmbus_channel_gpadl_body) +
 303                                   pfncurr * sizeof(u64);
 304                         msgbody = kzalloc(msgsize, GFP_KERNEL);
 305
 306                         if (!msgbody) {
 307                                 struct vmbus_channel_msginfo *pos = NULL;
 308                                 struct vmbus_channel_msginfo *tmp = NULL;
 309                                 /*
 310                                  * Free up all the allocated messages.
 311                                  */
 312                                 list_for_each_entry_safe(pos, tmp,
 313                                         &msgheader->submsglist,
 314                                         msglistentry) {
 315
 316                                         list_del(&pos->msglistentry);
 317                                         kfree(pos);
 318                                 }
 319
 320                                 goto nomem;
 321                         }
 322
 323                         msgbody->msgsize = msgsize;
 324                         gpadl_body =
 325                                 (struct vmbus_channel_gpadl_body *)msgbody->msg;
 326
 327                         /*
 328                          * Gpadl is u32 and we are using a pointer which could
 329                          * be 64-bit
 330                          * This is governed by the guest/host protocol and
 331                          * so the hypervisor gurantees that this is ok.
 332                          */
 333                         for (i = 0; i < pfncurr; i++)
 334                                 gpadl_body->pfn[i] = slow_virt_to_phys(
 335                                         kbuffer + PAGE_SIZE * (pfnsum + i)) >>
 336                                         PAGE_SHIFT;
 337
 338                         /* add to msg header */
 339                         list_add_tail(&msgbody->msglistentry,
 340                                       &msgheader->submsglist);
 341                         pfnsum += pfncurr;
 342                         pfnleft -= pfncurr;
 343                 }
 344         } else {
 345                 /* everything fits in a header */
 346                 msgsize = sizeof(struct vmbus_channel_msginfo) +
 347                           sizeof(struct vmbus_channel_gpadl_header) +
 348                           sizeof(struct gpa_range) + pagecount * sizeof(u64);
 349                 msgheader = kzalloc(msgsize, GFP_KERNEL);
 350                 if (msgheader == NULL)
 351                         goto nomem;
 352
 353                 INIT_LIST_HEAD(&msgheader->submsglist);
 354                 msgheader->msgsize = msgsize;
 355
 356                 gpadl_header = (struct vmbus_channel_gpadl_header *)
 357                         msgheader->msg;
 358                 gpadl_header->rangecount = 1;
 359                 gpadl_header->range_buflen = sizeof(struct gpa_range) +
 360                                          pagecount * sizeof(u64);
 361                 gpadl_header->range[0].byte_offset = 0;
 362                 gpadl_header->range[0].byte_count = size;
 363                 for (i = 0; i < pagecount; i++)
 364                         gpadl_header->range[0].pfn_array[i] = slow_virt_to_phys(
 365                                 kbuffer + PAGE_SIZE * i) >> PAGE_SHIFT;
 366
 367                 *msginfo = msgheader;
 368         }
 369
 370         return 0;
 371 nomem:
 372         kfree(msgheader);
 373         kfree(msgbody);
 374         return -ENOMEM;
 375 }
 376
 377 /*
 378  * vmbus_establish_gpadl - Estabish a GPADL for the specified buffer
 379  *
 380  * @channel: a channel
 381  * @kbuffer: from kmalloc or vmalloc
 382  * @size: page-size multiple
 383  * @gpadl_handle: some funky thing
 384  */
 385 int vmbus_establish_gpadl(struct vmbus_channel *channel, void *kbuffer,
 386                                u32 size, u32 *gpadl_handle)
 387 {
 388         struct vmbus_channel_gpadl_header *gpadlmsg;
 389         struct vmbus_channel_gpadl_body *gpadl_body;
 390         struct vmbus_channel_msginfo *msginfo = NULL;
 391         struct vmbus_channel_msginfo *submsginfo, *tmp;
 392         struct list_head *curr;
 393         u32 next_gpadl_handle;
 394         unsigned long flags;
 395         int ret = 0;
 396
 397         next_gpadl_handle =
 398                 (atomic_inc_return(&vmbus_connection.next_gpadl_handle) - 1);
 399
 400         ret = create_gpadl_header(kbuffer, size, &msginfo);
 401         if (ret)
 402                 return ret;
 403
 404         init_completion(&msginfo->waitevent);
 405
 406         gpadlmsg = (struct vmbus_channel_gpadl_header *)msginfo->msg;
 407         gpadlmsg->header.msgtype = CHANNELMSG_GPADL_HEADER;
 408         gpadlmsg->child_relid = channel->offermsg.child_relid;
 409         gpadlmsg->gpadl = next_gpadl_handle;
 410
 411
 412         spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
 413         list_add_tail(&msginfo->msglistentry,
 414                       &vmbus_connection.chn_msg_list);
 415
 416         spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 417
 418         ret = vmbus_post_msg(gpadlmsg, msginfo->msgsize -
 419                                sizeof(*msginfo));
 420         if (ret != 0)
 421                 goto cleanup;
 422
 423         list_for_each(curr, &msginfo->submsglist) {
 424                 submsginfo = (struct vmbus_channel_msginfo *)curr;
 425                 gpadl_body =
 426                         (struct vmbus_channel_gpadl_body *)submsginfo->msg;
 427
 428                 gpadl_body->header.msgtype =
 429                         CHANNELMSG_GPADL_BODY;
 430                 gpadl_body->gpadl = next_gpadl_handle;
 431
 432                 ret = vmbus_post_msg(gpadl_body,
 433                                      submsginfo->msgsize -
 434                                      sizeof(*submsginfo));
 435                 if (ret != 0)
 436                         goto cleanup;
 437
 438         }
 439         wait_for_completion(&msginfo->waitevent);
 440
 441         /* At this point, we received the gpadl created msg */
 442         *gpadl_handle = gpadlmsg->gpadl;
 443
 444 cleanup:
 445         spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
 446         list_del(&msginfo->msglistentry);
 447         spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 448         list_for_each_entry_safe(submsginfo, tmp, &msginfo->submsglist,
 449                                  msglistentry) {
 450                 kfree(submsginfo);
 451         }
 452
 453         kfree(msginfo);
 454         return ret;
 455 }
 456 EXPORT_SYMBOL_GPL(vmbus_establish_gpadl);
 457
 458 /*
 459  * vmbus_teardown_gpadl -Teardown the specified GPADL handle
 460  */
 461 int vmbus_teardown_gpadl(struct vmbus_channel *channel, u32 gpadl_handle)
 462 {
 463         struct vmbus_channel_gpadl_teardown *msg;
 464         struct vmbus_channel_msginfo *info;
 465         unsigned long flags;
 466         int ret;
 467
 468         info = kmalloc(sizeof(*info) +
 469                        sizeof(struct vmbus_channel_gpadl_teardown), GFP_KERNEL);
 470         if (!info)
 471                 return -ENOMEM;
 472
 473         init_completion(&info->waitevent);
 474
 475         msg = (struct vmbus_channel_gpadl_teardown *)info->msg;
 476
 477         msg->header.msgtype = CHANNELMSG_GPADL_TEARDOWN;
 478         msg->child_relid = channel->offermsg.child_relid;
 479         msg->gpadl = gpadl_handle;
 480
 481         spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
 482         list_add_tail(&info->msglistentry,
 483                       &vmbus_connection.chn_msg_list);
 484         spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 485         ret = vmbus_post_msg(msg,
 486                                sizeof(struct vmbus_channel_gpadl_teardown));
 487
 488         if (ret)
 489                 goto post_msg_err;
 490
 491         wait_for_completion(&info->waitevent);
 492
 493 post_msg_err:
 494         spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
 495         list_del(&info->msglistentry);
 496         spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
 497
 498         kfree(info);
 499         return ret;
 500 }
 501 EXPORT_SYMBOL_GPL(vmbus_teardown_gpadl);
 502
 503 static void reset_channel_cb(void *arg)
 504 {
 505         struct vmbus_channel *channel = arg;
 506
 507         channel->onchannel_callback = NULL;
 508 }
 509
 510 static int vmbus_close_internal(struct vmbus_channel *channel)
 511 {
 512         struct vmbus_channel_close_channel *msg;
 513         int ret;
 514
 515         /*
 516          * process_chn_event(), running in the tasklet, can race
 517          * with vmbus_close_internal() in the case of SMP guest, e.g., when
 518          * the former is accessing channel->inbound.ring_buffer, the latter
 519          * could be freeing the ring_buffer pages.
 520          *
 521          * To resolve the race, we can serialize them by disabling the
 522          * tasklet when the latter is running here.
 523          */
 524         hv_event_tasklet_disable(channel);
 525
 526         /*
 527          * In case a device driver's probe() fails (e.g.,
 528          * util_probe() -> vmbus_open() returns -ENOMEM) and the device is
 529          * rescinded later (e.g., we dynamically disble an Integrated Service
 530          * in Hyper-V Manager), the driver's remove() invokes vmbus_close():
 531          * here we should skip most of the below cleanup work.
 532          */
 533         if (channel->state != CHANNEL_OPENED_STATE) {
 534                 ret = -EINVAL;
 535                 goto out;
 536         }
 537
 538         channel->state = CHANNEL_OPEN_STATE;
 539         channel->sc_creation_callback = NULL;
 540         /* Stop callback and cancel the timer asap */
 541         if (channel->target_cpu != get_cpu()) {
 542                 put_cpu();
 543                 smp_call_function_single(channel->target_cpu, reset_channel_cb,
 544                                          channel, true);
 545         } else {
 546                 reset_channel_cb(channel);
 547                 put_cpu();
 548         }
 549
 550         /* Send a closing message */
 551
 552         msg = &channel->close_msg.msg;
 553
 554         msg->header.msgtype = CHANNELMSG_CLOSECHANNEL;
 555         msg->child_relid = channel->offermsg.child_relid;
 556
 557         ret = vmbus_post_msg(msg, sizeof(struct vmbus_channel_close_channel));
 558
 559         if (ret) {
 560                 pr_err("Close failed: close post msg return is %d\n", ret);
 561                 /*
 562                  * If we failed to post the close msg,
 563                  * it is perhaps better to leak memory.
 564                  */
 565                 goto out;
 566         }
 567
 568         /* Tear down the gpadl for the channel's ring buffer */
 569         if (channel->ringbuffer_gpadlhandle) {
 570                 ret = vmbus_teardown_gpadl(channel,
 571                                            channel->ringbuffer_gpadlhandle);
 572                 if (ret) {
 573                         pr_err("Close failed: teardown gpadl return %d\n", ret);
 574                         /*
 575                          * If we failed to teardown gpadl,
 576                          * it is perhaps better to leak memory.
 577                          */
 578                         goto out;
 579                 }
 580         }
 581
 582         /* Cleanup the ring buffers for this channel */
 583         hv_ringbuffer_cleanup(&channel->outbound);
 584         hv_ringbuffer_cleanup(&channel->inbound);
 585
 586         free_pages((unsigned long)channel->ringbuffer_pages,
 587                 get_order(channel->ringbuffer_pagecount * PAGE_SIZE));
 588
 589 out:
 590         hv_event_tasklet_enable(channel);
 591
 592         return ret;
 593 }
 594
 595 /*
 596  * vmbus_close - Close the specified channel
 597  */
 598 void vmbus_close(struct vmbus_channel *channel)
 599 {
 600         struct list_head *cur, *tmp;
 601         struct vmbus_channel *cur_channel;
 602
 603         if (channel->primary_channel != NULL) {
 604                 /*
 605                  * We will only close sub-channels when
 606                  * the primary is closed.
 607                  */
 608                 return;
 609         }
 610         /*
 611          * Close all the sub-channels first and then close the
 612          * primary channel.
 613          */
 614         list_for_each_safe(cur, tmp, &channel->sc_list) {
 615                 cur_channel = list_entry(cur, struct vmbus_channel, sc_list);
 616                 if (cur_channel->state != CHANNEL_OPENED_STATE)
 617                         continue;
 618                 vmbus_close_internal(cur_channel);
 619         }
 620         /*
 621          * Now close the primary.
 622          */
 623         vmbus_close_internal(channel);
 624 }
 625 EXPORT_SYMBOL_GPL(vmbus_close);
 626
 627 int vmbus_sendpacket_ctl(struct vmbus_channel *channel, void *buffer,
 628                            u32 bufferlen, u64 requestid,
 629                            enum vmbus_packet_type type, u32 flags, bool kick_q)
 630 {
 631         struct vmpacket_descriptor desc;
 632         u32 packetlen = sizeof(struct vmpacket_descriptor) + bufferlen;
 633         u32 packetlen_aligned = ALIGN(packetlen, sizeof(u64));
 634         struct kvec bufferlist[3];
 635         u64 aligned_data = 0;
 636         int ret;
 637         bool signal = false;
 638         bool lock = channel->acquire_ring_lock;
 639         int num_vecs = ((bufferlen != 0) ? 3 : 1);
 640
 641
 642         /* Setup the descriptor */
 643         desc.type = type; /* VmbusPacketTypeDataInBand; */
 644         desc.flags = flags; /* VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED; */
 645         /* in 8-bytes granularity */
 646         desc.offset8 = sizeof(struct vmpacket_descriptor) >> 3;
 647         desc.len8 = (u16)(packetlen_aligned >> 3);
 648         desc.trans_id = requestid;
 649
 650         bufferlist[0].iov_base = &desc;
 651         bufferlist[0].iov_len = sizeof(struct vmpacket_descriptor);
 652         bufferlist[1].iov_base = buffer;
 653         bufferlist[1].iov_len = bufferlen;
 654         bufferlist[2].iov_base = &aligned_data;
 655         bufferlist[2].iov_len = (packetlen_aligned - packetlen);
 656
 657         ret = hv_ringbuffer_write(&channel->outbound, bufferlist, num_vecs,
 658                                   &signal, lock, channel->signal_policy);
 659
 660         /*
 661          * Signalling the host is conditional on many factors:
 662          * 1. The ring state changed from being empty to non-empty.
 663          *    This is tracked by the variable "signal".
 664          * 2. The variable kick_q tracks if more data will be placed
 665          *    on the ring. We will not signal if more data is
 666          *    to be placed.
 667          *
 668          * Based on the channel signal state, we will decide
 669          * which signaling policy will be applied.
 670          *
 671          * If we cannot write to the ring-buffer; signal the host
 672          * even if we may not have written anything. This is a rare
 673          * enough condition that it should not matter.
 674          * NOTE: in this case, the hvsock channel is an exception, because
 675          * it looks the host side's hvsock implementation has a throttling
 676          * mechanism which can hurt the performance otherwise.
 677          */
 678
 679         if (((ret == 0) && kick_q && signal) ||
 680             (ret && !is_hvsock_channel(channel)))
 681                 vmbus_setevent(channel);
 682
 683         return ret;
 684 }
 685 EXPORT_SYMBOL(vmbus_sendpacket_ctl);
 686
 687 /**
 688  * vmbus_sendpacket() - Send the specified buffer on the given channel
 689  * @channel: Pointer to vmbus_channel structure.
 690  * @buffer: Pointer to the buffer you want to receive the data into.
 691  * @bufferlen: Maximum size of what the the buffer will hold
 692  * @requestid: Identifier of the request
 693  * @type: Type of packet that is being send e.g. negotiate, time
 694  * packet etc.
 695  *
 696  * Sends data in @buffer directly to hyper-v via the vmbus
 697  * This will send the data unparsed to hyper-v.
 698  *
 699  * Mainly used by Hyper-V drivers.
 700  */
 701 int vmbus_sendpacket(struct vmbus_channel *channel, void *buffer,
 702                            u32 bufferlen, u64 requestid,
 703                            enum vmbus_packet_type type, u32 flags)
 704 {
 705         return vmbus_sendpacket_ctl(channel, buffer, bufferlen, requestid,
 706                                     type, flags, true);
 707 }
 708 EXPORT_SYMBOL(vmbus_sendpacket);
 709
 710 /*
 711  * vmbus_sendpacket_pagebuffer_ctl - Send a range of single-page buffer
 712  * packets using a GPADL Direct packet type. This interface allows you
 713  * to control notifying the host. This will be useful for sending
 714  * batched data. Also the sender can control the send flags
 715  * explicitly.
 716  */
 717 int vmbus_sendpacket_pagebuffer_ctl(struct vmbus_channel *channel,
 718                                      struct hv_page_buffer pagebuffers[],
 719                                      u32 pagecount, void *buffer, u32 bufferlen,
 720                                      u64 requestid,
 721                                      u32 flags,
 722                                      bool kick_q)
 723 {
 724         int ret;
 725         int i;
 726         struct vmbus_channel_packet_page_buffer desc;
 727         u32 descsize;
 728         u32 packetlen;
 729         u32 packetlen_aligned;
 730         struct kvec bufferlist[3];
 731         u64 aligned_data = 0;
 732         bool signal = false;
 733         bool lock = channel->acquire_ring_lock;
 734
 735         if (pagecount > MAX_PAGE_BUFFER_COUNT)
 736                 return -EINVAL;
 737
 738
 739         /*
 740          * Adjust the size down since vmbus_channel_packet_page_buffer is the
 741          * largest size we support
 742          */
 743         descsize = sizeof(struct vmbus_channel_packet_page_buffer) -
 744                           ((MAX_PAGE_BUFFER_COUNT - pagecount) *
 745                           sizeof(struct hv_page_buffer));
 746         packetlen = descsize + bufferlen;
 747         packetlen_aligned = ALIGN(packetlen, sizeof(u64));
 748
 749         /* Setup the descriptor */
 750         desc.type = VM_PKT_DATA_USING_GPA_DIRECT;
 751         desc.flags = flags;
 752         desc.dataoffset8 = descsize >> 3; /* in 8-bytes grandularity */
 753         desc.length8 = (u16)(packetlen_aligned >> 3);
 754         desc.transactionid = requestid;
 755         desc.rangecount = pagecount;
 756
 757         for (i = 0; i < pagecount; i++) {
 758                 desc.range[i].len = pagebuffers[i].len;
 759                 desc.range[i].offset = pagebuffers[i].offset;
 760                 desc.range[i].pfn        = pagebuffers[i].pfn;
 761         }
 762
 763         bufferlist[0].iov_base = &desc;
 764         bufferlist[0].iov_len = descsize;
 765         bufferlist[1].iov_base = buffer;
 766         bufferlist[1].iov_len = bufferlen;
 767         bufferlist[2].iov_base = &aligned_data;
 768         bufferlist[2].iov_len = (packetlen_aligned - packetlen);
 769
 770         ret = hv_ringbuffer_write(&channel->outbound, bufferlist, 3,
 771                                   &signal, lock, channel->signal_policy);
 772
 773         /*
 774          * Signalling the host is conditional on many factors:
 775          * 1. The ring state changed from being empty to non-empty.
 776          *    This is tracked by the variable "signal".
 777          * 2. The variable kick_q tracks if more data will be placed
 778          *    on the ring. We will not signal if more data is
 779          *    to be placed.
 780          *
 781          * Based on the channel signal state, we will decide
 782          * which signaling policy will be applied.
 783          *
 784          * If we cannot write to the ring-buffer; signal the host
 785          * even if we may not have written anything. This is a rare
 786          * enough condition that it should not matter.
 787          */
 788
 789         if (((ret == 0) && kick_q && signal) || (ret))
 790                 vmbus_setevent(channel);
 791
 792         return ret;
 793 }
 794 EXPORT_SYMBOL_GPL(vmbus_sendpacket_pagebuffer_ctl);
 795
 796 /*
 797  * vmbus_sendpacket_pagebuffer - Send a range of single-page buffer
 798  * packets using a GPADL Direct packet type.
 799  */
 800 int vmbus_sendpacket_pagebuffer(struct vmbus_channel *channel,
 801                                      struct hv_page_buffer pagebuffers[],
 802                                      u32 pagecount, void *buffer, u32 bufferlen,
 803                                      u64 requestid)
 804 {
 805         u32 flags = VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED;
 806         return vmbus_sendpacket_pagebuffer_ctl(channel, pagebuffers, pagecount,
 807                                                buffer, bufferlen, requestid,
 808                                                flags, true);
 809
 810 }
 811 EXPORT_SYMBOL_GPL(vmbus_sendpacket_pagebuffer);
 812
 813 /*
 814  * vmbus_sendpacket_multipagebuffer - Send a multi-page buffer packet
 815  * using a GPADL Direct packet type.
 816  * The buffer includes the vmbus descriptor.
 817  */
 818 int vmbus_sendpacket_mpb_desc(struct vmbus_channel *channel,
 819                               struct vmbus_packet_mpb_array *desc,
 820                               u32 desc_size,
 821                               void *buffer, u32 bufferlen, u64 requestid)
 822 {
 823         int ret;
 824         u32 packetlen;
 825         u32 packetlen_aligned;
 826         struct kvec bufferlist[3];
 827         u64 aligned_data = 0;
 828         bool signal = false;
 829         bool lock = channel->acquire_ring_lock;
 830
 831         packetlen = desc_size + bufferlen;
 832         packetlen_aligned = ALIGN(packetlen, sizeof(u64));
 833
 834         /* Setup the descriptor */
 835         desc->type = VM_PKT_DATA_USING_GPA_DIRECT;
 836         desc->flags = VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED;
 837         desc->dataoffset8 = desc_size >> 3; /* in 8-bytes grandularity */
 838         desc->length8 = (u16)(packetlen_aligned >> 3);
 839         desc->transactionid = requestid;
 840         desc->rangecount = 1;
 841
 842         bufferlist[0].iov_base = desc;
 843         bufferlist[0].iov_len = desc_size;
 844         bufferlist[1].iov_base = buffer;
 845         bufferlist[1].iov_len = bufferlen;
 846         bufferlist[2].iov_base = &aligned_data;
 847         bufferlist[2].iov_len = (packetlen_aligned - packetlen);
 848
 849         ret = hv_ringbuffer_write(&channel->outbound, bufferlist, 3,
 850                                   &signal, lock, channel->signal_policy);
 851
 852         if (ret == 0 && signal)
 853                 vmbus_setevent(channel);
 854
 855         return ret;
 856 }
 857 EXPORT_SYMBOL_GPL(vmbus_sendpacket_mpb_desc);
 858
 859 /*
 860  * vmbus_sendpacket_multipagebuffer - Send a multi-page buffer packet
 861  * using a GPADL Direct packet type.
 862  */
 863 int vmbus_sendpacket_multipagebuffer(struct vmbus_channel *channel,
 864                                 struct hv_multipage_buffer *multi_pagebuffer,
 865                                 void *buffer, u32 bufferlen, u64 requestid)
 866 {
 867         int ret;
 868         struct vmbus_channel_packet_multipage_buffer desc;
 869         u32 descsize;
 870         u32 packetlen;
 871         u32 packetlen_aligned;
 872         struct kvec bufferlist[3];
 873         u64 aligned_data = 0;
 874         bool signal = false;
 875         bool lock = channel->acquire_ring_lock;
 876         u32 pfncount = NUM_PAGES_SPANNED(multi_pagebuffer->offset,
 877                                          multi_pagebuffer->len);
 878
 879         if (pfncount > MAX_MULTIPAGE_BUFFER_COUNT)
 880                 return -EINVAL;
 881
 882         /*
 883          * Adjust the size down since vmbus_channel_packet_multipage_buffer is
 884          * the largest size we support
 885          */
 886         descsize = sizeof(struct vmbus_channel_packet_multipage_buffer) -
 887                           ((MAX_MULTIPAGE_BUFFER_COUNT - pfncount) *
 888                           sizeof(u64));
 889         packetlen = descsize + bufferlen;
 890         packetlen_aligned = ALIGN(packetlen, sizeof(u64));
 891
 892
 893         /* Setup the descriptor */
 894         desc.type = VM_PKT_DATA_USING_GPA_DIRECT;
 895         desc.flags = VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED;
 896         desc.dataoffset8 = descsize >> 3; /* in 8-bytes grandularity */
 897         desc.length8 = (u16)(packetlen_aligned >> 3);
 898         desc.transactionid = requestid;
 899         desc.rangecount = 1;
 900
 901         desc.range.len = multi_pagebuffer->len;
 902         desc.range.offset = multi_pagebuffer->offset;
 903
 904         memcpy(desc.range.pfn_array, multi_pagebuffer->pfn_array,
 905                pfncount * sizeof(u64));
 906
 907         bufferlist[0].iov_base = &desc;
 908         bufferlist[0].iov_len = descsize;
 909         bufferlist[1].iov_base = buffer;
 910         bufferlist[1].iov_len = bufferlen;
 911         bufferlist[2].iov_base = &aligned_data;
 912         bufferlist[2].iov_len = (packetlen_aligned - packetlen);
 913
 914         ret = hv_ringbuffer_write(&channel->outbound, bufferlist, 3,
 915                                   &signal, lock, channel->signal_policy);
 916
 917         if (ret == 0 && signal)
 918                 vmbus_setevent(channel);
 919
 920         return ret;
 921 }
 922 EXPORT_SYMBOL_GPL(vmbus_sendpacket_multipagebuffer);
 923
 924 /**
 925  * vmbus_recvpacket() - Retrieve the user packet on the specified channel
 926  * @channel: Pointer to vmbus_channel structure.
 927  * @buffer: Pointer to the buffer you want to receive the data into.
 928  * @bufferlen: Maximum size of what the the buffer will hold
 929  * @buffer_actual_len: The actual size of the data after it was received
 930  * @requestid: Identifier of the request
 931  *
 932  * Receives directly from the hyper-v vmbus and puts the data it received
 933  * into Buffer. This will receive the data unparsed from hyper-v.
 934  *
 935  * Mainly used by Hyper-V drivers.
 936  */
 937 static inline int
 938 __vmbus_recvpacket(struct vmbus_channel *channel, void *buffer,
 939                    u32 bufferlen, u32 *buffer_actual_len, u64 *requestid,
 940                    bool raw)
 941 {
 942         int ret;
 943         bool signal = false;
 944
 945         ret = hv_ringbuffer_read(&channel->inbound, buffer, bufferlen,
 946                                  buffer_actual_len, requestid, &signal, raw);
 947
 948         if (signal)
 949                 vmbus_setevent(channel);
 950
 951         return ret;
 952 }
 953
 954 int vmbus_recvpacket(struct vmbus_channel *channel, void *buffer,
 955                      u32 bufferlen, u32 *buffer_actual_len,
 956                      u64 *requestid)
 957 {
 958         return __vmbus_recvpacket(channel, buffer, bufferlen,
 959                                   buffer_actual_len, requestid, false);
 960 }
 961 EXPORT_SYMBOL(vmbus_recvpacket);
 962
 963 /*
 964  * vmbus_recvpacket_raw - Retrieve the raw packet on the specified channel
 965  */
 966 int vmbus_recvpacket_raw(struct vmbus_channel *channel, void *buffer,
 967                               u32 bufferlen, u32 *buffer_actual_len,
 968                               u64 *requestid)
 969 {
 970         return __vmbus_recvpacket(channel, buffer, bufferlen,
 971                                   buffer_actual_len, requestid, true);
 972 }
 973 EXPORT_SYMBOL_GPL(vmbus_recvpacket_raw);