Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / hv / channel_mgmt.c
diff --git a/kernel/drivers/hv/channel_mgmt.c b/kernel/drivers/hv/channel_mgmt.c
new file mode 100644 (file)
index 0000000..0eeb1b3
--- /dev/null
@@ -0,0 +1,862 @@
+/*
+ * Copyright (c) 2009, Microsoft Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+ * Place - Suite 330, Boston, MA 02111-1307 USA.
+ *
+ * Authors:
+ *   Haiyang Zhang <haiyangz@microsoft.com>
+ *   Hank Janssen  <hjanssen@microsoft.com>
+ */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/wait.h>
+#include <linux/mm.h>
+#include <linux/slab.h>
+#include <linux/list.h>
+#include <linux/module.h>
+#include <linux/completion.h>
+#include <linux/hyperv.h>
+
+#include "hyperv_vmbus.h"
+
+/**
+ * vmbus_prep_negotiate_resp() - Create default response for Hyper-V Negotiate message
+ * @icmsghdrp: Pointer to msg header structure
+ * @icmsg_negotiate: Pointer to negotiate message structure
+ * @buf: Raw buffer channel data
+ *
+ * @icmsghdrp is of type &struct icmsg_hdr.
+ * @negop is of type &struct icmsg_negotiate.
+ * Set up and fill in default negotiate response message.
+ *
+ * The fw_version specifies the  framework version that
+ * we can support and srv_version specifies the service
+ * version we can support.
+ *
+ * Mainly used by Hyper-V drivers.
+ */
+bool vmbus_prep_negotiate_resp(struct icmsg_hdr *icmsghdrp,
+                               struct icmsg_negotiate *negop, u8 *buf,
+                               int fw_version, int srv_version)
+{
+       int icframe_major, icframe_minor;
+       int icmsg_major, icmsg_minor;
+       int fw_major, fw_minor;
+       int srv_major, srv_minor;
+       int i;
+       bool found_match = false;
+
+       icmsghdrp->icmsgsize = 0x10;
+       fw_major = (fw_version >> 16);
+       fw_minor = (fw_version & 0xFFFF);
+
+       srv_major = (srv_version >> 16);
+       srv_minor = (srv_version & 0xFFFF);
+
+       negop = (struct icmsg_negotiate *)&buf[
+               sizeof(struct vmbuspipe_hdr) +
+               sizeof(struct icmsg_hdr)];
+
+       icframe_major = negop->icframe_vercnt;
+       icframe_minor = 0;
+
+       icmsg_major = negop->icmsg_vercnt;
+       icmsg_minor = 0;
+
+       /*
+        * Select the framework version number we will
+        * support.
+        */
+
+       for (i = 0; i < negop->icframe_vercnt; i++) {
+               if ((negop->icversion_data[i].major == fw_major) &&
+                  (negop->icversion_data[i].minor == fw_minor)) {
+                       icframe_major = negop->icversion_data[i].major;
+                       icframe_minor = negop->icversion_data[i].minor;
+                       found_match = true;
+               }
+       }
+
+       if (!found_match)
+               goto fw_error;
+
+       found_match = false;
+
+       for (i = negop->icframe_vercnt;
+                (i < negop->icframe_vercnt + negop->icmsg_vercnt); i++) {
+               if ((negop->icversion_data[i].major == srv_major) &&
+                  (negop->icversion_data[i].minor == srv_minor)) {
+                       icmsg_major = negop->icversion_data[i].major;
+                       icmsg_minor = negop->icversion_data[i].minor;
+                       found_match = true;
+               }
+       }
+
+       /*
+        * Respond with the framework and service
+        * version numbers we can support.
+        */
+
+fw_error:
+       if (!found_match) {
+               negop->icframe_vercnt = 0;
+               negop->icmsg_vercnt = 0;
+       } else {
+               negop->icframe_vercnt = 1;
+               negop->icmsg_vercnt = 1;
+       }
+
+       negop->icversion_data[0].major = icframe_major;
+       negop->icversion_data[0].minor = icframe_minor;
+       negop->icversion_data[1].major = icmsg_major;
+       negop->icversion_data[1].minor = icmsg_minor;
+       return found_match;
+}
+
+EXPORT_SYMBOL_GPL(vmbus_prep_negotiate_resp);
+
+/*
+ * alloc_channel - Allocate and initialize a vmbus channel object
+ */
+static struct vmbus_channel *alloc_channel(void)
+{
+       static atomic_t chan_num = ATOMIC_INIT(0);
+       struct vmbus_channel *channel;
+
+       channel = kzalloc(sizeof(*channel), GFP_ATOMIC);
+       if (!channel)
+               return NULL;
+
+       channel->id = atomic_inc_return(&chan_num);
+       spin_lock_init(&channel->inbound_lock);
+       spin_lock_init(&channel->lock);
+
+       INIT_LIST_HEAD(&channel->sc_list);
+       INIT_LIST_HEAD(&channel->percpu_list);
+
+       return channel;
+}
+
+/*
+ * free_channel - Release the resources used by the vmbus channel object
+ */
+static void free_channel(struct vmbus_channel *channel)
+{
+       kfree(channel);
+}
+
+static void percpu_channel_enq(void *arg)
+{
+       struct vmbus_channel *channel = arg;
+       int cpu = smp_processor_id();
+
+       list_add_tail(&channel->percpu_list, &hv_context.percpu_list[cpu]);
+}
+
+static void percpu_channel_deq(void *arg)
+{
+       struct vmbus_channel *channel = arg;
+
+       list_del(&channel->percpu_list);
+}
+
+
+void hv_process_channel_removal(struct vmbus_channel *channel, u32 relid)
+{
+       struct vmbus_channel_relid_released msg;
+       unsigned long flags;
+       struct vmbus_channel *primary_channel;
+
+       memset(&msg, 0, sizeof(struct vmbus_channel_relid_released));
+       msg.child_relid = relid;
+       msg.header.msgtype = CHANNELMSG_RELID_RELEASED;
+       vmbus_post_msg(&msg, sizeof(struct vmbus_channel_relid_released));
+
+       if (channel == NULL)
+               return;
+
+       if (channel->target_cpu != get_cpu()) {
+               put_cpu();
+               smp_call_function_single(channel->target_cpu,
+                                        percpu_channel_deq, channel, true);
+       } else {
+               percpu_channel_deq(channel);
+               put_cpu();
+       }
+
+       if (channel->primary_channel == NULL) {
+               spin_lock_irqsave(&vmbus_connection.channel_lock, flags);
+               list_del(&channel->listentry);
+               spin_unlock_irqrestore(&vmbus_connection.channel_lock, flags);
+       } else {
+               primary_channel = channel->primary_channel;
+               spin_lock_irqsave(&primary_channel->lock, flags);
+               list_del(&channel->sc_list);
+               spin_unlock_irqrestore(&primary_channel->lock, flags);
+       }
+       free_channel(channel);
+}
+
+void vmbus_free_channels(void)
+{
+       struct vmbus_channel *channel;
+
+       list_for_each_entry(channel, &vmbus_connection.chn_list, listentry) {
+               vmbus_device_unregister(channel->device_obj);
+               free_channel(channel);
+       }
+}
+
+/*
+ * vmbus_process_offer - Process the offer by creating a channel/device
+ * associated with this offer
+ */
+static void vmbus_process_offer(struct vmbus_channel *newchannel)
+{
+       struct vmbus_channel *channel;
+       bool fnew = true;
+       bool enq = false;
+       unsigned long flags;
+
+       /* Make sure this is a new offer */
+       spin_lock_irqsave(&vmbus_connection.channel_lock, flags);
+
+       list_for_each_entry(channel, &vmbus_connection.chn_list, listentry) {
+               if (!uuid_le_cmp(channel->offermsg.offer.if_type,
+                       newchannel->offermsg.offer.if_type) &&
+                       !uuid_le_cmp(channel->offermsg.offer.if_instance,
+                               newchannel->offermsg.offer.if_instance)) {
+                       fnew = false;
+                       break;
+               }
+       }
+
+       if (fnew) {
+               list_add_tail(&newchannel->listentry,
+                             &vmbus_connection.chn_list);
+               enq = true;
+       }
+
+       spin_unlock_irqrestore(&vmbus_connection.channel_lock, flags);
+
+       if (enq) {
+               if (newchannel->target_cpu != get_cpu()) {
+                       put_cpu();
+                       smp_call_function_single(newchannel->target_cpu,
+                                                percpu_channel_enq,
+                                                newchannel, true);
+               } else {
+                       percpu_channel_enq(newchannel);
+                       put_cpu();
+               }
+       }
+       if (!fnew) {
+               /*
+                * Check to see if this is a sub-channel.
+                */
+               if (newchannel->offermsg.offer.sub_channel_index != 0) {
+                       /*
+                        * Process the sub-channel.
+                        */
+                       newchannel->primary_channel = channel;
+                       spin_lock_irqsave(&channel->lock, flags);
+                       list_add_tail(&newchannel->sc_list, &channel->sc_list);
+                       spin_unlock_irqrestore(&channel->lock, flags);
+
+                       if (newchannel->target_cpu != get_cpu()) {
+                               put_cpu();
+                               smp_call_function_single(newchannel->target_cpu,
+                                                        percpu_channel_enq,
+                                                        newchannel, true);
+                       } else {
+                               percpu_channel_enq(newchannel);
+                               put_cpu();
+                       }
+
+                       newchannel->state = CHANNEL_OPEN_STATE;
+                       channel->num_sc++;
+                       if (channel->sc_creation_callback != NULL)
+                               channel->sc_creation_callback(newchannel);
+
+                       return;
+               }
+
+               goto err_free_chan;
+       }
+
+       /*
+        * This state is used to indicate a successful open
+        * so that when we do close the channel normally, we
+        * can cleanup properly
+        */
+       newchannel->state = CHANNEL_OPEN_STATE;
+
+       /*
+        * Start the process of binding this offer to the driver
+        * We need to set the DeviceObject field before calling
+        * vmbus_child_dev_add()
+        */
+       newchannel->device_obj = vmbus_device_create(
+               &newchannel->offermsg.offer.if_type,
+               &newchannel->offermsg.offer.if_instance,
+               newchannel);
+       if (!newchannel->device_obj)
+               goto err_deq_chan;
+
+       /*
+        * Add the new device to the bus. This will kick off device-driver
+        * binding which eventually invokes the device driver's AddDevice()
+        * method.
+        */
+       if (vmbus_device_register(newchannel->device_obj) != 0) {
+               pr_err("unable to add child device object (relid %d)\n",
+                       newchannel->offermsg.child_relid);
+               kfree(newchannel->device_obj);
+               goto err_deq_chan;
+       }
+       return;
+
+err_deq_chan:
+       spin_lock_irqsave(&vmbus_connection.channel_lock, flags);
+       list_del(&newchannel->listentry);
+       spin_unlock_irqrestore(&vmbus_connection.channel_lock, flags);
+
+       if (newchannel->target_cpu != get_cpu()) {
+               put_cpu();
+               smp_call_function_single(newchannel->target_cpu,
+                                        percpu_channel_deq, newchannel, true);
+       } else {
+               percpu_channel_deq(newchannel);
+               put_cpu();
+       }
+
+err_free_chan:
+       free_channel(newchannel);
+}
+
+enum {
+       IDE = 0,
+       SCSI,
+       NIC,
+       MAX_PERF_CHN,
+};
+
+/*
+ * This is an array of device_ids (device types) that are performance critical.
+ * We attempt to distribute the interrupt load for these devices across
+ * all available CPUs.
+ */
+static const struct hv_vmbus_device_id hp_devs[] = {
+       /* IDE */
+       { HV_IDE_GUID, },
+       /* Storage - SCSI */
+       { HV_SCSI_GUID, },
+       /* Network */
+       { HV_NIC_GUID, },
+       /* NetworkDirect Guest RDMA */
+       { HV_ND_GUID, },
+};
+
+
+/*
+ * We use this state to statically distribute the channel interrupt load.
+ */
+static u32  next_vp;
+
+/*
+ * Starting with Win8, we can statically distribute the incoming
+ * channel interrupt load by binding a channel to VCPU. We
+ * implement here a simple round robin scheme for distributing
+ * the interrupt load.
+ * We will bind channels that are not performance critical to cpu 0 and
+ * performance critical channels (IDE, SCSI and Network) will be uniformly
+ * distributed across all available CPUs.
+ */
+static void init_vp_index(struct vmbus_channel *channel, const uuid_le *type_guid)
+{
+       u32 cur_cpu;
+       int i;
+       bool perf_chn = false;
+       u32 max_cpus = num_online_cpus();
+
+       for (i = IDE; i < MAX_PERF_CHN; i++) {
+               if (!memcmp(type_guid->b, hp_devs[i].guid,
+                                sizeof(uuid_le))) {
+                       perf_chn = true;
+                       break;
+               }
+       }
+       if ((vmbus_proto_version == VERSION_WS2008) ||
+           (vmbus_proto_version == VERSION_WIN7) || (!perf_chn)) {
+               /*
+                * Prior to win8, all channel interrupts are
+                * delivered on cpu 0.
+                * Also if the channel is not a performance critical
+                * channel, bind it to cpu 0.
+                */
+               channel->target_cpu = 0;
+               channel->target_vp = 0;
+               return;
+       }
+       cur_cpu = (++next_vp % max_cpus);
+       channel->target_cpu = cur_cpu;
+       channel->target_vp = hv_context.vp_index[cur_cpu];
+}
+
+/*
+ * vmbus_onoffer - Handler for channel offers from vmbus in parent partition.
+ *
+ */
+static void vmbus_onoffer(struct vmbus_channel_message_header *hdr)
+{
+       struct vmbus_channel_offer_channel *offer;
+       struct vmbus_channel *newchannel;
+
+       offer = (struct vmbus_channel_offer_channel *)hdr;
+
+       /* Allocate the channel object and save this offer. */
+       newchannel = alloc_channel();
+       if (!newchannel) {
+               pr_err("Unable to allocate channel object\n");
+               return;
+       }
+
+       /*
+        * By default we setup state to enable batched
+        * reading. A specific service can choose to
+        * disable this prior to opening the channel.
+        */
+       newchannel->batched_reading = true;
+
+       /*
+        * Setup state for signalling the host.
+        */
+       newchannel->sig_event = (struct hv_input_signal_event *)
+                               (ALIGN((unsigned long)
+                               &newchannel->sig_buf,
+                               HV_HYPERCALL_PARAM_ALIGN));
+
+       newchannel->sig_event->connectionid.asu32 = 0;
+       newchannel->sig_event->connectionid.u.id = VMBUS_EVENT_CONNECTION_ID;
+       newchannel->sig_event->flag_number = 0;
+       newchannel->sig_event->rsvdz = 0;
+
+       if (vmbus_proto_version != VERSION_WS2008) {
+               newchannel->is_dedicated_interrupt =
+                               (offer->is_dedicated_interrupt != 0);
+               newchannel->sig_event->connectionid.u.id =
+                               offer->connection_id;
+       }
+
+       init_vp_index(newchannel, &offer->offer.if_type);
+
+       memcpy(&newchannel->offermsg, offer,
+              sizeof(struct vmbus_channel_offer_channel));
+       newchannel->monitor_grp = (u8)offer->monitorid / 32;
+       newchannel->monitor_bit = (u8)offer->monitorid % 32;
+
+       vmbus_process_offer(newchannel);
+}
+
+/*
+ * vmbus_onoffer_rescind - Rescind offer handler.
+ *
+ * We queue a work item to process this offer synchronously
+ */
+static void vmbus_onoffer_rescind(struct vmbus_channel_message_header *hdr)
+{
+       struct vmbus_channel_rescind_offer *rescind;
+       struct vmbus_channel *channel;
+       unsigned long flags;
+       struct device *dev;
+
+       rescind = (struct vmbus_channel_rescind_offer *)hdr;
+       channel = relid2channel(rescind->child_relid);
+
+       if (channel == NULL) {
+               hv_process_channel_removal(NULL, rescind->child_relid);
+               return;
+       }
+
+       spin_lock_irqsave(&channel->lock, flags);
+       channel->rescind = true;
+       spin_unlock_irqrestore(&channel->lock, flags);
+
+       if (channel->device_obj) {
+               /*
+                * We will have to unregister this device from the
+                * driver core.
+                */
+               dev = get_device(&channel->device_obj->device);
+               if (dev) {
+                       vmbus_device_unregister(channel->device_obj);
+                       put_device(dev);
+               }
+       } else {
+               hv_process_channel_removal(channel,
+                       channel->offermsg.child_relid);
+       }
+}
+
+/*
+ * vmbus_onoffers_delivered -
+ * This is invoked when all offers have been delivered.
+ *
+ * Nothing to do here.
+ */
+static void vmbus_onoffers_delivered(
+                       struct vmbus_channel_message_header *hdr)
+{
+}
+
+/*
+ * vmbus_onopen_result - Open result handler.
+ *
+ * This is invoked when we received a response to our channel open request.
+ * Find the matching request, copy the response and signal the requesting
+ * thread.
+ */
+static void vmbus_onopen_result(struct vmbus_channel_message_header *hdr)
+{
+       struct vmbus_channel_open_result *result;
+       struct vmbus_channel_msginfo *msginfo;
+       struct vmbus_channel_message_header *requestheader;
+       struct vmbus_channel_open_channel *openmsg;
+       unsigned long flags;
+
+       result = (struct vmbus_channel_open_result *)hdr;
+
+       /*
+        * Find the open msg, copy the result and signal/unblock the wait event
+        */
+       spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
+
+       list_for_each_entry(msginfo, &vmbus_connection.chn_msg_list,
+                               msglistentry) {
+               requestheader =
+                       (struct vmbus_channel_message_header *)msginfo->msg;
+
+               if (requestheader->msgtype == CHANNELMSG_OPENCHANNEL) {
+                       openmsg =
+                       (struct vmbus_channel_open_channel *)msginfo->msg;
+                       if (openmsg->child_relid == result->child_relid &&
+                           openmsg->openid == result->openid) {
+                               memcpy(&msginfo->response.open_result,
+                                      result,
+                                      sizeof(
+                                       struct vmbus_channel_open_result));
+                               complete(&msginfo->waitevent);
+                               break;
+                       }
+               }
+       }
+       spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
+}
+
+/*
+ * vmbus_ongpadl_created - GPADL created handler.
+ *
+ * This is invoked when we received a response to our gpadl create request.
+ * Find the matching request, copy the response and signal the requesting
+ * thread.
+ */
+static void vmbus_ongpadl_created(struct vmbus_channel_message_header *hdr)
+{
+       struct vmbus_channel_gpadl_created *gpadlcreated;
+       struct vmbus_channel_msginfo *msginfo;
+       struct vmbus_channel_message_header *requestheader;
+       struct vmbus_channel_gpadl_header *gpadlheader;
+       unsigned long flags;
+
+       gpadlcreated = (struct vmbus_channel_gpadl_created *)hdr;
+
+       /*
+        * Find the establish msg, copy the result and signal/unblock the wait
+        * event
+        */
+       spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
+
+       list_for_each_entry(msginfo, &vmbus_connection.chn_msg_list,
+                               msglistentry) {
+               requestheader =
+                       (struct vmbus_channel_message_header *)msginfo->msg;
+
+               if (requestheader->msgtype == CHANNELMSG_GPADL_HEADER) {
+                       gpadlheader =
+                       (struct vmbus_channel_gpadl_header *)requestheader;
+
+                       if ((gpadlcreated->child_relid ==
+                            gpadlheader->child_relid) &&
+                           (gpadlcreated->gpadl == gpadlheader->gpadl)) {
+                               memcpy(&msginfo->response.gpadl_created,
+                                      gpadlcreated,
+                                      sizeof(
+                                       struct vmbus_channel_gpadl_created));
+                               complete(&msginfo->waitevent);
+                               break;
+                       }
+               }
+       }
+       spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
+}
+
+/*
+ * vmbus_ongpadl_torndown - GPADL torndown handler.
+ *
+ * This is invoked when we received a response to our gpadl teardown request.
+ * Find the matching request, copy the response and signal the requesting
+ * thread.
+ */
+static void vmbus_ongpadl_torndown(
+                       struct vmbus_channel_message_header *hdr)
+{
+       struct vmbus_channel_gpadl_torndown *gpadl_torndown;
+       struct vmbus_channel_msginfo *msginfo;
+       struct vmbus_channel_message_header *requestheader;
+       struct vmbus_channel_gpadl_teardown *gpadl_teardown;
+       unsigned long flags;
+
+       gpadl_torndown = (struct vmbus_channel_gpadl_torndown *)hdr;
+
+       /*
+        * Find the open msg, copy the result and signal/unblock the wait event
+        */
+       spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
+
+       list_for_each_entry(msginfo, &vmbus_connection.chn_msg_list,
+                               msglistentry) {
+               requestheader =
+                       (struct vmbus_channel_message_header *)msginfo->msg;
+
+               if (requestheader->msgtype == CHANNELMSG_GPADL_TEARDOWN) {
+                       gpadl_teardown =
+                       (struct vmbus_channel_gpadl_teardown *)requestheader;
+
+                       if (gpadl_torndown->gpadl == gpadl_teardown->gpadl) {
+                               memcpy(&msginfo->response.gpadl_torndown,
+                                      gpadl_torndown,
+                                      sizeof(
+                                       struct vmbus_channel_gpadl_torndown));
+                               complete(&msginfo->waitevent);
+                               break;
+                       }
+               }
+       }
+       spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
+}
+
+/*
+ * vmbus_onversion_response - Version response handler
+ *
+ * This is invoked when we received a response to our initiate contact request.
+ * Find the matching request, copy the response and signal the requesting
+ * thread.
+ */
+static void vmbus_onversion_response(
+               struct vmbus_channel_message_header *hdr)
+{
+       struct vmbus_channel_msginfo *msginfo;
+       struct vmbus_channel_message_header *requestheader;
+       struct vmbus_channel_version_response *version_response;
+       unsigned long flags;
+
+       version_response = (struct vmbus_channel_version_response *)hdr;
+       spin_lock_irqsave(&vmbus_connection.channelmsg_lock, flags);
+
+       list_for_each_entry(msginfo, &vmbus_connection.chn_msg_list,
+                               msglistentry) {
+               requestheader =
+                       (struct vmbus_channel_message_header *)msginfo->msg;
+
+               if (requestheader->msgtype ==
+                   CHANNELMSG_INITIATE_CONTACT) {
+                       memcpy(&msginfo->response.version_response,
+                             version_response,
+                             sizeof(struct vmbus_channel_version_response));
+                       complete(&msginfo->waitevent);
+               }
+       }
+       spin_unlock_irqrestore(&vmbus_connection.channelmsg_lock, flags);
+}
+
+/* Channel message dispatch table */
+struct vmbus_channel_message_table_entry
+       channel_message_table[CHANNELMSG_COUNT] = {
+       {CHANNELMSG_INVALID,                    0, NULL},
+       {CHANNELMSG_OFFERCHANNEL,               0, vmbus_onoffer},
+       {CHANNELMSG_RESCIND_CHANNELOFFER,       0, vmbus_onoffer_rescind},
+       {CHANNELMSG_REQUESTOFFERS,              0, NULL},
+       {CHANNELMSG_ALLOFFERS_DELIVERED,        1, vmbus_onoffers_delivered},
+       {CHANNELMSG_OPENCHANNEL,                0, NULL},
+       {CHANNELMSG_OPENCHANNEL_RESULT,         1, vmbus_onopen_result},
+       {CHANNELMSG_CLOSECHANNEL,               0, NULL},
+       {CHANNELMSG_GPADL_HEADER,               0, NULL},
+       {CHANNELMSG_GPADL_BODY,                 0, NULL},
+       {CHANNELMSG_GPADL_CREATED,              1, vmbus_ongpadl_created},
+       {CHANNELMSG_GPADL_TEARDOWN,             0, NULL},
+       {CHANNELMSG_GPADL_TORNDOWN,             1, vmbus_ongpadl_torndown},
+       {CHANNELMSG_RELID_RELEASED,             0, NULL},
+       {CHANNELMSG_INITIATE_CONTACT,           0, NULL},
+       {CHANNELMSG_VERSION_RESPONSE,           1, vmbus_onversion_response},
+       {CHANNELMSG_UNLOAD,                     0, NULL},
+};
+
+/*
+ * vmbus_onmessage - Handler for channel protocol messages.
+ *
+ * This is invoked in the vmbus worker thread context.
+ */
+void vmbus_onmessage(void *context)
+{
+       struct hv_message *msg = context;
+       struct vmbus_channel_message_header *hdr;
+       int size;
+
+       hdr = (struct vmbus_channel_message_header *)msg->u.payload;
+       size = msg->header.payload_size;
+
+       if (hdr->msgtype >= CHANNELMSG_COUNT) {
+               pr_err("Received invalid channel message type %d size %d\n",
+                          hdr->msgtype, size);
+               print_hex_dump_bytes("", DUMP_PREFIX_NONE,
+                                    (unsigned char *)msg->u.payload, size);
+               return;
+       }
+
+       if (channel_message_table[hdr->msgtype].message_handler)
+               channel_message_table[hdr->msgtype].message_handler(hdr);
+       else
+               pr_err("Unhandled channel message type %d\n", hdr->msgtype);
+}
+
+/*
+ * vmbus_request_offers - Send a request to get all our pending offers.
+ */
+int vmbus_request_offers(void)
+{
+       struct vmbus_channel_message_header *msg;
+       struct vmbus_channel_msginfo *msginfo;
+       int ret;
+
+       msginfo = kmalloc(sizeof(*msginfo) +
+                         sizeof(struct vmbus_channel_message_header),
+                         GFP_KERNEL);
+       if (!msginfo)
+               return -ENOMEM;
+
+       msg = (struct vmbus_channel_message_header *)msginfo->msg;
+
+       msg->msgtype = CHANNELMSG_REQUESTOFFERS;
+
+
+       ret = vmbus_post_msg(msg,
+                              sizeof(struct vmbus_channel_message_header));
+       if (ret != 0) {
+               pr_err("Unable to request offers - %d\n", ret);
+
+               goto cleanup;
+       }
+
+cleanup:
+       kfree(msginfo);
+
+       return ret;
+}
+
+/*
+ * Retrieve the (sub) channel on which to send an outgoing request.
+ * When a primary channel has multiple sub-channels, we try to
+ * distribute the load equally amongst all available channels.
+ */
+struct vmbus_channel *vmbus_get_outgoing_channel(struct vmbus_channel *primary)
+{
+       struct list_head *cur, *tmp;
+       int cur_cpu;
+       struct vmbus_channel *cur_channel;
+       struct vmbus_channel *outgoing_channel = primary;
+       int next_channel;
+       int i = 1;
+
+       if (list_empty(&primary->sc_list))
+               return outgoing_channel;
+
+       next_channel = primary->next_oc++;
+
+       if (next_channel > (primary->num_sc)) {
+               primary->next_oc = 0;
+               return outgoing_channel;
+       }
+
+       cur_cpu = hv_context.vp_index[get_cpu()];
+       put_cpu();
+       list_for_each_safe(cur, tmp, &primary->sc_list) {
+               cur_channel = list_entry(cur, struct vmbus_channel, sc_list);
+               if (cur_channel->state != CHANNEL_OPENED_STATE)
+                       continue;
+
+               if (cur_channel->target_vp == cur_cpu)
+                       return cur_channel;
+
+               if (i == next_channel)
+                       return cur_channel;
+
+               i++;
+       }
+
+       return outgoing_channel;
+}
+EXPORT_SYMBOL_GPL(vmbus_get_outgoing_channel);
+
+static void invoke_sc_cb(struct vmbus_channel *primary_channel)
+{
+       struct list_head *cur, *tmp;
+       struct vmbus_channel *cur_channel;
+
+       if (primary_channel->sc_creation_callback == NULL)
+               return;
+
+       list_for_each_safe(cur, tmp, &primary_channel->sc_list) {
+               cur_channel = list_entry(cur, struct vmbus_channel, sc_list);
+
+               primary_channel->sc_creation_callback(cur_channel);
+       }
+}
+
+void vmbus_set_sc_create_callback(struct vmbus_channel *primary_channel,
+                               void (*sc_cr_cb)(struct vmbus_channel *new_sc))
+{
+       primary_channel->sc_creation_callback = sc_cr_cb;
+}
+EXPORT_SYMBOL_GPL(vmbus_set_sc_create_callback);
+
+bool vmbus_are_subchannels_present(struct vmbus_channel *primary)
+{
+       bool ret;
+
+       ret = !list_empty(&primary->sc_list);
+
+       if (ret) {
+               /*
+                * Invoke the callback on sub-channel creation.
+                * This will present a uniform interface to the
+                * clients.
+                */
+               invoke_sc_cb(primary);
+       }
+
+       return ret;
+}
+EXPORT_SYMBOL_GPL(vmbus_are_subchannels_present);