[PATCH 2/2] Drivers: hv: vmbus: Fix a rescind issue

[kys@exchange.microsoft.com]

From: K. Y. Srinivasan <kys@microsoft.com>

The current rescind processing code will not correctly handle
the case where the host immediately rescinds a channel that has
been offerred. In this case, we could be blocked in the open call and
since the channel is rescinded, the host will not respond and we could
be blocked forever in the vmbus open call.i Fix this problem.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Cc: stable@vger.kernel.org
---
 drivers/hv/channel.c      |   10 ++++++++--
 drivers/hv/channel_mgmt.c |    7 ++++---
 include/linux/hyperv.h    |    1 +
 3 files changed, 13 insertions(+), 5 deletions(-)

diff --git a/drivers/hv/channel.c b/drivers/hv/channel.c
index 19f0cf3..ba0a092 100644
--- a/drivers/hv/channel.c
+++ b/drivers/hv/channel.c
@@ -659,22 +659,28 @@ void vmbus_close(struct vmbus_channel *channel)
 		 */
 		return;
 	}
-	mutex_lock(&vmbus_connection.channel_mutex);
 	/*
 	 * Close all the sub-channels first and then close the
 	 * primary channel.
 	 */
 	list_for_each_safe(cur, tmp, &channel->sc_list) {
 		cur_channel = list_entry(cur, struct vmbus_channel, sc_list);
-		vmbus_close_internal(cur_channel);
 		if (cur_channel->rescind) {
+			wait_for_completion(&cur_channel->rescind_event);
+			mutex_lock(&vmbus_connection.channel_mutex);
+			vmbus_close_internal(cur_channel);
 			hv_process_channel_removal(
 					   cur_channel->offermsg.child_relid);
+		} else {
+			mutex_lock(&vmbus_connection.channel_mutex);
+			vmbus_close_internal(cur_channel);
 		}
+		mutex_unlock(&vmbus_connection.channel_mutex);
 	}
 	/*
 	 * Now close the primary.
 	 */
+	mutex_lock(&vmbus_connection.channel_mutex);
 	vmbus_close_internal(channel);
 	mutex_unlock(&vmbus_connection.channel_mutex);
 }
diff --git a/drivers/hv/channel_mgmt.c b/drivers/hv/channel_mgmt.c
index ec5454f..c21020b 100644
--- a/drivers/hv/channel_mgmt.c
+++ b/drivers/hv/channel_mgmt.c
@@ -333,6 +333,7 @@ bool vmbus_prep_negotiate_resp(struct icmsg_hdr *icmsghdrp,
 		return NULL;
 
 	spin_lock_init(&channel->lock);
+	init_completion(&channel->rescind_event);
 
 	INIT_LIST_HEAD(&channel->sc_list);
 	INIT_LIST_HEAD(&channel->percpu_list);
@@ -898,6 +899,7 @@ static void vmbus_onoffer_rescind(struct vmbus_channel_message_header *hdr)
 	/*
 	 * Now wait for offer handling to complete.
 	 */
+	vmbus_rescind_cleanup(channel);
 	while (READ_ONCE(channel->probe_done) == false) {
 		/*
 		 * We wait here until any channel offer is currently
@@ -913,7 +915,6 @@ static void vmbus_onoffer_rescind(struct vmbus_channel_message_header *hdr)
 	if (channel->device_obj) {
 		if (channel->chn_rescind_callback) {
 			channel->chn_rescind_callback(channel);
-			vmbus_rescind_cleanup(channel);
 			return;
 		}
 		/*
@@ -922,7 +923,6 @@ static void vmbus_onoffer_rescind(struct vmbus_channel_message_header *hdr)
 		 */
 		dev = get_device(&channel->device_obj->device);
 		if (dev) {
-			vmbus_rescind_cleanup(channel);
 			vmbus_device_unregister(channel->device_obj);
 			put_device(dev);
 		}
@@ -936,13 +936,14 @@ static void vmbus_onoffer_rescind(struct vmbus_channel_message_header *hdr)
 		 * 2. Then close the primary channel.
 		 */
 		mutex_lock(&vmbus_connection.channel_mutex);
-		vmbus_rescind_cleanup(channel);
 		if (channel->state == CHANNEL_OPEN_STATE) {
 			/*
 			 * The channel is currently not open;
 			 * it is safe for us to cleanup the channel.
 			 */
 			hv_process_channel_removal(rescind->child_relid);
+		} else {
+			complete(&channel->rescind_event);
 		}
 		mutex_unlock(&vmbus_connection.channel_mutex);
 	}
diff --git a/include/linux/hyperv.h b/include/linux/hyperv.h
index f3e97c5..6c93366 100644
--- a/include/linux/hyperv.h
+++ b/include/linux/hyperv.h
@@ -708,6 +708,7 @@ struct vmbus_channel {
 	u8 monitor_bit;
 
 	bool rescind; /* got rescind msg */
+	struct completion rescind_event;
 
 	u32 ringbuffer_gpadlhandle;
 
-- 
1.7.1