Re: [PATCH v2] Drivers: hv: vmbus: Initialize unload_event statically

[Wei Liu]

On Tue, Apr 20, 2021 at 04:50:56AM +, Michael Kelley wrote:
> From: Andrea Parri (Microsoft)  Sent: Monday, April 
> 19, 2021 6:44 PM
> > 
> > If a malicious or compromised Hyper-V sends a spurious message of type
> > CHANNELMSG_UNLOAD_RESPONSE, the function vmbus_unload_response() will
> > call complete() on an uninitialized event, and cause an oops.
> > 
> > Reported-by: Michael Kelley 
> > Signed-off-by: Andrea Parri (Microsoft) 
> > ---
> > Changes since v1[1]:
> >   - add inline comment in vmbus_unload_response()
> > 
> > [1] 
> > https://lore.kernel.org/linux-hyperv/20210416143932.16512-1-parri.and...@gmail.com/
> > 
> >  drivers/hv/channel_mgmt.c | 7 ++-
> >  drivers/hv/connection.c   | 2 ++
> >  2 files changed, 8 insertions(+), 1 deletion(-)
> > 
> 
> Reviewed-by: Michael Kelley 

Applied to hyperv-next. Thanks.

RE: [PATCH v2] Drivers: hv: vmbus: Initialize unload_event statically

[Michael Kelley]

From: Andrea Parri (Microsoft)  Sent: Monday, April 19, 
2021 6:44 PM
> 
> If a malicious or compromised Hyper-V sends a spurious message of type
> CHANNELMSG_UNLOAD_RESPONSE, the function vmbus_unload_response() will
> call complete() on an uninitialized event, and cause an oops.
> 
> Reported-by: Michael Kelley 
> Signed-off-by: Andrea Parri (Microsoft) 
> ---
> Changes since v1[1]:
>   - add inline comment in vmbus_unload_response()
> 
> [1] 
> https://lore.kernel.org/linux-hyperv/20210416143932.16512-1-parri.and...@gmail.com/
> 
>  drivers/hv/channel_mgmt.c | 7 ++-
>  drivers/hv/connection.c   | 2 ++
>  2 files changed, 8 insertions(+), 1 deletion(-)
> 

Reviewed-by: Michael Kelley 

> diff --git a/drivers/hv/channel_mgmt.c b/drivers/hv/channel_mgmt.c
> index 4c9e45d1f462c..335a10ee03a5e 100644
> --- a/drivers/hv/channel_mgmt.c
> +++ b/drivers/hv/channel_mgmt.c
> @@ -826,6 +826,11 @@ static void vmbus_unload_response(struct
> vmbus_channel_message_header *hdr)
>   /*
>* This is a global event; just wakeup the waiting thread.
>* Once we successfully unload, we can cleanup the monitor state.
> +  *
> +  * NB.  A malicious or compromised Hyper-V could send a spurious
> +  * message of type CHANNELMSG_UNLOAD_RESPONSE, and trigger a call
> +  * of the complete() below.  Make sure that unload_event has been
> +  * initialized by the time this complete() is executed.
>*/
>   complete(_connection.unload_event);
>  }
> @@ -841,7 +846,7 @@ void vmbus_initiate_unload(bool crash)
>   if (vmbus_proto_version < VERSION_WIN8_1)
>   return;
> 
> - init_completion(_connection.unload_event);
> + reinit_completion(_connection.unload_event);
>   memset(, 0, sizeof(struct vmbus_channel_message_header));
>   hdr.msgtype = CHANNELMSG_UNLOAD;
>   vmbus_post_msg(, sizeof(struct vmbus_channel_message_header),
> diff --git a/drivers/hv/connection.c b/drivers/hv/connection.c
> index dc19d5ae4373c..311cd005b3be6 100644
> --- a/drivers/hv/connection.c
> +++ b/drivers/hv/connection.c
> @@ -26,6 +26,8 @@
> 
>  struct vmbus_connection vmbus_connection = {
>   .conn_state = DISCONNECTED,
> + .unload_event   = COMPLETION_INITIALIZER(
> +   vmbus_connection.unload_event),
>   .next_gpadl_handle  = ATOMIC_INIT(0xE1E10),
> 
>   .ready_for_suspend_event = COMPLETION_INITIALIZER(
> --
> 2.25.1

[PATCH v2] Drivers: hv: vmbus: Initialize unload_event statically

[Andrea Parri (Microsoft)]

If a malicious or compromised Hyper-V sends a spurious message of type
CHANNELMSG_UNLOAD_RESPONSE, the function vmbus_unload_response() will
call complete() on an uninitialized event, and cause an oops.

Reported-by: Michael Kelley 
Signed-off-by: Andrea Parri (Microsoft) 
---
Changes since v1[1]:
  - add inline comment in vmbus_unload_response()

[1] https://lkml.kernel.org/r/20210416143932.16512-1-parri.and...@gmail.com

 drivers/hv/channel_mgmt.c | 7 ++-
 drivers/hv/connection.c   | 2 ++
 2 files changed, 8 insertions(+), 1 deletion(-)

diff --git a/drivers/hv/channel_mgmt.c b/drivers/hv/channel_mgmt.c
index 4c9e45d1f462c..335a10ee03a5e 100644
--- a/drivers/hv/channel_mgmt.c
+++ b/drivers/hv/channel_mgmt.c
@@ -826,6 +826,11 @@ static void vmbus_unload_response(struct 
vmbus_channel_message_header *hdr)
/*
 * This is a global event; just wakeup the waiting thread.
 * Once we successfully unload, we can cleanup the monitor state.
+*
+* NB.  A malicious or compromised Hyper-V could send a spurious
+* message of type CHANNELMSG_UNLOAD_RESPONSE, and trigger a call
+* of the complete() below.  Make sure that unload_event has been
+* initialized by the time this complete() is executed.
 */
complete(_connection.unload_event);
 }
@@ -841,7 +846,7 @@ void vmbus_initiate_unload(bool crash)
if (vmbus_proto_version < VERSION_WIN8_1)
return;
 
-   init_completion(_connection.unload_event);
+   reinit_completion(_connection.unload_event);
memset(, 0, sizeof(struct vmbus_channel_message_header));
hdr.msgtype = CHANNELMSG_UNLOAD;
vmbus_post_msg(, sizeof(struct vmbus_channel_message_header),
diff --git a/drivers/hv/connection.c b/drivers/hv/connection.c
index dc19d5ae4373c..311cd005b3be6 100644
--- a/drivers/hv/connection.c
+++ b/drivers/hv/connection.c
@@ -26,6 +26,8 @@
 
 struct vmbus_connection vmbus_connection = {
.conn_state = DISCONNECTED,
+   .unload_event   = COMPLETION_INITIALIZER(
+ vmbus_connection.unload_event),
.next_gpadl_handle  = ATOMIC_INIT(0xE1E10),
 
.ready_for_suspend_event = COMPLETION_INITIALIZER(
-- 
2.25.1