STOP 0xEFFFFFFF Exception in OLE Servers with Multiple Clients

Article translations Article translations
Article ID: 195469 - View products that this article applies to.
This article has been archived. It is offered "as is" and will no longer be updated.
Expand all | Collapse all


When multiple clients are concurrently accessing an out of process COM server, client calls may fail unexpectedly. Debugging the server during such failures shows that the server is throwing exception 0xEFFFFFFF.

Recognizing this problem can be difficult because both the client and the server applications may continue running. In particular, this exception does not directly result in causing of the server or the client to stop responding (hang) or a crash.

Rather, the server simply returns a failure to the client. Whether the client continues functioning correctly after such a failure will depend entirely on how the client software is written. Well behaved client applications may silently retry the operation or log an error without being otherwise adversely affected. Clients that do not take care to check the return status of their function calls may experience access violations or other fatal errors.

To confirm whether this problem is the cause of a specific failure, it is necessary to debug the server application, set the debugger to break on this exception, and compare the resultant stack to the one below:
   RPCRT4! RpcRaiseException@4 + 49 bytes
   OLE32! ThreadInvoke(struct _RPC_MESSAGE *) + 524 bytes
   RPCRT4! DispatchToStubInC@12 + 52 bytes
   RPCRT4! RPC_INTERFACE::DispatchToStubWorker(struct _RPC_MESSAGE *,
                unsigned int,long *) + 292 bytes
   RPCRT4! RPC_INTERFACE::DispatchToStub(struct _RPC_MESSAGE *,
                unsigned int,long *) + 93 bytes
   RPCRT4! RPC_INTERFACE::DispatchToStubWithObject(struct _RPC_MESSAGE *,
                class RPC_UUID *,unsigned int,long *) + 176 bytes
   RPCRT4! WMSG_SASSOCIATION::DealWithRequestMessage(union _WMSG_MESSAGE *,
                union _WMSG_MESSAGE *,struct _RPC_MESSAGE *,
                class WMSG_SBINDING * *,unsigned int,int,int,int) + 826
   RPCRT4! WMSG_ADDRESS::HandleRequest(union _WMSG_MESSAGE *,
                class WMSG_SASSOCIATION *,struct _RPC_MESSAGE *,
                class WMSG_ENDPOINT *,int,int) + 146 bytes
                class WMSG_ENDPOINT * *,struct HWND__ * *,
                class WMSG_ASSOCIATION *,union _WMSG_MESSAGE * *) + 734
   RPCRT4! WMSG_ADDRESS::ReceiveLotsaCalls(void) + 837 bytes
   RPCRT4! RecvLotsaCallsWrapper(class WMSG_ADDRESS *) + 9 bytes
   RPCRT4! BaseCachedThreadRoutine(class CACHED_THREAD *) + 156 bytes
   RPCRT4! ThreadStartRoutine(class THREAD *) + 23 bytes
   KERNEL32! 77f04f44()


The OLE functions handling typelib marshaling are not completely multithread safe, resulting in a narrow window of opportunity for the exception to take place if two or more clients are making concurrent calls to the server.


To resolve this problem, obtain the latest service pack for Windows NT 4.0 or the individual software update. For information on obtaining the latest service pack, please go to:
For information on obtaining the individual software update, contact Microsoft Product Support Services. For a complete list of Microsoft Product Support Services phone numbers and information on support costs, please go to the following address on the World Wide Web:


Microsoft has confirmed this to be a problem in Windows NT version 4.0. This problem was first corrected in Windows NT 4.0 Service Pack 5.


Article ID: 195469 - Last Review: January 9, 2015 - Revision: 3.0
Applies to
  • Microsoft Windows NT Workstation 4.0 Developer Edition
  • Microsoft Windows NT Server 4.0 Standard Edition
kbnosurvey kbarchive kbhotfixserver kbqfe kbbug kbfix KB195469

Contact us for more help

Contact us for more help
Connect with Answer Desk for expert help.
Get more support from