Conference tuxedo::dce-products

2267.0. "rpc__krb_cn_cred_refresh called by rpc runtime?" by HGOSPS::MICKWIDLAM () Fri May 30 1997 03:57

    Hi,
    
    I have a customer with an rpc application intermittenly hang. They use
    DCE V1.3B. From the stack trace she found that her rpc call enters the
    client stub and the stub called rpc__cn_call_start and this called
    rpc__krb_cn_cred_refresh. Then it hung in mutex wait. We have a
    simplied stack dump from the customer:
    
    #0 msg_receive_trap(...)
    #1 msg_receive(...)
    #2 cma__vp_sleep(...)
    #3 cma__dispatch(...)
    #4 cma__int_mutex_block(...)
    #5 pthread_mutex_lock(...)
    #6 rpc__krb_cn_cred_refresh(...)
    #7 rpc__cn_call_start(...)
    #8 rpc_call_start(...)
    #9 op1_csr(...) ... bi_cstub.c <---- entered client stub from below
    #10 ((BIRpcClient*)...)->insert_terminal(t=const 36) BIRpcClient.C
    #11 RecBiServer::insert_terminal(...) RecBiServer.C
    #12 ((CThreadRec*)...)->Worker(...) CThreadRec.C
    #13 CThreadPool::thread_base_proc(...) CThreadPool.C
    #14 cma__thread_base(...)
    
    The application itself is not trying to refresh the login context as
    this is the job of another thread. So from customer's view, she don't
    expect to see the rpc__krb_cn_cred_refresh.
    
    Anyone seen similar kind of hang before?
    
    Thanks and regards,
    Mickwid.