Conference 49.910::kav30

    Hi Jean-Pierre,
    
    The KAV30$_ASBQUOTA exception can be raised in 1 of 2 places.  It can
    be raised during system initialization if the KAV30 kernel is unable to
    allocate the AST control block queue structure, or it can be returned
    by the KAV$DEF_AST routine at run-time if there are no free slots in
    the AST control block queue structure.   I suspect that your customer
    is seing the first flavour of this exception, if this is the case then
    increasing system region size by at least 128 pages should resolve this
    problem.
    
    Graham
          

    Hi Graham,
    
    First I thank you for your very quick answer.
    I gave this information to the customer. So he increased the system
    region size to 1500 pages as you suggest in your answer. Unfortunatly 
    that does not solve the problem. The customer runs his application which 
    run a while then he gets the same error. Assuming your answer it could be 
    caused by the second flavor suggested in your answer.
    So 
    1)  what must be at least the system region size ?
    2) what can we done to solve this problem if it's the second flavor
    cause ?
    
    Thanks for your help.
    
    Best regards,
     Jean-Pierre

    Hi Jean-Pierre,
    
    Without knowing more about the customer's application this is going to
    be difficult to pin-point what is going on.  Try setting system region
    size to 2000.
    
    Graham

    We have increased the system region over 2000 pages but the error still
    occurs. It seems that the problem occurs when many interrupts are used
    in the customer application . I mean that it seems that the problem is
    a cause of the number of interrups used.
    When did the official release delivered ?
    
    jean-pierre
    

    In fact the customer gets an exception showing the error number 5C04 .
    Is it the right number for the error message mentioned by the customer
    ?
    
    Jean-Pierre

    Here is the .dat file (the same used which worked with ELN-V3.2 and 
    KAV30-V1.0):
    
     characteristic /nofile /net=eza /noserver /emulator=none /debug=none
    /loader=5 /io_region=1600/interrupt_stack=6/target=24
    program prg_name /init /kernel=16 /mode=kernel
    device EZA /vec=%x130 /net_def
    device smp0 /vector=%x900 /noauto
    device rs0  /vector=%x940 /noauto
    device rs1  /vector=%x940 /noauto
    device rs2  /vector=%x940 /noauto
    device rs3  /vector=%x940 /noauto
    device rs4  /vector=%x940 /noauto
    device rs5  /vector=%x940 /noauto
    device rs6  /vector=%x940 /noauto
    device smp1 /vector=%x900 /noauto
    device smp2 /vector=%x900 /noauto
    device smp3 /vector=%x900 /noauto
    device smp4 /vector=%x900 /noauto
    device smp5 /vector=%x900 /noauto
    device smp6 /vector=%x900 /noauto
    terminal CONSOLE /scope
    
    
    Jean-Pierre

    I spoke with the customer this morning. He does not use the KAV$DEF_AST
    system service as he does not use AST's . We made some tries with an
    increased system region size (3000). It's the same result. The problem
    becomes more and more crucial for the customer. 
    May I know the link (which system resource is used) between the ASB
    quota and some ELN system resources? 
    Are there some KAV30 sources I could read ? Where ?
    
    Thanks for help.
    
    Jean-Pierre
    

    Hi Jean-Pierre,
    
    Does the customer's application use any of the following KAV30 system
    services?
    
	    KAV$IN_MAP
	    KAV$NOTIFY_FIFO
	    KAV$TIMERS
	    KAV$RTC
    
    I suspect that the behaviour that your customer is seeing is not a
    resource problem that can be corrected by increasing system
    parameters.  In order to further analyze the behaviour that your
    customer is observing, would it be possible for you to obtain a sample
    program from the customer that demonstrates this behaviour.
    
    Graham
    

    Hi Graham,
    
    Yes, you're right, the customer uses the service KAV$TIMERS and it
    seems to me that it could be the cause of the problem. The AST routine
    is correctly called when an AST occurs but it seems that the queue is
    never 'cleared' from the last AST which has been treated.
    As I've no KAV30 to make tests I asked the customer to try the example
    described in the documentation. He's actually doing that.
    
    MANY thanks for your help.
    
    Jean-Pierre

    Hi Jean-Pierre,
    
    Can you post the call to KAV$TIMERS that your customer's application
    makes as a reply to this note.
    
    Thanks in advance.
    
    Graham

    Hi Graham,
    
    Here are the informations you asked for.
    
    The customer's application has two timers.
    In his application the customer first call a subroutine named
    CLOCK_START which contains the following call:
    
      KAV$TIMERS (NULL,
    		  KAV$M_LOAD_CMR_CNT + KAV$M_START_TMR + KAV$M_REPEAT_TMR ,
    		  KAV$K_CTMR1,
    		  &t1_count,
    		  clock,NULL)
    
    Then after returned from this subroutine (called only once) he does the
    following:
    
      KAV$TIMERS (NULL,
    		  KAV$M_LOAD_TMR_CNT + KAV$M_START_TMR,
    		  KAV$K_CTMR2,
    		  &t2_count,
    		  T2_DONE,		<<<< should never be entered
    		  NULL)
    
    Then the customer has a loop where he does the following (at the end of
    this loop) :
    
      KAV$TIMERS (NULL,
    		  KAV$M_STOP_TMR + KAV$M_READ_TMR_CNT,
    		  KAV$K_CTMR2,
    		  &t2_val,
    		  NULL,NULL)
    
     then..
    
      KAV$TIMERS (NULL,
    		  KAV$M_RESET_TMR,
    		  KAV$K_CTMR2,
    		  &t2_count,
    		 NULL, NULL)
    
    
    
     Jean-Pierre
    	

    One question more...
    
    The ast routine CLOCK is entered each time the timer 1 expires but this
    ast routine never does a KAV$M_RESET_TMR . Is it important ?
    The customer explained also that the target 'speaks' with a host
    system througth the network. It happens that sometimes there is nomore
    dialog between these two nodes but we don't know why for now.. When this 
    occurs they halt the host and then they get the error mentioned in .0 
    after a while.
    
    Jean-Pierre