Conference helix::vaxeln

770.0. "crash in 6CCKER.EXE" by TKTVFS::KUSAKARI (CSC Tokyo) Thu May 29 1997 06:37

	Hello,

	I have a customer who use VAXeln V4.2.
	Although the version is very old,I'll be thankfull if someome
	advise me.

        The customer has often experienced system crash.
	I've investigated the crash log and found the crash module 
	in VAXeln kernel at last. As the customer use VAX6200 , so 6CCKER.EXE 
	is used in VAXeln kernel. I examined crash PC by using the console CCL
	examine command. The crash PC indicates the instruction of the module 
	WAIT.MAR as below.

	(ker$wait_any routine of WAIT.MAR)

110$:   movl    wcb$a_list(r8),wcb$a_next(r8) ; set address of next wcb
        bneq    120$                    ; if neq, a free wcb is available
        bsbw    ker$expand_process_wait ; no wcb, allocate additional wcb's
        blbs    r0,110$                 ; if lbs, continue - new wcb's available
        brb     150$                    ; unable to allocate additional wcb's
120$:   movl    wcb$a_list(r8),r8       ; get address of next wcb
===>    movb    r11,wcb$b_wait(r8)      ; set type of wait
        clrb    wcb$b_satisfied(r8)     ; set wait satisfied false
        movl    r1,wcb$a_object(r8)     ; set address of object
        movb    obj$b_type(r1), -       ; copy object type field
                wcb$b_object_type(r8)   ; into wcb


	The crash instruction is always "movb    r11,wcb$b_wait(r8) ".
	R8 has also always 41422042 in the crash console log. I think 
	"wcb$a_list(r8)" has bad value of 120$'s instruction. 
	"wcb$a_list(r8)" may be corrupted.

	Is this known problem? Does higher version fix this problem?

	Regards
	Yasuo

	
	

    There are no bugs in wait_any. It might be more helpful to see the code
    which is making the call and the actual text of the exception they're
    getting.

	I told the customer to investigate the code using wait_any.
	The customer ask me that if we have any VAX/ELN tool in order 
	to isolate the problem. It means the customer would like to know 
	which does the cause exist in the application or outside of the 
	application. Is there any good tool for such memory corruption?

	Thank you 
	Yasuo

	
	

>	The crash instruction is always "movb    r11,wcb$b_wait(r8) ".
>	R8 has also always 41422042 in the crash console log. I think 
>	"wcb$a_list(r8)" has bad value of 120$'s instruction. 
>	"wcb$a_list(r8)" may be corrupted.
    
    The 41422042 looks like data. In ASCII, that's "AB B". It could very
    well be a memory corruption problem. They may be running out of memory
    or reaching some virtual limit.
    
>	The customer ask me that if we have any VAX/ELN tool in order
> 	to isolate the problem.
    
    Use EDISPLAY and EDEBUG to monitor memory. I would suggest they watch
    their P0 usage especially. It might be a good experiement to increase
    the P0 limit in EBUILD and see if the same thing happens. 
    
    Also have them check their pool blocks. They may be running out.
    
    If they're using kernel mode, they may also need to adjust their kernel
    stack size.