Conference noted::hackers_v1

    Try SHOW PROC/PAGE in SDA (analyze/system).
    
    	Jon

    I need to be able to get the information from within a user mode
    program.
    

    You can't, since between when you get the answer and when you read
    it the page may have come or gone from the working-set.
    
    I know this is a hackers notes file, but I sure hope you aren't planning
    to ship something with hacks like this in it to customers...
    
    /Bevin

    No, this isn't going to customers... I'm working on test software,
    and I'm trying to control which pages are part of the user's working
    set and which aren't at various stages of a test case. Basically,
    I can control which pages are definitely in the test case at the
    start of the test case ($LKWSET) and by purging the remaining pages
    from the working set and by setting the PFC I can control (I think)
    which pages get brought in, but I was hoping there might be some 
    way to find out which pages used by the test case are in the 
    working set at the end of the test case.

    	There used to be (and may still be - I haven't looked at that
    part of the fiche for a long time) a system service for getting
    page table information. It was called $GETPTI (reasonably enough).
    
    	It was used by the linker to see which pages had the "modified"
    bit set, so it knew where it could create demand zero sections.
    You might try to find if it is still around and does what you want.
    
    	Though as pointed out, the information you want from the page
    tables may change behind your back before you can use it.

    I don't think you're going to be able to do it from user mode. If you
    can go to kernel mode, it's no problem -- I can give you some BLISS
    that will return the PTE of a specified address. But you have to be at
    IPL$_SYNC to do it (and have a spinlock in V5). 
    
    $GETPTI will not give you the information you want. What $GETPTI does
    is search your address space for N contiguous pages that have certain
    attributes. In V4 and before, it will only search for demand-zero
    pages. In V5, it will also search for unmapped pages (holes in the
    address space). It will not tell you if these pages are valid, which is
    what you want to know.
    
    	Jon

    Whatever happened to the old PFM (PageFault Monitor) utility?  As
    I recall, there is a bit reserved in the hardware PCB (i.e., in
    the VAX architecture itself) saying `this process is being monitored.'
    I think the PFM software used that.  There was a system service
    ($SETPFM) too, which may or may not still be around..  it is in
    V4 at least, although I suppose support and useability is somewhat
    problematical.
    
    The PFM wrote a file containing the PC and faulting address, at
    least.  It may have included the time and number of pages, but I
    don't recall.
    
    #ken	:-)}

    	I think the bit in the PCB just runs a line on the backplane. This
    allows a hardware monitor to know when the interesting process is
    current. I don't think it is much use with a software monitor.

Yup, the performance monitor bit in the PSL is meant for use in non-intrusive
hardware monitors, such as DIAMOND (DIAgnostic MONitoring Device) for the
11/780.

--PSW