Given the feedback from the _POSTAGE_STAMP_ kernel I'm going to
	see if I can compile up another targetted specifically at the
	KA630 - we should be able to make another few hundred K
	available :)

	Oh, and updated readme included at base:

-------------------------------------------------------------------------------
Wed Jul 14 1999: (before the NetBSD 1.4.1 release)                              

    Source used
        The source used should be the same as in the upcoming 1.4.1 release,
        and should be available as /pub/NetBSD/NetBSD-release.           

    Most important change from 1.4 
        A fix to avoid deadlocking the pagedaemon, so both paging 
        and swapping should work now as God intended. Many thanks to
        Chuck Silvers and Ragge.

    Outstanding issues
        Under heavy swapping random processes may get segfaults.
        Ragge is still working to track this down. You could expect
        to see this when compiling libc on a 6MB machine.

        Page tables are not swapped out. This causes the machine
        to use a little bit more memory than needed, but Ragge will
        fix it in the future.

    Contents of ftp://ftp.netbsd.org/pub/NetBSD/arch/vax/kernels/1.4.1/
        netbsd.generic.gz   Generic kernel
        netbsd.ps.gz        'postage stamp' kernel for KA410, 630 and 650
        _POSTAGE_STAMP_     Kernel config for above
        locore.diff         See note below

          Note: If you are using the _POSTAGE_STAMP_ config to configure
                and build your own kernel you will need to uncomment VAX43
                or 'cd /sys/arch/vax/vax ; patch < locore.diff'.

                David/absolute

        "Its just another ultimate battle of good against evil..."
-------------------------------------------------------------------------------

		David/absolute

         -=- They took away my irony, but they left my deceit  -=-

On Wed, 14 Jul 1999, Anders Magnusson wrote:

> > 
> > 	All in ftp://ftp.netbsd.org/pub/NetBSD/arch/vax/kernels/1.4.1/
> > 
> Great, David!
> 
> Some comments about the swap fixes though:
> 
> The problem with deadlocking the pagedaemon is gone, so both paging and
> swapping should work now as God intended. BUT:
> 
> Under heavy swapping random processes may get segfaults. I haven't 
> been able to track this down, it's quite hairy to find. This happens
> for example most likely when compiling libc on a 6MB machine.
> 
> Page tables are not swapped out. This causes the machine to use a little
> bit more memory than needed, but I will fix it in the future.
> 
> The fixes for this is both in the -current tree and in the 1.4.1 branch.
> 
> Many thanks (again) to Chuck Silvers who helped me out to understand
> how those parts of UVM worked, and how to do a good implementation
> of the page table handling.
> 
> -- Ragge
>