Driver design question

[Bill Frantz]

At 15:12 7/13/95 -0400, Jonathan Shapiro wrote:
>   Depending on your implementation, you can have the map you load when you
>   dispatch the domain be the same map as the kernel (losing all the
>   advantages of isolation), and you can probably contrive to even dispatch it
>   in supervisor state, making it a real kernel thread.
>
>My current plan was to dedicate some DIBs to such tasks, but not
>bother building a real domain for them.  I would mark these DIBS as
>"locked", which would prevent them from being decommitted (analogous
>to locked TLB entries).  The context switcher works against the DIB in
>any case.  This also avoids the need to check for supervisor state,
>since the PSW bits as captured in the DIB are assumed to be
>trustworthy.
>
>Does this raise any problems you can see?
>
>
>Jonathan

It may cost additional tests if there is any code that assumes that there
are the nodes of a domain when there is a DIB.  The test for "locked DIB"
may be a lot less frequent than the use of these nodes.  e.g. If there is
no keys node, then the domain can't do gate jumps.


-----------------------------------------------------------------
Bill Frantz                   Periwinkle  --  Computer Consulting
(408)356-8506                 16345 Englewood Ave.
frantz@netcom.com             Los Gatos, CA 95032, USA