Re: Task startup

[Johan Rydberg]

address@hidden (Niels Möller) wrote:

: > If you want to use the same startup protocol as for normal tasks,
: > you have to take things like fork() into account.  How should the
: > freshly created task know which pages it already has (as COW), in
: > what container they live?  All this information has to be passed
: > on the stack.  
: 
: I imagine containers will be passed in the same way as all other
: capabilities. Although I still only have a vague understanding of how
: the containers will work.

That is true, with the exception of a few special capabilities that is
used by the "startup glue code."   The glue code will only operate on
"raw" capabilities (server thread, cap id) since it is not linked against
libhurd-cap.   The startup code must be able to identify these caps.
That could ether be done by separate them from the othere caps, or list
the caps in a special order (say; primary container, task cap, others..)

-- 
Johan Rydberg, Free Software Developer, Sweden
http://rtmk.sf.net | http://www.nongnu.org/guss/