> If the server could install the
> mapping directly into the user's task, that would probably be better for
> performance

Why?  Not only would this require a system trap, but there are many
trust issues which you would need to resolve.  Anyway, pager_memcpy
only uses memcpy (as opposed to the far more efficient vm_map) for
unaligned and non-full page reads and writes.

> (is that how it will be done in L4?),

When issuing a receive ipc (or send and receive ipc), the client
provides a receive window.  The server returns an fpage which the
kernel installs into the receive window.



_______________________________________________
Help-hurd mailing list
[EMAIL PROTECTED]
http://mail.gnu.org/mailman/listinfo/help-hurd

Reply via email to