> I think that would probably break NSS/DCSS processing, since those can
appe=
> ar in multiple address spaces, and you'd need to fix all those references
t=
> oo.

That does indeed add complexity, but is solvable - for shared NSS/DCSS
pages, the "owning entity" is the NSS/DCSS itself, that's what we'd
point back to (they each have their own set of DAT structures).
Luckily, the "forward pointer" that needs to be found and updated
is in the PGMBK, which for shared segments, is shared across all
of the users who have it loaded (so we wouldn't need to visit each
user). For exclusive pages, since each user gets their own copy
(effectively of the PGMBK, not just the individual page) on first
reference, we'd point back to the user structures as for any other
private user owned page.

...

> Better to let Bill Holder worry about it.

Thanks.  I've got plenty of worrying going already, but this is
on the list (including all of the interesting ideas generated
from this discussion).  :-)

- Bill

Bill Holder, Senior Software Engineer
IBM z/VM Development, Memory Management, Endicott, NY
Phone:  607-429-3640        IBM TieLine: 620-3640

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For more information on Linux on System z, visit
http://wiki.linuxvm.org/

Reply via email to