🚀 go-pugleaf

In article <8106@pt.cs.cmu.edu> lindsay@MATHOM.GANDALF.CS.CMU.EDU (Donald Lindsay) writes:
>In article <43367@ames.arc.nasa.gov> lamaster@ames.arc.nasa.gov (Hugh LaMaster) writes:
>>1) Several posters have mentioned that there is some unspecified but obvious
>>(to them) major problem with using inverted page tables together with
>>memory mapped files.  ...
>
>The big problem is with sharing.

The ROMP, actually its mmu chip which is named Rosetta, handles this pretty
well despite having a reverse map page table with no provision for aliasing.
The high four bits of a memory address are looked up in a tiny fast memory
to get a segment number.  These segment numbers are global to the entire
system.  The file mapping we put into AIX maps a file to a segment, so if you
map a file into several processes there is no aliasing problem; they use the
same segment.

If you map a file copy-on-write, the aliasing problem does exist, and they
handled that at the VRM level via soft page faults.  The same problem occurs
after a fork, since the pages of the stack and data segments are physically
but not logically shared.  In that case, they made them copy-on-touch,
evidently because by the time they'd handled the soft page fault, the extra
effort involved in making a copy of the page wasn't all that great.
-- 
John R. Levine, Segue Software, POB 349, Cambridge MA 02238, +1 617 864 9650
johnl@esegue.segue.boston.ma.us, {ima|lotus|spdcc}!esegue!johnl
"Now, we are all jelly doughnuts."