Re: Questions about copy-on-write

[Rian Hunter]

Neal H. Walfield wrote:

> At Wed, 27 Oct 2004 15:14:16 -0400,
> Rian Hunter wrote:
>  
>
> > yes definitely, physmem would have to know who has the page COW and who 
> > really owns the page, otherwise how would it know who to make the 
> > physical copy for on a fault.
> >    
>
> This is wrong: if you have a frame with 10 users and one forces a copy
> what happens?  Well obviously a copy is made and the copy is
> associated with the task which wants to do that write which the other
> nine continue to refer to the original.  The last one to do a write
> gets the "original" frame--not the one who started the share.
>  
This is what I wrote? What I mean by "who really owns the page" is the 
task from which the paged was originally mapped. The task "who has the 
page COW" could be any one of the 10 users you refer to. Of course those 
pages don't refer to each other, they refer to the original that they 
mapped from.

> > i think the general understanding has currently bin that 
> > mapping/unmapping will be a privilegded operation. ie. only hurd servers 
> > should be doing, or most precisely mostly physmem. untrusted tasks won't 
> > really be unmapping/mapping, they will be using containers and closing 
> > containers, and dying. physmem will take care of the actual business of 
> > mapping and copying and etc.
> >    
>
> It is only convention which says that that tasks do not create map
> items or do grants.  Nothing else.  Tasks may do whatever they like.
>
>  
Yeah that was what I was trying to imply with "[something] only hurd 
servers _should_ be doing." "Should" just means what the current design 
current encourages, not strictly enforces. I also want to disclaim my 
"bin" typo: too much digging through the filesystem, :).
-rian