Re: [patch02/05]: Containers(V2)- Generic Linux kernel changes

[Rohit Seth <rohitseth@google.com>]

On Wed, 2006-09-20 at 13:27 +0200, Andi Kleen wrote:
> Rohit Seth <rohitseth@google.com> writes:
> >  					 */
> > +#ifdef CONFIG_CONTAINERS
> > +	struct container_struct *ctn; /* Pointer to container, may be NULL */
> > +#endif
> 
> I still don't think it's a good idea to add a pointer to struct page for this.

I thought last time you supported adding a pointer to struct page (when
you mentioned next gen slab will also consume page->mapping).  May be I
missed your point.

> This means any kernel that enables the config would need to carry this significant
> overhead, no matter if containers are used to not.
> 
Sure this is non-zero overhead but I think this is the logical place to
track the memory.

> Better would be to store them in some other data structure that is only
> allocated on demand or figure out a way to store them in the sometimes
> not all used fields in struct page.
> 

which one...I think the fields in page structure are already getting
doubly used. 

> BTW your patchkit seems to be also in wrong order in that when 02 is applied
> it won't compile.

Not sure if I understood that.  I've myself tested these patches on
2.6.18-rc6-mm2 kernel and they apply just fine.  Are you just trying to
apply 02....if so then that wouldn't suffice.

-rohit