Re: 230-objrmap fixes for 2.6.3-mjb2

[Andrea Arcangeli <andrea@suse.de>]

On Fri, Mar 26, 2004 at 03:25:23PM -0800, Andrew Morton wrote:
> Andrea Arcangeli <andrea@suse.de> wrote:
> >
> > On Fri, Mar 26, 2004 at 11:57:23AM +0000, Hugh Dickins wrote:
> > > Andrea's 2.6.5-rc2-aa4 (anon_vma): based on Martin's, but very
> > > likely safe since it does not use find_vma at all in swapout, and
> > > unuse_process downs mmap_sem as Martin's used to before.
> > 
> > Hugh, thanks for the review! I also don't see locking bugs in this area
> > in my tree and I like not to take the page_table_lock during vma
> > manipulations since I don't seem to need it.
> 
> It would be really, really nice if we could clean this crap up.  mmap_sem
> protects the vma tree, i_shared_sem protects the per-address_space vma
> lists and page_table_lock protects the pagetables.
> 
> Does this sound like something we can achieve?

yes, and I already achieved this in my tree, so if you plan to merge my
stuff you don't need to touch it in your tree (so it also avoids me to
reject on stuff I already did).

> (Could page_table_lock then become per-vma?)

I probably could make it per-vma, but it's low prio at this time.