Re: [PATCH v2] mm: Introduce new function vm_insert_kmem_page

[Matthew Wilcox <willy@infradead.org>]

On Tue, Oct 23, 2018 at 06:03:42PM +0530, Souptick Joarder wrote:
> On Tue, Oct 23, 2018 at 5:54 PM Matthew Wilcox <willy@infradead.org> wrote:
> > On Tue, Oct 23, 2018 at 05:44:32PM +0530, Souptick Joarder wrote:
> > > Instruction from Matthew  Wilcox who is supervising the entire vm_fault_t
> > > migration work :-)
> >
> > Hang on.  That was for the initial vm_fault_t conversion in which each
> > step was clearly an improvement.  What you're looking at now is far
> > from that.
> 
> Ok. But my understanding was, the approach of vm_insert_range comes
> into discussion as part of converting vm_insert_page into vmf_insert_page
> which is still part of original vm_fault_t conversion discussion.  No ?

No.  The initial part (converting all page fault methods to vm_fault_t)
is done.  What remains undone (looking at akpm's tree) is changing the
typedef of vm_fault_t from int to unsigned int.  That will prevent new
page fault handlers with the wrong type from being added.

I don't necessarily want to get rid of vm_insert_page().  Maybe it will
make sense to do that, and maybe not.  What I do want to see is thought,
and not "Matthew told me to do it", when I didn't.