Re: [PATCH v4 3.0-rc2-tip 2/22] 2: uprobes: Breakground page replacement.

[Srikar Dronamraju <srikar@linux.vnet.ibm.com>]

* Peter Zijlstra <peterz@infradead.org> [2011-06-10 01:03:32]:

> On Tue, 2011-06-07 at 18:28 +0530, Srikar Dronamraju wrote:
> > +/**
> > + * __replace_page - replace page in vma by new page.
> > + * based on replace_page in mm/ksm.c
> > + *
> > + * @vma:      vma that holds the pte pointing to page
> > + * @page:     the cowed page we are replacing by kpage
> > + * @kpage:    the modified page we replace page by
> > + *
> > + * Returns 0 on success, -EFAULT on failure.
> > + */
> > +static int __replace_page(struct vm_area_struct *vma, struct page *page,
> > +                                       struct page *kpage)
> 
> This is a verbatim copy of mm/ksm.c:replace_page(), I think I can
> remember why you did this, but the changelog utterly fails to mention
> why we need a second copy of this logic (or anything much at all).
> 

__replace_page is not exactly a copy of replace_page. Its a slightly
modified copy of replace_page. Here are the reasons for having this
modified copy instead of using the replace_page.

replace_page was written specifically for ksm purpose by Hugh Dickins.
Also Hugh said he doesnt like replace_page to be exposed for other uses.
He has plans for further modifying replace_page for ksm specific
purposes which might not be aligned to what we are using.

Further for replace_page, its good enuf to call page_add_anon_rmap()
However for uprobes needs we need to call page_add_new_anon_rmap().
page_add_new_anon_rmap() will add the page to the right lru list.

replace_page needs a reference to orig_pte while __replace_page doesnt
need.

I can add the same to Changelog but I am not sure it makes a good
reading. Hence I had skipped it.

-- 
Thanks and Regards
Srikar