Re: [PATCH 3/7] mm/thp: fix vma_address() if virtual address below file offset

[Hugh Dickins <hughd@google.com>]

On Tue, 1 Jun 2021, Matthew Wilcox wrote:
> On Tue, Jun 01, 2021 at 02:09:31PM -0700, Hugh Dickins wrote:
> >  static inline unsigned long
> > -__vma_address(struct page *page, struct vm_area_struct *vma)
> > +vma_address(struct page *page, struct vm_area_struct *vma)
> >  {
> > -	pgoff_t pgoff = page_to_pgoff(page);
> > -	return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
> > +	pgoff_t pgoff;
> > +	unsigned long address;
> > +
> > +	VM_BUG_ON_PAGE(PageKsm(page), page);	/* KSM page->index unusable */
> > +	pgoff = page_to_pgoff(page);
> > +	if (pgoff >= vma->vm_pgoff) {
> > +		address = vma->vm_start +
> > +			((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
> > +		/* Check for address beyond vma (or wrapped through 0?) */
> > +		if (address < vma->vm_start || address >= vma->vm_end)
> > +			address = -EFAULT;
> > +	} else if (PageHead(page) &&
> > +		   pgoff + (1UL << compound_order(page)) > vma->vm_pgoff) {
> 
> 	} else if (PageHead(page) &&
> 		   pgoff + compound_nr(page) > vma->vm_pgoff) {

Yes, that's better, thanks.

I was tempted to leave out the preliminary PageHead() test altogether;
but you didn't suggest that, and I've concluded that perhaps it makes
for good documentation, even though not strictly needed.

> 
> > +vma_address_end(struct page *page, struct vm_area_struct *vma)
> >  {
> > +	pgoff_t pgoff;
> > +	unsigned long address;
> > +
> > +	VM_BUG_ON_PAGE(PageKsm(page), page);	/* KSM page->index unusable */
> > +	pgoff = page_to_pgoff(page);
> > +	if (PageHead(page))
> > +		pgoff += 1UL << compound_order(page);
> > +	else
> > +		pgoff++;
> 
> Again, can use compound_nr here.  In fact, the whole thing can be:
> 
> 	pgoff += compound_nr(page);

Even nicer; and I've taken the liberty of just saying

	pgoff = page_to_pgoff(page) + compound_nr(page);

v2 of this patch coming up now, thanks.

Hugh