Re: [PATCH v2 3/7] mm/thp: fix vma_address() if virtual address below file offset

From: Hugh Dickins
Date: Fri Jun 04 2021 - 18:36:35 EST


On Fri, 4 Jun 2021, Matthew Wilcox wrote:
> On Thu, Jun 03, 2021 at 02:40:30PM -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 + compound_nr(page) > vma->vm_pgoff) {
>
> I think on 32-bit, you need ...
>
> pgoff + compound_nr(page) - 1 >= vma->vm_pgoff
>
> ... right?

Hey, beating me at my own game ;-) I'm pretty sure you're right (and
it's true that I first wrote this patch before becoming conscious of
the 32-bit MAX_LFS_FILESIZE issue); but caution tells me to think
some more and check some places before committing to that.

Thanks,
Hugh