On Mon, Dec 23, 2013 at 07:50:31AM -0700, Matthew Wilcox wrote:

quoted

quoted

quoted

+	pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
+	size = (i_size_read(inode) + PAGE_SIZE - 1) >> PAGE_SHIFT;
+	if (pgoff >= size)
+		return VM_FAULT_SIGBUS;
+	if ((pgoff | PG_PMD_COLOUR) >= size)
+		return VM_FAULT_FALLBACK;

I don't think it's necessary to fallback in this case.
Do you care about SIGBUS behaviour or what?

I'm looking to preserve the same behaviour we see with PTE mappings.  I mean,
it's supposed to be _transparent_ huge pages, right?

Speaking of which ... we also need to check if the PMD is entirely within
the VMA.  So, this is needed:

@@ -308,6 +308,11 @@ static int do_xip_pmd_fault(struct vm_area_struct *vma, uns
        /* Fall back to PTEs if we're going to COW */
        if ((flags & FAULT_FLAG_WRITE) && !(vma->vm_flags & VM_SHARED))
                return VM_FAULT_FALLBACK;
+       /* Fall back to PTEs if the mapping would extend outside the VMA */
+       if ((address & ~PMD_MASK) < vma->vm_start)
+               return VM_FAULT_FALLBACK;
+       if (ALIGN(address, PMD_SIZE) >= vma->vm_end)
+               return VM_FAULT_FALLBACK;
 
        pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
        size = (i_size_read(inode) + PAGE_SIZE - 1) >> PAGE_SHIFT;

I'll fold that into the next round of patches.