Re: [PATCHv5 22/36] mm, hugetlb: switch hugetlbfs to multi-order radix-tree entries

On Wednesday, November 30, 2016 9:16 PM Kirill A. Shutemov wrote:

On Wed, Nov 30, 2016 at 05:48:05PM +0800, Hillf Danton wrote:

quoted

On Tuesday, November 29, 2016 7:23 PM Kirill A. Shutemov wrote:

quoted

@@ -607,10 +605,10 @@ static long hugetlbfs_fallocate(struct file *file, int mode, loff_t offset,
 		}

 		/* Set numa allocation policy based on index */
-		hugetlb_set_vma_policy(&pseudo_vma, inode, index);
+		hugetlb_set_vma_policy(&pseudo_vma, inode, index >> huge_page_order(h));

 		/* addr is the offset within the file (zero based) */
-		addr = index * hpage_size;
+		addr = index << PAGE_SHIFT & ~huge_page_mask(h);

 		/* mutex taken here, fault path and hole punch */
 		hash = hugetlb_fault_mutex_hash(h, mm, &pseudo_vma, mapping,

Seems we can't use index in computing hash as long as it isn't in huge page size.

Look at changes in hugetlb_fault_mutex_hash(): we shift the index right by
huge_page_order(), before calculating the hash. I don't see a problem
here.

You are right. I missed that critical point.

thanks
Hillf