[PATCH] mm/mmap: Align the length parameter of munmap with hugepage size

[Haoran Jang]

From: Haoran Jiang <jianghaoran@xxxxxxxxxx>

munmap hugepge mappings, if the length of the range to munmap
is not aligned with hugepage size,munmap will fail.
In the hugetlb_vm_op_split function, an error will be returned
if startaddr+len is not hugepage size aligned.

before this patch:
in "tools/testing/selftests/mm/hugepage-mremap.c"
modify DEFAULT_LENGTH_MB to 3M,compile and run,
the following error message is displayed

-------------------------
running ./hugepage-mremap
-------------------------
TAP version 13
1..1
Map haddr: Returned address is 0x7eaa40000000
Map daddr: Returned address is 0x7daa40000000
Map vaddr: Returned address is 0x7faa40000000
Address returned by mmap() = 0x7cb34b000000
Mremap: Returned address is 0x7faa40000000
First hex is 0
First hex is 3020100
Bail out! mremap: Expected failure, but call succeeded

Signed-off-by: Haoran Jiang <jianghaoran@xxxxxxxxxx>
---
 mm/mmap.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/mm/mmap.c b/mm/mmap.c
index 83b4682ec85c..0b3a60bf9b6f 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2733,7 +2733,15 @@ int do_vmi_munmap(struct vma_iterator *vmi, struct mm_struct *mm,
 	if ((offset_in_page(start)) || start > TASK_SIZE || len > TASK_SIZE-start)
 		return -EINVAL;
 
-	end = start + PAGE_ALIGN(len);
+	vma = find_vma(mm, start);
+	if (!vma) {
+		if (unlock)
+			mmap_write_unlock(mm);
+		return 0;
+	}
+
+	end = start + ALIGN(len, vma_kernel_pagesize(vma));
+
 	if (end == start)
 		return -EINVAL;
 
-- 
2.43.0