hugetlbfs: fix offset overflow in hugetlbfs mmap
authorMike Kravetz <mike.kravetz@oracle.com>
Thu, 13 Apr 2017 21:56:32 +0000 (14:56 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 14 Apr 2017 01:24:21 +0000 (18:24 -0700)
If mmap() maps a file, it can be passed an offset into the file at which
the mapping is to start.  Offset could be a negative value when
represented as a loff_t.  The offset plus length will be used to update
the file size (i_size) which is also a loff_t.

Validate the value of offset and offset + length to make sure they do
not overflow and appear as negative.

Found by syzcaller with commit ff8c0c53c475 ("mm/hugetlb.c: don't call
region_abort if region_chg fails") applied.  Prior to this commit, the
overflow would still occur but we would luckily return ENOMEM.

To reproduce:

   mmap(0, 0x2000, 0, 0x40021, 0xffffffffffffffffULL, 0x8000000000000000ULL);

Resulted in,

  kernel BUG at mm/hugetlb.c:742!
  Call Trace:
   hugetlbfs_evict_inode+0x80/0xa0
   evict+0x24a/0x620
   iput+0x48f/0x8c0
   dentry_unlink_inode+0x31f/0x4d0
   __dentry_kill+0x292/0x5e0
   dput+0x730/0x830
   __fput+0x438/0x720
   ____fput+0x1a/0x20
   task_work_run+0xfe/0x180
   exit_to_usermode_loop+0x133/0x150
   syscall_return_slowpath+0x184/0x1c0
   entry_SYSCALL_64_fastpath+0xab/0xad

Fixes: ff8c0c53c475 ("mm/hugetlb.c: don't call region_abort if region_chg fails")
Link: http://lkml.kernel.org/r/1491951118-30678-1-git-send-email-mike.kravetz@oracle.com
Reported-by: Vegard Nossum <vegard.nossum@oracle.com>
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/hugetlbfs/inode.c

index 7163fe014b57f4e15813c1969958d5764b18e5af..dde861387a407810b35f73ff37c3ce6389048326 100644 (file)
@@ -136,17 +136,26 @@ static int hugetlbfs_file_mmap(struct file *file, struct vm_area_struct *vma)
        vma->vm_flags |= VM_HUGETLB | VM_DONTEXPAND;
        vma->vm_ops = &hugetlb_vm_ops;
 
+       /*
+        * Offset passed to mmap (before page shift) could have been
+        * negative when represented as a (l)off_t.
+        */
+       if (((loff_t)vma->vm_pgoff << PAGE_SHIFT) < 0)
+               return -EINVAL;
+
        if (vma->vm_pgoff & (~huge_page_mask(h) >> PAGE_SHIFT))
                return -EINVAL;
 
        vma_len = (loff_t)(vma->vm_end - vma->vm_start);
+       len = vma_len + ((loff_t)vma->vm_pgoff << PAGE_SHIFT);
+       /* check for overflow */
+       if (len < vma_len)
+               return -EINVAL;
 
        inode_lock(inode);
        file_accessed(file);
 
        ret = -ENOMEM;
-       len = vma_len + ((loff_t)vma->vm_pgoff << PAGE_SHIFT);
-
        if (hugetlb_reserve_pages(inode,
                                vma->vm_pgoff >> huge_page_order(h),
                                len >> huge_page_shift(h), vma,
@@ -155,7 +164,7 @@ static int hugetlbfs_file_mmap(struct file *file, struct vm_area_struct *vma)
 
        ret = 0;
        if (vma->vm_flags & VM_WRITE && inode->i_size < len)
-               inode->i_size = len;
+               i_size_write(inode, len);
 out:
        inode_unlock(inode);