userfaultfd: hugetlbfs: prevent UFFDIO_COPY to fill beyond the end of i_size

This oops:

  kernel BUG at fs/hugetlbfs/inode.c:484!
  RIP: remove_inode_hugepages+0x3d0/0x410
  Call Trace:
    hugetlbfs_setattr+0xd9/0x130
    notify_change+0x292/0x410
    do_truncate+0x65/0xa0
    do_sys_ftruncate.constprop.3+0x11a/0x180
    SyS_ftruncate+0xe/0x10
    tracesys+0xd9/0xde

was caused by the lack of i_size check in hugetlb_mcopy_atomic_pte.

mmap() can still succeed beyond the end of the i_size after vmtruncate
zapped vmas in those ranges, but the faults must not succeed, and that
includes UFFDIO_COPY.

We could differentiate the retval to userland to represent a SIGBUS like
a page fault would do (vs SIGSEGV), but it doesn't seem very useful and
we'd need to pick a random retval as there's no meaningful syscall
retval that would differentiate from SIGSEGV and SIGBUS, there's just
-EFAULT.

Link: http://lkml.kernel.org/r/20171016223914.2421-2-aarcange@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

1 files changed, 30 insertions, 2 deletions

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 424b0ef08a60..2d2ff5e8bf2b 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -3984,6 +3984,9 @@ int hugetlb_mcopy_atomic_pte(struct mm_struct *dst_mm,
                             unsigned long src_addr,
                             struct page **pagep)
 {
+        struct address_space *mapping;
+        pgoff_t idx;
+        unsigned long size;
         int vm_shared = dst_vma->vm_flags & VM_SHARED;
         struct hstate *h = hstate_vma(dst_vma);
         pte_t _dst_pte;
@@ -4021,13 +4024,24 @@ int hugetlb_mcopy_atomic_pte(struct mm_struct *dst_mm,
         __SetPageUptodate(page);
         set_page_huge_active(page);
 
+        mapping = dst_vma->vm_file->f_mapping;
+        idx = vma_hugecache_offset(h, dst_vma, dst_addr);
+
         /*
          * If shared, add to page cache
          */
         if (vm_shared) {
-                struct address_space *mapping = dst_vma->vm_file->f_mapping;
-                pgoff_t idx = vma_hugecache_offset(h, dst_vma, dst_addr);
+                size = i_size_read(mapping->host) >> huge_page_shift(h);
+                ret = -EFAULT;
+                if (idx >= size)
+                        goto out_release_nounlock;
 
+                /*
+                 * Serialization between remove_inode_hugepages() and
+                 * huge_add_to_page_cache() below happens through the
+                 * hugetlb_fault_mutex_table that here must be hold by
+                 * the caller.
+                 */
                 ret = huge_add_to_page_cache(page, mapping, idx);
                 if (ret)
                         goto out_release_nounlock;
@@ -4036,6 +4050,20 @@ int hugetlb_mcopy_atomic_pte(struct mm_struct *dst_mm,
         ptl = huge_pte_lockptr(h, dst_mm, dst_pte);
         spin_lock(ptl);
 
+        /*
+         * Recheck the i_size after holding PT lock to make sure not
+         * to leave any page mapped (as page_mapped()) beyond the end
+         * of the i_size (remove_inode_hugepages() is strict about
+         * enforcing that). If we bail out here, we'll also leave a
+         * page in the radix tree in the vm_shared case beyond the end
+         * of the i_size, but remove_inode_hugepages() will take care
+         * of it as soon as we drop the hugetlb_fault_mutex_table.
+         */
+        size = i_size_read(mapping->host) >> huge_page_shift(h);
+        ret = -EFAULT;
+        if (idx >= size)
+                goto out_release_unlock;
+
         ret = -EEXIST;
         if (!huge_pte_none(huge_ptep_get(dst_pte)))
                 goto out_release_unlock;