userfaultfd: shmem: __do_fault requires VM_FAULT_NOPAGE
authorAndrea Arcangeli <aarcange@redhat.com>
Fri, 10 Mar 2017 00:16:28 +0000 (16:16 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 10 Mar 2017 01:01:09 +0000 (17:01 -0800)
__do_fault assumes vmf->page has been initialized and is valid if
VM_FAULT_NOPAGE is not returned by vma->vm_ops->fault(vma, vmf).

handle_userfault() in turn should return VM_FAULT_NOPAGE if it doesn't
return VM_FAULT_SIGBUS or VM_FAULT_RETRY (the other two possibilities).

This VM_FAULT_NOPAGE case is only invoked when signal are pending and it
didn't matter for anonymous memory before.  It only started to matter
since shmem was introduced.  hugetlbfs also takes a different path and
doesn't exercise __do_fault.

Link: http://lkml.kernel.org/r/20170228154201.GH5816@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/userfaultfd.c

index 973607d..f62199b 100644 (file)
@@ -490,7 +490,7 @@ int handle_userfault(struct vm_fault *vmf, unsigned long reason)
                         * in such case.
                         */
                        down_read(&mm->mmap_sem);
-                       ret = 0;
+                       ret = VM_FAULT_NOPAGE;
                }
        }