userfaultfd: clear the vma->vm_userfaultfd_ctx if UFFD_EVENT_FORK fails

commit 0cbb4b4f4c44f54af268969b18d8deda63aded59 upstream.

The previous fix in commit 384632e67e ("userfaultfd: non-cooperative:
fix fork use after free") corrected the refcounting in case of
UFFD_EVENT_FORK failure for the fork userfault paths.

That still didn't clear the vma->vm_userfaultfd_ctx of the vmas that
were set to point to the aborted new uffd ctx earlier in
dup_userfaultfd.

Link: http://lkml.kernel.org/r/20171223002505.593-2-aarcange@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Reviewed-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fs/userfaultfd.c

@@ -570,11 +570,14 @@ out:
 static void userfaultfd_event_wait_completion(struct userfaultfd_ctx *ctx,
 					      struct userfaultfd_wait_queue *ewq)
 {
+	struct userfaultfd_ctx *release_new_ctx;
+
 	if (WARN_ON_ONCE(current->flags & PF_EXITING))
 		goto out;
 
 	ewq->ctx = ctx;
 	init_waitqueue_entry(&ewq->wq, current);
+	release_new_ctx = NULL;
 
 	spin_lock(&ctx->event_wqh.lock);
 	/*
@@ -601,8 +604,7 @@ static void userfaultfd_event_wait_completion(struct userfaultfd_ctx *ctx,
 				new = (struct userfaultfd_ctx *)
 					(unsigned long)
 					ewq->msg.arg.reserved.reserved1;
-
-				userfaultfd_ctx_put(new);
+				release_new_ctx = new;
 			}
 			break;
 		}
@@ -617,6 +619,20 @@ static void userfaultfd_event_wait_completion(struct userfaultfd_ctx *ctx,
 	__set_current_state(TASK_RUNNING);
 	spin_unlock(&ctx->event_wqh.lock);
 
+	if (release_new_ctx) {
+		struct vm_area_struct *vma;
+		struct mm_struct *mm = release_new_ctx->mm;
+
+		/* the various vma->vm_userfaultfd_ctx still points to it */
+		down_write(&mm->mmap_sem);
+		for (vma = mm->mmap; vma; vma = vma->vm_next)
+			if (vma->vm_userfaultfd_ctx.ctx == release_new_ctx)
+				vma->vm_userfaultfd_ctx = NULL_VM_UFFD_CTX;
+		up_write(&mm->mmap_sem);
+
+		userfaultfd_ctx_put(release_new_ctx);
+	}
+
 	/*
 	 * ctx may go away after this if the userfault pseudo fd is
 	 * already released.